This issue highlights the diversity of APL’s contributions to national and global challenges. The issue begins by discussing APL’s
role as the independent test and evaluation agent for an unmanned underwater vehicle family of systems, components of which have
been fielded by the U.S. Navy. The second article provides an introduction to the physics of fission chains and their detection, of vital
importance should a nuclear response team locate a potential nuclear object. The next article describes a process for balancing conflicts
among missile system performance goals, engagement support capabilities, and technology constraints. The fourth article details a
semiempirical model for leaves as a step toward improving Earth’s land characterization and removing radiative effects. The final article
highlights a technique for determining the atmosphere’s trace gas composition, which has broad applications for measuring gases globally.
The issue concludes with a list of the winners of APL’s Technical Achievement Awards and Prizes.
Editorial Board
Staff
Harry K. Charles Jr., Chair and Editor-in-Chief
Linda L. Maier-Tyler, Assistant Editor-in-Chief
James P. Christ, Member-at-Large
Miquel D. Antoine
Ra’id S. Awadallah
Benjamin H. Barnum
Jeffrey P. Hamman
David O. Harper
Mary Kae Lockwood
Zaruhi R. Mnatsakanyan
Peter P. Pandolfini
William K. Peter
Charles E. (Terry) Schemm
James C. Spall
Erin M. Richardson, Managing Editor
Kelly K. Livieratos, Assistant Managing Editor
Diane J. Dorsey, Art Director
Kenneth R. Moscati, Senior Illustrator
Editorial Support: Anne E. King and Peggy M. Moore
Illustration Support: Matthew D. Baughman,
Annie M. Marcotte, Shannon E. Nelson, and
Magda M. Saina
Front Cover Art: Kenneth R. Moscati
Inside Back Cover Art: Kenneth R. Moscati
Web Publisher: Erin M. Richardson
Photographers: Bill Rogers and Edward Whitman
Ex Officio
Clearance Coordinator: Bonnie Cubbage-Lain
Karen M. Gosnell
Erin M. Richardson
Production Assistant: Michele G. Tranter
The Johns Hopkins APL Technical Digest (ISSN 0270-5214) is published quarterly under the auspices of The Johns Hopkins University Applied Physics Laboratory (JHU/APL), 11100 Johns Hopkins Road, Laurel, MD 20723-6099. The objective of the publication is to provide a summary of unclassified individual
programs under way at JHU/APL. Subscriptions are free of charge but are provided only to qualified recipients (government employees and contractors, libraries, university faculty members, and R&D laboratories). Requests for individual copies, subscriptions, or permission to reprint the text and/or figures should be
submitted to the Office of the Editor-in-Chief at the above address. Phone: (240) 228-6588. E-mail: TechnicalDigest@jhuapl.edu.
The following abstracting services currently cover the Johns Hopkins APL Technical Digest: Chemical Abstracts; Current Contents; Science Citation Index;
Engineering Village; and the following CSA abstracts: Aerospace and High Technology Database; Aquatic Sciences and Fisheries Abstracts; Computer and
Information Systems Abstracts; Electronics and Communications Abstracts; Mechanical and Transportation Engineering Abstracts; Meteorological and Geoastrophysical Abstracts; and Oceanic Abstracts.
Postmaster: Send address changes to the Johns Hopkins APL Technical Digest, Rm. 2-204, 11100 Johns Hopkins Road, Laurel, MD 20723-6099. Periodical postage
paid at Laurel, MD.
© 2014 by The Johns Hopkins University Applied Physics Laboratory.
The Johns Hopkins APL Technical Digest can be accessed electronically at www.jhuapl.edu/techdigest.
The electronic version may include multimedia capabilities for enhanced visualization of some concepts.
Johns Hopkins APL
TECHNICAL DIGEST
Volume 32, Number 5
APL Research and Development
752 Unmanned Underwater Vehicle Independent Test and Evaluation
William P. Ervin, J. Patrick Madden, and George W. Pollitt
762 Neutrons: It Is All in the Timing—The Physics of Nuclear Fission Chains and Their Detection
William A. Noonan
774 Missile Concept Optimization for Ballistic Missile Defense
Alan J. Pue, Richard J. Hildebrand, Daniel E. Clemens, Jonah R. Gottlieb, James M. Bielefeld,
and Timothy C. Miller
787 Improving Earth Background Characterization Through Modeling and Measurements of Leaf
Bidirectional Reflectivity
Shadrian B. Strong, Michael E. Thomas, Andrea M. Brown, and Elena Y. Adams
803 The MSX/UVISI Stellar Occultation Experiments: Proof-of-Concept Demonstration of a New
Approach to Remote Sensing of Earth’s Atmosphere
Ronald J. Vervack Jr., Jeng-Hwa Yee, William H. Swartz, Robert DeMajistre, and Larry J. Paxton
MISCELLANEA
822 APL Achievement Awards and Prizes
Linda L. Maier-Tyler
Unmanned Underwater Vehicle
Independent Test and Evaluation
William P. Ervin, J. Patrick Madden, and George W. Pollitt
he Johns Hopkins University Applied Physics Laboratory (APL) has a long
history of contributing to unmanned undersea or underwater vehicle (UUV)
programs sponsored by several Navy acquisition program offices. Those
contributions span the systems engineering realm, including leadership of independent test
and evaluation for prototypes and systems fielded for military use. One of the most enduring relationships has been with Program Manager Naval Sea Systems Command (Expeditionary Missions) (PMS-408) for its acquisition of UUVs applied to mine countermeasures
(MCM) missions. Since 2002, APL has served as the independent test and evaluation agent
for the Mk 18 UUV family of systems. The fielding of key components of the Mk 18 UUV
family of systems was accelerated as part of an Office of the Secretary of Defense “FastLane” program to meet an operational need in theater. As a result, Commander Fifth Fleet
now has an improved operational MCM capability, including advanced sensors.
OVERVIEW
In December 2011, the Office of the Secretary of
Defense approved a Fast-Lane initiative to provide
Mk 18 Mod 2 Kingfish unmanned underwater vehicle
(UUV) systems and associated sensors and upgrades
to Commander Fifth Fleet (C5F) on an accelerated
basis. Seven months later, in July 2012, wave 1 of the
Mk 18 Mod 2 Kingfish UUVs arrived in the C5F area of
responsibility to begin search, classify, and map missions
as part of a phased incremental-capability rapid-fielding
plan that included an extended user operational evaluation system (UOES) period in theater. The purpose of
752
the UOES period was to develop mine countermeasures
(MCM) concepts of operations (CONOPS) for integration with other MCM platforms in theater and to
receive operator feedback that could be used to improve
the design. A second wave of Mk 18 Mod 2 UUVs
arrived in theater in February 2013. The third wave
arrived in October 2013 and included more UUVs and
ancillary equipment. Advanced sensors and command
and control technologies were demonstrated in theater
in November 2013. After undergoing operational testing in February and April 2014, respectively, they were
JOHNS HOPKINS APL TECHNICAL DIGEST, VOLUME 32, NUMBER 5 (2014)
provided as operational capabilities. The rapid delivery
of these capabilities to meet the commander’s operational need was made possible by several factors, including the following:
• Strict adherence to identified measurable and testable user requirements
Acquisition Category (ACAT) IV in terms of programmatic expenditures] incrementally developed, tested,
and fielded leading-edge unmanned vehicle and command and control technologies using both military and
civilian crewing philosophies. As the Navy transitions
to increased reliance on unmanned systems, as well as to
the operational integration of unmanned and manned
systems in the underwater domain, future acquisition
programs might consider adopting aspects of this program’s organization, development, testing, and fielding
practices to help navigate the acquisition pipeline.
• Outstanding testing and feedback support from
operational units
BACKGROUND
• A technologically mature system design when the
Fast-Lane initiative was approved
• Strong program office leadership of a multi­
organizational integrated product team (IPT)
• A competitive manufacturer selection process
• A “build a little, test a little, field a little” development process
• Responsive in-service engineering agent (ISEA)
support
The Mk 18 Mod 2 systems in theater are being operated by civilian contractor crews led by a government
civilian. The crews and their leadership are under the
administrative and operational command of the C5F
Explosive Ordnance Disposal (EOD) and MCM task
force commanders, respectively. The civilian crews in
the C5F area of responsibility will be replaced by military crews.
Advances in unmanned ground vehicles have
reduced human casualty risk during EOD operations in
Afghanistan and Iraq. Investment in UUV technologies is considered particularly important for the maritime environment because UUVs can “get the man out
of the minefield” for some, if not all, required missions.
The Mk 18 Mod 2 Kingfish UUV program is one of the
acquisition community’s initiatives to meet the fleet
mission need to conduct EOD MCM operations more
safely, efficiently, and effectively against a wide spectrum of current and anticipated threats in a variety of
operational environments. UUVs of various sizes, with
increasing levels of autonomy, sensor capability, and payload composition, comprise a rapidly expanding part of
the “toolbox” available to address underwater domain
mission requirements.
The Mk 18 Mod 2 was preceded by other systems
used for hydrographic surveys and harbor defense and by
the militarized remote environmental measuring units
(REMUS), which were used in 2003 as part of the clearance of Umm Qasr, Iraq, during Operation Iraqi Freedom.
According to Captain Michael Tillotson, Commander,
Naval Special Operations Task Force 56 during Operation Iraqi Freedom, “If we didn’t have UUVs, you could
multiply the time to clear the [Umm] Qasr area by twoand-a-half, an additional 20 days.”1 In less than 15 years,
this relatively small acquisition program [less than
JOHNS HOPKINS APL TECHNICAL DIGEST, VOLUME 32, NUMBER 5 (2014)
APL has a long history contributing to UUV programs sponsored by the Defense Advanced Research
Projects Agency, the Office of Naval Research (ONR),
and several Navy acquisition program offices. Those
critical contributions span the systems engineering
realm, including leadership of independent test and
evaluation for prototypes and systems fielded for military
use. One of those contributions was to assist in the preparation of The Navy UUV Master Plan in 2000, which
was updated in 20042 and identified nine high-priority
UUV missions:
• Intelligence, surveillance, and reconnaissance
• Mine countermeasures
• Anti-submarine warfare
• Inspection/identification
• Oceanography
• Communication/navigation network nodes
• Payload delivery
• Information operations
• Time-critical strike
To perform these missions, The Navy UUV Master
Plan characterized vehicle systems into four general
classes:
• Man-portable vehicle class: Vehicles of approximately 25–100 lb displacement with 10–20 hours of
mission endurance; no specific hull shape identified
• Lightweight vehicle class: 12.75-in.-diameter
vehicles of approximately 500 lb displacement,
20–40 hours endurance, with payloads of 6–12 times
the size of a man-portable vehicle
• Heavyweight vehicle class: 21-in.-diameter vehicles,
up to 3000 lb displacement, 40–80 hours endurance,
with 2 times payload capacity of lightweight vehicle
class, suitable for launch from submarines
753­­­­
W. P. ERVIN, J. P. MADDEN, AND G. W. POLLITT
Year
2000
was reached between PMS-408
and COMOPTEVFOR, APL
REMUS technology (NOAA, ONR, WHOI)
was tasked to take over the role
REMUS production (Hydroid, Inc.)
of independent test and evaluaMk 14 Mod 0 SAHRV (NSW)
tion agent (IT&EA) for followSculpin preliminary capability (PMS-408)
on non-ACAT small UUV
Mk 18 Mod 1 Swordfish (PMS-408)
programs. Since 2002, APL has
served as the IT&EA for the origMk 18 Mod 2 Kingfish (PMS-408)
inal very shallow water (VSW)
Fast-Lane initiative (OSD)
UUV program, the Bottom
Mk 18 improvements (ONR, PMS-408)
UUV Localization System, and
the Mk 18 FoS program. Figure 1
Figure 1. UUV acquisition time line. NOAA, National Oceanic and Atmospheric Administraprovides a chronology of system
tion; NSW, Naval Special Warfare Program Office; PMS-408, Program Manager Naval Sea Sysdevelopment that culminated in
tems Command (Expeditionary Missions); OSD, Office of the Secretary of Defense; SAHRV,
the initiation of the Mk 18 FoS
semi-autonomous hydrographic reconnaissance vehicle.
program of record.
Design changes and improvements made to the SAHRV and Sculpin, as a result of
• Large vehicle class: Vehicles with approximately
user feedback, enabled a running start for the Mk 18
10 long tons displacement and suitable for launch
FoS. Figure 2 shows the relative sizes of the Mk 18 Mod 1
from surface ships (e.g., littoral combat ship) and
man-portable and Mk 18 Mod 2 lightweight vehicles.
submarines
The Mk 18 Mod 1 vehicles are designed to be
The lightweight vehicle-class Mk 18 Mod 2 Kingfish
launched and recovered by operators in small boats such
UUV is a larger, extended-range version of the Mk 18
as the 4.7-m combat rubber raiding craft (CRRC) or 7-m
Mod 1 Swordfish man-portable search, classify, and
map system currently deployed in several operational
theaters. In accordance with Secretary of the Navy
Instruction 5000.2E,3 both systems were developed as
Abbreviated Acquisition Programs (AAPs) under the
sponsorship of OPNAV N957 and guidance of PMS-408
using an informal IPT organization from requirements
development through system development, developmental testing, user evaluation, and operational fielding.
The Mk 18 family of systems (FoS) is based on
REMUS vehicles built by Hydroid, Inc., a subsidiary
of Kongsberg Maritime. The Mk 18 Mod 1 and Mod 2
vehicles are REMUS 100 and REMUS 600 vehicles,
respectively, where the number denotes the rated depth
of the vehicle in meters.
Figure 2. Mk 18 Mod 1 and Mod 2 UUVs. (Image courtesy of
REMUS UUV technologies originated in the early
Space and Naval Warfare System Center, Pacific.)
1990s at the Woods Hole Oceanographic Institution in
Massachusetts. UUV systems based on REMUS vehicles
are in use by the navies of the United States, United
Kingdom, and others that leverage the REMUS UUV
family of vehicles. The Navy tested and fielded earlier
versions of small man-portable REMUS UUVs, known
as the SAHRV, and the Sculpin (a predecessor of the
Swordfish). REMUS vehicles are also in use by commercial, oceanographic, and academic organizations in
several countries.
Commander Operational Test and Evaluation Force
(COMOPTEVFOR, designated by the Chief of Naval
Operations to be the Navy’s sole independent agency for
operational test and evaluation of ACAT I through IV
programs) conducted the operational evaluation of the
Figure 3. Recovery of Mk 18 Mod 1.
SAHRV vehicles in the late 1990s. After an agreement
1990
754
2010
JOHNS HOPKINS APL TECHNICAL DIGEST, VOLUME 32, NUMBER 5 (2014)
UNMANNED UNDERWATER VEHICLE INDEPENDENT TEST AND EVALUATION
Figure 4. Launch preparations for Mk 18 Mod 2.
rigid hull inflatable boat (RHIB). The Mk 18 Mod 2
vehicles are designed to be launched and recovered from
11-m RHIBs using a specially built launch and recovery
system or by crane from a ship. Figure 3 shows a Mod 1
vehicle being recovered from a CRRC by EOD Mobile
Unit One (EODMU-1) operators during factory acceptance testing in 2006. Figure 4 shows two operators from
EODMU-1 rigging the 11-m RHIB Mk 18 Mod 2 launch
and recovery system before a vehicle launch during the
February 2011 user evaluation testing.
The Mk 18 Mod 1 was required to address threats in
the VSW and some parts of the shallow water region.
The Mk 18 Mod 2 was also required to conduct operations in the VSW and shallow water region, but in
addition, the fleet requested characterization of system
performance in deeper water. Figure 5 shows the standard operating regions by depth for MCM operations.4
In December 2011, C5F submitted a request for
additional expeditionary underwater MCM operations
capabilities, and an Office of the Secretary of Defense
(OSD)-funded “Fast-Lane” program was established to
accelerate the transition of existing and planned Mk 18
Mod 2 UUV capabilities into theater as soon as possible.
In November 2012, unrelated to the Fast-Lane initiative and as part of the maturing of the Mk 18 FoS,
OPNAV 957 established a consolidated requirements
document for search-based UUVs in support of expeditionary operations and an ACAT IV program to
continue the development of UUV underwater MCM
capabilities. OPNAV 957 is currently coordinating the
development of the Lightweight Expeditionary MCM
UUV (LEMUUV) capability development document.
Key participants in the development of Mk 18 Mod 2
system capabilities are listed in Fig. 6.
FACTORS FOR SUCCESSFUL DEVELOPMENT,
TESTING, AND FIELDING
Several factors enabled the rapid delivery of Mk 18
Mod 2 capability in theater to meet the commander’s
operational need.
Technologically Mature System Design
Much like the Mod 1 system development that leveraged the REMUS 100 vehicle technologies (including
the propulsion design, battery power supply, sensors, navigation capabilities, and vehicle interface program), the
Mk 18 Mod 2 system development leveraged the Mk 18
Mod 1 development and programmatic documentation
(acquisition plan, requirements document, and performance specification) and the existing REMUS 600
vehicle technologies. The Mod 2 vehicles had already
demonstrated reliable operations before mission testing
started. The importance of starting operator testing with
a reliable vehicle cannot be
overstated. Fleet operators
will not use an unreliable
system despite any promise
of new capability.
Because the vehicles and
supporting equipment (e.g.,
laptops, software applications, and vehicle communications) were reliable,
fleet, government civilian,
and contractor operators
had the opportunity to use
the system in operationally
relevant environments and
provided early feedback.
Anti-invasion
Buried/
Bottom
Moored
Moored
Floating
Rising
This led the developer and
partially buried
influence
contact influence
contact
influence
IPT membership to recognize, early in the process,
Figure 5. Littoral mine threats (Image courtesy of Office of the Chief of Naval Operations Expedithe importance of operator
tionary Warfare Directorate). CLZ, craft landing zone.
training, the need to address
JOHNS HOPKINS APL TECHNICAL DIGEST, VOLUME 32, NUMBER 5 (2014)
755­­­­
W. P. ERVIN, J. P. MADDEN, AND G. W. POLLITT
requirements and feedback
on design and CONOPS as
well as lessons learned from
the development and operational testing of the SAHRV
Programmatic guidance
and Sculpin vehicles.
• PMS-408
Emphasis was placed on
designing
modular payloads
Fleet users
so that systems could later
• EODMU-1
• NOMWC
be upgraded as sensor, vehi• MDSU-2
cle navigation and stability,
• C5F civilian crews
and battery technologies
improved. It was recognized early on that sensor
maturity would not support
Contractor support
Government laboratories
• ITT Exelis (acquisition support)
operations in the most rig• SSC-San Diego, CA (developmental
• Hydroid (vehicles)
testing, FoS integration)
orous of environments or in
• ARL:UT (ATLAS)
• NSWC-Panama City, FL (ISEA, tactics)
every sea state against every
•
ARL:PSU
(SSAM)
• NSWC-Carderock, MD (battery)
• Orca Maritime (CONOPS)
• NUWC-Newport, RI (autonomous functions)
possible threat, so realistic,
• JHU/APL (IT&EA)
measurable, and testable
ACRONYMS: ARL:PSU, Applied Research Laboratories, Pennsylvania State University; ARL:UT, Applied Research Laboratories, The University of
requirements were estabTexas at Austin; ATLAS, Autonomous Topographic Large Area Survey; MDSU-2, Maritime Diving and Salvage Unit Two; NMAWC, Naval Mine and
lished for what were to be
Anti-Submarine Warfare Command; NOMWC, Naval Oceanography Mine Warfare Center; NSWC, Naval Surface Warfare Center; NUWC, Naval
Undersea Warfare Center; OPNAV N957, Chief of Naval Operations, Expeditionary Combat Branch; SSAM, Small Synthetic Aperture Minehunter;
considered first-generation
SSC, Space and Naval Warfare Systems Center.
vehicles and sensors.
The program manager
Figure 6. Mk 18 FoS IPT.
actively championed the
importance of evaluating the systems’ operational suithuman machine interface issues to facilitate operator
ability aboard ship and for use from all the expected
use, and the importance of having sufficient spares and
launch platforms (e.g., amphibious ship well decks, 7-m
logistics in place to support sustained operations.
and 11-m RHIBs, CRRCs, and piers). Figure 7 shows
launch and recovery operations of the Mk 18 Mod 1
Program Office Leadership
system during an EODMU-1 command exercise aboard
The program manager worked closely with OPNAV,
USS Denver, amphibious transport dock 9 (LPD-9) in
the Navy requirements community, ONR, prospective
December 2006.
developers including Hydroid, government laboratories,
The program manager also recognized the need to
and the fleet to forge a dedicated team with identifiprovide a consistent and interoperable way to provide
able goals and milestones. The program manager diccommand and control of multiple and different
tated that a phased, walk first–run later development
UUVs and, after a thorough industry search, funded
approach would be undertaken. The importance of getdevelopment of the Common Operator Interface Navyting something reliable into the hands of operators for
EOD (COIN) software to conduct mission planning
evaluation and feedback was emphasized from the start.
and post-mission analysis (PMA). The COIN output
For the Mk 18 Mod 1 system, the size and weight of the
was compatible with the Navy’s standard mine warfare
vehicle was directly driven by the operator’s ability to
tactical decision aid known as the Mine Warfare and
handle it during launch and recovery and during transEnvironmental Decision Aids Library (MEDAL).
port to and from the launch platform. Any manufacturer
MEDAL compatibility made the UUVs interoperable
request to increase the size or weight of the vehicle was
with the larger Navy’s activities. The program manager
contingent on operator agreement and demonstration
decreed that for any UUV manufacturer to compete
that the added weight or size could be safely integrated.
for future production opportunities, the vehicles
Enabled by the program manager, IPT members,
must be able to exchange information via the COIN
including the small APL team (one to three part-time
system. The policy encouraged UUV manufacturers to
personnel at any time over a 12-year period), were
make their systems compatible with COIN for mission
afforded early access to the technology in development
planning and PMA.
by ONR and the manufacturers. IPT members were
The program manager worked closely with ONR
educated on the maturity of existing and near-term
to encourage the development of the next-generation
sensor technologies and vehicle endurance characterissensor technologies so that UUVs could operate in more
tics. IPT members were also exposed early to operator
challenging environments and against a wider array of
Fleet requirements
• OPNAV N957
• C5F
• NMAWC
756
Research and development
• ONR
JOHNS HOPKINS APL TECHNICAL DIGEST, VOLUME 32, NUMBER 5 (2014)
UNMANNED UNDERWATER VEHICLE INDEPENDENT TEST AND EVALUATION
Figure 7. RHIB launch and recovery aboard an amphibious ship.
threats. Desired sensor technology enhancements and
increased endurance requirements eventually led to
increases in the size and weight of the vehicles and to
the initiation of the Mod 2 program.
Measurable and Testable User Requirements
The initial requirements for the VSW MCM UUV
were drafted by acquisition and technical subject-matter experts, followed by early input from the operational community. Formal requirements documentation
was then influenced by experience gained during the
SAHRV operations and the first VSW MCM UUV
UOES period with the REMUS 100 system that later
became known as Sculpin, as well as by expected
CONOPS and knowledge of the existing and near-term
threat. Early exposure to SAHRV and Sculpin lessons
learned, near-term UUV and sensor technologies, and
user requirements and CONOPS enabled the IPT to
develop measurable and testable requirements.
APL, as the IT&EA, was part of the vetting of the
requirements documentation and stressed that, if the
criteria for the measurement and testing of a requirement were not clearly delineated in writing, the requirement should be rewritten until it was both measurable
and testable. The program manager supported this philosophy and ultimately mediated and resolved several
discussions where the operators, developers, and testers
differed on the interpretation of the requirements. APL
later used these requirements to prepare and deliver
test plans, conduct user evaluation during several test
periods from 2002 to the present, analyze results, and
characterize the system’s operational effectiveness and
suitability in several formal reports.
A good example of a challenging test requirement
was the “Probability of Classifying a non-mine as a
mine (PCXM).” This metric was recommended over
using the more widely known MCM metric “non-mine
density for classification” because PCXM could be more
objectively measured during the brief user evaluation
periods. The UOES experience with the REMUS 100
JOHNS HOPKINS APL TECHNICAL DIGEST, VOLUME 32, NUMBER 5 (2014)
UUVs indicated that PCXM was not as straightforward
as it sounded. Images provided by early-generation sidescan sonars made some commonly accepted non-mine
objects appear mine-like. As a result, objects that did not
produce mine-like returns, known as distractors, were
placed on the ocean bottom among the exercise mines
to unambiguously assess the PCXM metric. Although
false contact density was not an imposed requirement,
because it varies widely with the environment, the
metric was routinely included in test reports.
To ensure an understanding of the UUV system’s
capabilities, the IT&EA participated during key parts of
UOES periods, such as command exercises where the
UUV system was used operationally. In addition, developmental testing personnel invited APL staff to observe
whenever fleet operators were operating the vehicles.
The early use of the system by fleet operators provided
valuable feedback on the expected CONOPS and an
appreciation by the IPT members for the “must-haves”
versus the “nice-to-haves,” and it also enabled realistic
planning for the subsequent user evaluations. During
user evaluations, all UUV operations were planned,
conducted, and analyzed by fleet operators. User evaluation is to an AAP what an operational evaluation is to
an ACAT program.
System performance specifications, key performance parameters, and critical operational issues were
identified and approved by the program manager after
endorsement by members of the IPT. Although requirements were informed by existing intelligence agency
threat characterizations, the system requirements did
not require threshold performance against every conceivable underwater threat. Provisions were made in the
acquisition program testing plan to characterize performance against more challenging threats and environments without mandating specific performance against
all anticipated threats.
APL staff also leveraged the COMOPTEVFOR
SAHRV program documentation and testing methodology to help ensure that user requirements for operational effectiveness and operational suitability were
adequately represented in the requirements and testing
documentation.
Outstanding Operator Support
Outstanding support was provided during all
UOES, developmental testing, and user evaluation
testing periods by the EODMU-1 UUV platoon (previously known as Naval Special Clearance Team One)
and later by other military and civilian UUV operators from Naval Oceanography Mine Warfare Center,
Maritime Diving and Salvage Unit Two, and Space
and Naval Warfare Systems Center Pacific. Operators willingly executed a myriad of testing operations.
Operators diligently completed surveys and interviews
757­­­­
W. P. ERVIN, J. P. MADDEN, AND G. W. POLLITT
and attended and provided regular formal presentations at program reviews. This feedback was essential
for the program. The fleet’s buy-in to the objectives of
the testing program was critical to the identification
and implementation of UUV improvements. Figure 8
shows EODMU-1 personnel during the February 2011
user evaluation rigging Mk 18 Mod 2 vehicles onto an
11-m RHIB pier side at Space and Naval Warfare Systems Center Pacific.
All user evaluation operations were conducted as
blind tests. Similar to real-world operations, UUV
operators were provided a mission and asked to plan,
search, and report results of operationally representative missions. Tests were designed to represent a wide
variety of operational environments. For example, maximum-endurance and short-duration operations were
conducted during daylight and at night, in sea states
that varied from 1 to 3, in various bottom types, using
a variety of launch and recovery platforms, and with
little to no advance notice on the detailed tasking to be
executed. Operators had the flexibility to determine the
battle rhythm so long as all vehicle operations and PMA
were completed during the evaluation. Figure 9 shows an
EODMU-1 operator conducting PMA.
Testing and feedback from operators led to UUV
system requirements such as improved planning and
PMA software functionality; more rugged and waterresistant computers with larger, more viewable screens;
tamper-proof vehicle design; modifications to the battery charging and safety considerations; P-code Global
Positioning System (GPS), vehicle launch, and recovery handling modifications; modified vehicle lighting
for low-visibility operations; more extensive training;
and system documentation including detailed operator and maintenance manuals. For the Mk 18 Mod 2
program, operator feedback was used to design, deliver,
and upgrade launch and recovery systems for individual
vehicles onto 11-m RHIBs and for 11-m RHIBs carrying
Mk 18 Mod 2 vehicles onto ships.
Figure 10 shows civilian operators conducting Mk 18
Mod 2, 11-m RHIB stern gate launch and recovery system
Figure 8. Mk 18 Mod 2 pier side loading onto 11–m RHIB.
758
Figure 9. EODMU-1 operator conducting PMA.
Figure 10. Mk 18 Mod 2 11-m RHIB stern gate launch and recovery system.
operations during April 2013 testing aboard USS Ponce,
Afloat Forward Staging Base Interim 15 [AFSB(I) 15],
in the Arabian Gulf, with APL participation as IT&EA.
The alternative to the stern gate launch and recovery is
craning the 11-m RHIB off the ship as shown in Fig. 11.
Competitive Manufacturer Selection Process
After a broad agency announcement, multiple companies participated in a demonstration at the Naval
Amphibious Base, Little Creek, Virginia, to enable the
Navy to select potential UUVs to enter into the acquisition process. This demonstration and down-selection
allowed PMS-408 to begin a UOES period with two
technologically mature man-portable UUVs.
The UOES period and Mk 18 Mod 1 developmental test periods yielded two UUV systems provided by
different manufacturers that were assessed and determined sufficiently ready to proceed to user evaluation
to support a production decision. For the user evaluation in 2004, vehicles from the two manufacturers were
tested in the same minefields over a rigorous 6-week
test period. Fleet operators from Naval Special Clearance Team One (later called EODMU-1) operated the
vehicles in daytime and nighttime conditions similar to
JOHNS HOPKINS APL TECHNICAL DIGEST, VOLUME 32, NUMBER 5 (2014)
UNMANNED UNDERWATER VEHICLE INDEPENDENT TEST AND EVALUATION
Figure 11. Shipboard crane operations for Mk 18 Mod 2 11-m RHIB.
those anticipated during amphibious operations. Environmental testing (vibration, shock, and temperature)
was later conducted to ensure that both vehicles were
capable of sustained operations in more challenging
environments.
After review of the developmental testing, user evaluation, and environmental testing results in 2005, APL
as the IT&EA endorsed the Hydroid vehicle as operationally effective and suitable, and the program manager selected the REMUS 100 vehicle for production
of several systems (consisting of three vehicles each)
that later became known as Mk 18 Mod 1 systems. The
Mk 18 Mod 2 was developed as an engineering change
to the Mod 1 because the REMUS technologies were
readily scalable.
The program manager continued to investigate other
technologies, some provided by foreign manufacturers,
to improve performance of UUVs or their supporting
software. To avoid interoperability (and cost) issues, the
program manager hired a software consultant to oversee
the selection and development of a mission-planning
and PMA software package that would enable any willing manufacturer’s vehicle to be interoperable with fleet
JOHNS HOPKINS APL TECHNICAL DIGEST, VOLUME 32, NUMBER 5 (2014)
needs for required analysis and reports. COIN, initially
developed by SeeByte, a small foreign company, was
chosen as the software technology, and the government
team worked to purchase the rights to the software and
sufficient licenses to support operations.
“Build a Little, Test a Little, Field a Little” Development
Process
As indicated above, the program manager determined that a phased, walk first–run later development
approach would be used. This translated first into the
identification of reliable trucks (Mod 1 and Mod 2 vehicles) to haul payloads (sensors and navigation equipment) in the required operating environments. Once the
reliable trucks were chosen, focus shifted to incremental
delivery of increasingly capable sensors and navigation
equipment. In several cases, testing revealed incomplete
sensor-to-vehicle or navigation-to-vehicle integration
and less-than-anticipated improvement in navigation
or sensor PMA results. End-to-end system performance
observations were identified during UOES or dedicated
developmental testing or user evaluation periods.
759­­­­
W. P. ERVIN, J. P. MADDEN, AND G. W. POLLITT
PMS-408 applied the acquisition concept of low-rate
initial production to order small quantities of vehicles
to support improvement testing and, when system performance was characterized as meeting requirements,
additional quantities of systems were procured. The
importance of delivering reliable equipment to the hands
of operators for evaluation and feedback was emphasized
from the start.
• Interoperability and modularity
Responsive ISEA Support
The Mk 18 FoS program has a plan to continue
upgrading vehicle and sensor technologies along with
adding communications/networking and autonomy as
the systems mature into future increments. Table 1 illustrates the incremental capability improvement approach
that the program office has implemented to the baseline capabilities to meet fleet requirements. Currently,
the baseline Mod 1 (man-portable) and Mod 2 (lightweight) UUVs have been tested and are deployed.
Improved modular sensors for the Mk 18 Mod 2 system,
which is now formally designated as an ACAT IV program, have also been delivered to the fleet. Mk 18 FoS
testing and evaluation efforts are ongoing in parallel to
develop and incrementally deliver capabilities for autonomy, command and control and sensor improvements,
and advanced sensors across the future-year defense
program. These incremental upgrades leverage technologies previously demonstrated by ONR and other
science and technology investments. Concurrently, science and technology efforts for future UUV capabilities
are ongoing. Depending on the success of these investments, they may be implemented in the form of future
block upgrades to the Mk 18 Mod 2 or as a future Mk 18
Mod 3 UUV program. Near-term to midterm initiatives
planned for the Mk 18 UUVs include introduction of
an internal payload computer and supporting architecture to enable automated target recognition, autonomy,
and additional plug-and-play payloads. These initiatives,
along with regularly improved sensors, will build on the
baseline capabilities of the systems for use in more complex operational environments.
The Naval Surface Warfare Center Panama City
Division is the ISEA for both the Mod 1 and Mod 2 programs. A small Naval Surface Warfare Center Panama
City Division staff works closely with the operators and
the manufacturer to manage the repair, component
upgrade, and replacement of UUVs.
The maintenance philosophy, instituted by the program manager, is to provide vehicles as part of a system.
For the Mk 18 Mod 1 and Mod 2 program, a system consists of three vehicles. A multi-vehicle system with a tailored onboard repair parts kit allows the forward-deployed
units to ensure that at least two UUVs are available and
to perform organizational-level repairs to the third. Additional spares are kept at the depot level (manufacturer’s
facility) and are shipped by commercial or military air
to facilitate quick turnaround. In some cases, it is more
efficient to swap entire systems of vehicles.
For both the Mod 1 and Mod 2, there is no intermediate maintenance facility. The operating command has
a stockpile of spare parts, and the manufacturer is under
contract to provide maintenance support if the repair
is beyond operator capability. Manufacturer responsiveness for maintenance has improved under the guidance
of the ISEA thereby meeting fleet needs.
The ISEA and operators normally rely on express
commercial shipping to transport Mk 18 Mod 1 whole
man-portable vehicles for repair or to ship parts that
can be installed when a Hydroid representative is
present for on-site repairs. The larger Mk 18 Mod 2 is
transported aboard military aircraft, but spare parts
are shipped using express
commercial shipping.
• Communications systems, spectrum and resilience
• Security
• Persistent resilience
• Autonomy and cognitive behavior
• Weaponry
Table 1. Mk 18 Mod 2 incremental improvement milestones
LOOKING TO
THE FUTURE
The 2013 Department
of Defense Unmanned
Systems Integrated Roadmap FY2013–20385 identifies six technology areas
to enhance capability and
reduce cost:
760
Phase
Activity
Prototype
Initial production system
Production system
Synthetic aperture sonar
Forward-looking sonar
LEMUUV improvement increment 1
LEMUUV improvement increment 2
LEMUUV improvement increment 3
Used for requirement compliance test and evaluation
System 0, block A vehicles
Follow-on systems, block A+ vehicles
Synthetic aperture sonar sensor module integration
Forward-looking sonar sensor module integration
August 2014
Autonomy and optics enhancement
Command and control and sensor improvements
Multi-sensor integration
JOHNS HOPKINS APL TECHNICAL DIGEST, VOLUME 32, NUMBER 5 (2014)
UNMANNED UNDERWATER VEHICLE INDEPENDENT TEST AND EVALUATION
Figure 12. Civilian crew recovers Mk 18 Mod 2 UUV in CRRC
during overseas sensor testing.
Due largely to successes in fielding the Mk 18 Mod 2
UUV and advanced sensors in support of the OSD FastLane initiative, efforts are underway to procure more
UUV systems and more advanced capabilities than were
originally planned under the AAP strategy. In accordance with acquisition policies, Staff of the Chief of
Naval Operations (OPNAV N957) and PMS-408 have
transitioned from AAP processes to the Joint Capabilities Integration and Development System (JCIDS) process for future development and procurement efforts.
Consequently, future incremental upgrades for the
Mk 18 Mod 2 UUV program will be implemented using
ACAT IV-level program management guidance that
adds rigor and discipline to the development, testing,
and fielding practices.
APL remains ready to support future testing as evidenced by recent experience in early 2014 with one of the
most intense periods of UUV independent test and evaluation for PMS-408. A two-person team, with reach-back
to a third person at APL, deployed overseas for a twoweek evaluation of a synthetic aperture sonar followed
seven weeks later by a two-week stateside evaluation of
a forward-looking sonar. The tempo included finalizing
test plans, conducting test readiness reviews, coordinating placement of exercise mines, performing analysis, presenting quick-look results, and summarizing requirement
compliance test and evaluation results to support production decisions within a few weeks after completing each
event. Figure 12 shows civilian crews conducting Mk 18
Mod 2 operations from a CRRC during advanced sensor
testing conducted overseas in February 2014. Figure 13
shows sensor testing stateside in April 2014.
SUMMARY
The Mk 18 Mod 2 program is an example of an AAP
that successfully responded to a rapid fielding request by
a fleet commander. From the perspective of the IT&EA,
the history of the Mk 18 program indicates that, if “the
build a little, test a little, field a little” process is followed,
there is confidence that the team assembled by PMS-408
will meet schedule and system integration challenges
and continue to provide useful capabilities to the fleet.
REFERENCES
1U.S.
Figure 13. Civilian crew recovering UUV in 11-m RHIB with
forward-looking sonar module.
The Authors
Navy Office of Information, Rhumb Lines, Straight Lines to Navigate By: Talking Points (7 Nov 2003).
2Department of the Navy, The Navy Unmanned Undersea Vehicle
(UUV) Master Plan, 9 Nov 2004.
3Department of the Navy, Implementation and Operation of the Defense
Acquisition System and the Joint Capabilities Integration and Development System, SECNAVINST 5000.2E (1 Sep 2011).
4N852 Mine Warfare Branch, Brief to the Expeditionary Warfare Conference by CAPT Mark Rios, 4 Oct 2010 (document in possession of
corresponding author).
5Department of Defense, Unmanned Systems Integrated Roadmap
FY2013–2038 (2013).
William P. Ervin has test and evaluation experience with combat, command and control, and unmanned systems. J. Patrick Madden has
contributed to several UUV programs including test and evaluation for the Defense Advanced Research Projects Agency and the Navy.
George W. Pollitt is the APL subject matter expert for Mine Warfare and currently serves as the IT&EA for the Mk 18 UUV FoS. For further
information on the work reported here, contact George Pollitt. His e-mail address is george.pollitt@jhuapl.edu.
The Johns Hopkins APL Technical Digest can be accessed electronically at www.jhuapl.edu/techdigest.
JOHNS HOPKINS APL TECHNICAL DIGEST, VOLUME 32, NUMBER 5 (2014)
761­­­­
Neutrons: It Is All in the Timing—The Physics of
Nuclear Fission Chains and Their Detection
William A. Noonan
hould a nuclear response team locate a potential threat object, it
will be imperative for the team to quickly determine whether or not
the object is an actual nuclear device. Only a nuclear device or a large
amount of special nuclear material will sustain a significant number of fission chains.
Therefore, the detection of fission chains constitutes a “smoking gun.” This article is
a technical introduction to the physics of fission chains and their detection using neutron time-correlation methods; it also touches on some of the Johns Hopkins University Applied Physics Laboratory’s efforts to bring this important detection capability to
the field.
INTRODUCTION
Since the 1970s the U.S. government has been concerned about the “loose nukes” scenario, in which a
state with nuclear weapons loses control of one and it
is used for a terrorist attack. In recent decades, several
trends have heightened this fear. There has been widespread proliferation of nuclear weapons technology,
most famously by the A. Q. Khan network.1 Special
nuclear material (SNM), the key component of nuclear
weapons, has also proliferated. In the aftermath of the
Soviet Union’s dissolution, materials protection, control, and accountability for SNM in the former Soviet
Union was poor. In the years since, materials protection, control, and accountability has been strengthened but not entirely fixed,2 and there have been
several documented cases of SNM disappearing from
Russian stockpiles.
762
A third troubling trend has been the steadily growing number of nuclear weapon-owning states that are
not signatories of the Treaty on the Non-Proliferation of
Nuclear Weapons and that might act to further weaken
the non-proliferation regime. Specifically, Pakistan has
a growing stockpile of weapons but questionable political stability, and the government is having trouble contending with local terrorist organizations. Here, the
loose nuke scenario is particularly worrisome. In addition, the Pakistani government has been complicit in
the proliferation of nuclear weapon technology via the
now-defunct smuggling network set up by A. Q. Khan,
the founder of its nuclear weapons program. North
Korea has also demonstrated a nuclear capability while
behaving as a seemingly irrational international actor,
heedless of international norms and unmoored by inter-
JOHNS HOPKINS APL TECHNICAL DIGEST, VOLUME 32, NUMBER 5 (2014)
national commitments. Finally, Iran, a known sponsor
of terrorism, is arguably working toward a nuclear capability, although recent diplomatic developments might
ultimately limit the Iranian program.
Although it is impossible to estimate the likelihood
that a terrorist organization could actually acquire a
functioning nuclear device, the possibility exists via at
least two different pathways.3, 4 Terrorists could steal
a state weapon or they could acquire SNM and proliferated nuclear technology and then build their own
improvised nuclear device.
Thus, the U.S. government maintains an assortment
of assets and mechanisms for interdicting the movement
of illicit SNM or an actual nuclear device. The U.S.
Customs and Border Patrol operates radiation portal
monitors at ports of entry throughout the United States.
Radiation portal monitors are also operated in ports
around the world, in collaboration with host nations as
part of Customs and Border Patrol’s Container Security
Initiative and the Department of Energy’s Megaports
project.5 In addition, a string of sensors was emplaced
along the borders around Russia as part of the Department of Energy’s Second Line of Defense program.6 This
patchwork of detection capabilities is coordinated under
the Global Nuclear Detection Architecture managed by
the Domestic Nuclear Detection Office in the Department of Homeland Security.7
Should any of these portals or sensors detect the passage of suspicious radiological material, a special rapid
response team could be deployed to assess the threat.
Alternatively, there might be an intelligence cue for the
movement of SNM or a nuclear device, in which case
the U.S. government would deploy national assets to
search for, locate, and asses the threat.
Once the threat object is located, the response team
would use diagnostic instrumentation to determine the
level of threat it poses. The most pressing question for
(a)
0.40
To understand how fission chains can be detected, it
is first necessary to delve into some aspects of the physics of fission chains. As is widely known, certain isotopes of the heaviest elements can fission into smaller
nuclei. This process can happen spontaneously, or it
can be induced by a free neutron colliding with the
nucleus. When a nucleus fissions, it emits zero, one, or
more fission neutrons with probabilities that depend on
the particular isotope involved; on whether the fission is
spontaneous or induced; and in the case of induced fission, on the kinetic energy of the incident neutron. The
number of neutrons emitted in a given fission event is
called its “multiplicity,” and the probability distribution
of different multiplicities is called a “multiplicity distribution.” The multiplicity distributions for the spontaneous and induced fission of various important isotopes are
plotted in Fig. 1.
It is possible for the neutrons emitted by the fission of
one nucleus to collide with and induce fissions in other
nuclei, which in turn emit more neutrons. But in order
for this phenomenon to create a self-sustaining nuclear
chain reaction, the probability that fission neutrons
successfully induce additional fissions needs to be suf-
vs
U-238 2.21
Pu-240 2.16
Cf-252 3.76
U-238
Pu-240
Cf-252
0.35
0.30
0.25
0.20
0.15
0.35
vi
U-235 2.52
U-238 2.43
Pu-239 3.01
U-235
U-238
Pu-239
0.30
0.25
Probability
Probability
THE PHYSICS OF FISSION CHAINS
(b)
0.45
0.20
0.15
0.10
0.10
0.05
0.05
0.00
the response team would be whether or not the object
was an actual nuclear device. Because the presence of
nuclear chain reactions is the sine qua non of a nuclear
device, their detection would provide the smoking gun
that escalates the national response.
By their nature, these small rapid response teams necessarily have a limited load-out of equipment. The Johns
Hopkins University Applied Physics Laboratory (APL)
is developing technology that would give these teams
the ability to detect fission chains without increasing
their load-out.
0
1
2
3
4
Multiplicity, vs
5
6
7
0.00
0
1
2
3
4
Multiplicity, vi
5
6
Figure 1. Multiplicity distributions and average multiplicity for various important isotopes for (a) spontaneous fission and (b) fission
induced by 1-MeV neutrons.8, 9
JOHNS HOPKINS APL TECHNICAL DIGEST, VOLUME 32, NUMBER 5 (2014)
763­­­­
W. A. NOONAN
the average number of fission neutrons in the first genficiently high. Although there are many isotopes that
are fissionable, there are relatively few isotopes for which
eration of a fission chain is then p v i (see Fig. 2). Because
this probability is high enough to sustain chain reaceach of these first-generation neutrons will induce their
tions, and these select few are called “fissile.”
own fission with probability p, there are p $ p v i secondConsider the case of uranium. The largest constituent
generation fissions and ^p v ih2 second-generation fission
of naturally abundant uranium is uranium-238 (U-238),
neutrons. Thus, there are ^p v ihn neutrons in the nth genwhich is fissionable but not fissile. The kinetic energies
eration, and the total number of neutrons produced by
of most fission neutrons are not high enough to induce
the fission chain is
fission in U-238. Furthermore, U-238 has a strong pre1
1
dilection for absorbing intermediate energy neutrons,
,(1)
N = 1 + p v i + ^p v ih2 + ^p v ih3 + f = 1 – p v =
1–k
preventing them from inducing further fissions. (Incii
dentally, plutonium-239, or Pu-239, used in both power
where k = p v i is the multiplication factor. If k ≥ 1, then
reactors and nuclear weapons, is produced using this
the series in Eq. 1 diverges, and the assembly is said to be
absorption reaction: U-238 is placed inside a nuclear
either critical or supercritical.
reactor, where it absorbs neutrons and coverts to U-239.
Each of these N fission neutrons has probability 1 – p
U-239 is radioactively unstable and quickly decays into
Pu-239.) On the other hand, a minor constituent, U-235,
of escaping the nuclear assembly, so the average number
can be induced to fission by neutrons with any kinetic
of neutrons from a fission chain that manage to escape is
energy, and so it can sustain a chain reaction if there is
1–p
(2)
^1 – p h N = 1 – p v = M L ,
enough of it present. Because naturally abundant urai
nium consists of 99.3% U-238, there is too little U-235
where ML is called the leakage multiplication.
and too much neutron-absorbing U-238 to sustain chain
Note that this model does not take geometry comreactions. Therefore, uranium needs to be enriched in its
pletely into account. For instance, a neutron born near
U-235 content before it can be used in a nuclear reactor
the surface of a nuclear assembly will have a greater probor weapon. Weapons require highly enriched uranium
ability of escaping than a neutron born at the center,
(HEU), whereas reactors require lower enrichments and
and so p is not really a constant. This effect is often
run on low enriched uranium.
taken into account approximately by using a semiemWe can construct a simple model for the average
pirical effective value for k (keff  k) in Eq. 1. Regardless
number of neutrons released by a single chain reaction
(also called a “fission chain”). Any given free neutron can
of whether or not the theoretical value (k = p v i) is used,
suffer one of several fates. It could,
of course, induce a fission; it could
be absorbed without inducing a
fission; or it could do neither and
escape from the nuclear material
altogether. Let p be the probability
that a given neutron will induce a
fission. The value of p depends on
the size and shape of the nuclear
assembly, the type of nuclear
Sf
material being used, and the
fraction of fissile isotopes in the
nuclear material. Furthermore,
nuclear explosives are deliberately
contrived such that the probability of absorption without inducing
fission is small, so we will neglect
absorption in our simple model.
Thus the probability of a neutron
p
pvi
p pνi
( pνi) 2
escaping from the nuclear assembly is simply 1 – p.
If a neutron induces a fission, Figure 2. The evolution of a fission chain seeded by a single neutron. The average numbers
on average v i neutrons will be of induced fissions occurring in the first and second generations are p and p  pvi, respecreleased. Taking into account that tively, and the numbers of fission neutrons are pvi and ^ p vih2, where p is the probability that
a neutron will successfully induce a given neutron will induce a fission and vi is the average number of neutrons emitted by
a fission only with probability p, an induced fission.
•
764
JOHNS HOPKINS APL TECHNICAL DIGEST, VOLUME 32, NUMBER 5 (2014)
THE PHYSICS OF FISSION CHAINS AND THEIR DETECTION
Free neutrons are needed
to seed fission chains, and
they can come from three
SNM
Moderator
sources. First there is spontaneous fission (indicated by
Sf in Fig. 3), but the source
of spontaneous fission neuSf
trons does not have to be the
same fissile isotope that supports the chain reaction. For
(,n)
example, a nuclear assembly
HDPE
made of HEU is composed
of primarily U-235, but
because the U-235 has a
Figure 3. A schematic representation of processes occurring inside a multiplying nuclear assemmuch longer spontaneous
bly and the detection of the neutrons that leak out. Spontaneous fissions (Sf ) produce multiple
fission half-life than U-238,
neutrons and (a,n) reactions produce only one. These neutrons or neutrons from cosmic showers
most of the spontaneous
seed multiple fission chains. Neutron trajectories are shown as crooked lines (indicating multiple
fission neutrons come from
scatterings), and the branching indicates induced fissions producing more neutrons. Neutrons that
U-238, even though it is a
leak out of the nuclear assembly fly largely unimpeded to the He-3 neutron detector, where they
minor constituent.
scatter many times in the surrounding high-density polyethylene (HDPE) until they slow down to
The second source of
thermal energies, at which point they react with the He-3 in the proportional counter tube at the
neutrons
are so-called
center and produce electrical detection pulses.
(a,n) reactions. Not only
does SNM radioactively
decay by spontaneous fission, it also decays by emitting
people often talk about the “k-effective” of a multiplying
a-particles. These a-particles can react with certain
nuclear assembly.
lighter nuclei, producing neutrons. For example, there
Now suppose we start with a spontaneous fission. On
may be trace amounts of oxygen in the metallic SNM
average, a spontaneous fission event emits v s neutrons.
used in a nuclear device, or the SNM may be in oxide
Each of these neutrons seeds its own fission chain, so
form, as it often is in nuclear reactor fuel. Naturally
then the total number of fission neutrons produced on
abundant oxygen is composed of 0.2% oxygen-18, which
average is
can undergo the reaction
vs
NT =
,
(3)
1 – k eff
18O + a → 21Ne + n,
(4)
and N L = ML v s of those neutrons leak from the assemproducing neutrons that can seed fission chains.
bly. Thus the original v s neutrons are multiplied by a
The third source of neutrons comes from cosmic
factor of (1 – keff)–1 overall and by a factor of ML from
showers that impinge on the nuclear assembly. There
the viewpoint of the leakage neutrons. Because only the
is a flux of high-energy charged particles, consisting
leaked neutrons can actually be observed, the leakage
mostly of protons (~90%) coming from space that strike
multiplication is the property that is of central interest.
Earth’s upper atmosphere. These particles interact with
This simple model predicts only the average number
atmospheric atomic nuclei and generate a cascade of eleof neutrons produced in a fission chain that leak from
mentary particles. The particles in this “air shower” that
the nuclear assembly. In reality, this is a random number
reach the ground are mostly protons, neutrons, electrons,
with a probability distribution that depends parametrigamma-rays, and muons. At ground level, cosmic-ray
cally on the induced fission probability p. The mathshowers typically have a short duration: less than 100 ns
ematical theory of branching processes describes the
or so. So-called extensive air showers (tens to thousands
evolution of populations whose members reproduce and
of meters wide and N ~104–109) are infrequent but
die according to probabilistic laws, and the neutron popintense. Smaller air showers occur more frequently, with
ulation in a fission chain can be adequately described by
the frequency growing rapidly with shrinking size. The
this general theory. In particular, techniques from the
flux of particles at ground level depends on many factheory of branching processes can be used to calculate
tors including latitude, air pressure, and solar activity,
the probability distribution of the number of neutrons
but the average flux of neutrons is typically 80 m–2 s–1
leaked by a fission chain.10–13
and 1 m–2 s–1 for protons.14
Next let us turn to the sequence of events occurCosmic neutrons and protons that strike uranium
ring inside a multiplying nuclear assembly and its time
dependence (depicted schematically in Figs. 3 and 4).
nuclei in the nuclear assembly can knock neutrons free
3He
Cosmic neutron
JOHNS HOPKINS APL TECHNICAL DIGEST, VOLUME 32, NUMBER 5 (2014)
765­­­­
W. A. NOONAN
from the nuclei through a process called nuclear spalltime-of-flight delays. This last phenomenon is particularly prominent if the nuclear assembly is surrounded by
ation. These spallation neutrons have an energy speclow-atomic-mass material (called a moderator), which is
trum very similar to fission neutrons, and they can go on
an efficient scatterer of neutrons. An example would be
to induce fission reactions. Alternatively, cosmic protons
neutrons passing through high explosive before reaching
and neutrons can interact with nearby structural steel,
the detector.
also producing spallation neutrons that go on to strike
the nuclear assembly and initiate fission chain reactions.
A single spontaneous fission, (a,n) reaction, or cosmic
DETECTION SCHEMES FOR FISSION CHAINS
neutron incidence is generically called a “source event,”
and these events occur completely randomly in time
The upshot is that when fission chains are present,
(depicted as ’s in Fig. 4a). That is, their occurrence
neutron detections occur in bursts. Although fission
times are a Poisson process (see for instance Ref. 15), and
chains are spread out and occur at completely random
they are governed by the Poisson statistics of classical
times, each chain releases a burst of neutrons over a relaradioactive decay.
tively short interval. This “burstiness” is the signature
Each source event liberates one or more neutrons, and
we use to discern the presence of fission chains, and we
each neutron either escapes from the nuclear assembly
refer to it as neutron time-correlation.
with probability 1 – p, or it seeds its own fission chain.
Unfortunately, fission chains are not uniquely
In turn, each neutron in the fission chain either escapes
indicated by neutron time-correlations; there are
with probability 1 – p or perpetuates the chain. In this
non-multiplying, non-SNM sources that also exhibit
way, a random number of neutrons from each chain leak
correlations. As we have noted already, there are isofrom the nuclear assembly (depicted by bars of random
topes that can still spontaneously fission or undergo
heights in Figs. 4b–4d). The fission neutrons that leak
induced fission—even if they cannot support fission
from the nuclear assembly eventually reach an external
chains—and when they fission, they also release one
neutron detector after suffering a random delay. (These
or more neutrons in a short burst. Both californium-252
randomly delayed arrival times
are depicted by dots in Figs. 4c (a)
and 4d.)
There are several causes for
this delay. First, fission chains
Time
take a finite amount time to
grow, wither, and die, and a (b)
leaked neutron could have
been born at any time during
this life span. Second, fission
Time
neutrons are emitted with a (c)
spectrum of kinetic energies.
T
Hence, the leaked neutrons
travel with a random distribution of velocities and require
Time
a random amount of time to
∆t
cross the distance separat- (d)
ing the nuclear assembly and
an external neutron detector. Finally, a leaked neutron
Time
may scatter multiple times on
Gate
its journey from the nuclear
assembly to the external detec- Figure 4. A schematic representation of the time dependence of events occurring inside a multor. When the neutron scatters, tiplying nuclear assembly. (a) The crosses (×) are source events, e.g., spontaneous fissions or
it gives up a random amount of (a,n) reactions, and they occur at completely random times (a Poisson process). (b) These source
energy, further broadening the events seed fission chains that produce random numbers of neutrons, represented by vertical
distribution of neutron ener- bars with random heights. (c) Not all of these neutrons leak from the assembly and are successgies. Moreover, the length of fully detected, but those that do are detected after random delays from the source event times.
the path traveled on the way The neutron detection times are indicated by filled circles. T is the total time spread of detection
to the detector is increased by times for neutrons emitted by a single source event. (d) The time axis is partitioned into contigua random amount. Both effects ous gates Dt seconds wide for generating the gated-count distributions used by one of the data
broaden the distribution of analysis methods.
766
JOHNS HOPKINS APL TECHNICAL DIGEST, VOLUME 32, NUMBER 5 (2014)
THE PHYSICS OF FISSION CHAINS AND THEIR DETECTION
(Cf-252) and depleted uranium (U-238) are examples that are used industrially and that will
be encountered in the field. Therefore, any detection technique will have to distinguish
these materials from genuine SNM.
A Simple Detection Scheme
In principle, the most straightforward way to observe these time-correlations is to simply
record the detection time of each neutron and then scan the resulting record, looking for the
clumping of detection times. The multiplicity distributions for neutrons emitted by Cf-252 and
U-238 spontaneous fission are plotted in Fig. 1. In the absence of fission chains, the number of
neutron detections in each clump or burst would have that same distribution. Thus, the average numbers of neutrons emitted and detected would be 3.76 and 2.01, respectively. On the
other hand, in fissile material, thanks to neutron multiplication occurring in fission chains,
these averages would be about ML times higher. Therefore, multiplying SNM assemblies with
ML somewhat larger than unity could be distinguished from nonfissile material by computing
the average burst size in the data record, setting a threshold somewhat higher than ~3.76, and
then determining whether the average burst size exceeds this threshold.
However, there are complications. Not every neutron leaking from the nuclear assembly
will strike a detector, and not every neutron striking a detector will actually produce a detection pulse. The fraction, , of neutrons that is actually detected is called the absolute detection efficiency, and the average number of neutrons detected in each burst is reduced by this
factor. If the detection efficiency is known, then the method just described can still be used
if we scale the detection threshold by .
Unfortunately, there will never be an opportunity to independently calibrate the absolute
detector efficiency in an actual field deployment. Thus, a more sophisticated approach to
account for detector efficiency is required. Let N be the random number of neutrons leaked
from a single source event. From our discussion in the previous section, we know that the
probability distribution of this number depends parametrically on the induced fission probability p, i.e., Pr ^N = nh = Pn ^ p h . Let M be the random number of neutrons actually detected.
The probability of detecting any one leaked neutron is simply . Therefore, the probability
of detecting m out of n leaked neutrons follows the binomial distribution. This probability is,
in fact, the conditional probability of M = m given N = n, so
6
@
n
Pr 6M = m N = n@ = c mm ^1 – hn – m.
m
(5)
By applying the law of total probability, we can then calculate the probability distribution,
Qm, for the number of detected neutrons:
n
Q m ^p, h = Pr 6M = m@ = / Pr 6M = m N = n@ Pr 6N = n@ = / c mm ^1– hn – m Pn ^ p h. (6)
m
n
n
Because we can calculate Pn using the theory of branching processes, Eq. 6 yields a theoretical expression for Qm that depends on two parameters, p and . This expression is the basis
of a detection technique that handles unknown detector efficiencies. Rather than analyzing
the time series of neutron detections to find merely the average size of a burst, instead histogram the number of neutrons in each burst (i.e., the “burst count”) to make an empirical
estimate of the probability distribution of M. Then fit the theoretical expression for Qm(p,)
to estimate the parameters p and . Finally, use k eff = p est v i to find k-effective, and then use
Eq. 2 to calculate the leakage multiplication, ML.
A computationally simpler alternative to calculating and fitting Qm(p,) directly is to compute its first two combinatorial moments instead. These two theoretically derived moments
(which are necessarily dependent on p and  also) are equated with the corresponding combinatorial moments of the measured distribution of M. The resulting two equations are then
solved for p and .
In statistics, the shape of a distribution function is often characterized by its moments.
There are several different classes of moments that can be used, depending on the context
JOHNS HOPKINS APL TECHNICAL DIGEST, VOLUME 32, NUMBER 5 (2014)
767­­­­
W. A. NOONAN
of the problem. The most common is the set of central
moments (which includes the mean and the variance),
but other classes of moments are the factorial moments,
cumulants, and combinatorial moments. The jth combinatorial moment of a discrete random variable N is
defined as:
N
n!
M^ j h = c m = /
Pr 6N = n@ . (7)
j
n $ j j! ^ n – jh !
The last case of large detection time spreads due to
the presence of moderator occurs often, because the
most commonly deployed detector type incorporates
moderator in its design. The heart of this detector type
is a proportional counter tube pressurized with helium-3
(He-3) gas. (Refer to Fig. 3 for a simplified drawing.)
Incident neutrons react with the He-3 nuclei through
an (n,p) reaction:
In this alternative approach, the shape of the
burst-count distribution, Qm, is characterized by its
moments—a common strategy in statistics. Because
branching processes are essentially combinatorial in
nature, the combinatorial moments have a relatively
simple algebraic expression. This is the reason that combinatorial moments are used instead of the more familiar
central moments, mean and variance.
Overlapping Fission Chains
A second complication occurs when the clumps of
neutron detections from different fission chains overlap
in time. When this happens, it is impossible to tell when
one fission chain ends and the next chain begins. Thus,
it becomes impossible to parse the time series of neutron
detections into separate chains and then histogram the
burst sizes to estimate Qm.
Overlapping detections of fission chain neutrons
occur when the typical time between source events (e.g.,
the spontaneous fissions that seed the fission chains) is
smaller than the time spread, T, of the randomly delayed
neutron detections (Fig. 4c). This circumstance arises for
nuclear assemblies with high spontaneous fission rates.
It can also occur in situations in which the time spread
of detection delays becomes excessive—for example,
when the neutrons pass through a lot of moderator.
(a)
(8)
The 3He and p reaction products fly apart at high velocities, carrying away the 0.764 MeV of energy released in
the reaction and, by ionizing the helium gas, produce an
electrical pulse at the tube’s output. However, this reaction has a high probability of occurring only for slowly
moving neutrons with energies 10 eV, whereas fission
neutrons are fast moving, with energies on the order
of 1 MeV. To combat this disparity, the proportional
counter tube is encased in hydrogen-rich high-density
polyethylene (HDPE) moderator. Incident neutrons
repeatedly scatter in this moderator, losing energy with
each collision, until they slow down to low energies, at
which point they random-walk into the central tube and
readily react with the He-3.
Figure 5 plots probability distributions for the time
spreads, T, in the detection delays when a moderated
He-3 detector (a) is used versus a fast detector that does
not incorporate a moderator (b). This plot is generated
from experimental data taken on a bare HEU object,
and it shows that the moderator used in the He-3 detector significantly increases the spread in detection times.
For the He-3 detector used in this experiment, detection delays extend out to about 150 ms. Therefore, if the
spontaneous fission rate of a nuclear assembly were to be
(b)
0.08
0.045
0.07
0.040
0.035
0.06
0.05
Probability
Probability
n + 3He → 3H + p + 0.764 MeV.
0.04
0.03
0.025
0.020
0.015
0.02
0.010
0.01
0
0.030
0.005
0
50
100
150
200
Burst duration (µs)
250
300
0
0
10
20
30
40 50 60 70
Burst duration (ns)
80
90 100
Figure 5. Probability distributions for the time spread, T, in the detection times for neutrons from the same fission chain, measured on
the same HEU object using a moderated He-3 detector (a) and a fast neutron detector (b). Note the change in timescale—from microseconds to nanoseconds.
768
JOHNS HOPKINS APL TECHNICAL DIGEST, VOLUME 32, NUMBER 5 (2014)
THE PHYSICS OF FISSION CHAINS AND THEIR DETECTION
above ~1000 s–1 (which is not unrealistic in some cases),
then more than 10% of the fission chains would occur
at times separated by less than 100 ms, and overlapping
fissions chains would be a problem.
For moderated He-3 detectors, a different approach
for quantifying neutron time correlations is needed.
This second approach dispenses with parsing the record
of neutron detection times into fission chains. Instead,
the time axis is partitioned into consecutive “gates”
(Fig. 4d), and the number of neutron detections in each
gate is tallied and histogrammed. The result is an empirical estimate of the probability distribution for getting K
neutron detections within a gate Dt seconds wide. An
experimentally measured gated-count distribution for a
multiplying assembly is plotted in Fig. 6; also plotted is
the best fit Poisson distribution. Clearly the probability
distribution is non-Poisson.
On the other hand, there are industrial sources that
produce neutrons solely from (a,n) reactions, such as
americium-beryllium (AmBe) sources. In these sources,
one and only one neutron is produced from each source
event (Fig. 4a), and so the neutron count distribution will
follow the Poisson distribution characteristic of a-decay.
This suggests that deviations from Poisson statistics can
serve as an alternative signature for fission chains.
Building on the theoretical expression for Qm(p,),
it is possible to derive the probability distribution,
Pr 6K = k@ = R k ^p, , t, S f h, for the random number
of neutrons detected within a gate Dt seconds wide.16
This theoretical distribution for K is compared with the
empirically measured histogram to determine the source
leakage multiplication—just as the empirical histogram
for M is compared to the theoretical distribution Qm(p,)
in the burst-count method.
There are a number of parallels between the two
methods but also some differences. Like Qm, Rk depends
on p and , but it also depends on the gate width, Dt,
and the spontaneous fission rate, Sf. As with the first
method, the mathematical computations are considerably simplified if we work with the moments of Rk instead
of the full distribution. However, because Rk depends
on more parameters, more than two moments need
to be computed. In the burst-count method, we compared the combinatorial moments of the theoretical and
empirically measured distributions. In the gated-count
method, it turns out that we instead equate the “correlation” moments of the theoretical and empirical count
distributions. These correlation moments are defined
recursively in terms of the combinatorial moments, M(j),
and the first three are:
Y1 = M^1h = K
Y2 = M^2h –
1
2!
Y12
Y3 = M^3h – Y2 Y1 –
.
1
3!
(9)
Y13
Probability
The use of these arcane statistical moments may seem
abstruse, but they actually have a simple physical interpretation.11, 17 Suppose we open a gate and count K neutrons
arriving within the gate time. From Eq. 9, we see that
M(1) and Y1 are simply the expected number of neutrons.
Now, any particular group of q
0
neutrons selected from the K
10
neutrons detected might have
–1
come from the same fission
10
chain and are therefore “corre–2
lated.” On the other hand, the
10
neutrons in the group might
–3
have come from two or more
10
different fission chains and,
–4
therefore, have been acciden10
tally grouped. The qth correlation moment, Yq, is simply the
10–5
expected number of q-tuples
10–6
of correlated neutrons in any
given gate.
10–7
Given this physical interpretation, it is easy to see
10–8
why combinatorial moments
Data
are used in the burst-count
Best-fit Poisson distribution
10–9
method, whereas correlation
moments are used in the gated10–10
count method. Because we can
0
5
10
15
20
25
30
Multiplicity, m
parse neutron detections into
their respective fission chains
in the burst-count method, we
Figure 6. Measured neutron gated-count distribution from an HEU object for a 500-ms wide
gates (blue) and the best-fit Poisson distribution (red).
know that the resulting count
JOHNS HOPKINS APL TECHNICAL DIGEST, VOLUME 32, NUMBER 5 (2014)
769­­­­
W. A. NOONAN
Then, as was done in the burst-count method, we use
the estimated value for the induced fission probability,
770
Y3F
Y2F
0.35
0.12
distribution and its combiU3F
natorial moments pertain
to only correlated neutrons.
0.30
0.10
This is not the case for the
gated-count method, and so
U2F
the combinatorial moments
0.08
0.25
have to be corrected for
accidental correlations. The
0.20
0.06
corrected moments are the
correlation moments.
One last set of moments
0.15
0.04
is frequently used when
analyzing gated-count distributions: the Feynman
0.10
0.02
moments, YqF  Yq / Y1, which
are essentially normalized
correlation moments. These
0
0.50
are so named because during
the Manhattan Project in
–0.02
0
World War II, Feynman,
0
500
1000 1500 2000
2500 3000
3500 4000 4500 5000
Serber, and de Hoffman
Gate width (µs)
did the original work on
neutron time correlations Figure 7. Feynman moments Y (blue) and Y (green) vs. gate width, measured on an HEU object.
2F
3F
while studying neutron fluctuations in the so-called
p, in Eq. 2 to arrive at an estimate for the leakage mul“water boiler” at Los Alamos.18 In their analysis, which
is attributed to Feynman, they defined the quantity that
tiplication ML.
we now call Y2F.
To estimate p and Sf, we use the correlation moments
Finally, the reason why the gated-count distribution,
of Rk(p,,Dt,Sf), which also necessarily depend on p, e,
Rk(p,,Dt,Sf), depends parametrically on the gate width,
Dt, and Sf. The dependence on gate width, Dt, can be
Dt, can be understood by referring to Fig. 4d. Because of
eliminated by working with the asymptotic values of
the spreading of neutron detection times, it is possible
these moments:
that a fission chain occurring within one gate will have
U q ^p,, S f h = lim Yq ^p, , t, S f h
t"3
.
(10)
some of its neutrons detected in subsequent gates. Similarly, it is also possible to count neutrons from fission
These asymptotic values are estimated from the empirichains actually occurring in earlier gates. Clearly, the
cally measured count distributions and equated with
smaller the gate width, the more significant this “leaktheir theoretical expressions. Doing this for the first
age” effect will be. This phenomenon can be observed in
three correlation moments gives us three equations in
the data plotted in Fig. 7, which is a graph of the Y2F and
three unknowns, which can be solved for p, , and Sf.19
Y3F moments versus gate width for measurements made
on an HEU object. As the gates become wider, the relative significance of this effect becomes smaller, and the
INSTRUMENTATION
Y2F and Y3F values approach horizontal asymptotes.
Instruments called multiplicity counters have been
used advantageously in materials accountability and
Putting It All Together
international safeguards applications for a couple of
Ultimately, we wish to find both the leakage muldecades now,9 and they are available commercially
tiplication, which will indicate the presence of fission
(Canberra, http://www.canberra.com; and Ortec, http://
chains, and the total mass of SNM present. We do this
ortec-online.com). These instruments use moderated
by estimating the values of parameters p and Sf from
He-3 neutron detectors and therefore use the gatedthe measured gated-count distribution, Rk(p,,Dt,Sf).
count method for characterizing neutron time correlaThe estimate for the spontaneous fission rate, Sf , along
tions described above. Their design is usually carefully
with the known spontaneous fission half-life and mateoptimized in several respects:
rial density will give us an estimate for the SNM mass.
• Neutron detection efficiency is maximized, because
Yq  q, and so higher-order correlation moments
JOHNS HOPKINS APL TECHNICAL DIGEST, VOLUME 32, NUMBER 5 (2014)
THE PHYSICS OF FISSION CHAINS AND THEIR DETECTION
quickly become immeasurably small at low efficiencies. Most designs aim for 40–60% efficiency, which
is achieved by surrounding the sample with neutron
detectors. Therefore, these geometries are completely closed, with the sample lying in a cavity at
the center of the instrument.
• Variation of neutron detection efficiency versus the
precise location of the sample inside the sample
cavity is minimized.
• Variation of neutron detection efficiency versus neutron energy is minimized. Fission neutrons are emitted with a continuous spectrum of energies, yet the
models described above take into account neither
this fact nor any energy dependence in the detection
efficiency. However, if the instrument’s response is
designed to be independent of neutron energy, this
oversimplification in the model becomes a nonissue.
• Detector dead time is minimized by using a highly
segmented design. The He-3 proportional counter tubes produce electrical pulses that last several
microseconds. If a second neutron enters the tube
during this time, it will not be detected, and this
presents a serious problem. The whole purpose of
these instruments is to detect time-correlated neutrons that arrive in bursts; yet it is precisely these
closely spaced neutrons that will be missed because
of dead time. Multiplicity counters mitigate this
problem by using many (15~50) independent channels of neutron detection, spread around the circumference of the instrument in a form of spatial
multiplexing. Neutrons may be emitted in short
bursts, but they leave the sample traveling in all
directions. Thus, the odds of two neutrons striking
the same channel within a few microseconds of each
other are kept small.
• Immunity to background neutrons from the environment is maximized. The neutron background
can be highly correlated and might fool the instrument into thinking that it is detecting higher levels
of correlations from the sample. To guard against
this problem, the detector can be shielded against
external neutrons—a task made possible by placing
the sample on the inside.
These instruments are used in industrial settings
and can be quite large. For instance they may need
to accommodate a drum of plutonium waste. Consequently, they are not man-portable and are not usable by
a rapid-response deployment team. Furthermore, if such
an instrument were to be used in a field deployment,
the object under test would have to be lifted and placed
inside the instrument—something a team would not do
with a nuclear threat object.
JOHNS HOPKINS APL TECHNICAL DIGEST, VOLUME 32, NUMBER 5 (2014)
In response, man-portable multiplicity counters have
been specially developed. However, the constraints
imposed by portability forces trade-offs in some of the
design optimizations listed above. The most significant
change is that the neutron detectors can no longer surround the sample or object under test. This means that
the detection efficiency will be much lower and uncontrolled, except perhaps in terms of energy dependence.
Moreover, because of the open sample-detector geometry, it is not possible to shield the instrument against
the neutron background.
Ad Hoc Multiplicity Counters
Although deployable, existing man-portable multiplicity counters are single-purpose instruments, and
they necessarily add to the limited equipment load-out
of a rapid-response team. APL is seeking to minimize
this load-out, retaining its core capabilities while adding
multiplicity counting for detecting fission chains.
The underlying approach is to concentrate on
answering the most pressing question—is the threat
object multiplying or not?—and to leave detailed assay
to larger follow-on teams. Concentrating on this more
limited question opens the possibility of using cruder
instrumentation than is needed for traditional multiplicity counting assay.
Rapid-response deployment teams already carry neutron detectors with them, built into radiation backpacks
or as discrete sensors, but these neutron detectors are
configured as simple counters that measure only the
average neutron count rate. Because these neutron
detectors are already part of the load-out, the most convenient way to add neutron time-correlation capability
is to retrofit the existing deployment kit with electronics
that allow operators to gang individual neutron detectors together into a single ad hoc multiplicity counter.
The data from this ad hoc instrument will necessarily
be lower fidelity than those obtained from a purposebuilt assay instrument. Additionally, an operator in the
field will seldom have the luxury of arranging the individual neutron counters optimally, further degrading the
quality of the data. Therefore, novel algorithms are also
needed to take the necessarily low-fidelity data from this
ganged instrumentation and answer the limited yes/no
question on neutron multiplication in the threat object.
In an effort to explore the viability of this approach,
APL has developed the electronics for integrating arbitrary individual neutron counters into an ad hoc multiplicity counting instrument. In collaboration with the
Idaho National Laboratory, APL has also just completed
a measurement campaign in which two representative
ad hoc multiplicity counters were fielded against a multiplying HEU test object. The purpose of this campaign
was to gather data from a wide range of measurement
scenarios to systematically map out the performance
771­­­­
W. A. NOONAN
envelope of these ad hoc multiplicity counters. The test
object20 was configurable so that the k-effective was
variable from approximately 0.4 to 0.7. The test object
was also sometimes configured with neutron reflectors made of steel, tungsten, and HDPE, which further
boosted the k-effective up to 0.84. A radioisotope neutrons source, such as Cf-252, could also be inserted into
the test object to boost the effective spontaneous fission
rate. This allowed us to independently vary the multiplication and the rate at which fission chains were created.
Altogether, varying the source configuration in all these
ways allowed us to cover a wide range of source physics.
In addition, we made measurements over a range of different detector arrangements and detector standoff distances (which affect the all-important neutron detection
efficiency, ).
The data that were collected in this series of measurements provide the foundation for developing binary
classifier algorithms for determining whether or not
the threat object under measurement is supporting significant fission chains. These algorithms will ascertain,
with a stated level of statistical confidence, whether the
object is exhibiting neutron multiplication above some
exigent threshold, and they must be robust against the
marginal data quality expected from an ad hoc instrument operated in less than ideal conditions in the field.
Fast Neutron Detectors
Another class of detectors, called proton recoil scintillators, can detect fast neutrons directly, needing no
moderation. Free of the attendant moderation time,
these detectors have a fast response. Therefore, these
detectors have advantages for measuring neutron time
correlations. Provided that the spontaneous fission
rate is not too large, it is possible with these detectors
to implement the burst-count method for determining
neutron multiplication described first.
Even if the gated-count method is used, the fast
detector response still has the advantage of needing a
shorter measurement time. Because these fast detectors
suffer no moderation time, the Y2F(Dt) and Y3F(Dt) curves
reach their asymptotes at much shorter gate widths. The
statistical error in estimating YqF is proportional to √N,
where N is the number of gate periods for which data
are collected. Thus, specifying the required error fixes
N. Furthermore, reducing the maximum gate width by
using a fast-response detector also reduces the total measurement time, T = NDt.
Proton recoil scintillators are organic materials that
are rich in hydrogen. When an incident neutron collides
with one of these hydrogen nuclei, enough kinetic energy
is imparted to it that it is ejected from its parent molecule.
This recoil proton then traverses the scintillator, leaving a trail of ionization and molecular excitations in its
wake. These excited molecules subsequently decay back
772
to their ground states, emitting photons and producing a
light pulse. The light pulses are then collected and converted into an electrical detection pulse by a high gain
photodetector, such as a photomultiplier tube.21
These scintillators are also sensitive to gamma-rays
via a parallel mechanism: incident gamma-rays Compton scatter off electrons in the scintillator and eject
them from their parent molecule. Like the recoil protons, these Compton electrons also leave a path of ionization and molecular excitations.
The scintillation pulses produced by recoil protons
decay more slowly than the pulses produced by Compton electrons. This difference in pulse length forms the
basis of pulse shape discrimination for distinguishing
incident neutrons from incident gamma-rays. Unfortunately, pulse shape discrimination is not error free, and
one gamma-ray in 103 or 104 may be misclassified as a
neutron. Since the gamma flux is often much higher
than the neutron flux, then the number of misclassified
gamma-rays can be a substantial fraction of the neutron counts, leading to errors. By comparison, only one
gamma-ray in 106 or 107 causes a false neutron count in
moderated He-3 detectors.
Not all scintillators exhibit a significant difference
in the pulse lengths for neutrons versus gamma-rays.
Unfortunately, scintillators that have good pulse shape
discrimination properties typically do not lend themselves to building rugged deployment gear for many reasons. They are either liquids or fragile crystals. Most are
flammable, many with low flash points, and they can be
carcinogenic. However, recently there has been renewed
interest in plastic scintillators capable of pulse shape
discrimination.22, 23 Furthermore, a plastic developed
by Lawrence Livermore National Laboratory24 is now
available commercially (Eljen Technology, http://www.
eljentechnology.com). Plastic has many advantages for
making deployable gear. It is not flammable; being plastic, it is intrinsically rugged; and it can be molded into
any size or shape.
Therefore, APL is investigating these new plastics
for potential use in field deployable systems. However, it
remains to be seen whether the benefits of fast response
outweigh the downside of gamma misclassifications that
come with pulse shape discrimination.
SUMMARY
In this current period of increased nuclear proliferation risks, the United States is implementing a multilayered defense aimed at interdicting a smuggled nuclear
device. If a potential threat object is found, it is imperative that the exact level of threat be assessed quickly.
Because the presence of fission chains is hard evidence
of either an actual nuclear device or a significant quantity of SNM, detecting fission chains is invaluable for
this threat assessment.
JOHNS HOPKINS APL TECHNICAL DIGEST, VOLUME 32, NUMBER 5 (2014)
THE PHYSICS OF FISSION CHAINS AND THEIR DETECTION
The technology for detecting fission chains has been
in use for several decades for various applications and
has roots going back to the Manhattan Project in World
War II. However, the standard instrumentation does not
lend itself to field deployment by a small, rapid-response
team that might be dispatched to assess a threat. APL
is exploring several avenues to bring to the field better
neutron time-correlation techniques for detecting fission chains.
ACKNOWLEDGMENTS: The neutron time-correlation effort
at APL has benefited from the efforts of many APL staff
members, including Sungshan (Cliff) Chiang, Jeffrey
Kalter, Christopher Lavelle, Eric Miller, Wayne Shanks, and
Erin van Erp. The test data on HEU was collected at the
Idaho National Laboratory in collaboration with David
Chichester, Mathew Kinlaw, and Scott Watson. The author
thanks Eric Miller for the data reduction and analysis that
went into Fig. 5b. The author also thanks Stanley Puchalla
and Robert Mayo for many hours of helpful discussions
and for reading various drafts of the manuscript. Any
errors, however, are entirely the author’s.
REFERENCES
1Frantz,
D., and Collins, C., The Nuclear Jihadist: The True Story of the
Man Who Sold the World’s Most Dangerous Secrets . . . And How We
Could Have Stopped Him, Twelve, Hachette Book Group, New York
(2007).
2Wolf, A. F., Nonproliferation and Threat Reduction Assistance: U.S.
Programs in the Former Soviet Union, Congressional Research Service
Report no. RL31957 (6 Mar 2012).
3Montgomery, E. B., “Nuclear Terrorism: Assessing a Threat, Developing a Response,” Strategy for the Long Haul Series, Center for Strategic and Budgetary Assessments, Washington, DC (2009).
4Masse, T., Nuclear Terrorism Redux: Conventionalists, Skeptics, and the
Margin of Safety, Johns Hopkins University Applied Physics Laboratory, Laurel, MD (2009).
5Combating Nuclear Smuggling: Megaports Initiative Faces Funding and
Sustainability Challenges, United States Government Accountability
Office Report no. GAO-13-37 (Oct 2012).
6SLD Implementation Strategy: Revision B, April 2006, Office of the
Second Line of Defense, National Nuclear Security Administration
(2006).
7Shea, D. A., The Global Nuclear Detection Architecture: Issues for
Congress, Congressional Research Service Report no. RL34754
(25 Mar 2009).
The Author
8Verbeke,
J. M., Hagmann, C., and Wright, D., Simulation of Neutron and Gamma Ray Emission from Fission, Report no. UCRLAR-228518, Lawrence Livermore National Laboratory, Livermore,
CA (2009).
9Ensslin, N., Harker, W. C., Krick, M. S., Langner, D. G., Pickrell,
M. M., and Stewart, J. E., Application Guide to Neutron Multiplicity
Counting, Report no. LA-13422-M, Los Alamos National Laboratory,
Los Alamos, NM (1998).
10Böhnel, K., “The Effect of Multiplication on the Quantitative Determination of Spontaneously Fissioning Isotopes by Neutron Correlation Analysis,” Nucl. Sci. Eng. 90(1), 75–82 (1985).
11Prasad, M. K., and Snyderman, N. J., “Statistical Theory of Fission
Chains and Generalized Poisson Neutron Counting Distributions,”
Nucl. Sci. Eng. 172(3), 300–326 (2012).
12Pazsit, I., and Pal, L., Neutron Fluctuations: A Treatise on the Physics of
Branching Processes, Elsevier, New York (2008).
13Enqvist, A., Pazsit, I., and Pozzi, S., “The Number Distribution of
Neutrons and Gamma Photons Generated in a Multiplying Sample,”
Nucl. Instr. Methods A 566(2), 598–608 (2006).
14Grieder, P. K. F., Cosmic Rays at Earth: Reference Manual and Data
Book, Elsevier Science Ltd., Amsterdam (2001).
15Kingman, J. F. C., Poisson Processes, Oxford University Press, New
York (1993).
16Hage, W., and Cifarelli, D. M., “Correlation Analysis with Neutron
Count Distributions in Randomly or Signal Triggered Time Intervals
for Assay of Special Fissile Materials,” Nucl. Sci. Eng. 89(2), 159–176
(1985).
17Uhrig, R. E., Random Noise Techniques in Nuclear Reactor Systems,
The Ronald Press Co., New York, pp. 60–65 (1970).
18Feynman, R. P., de Hoffman, F., and Serber, R., Intensity Fluctuations of a Neutron Chain Reaction, Report no. LA-256, Los Alamos
National Laboratory, Los Alamos, NM (1944).
19Cifarelli, D. M., and Hage, W., “Models for a Three-Parameter Analysis of Neutron Signal Correlation Measurements for Fissile Material
Assay,” Nucl. Instr. Methods A 251(3), 550–563 (1986).
20Chichester, D., and Kinlaw, M., “The MARVEL Assembly for Neutron Multiplication,” Appl. Radiat. Isot. 80, 42–48 (2013).
21Knoll, G. F., Radiation Detection and Measurement, 3rd ed., John Wiley
& Sons, New York, pp. 553–565 (2000).
22Hamel, M., Blanc, P., Rocha, L., Normand, S., and Pansu, R.,
“Study and Understanding of n/ Discrimination Processes in
Organic Plastic Scintillators,” in Chemical, Biological, Radiological, Nuclear, and Explosives (CBRNE) Sensing XIV, Augustus Way
Fountain III (ed.), Proc. of SPIE, Vol. 8710, SPIE, Bellingham, WA,
87101F (2013).
23Hutcheson, A., Simonson, D. L., Christophersen, M., Phlips, B. F.,
Charipar, N. A., et al., “Neutron/Gamma Pulse Shape Discrimination
(PSD) in Plastic Scintillators with Digital PSD Electronics,” in Chemical, Biological, Radiological, Nuclear, and Explosives (CBRNE) Sensing
XIV, Augustus Way Fountain III (ed.), Proc. of SPIE, Vol. 8710, SPIE,
Bellingham, WA, 87101K (2013).
24Zaitseva, N., Rupert, B. L., Pawelczak, I., Glenn, A. H., Martinez, P.,
et al., “Plastic Scintillators with efficient Neutron/Gamma Pulse
Shape Discrimination,” Nucl. Instr. Methods A 668, 88–93 (2012).
William A. Noonan is a member of the APL Principal Professional Staff. Originally trained as a plasma physicist, he
worked as a researcher at the Naval Research Laboratory and the University of Maryland. He also did a stint in the MEMS
industry before joining the Department of Energy Remote Sensing Laboratory to serve on nuclear/radiological emergency
response teams. He joined the APL technical staff in 2007. His e-mail address is william.noonan@jhuapl.edu.
The Johns Hopkins APL Technical Digest can be accessed electronically at www.jhuapl.edu/techdigest.
JOHNS HOPKINS APL TECHNICAL DIGEST, VOLUME 32, NUMBER 5 (2014)
773­­­­
Missile Concept Optimization for Ballistic
Missile Defense
Alan J. Pue, Richard J. Hildebrand, Daniel E. Clemens, Jonah R. Gottlieb,
James M. Bielefeld, and Timothy C. Miller
fundamental trade in the development of a Ballistic Missile Defense interceptor missile concept is the quality of the engagement support system that provides
threat state estimates versus the interceptor kinematic and terminal guidance capabilities. This article describes a methodology to optimally balance the conflicts among
system performance goals, engagement support capabilities, and missile technology
constraints. A key factor is the ability of the missile and the final-stage “kill vehicle” to
remove the system errors caused by engagement support threat tracking and interceptor fly-out errors. The kill vehicle divert maneuver and guidance capabilities must be
sized to remove system errors and accomplish intercept, but this need conflicts with the
objective to maximize the missile kinematic capability to reach potential threat trajectories. An optimized missile configuration minimizes the time from launch to intercept
while ensuring the seeker and divert maneuver capabilities needed to remove system
errors. The methodology described here provides a unified high-fidelity approach to
missile concept development.
INTRODUCTION
Air and missile defense weapon system designs are
typically based on an architecture that integrates an
acquisition and tracking sensor used for fire control, a
battle management command and control system, and
guided missiles. Figure 1 illustrates the major engagement components and events that may comprise a
Ballistic Missile Defense (BMD) engagement. The
774
detection and tracking of a threat missile may involve
several sensors, which could be space, land, and/or sea
based. Remote sensor data may be received, processed,
and transmitted to a launch platform via a battle management, command, control, and communications
(BMC3) node. Upon determination of an engagement
solution, the defending missile is launched and guided
JOHNS HOPKINS APL TECHNICAL DIGEST, VOLUME 32, NUMBER 5 (2014)
Figure 1. Notional BMD engagement.
on an intercept path until a kill vehicle (KV) is released
for the final phase of the mission. The seeker on the
KV acquires the threat lethal object and establishes a
track for guidance. The KV’s divert system removes the
trajectory errors that remain after the earlier portions of
flight and responds to guidance commands derived from
seeker measurements. For exoatmospheric intercepts,
the onboard sensor is typically an IR seeker, whereas a
RF seeker is often used for endoatmospheric intercepts.
The development of a new weapon system concept
typically involves a series of trades that derive fire control sensor, BMC3, and missile key performance parameters. These performance parameters characterize, for
example, the detection range and track accuracy of
the fire control sensor, communications time delay, the
time of flight of the missile, and the ability of the missile terminal guidance to remove system errors. Depending on the particular weapon system needs, the tradable
parameters may include all aspects of the fire control
sensor, BMC3, and missile. In some cases, only missile
parameters may be tradable as constrained by an existing launch system. In all cases, the starting point is the
definition of a mission, the threat characteristics, oper-
JOHNS HOPKINS APL TECHNICAL DIGEST, VOLUME 32, NUMBER 5 (2014)
ating constraints such as potential missile launch locations, and the portions of the system that must remain
unchanged because of programmatic decisions.
Given a definition of the system context, the system
engineering challenge is to appropriately balance the
choice of key performance parameters with technology constraints, technical risk, program schedule, and
cost. In addition, a system architecture that includes the
participating components and the interfaces must be
defined. The combination of key performance parameters and system architecture constitutes a system concept
that ultimately evolves to a set of system and component
requirements as more detailed analyses are conducted.
System concepts are often developed in an iterative
fashion, with relatively simple performance models of
system components. Depending on the operational
context, measures of effectiveness are selected to ensure
that the derived concept meets an operational need. For
example, a common BMD measure is the defended area
versus the threat launch area denied. The ship operating area or locations of land-based missile launchers are
other important operational considerations. A system
performance model can then be applied to determine
775­­­­
A. J. PUE ET AL.
collection of possible intercept points will be referred
to as the “kinematic battle
space.” The kinematic capaLaunch sites
Remote sensors
bility of the missile is mostly
constrained by missile mass
and dimension limits defined
Launcher
BMC3
by the launcher system. Given
constraints
these constraints, the booster
Missile mass and
Time to launch Quality of service
can be optimized to minimize
size allowance
time of flight to a space point
as a function of staging, conBooster
Kinematic reach
trol options, and KV mass.
In addition to the basic kineKV mass allowance
matic capability of the misKV size allowance
Error
sile, kinematic reach is also
KV
Handover error containment
constrained by the minimum
allocation
timeline to missile launch
Communications
after threat launch, which is
dictated by remote sensor and
DACS*
Control
Seeker
BMC3 capabilities.
* Divert and attitude control system
“Error containment” refers
System context
to the ability of the misMissile configuration
Lethality
sile to remove the predicted
Missile Trade Space
Missile performance
intercept point error caused
by sensor track error, system
time delays, and missile guidFigure 2. BMD system and missile dependencies.
ance and navigation errors.
the system parameters that provide the needed system
It also includes the missile pointing error, which must
performance while satisfying technology capability
be contained by the seeker field of view (FOV) for inilimits. Because of the many parameters involved, this
tial acquisition of a threat object. Some of these errors
can become a very complex and iterative process that
can be managed by the upper stage of the booster stack,
but much of the error, called handover error, must be
crosses many disciplines.
removed by the KV. The required probability of error
This article describes an integrated multidisciplinary
containment is allocated from an overall probability of
optimization approach that is based on a flow-down of
single-shot kill.
the key component characteristics from the operational
The final performance attribute, “lethality,” refers to
needs and constraints. The focus is on the defending
the endgame guidance accuracy of the KV to effect a
missile. The other portions of the system are treated in
lethal hit. The probability of lethal hit is also allocated
a simpler way, but the approach covers all of the critical
from a total probability of single-shot kill.
performance parameters across the entire system.
Figure 2 illustrates the basic dependencies for the
missile concept development. The missile trade space
MISSILE CONCEPT OPTIMIZATION
is bounded by a system context definition including the
mission, threat, launch sites, launcher constraints, sensor
Paralleling the performance attributes in Fig. 2, opticapabilities, and BMC3 assumptions. The major compomization of the missile concept can be separated into
nents of the missile concept are (i) the booster propulthree loosely coupled subproblems as shown in Fig. 3: (i)
sion stack, (ii) the control systems used for each booster
kinematic reach, (ii) error containment, and (iii) lethalstage, (iii) the communications with the ground weapon
ity. Decoupling is possible because optimization of the
control system, (iv) the KV propulsion for divert and attibooster configuration to minimize flight time depends
tude control, and (v) the KV seeker. Each of these major
on KV mass but not the specific KV configuration.
components is described by a set of key performance
Kinematic reach is the most fundamental performance
parameters, which are discussed in later sections.
criterion because the threat trajectory must be within
Besides the five major missile components, Fig. 2 also
both the missile range and speed capabilities for the
highlights the principal performance attributes. “Kineintercept to be possible. For specified missile launcher
matic reach” refers to the ability of the missile to reach a
and system timeline constraints, the booster can be optithreat in time and space after launch of the threat. The
mized to maximize the reach to threat trajectories for a
Mission and threat space
776
JOHNS HOPKINS APL TECHNICAL DIGEST, VOLUME 32, NUMBER 5 (2014)
SYSTEM OPTIMIZATION FOR BMD MISSILE CONCEPT DEVELOPMENT
able missile size and mass trade
space given launcher constraints.
Launcher
For each of these booster conconstraints
cepts, the mass and volume of
KV
Kinematic
Error
KV
Lethality
mass
reach
containment
the KV is allowed to vary paraconfiguration
limit
metrically. Once the configuraBooster
Seeker
KV
DACS
Seeker mass
KV
configuration
resolution
tions are developed, a coverage
mass
mass and ∆V
and aperture
acceleration
analysis identifies the maximum
KV mass that can be tolerated
Optimized
Optimized
Optimized
for each concept and still meet
booster
KV
endgame
the threat trajectory engagement
goals. This analysis will establish
Figure 3. Key missile trade flow.
missile kinematic and KV mass
KV mass limit. Optimization of the booster configurathresholds for each missile concept. The missile kinetion for a given KV mass is discussed in the next section.
matic threshold can be expressed in terms of a minimum
Given kinematic reach, the system errors must be
booster burnout velocity (Vbo).
removed to intercept the threat. A larger KV mass
degrades missile kinematic reach but allows more capable
Booster Optimization
seeker and divert and attitude control system (DACS)
The booster concept is developed with a multi­
capabilities to remove handover error. Thus, the second
disciplinary
system-level missile design optimization tool
optimization is to minimize KV mass while still achieving
called
ORION
(Optimization of Rockets for Intercept
error containment. This optimization mostly becomes a
OperatioNs),
which
was written at the Johns Hopkins
trade between the DACS mass and the seeker mass. DACS
University
Applied
Physics
Laboratory (APL). ORION
mass translates to KV divert maneuver performance (i.e.,
integrates
physics-based
and
empirically benchmarked
acceleration or velocity change, V), whereas seeker mass
models
of
propulsion,
aerodynamics,
payload packagtranslates to seeker acquisition range. As seeker acquisiing,
and
vehicle
kinematics
for
singleor
multi­objective
tion range is increased, less divert performance is needed
booster
optimization
and
relies
primarily
on genetic
because more time is available to remove error.
algorithms to determine the optimal solution.
There is some coupling between the booster and KV
Modern computational resources have now enabled
optimization problems. The booster and KV capabilities
multidisciplinary, system-level analysis and design optiare both optimized when the kinematic reach and error
mization.1 In a multidisciplinary design optimization
containment are brought into balance. If error containapproach, complex system models are developed by
ment cannot be achieved for certain trajectories given
integrating detailed models of various subsystems early
the KV mass limit, then some kinematic reach may need
in the design phase. Subsystem design parameters are
to be sacrificed to bring the concept into balance. The
then varied, with their interactions observed at the
goal is to ensure that errors are contained for all potensystem level, leading to truly optimized system designs.
tial intercept points. Conversely, if all of the mission
ORION, a multidisciplinary design optimization tool
threat trajectories are reached with excessive containfor missile propulsion systems, provides the capability
ment margin, then other portions of the system design
to comprehensively observe the impacts of missile subsuch as engagement support quality or KV mass and missystem interactions earlier in the design evolution than
sile size might be relaxed.
previously possible.
Once a basic KV configuration is determined, endORION integrates detailed models for propulsion,
game lethality depends on the ability to determine and
nosecone design and payload packaging, aerodynamics,
steer out remaining guidance errors. This is mostly a
and vehicle kinematics to facilitate system-level optitrade between seeker resolution and acceleration capamization of the missile component design parameters.
bility of the KV. The result of this trade can affect the
Coupling of the integrated missile model to various optiKV optimization because both the seeker resolution
mization algorithms provides the capability for multiand KV acceleration parameter selections might influence KV mass, which may require a rebalance of the KV
variable, multiobjective optimization. Figure 4 illustrates
DACS and seeker capabilities.
the basic approach used in ORION and a three-stage
missile example. The optimization variables include the
length, Lg, outer diameter, Dgo, and inner diameter, Dgi,
KINEMATIC REACH
of the propellant grains for each stage; the length, Ln,
exit diameter, De, and throat diameter, Dt, of the nozzles;
To establish missile kinematic performance requirethe length, Lnose, and radius, Rnose, of the nosecone; and
ments for a given mission, the first step is to develop
several optimized booster concepts that span the allowthe outer diameter, DGS, of the guidance system. The
Threat trajectories, missile
launch locations, and
engagement support quality
JOHNS HOPKINS APL TECHNICAL DIGEST, VOLUME 32, NUMBER 5 (2014)
777­­­­
A. J. PUE ET AL.
Variables
Propulsion
• Propellant grain geometry,
dimensions, and burn rates
• Nozzle length, throat and exit diameter
Objective
Optimized
configuration
Minimize time of flight
• Vertical trajectory
Nosecone and payload
• Nosecone length and bluntness
• Guidance system outer diameter
Lnose
Lg, 3
Dt, 3
De, 3
Dt, 2
Lg, 2
De, 2
Lg, 1
Dt, 1
De, 1
Rnose
DGS
Dgo, 3
Dgi, 3
Ln, 3
Dgi, 2
Dgo, 2
Ln, 2
Dgi, 1
Dgo, 1
Ln, 1
Figure 4. ORION booster optimization.
total number of design variables is 15 for a two-stage
missile and 21 for a three-stage missile.
Propulsion Model
The propulsion system model uses physics-based
and empirically benchmarked calculations to provide
the capability for medium-fidelity stage and motor
characterization. A solid rocket motor modeling and
design tool developed by APL, called ARIES (Analysis of Rockets for Initial Exploratory Studies), is used for
this purpose. The motor/stage model accepts an input
list whose components generally fall into one of four
categories: (i) propellant; (ii) nozzle assembly; (iii) case
assembly; or (iv) stage assembly. The inputs consist of
propellant formulation and ballistic properties, as well
as certain dimensions, masses, and material properties
of various components. ARIES calculates motor component dimensions and masses and, in addition, calculates motor interior ballistics, producing time traces of
chamber pressure, thrust, and expelled propellant mass.
ARIES primarily follows textbook principles for design
calculations and performance predictions. Motor performance is calculated on the basis of a lumped-parameter
ballistics code developed at APL.
Nosecone and Aerodynamics
The nosecone model in ORION allows the user to
choose from five different standard nosecone shapes:
conical, tangent ogive, Kármán ogive, LV-Haack, and
power law. Nosecone length, base diameter, bluntness,
thickness, and material density are input. ORION then
solves for the outer mold line, surface area, and mass of
the nosecone. Thermal analysis is performed separately
to ensure the nosecone provides adequate thermal protection of the KV.
Aerodynamic calculations are performed using Missile DATCOM, an industry standard for preliminary
778
aerodynamic analysis. The geometry and dimensions
of the nosecone and propulsion stages, as well as fins
or other aerodynamic surfaces that exist in the design
space, are flowed to DATCOM. DATCOM then returns
sets of tables of aerodynamic coefficients (e.g., axial or
normal force coefficient versus Mach number) for use in
the flight model.
Optimization Algorithms
To perform optimization, ORION primarily relies
on a genetic algorithm (GA), which is a non-gradientbased method capable of evaluating continuous and
discontinuous systems. GAs have been shown to be successful in the single- and multiobjective design optimization of various types of aerospace systems and their
components. GAs have been applied to the design of
airfoil and nosecone shapes,2 unmanned aerial vehicles,3 spacecraft orbital dynamics,4 hybrid rockets, space
launch vehicles,5 and missiles.6 The GA does not use
gradients to determine the direction of parameter variation for a single design point but instead evaluates a
population distributed over the entire parameter space.
Thus, compared with gradient-based optimization algorithms, the GA is more likely to avoid local optima and
find the global optimum. However, the evaluation of so
many more points leads to increased convergence times.
In addition, because the process of population evolution
is stochastic, the exactness of the converged solution, as
well as repeatability of the algorithm execution, cannot
be guaranteed. Each time the algorithm is executed, a
slightly different solution could be obtained.
In addition to the GA, the pattern search algorithm (PSA), another nongradient-based optimization
method, can also be used. The PSA requires a starting
point and evaluates multiple nearby design points to find
one where the objective function is improved compared
with the current point. That point then becomes the
JOHNS HOPKINS APL TECHNICAL DIGEST, VOLUME 32, NUMBER 5 (2014)
SYSTEM OPTIMIZATION FOR BMD MISSILE CONCEPT DEVELOPMENT
starting point for the next iteration, and the process is
repeated until no improvement is observed. Compared
with the GA, the PSA typically requires fewer function
evaluations to converge to a solution, leading to faster
convergence times, and the solution is sometimes more
accurate. However, the PSA is sensitive to the choice of
initial point and can be less likely to find global optima
than the GA.
Hybrid algorithms can combine the advantages of
various algorithms while avoiding some of their dis­
advantages. In ORION, a hybrid algorithm, consisting
of a combination of the GA and PSA, can also be executed. In this hybrid, the GA is executed first and the
result becomes the starting point for the PSA. In this
sense, the GA, which is highly likely to converge to a
point very close to the global optimum, is used to find a
very good starting point for the PSA, which is in most
cases capable of higher local accuracy. This type of strategy is excellent for highly nonlinear systems with many
variables in which many local optima exist.
Inert Mass Growth
Following typical U.S. engineering practices, inert
mass growth is observed during the evolution of aerospace
system designs, and this growth adversely affects the performance of those systems. Frequently, initial estimates
of component masses are aggressively low, resulting in
overly optimistic preliminary performance estimates. As
the component and system designs mature and designers
better understand the impacts of the component requirements, mass inevitably increases and is accompanied by
decreasing performance. To avoid overly aggressive performance estimates, mass growth projections must be
included early in the design. To account for inert mass
growth, ORION uses a standard developed by the American Institute of Aeronautics and Astronautics (AIAA).7
In the methodology provided by this standard, mass
growth allowance is included at any time in the system
design phase by accurately assessing the current maturity
of the individual components and assigning a certain
percentage of additional mass to the current component
mass estimate. The value of the percentage depends on
both the type of component (i.e., electrical, structural,
thermal, propulsion, etc.) and the maturity of that component (i.e., early estimate, layout, preliminary design,
released design, existing hardware, and actual mass). This
mass growth allowance accounts for component mass
increases necessary to meet existing requirements. However, the standard also calls for the inclusion of additional
mass margin to account for potential omissions or revisions of existing requirements. The assigned percentages
for mass growth allowance and mass margin decrease as
the program matures. ORION accounts for inert mass
growth by assigning a user-input percentage increase to
the masses determined by physical calculations.
JOHNS HOPKINS APL TECHNICAL DIGEST, VOLUME 32, NUMBER 5 (2014)
Boost Control
The design of the system used to maintain airframe
stability throughout flight has a significant impact on
the overall missile concept. Traditional types of control
systems include aerodynamic surfaces, attitude control
systems (ACSs), thrust vector control systems, or some
combination thereof. The level of control required will
determine the actuator type, size, and mass, which in
turn will impact the overall missile kinematic performance. Thus, control system design is coupled to the
booster optimization process described above.
Key events that drive the design of the control system
are stage separation, coast periods, and upper-stage
maneuvers. A stage separation occurs when a spent stage
separates from the rest of the missile, which induces
destabilizing conditions in the form of tip-off angles and
angular rates. The control system must maintain airframe control during stage separation. This function is
called capture.
First-stage separation occurs during a portion of flight
when the dynamic pressure is near its highest and is particularly stressing on the actuators. They must provide a
high force in a short time to maintain stability. This is
especially important if the airframe is aerodynamically
unstable, because unchecked tip-off angles and rates can
quickly lead to a loss of control. Other events that drive
control system requirements are wind gust disturbance
rejection and upper-stage maneuvers such as nosecone
deployment.
Both operating conditions and key event times are
needed to define the control system requirements. Missile flight is simulated to characterize each event, and
analyses are parametrically run to develop control performance requirements. A control system is chosen and
sized for each stage, and its mass is added to the missile in the appropriate locations. This mass addition will
change the missile flight characteristics, which in turn
modify the control requirements, making the determination of control system requirements a highly iterative
and coupled problem.
Once the control requirements are established, it is
then necessary to parametrically define the control performance requirements and associated hardware size and
mass versus dynamic pressure. With those relationships
defined, ORION can generate missile staging concepts
that reflect a properly controlled missile.
An example of this coupled process can be seen on
the sizing of the second-stage ACS. Total impulse is
driven by the control required during capture and the
second-stage coast. The second-stage coast occurs after
the motor has burned out, and the ACS must provide
control without the aid of the thrust vector control
system. For each possible missile trajectory, coast time
and the corresponding required impulse is determined.
Combined with the impulse required for capture, ACS
mass is estimated as a function of total impulse. That
779­­­­
A. J. PUE ET AL.
Velocity (km/s)
Altitude (km)
2-stage
3-stage
41˝
38˝
70˝
54˝
g-Load
Dynamic pressure (psf)
266˝
266˝
79˝
155˝
95˝
0
20
40
60
80
Flight time (s)
100
120
0
20
40
60
80
Flight time (s)
100
120
22.5˝
22.5˝
Figure 5. Vertical flight model results for the two- and three-stage optimized concepts.
relationship is used in ORION to model the interaction
of missile performance and control system design in the
booster optimization.
Booster Optimization Example
In this example, the dimensions of the rocket motor
propellant grains, nozzles, nosecone, and payload were
considered to be the primary drivers of size and performance. These parameters were optimized with the
objective of maximizing Vbo. Two- and three-stage missiles are considered. For each stage, the composite motor
case, the submerged contoured nozzle, the ACS, and the
thrust vector control system are assumed. The propellant
is assumed to be a conventional aluminized composite
with ballistic properties typical of such propellants and
the grain geometry assumed to be an internal burn tube.
Figure 5 shows a comparison of flight performance
for the two- and three-stage variants of a 22.5-in.diameter configuration. Although the two configurations produced similar results for Vbo, the three-stage
variant has lower peak dynamic pressure and lower axial
acceleration. Consideration of lower-level requirements
and kinematic performance throughout the battle space
may lead to one choice or the other.
ERROR CONTAINMENT
Referring to Fig. 3, the optimized booster
configuration and associated kinematic analysis provides
a KV mass limit, which allows the interceptor to meet
range and time-of-flight requirements. The next step
in the optimization process is to balance the capability
between the IR seeker (and associated avionics package)
and a propulsion system that provides the maneuver
capability.
780
DACS operation and capability are driven by several
key events determined by the KV release time and seeker
functions. These events can include a divert maneuver
using remote sensor data before the operation of the
seeker, a divert maneuver after seeker acquisition, a
divert maneuver after seeker discrimination, and a divert
maneuver just before intercept. Of these divert events,
the discrimination divert can be the most demanding
and can be reduced by the incorporation of a more capable IR seeker. One of the critical factors in the concept
design is the allocation of impulse to the possible divert
maneuvers.
KV Configuration
The most commonly used KV configuration consists
of a hard-mounted seeker and a cruciform DACS. The
divert system provides the lateral motion for the KV, and
the ACS provides the angular control to stabilize seeker
pointing and control divert direction. The design of the
ACS can be simplified if the center of gravity of the KV
is aligned with the divert thrusters and remains aligned
throughout operation. This can generally be achieved
by positioning some of the avionics components aft of
the DACS. This is called a split KV configuration as
opposed to a unitary layout. Here the trade is between
DACS and KV packaging complexity. The split KV
design is attractive because the DACS is often the highest-risk assembly on the KV, and the entire missile. An
example configuration is illustrated in Fig. 6.
DACS Constraints
For the traditional DACS, the two primary propellant options are hypergolic liquids and solids. Hypergolic
propellants typically consist of a fuel and an oxidizer
JOHNS HOPKINS APL TECHNICAL DIGEST, VOLUME 32, NUMBER 5 (2014)
SYSTEM OPTIMIZATION FOR BMD MISSILE CONCEPT DEVELOPMENT
Attitude
control
Seeker
Divert
thrusters
Figure 6. Example KV configuration.
that spontaneously ignite when they come into contact
with each other. In addition, they are extremely toxic
and/or corrosive, making them difficult to handle. Thus,
liquid fuels have handling and safety concerns, which
lead to higher infrastructure and leakage mitigation
costs. On the other hand, a liquid-propellant DACS can
be designed to ignite reliably and repeatedly, and it is a
relatively mature technology.
There are four major types of solid-propellant DACSs
(SDACS). The first is an extinguishable system, which
can be stopped and started as required. Among the
options, the extinguishable system is the least mature
technology (lowest technology readiness level, or TRL)
and highest risk. The second type SDACS uses multiple pulses. In this system, two or more divert pulses
are contained in a single pressure vessel. This design is
a medium TRL and risk option. The third option is a
modular multiple gas generator design. The generators
can be fired in pairs for each divert event to keep the
center of gravity aligned with the divert plane as discussed in the previous section. For example, three divert
events require six gas generators. This approach has a
higher TRL and lower risk than the first two options.
One of the drawbacks of this design is the low packaging efficiency, which results in a larger DACS space
envelope compared with the other options. The fourth
type, throttleable SDACS, is similar to the extinguishable system except the thrust can only be turned down
to a lower level rather than completely turned on and
off. This type of system has the highest TRL and lowest
risk, but that can depend on the specific requirements.
The final selection of a DACS configuration depends on
the required operating time, divert capability, and mass
while considering risk and cost.
Seeker Constraints
The IR seeker detects, acquires, and tracks objects of
interest and selects which object should be intercepted.
At a basic level, the IR sensor consists of optical com-
JOHNS HOPKINS APL TECHNICAL DIGEST, VOLUME 32, NUMBER 5 (2014)
ponents that focus IR radiation, which is emitted or
reflected from distant threats, onto an array of IR sensor
elements, or pixels, that make up a focal plane array
(FPA). There are several sensor and threat properties, or
parameters, that will affect the design of the IR sensor.
These parameters include:
• Aperture: The physical aperture diameter is the
diameter of the IR sensor. A larger aperture will
improve seeker performance for two reasons. First,
more IR radiation will be accepted into the sensor if
the aperture is larger, increasing sensitivity. Second,
the ability of an optical system to focus radiation
onto a small spot will improve with larger aperture,
so seeker resolution will also improve with larger
aperture. However, a larger aperture will require a
larger and more massive seeker and thereby a more
massive KV. Note the design of the optical system
may cause blockage, which reduces the effective size
of the aperture.
• Waveband: The IR sensor detects radiation in the
IR region of the electromagnetic spectrum, which
extends from ~2 microns to a few tens of microns.
Threats will emit IR radiation according to the
blackbody radiation equation, with the wavelength
of the peak of the emission spectrum depending
on the threat temperature. Colder threats will emit
radiation that peaks at longer wavelengths. For
example, room temperature threats, around 300 K,
will emit a spectrum that peaks around 8–9 microns,
so threat properties must be considered in selecting
the seeker operating waveband.
• FOV: The FOV is the angular extent observed by the
seeker. A wider FOV allows the seeker to simultaneously observe objects with increased spacing or to
find an object with increased location uncertainty.
The former capability will affect the time at which
threat selection must be accomplished, whereas the
latter will affect the capability of the interceptor to
contain a threat within its FOV at acquisition.
• Instantaneous FOV (IFOV): The IFOV is the
angular width observed by a single pixel of the sensor
array. A smaller IFOV is generally better because it
will allow increased resolution, which will enable
the seeker to resolve multiple threats earlier, allowing more time for endgame guidance.
• Number of pixels or FPA format: For a square
array, the number of pixels in one dimension is given
by the FOV divided by the IFOV: Npixels = FOV/
IFOV. Because it is desired to maximize FOV and
minimize IFOV, a large number of pixels is advantageous. However, very-large-format IR arrays are more
expensive to manufacture, so the maximum number
781­­­­
A. J. PUE ET AL.
• Noise sensitivity: The ability of an IR sensor to
detect a given threat at a desired range will depend
on its sensitivity, which is determined by its noise
characteristics. There are several sources of noise
that affect IR sensors, including shot (photon) noise,
thermal emissions of the seeker mechanical and
optical parts, detector dark current, stray light in the
optical path, amplifier and readout noise, defective
detector pixels, and quantization noise.
There are several figures of merit that are typically
used to describe noise or sensitivity for IR sensors. One
common measure is the noise equivalent irradiance
(NEI), which is the flux density at the entrance of the
optical system that produces an output with a signal-tonoise ratio (SNR) equal to 1.0 (output signal = system
noise). NEI characterizes the sensitivity of the sensor
system to a point (unresolved) source. NEI is expressed
in watts per square centimeter or photons per second
per square centimeter, and a lower number is better. To
remove aperture dependency from the noise figure, a
normalized parameter, NCA = (NEI × clear aperture),
may be used.
Seeker Acquisition Range
782
D0
DC
DTV
DTVS
Number of frames
summed
No translation and vibration motion
Translation present, no vibration motion
Translation and vibration motions
With translation and vibration motions,
and frame summing
End here
with aperture
Object radiance
(W/sr)
Operation range, R (km)
Irradiance, L (W/cm 2)
DT (cm)
DT (cm)
Translation rate
(pixel/frame)
Irradiance, L (W/cm 2)
To optimize the KV configuration, the seeker performance
parameter trade space must be
characterized and related to the
mass of the seeker. The relationships between the key seeker
parameters, which determine
acquisition range, are shown in
nomograph form in Fig. 7.
One begins at the lower left
by specifying range requirements
for acquisition and discrimination and ends at the upper right
with an aperture requirement
to meet the required ranges. To
move from the start to the end
of the Fig. 7 nomogram requires
five steps.
In the first step, shown in
the lower-left graph, the irradiance at the seeker aperture is
calculated at the required range
versus threat radiance. A proper
choice of waveband to match the
peak of the threat radiation spectrum will result in higher threat
radiance and thus greater irradiance at the seeker at a
given range.
In the second step, shown in the middle-bottom
graph, the irradiance found in the previous step is used
to determine the first-stage aperture D0, which is the
aperture required to meet the range requirement under
the assumption of no translation or vibration motion
and no frame summing. The input parameters at this
stage are the SNR required to perform the considered
seeker function (either acquisition or discrimination)
and a noise figure for the sensor.
In the third step, shown in the middle-top graph, the
effect of translational motion of objects in the seeker
FOV is calculated. The major source of translational
motion is typically a forced movement of the seeker to
mitigate the effect of malfunctioning pixels in the FPA.
Some may have no readout signal (dead pixels), and
some may produce a very high readout signal at all times
(hot pixels). The translational rate will affect how much
energy is contained on a pixel in a frame time. The
result of the calculation at this point is DT, the required
aperture with translation present, but with no vibration
motion and no frame summing.
In the fourth step, shown in the lower right graph,
the effect of vibrational motion, also known as jitter,
is calculated. Jitter occurs because of high-frequency
vibrational motion of the KV. The amount of jitter will
depend on the DACS system used and associated pointing accuracy. Jitter is usually described in terms of peakto-peak amplitude, in units of pixels. The result of the
D TVS (cm)
of pixels in one dimension is typically a few hundred. This means that balancing the requirements
for large FOV and small IFOV is a critical factor in
seeker design.
1.0
0.75
0.50
D0 (cm)
1
4
9
25
Vibration amplitude,
peak to peak (pixel)
0.50
0.75
1.0
D TV (cm)
SNR or probability
Assumed NCA
C
NCA 2
NCA1
B
A
D0 (cm)
Start here
with range
Figure 7. Seeker parameter nomogram.
JOHNS HOPKINS APL TECHNICAL DIGEST, VOLUME 32, NUMBER 5 (2014)
SYSTEM OPTIMIZATION FOR BMD MISSILE CONCEPT DEVELOPMENT
FOV (°)
rm
at (
e.g
f-number
., 2
5
6×
Computed across
battle space
– Sensor support level
– Missile errors
– Threat
256
)
FOV =
Seeker aperture (cm)
x
F
=
x
f-number × D
F, Focal length
D, Aperture
x, FPA size
FPA
x
Select FOV and
search strategy
(FOR)
Seeker range (km)
Seeker Field-of-Regard
Containment
Fo
Seeker range (km)
calculation at this point is
DTV, the required aperture
with translation and vibration present.
In the final step, the
effect of frame summing to
reduce noise is included.
This has the effect of reducing the required aperture,
depending on the number
of frames summed for each
measurement.
To acquire the threat
FOV
object, the object must be
within both the seeker detection range and the seeker
Seeker aperture (cm)
Probability (FOR containment)
FOV. The KV is commanded
to point in the direction of
Figure 8. Seeker FOR containment relationships.
the estimated threat object
location as provided by the
stack has been balanced with the mass of the KV, the
fire control sensor. However, pointing errors caused by
next step is to balance the KV internally between the
threat tracking errors and KV navigation errors must
seeker and DACS. The objective is to ensure that error
be contained within the seeker FOV with high probcontainment can be achieved given the KV mass limit or,
ability. The KV first attempts to acquire threats within
if necessary, to rebalance KV mass with kinematic perits seeker FOV but may perform an angular search to
formance. The result is an optimized KV configuration.
achieve a larger field of regard (FOR).
The parameter relationships in developing the FOV/
Analyzing the Battle Space
FOR containment requirement are expressed by the
nomogram shown in Fig. 8. It begins with a roll-up of
Before KV optimization can be accomplished, key
Fig. 7 in the lower-left corner to relate aperture to range.
parameters related to the geometry between the interThe FOV/FOR that can be achieved for any apercepting missile and the threat must be determined.
These parameters include: (i) handover error, expressed
ture is limited by seeker design parameters such as FPA
as initial zero effort miss; (ii) the closing velocity, which
format and f-number. The f-number is the focal length
is the relative velocity between the intercepting missile
divided by the effective aperture diameter. There are
2
2
and the threat; (iii) missile third stage burnout time;
two commonly used formats, 256 and 512 , and prac+
(iv) threat burnout time; and (v) missile time of flight.
tical f-numbers vary from 1.2 (expensive) to 3.0 (large
To extract these key engagement parameters, the battle
and heavy seeker). The relationship of the FOV to these
space (all possible combinations of intercept, threat
parameters is illustrated in the upper portion of Fig. 8.
launch, and threat impact points) must be analyzed using
The FOV/FOR to contain KV pointing error and
an engagement simulation, which computes all possible
threat location uncertainty at a seeker acquisition range
intercepts where the intercepting missile can kinematiis shown in the KV seeker FOR containment calculacally reach the threat. The simulation also determines
tions in the lower-right portion of the nomogram. The
the maximum possible time window during which the
FOV/FOR requirement is evaluated for all kinematically
missile can be launched to have a successful intercept.
feasible intercepts in the threat intercept battle space for
This window is called the launch window.
a given seeker acquisition range.
For a given weapon system, there is usually a minimum acceptable launch window to allow for a successKV Optimization
ful engagement. For example, if the maximum launch
The previous sections described how DACS and
window for a particular engagement is 120-s long and
seeker technology constraints can be related to the
the minimum acceptable window is 45 s, then the
engagement planner would find the best 45-s window
basic performance parameters of the KV trade space. As
within the 120-s time period. For this concept analysis,
shown in Fig. 3, after the performance of the booster
JOHNS HOPKINS APL TECHNICAL DIGEST, VOLUME 32, NUMBER 5 (2014)
783­­­­
A. J. PUE ET AL.
is computed as a function of
seeker range capability. The
containment probability is
the product of divert and FOV
containment probabilities.
Cabapility
Moving from the lowerleft to the upper-left subMargin
KV
plot, this is where the DACS
(res mass
ulti
performance verses mass is
Need
ng (kg)
Vbo
)
described. Multiple curves
can be generated for differSeeker mass (kg)
Seeker range (km)
ent KV masses as a function
of the seeker mass. The curves
will depend on whether an
Threat
SDACS or LDACS is assumed.
assumptions
The relationship between the
upper-left and -right design
ion
ns
o
rat
points identifies how much
i
t
u
)
p ng
fig
sum mmi
on
s
margin, if any, exists between
c
a
u
er
ker e s
ek
the required and delivered
See ., fram
Se
(e.g
amount of divert.
Seeker mass (kg)
Seeker range (km)
Although the nomogram
provides a useful tool for disFigure 9. Divert containment relationships.
playing different KV options
versus system capabilities,
“best” is defined as the launch window that requires the
it does not directly provide an answer to the best split
between optics and divert capability. From a performinimum amount of divert. This methodology is used
mance perspective, the lightest KV that accomplishes
to select a minimum launch window for each and every
the mission is desired. If margin is found, requirements
engagement. Then, for the purpose of divert sizing, the
on other portions of the system may be relaxed. To find
single most stressing engagement is selected from all
the lowest mass KV, an automated tool was developed at
of the minimum launch windows. The key parameters
APL that basically traces all possible paths around the
that define this trajectory are used in the sizing of the
nomogram in Fig. 9, with a specified margin between the
required KV divert.
required and delivered divert.
Each of the points along the x axis in Fig. 10 repreSeeker versus DACS
sents a different trace around the nomogram. The figure
With the key parameters associated with the most
shows the subassembly masses of the seeker and avionics
stressing trajectory established, it is now possible to balas well as the DACS. The KV mass is the summation
ance the KV mass between the seeker and DACS.
The relationships between the divert containment
parameters are illustrated by the nomogram shown in
Optimum
Fig. 9. The seeker aperture size is first selected in the
KV mass
KV mass
lower-left corner of the nomogram. The first plot relates
the seeker mass to the aperture size. Multiple curves can
be generated for different mass margin philosophies.
DACS mass
Moving to the lower-right plot, the seeker performance
is given as a function of aperture size. This is a roll-up of
the nomogram shown in Fig. 7.
Moving on to the upper-right plot, this is where the
required divert is related to seeker performance. Of the
Seeker and avionics mass
four plots in this KV nomogram, the upper-right plot is
the most complex to generate. This is where many of the
Optimum
seeker aperture
system assumptions on target set, mission geometry, and
Seeker aperture
engagement support quality are introduced. For these
assumptions and throughout the battle space, the total
divert (V) needed to satisfy a containment probability
Figure 10. KV mass optimization.
KV mass
Seeker aperture (cm)
Seeker aperture (cm)
KV divert ∆V
need (m/s)
KV divert ∆V
capability (m/s)
Computed across battle space
– Sensor support level
– Missile errors
– Threat
– Containment probability
784
JOHNS HOPKINS APL TECHNICAL DIGEST, VOLUME 32, NUMBER 5 (2014)
SYSTEM OPTIMIZATION FOR BMD MISSILE CONCEPT DEVELOPMENT
Strike
angle
θextent (rad)
Seeker aperture = D
Size
an
gl
θextent (rad)
Range vs. θextent
N = no. of pixels to recognize
target object
θextent = N*IFOV
e
r ik
e
IFOV < blur ( 2.44λ )
D
IFOV (rad)
Aimpoint recognition range (m)
FOV vs. IFOV
Now that concepts for
the major components of the
IFOV (rad)
KV have been developed,
a physical layout of the KV
must be realized to ensure a
feasible KV size. This can be
done using a solid modeling tool, such as Pro/ENGINEER
or SolidWorks. In this way, the overall package is
developed and visually checked and mass properties are
determined. These mass properties are then applied in a
simple six-degree-of-freedom simulation to size the ACS.
The ACS is sized to maintain stability after the KV is
ejected from the upper stage, perform the roll maneuvers
before specific divert events, maintain control during
divert events, and perform the seeker pointing functions.
LETHALITY CONSTRAINTS
The final trade area in Fig. 3 is the seeker resolution
(IFOV) versus KV acceleration needed to ensure a hit
accuracy that provides the desired level of lethality.
Figure 11 shows the form of the nomograph that
illustrates the trades between IFOV and KV acceleration versus probability of hit. Given the FOV that is
determined by handover error containment analysis, the
lower left plot of Fig. 11 shows the relationship between
FOV and IFOV versus FPA format. Given IFOV, the
upper-left plot of Fig. 11 defines the angular extent of
the threat object as a function of seeker aperture and
the number of pixels, N, needed for recognition. The
aimpoint recognition range for an engagement is the
target projected length perpendicular to the line of
sight divided by angular resolution required for aimpoint
selection. Given the closing velocity of the encounter
and the aimpoint selection range, the time-to-go at aimpoint selection is calculated. The acceleration needed to
accurately hit the threat is estimated from the projected
JOHNS HOPKINS APL TECHNICAL DIGEST, VOLUME 32, NUMBER 5 (2014)
P hit
Format
S tr
ike
ang
le
P hit
FOV (°)
Physical Packaging and
ACS Sizing
θextent vs. IFOV
St
of the two. From this plot,
the optimum seeker aperture size and resulting minimum KV mass can easily be
obtained. It is worth noting
that the KV mass is highly
dependent on the level of
engagement system support
and the required target set.
Thus, a larger system balancing may consider the
trade between engagement
support parameters and the
resulting missile configuration and performance.
Conditions:
– KV divert
acceleration limit
– Strike angle
– Closing velocity
Aimpoint recognition range (m)
Figure 11. Lethality trade space.
miss and the time-to-go at aimpoint selection less processing delays. Finally, a terminal guidance simulation is
used to calculate the probability of hit (Phit) shown in
the lower-right plot of Fig. 11.
CONCLUSIONS
Development of a missile concept in the context of
a larger engagement support system involves a complex
set of highly coupled trades. We have defined a systematic process that is logically decoupled and allows
an overall performance optimization and balancing of
conflicting constraints. Currently, the methodology is
split into three major steps. The first is the optimization of booster kinematics, the second is optimization
of KV mass to remove system errors, and finally, the
third optimizes endgame lethality. Future research can
explore ways to integrate these steps and include more
detailed representations of the external sensors, command and control architecture, and engagement support functions. This can then lead to a more automated
and simultaneous optimization of both engagement support and missile parameters.
ACKNOWLEDGMENTS: The authors acknowledge the contributions of Walt Dyer, Isaac Bankman, Lisa Francisco, and
Wayne Pavalko to this work.
REFERENCES
1Brown,
N., and Olds, J., “Evaluation of Multidisciplinary Optimization Techniques Applied to a Reusable Launch Vehicle,” J. Spacecr.
Rockets 43(6), 1289–1300 (2006).
785­­­­
A. J. PUE ET AL.
2Boria,
F., Stanford, B., Bowman, S., and Ifju, P., “Evolutionary Optimization of a Morphing Wing with Wind-Tunnel Hardware in the
Loop,” AIAA J. 47(2), 399–409 (2009).
3Shiau, J., Ma, D., and Chiu, C., “Optimal Sizing and Cruise Speed
Determination for a Solar-Powered Airplane,” J. Aircraft 47(2), 622–
629 (2010).
4Sentinella, M., and Casalino, L., “Hybrid Evolutionary Algorithm for
the Optimization of Interplanetary Trajectories,” J. Spacecr. Rockets
46(2), 365–372 (2009).
5Bayley, D., Hartfield, R., Burkhalter, J., and Jenkins, R., “Design Opti-
mization of a Space Launch Vehicle Using a Genetic Algorithm,”
J. Spacecr. Rockets 45(4), 733–740 (2008).
6Anderson, M., Burkhalter, J., and Jenkins, R., “Design of a Guided
Missile Interceptor Using a Genetic Algorithm,” J. Spacecr. Rockets
38(1), 28–35 (2001).
7American Institute of Aeronautics and Astronautics, “Mass Properties Control for Space Systems,” AIAA Standard S-120-2006
(2006).
The Authors
Alan J. Pue serves as Technical Advisor for Missile Defense Agency/DVM, sponsor of the Standard Missile 3 (SM-3)
Block IIB concept development. Dr. Pue is the Chief Scientist in the Air and Missile Defense Sector (AMDS). He led
efforts to develop the overall systems engineering approach used for the SM-3 Block IIB Government Reference Concept.
Richard J. Hildebrand is the Assistant Group Supervisor of the Mechanical and Aeronautical Engineering Group in
AMDS and the Chief Engineer for the APL SM-3 Block IIB project. In addition to his responsibilities as Chief Engineer,
he developed the KV sizing tool and developed the conceptual KVs. Daniel E. Clemens is a section supervisor in the
Mechanical and Aeronautical Engineering Group. Dr. Clemens developed the missile booster optimization program,
ORION, for the SM-3 Block IIB project. Jonah R. Gottlieb is a mechanical engineer in the Mechanical and Aeronautical
Engineering Group. He analyzed booster control hardware options and alternative KV configurations. James M. Bielefeld is an electrical engineer in the Guidance, Navigation, and Control Group in AMDS. Dr. Bielefeld conducted battle
space analyses to assess seeker FOV error containment and endgame lethality. Timothy C. Miller is a section supervisor
in the Electro-Optical and Infrared Seekers and Sensor Systems Group in AMDS. Dr. Miller analyzed and developed
alternative KV seeker concepts. For further information on the work reported here, contact Alan Pue. His e-mail address
is alan.pue@jhuapl.edu.
The Johns Hopkins APL Technical Digest can be accessed electronically at www.jhuapl.edu/techdigest.
786
JOHNS HOPKINS APL TECHNICAL DIGEST, VOLUME 32, NUMBER 5 (2014)
Improving Earth Background Characterization
Through Modeling and Measurements of Leaf
Bidirectional Reflectivity
Shadrian B. Strong, Michael E. Thomas, Andrea M. Brown, and
Elena Y. Adams
he Earth albedo, or the ratio of the upwelling to down-welling radiative flux at
the surface, consolidates both wavelength- and angle-dependent information,
often oversimplifying surface properties. Accurate characterization of this
parameter is important for many Johns Hopkins University Applied Physics Laboratory (APL) programs. Within the albedo value is the bidirectional reflectance
distribution function (BRDF), which describes the angle dependency of reflectance and
emission relative to observation angle, solar angles, and wavelength. As a first step
toward improving land characterization of Earth and assisting with removal of land
radiative effects to reveal pristine target signatures, we developed a semiempirical BRDF
model for leaves. Leaves are easy to acquire, easy to integrate into an existing facility,
and highly seasonally and regionally dependent. Typical leaf reflectances used in modeling cover a wavelength of ~500–2000 nm. We extended this database from 250 to
10,000 nm to examine the radiative impact of surface albedo. In this article, we present the model and compare it with quantitative observations for red maple, dogwood,
and white oak leaves. Implementing the default parameterization of a maple leaf in the
Moderate Resolution Radiative Transfer (MODTRAN) radiative transfer code results in a
retrieved radiance error of 1–3% relative to our improved BRDF model. This error is outside the required absolute accuracy for environmental and atmospheric models used at
APL and highlights the need to use our new parameterization for environmental studies.
INTRODUCTION
Reflected solar radiation from Earth is directly measurable by satellite instrumentation. The reflectance
observed by a spacecraft sensor contains both surface
JOHNS HOPKINS APL TECHNICAL DIGEST, VOLUME 32, NUMBER 5 (2014)
and atmospheric components and depends on the direction of solar illumination and angle of observation.
Whether the observational intent is to gather infor-
787­­­­
S. B. STRONG ET AL.
mation about the atmosphere specifically or to identify
(<10%) result in large thermal changes (>80%).6 An
absolute accuracy of 0.02–0.05 for albedos is required
embedded surfaces or targets, decoupling the effects
for climate, biochemical, and hydrological models (e.g.,
of these contributions is critical to accurately deriving
Ref. 6). Models used at APL through NASA Earth scithermal and intensity information for objects of interest.
ence programs require this accuracy to assess anthroAs discussed by Hudson et al.,1 this decoupling is chalpogenic and nonanthropogenic contributions to the
lenging because the atmospheric and surface interaction
Earth climate system. APL’s National Security Space
comprises multiple scattering effects.
and Air and Missile Defense Mission Areas require
It is critical to extricate atmospheric, surface, and
this accuracy to assess the radiative signature of the
target properties for a broad range of applications. For
atmosphere on target detection, identification, and sigexample, both Johns Hopkins University Applied Physics
nature analysis.
Laboratory (APL) Air and Missile Defense and National
Earth surface BRDFs are highly dependent on land
Security Space Mission Area applications require accusurface characteristics, including vegetation type (or lack
rate, radiometric characterization of specific man-made
thereof), soil, moisture, and physical state (solid, liquid,
targets. Physical changes in the target (e.g., material, temgas). BRDFs, and consequentially albedos, may be physiperature, and emissivity changes) must be disentangled
cally modified through seasonal changes, anthropogenic
from the physical changes of Earth’s background. Neglectparticulate deposition (such as soot, aerosols, etc.), agriing changes in the atmosphere or environment may result
cultural practices, deforestation, urbanization, and daily
in attributing natural variability to changes in the target.
meteorologically driven actions (wind, erosion, etc.). As
Ignoring surface scattering and reflection properties has
the radiometric interface between the land surface and
led to uncertainties as high as 15–20% in radiance and
the atmosphere, the albedo both defines the lower boundsurface brightness retrievals.2 For APL’s Civil Space Misary for atmospheric radiative transfer and details the
sion Area, it is essential to distinguish between short-term
total shortwave energy input into the biosphere.7 Surface
(hourly, daily) cloud reflection and long-term (seasonal,
3
albedo is an important parameter used to investigate the
annual) ice reflection. This knowledge of surface reflecnet change between incoming radiation energy and outtivity, input into general circulation models for climate
going radiation energy in the Earth climate system (e.g.,
change investigation, has a substantial impact on model
Ref. 7). At the most basic level, understanding changes in
predictions.4 In particular, assumed land surface reflectivthe BRDFs of a variety of natural surfaces (e.g., leaf/tree
ity directly affects global surface energy and water baltype, ice, sand, rock) will enable more accurate derivaance in these simulations. Deviations between satellite
tions of surface albedos.
observations and model assumptions of surface reflectivAs a first step toward creating an improved dataity reveal errors >25% in surface reflectance.5
base of natural Earth surface BRDFs for use in APL
The percent reflection of Earth, as viewed by a satelNational Security Space and Civil Space Mission Area
lite, is referred to as albedo, or the ratio of the upwelling to down-welling radiative
Chopper
flux at Earth’s surface. The
albedo is related to the inteAttenuation
0.633 µm
optics
gral of the wavelength-depenSignal detector
dent bidirectional reflectance
Reference
distribution function (BRDF),
detector
0.405 µm
which describes the angle
dependency of reflectance
Sample
Mirror
rotation
(and emission) relative to
1.064 µm
focusing
stage
system
observation angle and solar
angles. Often, the intrinsic
3.390 µm
BRDF of a surface is lost in the
observation of a single-value
albedo because the albedo
10.6 µm
Lock-in
Lock-in
is a summation or angledetector
detector
averaged BRDF. There is also
Motion
an inherent loss of spectral
Computer
controller
dependency in albedo. Little
has been done to quantify
the impact of subtle changes Figure 1. Experimental setup of the AMD BRDF system. Inputs are five laser sources shown on
in BRDF on the derivation of the left. Inputs travel through a series of optical components (middle). Shown on the right is the
albedo, but it has been shown position of the sample on a rotating stage and mobile sensor unit. Reproduced from Ref. 15.
that small changes in albedo © 2010 Society of Photo Optical Instrumentation Engineers.
788
JOHNS HOPKINS APL TECHNICAL DIGEST, VOLUME 32, NUMBER 5 (2014)
MODELING AND MEASUREMENTS OF LEAF BIDIRECTIONAL REFLECTIVITY
applications, we have developed a semiempirical BRDF
model for leaves using the APL Air and Missile Defense
Sector’s BRDF laboratory facility with parameters from
the Leaf Optical Properties EXperiment (LOPEX)
database.8 Sponsored jointly by APL National Security Space Mission Area and Research and Exploratory
Development Department independent research and
development funding in 2009, we developed a model
for leaves specifically, but the model has potential for
broad application. Leaves are easy to acquire, easy to
integrate into the existing APL facility apparatus, and
highly seasonally and regionally dependent. They cover
~10% of Earth.
By investigating BRDF variability in leaves as a function of species, water content, chlorophyll content, and
thickness, we begin to understand how a more complex
system, such as a tree or canopy, may be parameterized.
An initial simple analysis of the BRDF of individual
leaves is required before we can extrapolate and study
the signatures of trees and forests: we anticipate the
BRDF of entire trees and forests would be highly complex and variable but would maintain an underlying
signature associated with plant type. Leaf reflectance is
partly due to roughness with varying density, dimension,
and refractive indices and partly due to scattering off
cell surfaces.9 The leaf is the primary energy-harvesting
element of a plant and provides the energy necessary
to drive the conversion of CO2 into plant sustenance.10
Remote sensing observations of changes in leaf reflectivity and emission can provide information about plant
type (e.g., detection of invasive plant species)11 and
information about changes in ecosystems (e.g., Ref. 12).
We have compared our model and observations with
the Leaf Experimental Absorptivity Feasibility MODel
(LEAFMOD)10 and found them to agree favorably. We
have parameterized leaf BRDFs for the 0.3- to 3.5-µm
wavelength range for deciduous leaves: maple, dogwood,
and oak leaves. We anticipate
deciduous tree leaves to have
simpler BRDFs, facilitating
preliminary analysis. Further study would incorporate
coniferous species. The leaf
parameterization has been
implemented for inclusion
in the Moderate Resolution
Radiative Transfer (MODTRAN) model.13
120
90
0.1
60
0.01
150
30
110–3
Laser
10–4
180
0
Sample
210
330
240
300
270
Figure 2. The leaf specimen is vertically attached to the rotation
stage in the center of the apparatus. The laser beam is then incident on the sample, and the resultant reflection/scatter is measured at the detector.
ing from man-made surfaces such as ceramics, paints,
and metals. There are currently many laser wavelengths
available for measurements from UV to the longwave
infrared (LWIR), although only five are depicted in
Fig. 1. Free-space mirrors are used to manually select the
wavelength of interest for each series of measurements.
A lock-in detector is used to separate the reflected signal
from background light, and a chopper is used in the path
of the continuous wave lasers for this purpose. Mirrors
are used to focus the laser light onto the signal detector,
which is mounted on an extension arm attached to a
rotation stage. The sample is placed in the center of the
detector circle on the second rotation stage, which controls the incident angle of the laser light. The leaf specimen is vertically attached to the second rotation stage
(Fig. 2). Figure 1 depicts the layout of the BRDF facility
EXPERIMENTAL SETUP
The Air and Missile
Defense Sector’s BRDF laboratory system is shown in
Fig. 1. The system is typically
used to analyze the scatter-
Figure 3. Spring/summer flowering dogwood leaf (top left), red maple leaf (top center), and
white oak leaf (top right) and fall maple leaf (bottom left) and oak leaf (bottom right).
JOHNS HOPKINS APL TECHNICAL DIGEST, VOLUME 32, NUMBER 5 (2014)
789­­­­
S. B. STRONG ET AL.
0.8
MODTRAN4 maple leaf albedo
Green paint spectrum (reflectance)
APL dry maple leaf (reflectance)
APL fresh maple leaf (reflectance)
0.7
0.6
Reflectance
0.5
0.4
0.3
0.2
0.1
0
0
0.5
1.0
1.5
Wavelength (µm)
2.0
2.5
Figure 4. Spectrum of the default MODTRAN4 maple leaf reflectance (blue), spectrum of a green paint sample (red), dry maple
leaf reflectance spectrum (green), and fresh maple leaf reflectance spectrum (black). Notably, the dry and fresh samples have
distinctly different chlorophyll “red edges” near 0.750 µm, such
that the dry leaf (green) has a smoother slope between 0.5 and
1 µm relative to the sharp, near-discontinuity of the fresh leaf
(black). Differences in water content and leaf thickness account
for amplitude changes. The MODTRAN4 spectrum has a coarser
resolution and does not enable the model user to differentiate
between dry and fresh leaves. For comparison, we have included
a green paint sample, which mimics features of the leaves, including the chlorophyll bump near 0.5 µm.
with some of the available laser wavelengths represented.
A current list of available wavelengths includes: 0.375,
0.405, 0.532, 0.633, 0.635, 0.760, 0.830, 0.905, 1.064, 1.55,
3.39, and 10.6 µm and a scanning optical parametric
oscillator that includes wavelengths from 1.4 to 1.9 and
2.4 to 3.8 µm. APL recently acquired a vacuum chamber
with optical ports spaced at 20° increments in the scattering plane. This chamber facilitates the measurement
of high-temperature BRDFs using either a resistive heating element to heat the sample from the back or a 40-W
CO2 laser to heat the sample from the front.
Three different species of leaves were analyzed: (i) red
maple (Acer rubrum), (ii) flowering dogwood (Cornus
florida), and (iii) white oak (Quercus alba), shown in
Fig. 3. Leaf samples were collected from APL’s main
campus. Both “fresh” and dry leaves were analyzed. In
the case of fresh leaves, a sample from each of the species
was selected immediately before measurements were to be
790
conducted. This was done to ensure that minimal water
was lost (due to drying of the leaves and decomposition).
Loss of water was shown to affect the observed absorption and reflectance spectrum. This may be observed in
Fig. 4 (green for dry maple leaf and black for fresh maple
leaf). For comparison, green paint is overplotted in red,
and the assumed MODTRAN maple leaf spectrum is
overplotted in blue. Notably, the dry maple leaf exhibits both a high reflectance relative to the fresh leaf and
a degradation in the chlorophyll feature near 500 nm.
The MODTRAN spectrum most closely replicates the
green-paint spectrum and neither of the two maple leaf
samples. Angle-resolved reflectance data were collected
for each sample at three different sample rotary angles at
wavelengths of 405, 633, 1064, and 3390 nm.
Spectrally resolved total integrated reflectance [TIR;
also called hemispherical directional reflectance (HDR)]
and transmittance [also called hemispherical directional
transmittance (HDT)] data were also collected for each
leaf species by using a grating spectrometer to cover the
UV, visible, and near-infrared (NIR) regions of the spectrum, and a Fourier transform infrared spectrometer was
used to cover the midwave infrared (MWIR) and LWIR
regions of the spectrum.
BRDF MODELING
The BRDF is defined as the ratio of backwardreflected flux into the differential solid angle, , to the
incident flux, i,
^
i,
r,
h=
1 d
i d
r
r
=
Ir ^
Incident
light
i,
ht r $ nt
. (1)
i^ h
r,
Specular
i
r
Diffuse
Lambertian
t
Near
specular
Figure 5. Experimental geometry and graphical depiction of
scattering models for in-plane measurements. Reproduced from
Ref. 15. © 2010 Society of Photo Optical Instrumentation Engineers.
JOHNS HOPKINS APL TECHNICAL DIGEST, VOLUME 32, NUMBER 5 (2014)
MODELING AND MEASUREMENTS OF LEAF BIDIRECTIONAL REFLECTIVITY
(i,r,) is called the BRDF. For a flat reflecting surface, t r $ nt = cosr. The inner
product equals 1 for a spherical surface or a point reflector.
For a homogenous, uniform surface with randomly oriented surface roughness, the
BRDF will display no dependence in the  direction (out of the plane reflection). Consequentially, scatter from the sample is unchanged under rotation by the angles i and
r, and the phase function is a function of i and r only. This is assumed to be the case
for the samples analyzed in this study, and all measurements and analyses are performed
for in-plane scatter (i.e., the case of  = 0°).
Integrating the BRDF over the entire hemisphere results in the hemispherical TIR
(also called directional hemispherical reflectance), TIR. The values of TIR range from
0 to 1. The modeled BRDF is a product of the TIR and a normalized angle-dependent
phase function, P(i,r). All phase functions are normalized to 1, becoming equivalent
to a probability density function, such that one may determine the probability of observing a scattered ray in a particular direction given an incidence angle. The BRDF may
be written as
BRDF^ i, r h = TIR ^ ih Pr ^ i, rh .
(2)
Physically based empirical models are fit to the BRDF experimental data to extend a
limited set of in-plane scattering data to all possible incident and scattered angles. The
model also compensates for the dropout in experimental signal caused by the detector
arm blocking the incident laser beam from striking the sample for a small range of scattering angles (see Fig. 10). The BRDF model is separated into specular (S), near-specular
(NS), diffuse (D), and Lambertian (L) components. Each of these four components follows a separate angular dependence [P()] and together represent the reflected beam
from the interface reflection (specular) and the surface and bulk scattered light.14 The
BRDF is modeled as a sum of all four components, in other words,
BRDF^ i, r, h = S ^ i, h Pinst ^ i, r, h + NS ^ i, h PNS ^ i, r, h
.
+ D ^ i, h PD ^ i, r, h + L ^ h PL ^ r, h
(3)
It is assumed that these different types of reflected, transmitted, and scattered light
are independent. Specular components (PS) represent the contributions from the
Absorption coefficient (1/cm)
1.2103
Chlorophyll a model
Chlorophyll b model
Chlorophyll a in ether
Chlorophyll a in MeOH
Chlorophyll b
Carotenoid model
800
400
0
250
350
450
550
Wavelength (nm)
650
750
Figure 6. Comparison of chlorophyll a and b absorption coefficient to uncalibrated published spectra.
JOHNS HOPKINS APL TECHNICAL DIGEST, VOLUME 32, NUMBER 5 (2014)
791­­­­
S. B. STRONG ET AL.
reflected and transmitted rays for a perfectly flat surface and are calculated by applying
the Law of Reflection (i.e., r = –i). Near-specular components (PNS) represent weak
or single scatter phenomena that are only slightly spread from the specular direction
primarily because of minor surface roughness. Diffuse components (PD) represent the
effects of surface roughness and bulk scatter that are strongly influenced by multiple
scatter and are partially coherent relative to the incident light. The Lambertian components (PL) represent random rough surface and bulk scatter that is entirely incoherent
with the incident light.15 Examples of all four types of reflection are depicted in Fig. 5.
Because of the diffuse nature of the samples studied, only the diffuse and Lambertian
components are required to model the reflectance from the three species of leaves that
were studied.
The diffuse phase function is given by
PD = ^ r, i, m, h = N D
^ cos ihm – 1
cos r ,
^sin r + sin ihm + ^ cos ihm
(4)
where  is the half angle of the reflected beam, m is an exponent chosen to best fit
the observed data, and ND is a normalization factor.15–17 The hemispheric Lambertian
phase function is given by
1
PL ^ r h = cos r .
(5)
The integrated quantities for each component represent the fraction of incident light
that is reflected under each phase function. The TIR for the entire BRDF would then
be the sum of each individual component integrated reflectances, or
TIR ^ i, h = L ^ i, h + D ^ i, h + NS ^ i, h + S ^ i, h # 1 .
(6)
The TIR for the complete BRDF follows the total power law,
1 = ^ TIR ^ i, h + TIT ^ i, hh + abs ^ i, h = TIS ^ i, h + abs ^ i, h ,
1.510
(7)
3
Background model
Absorption coefficient (1/)
Fall background
1103
500
0
250
350
450
550
650
750
Wavelength (nm)
Figure 7. Required background absorption to fit observed leaf spectra. The fall leaf background is
mostly due to decomposing chlorophyll.
792
JOHNS HOPKINS APL TECHNICAL DIGEST, VOLUME 32, NUMBER 5 (2014)
MODELING AND MEASUREMENTS OF LEAF BIDIRECTIONAL REFLECTIVITY
Table 1. Content of different species composing the leaf absorption coefficients as implemented in the
APL model for maple, dogwood, and white oak
Sample
Fall red
maple
Fall white
oak
Spring
dogwood
Spring
holly
Water
Chlorophyll
Dry
leaf
matter
Carot- Decomposenoids ing chlorophyll
Concentra- Scatter
tion of scat- radius
ter centers
(mm)
(no. per cm3)
Leaf
thickness
(cm)
0.135
0.005
0.09
0.1
0.67
2.9 × 108
1.95
0.015
0.135
0.005
0.09
0.1
0.67
2.9 × 108
1.95
0.037
0.6
0.27
0.03
0.1
0
0.75 × 108
1.7
0.033
0.6
0.27
0.055
0.075
0
0.75 × 108
1.7
0.037
Mixing ratios listed by concentration. The listed radius is for an effective sphere.
where TIT is the total integrated transmittance, rTIR is the TIR, and abs is the absorptance. Following Kirchhoff’s law of thermal radiation, the emissivity () of a material is
equal to its absorptance, therefore
^ i, h = 1 – ^
TIT ^ i,
h+
TIR ^ i,
hh = 1 –
TIS ^ i,
h.
(8)
The BRDF for the leaves, or any opaque material, provides an estimate of the unidirectional emissivity of a material at a particular incident angle and wavelength.15, 17
The Lambertian BRDF model is a variation on a radiative transfer model with a twostream approximation (see the Appendix A for further description). The model uses the
Kubelka–Munk equation to account for the relationship between absorption concentration and reflectance in the leaf specimens, modeling the wavelength-dependent reflectance and transmittance of the leaf. The model takes as input four leaf constituents that
act as the main absorbers in the leaf samples: chlorophylls (types a and b), water, carotenoids, and background material or “dry leaf matter.” Figures 6 and 7 plot the absorption coefficient of these constituents. The model is designed such that the absorption
coefficients of each of the leaf components may be varied to fit the measured data from
the UV through the LWIR. The mixture coefficients used for each different tree species can be found in Table 1. The fall leaf background is believed to be decomposing
1104
30
Absorption
1103
20
100
10
10
1.0
0
0.1
1
10
Absorption coefficient (1/cm)
Scatter coefficient (1/cm)
Scatter
0.1
100
Wavelength (µm)
Figure 8. Scatter coefficient and net absorption coefficient for a spring/summer green leaf.
JOHNS HOPKINS APL TECHNICAL DIGEST, VOLUME 32, NUMBER 5 (2014)
793­­­­
S. B. STRONG ET AL.
0.8
Maple
Measured UV-NIR front
0.7
Laser TIR
APL model
0.6
HDR
0.5
0.4
0.3
example of this comparison
is shown in Fig. 9, where
the blue squares are TIRs
calculated from the measured BRDFs for a maple
leaf, and the red line is the
TIR from the spectrometer
measurement, also known
as the HDR.
EXPERIMENTAL
RESULTS
0.2
The phase functions
explained in the previous
section (Eqs. 4 and 5) have
0
been used to model the
3
4
5
100
110
110
110
Wavelength (nm)
scattering from the three
leaf samples at the different
Figure 9. The APL leaf model fit (black) to the spectrometer measurements (red). The BRDF TIR
incident angles and wavemeasurements for the spring red maple are overplotted with blue squares.
lengths. Figure 10 shows the
reflectance of a red maple
leaf at 405 nm at three different incident angles: 5°, 20°,
chlorophyll. Also listed in Table 1 are the leaf thickand 60°. The red crosses indicate measured values, and
ness, the most common particle radius, and the number
the dotted blue lines are the modeled function. The
of scatter centers within each leaf. This information is
reflectance of leaves is often modeled as purely Lamberused to compute the scatter coefficient based on anomatian, but analysis at an incident angle of 60° in Fig. 10
lous diffraction approximation (ADA). For this calculaindicates that there is a diffuse component of the scattion, it is assumed that the particles are spherical. It is
tered light that retains directional coherence.
interesting to note that the size of chloroplasts within a
leaf cell match the size of the
0.1
0.1
scattering particles required by
the model. The index of refraction of the scattering particles
is assumed to be 1.6. For green
leaves, the background is pre0.01
0.01
dominantly water. The combination of Kubelka–Munk
theory and ADA make this
approach
computationally
0.001
0.001
50
0
50
50
0
50
efficient. A plot of the scatter
Reflection angle (°)
Reflection angle (°)
coefficient and net absorption
0.1
coefficient for a spring/summer
leaf is presented in Fig. 8. A
measured TIR spectrum is then
used to empirically determine
0.01
these parameters for each leaf.
An example of the fit obtained
by using this model is shown in
Fig. 9 for a spring maple leaf.
0.001
The TIR calculated from
50
0
50
the angularly resolved BRDF
Reflection angle (°)
model is compared with the
wavelength-dependent reflec- Figure 10. Measured (red crosses) and modeled (blue dotted lines) BRDF for a maple leaf at
tance that was measured using 405 nm for incident angles (clockwise from top) of 5°, 20°, and 60°. Dropout in experimental
a grating spectrometer. An BRDF is caused by the detector arm blocking the incident laser for a small range of angles.
BDRF (sr –1)
BDRF (sr-1)
BDRF (sr –1)
0.1
794
JOHNS HOPKINS APL TECHNICAL DIGEST, VOLUME 32, NUMBER 5 (2014)
MODELING AND MEASUREMENTS OF LEAF BIDIRECTIONAL REFLECTIVITY
0.1
0.1
0.01
0.001
100
0
50
50
Scattering angle (°)
0.1
100
3390 nm
0.01
0
50
50
Scattering angle (°)
50
0
50
Scattering angle (°)
0.1
1064 nm
0.1
0.001
100
0.01
0.001
100
100
BDRF (sr –1)
BDRF (sr –1)
633 nm
BDRF (sr –1)
BDRF (sr –1)
405 nm
0.01
0.001
100
100
50
0
50
Scattering angle (°)
100
Figure 11. Measured (red crosses) and modeled (blue dotted lines) BRDF for a maple leaf at an
incident angle of 20° for wavelengths of 405, 633, 1064, and 3390 nm.
0.1
Maple
BDRF (sr –1)
BDRF (sr –1)
0.1
0.01
0.01
Dogwood
BRDF of the three different leaves at 405 nm and for
an incident angle of 20° is
shown in Fig. 12. Notably,
the red maple appears to
scatter closest to Lambertian, whereas the dogwood,
at 405 nm, has the least
Lambertian profile. The
parameters used to create
the modeled BRDF curves
in Figs. 10–12 are summarized in Table 2.
Figures 13 and 14 compare
the Kubelka–Munk model
fit with the experimental
data on the spring/summer
dogwood leaf and fall oak
leaf, respectively. The model
parameters are determined
by the reflectance measurement and checked with the
transmittance measurement.
Good agreement is demonstrated in both cases.
RADIATIVE TRANSFER
ANALYSIS
BDRF (sr –1)
To assess the radiative
impact of changes in the
BRDF for leaf species, we
input into MODTRAN518
0.001
0.001
our fresh and dry maple
100 50
0
50
100
100 50
0
50
100
leaf
integrated
reflecScattering angle (°)
Scattering angle (°)
tance (BRDF) values and
0.1
compare with the default
Oak
MODTRAN parameterization. Additionally, we investigate the radiative signature
calculated from the LOPEX8
0.01
maple leaf parameterization
(Fig. 15). The LEAFMOD10
calculations for reflectance
for a variety of leaf thick0.001
100 50
0
50
100
nesses are plotted in Fig. 15.
Scattering angle (°)
Because they are comparable in the shortwave
Figure 12. Measured (red crosses) and modeled (blue dotted lines) BRDF for an incident angle of
(wavelength <2 μm) and
20° at a wavelength of 405 nm for a maple leaf, a dogwood leaf, and an oak leaf.
minimal LEAFMOD data
are available at longer waveFigure 11 shows the BRDF of the red maple leaf
lengths, we do not implement these LEAFMOD profiles
at the four different wavelengths: 405, 633, 1064, and
in MODTRAN.
3390 nm. The red crosses indicate measured values, and
MODTRAN5 was run for an early-spring, zenith
angle of 0° configuration at the APL latitude and lonthe dotted blue lines are the modeled function. The
JOHNS HOPKINS APL TECHNICAL DIGEST, VOLUME 32, NUMBER 5 (2014)
795­­­­
S. B. STRONG ET AL.
reflectance, derived from the
BRDF in grey. The APL dray
maple leaf is presented in
green. The LOPEX maple leaf
Sample
 (nm)
i (°)
L
D
m
 (°)
in Fig. 15 is shown in magenta,
Maple
405
5
0.045
0.015
2
30
and the default MODTRAN5
Maple
405
20
0.045
0.015
2
30
maple profile is in black. MiniMaple
405
60
0.045
0.025
3
25
mal to no radiance difference
was observed at longer waveMaple
405
15
0.030
0.020
2
30
lengths (>2.5 μm), despite proMaple
633
20
0.045
0.015
2
30
viding MODTRAN with a
Maple
1064
20
0.500
–
–
–
more comprehensive spectral
Maple
3390
15
0.02
0.01
2
30
reflectance to 10 μm (Fig. 9).
Most notably, the APL leaf
Maple
405
20
0.045
0.015
2
30
models
both exhibit greater
Dogwood
405
20
0.023
0.03
2
40
radiance
than LOPEX and
Oak
405
20
0.025
0.03
2
30
MODTRAN default between
400 and 700 nm. There is also
0.8
0.8
a slight enhancement between
Dogwood
Dogwood
Measured HDR front
Measured HDR front
Measured HDR back
Measured HDR back
1.2 and 1.7 μm from the APL
Laser HDR back
Laser HDR back
0.6
0.6
dry maple leaf. If only the dry
Model
Model
and fresh APL leaf radiances
0.4
0.4
are examined, we find that the
fresh leaf appears to result in a
0.2
0.2
greater radiance than the dry
leaf for wavelengths <700 nm,
0
0
but the dry leaf exhibits a
100
1103
1104
1105
100
1103
1104
1105
Wavelength (nm)
Wavelength (nm)
higher radiance from 1.2 to
1.7 μm. Although these radiFigure 13. Comparison of Kubelka–Munk model fit to experimental data for a spring/summer
ance signatures are subtle, we
dogwood leaf.
have yet to fully investigate
the impact on net albedo esti0.8
0.8
mates. We also demonstrate
Measured UV-NIR front
Measured UV-NIR
Fall oak
Fall oak
Measured UV-NIR back
differences from a statistically
Measured MWIR-LWIR
Measured MWIR-LWIR
Laser TIT
0.6
0.6
front
low sample, and further quanAPL model
Measured MWIR-LWIR
back
titative impact is necessary
Laser TIR
0.4
0.4
APL model
with more leaves and leaf clusters. It should be noted that in
0.2
0.2
all instances, the reflectance
0
0
values supplied are applicable
100
1103
1104
1105
100
1103
1104
1105
for a single leaf. The impact
Wavelength (nm)
Wavelength (nm)
of canopies has yet to be constrained in this current work
Figure 14. Comparison of Kubelka–Munk model fit to experimental data for a fall oak leaf.
but is ultimately critical for a
complete understanding of the
gitude and noon hour. A precipitable water vapor value
remotely sensed signature of Earth’s surface from space.
of 2 cm (NASA MODerate-resolution Imaging SpecWe anticipate that a small radiative signature from an
troradiometer, MODIS) and column ozone amount of
individual leaf would manifest as a larger signature for
240 Dobson Unit (NASA Ozone Monitoring Instrua tree or forest, consequentially complicating environment) were implemented with a midlatitude summer
mental signature analysis, target detection, and climate
meteorological profile and rural aerosol contribution
analysis. Integrating the full spectra (0–10 μm) plotted
(23-km visibility). MODTRAN cards 4A, 4L1, and 4L2
in Figs. 16 and 17 results in the following radiances: (i)
were implemented. The only change between the analyAPL fresh maple, 0.0102 W/cm2-sr, (ii) APL dry maple,
ses of the differing maple reflectances was the specific
0.0101 W/cm2-sr, (iii) MODTRAN default, 0.0099 W/
spectral reflectance file used. Figure 16 displays the
cm2-sr, and (iv) 0.0098 W/cm2-sr. Relative to the APL
results of implementing the APL fresh maple integrated
dry maple, these percent differences are on the order of
HDR
HDT
HDT
HDR
Table 2. Parameters used in modeled fits to BRDF leaf data in Figs. 10–12
796
JOHNS HOPKINS APL TECHNICAL DIGEST, VOLUME 32, NUMBER 5 (2014)
MODELING AND MEASUREMENTS OF LEAF BIDIRECTIONAL REFLECTIVITY
0.6
Reflectance
0.5
0.4
0.3
Experimental
Leaf thickness 0.018 cm
Leaf thickness 0.020 cm
Leaf thickness 0.023 cm
Leaf thickness 0.025 cm
Leaf thickness 0.028 cm
LOPEX
0.2
0.1
0
400
600
800
1000
1200
1400
1600
1800
2000
2200
2400
Wavelength (nm)
Path radiance (W/cm2/sr1/cm1)
Figure 15. Fresh maple leaf reflectance comparison against the LEAFMOD model and the LOPEX
data. The APL experimental results are plotted in black. LOPEX data are presented in green. The
other colors indicate reflectances due to different leaf thicknesses.
–7
610
APL fresh maple leaf
APL dry maple leaf
LOPEX maple
Default MODTRAN maple
5
4
3
2
1
0
0.2
0.4
0.6
0.8
1.0
1.2
Wavelength (µm)
1.4
1.6
1.8
2.0
Path radiance (W/cm2/sr1/cm1)
Figure 16. UV-NIR spectral radiance differences between several maple leaf models: APL results
(green), default MODTRAN5 (black), and LOPEX model dry maple leaf (magenta). These results were
derived by modifying MODTRAN5 parameterization of maple leaf reflectance.
10–7
5
APL fresh maple leaf
APL dry maple leaf
4
3
2
1
0.4
0.6
0.8
1.0
1.2
Wavelength (µm)
1.4
1.6
1.8
Figure 17. APL dry (green) and fresh (grey) radiance products from MODTRAN5. The fresh leaf
appears to have a greater signature from 400 to 700 nm, whereas the dry leaf begins to exhibit and
enhancement at 1.2 μm.
JOHNS HOPKINS APL TECHNICAL DIGEST, VOLUME 32, NUMBER 5 (2014)
797­­­­
S. B. STRONG ET AL.
1% for the APL fresh maple, 2% for the default MODTRAN case, and 3% for the LOPEX data. Future higher
spectral resolution in the APL reflectances may also
impact these contributions.
CONCLUSION
As a first step toward creating an improved database of natural Earth surface BRDFs for use in critically
decoupling the Earth background from target and atmospheric signatures, we have developed a semiempirical
BRDF model for leaves using the APL Air and Missile
Defense Sector’s BRDF laboratory facility. Sponsored
jointly by National Security Space Mission Area and
Research and Exploratory Development Department
independent research and development funding in 2009,
our model was developed for leaves but has potential for
broad application. Leaves are easy to acquire, easy to
integrate into the existing APL facility apparatus, and
highly seasonally and regionally dependent.
Assumed BRDF signatures of leaves impact the
atmospheric radiative signature. We have shown that
implementing a simple, default MODTRAN model for
maple leaf reflectances may underpredict the retrieved
radiance signature by 1–3%, based on a statistically limited sample size, and assuming single leaf properties in
all comparison analysis. Although this underprediction
is small, we anticipate that the retrieved radiance error
will increase with increasing leaf number and with the
existence of multiple trees. Characterizing the signature
of an individual leaf for a specific tree type is important
at a fundamental level, prior to identifying the signature of an entire forest. We find that the BRDF model fit
the measured BRDF for the leaves. Leaf BRDF is highly
dependent on moisture content, chlorophyll content,
and orientation (front side or back side). We observe
shortwave scattering that seems consistent with scattering off chloroplasts within the individual cells.
At 500 nm, our model reveals a ~0.1 difference in
integrated reflection relative to LEAFMOD predictions
for maple leaves and, at specific wavelengths, >0.2 difference in reflectance (near 2 μm). A 0.1 difference
in reflectance can be easily obtained by observing a
fall and spring leaf. Because approximately 0.02–0.05
absolute accuracy is required for climate, biogeochemical, and hydrological models,6 precise parameterization
of seasonally dependent leaf reflectances is important.
Ultimately decoupling the radiative signature of leaves
from atmospheric signatures and other Earth surface
materials will improve APL’s environmental characterization capability within the National Security Space,
Civil Space, and Air and Missile Defense Mission Areas.
In the real world, we expect leaves to cluster in canopies and on the ground, presenting additional complications that were unexplored in this preliminary research.
Consequentially, we move toward incorporating such
798
properties in the APL BRDF facility. We anticipate the
next steps to include examination of the BRDF of leaf
clusters and coniferous leaf BRDFs. We anticipate this
information to be critical for APL to continue participation in NASA Earth climate science and environmental
remote sensing and DoD target signature analysis. The
fundamental analysis of the radiative impact of leaves
would facilitate the development of instrumentation for
airborne and space-based platforms for both Civil Space
Mission Area and National Security Space Mission
Area applications for signature exploitation: natural and
man-made object identification and characterization
and synthetic paint/material development (to mimic
plant/leaf-like characteristics).
ACKNOWLEDGMENTS: We thank summer intern Jessica
Makowski for her assistance with the leaf analysis. Thanks
to Barry Ganapol and Roberto Furfaro with the University of Arizona for their expertise and assistance with
LEAFMOD analysis and comparison. This work was funded
through the APL independent research and development program and was co-sponsored by APL’s Research
and Exploratory Development Department (previously
the Milton Eisenhower Research Center) and the National
Security Space Mission Area.
REFERENCES
1Hudson,
S. R., Warren, S. G., Brandt, R. E., Grenfell, T. C., and
Six, D., “Spectral Bidirectional Reflectance of Antarctic Snow: Measurements and Parameterization,” J. Geophys. Res. 111(D18), 1–19
(2006).
2Lee, T. Y., and Kaufman, Y. J., “Non-Lambertian Effects on Remote
Sensing of Surface Reflectance and Vegetation Index,” IEEE Trans.
Geosci. Remote Sens. 24(5), 699–708 (1986).
3Hudson, S. R., and Warren, S. G., “An Explanation for the Effect of
Clouds over Snow on the Top-of-Atmosphere Bidirectional Reflectance,” J. Geophys. Res. 112(D19202), 1–11 (2007).
4Oleson, K. W., Bonan, G. B., Schaaf, C., Gao, F., Jin, Y., and
Strahler, A., “Assessment of Global Climate Model Land Surface
Albedo Using MODIS Data,” Geophys. Res. Lett. 30(8), 1443 (2003).
5Wang, Z., Zeng, X., Barlage, M., Dickenson, R. E., Gao, F., and
Schaaf, C. B., “Using MODS BRDF and Albedo Data to Evaluate
Global Model Land Surface Albedo,” AMS J. Hydrometeorol. 5(1),
3–14 (2004).
6Henderson-Sellers, A., and Wilson, M. F., “Surface Albedo Data for
Climatic Modeling,” Rev. Geophys. 21(8), 1743–1778 (1983).
7Lucht, W., Schaaf, C. B., and Strahler, A. H., “An Algorithm for the
Retrieval of Albedo from Space Using Semiempirical BRDF Models,”
IEEE Trans. Geosci. Remote Sens. 38(2), 977–998 (2000).
8Hosgood, B., Jacquemoud, S., Andreoli, G., Verdebout, J., Pedrini, G.,
and Schmuck, G., Leaf Optical Properties EXperiment 93 (LOPEX93),
Ispra, Italy: European Commission (1995).
9Grant, L., “Diffuse and Specular Characteristics of Leaf Reflectance,”
Remote Sens. Environ. 22(2), 309–322 (1987).
10Ganapol, B. D., Johnson, L. F., Hammer, P. D., Hlavka, C. A., and
Peterson, D. L., “LEAFMOD: A New Within-Leaf Radiative Transfer
Model,” Remote. Sens. Environ. 63(2), 182–193 (1998).
11Huang, C., and Asner, G. P., “Applications of Remote Sensing to
Alien Invasive Plant Studies,” Sensors 9(6), 4869–4889 (2009).
12Su, M. R., Yang, Z. F., Chen, B., and Ulgiati, S., “Urban Ecosystem
Health Assessment Based on Emergy and Set Pair Analysis—A
Comparative Study of Typical Chinese Cities,” Ecol. Modell. 220(18),
2341–2348 (2009).
JOHNS HOPKINS APL TECHNICAL DIGEST, VOLUME 32, NUMBER 5 (2014)
MODELING AND MEASUREMENTS OF LEAF BIDIRECTIONAL REFLECTIVITY
13Berk,
A., Bernstein, L. S., and Robertson, D. C., “MODTRAN: A
Moderate Resolution Model for LOWTRAN 7,” Report prepared for
the U.S. Air Force Geophysics Laboratory by Spectral Sciences, Inc.,
Burlington, MA, http://www.dtic.mil/dtic/tr/fulltext/u2/a214337.pdf
(1989).
14Kubelka, P., and Munk, F., “Ein Beitrag zur Optik der Farbenstricke”
(“An Article on Optics of Paint Layers”), Z. tech. Physik 12: 593–601
(1931).
15Brown, A. M., Hahn, D. V., Thomas, M. E., Brown, D. M., and
Makowski, J., “Optical Material Characterization through BSDF Measurement and Analysis,” in Optical Material Characterization through
BSDF Measurement and Analysis, Proc. SPIE, Vol. 7792, Z.-H. Gu and
L. M. Hanssen (eds.), SPIE, Bellingham, WA, pp. 779211-1–779211-11
(2010).
16Thomas, M. E., Blodgett, D. W., and Hahn, D. V., “Analysis and Representation of BSDF and BRDF Measurements,” in Optical Diagnostic
Methods for Inorganic Materials III, Proc. SPIE, Vol. 5192, L. M. Hanssen (ed.), SPIE, Bellingham, WA pp. 158–167 (2003).
17Duncan, D. D., Hahn, D. V., and Thomas, M. E., “Physics-Based
Polarimetric BRDF Models,” in Optical Diagnostic Methods for Inorganic Materials III, Proc. SPIE, Vol. 5192, L. M. Hanssen (ed.), SPIE,
Bellingham, WA, pp. 129–140 (2003).
18Berk, A., Anderson, G. P., Bernstein, L. S., Acharya, P. K., Dothe, H.,
et al. “MODTRAN4 Radiative Transfer Modeling for Atmospheric
Correction,” in Optical Spectroscopic Techniques and Instrumentation
for Atmospheric and Space Research III, Proc. SPIE, Vol. 3756, A. M.
Larar (ed.), SPIE, Bellingham, WA, pp. 348–353 (1999).
19Joseph, R. I., and Thomas, M. E., “How Accurate Is the KubelkaMunk Theory of Diffuse Reflectance? A Quantitative Answer,” in
Reflection, Scattering, and Diffraction from Surfaces III, Proc. SPIE,
Vol. 8495, L. M. Hanssen (ed.), SPIE, Bellingham, WA, pp. 84950I-1–
84950I-9 (2012).
APPENDIX
x
RANDOM DIFFUSE LIGHT PROPAGATION AND
KUBELKA–MUNK THEORY
Consider an infinite turbid thin film that is applied to a substrate with a Lambertian surface reflectance of Lsub. Because
of random multiple scatter, the flow of random diffuse light
flux within the film can only be in two directions perpendicular to the film surface, up and down, because rays to the side
ultimately get redirected to the up or down direction. This concept is illustrated in Fig. 18. The downward flux in the film is
attenuated by both absorption and scatter. The backscattered
component removed from the downward flow is ultimately redirected upward. The same will be true for the net upward flow of
flux. Differential equations describing this random diffuse light
propagation are stated by Eqs. 9 and 10:14
d
Film
0
Substrate
Figure 18. Two-flux diffuse light propagation within a
film on a substrate.
back
–d down ^ x h = – ` abs + back
sca j down ^ x h dx + sca up ^ x h dx (9)
back
d up ^ x h = – ` abs + back
sca j up ^ x h dx + sca down ^ x h dx .
(10)
and
The absorption coefficient, abs, and bulk scatter coefficient,  back
sca , are assumed independent of position but depend on wave
number, n. The second term in the preceding equations accounts for random multiple scatter such that the effective forwardto-back scatter ratio is 1, even though the forward-to-back scatter ratio for a single particle may not be 1. It is a source term
in the coupled radiation transfer equations. Thus,  back
sca in the above equations represents backscattered loss and is one-half
the regular scatter coefficient, sca. The random diffuse reflectance just above the film and the random diffuse transmittance
below the film are sought.
Differentiate Eq. 9 and use Eq. 10 to obtain
d 2 down
dx 2
– 2 down = 0 ,
(11)
2
2
back
where  2 =  ext
– `  sca
j and  ext =  abs +  back
sca . A similar equation is obtained by differentiating Eq. 10, thus
d 2 up
dx 2
– 2 up = 0 .
(12)
The general solutions to the above homogeneous Helmholtz equations are
JOHNS HOPKINS APL TECHNICAL DIGEST, VOLUME 32, NUMBER 5 (2014)
799­­­­
S. B. STRONG ET AL.
down ^ x h = Ae –x + Be x (13)
up ^ x h = Ce –x + De x ,
(14)
and
where
 =  back
sca
1
– 1 =  abs
a L2
 back
 back
1 + 2 sca , and a L = sca .
abs
ext
(15)
The random diffuse albedo is aL and represents the back hemisphere scatter loss. If the forward-to-back ratio is 1, then,
 sca
 back
sca = 2 .
(16)
For this reason, care must be exercised when comparing the backscatter coefficient in the Kubelka–Munk model to the regular single scatter coefficient and the Kubelka–Munk albedo to the regular albedo. Furthermore, Kubelka–Munk theory is an
approximation to a full radiation transfer equation solution in the diffuse limit. It is also expected that the average photon
path is greater than the film thickness. Thus, the absorption coefficient should be scaled also. A comparative study produced
corrections to bring Kubelka–Munk theory into close agreement with numerical solutions of the radiation transfer equation
in the diffuse limit. The Kubelka–Munk or Lambertian albedo is defined in the following manner:19
p
p sca
q  sca
.
(17)
a L = q + p =
p
abs
sca
 abs + q  sca
Knowing that p = 0.5, it was found that q = 1.22.
The random diffuse reflectance and random diffuse transmittance at some distance within the film are defined as
L ^x h =
up ^ x h
^xh
and L ^ x h = down
down ^ d h
down ^ d h
0 # x # d .
(18)
The boundary condition at the interface between the film and the substrate requires
Lsub = e
up ^0 h
o = C + D ,
A+B
down ^0 h
(19)
where Lsub is the Lambertian TIR of the substrate. Solving for the coefficients in Eq. 19, the following results are obtained
for a film on an opaque substrate:
L ^d h =
and
b – – Lsub –2d
e
b + – Lsub
b – Lsub –2d
1– –
e
b + – Lsub
b– – b+
b –
e 1 – b – – Lsub o e –d
+
Lsub
,
L ^0h =
b – – Lsub –2d
e
1–
b + – Lsub
where
1
b != a !
L
1
– 1 (Note: b+b_ = 1). a L2
(20)
(21)
(22)
In practice, the transmittance is more meaningful when there is no substrate, thus Lsub = 0. Then, for the case of a slab (e.g.,
a window),
800
JOHNS HOPKINS APL TECHNICAL DIGEST, VOLUME 32, NUMBER 5 (2014)
MODELING AND MEASUREMENTS OF LEAF BIDIRECTIONAL REFLECTIVITY
1
–d
2 –1e
a
L ^0h =
b + – b – e –2d
2
and
L ^d h =
1 – e –2d
.
b + – b – e –2d
(23)
(24)
Using the total power law, the random diffuse absorptance becomes
L ^0 h = 1 – L ^0 h – L ^ d h .
(25)
For the case of an opaque bulk material (such as a rough metallic surface), let d go to infinity. Then the diffuse reflectance
reduces to
 L ^d h =
1
= b – .
b+
(26)
ANOMALOUS DIFFRACTION APPROXIMATION
ADA yields a computationally efficient and robust approximation to Mie theory in the region of large spherical particles (x
large) and for particle refractive index that closely matches the background (van de Hulst). This is often the case in ocean
particle scattering and for water-based aerosols. It is based on plane wave propagation and Huygen’s principle. It is also
assumed that reflection and refraction can be ignored [that is, (m – 1) is small]. Thus the theory emphasizes diffraction and
interference effects that often dominate particle scatter phenomenon. Other shapes besides spheres are also possible, and
computing the extinction cross-section given a size distribution function is much faster.
The theoretical foundation begins with an incident electric field plane wave illuminating an arbitrarily shaped particle, thus
E i ^ z h = E i0 exp ^–jk l0 n 0 zh a x .
(27)
Inside the particle of refractive index n1 the plane wave becomes
E s ^ z h = E i0 exp ^–jk l0 n 1 zh a x .
(28)
Using the above fields, one may determine that the scatter amplitude leads to
kl 2
S ^0 h = 2
## ^1 – e–jkl n ^m – 1hz^x, yhhE^x, yhdxdy ,
0 0
(29)
where E(x,y) represents the projected area of the particle to the incident plane wave, z(x,y) is the path through the particle
n
and m = n 1 . In general, m is complex (m = mr – jmi). The corresponding extinction cross-section becomes
0
C ext = 2 Re ;
## ^1 – e–jkl n ^m – 1hz^x, yhhE^x, yhdxdyE .
0 0
(30)
For the case of a spherical particle, the chord length is computable. The resulting extinction efficiency is
1 e –w^x, mh e –w^x, mh – 1
Q ext ^x, mh = 4P Re = 2 +
+
G ,
w ^ x, mh
w ^x, mh2
(31)
where w(x,m) = –j2x(m – 1) for a nonabsorbing spherical particle of radius a ( real) is
4
4
Q ext ^ h = Q sca ^ h = 2 –  sin ^ h + 2 ^1 – cos ^ hh ,

(32)
2 n a
where = 2x ^m –1 h, x = k l a = 0 , a is the radius of the spherical particle, n1 is the refractive index of the particle,
and n0 is the refractive index of the background medium.
JOHNS HOPKINS APL TECHNICAL DIGEST, VOLUME 32, NUMBER 5 (2014)
801­­­­
S. B. STRONG ET AL.
m = 1.33
m = 1.55
4
3
Qsca
Qext
3
2
1
0
0
5
0
10 15 20 25 30 35 40
Size parameter x
1
5
10 15 20 25 30 35 40
Size parameter x
m = 1.33
m = 1.33-0.01i
m = 1.33-0.03i
m = 1.33-0.1i
3
Qext
2
0
4
m = 1.33-0.00i
m = 1.33-0.01i
m = 1.33-0.03i
m = 1.33-0.10i
3
Qext
2
1
4
0
m = 1.33
m = 1.55
4
2
1
0 10
20 30 40 50 60 70 80
Size parameter x
0
0
10 20 30 40 50 60 70 80
Size parameter x
Figure 19. Comparison between Mie theory (left) and ADA (right) for spheres.
The Authors
Shadrian B. Strong, presently in the Air and Missile Defense Sector, is a Co-Investigator (Co-I) for the BRDF independent research and development project. Dr. Strong’s expertise is in environmental/atmospheric characterization, remote
sensing, and EO/IR performance analysis for space systems. Michael E. Thomas is a principal staff engineer in the Air
and Missile Defense Sector. He assisted with algorithm development and analysis of BRDF observations. Dr. Thomas
is a specialist in electromagnetic theory, optical propagation, and quantum electronics and has research interests in
measurement and theoretical modeling of atmospheric propagation and remote sensing, optical properties of solids, and
high-pressure gases. Andrea M. Brown is in the Asymmetric Operations Sector and is an expert in atmospheric characterization and aerosol modeling. Dr. Brown’s expertise has been invaluable in BRDF analysis, algorithm development,
and observations. Elena Y. Adams, Co-I for the BRDF project, is a systems engineer in APL’s Space Exploration Sector
and helped with project formulation and research. For further information on the work reported here, contact Shadrian
Strong. Her e-mail address is shadrian.strong@jhuapl.edu.
The Johns Hopkins APL Technical Digest can be accessed electronically at www.jhuapl.edu/techdigest.
802
JOHNS HOPKINS APL TECHNICAL DIGEST, VOLUME 32, NUMBER 5 (2014)
The MSX/UVISI Stellar Occultation
Experiments: Proof-of-Concept Demonstration
of a New Approach to Remote Sensing of
Earth’s Atmosphere
Ronald J. Vervack Jr., Jeng-Hwa Yee, William H. Swartz,
Robert DeMajistre, and Larry J. Paxton
primary means of monitoring the atmosphere on a global
scale is remote sensing of Earth from space, and occultation methods based on observing changes in a star’s
spectrum as it sets through the atmosphere have proven valuable in this endeavor.
Two occultation techniques—focused respectively on atmospheric extinction and refraction—have been treated separately historically, but the combination of the two offers
the possibility of higher accuracy, greater altitude coverage, and simultaneous measurement of primary and trace gases, both of which are important in the chemical balance of
the atmosphere. Using data from the Ultraviolet and Visible Imagers and Spectrographic
Imagers (UVISI) on the Midcourse Space Experiment (MSX) satellite, we have developed
a combined extinctive/refractive stellar occultation technique that relates the changes in
measured stellar intensity to the height-dependent density profiles of atmospheric constituents and have demonstrated its viability for retrieving both primary and trace gases
through the analysis of approximately 200 stellar occultations. This self-calibrating technique is broadly applicable and suitable for measurement of gases that are otherwise
difficult to quantify on a global basis. In this article, we summarize the development of
the technique, its validation through comparison to other measurements, and its application to the study of Earth’s overall atmosphere and lower atmosphere in particular.
INTRODUCTION
The Earth’s atmospheric composition, temperature,
and ability to cleanse itself chemically have coevolved
with biological life on geological timescales.1 On much
JOHNS HOPKINS APL TECHNICAL DIGEST, VOLUME 32, NUMBER 5 (2014)
shorter timescales, however, anthropogenic activities have effected changes in the atmosphere. Much
of the observed change is associated with minor atmo-
803­­­­
VERVACK ET AL.
spheric constituents, referred to as trace gases, and their
impact on atmospheric chemistry. Although many of
the Earth’s trace gases exist naturally, the rapid rise in
the concentrations of both native and nonnative gases
and their observed effects on the atmosphere’s chemical
balance over the past 150 years or so is coincident with
industrialization and the advent of modern agricultural
practices.2 Whereas studies assessing the long-term consequences of these changes are ongoing, a significant
obstacle is often a lack of regular, global measurements
of the concentrations of these trace gases in the atmosphere, particularly in the lower atmosphere (less than
50 km altitude) where their concentrations and effects
are most significant. Measurement of emission spectra
from atoms and molecules is often used to probe atmospheric densities, but many of the important trace gases
in the atmosphere emit such radiation weakly or not
at all, or their emission signatures are convolved with
stronger emission from other atoms or molecules and
cannot be separated. As such, these trace gases can only
be detected through in situ sampling or via remote observation of their attenuation or scattering of light.
Stellar occultation is one of the most promising techniques to meet the needs of atmospheric scientists, policy
makers, and enforcement agencies tasked with understanding and regulating man-made contributions to the
trace gas atmospheric budget. Occultation techniques
have been used for many years to study the atmospheres
of Earth, other planets, and their satellites. Because these
techniques are generally based on relative measurements
(i.e., compared to measurements in the absence of the
occulting atmosphere), they tend to be less sensitive to
Wavelength range 100–900 nm
Generally retrievable
Under certain conditions
O2
Altitude (km)
200
60
Thermosphere
150
100
50
Aerosols
O3
Total
(N2,O2)
NO2
H2O
0
OCIO
SO2
NO3
Mesosphere
Stratosphere
Troposphere
40
Wavelength range
0.8–2.5 m
0.8–5.0 m
CO2 O2
50
Altitude (km)
250
instrument degradation and changes in calibration than
other remote sensing methods. Although these techniques are easy to implement in principle, technical
hurdles exist. As these challenges have been overcome,
occultation techniques have proven to be an excellent
means of exploring the overall structure of both terrestrial and planetary atmospheres.
Using occultation methods to probe Earth’s lower
atmosphere is complicated by the effects of both
refraction and extinction as light passes through the
densest part of the atmosphere. To demonstrate a new
approach to this challenge, we took advantage of the
Ultraviolet and Visible Imagers and Spectrographic
Imagers (UVISI) suite of instruments onboard the
Midcourse Space Experiment (MSX) spacecraft to
conduct stellar occultation experiments in which we
combined techniques suited to measuring refraction and
extinction independently into a single, self-consistent
occultation method that leverages the benefits of both
techniques. This demonstration led to the development
of an instrument concept suitable for flight on a small
space platform.
The potential measurement capabilities of this combined stellar occultation technique are illustrated in
Fig. 1. The left panel shows the retrievable altitude ranges
for several gas and aerosol constituents in the terrestrial
atmosphere that can be measured over far-UV to visible
wavelengths, whereas the right panel shows the altitude
ranges for gases that can be measured in the near-IR.
These altitude ranges are derived from an analysis of
the extinction spectra for the various species (a term
commonly applied to atmospheric constituents) and the
CH4 N O
2
H2O
30
20
O2
CO
10
0
Figure 1. Illustration of the measurement capabilities of the combined occultation technique. The left panel shows the retrievable altitude ranges for several species in the terrestrial atmosphere that can be measured over the wavelength range of 100–900 nm. The blue
bars represent ranges that are generally measurable under all conditions, whereas the red bars represent ranges that vary depending
on the atmospheric conditions (e.g., night/day, presence of high-altitude clouds) and the target star involved (e.g., spectral type, magnitude). Colored bars along the right y axis denote the altitude ranges of various atmospheric regions. The right panel shows the retrievable altitude ranges for several terrestrial species that can be measured if coverage is extended into near-IR wavelengths: 0.8–2.5 m
(blue bars) or 0.8–5.0 m (blue+red bars). In both panels, the actual lower limits of the altitude ranges are determined by cloud-top
heights at the time of the occultation; the limits shown are for clear skies. Note that the upper limits on CO2 and O3 in the right panel are
truncated to 50 km to allow the panel to focus on the lower altitudes.
804
JOHNS HOPKINS APL TECHNICAL DIGEST, VOLUME 32, NUMBER 5 (2014)
MSX/UVISI STELLAR OCCULTATION EXPERIMENTS
capabilities of an instrument optimized for their measurement, particularly in the lower atmosphere where
refraction effects must be accounted for in extinctive
occultation observations. We will explore the basis for
these altitude ranges as we discuss our proof-of-concept
measurements with MSX/UVISI throughout this article.
Owing to the low altitudes that must be reached
before near-IR extinction is significant, an occultation
technique combining extinctive and refractive measurements is both ideal and necessary: ideal because
measuring the bulk atmosphere and the trace species simultaneously provides a complete picture of the
atmosphere, and necessary because refraction effects are
intertwined with the extinction and must be separated.
Several hundred stars are bright enough and of an appropriate spectral class for far-UV and visible wavelength
occultations. As wavelength increases, there are fewer
stars of sufficient brightness or spectral class; approxi-
mately 50 remain suitable out to 2.5 m, and only 10–20
are viable out to 5 m. We can, however, take advantage
of the slow motion of Earth relative to the background
stars—it takes many days for Earth to progress in its
orbit such that a given star is no longer available as an
occultation source. Thus, a given star may be used many
times when it is observable, and the additional spectral
coverage out to near-IR wavelengths is worth consideration given the extended range of altitudes and species
it enables.
We first give a brief historical perspective on stellar
occultation techniques to set the stage for a description
of the MSX/UVISI stellar occultation experiments. We
then highlight the success of these proof-of-concept
measurements and explore the future possibilities for the
combined stellar occultation technique as a means of
probing Earth’s atmosphere.
Extinctive
Refractive
Star
Io(λ)
I1(λ)
I2(λ)
I3(λ)
I
Io
Star
λ
End of
atmosphere
Refraction
angle
Photometric
Relative motion
of Earth or sensor
Central
Flash
Star
0
Flux 1
Figure 2. Geometric illustrations of the principles of various stellar occultation methods. In all three panels, the star is to the left and far
enough away that the incoming rays can be considered parallel. The upper left panel shows the geometry of an extinctive occultation,
in which the flux Io() from a star is progressively attenuated as the line of sight to the star moves deeper into the atmosphere. The ratio
of the measured flux I at each point to the stellar flux is I/Io, known as the atmospheric (or stellar) transmission. Transmission values run
from 0 (full extinction) to 1 (no extinction) and are a function of wavelength because the extinction by the atmosphere depends on
the wavelength. The upper right panel shows the geometry of a refractive occultation, in which the angle between the observed ray
and the unrefracted ray (the incoming stellar ray) is measured. The deeper in the atmosphere the line of sight penetrates, the larger the
measured refraction angle will be. The lower panel shows the geometry of a photometric occultation, which is a variant on the refractive
occultation. In this case, rather than measure the refraction angle of the star, the total flux from the star is measured. As the rays from
the star diverge due to refraction, the relative flux measured at the sensor drops from a value of 1 outside the atmosphere to 0 when the
star is blocked by the planet. This is because diverging rays fall outside the sensor field of view, and the greater the divergence, the less
flux that is measured. The central flash often seen in the total flux is caused by the convergence of rays refracted around opposite sides
of the planet even though the star itself is completely blocked.
JOHNS HOPKINS APL TECHNICAL DIGEST, VOLUME 32, NUMBER 5 (2014)
805­­­­
VERVACK ET AL.
THE STELLAR OCCULTATION TECHNIQUE
Two general categories of occultations exist: extinctive and refractive. In both, radiation from a source is
affected by physical processes in the atmosphere, and
measurements of the outgoing radiation are used to
infer atmospheric properties. In the following sections,
we briefly describe the techniques that have evolved for
these two types of occultations. Although the Sun, stars,
and radio signals have all been used as the sources in
occultations, we focus on stellar occultations for reasons
that will be made apparent in the discussion.
Extinctive Stellar Occultations
The principles of extinctive stellar occultations are
similar to those of classical absorption spectroscopy.
As shown in the upper left of Fig. 2, a star is used as
the light source, a spectrometer serves as the detector,
and the intervening atmosphere acts as the absorption
cell. The spectrometer is most commonly mounted on
a spacecraft but may also be on Earth’s surface or on a
balloon, in which case the effective absorption path is
half that shown in the figure. Starlight passing through
the atmosphere is attenuated because atmospheric constituents either absorb or scatter the incoming light.
Although the physical process involved is different,
the net effect of extinction by scattering (molecular/
Rayleigh or aerosol/Mie scattering) is similar to that
by absorption (i.e., scattering “out of the beam” reduces
light intensity far more than scattering “into the beam”
increases it); therefore, the term extinction as used here
refers to any absorption or scattering process. As the
line of sight to the star moves deeper into the atmo-
sphere, the light is progressively attenuated because the
effective extinction path length increases. At the same
time, changes in the densities of the species responsible
for extinction along the path are recorded in the measured spectra. The ratio of an attenuated spectrum, I,
to the unattenuated spectrum, Io, is referred to as the
atmospheric (or stellar) transmission and is the fraction
of light that passes through the atmosphere as a function of wavelength. The minimum altitude to which a
ray penetrates is referred to as the tangent point height
or tangent point altitude of the measurement, or more
simply as the tangent point. Note that in this article,
we use altitude to refer to the actual altitudes relevant to
atmospheric density profiles, whereas tangent point altitude or height is used to refer to the minimum altitude
of an observation. The distinction is subtle but will be
made clear as we discuss the observations.
Because the extinction cross sections of most species
are wavelength dependent, spectral measurements of the
transmission represent a “fingerprint” of the relevant
species. Thus, they are diagnostic of the atmospheric
composition, and densities of the species involved can
be determined from the observed transmission spectra.
Figure 3 illustrates the wavelength dependence of the
altitude at which the density of several species results in
a specific level of extinction, in this case the  = 0.1 level.
Extinction in the atmosphere is generally proportional
to the function exp(–()), where  is the wavelength
and  is referred to as the optical depth; thus,  = 1 refers
to the point in the atmosphere at which the total flux
from a star at a given wavelength has decreased to 1/e
of its unattenuated value. Because the transmission is
related to the optical depth, this figure is also illustrative
200
Altitude (km)
150
O2
100
O3
NO3
50
NO2
0
100
H2O
SO2
NO2
OClO
200
300
400
500
Wavelength (nm)
600
700
800
900
Figure 3. Altitudes at which the line-of-sight optical depth for several species in the terrestrial atmosphere reaches the  = 0.1 level
( = 0.01 for OClO and NO3) as a function of wavelength. Because extinction is generally proportional to the function exp(–), the  = 1
level is that point in the atmosphere at which the total flux from a star at a given wavelength has decreased to 1/e of its unattenuated
value. Because the extinction character of different atmospheric species varies with wavelength, plots such as this are helpful in designing occultation instruments because they highlight where extinction is most likely and indicate the altitude ranges in the atmosphere
that it is possible to probe (e.g., see Fig. 1).
806
JOHNS HOPKINS APL TECHNICAL DIGEST, VOLUME 32, NUMBER 5 (2014)
MSX/UVISI STELLAR OCCULTATION EXPERIMENTS
of the range of altitudes over which the density profiles
of such species may be inferred from transmission spectra. Extinction as a function of wavelength and altitude
for the species in Fig. 3 and for other species form the
basis of the altitude ranges illustrated in Fig 1. With a
properly designed instrument, accurate retrievals can
generally be carried out over transmissions ranging from
1 to 99% ( ranges of roughly 0.01–4.5).
The possibility of using stars as the source in an
extinctive occultation observed by a space-based platform was first suggested by Hays and Roble.3 The first
stellar occultation measurements of the terrestrial atmosphere came a few years later when a spectrometer on
the Orbiting Astronomical Observatory (OAO)-2 satellite was used to infer density profiles for molecular
oxygen (O2) and ozone (O3) in the thermosphere and
upper mesosphere.4–6 Since OAO-2, extinctive stellar
occultations have yielded densities for a number of species in Earth’s atmosphere, including O2, O3, NO, NO2,
NO3, N2O, H2, CH4, CO, H2O, Cl, and OClO.7–10 As
mentioned above, because many of these species do not
exhibit emission features that are directly indicative of
their density, the stellar occultation technique provides
a unique insight into atmospheric composition. The
technique has also been a primary means of probing
the atmospheres of the other planets and their satellites. The ultraviolet spectrometers on Voyagers 1 and 2
observed stellar occultations by the atmospheres of Jupiter, Saturn, Uranus, Triton, and Ganymede.11, 12
Refractive Stellar Occultations
Refractive stellar occultations occur because density gradients in the atmosphere lead to refraction, or
bending, of the path of incoming starlight. This results
in the light following curved paths through the atmosphere, which leads to differences between the star’s true
(geometric) and apparent (refracted) positions. Measurements of the degree to which the path of the incoming
starlight is changed provide the bulk properties of the
atmosphere (i.e., total density, pressure, temperature).
Pannekoek13 first realized the potential of refraction for
studying planetary atmospheres, but roughly 50 years
would pass before technology progressed to the point
that useful observations could be made.
The primary refractive stellar occultation technique is
the photometric approach (see the lower panel in Fig. 2),
which involves visible-light observations of the occultation of stars by a planetary atmosphere using ground, aircraft, or space-based telescopes. This approach has been
applied primarily to the study of the atmospheres of the
other planets. As a star passes behind a planet, light rays
passing through deeper regions of the atmosphere are
refracted more than rays at higher altitudes. This results
in a divergence of the incoming parallel light from the
star. At the detector, the divergence appears as an atten-
JOHNS HOPKINS APL TECHNICAL DIGEST, VOLUME 32, NUMBER 5 (2014)
uation of the light as a function of time (referred to as
refractive attenuation) as more and more of the total
flux from the star is refracted out of the field of view of
the sensor. Photometric observations of this attenuation
yield the so-called occultation light curve, which may
be inverted to retrieve the atmospheric density, pressure,
and temperature.
The first useful application of this technique was by
Baum and Code,14 who determined the mean molecular weight of Jupiter’s atmosphere from an occultation of
the star  Arietis. Since then, numerous occultations of
stars by other planets have been observed, and the technique has evolved into a powerful method for remotely
probing planetary atmospheres. The literature is extensive; however, a representative sample may be found in
Refs. 15–18. Summaries of the general methods and
much of the early work are given in Refs. 19 and 20.
Often, these photometric occultations yield light
curves that exhibit rapid fluctuations in intensity. These
fluctuations are caused by a refractive phenomenon
known as scintillation, which arises from small-scale
variations in the density profile at points along the line
of sight. Such variations may be caused by atmospheric
turbulence or by the presence of atmospheric waves.
When combined with models of such variations, measurements of the scintillation have proven useful in
inferring the small-scale atmospheric structure.16
Earth’s Lower Atmosphere
Historically, extinctive stellar occultations have been
applied to high altitudes for three reasons. First, the
species involved (e.g., O2 and O3 for Earth) have strong
spectral signatures at UV wavelengths, which are completely absorbed by the atmosphere before the lower altitudes are reached. Second, the presence of atmospheric
emissions (e.g., airglow) from other species that occur
in the instrumental bandpass used in the occultation
experiment leads to an additional source that competes
with the stellar signal. Once the line of sight to the star
has passed below the atmospheric level of such emission,
which ranges in altitude depending on the wavelength
and species involved, the emission is always present as
a contaminating source. Separation of the stellar and
emission sources is therefore necessary and can be
complicated. Finally, there are no significant refractive effects at high altitudes, allowing a straightforward
interpretation of the observed transmission.
Refractive stellar occultations have been limited
almost exclusively to the study of other planets. The
primary reason for this is that the long distances
involved allow for a greater separation of the diverging
rays and a more easily measured change in the stellar
intensity. At the same time, the choice of visible wavelengths for these occultations generally excludes the
need to consider extinction of the starlight within the
atmosphere itself.
807­­­­
VERVACK ET AL.
These extinctive and refractive methods, however,
are complementary, and consideration of both processes
would allow stellar occultation measurements to probe
a larger altitude range more accurately than either one
taken individually. In particular, the idea of using extinctive stellar occultations to probe Earth’s lower atmosphere by considering the effects of refraction on the
measurements was first proposed by Hays and Roble.3
They favored stellar occultations over solar occultations, which have been the primary occultation method
of studying Earth’s lower atmosphere, for a number of
reasons. Stellar occultations are not limited to the terminator and allow for hundreds of occultations per day
with near-global coverage versus the 30 or so daily using
solar occultations. Stellar occultations can also yield
higher vertical resolution owing to the point nature of
the source; the Sun subtends an angular range of half a
degree, which corresponds to about 30 km in altitude.
Solar occultations do yield higher signal-to-noise ratio
observations owing to the intensity of the Sun relative
to stars, and solar occultations are generally immune to
the effects of atmospheric emissions because the Sun is
so much brighter than the atmosphere. However, in the
lower atmosphere where refraction is significant, stellar occultations have the advantage over solar occultations in that the refractive effects can be treated more
easily for stars (as point sources) than for the Sun (as an
extended source).
Although only a theoretical study, the Hays and
Roble work demonstrated that the extinctive stellar occultation technique could be a powerful tool for
remotely sensing Earth’s atmosphere. However, the
implementation requirements for this technique, especially in the lower atmosphere, are rather stringent.
Because stars are often a relatively low-intensity source,
the instrument has to have a high signal-to-noise ratio
to allow occultations by moderately bright stars, thereby
maximizing the number of available stars. A high-precision, two-axis system is needed to acquire and track
the star as it sets through the atmosphere, including the
apparent motion induced by refraction. To infer altitude profiles, multi-wavelength measurements at moderate to high spectral resolution are needed to separate
the spectral signatures of the various atmospheric species. The presence of atmospheric emissions requires a
hyperspectral imaging system with both spectral and
spatial dimensions. Finally, to track the star even in the
presence of refraction, the spacecraft, instrument, or
both must have excellent stability and jitter compensation. Reducing this idea to practice had to wait until
the MSX opportunity.
MSX/UVISI STELLAR OCCULTATIONS
The MSX satellite—designed, built, and tested at
the Johns Hopkins University Applied Physics Labo-
808
ratory (APL) for the Ballistic Missile Defense Organization—was launched into a 900-km, circular,
near-Sun-synchronous orbit on April 24, 1996. As a
three-axis stabilized spacecraft, MSX was capable of
extraordinarily accurate and stable pointing. Inertial
targets such as stars were observable with an absolute
accuracy approaching ~100 rad and a relative stability
of ~10 rad (1 ) over a time span of several minutes.
The UVISI instrument package consisted of four visible
and UV imagers (two narrow field-of-view and two wide
field-of-view) and five spectrographic imagers (SPIMs)
covering the wavelength range of 120–900 nm. All nine
instruments were co-aligned within 50 rad along the
long axis of the spacecraft and mounted on a common
optical bench to maintain the co-alignment. Thorough
descriptions of MSX and the UVISI instruments may be
found in Refs. 21–23.
We took advantage of the capabilities of MSX and
UVISI to test the suggestions of Hays and Roble3 and
thereby demonstrate the viability of using stellar occultations for the retrieval of atmospheric composition,
particularly in the lower atmosphere. Although the
MSX spacecraft and the UVISI suite of instruments
were not specifically designed for stellar occultations,
their combination fortuitously satisfied the stringent
requirements necessary to conduct these experiments.
Rather than simply use models or climatological information to treat the effects of refraction on the extinctive measurements, the MSX/UVISI experiments
focused on a combined extinctive and refractive stellar occultation technique in which SPIMs were used
to measure the wavelength-dependent atmospheric
extinction of starlight while a co-aligned imager (IVN)
was used to measure the atmospheric refraction along
the same line of sight.
The key element in this new approach to stellar occultations was the direct measurement of the refraction
angle of the star at visible wavelengths, as suggested by
Jones et al.24 Although simple in concept, this method
of refraction angle measurement has previously been
used in only a few, mostly serendipitous, instances.25, 26
The MSX/UVISI experiments represent the first systematic implementation of the technique. The refraction angles themselves are used to infer atmospheric
refractivity and subsequently the bulk atmospheric density, pressure, and temperature. Although the temperature profiles determined from such measurements are
unlikely to ever be as accurate as profiles derived from
radio occultation methods such as those used in GPS/
MET (Global Positioning System/Meteorology),27, 28
the advantage of this visible-light technique is that the
refraction angles are measured simultaneously along the
same line of sight as the extinction measurements.
By simultaneously measuring both refraction and
extinction effects on the stellar spectrum, the extinction
measurements were able to probe the atmospheric com-
JOHNS HOPKINS APL TECHNICAL DIGEST, VOLUME 32, NUMBER 5 (2014)
MSX/UVISI STELLAR OCCULTATION EXPERIMENTS
position accurately at the low altitudes where refraction
is significant. Measurements of refraction provided the
actual light path through the atmosphere and the bulk
atmospheric properties, both of which are necessary to
determine extinction accurately in the refractive region
of the lower atmosphere. More importantly, knowledge
of the total density profile allowed the effects of Rayleigh scattering in the transmission spectra to be handled accurately. The simultaneous measurement of both
constituent and total density profiles also allowed for
direct calculation of volume mixing ratios, a key quantity that typically has to be inferred and is therefore
subject to increased uncertainty.
The design of the MSX/UVISI stellar occultation
experiments and the associated extinction and refraction retrieval algorithms are discussed in detail in a
three-paper series.29–31 A full discussion of the combined
technique, the development of the retrieval algorithms,
and a rigorous assessment of the uncertainties may be
found in these papers. In this article, we present an overview of the technique and analysis methods along with
a summary of several results that highlight the potential
of the combined extinctive and refractive stellar occultation technique for studying Earth’s atmosphere.
Typical MSX/UVISI Occultation Event
line of sight to the star (dashed line) at that point actually had a tangent point altitude below the horizon.
The tangent point altitude of the direct line is referred
to as the geometric tangent point altitude and can be
negative. When refraction is not important (i.e., above
35 km), the apparent and geometric tangent point altitudes are the same.
As illustrated in Fig. 4, airglow emissions are an
issue. These first appear in the bottom edge of a spectrographic slit as it descends through the altitudes at
which the emissions originate (e.g., the airglow layer).
The peak in the emission layer signal will continue
to move up the slit, eventually disappearing above
the top edge. However, once the line of sight passes
through the airglow layer, the stellar signal is superimposed on an ever-present background airglow signal.
Because each spatial pixel covers a field of view of
~0.025° (about 1.5 km projected onto the limb), the
UVISI SPIMs are essentially used as large-aperture
photometers (~100 cm2) with an effective field of view
of 0.1°  0.025°. With a maximum star setting rate of
roughly 0.07°/s (~3 km/s), the limb airglow emission at
a given tangent point altitude is sampled for approximately 10 s. On the ground, this multiple sampling of
the airglow allows us to “shift” the SPIM pixels spatially
and co-add them in a manner similar to that used for
a time-delay and integration system, thereby improving
the signal-to-noise ratio of the observations and subsequent accuracy of the separation of the stellar and
airglow signals.
Figure 5 shows a series of images acquired at different
times during a typical MSX/UVISI occultation to illustrate the extinction of stellar light with altitude. The
panels show composite SPIM spectral images together
The geometry of a typical MSX/UVISI stellar occultation event is shown in Fig. 4. The star was acquired at
high tangent altitudes and the boresights of the UVISI
instruments remained fixed on the star’s inertial position during the entire occultation. The entrance slits
to the five spectrographs were held vertical with respect
to the horizon. Several hundred unattenuated Io spectra were obtained at high
altitudes before any atmo1
spheric effects began. As the
2
1
star set through the atmo3
2
1
4
sphere, its spatial position
3
4
in the spectrograph images
0.1°
was unchanged until refrac5
5
tion set in at altitudes below
UVISI
Airglow
roughly 35 km, after which
SPIM
layer
images
the image drifted vertically
MSX orbit
up the slit until it slowly
disappeared from the field
of view. For a star inertially Figure 4. Schematic illustration of the MSX/UVISI stellar occultation geometry. The UVISI instrufixed in the center of the ment is oriented so that the slit is vertical with respect to the atmosphere horizontal. The center of
slit, roughly 0.5° of motion the slit is fixed on the star’s inertial position throughout the occultation; therefore, the star itself
due to refraction could be can refract up to 0.5° before disappearing from the field of view. The 1° field of view in the vertiobserved before the star’s cal direction corresponds to ~60 km at the limb. The superposition of the background airglow
image reached the edge of (magenta) and stellar signals once the line of sight has passed through the airglow layer is illusthe frames, corresponding trated. The minimum height of the solid rays above Earth’s surface (blue) reflects the apparent
to apparent tangent point tangent point altitude, whereas the dashed line between the spacecraft and the star is an example
altitudes of 7–8 km. As seen of the geometric tangent point altitude, which can be negative. In the absence of refraction, the
in path 5 of Fig. 4, the direct two quantities are equal. (Modified from a figure published in Ref. 29.)
JOHNS HOPKINS APL TECHNICAL DIGEST, VOLUME 32, NUMBER 5 (2014)
809­­­­
VERVACK ET AL.
Spectral Image
300
10,000
280
1,000
260
220
0.010
10
(a)
1.000
0.100
100
240
Image of star
0.001
140
10,000
1.000
120
1,000
100
100
0.100
80
10
Tangent Height (km)
60 (b)
1
80
Brightness(R/nm)
1
200
10,000
60
1,000
40
100
20
10
0
–20 (c)
0.010
0.001
1.000
0.100
0.010
1
0.001
40
10,000
1.000
20
1,000
0.100
0
100
–20
–40 (d)
10
200
400
600
Wavelength (nm)
800
1
0.010
0.001
Figure 5. Series of UVISI images during a typical occultation event. The images on the left are
from the spectrographic imagers and illustrate how the star’s spectrum (the bright horizontal line
in the center of each image) is attenuated as a function of wavelength (left to right) and altitude
(top to bottom). Note that the tangent height scale changes with each panel to be centered on
the tangent height of the unrefracted star (i.e., its inertial position). Thus, the spectrum has moved
“up” relative to the panel center in the final panel. Bright lines in the vertical direction are emissions
from Earth’s atmosphere (i.e., the background airglow). The images on the right are from the IVN
imager and show the position of the star corresponding to the spectrographic imager measurements on the left. Note that the star moves up away from the image center as refraction begins.
The refraction angle is the angle between the center and measured positions. (Modified from a
figure published in Ref. 29.)
with an IVN image acquired simultaneously. Scene
brightness is represented by a color scale as a function
of geometric tangent altitude (y axis) and wavelength (x
axis) in the SPIM images. The star appears as a bright,
narrow, horizontal band in the composite SPIM images
because it is a point source emitting a continuous spectrum at all wavelengths, whereas the airglow signal
covers an extended portion of the slit vertically owing
to the diffuse nature of the airglow emission. The ragged
horizontal edges of the SPIM images result from small,
well-known misalignments of the five SPIM boresights
plus corrections for optical aberrations when referencing
the stellar signal at each wavelength to a common tan-
810
gent point altitude. Black areas in Fig. 5 correspond to
altitudes outside the SPIM fields of view, and a decrease
in the SPIM 5 sensitivity results in the weak signal
at the longest wavelengths of the spectral range. The
star of interest is the bright dot near the center of each
IVN image.
The data in Fig. 5a correspond to a geometric tangent
altitude of 245 km. There is essentially no atmospheric
attenuation of the star at this altitude; however, several
airglow emissions are visible in the spectrum [e.g., geocoronal H Lyman  at 121.6 nm, nightglow O(1S) “green
line” at 557.7 nm, and nightglow O(1D) “red lines” at
630.0 and 636.4 nm]. Spectra from such altitudes and
JOHNS HOPKINS APL TECHNICAL DIGEST, VOLUME 32, NUMBER 5 (2014)
MSX/UVISI STELLAR OCCULTATION EXPERIMENTS
higher provide the unattenuated stellar Io spectra needed
to determine the atmospheric transmission.
In Fig. 5b, the line of sight has descended to a tangent
altitude near 95 km. Complete attenuation of the star
by O2 is evident in the Schumann-Runge continuum
and bands shortward of 200 nm, whereas absorption
by O3 in the Hartley/Huggins bands is just beginning
near 255 nm. Emissions in Earth’s mesospheric airglow
layer are now visible [e.g., O(1S) “green line,” O2
Atmospheric (0-0) band at 762 nm, several OH Meinel
bands in the near-IR]. The mesospheric emissions
cover only the lower portion of the SPIM images
because the upper part of the slit still views tangent
heights above the mesospheric airglow emission layer.
Even though the stellar signal short of 200 nm is gone,
the H Lyman  geocoronal line is still present owing
to emission that originates above the tangent point
and extends into the geocorona. From the geometry of
Fig. 4, it can be seen that this “near-field” source also
exists for the airglow layer, explaining why many of the
airglow emissions are only moderately attenuated as the
occultation progresses.
Figure 5c shows the image when the star has reached
a geometric tangent height of 20 km. In the lower
stratosphere, the O3 column density has increased to
the point that absorption in the Hartley/Huggins bands
is complete, whereas absorption in the relatively weak
Chappuis band near 600 nm has now become significant. The star has also moved in the SPIM and IVN
images, shifting slightly upward as a result of refraction.
Note that the second star in the upper right corner of
the IVN image is at a higher tangent height and has not
yet been refracted. Finally, Rayleigh extinction, refractive attenuation, and scintillation effects are becoming
increasingly important as evident in the overall reduction of the stellar signal.
In Fig. 5d the geometric tangent altitude of the star
has descended to 0 km, but refraction is sufficient to
keep the star visible above the horizon. The refractive
shift of the star’s position is evident not only in the
IVN image but also in the SPIM data where the bright
line delineating the stellar spectrum has moved upward
relative to the slit center, which is fixed on the star’s
inertial position. The star’s apparent tangent height is
roughly 13 km as shown in SPIM 5. The stellar signal
continues to weaken and is visible only at the longer
wavelengths. Both Rayleigh scattering and refractive
attenuation act to block the majority of the stellar
signal at tropospheric altitudes. At the same time, the
relative constancy of the “near-field” airglow emissions
from panel to panel results in a comparable, or even
brighter in the case of 762 nm, signal relative to the
star. This ultimately results in a decreased signal-tonoise ratio in the wavelength channels of the stellar
transmission spectra contaminated by airglow once the
two signals are separated.
JOHNS HOPKINS APL TECHNICAL DIGEST, VOLUME 32, NUMBER 5 (2014)
Each MSX/UVISI occultation consists of images
such as those presented in Fig. 5. Before the atmospheric
retrievals, the spectral images must be processed to
remove a number of instrumental effects (e.g., changes
in gain with altitude, dark subtraction), and the stellar
spectra must be separated from the contaminating airglow background. Accomplishing this requires knowledge of the true stellar position in the SPIM images,
which is calculated from the measured position of the
star in the IVN images and known relationships in the
relative alignments of the various instruments (thereby
emphasizing the need to both know and maintain the
relative co-alignment of the SPIMs and imager).
Data Reduction and Analysis
The data reduction and analysis process, as implemented in separate subsystems on MSX, is shown
schematically in Fig. 6 and can be broken down into
four general steps. First, the IVN images are processed
to determine the angles through which the star is
refracted at all altitudes. This is accomplished using the
known inertial position of the star and the measured
positions of the star in the IVN images. Second, the
SPIM spectra are processed to yield stellar transmission
spectra. The most important aspect of this process is
the separation of the stellar and airglow signals, which
is possible owing to accurate measurement of the star’s
true position in the IVN and SPIM images, knowledge
IVN images
Co-aligned
SPIM images
Star
locations
Airglow
Refraction angles
Density profile
Pressure profile
Temperature
profile
Star
Star absorption ray path
Refractive attenuation correction
Rayleigh scattering
Absorption cross sections
Scintillation
profile
Composition
profiles
Figure 6. Schematic illustration of the combined stellar occultation data analysis process. Refraction information from the IVN
images yields the bulk atmospheric properties of total density,
pressure, and temperature. The SPIM images are processed to
yield separate images of the star and background airglow, and
the star images are then processed to yield the composition profiles for various species. The feedback of the bulk properties to
the composition retrievals is indicated, and it is that feedback that
allows the accurate determination of profiles in the lower atmosphere. (Reproduced from Ref. 29.)
811­­­­
VERVACK ET AL.
300
1.00
200
0.10
50
0
200
400
600
Wavelength (nm)
800
0.01
1,000
200
100
100
10
Brightness (R/nm)
100
Geometric tangent height (km)
150
Transmission
Geometric tangent height (km)
10,000
1
0
200
400
600
Wavelength (nm)
800
Figure 7. Stellar transmission (left) and airglow (right) spectra as a function of geometric tangent height and wavelength as observed
by MSX/UVISI during the typical occultation event described in the text. Notice how the extinction edge of the starlight as a function
of wavelength and tangent height matches the shape of the optical depths of the species in Fig. 3. Geometric tangent height refers to
the tangent height of the unrefracted (i.e., inertial) stellar position, which is why the altitudes in the left panel still show stellar signal at
negative altitudes (i.e., the star has gone below the horizon but refraction of the light allows signal to still reach the instruments). Note
the different y axis scales in the two panels. (Reproduced from Ref. 29.)
of the spatial point spread functions of the five SPIMs,
and the multiple sampling of the airglow data afforded
by the spatial imaging capabilities of the SPIMs. Equally
important in this process is the determination of the
spectral uncertainties given that the stellar and airglow
signals come from two different statistical distributions.
Third, the bulk atmospheric properties (density, pressure, and temperature) are retrieved from the refraction angles using the technique developed by Vervack
et al.31 Finally, the O3 profile is retrieved from the stellar
transmission spectra using the technique described by
DeMajistre and Yee.30 It is important to note that the
bulk properties from the refraction retrieval feed into
the extinction retrieval, thereby allowing the proper
temperature-dependent cross sections to be used, the
actual refractive path to be calculated, and the Rayleigh
scattering to be accurately accounted for when inferring
the O3 profile.
Although each step is executed separately, interconnections with the other steps exist and are critical to
the combined occultation retrieval process. The position of the star in the high-angular-resolution (0.005°/
pixel) IVN images not only yields refraction angles with
sub-pixel accuracy but also provides the exact location
of the star in the SPIM slits. Knowledge of the true position of the star in all the SPIM images is necessary in
separating the airglow spectra from the stellar spectra,
but this is especially the case once the star begins to
move in the slit because of atmospheric refraction. The
index of refraction profile retrieved directly from the
observed refraction angles allows determination of the
actual light path in the atmosphere for extinction calculations as well as the retrieval of the bulk atmosphere
812
density and temperature profiles. Finally, the retrieved
bulk density profile constrains the amount of Rayleigh
scattering inferred from the transmission spectra.
Figure 7 shows the separated transmission (left
panel) and airglow volume emission (right panel) spectra for the typical occultation event. Absorption by O2
and O3 is clearly evident in the transmission spectra
near 150 nm (thermospheric O2, Schumann-Runge
continuum), 255 nm (mesospheric O3, Hartley band),
and 600 nm (stratospheric/upper tropospheric O3,
Chappuis band). Atmospheric scintillation causes the
signal intensity to fluctuate below 40 km and results
in the horizontal stripes in the atmospheric transmission. Because of atmospheric refraction, the star
is observed even after the geometric line of sight has
passed below the horizon (i.e., at negative geometric
tangent heights). Abrupt changes in the noise characteristics and visual appearance of both the transmission and airglow spectra (e.g., at the juncture between
SPIMs 3 and 4 near 380 nm) are caused by variations
in wavelength sampling interval, instrument sensitivity, and resolution between the SPIMs. The peculiar
structure near 760 nm is caused by difficulty in separating the stellar signal from the strong O2 Atmospheric
band emission.
Density profiles for the various species can be retrieved
from transmission spectra such as those in Fig. 7. Above
60 km or so, extinction is effectively only due to absorption by O2 and O3, there are no refraction effects, and the
airglow contamination is relatively weak at the shorter
wavelengths where extinction is present. Retrievals at
these geometric tangent altitudes are thus relatively
straightforward. Below 60 km, however, absorption from
JOHNS HOPKINS APL TECHNICAL DIGEST, VOLUME 32, NUMBER 5 (2014)
MSX/UVISI STELLAR OCCULTATION EXPERIMENTS
JOHNS HOPKINS APL TECHNICAL DIGEST, VOLUME 32, NUMBER 5 (2014)
30
25
20
Altitude (km)
10
15
100
10
180
0
200
220
240
Temperature (K)
260
O3 volume mixing ratio (ppm)
1
2
3
4
280
5
40
35
30
10
25
20
Altitude (km)
The results of the refraction and extinction retrievals for a typical occultation are presented in Fig. 8,
with the temperature profile shown in the top panel
and the O3 density profile shown in the bottom panel.
Both MSX/UVISI profiles (red) are compared to profiles measured by sonde at the McMurdo, Antarctica,
ground-based station (blue). McMurdo was nearly coincident in both time and space to the tangent point of
the MSX/UVISI occultation. Note that the retrieved
temperature profile does not extend all the way to the
surface because the star refracted out of the imager field
of view at apparent tangent altitudes between 6 and
7 km. In a follow-on instrument, direct tracking of the
refracted star position or a larger field would capture
the full range of motion in the refracted stellar position and enable temperature profile retrievals closer to
the surface (most likely limited by cloud heights when
clouds are present). Also shown in the bottom panel
is the MSX/UVISI-determined O3 mixing ratio profile
calculated using the density profile from the refraction retrieval. A significant advantage of the combined occultation technique is the direct calculation
of mixing ratios using a simultaneously measured total
density profile, providing a more accurate measure of
mixing ratio than current methods that rely on model
total density profiles.
Another significant point regarding Fig. 8 is that the
temperature and O3 profiles are plotted versus both altitude and the retrieved pressure. This illustrates another
important advantage of the combined extinctive/refractive technique: the pressure–altitude relationship is
firmly known. The altitude grid used in the density profile retrievals is accurately established from the observation geometry, whereas the pressure profile is integrated
directly from the measured total density profile and need
not be inferred, as is commonly required in alternate
methods. Such accurate knowledge of the pressure–altitude relationship increases the utility of the retrieved
profiles in atmospheric models based on either pressure
or altitude.
As with any measurements, there are several types
of uncertainty that must be considered in the retrievals. Systematic uncertainty results from approximations
made in the retrieval algorithms, imperfect knowledge
of instrument characteristics, and the various assumptions incorporated into the retrievals. These are difficult to evaluate and often lead to biases when compared
35
Pressure (mbar)
Examples of Retrieved Profiles
40
Pressure (mbar)
several species is intertwined with Rayleigh and aerosol
scattering, refraction begins to have significant effects,
and airglow contamination is more evident. Although
these factors complicate the retrievals, the composition
and structure of the atmosphere can be determined by
using the data from both the SPIMs and IVN.
15
100
10
0
1
2
3
O3 number density (1012 cm–3)
4
Figure 8. Comparison of temperature and O3 profiles retrieved
from a typical MSX/UVISI stellar occultation event (red) to the
profiles measured by sonde at the McMurdo, Antarctica, groundbased station (blue). The agreement is good at all altitudes,
including some of the finer altitude structure in the density profile and particularly in light of the different spatial scales (i.e.,
in situ vs. a limb observation). The bottom panel also shows the
MSX/UVISI O3 mixing ratio profile (green) calculated using the
total density profile retrieved from the refraction measurements.
Uncertainties (1-) in the profile are shown in both panels as
horizontal error bars; pressure/altitude uncertainties are on the
order of the size of the symbols or smaller and are not visible at
this scale.
to results from other methods. Statistical uncertainty
results from the uncertainty in the measured transmission spectra and stellar positions as they propagate through the retrievals into the uncertainty of the
derived profiles. For example, the finite precision of
the MSX/UVISI total density and temperature profiles
provided by the refraction measurements will have an
813­­­­
VERVACK ET AL.
200
Altitude (km)
impact on the retrieved density profiles for the atmospheric species that can be properly accounted for (see
the papers by DeMajistre and Yee30 and Vervack et al.31
for detailed discussions of the uncertainty sources and
error analysis). As can be seen in the resulting statistical uncertainties in Fig. 8, however, the typical MSX/
UVISI occultation event provides fairly accurate results
despite the fact that MSX/UVISI was not optimized for
these observations; an optimized instrument would do
even better.
150
100
50
106
SCIENTIFIC RESULTS
Wide-Ranging Altitude Coverage
During May 5–7, 1997, MSX/UVISI obtained 15
stellar occultations at a single latitude but spaced longitudinally. Shown in Fig. 9 are the weighted mean
density profiles for O2 and O3 for this series of occultations. Although previous stellar occultation experiments
have also yielded O2 and O3 profiles, the MSX/UVISI
results are significant because of the large altitude range
they cover.
Earlier measurements4, 5 of O2 were limited to altitudes above 100 km or more. The MSX/UVISI results
extend down to roughly 50 km owing to two factors.
First, the wavelength coverage of the SPIMs extends
to longer wavelengths with moderate spectral resolution, allowing use of the Schumann-Runge bands and
the Herzberg continuum. Second, the SPIMs generate
hyperspectral images that allow for the separation of
the airglow signal from the stellar transmission. The
spatial dimension of these images provides measurement of the airglow at altitudes above and below the
line of sight to the star such that the airglow signal can
be determined and removed from the pixels containing
the stellar signal. Once the airglow signal is removed,
the Schumann-Runge band and Herzberg continuum
regions of the spectrum represent a purely stellar signal,
and the O2 profile can extend down into the mesosphere. Because the bulk density profile inferred from
the refraction measurements can be converted to the
equivalent O2 density owing to the well-mixed nature
of the lower atmosphere, the O2 density profile can be
effectively measured from 200 km or so down to the surface or cloud top heights with only a small gap between
roughly 50 and 30 km.
814
1010
1012
1014
O2 number density (cm–3)
1016
109
1010
1011
O3 number density (cm–3)
1013
100
80
Altitude (km)
Between April 1996 and March 2000, MSX/UVISI
observed roughly 200 stellar occultations. Owing to the
scattered nature of the occultations in both time and
location, general scientific studies have been limited to
a few cases. Nevertheless, these cases have allowed for
a quantitative demonstration of the capabilities of the
combined extinctive and refractive stellar occultation
technique.
108
60
40
20
0
107
108
1012
Figure 9. O2 and O3 profiles retrieved from a series of 15 MSX/
UVISI stellar occultations. The dots represent the individual profiles, whereas the solid lines represent the weighted average profiles and their associated 1- uncertainties. The change in slope of
the O2 profile near 120 km corresponds to the altitudes at which
O2 is no longer well mixed with the other atmospheric gases and
begins to follow its own scale height. The O3 profile clearly shows
the main O3 peak in the stratosphere near 25–30 km altitude as
well as the secondary peak in the upper mesosphere/lower thermosphere near 90 km altitude. The ability to measure these two
profiles from the lower atmosphere to the highest altitudes possible is significant owing to the chemical connections between
the two species and their importance in understanding physical
processes in Earth’s atmosphere.
Similar statements may be made regarding the O3 profiles in Fig. 9. Previous measurements from stellar occultation experiments6 only covered the altitudes down
to roughly 60 km. The combination of the extended
wavelength range and imaging capabilities of the UVISI
SPIMs with the tracking capabilities of the MSX spacecraft allow for the removal of the airglow signal and measurement of the extinction in the Chappuis band of O3
at visible wavelengths even when the star experienced
significant refraction in the lower atmosphere. The true
key to the success of these O3 measurements, however,
is the combination of both extinctive and refractive
measurements because the bulk properties inferred from
JOHNS HOPKINS APL TECHNICAL DIGEST, VOLUME 32, NUMBER 5 (2014)
MSX/UVISI STELLAR OCCULTATION EXPERIMENTS
Intercomparison with Other Measurements
the refraction data allow accurate treatment of Rayleigh
scattering, the refractive attenuation of the starlight,
and the change in the actual light path caused by refractive bending. Without the refraction measurements,
using stellar occultations to measure lower atmospheric
O3 would rely on a climatological model, thereby reducing the accuracy of the results.
To investigate the quality of the retrieved MSX/
UVISI O3 profiles and overall performance of the combined occultation method, a search for coincidences
with ground-based facilities and measurements by spacebased solar occultation instruments such as SAGE
8
40
35
6
Altitude (km)
MSX O3 density (1012 cm–3)
Correlation = 0.898
4
15–20 km
20–25 km
25–30 km
30–35 km
35–40 km
2
0
0
6
2
4
Ground-based O3 density (1012 cm–3)
30
25
20
15
–100
–50
0
50
100
Relative [(MSX-ground-based)/MSX] difference (%)
8
8
40
35
6
Altitude (km)
MSX O3 density (1012 cm–3)
Correlation = 0.948
4
15–20 km
20–25 km
25–30 km
30–35 km
35–40 km
2
0
0
6
2
4
SAGE II O3 density (1012 cm–3)
30
25
20
8
15
–100
–50
0
50
Relative [(MSX-SAGE II)/MSX] difference (%)
100
Figure 10. Comparison of MSX/UVISI-derived O3 values to ground-based sonde (top panels) and SAGE II data (bottom panels) for a
number of occultations. The left panels show point-by-point comparisons of the absolute densities measured at the same altitudes (all
ground-based and SAGE II data have been interpolated onto the 1-km altitude grid of the MSX/UVISI data), with the diagonal lines representing perfect agreement between MSX and the other data set. To illustrate the altitudes to which each point corresponds, the data
have been color-coded according to the legends. For example, the dark green circles represent the points at altitudes between 15 and
20 km (i.e., 15, 16, 17, 18, 19, and 20). Note that there has not been any binning of the data over these ranges; the ranges are merely an
indicator of the altitude region from which the data originate so that systematic biases at certain altitudes can be identified more easily.
The right panels show the differences between the datasets as a function of altitude. At each altitude, the dots represent the point-bypoint relative differences between MSX and the other data set. The solid red line is the weighted mean difference profile with the error
bars on the solid red line representing the uncertainty in the weighted mean. The dashed red lines are the average 3- uncertainties
inferred from the measurement uncertainties as propagated through the relative difference process of subtraction and division. In calculating these average uncertainties, both the uncertainties in the MSX/UVISI and the (properly interpolated) SAGE II data were used;
however, for the ground-based comparisons, zero uncertainty in the comparison data was assumed because most of the ground-based
data were provided without such information and we wished to treat all the ground-based data equally. The dashed blue lines are the
3- uncertainties calculated from the distribution of the relative difference points at each altitude without any weighting by the uncertainties in the relative differences. Thus, under the condition that the uncertainty estimates truly reflect the respective measurement
uncertainties, the differences between the red and blue dashed lines are potentially a reflection of the geophysical variance among the
various locations and times at which the measurements were made. (Reproduced from Ref. 33.)
JOHNS HOPKINS APL TECHNICAL DIGEST, VOLUME 32, NUMBER 5 (2014)
815­­­­
VERVACK ET AL.
(Stratospheric Aerosol and Gas Experiment) II was conducted. Search windows of ±5° latitude, ±20° longitude,
and ±24 h UTC were used. Although the longitude
window may seem large, most coincidences occur from
mid-latitudes to polar latitudes such that the separation
in absolute distance is smaller than that from an equatorial standpoint. The majority of the coincidences are
well within the allowed window regardless, and similar
windows have been used in validating SAGE II data.32
These search parameters yielded 17 ground-based and 20
SAGE II coincidences.
When comparing measurements from different
techniques, the context of the measurement scenarios
must be factored in. Time, altitude, and instrument
resolutions vary and affect the comparisons. Horizontal resolution differs greatly between ground-based and
space-based measurements, with the former being fairly
localized in extent whereas the latter can have a horizontal resolution of several hundred kilometers across
the limb. The choice of absorption and scattering cross
sections is another potential source of differences, due
to both absolute calibration issues in the laboratory
measurements and the temperature dependence of these
data (highlighting the need for an accurate temperature
profile as provided by the combined technique). Systematic comparisons of various measurement techniques
are thus a common means of assessing the differences
and similarities.
As can be seen in Fig. 10, comparisons between
MSX/UVISI and “ground-based” (primarily sondes,
top panels) O3 density measurements are excellent.
Although there is some scatter in the 45° plot (top
left panel), there is no clearly evident systematic bias.
Further evidence for good agreement is provided by
the weighted mean difference profile in the top right
panel of Fig. 10, which is more or less centered on the
zero line. Comparisons to space-based observations by
SAGE II (bottom panels), one of the standards in O3
measurements, are also strikingly good. Near the O3
peak between 20 and 36 km (retrievals are generally
most accurate near the density peaks), the differences
are better than 5%, although a slight systematic bias is
evident in that the MSX/UVISI values are generally
larger than the SAGE II values (i.e., positive residuals in the weighted mean). At higher altitudes, there
is a systematic bias toward negative differences, but
the agreement is still better than 10%. A larger systematic difference is present at low altitudes, with the
bias toward positive differences. The source of these
high- and low-altitude biases may be related, in part,
to the use of the Sun as the source for the SAGE II
measurements. Because the Sun subtends a finite range
of altitudes in the atmosphere, it is difficult to disentangle the absorption at a specific altitude. In contrast,
the stars used by MSX/UVISI are point sources and
provide better altitude resolution. The rapid increase
in the uncertainty of the SAGE II differences at low
altitudes is due to highly uncertain retrievals from the
SAGE II data, which suffer at the lowest altitudes once
the line of sight in their observations falls below the
O3 peak density. SAGE II uncertainties at the lowest
altitudes can approach the level of the measurements
themselves, whereas the uncertainty in the combined
technique results is significantly smaller (see Fig. 8).
Issues with knowledge of the refractive state of the
atmosphere and the total density and temperature,
which are not measured simultaneously by SAGE II
(their results generally incorporate climatological
model profiles), likely play into the differences at the
lowest altitudes as well.
A similar comparison was made between the
retrieved MSX/UVISI temperature profiles and profiles
240
30
25
220
Altitude (km)
MSX temperature (K)
Correlation = 0.950
200
10–15 km
15–20 km
20–25 km
25–30 km
180
180
220
200
Ground-based temperature (K)
20
15
240
10
–15
–10
–5
0
5
10
(MSX-ground-based) temperature (K)
15
Figure 11. Comparison of MSX/UVISI-derived temperature values to ground-based sonde data. The panels are similar to those in Fig. 10
with the exception that the right panel shows absolute rather than relative differences. The solid red, dashed red, and dashed blue lines
represent the equivalent quantities as in the relative O3 difference plots of Fig. 10. (Reproduced from Ref. 33.)
816
JOHNS HOPKINS APL TECHNICAL DIGEST, VOLUME 32, NUMBER 5 (2014)
MSX/UVISI STELLAR OCCULTATION EXPERIMENTS
0
31
30
Mar 1
0
Feb 1
27
20
21
22
23
24
MSX/UVISI values is indicated at most altitudes above
20 km, which is possibly related to two causes: refraction angle uncertainties and sonde capabilities. The
refraction angle uncertainties approach the size of the
refraction angles themselves at these altitudes. Vervack et al.31 noted this as a reasonable indicator of the
limits of the visible-light refraction angle approach, but
this is well captured in the statistical uncertainty of
the MSX/UVISI temperature retrievals and should not
introduce a significant systematic bias. Also at these
altitudes, the measurement capabilities of the sondes
begin to reach their limits, and the measurements are
correspondingly less accurate, which may introduce a
systematic trend if they tend to underestimate the temperatures. Over the altitude range where the observations from both data sets are best known, however, the
agreement is quite good.
The results in Figs. 10 and 11 show that the combined
occultation technique compares well with established,
routine methods and instruments used to measure O3
and temperature and that it represents a highly valuable, complementary means of monitoring atmospheric
O3. Further details on the intercomparison study may be
found in the paper by Vervack et al.33
17
10
11
12
13
14
15
16
8
9
0
600
1
2
3
4
5
6
7
from ground-based measurements (Fig. 11). As with
the O3 comparisons, the MSX/UVISI data compare
well with the ground-based temperature data, particularly given the in situ nature of the ground-based data
compared to the wider horizontal scale of the occultations. The larger scatter around 220 K in the 45° plot
(left panel) is primarily caused by three temperature
profiles obtained at the Lauder, New Zealand, monitoring station. Because the MSX/UVISI data agree well
with other Lauder profiles, this subset of poor agreement is somewhat puzzling, but they are included to
avoid systematic biasing of the intercomparison. It
may be noteworthy that stars for two of these MSX/
UVISI occultations were particularly dim, of a poor
spectral type for UV/visible occultations, or both. The
increased scatter at higher altitudes is likely due in part
to large uncertainties in the sonde data near the maximum altitudes of the balloons.
The absolute temperature difference plot in the right
panel of Fig. 11 shows excellent agreement between
the MSX/UVISI and ground-based data. The majority of the absolute differences are less than 5 K at all
altitudes, and the differences shown by the weighted
mean profile are better than 2 K. A slight bias to larger
0
1.00
0.75
0
0.50
0.25
0
500
–0.25
–0.50
–0.75
–1.00
–1.25
0
450
Ozone change (ppmv)
0.00
0
Potential temperature (K)
550
–1.50
–1.75
400
–2.00
30
40
50
Time (day of year)
60
Figure 12. Changes in the O3 mixing ratio in the lower stratosphere within the Arctic polar vortex
as measured by MSX/UVISI during the SOLVE campaign. Altitude is represented in terms of potential temperature. The times of the MSX/UVISI stellar occultations are denoted by diamonds along
the upper and lower abscissas and identified by profile number. Vertical dotted lines indicate
February 1 and March 1. The dashed lines denote the ensemble average diabatic descent of air
parcels within the vortex passing through the tangent points of the MSX/UVISI observations.
(Reproduced from Ref. 36.)
JOHNS HOPKINS APL TECHNICAL DIGEST, VOLUME 32, NUMBER 5 (2014)
817­­­­
VERVACK ET AL.
The SOLVE Campaign
818
(a)
4
3
2
575 K (on March 4)
Slope: 0.0012 (+/–0.0028) ppmv/day
(b)
4
Ozone (ppmv)
We used MSX to demonstrate the scientific utility
of this measurement technique during the SAGE III
Ozone Loss and Validation Experiment (SOLVE) campaign. During wintertime over the poles, a dynamically
isolated, continental-scale vortex forms in the stratosphere. Within this vortex, more popularly known as
the “ozone hole,” air cools radiatively and sinks. Halogen species become chemically activated on the surfaces
of polar stratospheric cloud particulates and catalytically
destroy O3 with the return of sunlight in the late winter/
early spring. Between November 1999 and March 2000,
the SOLVE campaign was conducted simultaneously
with the Third European Stratospheric Experiment on
Ozone 2000 (THESEO 2000), based in Kiruna, Sweden
(68°N, 20°E). The campaign was focused on coordinated observations from aircraft-, balloon-, ground-, and
space-based instruments with the goal of assessing quantitatively the rate of high-latitude stratospheric O3 loss
in order to gain a better understanding of the processes
controlling the loss of stratospheric O3 at mid- to high
northern latitudes.
There were 31 MSX/UVISI stellar occultation observations conducted during the SOLVE campaign, from
January 23 through March 4, 2000. Twenty-five of these
occultations occurred within the Arctic vortex, and the
results of the O3 retrievals in the lower stratosphere for
these in-vortex profiles are shown in Fig. 12 as a function of potential temperature (roughly corresponding to
altitude). Rather than absolute mixing ratios, this figure
shows the change in O3 mixing ratio as a function of
time (measurement times are indicated at the top and
bottom of the figure). The extent of subsidence within
the vortex is indicated by the ensemble average descent
rates for air parcels (averaged over 25-K intervals and
indicated by the dashed lines); an overall trend in O3
loss is observed, particularly below 500 K.
Within the lower stratosphere, O3 mixing ratios
increase with altitude; any subsidence will thus tend
to increase the observed O3 mixing ratios with time.
Such increases may partly mask the observable effects
of actual chemical O3 destruction. Therefore, to make
quantitative inferences about chemical O3 loss, the
dynamical effects within the vortex must be accounted
for in O3 retrievals such as those in Fig. 12. The Goddard
Space Flight Center trajectory model34 was used to conduct diabatic trajectory calculations to follow individual
air parcels from each occultation profile forward to
March 4, the date of the final occultation. Each retrieved
O3 profile was interpolated onto diabatic surfaces intersecting a 5-K grid using pairs of adjacent trajectories.
The grid was established using the potential temperatures on March 4 from the trajectory models. The results
on three such surfaces, intersecting the 575-, 500-, and
425-K potential temperature surfaces on March 4, are
shown in Fig. 13. Each plot represents cross sections par-
3
2
500 K (on March 4)
Slope: –0.0210 (+/–0.0020) ppmv/day
4
(c)
3
2
425 K (on March 4)
Slope: –0.0315 (+/–0.0014) ppmv/day
1
30
40
50
Time (day of year)
60
Figure 13. Photochemical O3 loss within the polar vortex along
three particular diabatic surfaces as observed by MSX/UVISI
during the SOLVE campaign. The error bars denote the 1-
uncertainties of the occultation measurements and retrieval and
do not include the uncertainty in the trajectory calculations. The
average daily loss rates for the period (i.e., the absolute values of
the slopes resulting from weighted linear regressions through
the observations on each surface) are also shown, along with the
uncertainties in the fits. (Reproduced from Ref. 36.)
allel (on average) to the diabatic descent curves of Fig. 12
(dashed lines). The slopes of the fits to the measurement
points in Fig. 13 correspond to the average daily O3 loss
rates at each level during the SOLVE campaign period.
Although little net change is observed in the O3 mixing
ratio at 575 K, increasing loss with decreasing altitude
is observed in the trend to lower potential temperatures, with a maximum loss near 425 K. A maximum
O3 loss of about 1 ppmv at 400–500 K (~16–21 km) was
determined and corresponds to an average loss rate of
~0.024 ppmv/d, in agreement with other SOLVE determinations of O3 loss.35 Further details on the SOLVE
study may be found in the paper by Swartz et al.36
JOHNS HOPKINS APL TECHNICAL DIGEST, VOLUME 32, NUMBER 5 (2014)
MSX/UVISI STELLAR OCCULTATION EXPERIMENTS
FUTURE CHALLENGES
Although MSX/UVISI was not optimized for stellar
occultations, the MSX experiments successfully reduced
to practice the techniques outlined in this article. MSX
provided a valuable opportunity to test the combined
extinctive and refractive stellar occultation technique
through many observations and validated the technique. Even in this nonoptimized state, the MSX/UVISI
data have yielded atmospheric profiles that compare well
with established ground-based and space-based measurements.33 In particular, the excellent agreement between
the ground-based data and the MSX/UVISI results,
both in O3 and temperature, provides strong evidence
that an instrument specifically designed for implementation of the combined occultation technique can provide
O3 profiles down into the troposphere, with accuracy
at least comparable to that of current techniques but
offering the prospect of global measurements covering a
greater range of locations and times. Ground-based data
are limited to a few locations on Earth, and instruments
such as SAGE II, which uses the Sun as an occultation
source, are limited to the terminator region and cannot
probe the atmosphere globally on short timescales; a
stellar occultation method does not suffer from these
limitations. The combined technique is thus a viable
method for probing Earth’s lower atmosphere for both
composition and bulk properties and has the potential
to become a powerful new observational technique for
global monitoring of lower atmospheric O3 and other
trace gas species.
Daily, global monitoring of the atmosphere via stellar
occultations has been implemented on spacecraft platforms that flew after MSX [e.g., by the GOMOS (Global
Ozone Monitoring by Occultation of Stars) instruStar simulator
ment on Envisat-137]. However, these instruments were
not designed to exploit the full potential of the stellar
occultation technique and in particular were not able
to probe the important lower atmosphere accurately.
The MSX/UVISI experiments have demonstrated that
the combined technique can provide the global measurements needed by the scientific and policy-making
communities, with accuracy equivalent to or with an
optimized instrument better than the current state of
the art. Thus, an optimally designed stellar occultation
instrument that packages all the necessary elements in a
single box is required to take this new technique to the
next level.
Our group at APL has addressed this need. Through
a NASA Instrument Incubator Program project, the
design of an optimized instrument, known as STARS
(STellar Absorption and Refraction Sensor) has been
conducted.38 The goals of this project were three-fold.
The first goal was to perform a trade-off analysis between
measurement capabilities and instrument requirements
to ensure that the final design met the desired measurement objectives as shown in Fig. 1. The second goal was
to design an instrument independently capable of performing the stringent tracking requirements necessary
for the combined stellar occultation technique, thereby
allowing its flight on any number of spacecraft platforms
offering only modest tracking capabilities. The third
and final goal was to test the critical tracking subsystem components in the laboratory (see Fig. 14) in order
to increase the technology readiness levels of these elements to flight proposal status. Unfortunately, despite
the success of the STARS project in achieving all its
goals, shifting priorities in Earth science led to a lack of
flight opportunities.
STARS test instrument
Elevation gimbal
(behind)
Primary
optics
Fast steering mirror
(simulates S/C jitter)
Beam splitter
Star point
source
Motion stage
support structure
Slit image
CCD detector
Imager
Fast steering mirror
(jitter compensation)
Linear motion stage
(simulates setting star)
Star simulator
control computer
Azimuth gimbal
Star intensity
attenuator
Electronics
Figure 14. Laboratory setup for testing of the tracking capabilities of the STARS instrument. (Modified from a figure published in Ref. 38.)
JOHNS HOPKINS APL TECHNICAL DIGEST, VOLUME 32, NUMBER 5 (2014)
819­­­­
VERVACK ET AL.
More recently, a renewed emphasis on Earth science
has emerged from NASA in the form of the Earth Venture line of missions. These small to medium, low-cost
missions are ideal for stellar occultation measurements;
however, the STARS instrument is too large for these
missions. The current challenge, therefore, is to redesign the STARS system for implementation on the next
generation of these smaller satellites. Enabling technologies in the form of uncooled IR detectors, advanced
microprocessors and focal plane elements, and miniaturized electronics need to be leveraged to reduce the
instrument size and enable this powerful remote sensing
technique that has the advantage of both low-cost and
low-resource requirements.
From the proof-of-concept demonstrations with
MSX/UVISI to the breadboard optimization of STARS,
APL is positioned on the forefront of this exciting new
method for remotely sensing Earth’s lower atmosphere
and beyond.
ACKNOWLEDGMENTS: We thank Frank Morgan, James
Carbary, Gerry Romick, Danny Morrison, Steve Lloyd,
Phil DeCola, Don Anderson, and Kris Kumar for their part
in fostering the combined occultation technique. We
acknowledge the MSX team at APL for their roles in the
operation, scheduling, data processing, and calibration of
the MSX spacecraft and instruments, as well as the efforts
of the STARS instrument development team. The groundbased data were obtained via FTP from the World Ozone
and Ultraviolet Data Centre provided by the Meteorological Service of Canada, Environment Canada. The SAGE II
data (Version 6.0) were obtained by CD from the NASA
Langley Research Center Atmospheric Sciences Data
Center. We thank all the investigators involved for making
these data available to the scientific community. Funding
for the data acquisition and operation of the satellite has
been through the DoD’s Ballistic Missile Defense Organization. The MSX/UVISI stellar occultation analysis was
supported by internal APL basic research funding and
NASA Grants NAG5-7552, NAG5-9988, and NAG-12270 to
APL. The development of the STARS instrument was supported by NASA Contract NAS5-97271 to APL. The ideas
summarized in this paper are captured in greater detail in
Refs. 29–31, 33, and 38.
REFERENCES
1Graedel,
T. E., and Crutzen, P. J., “The Changing Atmosphere,” in
Managing Planet Earth, Readings from Scientific American, Freeman,
New York (1990).
2Houghton, J. T., Ding Y., Griggs, D. J., Noguer, M., van der Linden,
P. J., et al. (eds.), Climate Change 2001: The Scientific Basis, Intergovernmental Panel on Climate Change, Cambridge University Press,
New York (2001).
3Hays, P. B., and Roble, R. G., “Stellar Spectra and Atmospheric Composition,” J. Atmos. Sci. 25(6), 1141–1153 (1968).
4Hays, P. B., Roble, R. G., and Shah, A. N., “Terrestrial Atmospheric
Composition from Stellar Occultations,” Science 176(4036), 793–794
(1972).
820
5Hays,
P. B., and Roble, R. G., “Stellar Occultation Measurements of
Molecular Oxygen in the Lower Thermosphere,” Planet. Space Sci.
21(3), 339–348 (1973).
6Hays, P. B., and Roble, R. G., “Observation of Mesospheric Ozone at
Low Latitudes,” Planet. Space Sci. 21(2), 273–279 (1973).
7Riegler, G. R., Atreya, S. K., Donahue, T. M., Liu, S. C., Wasser, B.,
and Drake, J. F., “UV Stellar Occultation Measurements of Nighttime
Equatorial Ozone,” Geophys. Res. Lett. 4(4), 145–148 (1977).
8Atreya, S. K., “Measurement of Minor Species (H , Cl, O , NO) in the
2
3
Earth’s Atmosphere by the Occultation Technique,” Adv. Space Res.
1(9), 127–141 (1981).
9Roscoe, H. K., Freshwater, R. A., Wolfenden, R., Jones, R. L., Fish,
D. J., et al., “Using Stars for Remote-Sensing of the Earth’s Stratosphere,” Applied Optics 33(30), 7126–7131 (1994).
10Renard, J.-B., Pirre, M., Robert, C., Moreau, G,. Huguenin, D., and
Russell, J. M. III, “Nocturnal Vertical Distribution of Stratospheric
O3, NO2 and NO3 from Balloon Measurements,” J. Geophys. Res.
101(D22), 28,793–28,804 (1996).
11Broadfoot, A. L., Sandel, B. R., Shemansky, D. E., Atreya, S. K.,
Donahue, T. M., et al., “Ultraviolet Spectrometer Experiment for the
Voyager Mission,” Space Sci. Rev. 21(2), 183–205 (1977).
12Smith, G. R., and Hunten, D. M., “Study of Planetary Atmospheres
by Absorptive Occultations,” Rev. Geophys. 28(2), 117–143 (1990).
13Pannekoek, A., “Über die Erscheinungen, Welche bei Einer Sternbedeckung Durch Einen Planeten Auftreten,” Astron. Nachr. 164(1),
5–10 (1904).
14Baum, W. A., and Code, A. D., “A Photometric Observation
of the Occultation of  Arietis by Jupiter,” Astron. J. 58(1208),
108–112 (1953).
15Wasserman, L. H., and Veverka, J., “On the Reduction of Occultation
Light Curves,” Icarus 20(3), 322–345 (1973).
16French, R. G., and Lovelace, R. V. E., “Strong Turbulence and
Atmospheric Waves in Stellar Occultations,” Icarus 56(1), 122–146
(1983).
17Elliot, J. L., Dunham, E. W., Bosh, A. S., Slivan, S. M., Young, L. A.,
et al., “Pluto’s Atmosphere,” Icarus 77(1), 148–170 (1989).
18Hubbard, W. B., Sicardy, B., Miles, R., Hollis, A. J., Forrest, R. W.,
et al., “The Occultation of 28 Sgr by Titan,” Astron. Astrophys.
269(1–2), 541–563 (1993).
19Hunten, D. M., and Veverka, J., “Stellar and Spacecraft Occultations
by Jupiter: A critical Review of Derived Temperature Profiles,” in Jupiter, T. Gehrels (ed.), University of Arizona Press, Tucson, pp. 247–283
(1976).
20Elliot, J. L., “Stellar Occultation Studies of the Solar System,” Ann.
Rev. Astron. Astrophys. 17, 445–475 (1979).
21Mill, J. D., O’Neil, R. R., Price, S., Romick, G. J., Uy, O. M., et al.,
“Midcourse Space Experiment: Introduction to the Spacecraft,
Instruments, and Scientific Objectives,” J. Spacecr. Roc. 31(5), 900–
907 (1994).
22Carbary, J. F., Darlington, E. H., Harris, T. J., McEvaddy, P. J., Mayr,
M. J., et al., “Ultraviolet and Visible Imaging and Spectrographic
Imaging Instrument,” Appl. Opt. 33(19), 4201–4213 (1994).
23Paxton, L. J., Meng, C.-I., Anderson, D. E., and Romick, G. J.,
“MSX—A Multiuse Space Experiment,” Johns Hopkins APL Tech. Dig.
17(1), 19–34 (1996).
24Jones, L. M., Fischbach, F. F., and Peterson, J. W., “Satellite Measurements of Atmospheric Structure by Refraction,” Planet. Space Sci.
9(6), 351–352 (1962).
25Grechko, G. M., Gurvich, A. S., Lyakhov, V. A., Savchenko, S.
A., and Sokolovskiy, S. V., “Results of an Investigation of Refraction during the Third Expedition on the Salyut-6 Orbiter” (English
translation), Izv. Russ. Acad. Sci. Atmos. Oceanic Phys. 17, 835–841
(1981).
26White, R. L., Tanner, W. E., and Polidan, R. S., “Star Line-of-Sight
Refraction Observations from the Orbiting Astronomical Observatory Copernicus and Deduction of Stratospheric Structure in the
Tropical Region,” J. Geophys. Res. 88(C13), 8535–8542 (1983).
27Ware, R., Exner, M., Feng, D., Gorbunov, M., Hardy, K., et al., “GPS
Sounding of the Atmosphere from Low Earth Orbit: Preliminary
Results,” Bull. Amer. Meteor. Soc. 77(1), 19–40 (1996).
28Kursinski, E. R., Hajj, G. A., Bertiger, W. I., Leroy, S. S., Meehan,
T. K., et al., “Initial Results of Radio Occultation Observations of
Earth’s Atmosphere Using the Global Positioning System,” Science
271(5252), 1107–1110 (1996).
JOHNS HOPKINS APL TECHNICAL DIGEST, VOLUME 32, NUMBER 5 (2014)
MSX/UVISI STELLAR OCCULTATION EXPERIMENTS
29Yee,
J.-H., Vervack, R. J. Jr., DeMajistre, R., Morgan, F., Carbary, J.
F., et al., “Atmospheric Remote Sensing Using a Combined Extinctive and Refractive Stellar Occultation Technique: 1. Overview and
Proof-of-Concept Observations,” J. Geophys. Res. 107(D14), ACH
15-1–ACH 15-13 (2002).
30DeMajistre, R., and Yee, J.-H., “Atmospheric Remote Sensing Using a
Combined Extinctive and Refractive Stellar Occultation Technique:
2. Inversion Method for Extinction Measurements,” J. Geophys. Res.
107(D15), ACH 6-1–ACH 6-18 (2002).
31Vervack, R. J. Jr., Yee, J.-H., Carbary, J. F., and Morgan, F., “Atmospheric
Remote Sensing Using a Combined Extinctive and Refractive Stellar
Occultation Technique: 3. Inversion Method for Refraction Measurements,” J. Geophys. Res. 107(D15), ACH 7-1–ACH 7-19 (2002).
32Veiga, R. E., Cunnold, D. M., Chu, W. P., and McCormick, M. P.,
“Stratospheric Aerosol and Gas Experiments I and II Comparisons
with Ozonesondes,” J. Geophys. Res. 100(D5), 9073–9090 (1995).
33Vervack, R. J. Jr., Yee, J.-H., DeMajistre, R., and Swartz, W. H., “Intercomparison of MSX/UVISI-Derived Ozone and Temperature Profiles
with Ground-Based, SAGE II, HALOE, and POAM III Data,” J. Geophys. Res. 108(D22), ACH 2-1–ACH 2-19 (2003).
34Schoeberl,
M. R., and Newman, P. A., “A Multiple-Level Trajectory
Analysis of Vortex Filaments,” J. Geophys. Res. 100(D12), 25,801–
25,815 (1995).
35Newman, P. A., Harris, N. R. P., Adriani, A., Amanatidis, G. T.,
Anderson, J. G., et al., “An Overview of the SOLVE-THESEO 2000
Campaign,” J. Geophys. Res. 107(D20), SOL 1-1–SOL 1-26 (2002).
36Swartz, W. H., Yee, J.-H., Vervack, R. J. Jr., Lloyd, S. A., and
Newman, P. A., “Photochemical Ozone Loss in the Arctic as Determined by MSX/UVISI Stellar Occultation Observations during the
1999/2000 Winter,” J. Geophys. Res. 107(D20), SOL 39-1–SOL 39-10
(2002).
37Bertaux, J. L., Mégie, G., Widemann, T., Chassefière, E., Pellinen,
R., et al., “Monitoring of Ozone Trend by Stellar Occultations: The
GOMOS Instrument,” Adv. Space Res. 11(3), 237–242 (1991).
38Yee, J.-H., Morrison, D., Murphy, G. A., Morgan, M. F. III,
Humm, D. C., et al., “STARS: the Stellar Absorption and Refraction Sensor,” in Optical Spectroscopic Techniques, Remote Sensing, and
Instrumentation for Atmospheric and Space Research IV, Proc. SPIE
Vol. 4485, A. M. Larar and M. G. Mlynczak (eds.), SPIE, Bellingham,
WA, pp. 51–59 (2002).
The Authors
Ronald J. Vervack Jr. was the lead on the refraction observation analysis for the MSX/UVISI stellar occultation experiments and also lead on the intercomparisons to measurements made by other spacecraft and ground-based facilities. He
is a member of the APL Senior Professional Staff in the Space Exploration Sector’s Geospace and Earth Science (SRG)
Group. Jeng-Hwa (Sam) Yee is Branch Scientist for Earth Science in the Space Exploration Sector and was responsible
for the overall conduct and analysis of the MSX/UVISI stellar occultations. William H. Swartz is Assistant Group
Supervisor of the Space Exploration Sector’s SRG Group and was the lead on the collaborative effort during the international SOLVE campaign. Robert DeMajistre is Assistant Group Supervisor of the Space Exploration Sector’s Science
Applications (SIS) Group and was the lead on the extinction observation analysis for the MSX/UVISI stellar occultation
experiments. Larry J. Paxton is the SRG Group Supervisor and was instrumental in identifying the scientific potential
of MSX/UVISI for Earth remote sensing. For further information on the work reported here, contact Ron Vervack. His
e-mail address is ron.vervack@jhuapl.edu.
The Johns Hopkins APL Technical Digest can be accessed electronically at www.jhuapl.edu/techdigest.
JOHNS HOPKINS APL TECHNICAL DIGEST, VOLUME 32, NUMBER 5 (2014)
821­­­­
APL Achievement Awards and Prizes
Linda L. Maier-Tyler
hallenges to our national security are increasing. Terrorism,
climate and demographic changes, resource protection and
cybersecurity, and the impact of extreme economic events challenge our current and future national security environment. To combat these challenges, accelerating scientific discovery and technological innovation is a high priority.
The Johns Hopkins University Applied Physics Laboratory (APL) has a proud history of
applying state-of-the-art science and technology to the national security requirements
of the time as well as to the challenges of tomorrow.
For the past seven decades, APL’s investment in its
science and technology enterprise has been central to
ensuring its position as a leader among premier engineering organizations. To promote initiative and reward
exceptional work, APL conducts annual awards programs to recognize individuals whose work from the
previous year advances science, technology, and education through tangible achievement in technical publications, independent research and development (IR&D)
projects, and out-of-the-box innovative thinking that
leads to the invention of new technologies. The Invention of the Year Award, the Government Purpose Innovation Award, the Ignition Grant Prize for Innovation,
the R. W. Hart Prizes, the Publication Awards, and
new this year, the Outstanding Mission Accomplishment Awards, all support APL’s resolve to foster and
bring forth new technologies and concepts to meet the
nation’s critical challenges.
The term invention specifically refers to making new
discoveries and designs or coming up with new ways of
822
doing things. Invention is the technical part of innovation, involving the development of an idea or discovery to
the point at which it works theoretically. The Invention
of the Year Awards program encourages new technology
and innovation at APL and identifies the top technology from the preceding year. Invention disclosures are
judged by an independent review panel of technical and
business consultants, technology transfer professionals,
and intellectual property attorneys. Judges base their
selections of the winning technologies on creativity,
novelty, improvement to existing technology, commercial potential, and probable benefit to society. In 2011, a
Government Purpose Innovation Award was established
to recognize an invention that specifically meets a critical need of a sponsor. For the 2013 competition, APL
researchers disclosed 230 inventions. Of these, one was
selected for the Invention of the Year Award and one
received the Government Purpose Innovation Award.
Master Inventor Awards are rare and honor staff who
can meet the criterion of at least 10 U.S. patents for their
JOHNS HOPKINS APL TECHNICAL DIGEST, VOLUME 32, NUMBER 5 (2014)
APL work. This year one recipient received the prestigious award for 10 U.S. patents issued during his employment at APL. He is only the 25th person in the history
of the Laboratory to qualify for this award.
Creativity is wasted if there is no process in place
to take ideas and turn them into something that has
market potential. The Ignition Grant Prize for Innovation was established to help APL staff explore innovative ideas outside of APL’s traditional programs. Open
to all staff, challenges are posted during several cycles
held throughout the year and ideas are submitted for
solutions. The winning ideas from each cycle are determined by popular vote, and the finalists receive funding
to develop their ideas. The Management Forum narrows
the field of awarded grants to a top few nominees, and
then APL staff vote for the top award on the basis of
each idea’s creativity and potential for impact. Of all
those who submitted ideas in 2013, one finalist received
the Ignition Grant Prize for Innovation.
The R. W. Hart Prize for Excellence in Independent
Research and Development recognizes significant contributions to the advancement of science and technology. Sectors and departments recommend candidates,
and the Management Forum judges the nominations on
the quality and importance of the work to APL. Prizes
are awarded in two categories: one for the best research
project and the other for the best development project. From those projects active in 2013, one prize was
awarded in the research category and two prizes were
awarded in the development category.
This year APL established the Outstanding Mission
Accomplishment Awards to recognize major achievement in mission-oriented programs and projects. Awards
are given in each of two categories, Outstanding Mission Accomplishment for a Current Challenge and
Outstanding Mission Accomplishment for an Emerging Challenge. For both types of awards, a review team
JOHNS HOPKINS APL TECHNICAL DIGEST, VOLUME 32, NUMBER 5 (2014)
of top managers and executives from APL’s sectors and
mission areas solicits nominations for technical accomplishments achieved during the previous year. Entries
are judged on technical excellence and potential impact.
Two winners were selected for the Outstanding Mission Accomplishment for a Current Challenge: one for
defense operations and one for intelligence operations.
One program won honors for contributions to an emerging challenge.
Professional publication in refereed journals is central to scientific communication and is the medium by
which important results and innovations are promptly
and accurately recorded and disseminated. To encourage
and reward exceptional scholarship, the Editorial Board
of the Johns Hopkins APL Technical Digest established
the Publication Awards competition both to encourage
professional writing and to recognize outstanding publications by the APL professional staff. Departments and
sectors may submit up to two nominations in each of six
categories. Judges base their selections on significance
and clarity, with considerably greater weight given to the
significance of the work in advancing science, engineering, or the mission of the Laboratory. In 2014, six technical departments and sectors submitted 36 publications
from those published in 2013. Of these, six publications,
including one book, won honors.
The hard work, excellence, and innovation represented in these technical awards are the qualities that
have shaped APL’s character since the early days of
World War II. Promoting and rewarding scientific discovery and technological innovation is a high priority
for the Laboratory. The work of these outstanding individuals not only represents APL’s best but also enhances
APL’s capacity to meet evolving challenges to our longterm national security. The recipients’ names, along
with the titles of their inventions, projects, and publications are displayed on the following pages.
823­­­­
L. L. MAIER-TYLER
INVENTION OF THE YEAR AWARD FOR 2013
For “System and Method to Rapidly Design Viral Vaccines to Prevent Vaccine Failure”
Existing vaccines are designed to protect against viruses
that are already infecting humans and animals. Viruses
quickly adapt to resist vaccines and immune systems,
and new vaccines currently take years to develop. This
award-winning technology speeds up this process and
can predict new viruses before they exist.
Andrew B. Feldman, Principal Professional Staff,
Research and Exploratory Development Department
(REDD), Ph.D., Harvard Univ., Physics; Jeffrey S. Lin,
APL Principal Professional Staff, REDD, M.S., Johns
Hopkins Univ., Computer Science
From left to right, Jeffrey Lin and Andrew Feldman.
GOVERNMENT PURPOSE INNOVATION AWARD FOR 2013
For “Aircraft and Sensor Product Geo-Registration in GPS-Denied Environments”
Two separate algorithms were developed to perform georegistration for aerial surveillance and reconnaissance in
a GPS-denied environment. The “scan-to-reference”
registers a single radar scan to a reference image, and the
“scan-to-scan” registers the current radar scan to a previous one. The implementation of these two algorithms
significantly improves the geo-location accuracy.
Mason M. Baron, Principal Professional Staff, Air and
Missile Defense Sector, M.S., Johns Hopkins Univ.,
Electrical and Computer Engineering; Gregory H. Barr,
Associate Professional Staff, Force Projection Sector,
B.S., Univ. of Maryland, College Park, Mechanical
Engineering; James G. Cochran, Senior Professional
Staff, Asymmetric Operations Sector, M.S., Georgia
Institute of Technology, Electrical Engineering
From left to right, Gregory Barr and Mason Baron. Not pictured: James Cochran.
824
JOHNS HOPKINS APL TECHNICAL DIGEST, VOLUME 32, NUMBER 5 (2014)
APL ACHIEVEMENT AWARDS AND PRIZES
MASTER INVENTOR AWARD FOR 2013
Rengaswamy “Srini” Srinivasan, a battery specialist and electrochemist, was awarded the Master Inventor Award in recognition for his 10 U.S. patents issued while he was employed at APL. He is only the 25th person in the history of the Laboratory
to qualify for this prestigious award.
Srini Srinivasan.
IGNITION GRANT PRIZE FOR INNOVATION FOR 2013
For “APL Maker Exploitation, Maker Movement or ‘MEME’ ”
The Maker Movement embraces do-it-yourself
techniques and a learning culture in which
makers teach one another skills such as 3-D
printing, programming hardware, rapid prototyping, and laser cutting. MEME brings this
movement to APL by promoting the culture
and providing resources to staff.
Robert Osiander, Principal Professional
Staff, Research and Exploratory Development Department (REDD), Ph.D., Technical
Univ. of Munich, Physics; Kimberly M. Griffin, Senior Professional Staff, Space Exploration Sector (SES), B.S., Virginia Polytechnic
Institute and State Univ., Finance; Robert A.
Berardino, Senior Professional Staff, SES,
B.S., Strayer Univ., Computer and Information System; Colin J. Taylor, Associate Professional Staff, REDD, B.S., Univ. of Maryland
Baltimore County, Computer Science
From left to right, Robert Osiander, Colin Taylor, Robert Berardino,
and Kimberly Griffin.
JOHNS HOPKINS APL TECHNICAL DIGEST, VOLUME 32, NUMBER 5 (2014)
825­­­­
L. L. MAIER-TYLER
R. W. HART PRIZES FOR 2013
Excellence in Research
For “High-Energy Laser (HEL) Effects on Space Systems and Materials”
This project started from what is known and modeled of a high-energy laser encounter with a spacecraft. Using experimental
data and physics modeling, the team examined the effects of a laser attack on the materials and functions of a spacecraft.
This new understanding and modeling capability will enable the design of spacecraft defenses and protections against lasers.
Kaushik A. Iyer, Senior Professional Staff, Space Exploration Sector (SES), Ph.D., Vanderbilt Univ., Materials Science
and Engineering; John J. Aiello, Senior Professional Staff, SES, M.S., Johns Hopkins Univ., Applied Physics; O. Manny
Uy, Principal Professional Staff, Research and Exploratory Development Department, B.S., Case Institute of Technology,
Chemical Engineering
Excellence in Development
For “Offensive Operations in an Anti-Access/Area Denial (A2AD) Environment”
This multifaceted project featured design, simulation, and field experiments for novel, affordable counters to A2AD threat
environments. Significant accomplishments were made in the offensive areas of large-volume fire swarming vehicle control,
small-platform confusers, and joint electronic attack/cyber operations, and to APL capabilities to model complex multi­
domain system-versus-system interactions.
David M. Van Wie, Principal Professional Staff, Force Projection Sector (FPS), Ph.D., Univ. of Maryland, College Park,
Aerospace Engineering; Jeffrey D. Barton, Principal Professional Staff, FPS, M.S., Johns Hopkins Univ., Mathematics/
Applied Mathematics; Cameron K. Peterson, Senior Professional Staff, FPS, Ph.D., Univ. of Maryland, College Park, Aerospace Engineering; Hans P. Widmer, Principal Professional Staff, FPS, M.S., Johns Hopkins Univ., Engineering and Applied
Physics of Biomedicine; Mark A. Oursler, Principal Professional Staff, FPS, M.S., Univ. of Virginia, Mechanical Engineering; Treven P. Wall, Senior Professional Staff, FPS, Ph.D., Cornell Univ., Mathematics; Brian L. Geesaman, Principal Professional Staff, FPS, M.S. Johns Hopkins Univ., Electrical Engineering; Coire J. Maranzano, Senior Professional Staff, FPS,
Ph.D., Univ. of Virginia, Systems Engineering; Daniel J. Silvera, Senior Professional Staff, Asymmetric Operations Sector,
M.S., Johns Hopkins Univ., Electrical and Computer Engineering; Edmund H. Nowicki, Principal Professional Staff, FPS,
M.S., Pennsylvania State Univ., Engineering Science
For “Agile Infrared Scene Projector on Carbon Nanotubes”
An array of vertically aligned carbon nanotubes was fabricated to create a new class of infrared scene projectors with highly
responsive frame rates and very high resolution. The resulting technology will enable Standard Missile-3 seekers to be tested
against highly advanced emulations of future ballistic missile threats. The technology also offers to advance high-resolution
image sensor capabilities.
Raul Fainchtein, Principal Professional Staff, Air and Missile Defense Sector (AMDS), Ph.D., Univ. of Texas, Physics;
David M. Brown, Senior Professional Staff, AMDS, Ph.D., Pennsylvania State Univ., Electrical Engineering; Karen M.
Siegrist, Senior Professional Staff, AMDS, Ph.D., Univ. of Maryland, College Park, Physics; Ryan P. DiNello-Fass, Senior
Professional Staff, AMDS, M.S., Columbia Univ., Electrical Engineering; Terry E. Phillips, Principal Professional Staff,
Research and Exploratory Development Department (REDD), Ph.D., Johns Hopkins Univ., Chemistry; Andrew H. Monica,
Senior Professional Staff, REDD, Ph.D., Georgetown Univ., Physics; David M. Deglau, Associate Professional Staff, REDD,
M.S., Johns Hopkins Univ., Chemical and Biomolecular Engineering
826
JOHNS HOPKINS APL TECHNICAL DIGEST, VOLUME 32, NUMBER 5 (2014)
APL ACHIEVEMENT AWARDS AND PRIZES
OUTSTANDING MISSION ACCOMPLISHMENT AWARDS FOR 2013
Current Defense Operational Challenge
For “Minotaur Mission Processor”
Minotaur is a transformational capability for the Navy, Coast Guard, and Customs and Border patrol. It enables interconnecting a diverse set of sensors and data sources across a wide operational area for automatic tracking, identification, and
interdiction support, especially for airborne surveillance of maritime surface traffic.
Mason M. Baron, Principal Professional Staff, Air and Missile Defense Sector (AMDS), M.S., Johns Hopkins Univ., Electrical and Computer Engineering; Weston R. Boyd, Associate Professional Staff, AMDS, M.S., Johns Hopkins Univ.,
Computer Science; Daniel J. Christine, Senior Professional Staff, AMDS, M.S., Johns Hopkins Univ., Computer Science;
James G. Cochran, Senior Professional Staff, Asymmetric Operations Sector (AOS), M.S., Georgia Institute of Technology, Electrical Engineering; Michael A. Delaney, Principal Professional Staff, AMDS, M.S., Naval War College, National
Security and Strategic Studies; Scott D. Heitkamp, Senior Professional Staff, Force Projection Sector, M.S., Johns Hopkins
Univ., Computer Science; Larry W. Nemsick, Principal Professional Staff, AOS, M.S., Univ. of Maryland, College Park,
Electrical Engineering; Conor R. Scott, Associate Professional Staff, AMDS, B.S., Virginia Polytechnic Institute and State
Univ., Computer Science/Mathematics; Mark A. Swana, Principal Professional Staff, AOS, M.S., Johns Hopkins Univ.,
Computer Science
From left to right, Conor Scott, Mason Baron, Mark Swana, Michael Delaney,
Larry Nemsick, Weston Boyd, Nicholas Francoski, and Scott Heitkamp. Not pictured:
Daniel Christine and James Cochran.
JOHNS HOPKINS APL TECHNICAL DIGEST, VOLUME 32, NUMBER 5 (2014)
827­­­­
L. L. MAIER-TYLER
Emerging Challenge
For “Naval Integrated Fire Control–Counter Air (NIFC-CA)”
NIFC-CA was conceived by APL in the 1970s and
developed over decades to greatly extend the engagement range against incoming air threats by networking both ships and aircraft. As the Technical Direction
Agent for three of the four elements of this family of
systems, AEGIS, the new Standard Missile-6, and the
Cooperative Engagement Capability, APL played a lead
role in demonstrating engagement of air threats beyond
the firing ship’s horizon, even over land. Testing this year
culminated in a major milestone toward fleet operation.
Jeffrey C. Mitchell, Senior Professional Staff, Air and
Missile Defense Sector (AMDS), M.S., Johns Hopkins
Univ., Engineering; William C. Hughes, Principal Professional Staff, AMDS, Dwayne A. Hawbaker, M.S.,
Virginia Polytechnic Institute and State Univ., Electrical
Engineering; Kenneth A. Plantz, Principal Professional
Staff, AMDS, M.S., Michigan State Univ., Mathematics/ Applied Mathematics; Matthew J. Kazanas, Principal Professional Staff, AMDS, M.S., Georgia Institute of
Technology, Electrical Engineering; Lorenzo R. Brooks,
Senior Professional Staff, AMDS, B.S., Tuskegee University, Electrical Engineering
From left to right, Kenneth Plantz, Dwayne Hawbaker, and
Jeffrey Mitchell. Not pictured: William Hughes, Matthew Kazanas, and Lorenzo Brooks.
PUBLICATION AWARDS FOR 2013
Author’s First Paper in a Journal or Proceedings
For “Empirical Reconstruction of Storm Time
Steady Magnetospheric Convection Events,” Journal of Geophysical Research Space Physics 118(10),
6434–6456 (2013).
The Tsyganenko magnetic field models have been legendary for 20 years in space physics. Now, APL has
brought the models to new levels of usability and predictive power. Detailed features of Earth’s otherwise invisible magnetic field are now accessible, and these features are revealing new modes of global magnetospheric
configuration.
Grant Stephens.
828
Grant K. Stephens, Associate Professional Staff, Space
Exploration Sector (SES), M.S., Johns Hopkins Univ.,
Applied Physics; Mikhail I. Sitnov, Senior Professional Staff, SES, Ph.D., Moscow State Univ., Physics; J. Kissinger (non-APL staff); N. A. Tsyganenko
(non-APL staff); R. L. McPherron (non-APL staff);
Haje Korth, Senior Professional Staff, SES, Ph.D.,
Technical Univ. of Braunschweig, Physics; Brian J.
Anderson, Principal Professional Staff, SES, Ph.D.,
Univ. of Minnesota, Physics
JOHNS HOPKINS APL TECHNICAL DIGEST, VOLUME 32, NUMBER 5 (2014)
APL ACHIEVEMENT AWARDS AND PRIZES
Outstanding Paper in the Johns Hopkins APL Technical Digest
The Walter G. Berl Award
For “Implementing Genome-Informed Personalized Medicine in the US Air Force Medical Service via the
Patient-Centered Precision Care Research Program,” Johns Hopkins APL Technical Digest 31(4), 333–344 (2013)
This paper describes an approach for incorporating genomic information into personalized health care for active duty Air
Force and DoD personnel. It describes the overall program, the clinical utility study infrastructure, and other components
critical to implementation such as provider education, policy issues, and integration with the electronic healthcare record.
Christopher E. Bradburne, Senior Professional Staff, Asymmetric Operations Sector (AOS), Ph.D., George Mason Univ.,
Bioscience; Lucy M. Carruth, Senior Professional Staff, AOS, Ph.D., Wake Forest Univ., Microbiology; John H. Benson,
Principal Professional Staff, Research and Exploratory Development Department (REDD), M.S., Univ. of Maryland, College
Park, Civil Engineering; Jeffrey S. Lin, Principal Professional Staff, REDD, M.S., Johns Hopkins Univ., Computer Science; Ashok Sivakumar, Senior Professional Staff, REDD, M.E., Massachusetts Inst. of Technology, Electrical Engineering;
Ruth A. Vogel, Senior Professional Staff, REDD, B.S., Univ. of North Dakota, Epidemiology
Outstanding Research Paper in an Externally Refereed Publication
For “Evidence for Water Ice Near Mercury’s North Pole from MESSENGER Neutron Spectrometer Measurements,” in Science 339(6117), 292–2906 (2013).
This paper describes the first direct measurement of ice deposits in Mercury’s permanently shadowed polar craters. These
ground-breaking results include the location, purity, stratigraphy, and amount of these deposits. These results fulfill important
MESSENGER mission goals and provide information on the Mercurial impact flux.
David J. Lawrence, Principal Professional Staff, Space Exploration Sector (SES), Ph.D., Washington Univ., Physics; William C. Feldman (non-APL staff); John O. Goldsten, Principal Professional Staff, SES, M.S., Johns Hopkins Univ., Electrical Engineering; Sylvestre Maurice (non-APL staff); Patrick N. Peplowski, Senior Professional Staff, SES, Ph.D., Florida
State, Experimental Nuclear Physics; Brian J. Anderson, Principal Professional Staff, SES, Ph.D., Univ. of Minnesota,
Physics; David Bazell, Senior Professional Staff, SES, Ph.D., Univ. of Maryland, College Park, Physics; Ralph L. McNutt Jr.,
Principal Professional Staff, SES, Ph.D., Massachusetts Inst. Technology, Physics; Larry R. Nittler (non-APL staff); Thomas
H. Prettyman (non-APL staff); Douglas J. Rodgers, Senior Professional Staff, SES, Ph.D., Univ. of Delaware, Physics; Sean
C. Solomon (non-APL staff); Shoshana Z. Weider (non-APL staff)
Outstanding Development Paper in an Externally Refereed Publication
For “Bayesian Context-Dependent Learning for Anomaly Classification in Hyperspectral Imagery,” IEEE Transactions on Geoscience and Remote Sensing 52(4), 1969–1981 (2013)
Automatic context-based target discrimination is a significant step forward in the remote sensing community. This paper proposes an algorithm for exploiting environmental context in hyperspectral imagery and demonstrates its efficacy for airborne
countermine operations. Results demonstrate a substantial improvement over the current standard by exploiting context.
Christopher R. Ratto, Senior Professional Staff, Force Projection Sector, Ph.D., Duke Univ., Electrical and Computer
Engineering; Kenneth D. Morton Jr. (non-APL staff); Leslie M. Collins (non-APL staff); Peter A. Torrione (non-APL staff)
JOHNS HOPKINS APL TECHNICAL DIGEST, VOLUME 32, NUMBER 5 (2014)
829­­­­
L. L. MAIER-TYLER
Outstanding Professional Book
For Wireless Networking: Understanding Internetworking Challenges, Wiley–IEEE Press. Hoboken, New Jersey (2013)
This book provides a concise yet comprehensive summary of the key current and emerging technologies that make up the
commercial wireless networking landscape. It bridges the often-disparate communities of wireless networking and cellular
technologies in a novel way.
Jack L. Burbank, Principal Professional Staff, Asymmetric Operations Sector (AOS), M.S., North Carolina State Univ., Electrical Engineering; Julia Andrusenko, Senior Professional Staff, AOS, M.S., Drexel Univ., Electrical Engineering; Jared S.
Everett, Associate Professional Staff, AOS, M.S., North Carolina State Univ., Electrical Engineering; William T. M. Kasch,
Senior Professional Staff, AOS, M.S., Johns Hopkins Univ., Electrical Engineering
Outstanding Special Publication
For “A Continuum Manipulator Made of Interlocking Fibers,” Proceedings of the 2013 IEEE International Conference on Robotics and Automation, Karlsruhe, Germany, May 6–10, 2013, IEEE, pp. 4008–4015 (2013).
This work clearly presents a novel, biologically inspired approach to manipulation with applications to computer-assisted
surgical systems. Manipulator modeling and associated equations are described and justified. Prototype hardware is detailed
and design parameters are provided. A straightforward method for controlling the manipulator is described.
Matthew S. Moses (non-APL staff); Michael D. M. Kutzer, Senior Professional Staff, Research and Exploratory Development
Department (REDD), Ph.D., Johns Hopkins Univ., Mechanical Engineering; Hans Ma (non-APL staff); Mehran Armand,
Principal Professional Staff, REDD, Ph.D., Univ. of Waterloo, Mechanical Engineering
830
JOHNS HOPKINS APL TECHNICAL DIGEST, VOLUME 32, NUMBER 5 (2014)
This issue of the Johns Hopkins APL Technical Digest is a testament to the diverse contributions the Laboratory makes to solving critical challenges. The issue covers technical topics ranging from ballistic missile
defense system optimization and unmanned underwater vehicle testing to nuclear radiation source detection, remote sensing, and the development of a model and measurement techniques for leaf reflectivity with
global implications for environmental assessment.