Interferometric Synthetic Aperture Radar
(IFSAR)
Applications to Crisis Support
By
Carlton Daniel
U.S. Army Topographic Engineering Center
Terrain Data Representation Branch
Robert Yoha
Integrated GeoData Management Unit
Office of Technology Assessment, Planning and Development
California Department of Conservation
Key Words: DEM, GPS, IFSAR, INS, STAR-3i
Abstract
The state of the art in exploiting interferometric synthetic aperture radar (IFSAR) for terrain
information is advancing rapidly, and provides significant potential for use in crisis support
operations. Unlike conventional SAR imagery, IFSAR data permits the generation of rectified
SAR images co-registered with an accurate Digital Elevation Model (DEM). In addition, this
imagery can have an absolute geographic accuracy of 3 meters RMS or less. The rapidity, with
which IFSAR data can be collected and processed over wide areas, and the all-weather, day-night
capability offers significant potential for providing direct support to crisis situations.
Introduction
Cognizant of expanding capabilities in radar interferometry, the U.S. Department of Defense
began an aggressive program to pursue the acquisition of highly accurate computerized terrain
data utilizing IFSAR in 1992 under the sponsorship of the Advanced Research Project Agency
(ARPA) with the Topographic Engineering Center (TEC), U.S. Army Corps of Engineers as the
executive agent. This program, titled Interferometric Synthetic Aperture Radar for Elevations
(IFSAR-E), has been executed by the Environmental Research Institute of Michigan (ERIM) and
resulted in the fabrication of an interferometric radar integrated with a Geographic Positioning
System (GPS) and Inertial Navigation System (INS) systems on a Learjet 36A. NASA’s Jet
Propulsion Laboratory (JPL) at the California Institute of Technology developed processing
software and the ground-processing environment. This software and ground processing
capability has now been transitioned to Intermap Technologies, the system is now referred to as
the STAR-3i.
These efforts represent a convergence of technology development by a large number of
investigators, only a few of whom are referenced above, and a pressing need for low cost, fine
resolution, highly accurate terrain elevations for a wide variety of military and civil applications.
The STAR-3i system for collecting DEMs at low cost is just now becoming available for
application to civil problems and it is anticipated that significant data collections will occur in the
next few years. Ongoing activities within the TEC are quantifying the performance of IFSAR
techniques for the Federal Emergency Management Agency (FEMA). This paper discusses a
FEMA-sponsored STAR-3i IFSAR data collection of the Sacramento and San Joaquin Valleys,
for suitability to flood plain applications and emergency management support.
Existing Topographic Maps are Inadequate for Crisis Management
During the winter of 1996-97, severe flooding occurred in the Sacramento and San Joaquin
Valleys in California. Many communities experienced extensive property damage due to
flooding. During the height of these floods, a situation developed which demonstrated the
shortcomings of existing paper topographic maps and pointed to a need for computer modeling of
a flooding situation.
The problem began with a levee break on the West Side of the Sutter Bypass, which developed
into an unforeseen situation. The Sutter Bypass is located between the Sacramento and Feather
Rivers, west of the towns of Yuba City and Marysville, which are located 64 kilometers north of
Sacramento, California. During periods of high water levels the Sutter Bypass is used to divert
floodwaters from the Sacramento River. About 45 kilometers downstream and shortly after its
junction with the Feather River, the Sutter Bypass is again joined with the Sacramento River. At
this junction, high floodwater from both rivers then flow into the Yolo Bypass, which diverts the
combined flow around the city of Sacramento, California’s capitol.
On January 4, 1997, a section of levee failed on the Sutter Bypass about 26 kilometers south east
from the town of Meridian which is located in an agricultural basin between the Bypass and the
Sacramento River; directly west of the Yuba City and Marysville area which was experiencing
flooding from the Feather River. Water flowing through the levee break begun inundating
farmland as it flowed southward. Plans were begun to locate the proper point to make a breach
cut in the levee, downstream from the break, to drain the water back into the Bypass. As the
inundation flowed southward it encountered a topographic rise that caused the flood water to also
flow towards Meridian, threatening to flood the town. Heavy equipment was brought in to
construct an eastern levee to protect the town from the rising water. After the flood level
stabilized, a breach cut was finally made to drain the floodwaters out of the basin and back into
the Bypass. The floodwaters came almost to the top of the hastily constructed levee around
Meridian, but the town was saved.
California Department of Conservation (CalDoC) staff were assisting Office of Emergency
Services, Department of Water Resources and other responding agencies, with providing 2-D
computerized support maps using a geographic information system (GIS) to respond to this
situation. The question was asked, “could a computer model have been used to predict the
inundation area as the water flowed through the levee break?” Also, “could a computer model
have determined how high to construct the Meridian levee?” The answer was, “yes but better
maps were needed with more accurate elevations.”
A DEM combined with hydrologic computer modeling could have been put to much use in this
situation. Using IFSAR to map the terrain and create a DEM, the topographic rise may have been
mapped. In addition, by using a computer model, the new flooding could have been projected into
the adjoining area to determine how high the levee needed to be to protect the town of Meridain.
The problem was posed how to acquire a DEM of the flood-prone areas in the Sacramento and
San Joaquin Valleys?
Based on experience modeling terrain with IFSAR generated DEMs and from having worked
with ERIM IFSAR-E data, CalDoC proposed a flight plan to conduct an IFSAR data collection of
the Sacramento and San Joaquin Valleys. The goal would be to develop a DEM of the major
river channels and their levees, along with the adjacent flood prone areas. This could be used for
advance planning and improved response in real time. The intent was that by using an IFSAR
DEM along with GIS, one would be able to model the extent of flooding into adjacent areas if a
levee were to be breached. An accurate DEM of flood prone areas would provide improved
planning and faster response.
Potential for Utilization of IFSAR for Flood Plain Mapping
A proposal was developed by CalDoC to use the STAR-3i airborne X-band radar terrain mapping
system operated by Intermap, to acquire the IFSAR DEM data. The proposal was funded by
FEMA, with TEC as the project manager and CalDoC providing user input. The planned project
area is shown in Figure 1.
FEMA Flood Project
Radar Terrain Data
Figure 1. Location Map of IFSAR DEM Data
Coverage of the flood data set extends from north of Sacramento south to Fresno, following the
Sacramento and San Joaquin Rivers. The project area covers about 17,000 square kilometers (or
6,564 square miles). Data was initially collected during September 1997 in two flights, one each
for a northern and southern area. Strong turbulence caused the southern mission to be aborted; it
was reflown in July 1998. Intermap began processing data in early 1998 with completion planned
for January 1999.
The STAR-3i System
Traditional SAR systems from the past only gave 2-dimensional views of the Earth and included
geometric distortions inherent in slant range SAR data. IFSAR was developed to provide an
elevation component to SAR imagery. The additional information from interferometric
techniques provides a 3-dimensional view of the Earth and removes some of the geometric
distortions.
Three files are generated from the IFSAR instrument: a magnitude file, correlation file, and an
elevation file. The magnitude file is a backscatter image that provides information on the shape
of features, as well as terrain texture. The correlation image provides information on surface or
volume backscatter. The elevation data, or DEM, provide information on terrain elevation and
height of features.
The STAR-3i system consists of two X-band radar antennas mounted in a model 36A Learjet.
Data collection from the twin antennas occurs simultaneously. The set of acquired data are
“interfered” by a digital correlation process to extract terrain height data that are used to
geometrically correct the radar image. STAR-3i uses post-processed Differential GPS (DGPS)
data, together with on-board laser-based inertial measurement data, to obtain highly accurate
positioning control. Terrain height and positioning data are enhanced by calibration of the
baseline (the distance between the two antennas). The accuracy of the positioning information
and calibration is such that no in-scene control points are required. The only requirement is that a
ground-based GPS receiver must be located within 200 kilometers of the data collection site so
that DGPS processing can take place.
The STAR-3i is typically flown at 12,000 meters and acquires a 10-kilometer wide swath of 2.5meters resolution on the ground. The system has been designed to collect DEMs at a rate of 100
square kilometers per minute with 3-meter accuracy. Improved DEM accuracy is achieved by
reducing the aircraft altitude to 6,000 meters, which reduces the swath width to 6 kilometers. At
this lower aircraft height, ground resolution stays the same; however, the signal-to-noise ratio is
one-half that of the higher altitude, thereby improving precision in the vertical direction.
Central California Valley IFSAR DEM
Figure 2. Highways, Aqueducts, Stream Channel
Interstate 5 is shown between the California Aqueduct to the left with the Delta-Mendota Canal to
the right, crossing the well-defined stream channel of Orestimba Creek. Individual trees are
apparent within the upper stream channel. It is possible to discern various stream channels, which
do not appear on the USGS topographic map. Note the grass covered hilly terrain and its contact
line with the alluvium of the agricultural valley floor. Numerous orchard canopies and crops are
apparent. Towards the bottom is an east-west levee and canal, which serve as a waste drain for
the Delta-Mendota Canal. This area is about 8 kilometers south of Crows Landing on the eastern
edge of the San Joaquin Valley.
Figure 3. Urban Area with Water Features
The city of Stockton is evenly illuminated by the radar through careful alignment of the aircraft
flight path. Note the well define riverbanks and water features amongst the farmland and urban
built-up areas. The major channel is the San Joaquin River, which meanders in at the bottom of
the image and flows northwesterly towards the Sacramento Delta. The river in this area has been
dredged into a deep-water channel to serve the Port of Stockton, note the turning basin and the
“Y” shape at its end. A straight canal cuts eastward ending in Yosemite Lake in the middle of a
residential area. The Calaveras River flows from the upper right corner southwesterly to the San
Joaquin River. Interstate Highway 5 cuts from north-to-south across these water features, the
bridges do not appear in this image. A grouping of warehouses are easily identified on Rough
and Ready Island formed by the Burns Cutoff and the San Joaquin River. A series of high voltage
electrical towers appear on the left. A discerning eye can observe three separate lines of towers,
where they split in the southwest.
Figure 4. Stream Features
The Stanislaus River flows from east to west in this image located midway between Modesto and
Oakdale, about 35 kilometers south east of the city of Stockton. The urban pattern of the rural
town of Riverbank can be carefully observed in the lower center below the river. An old railroad
alignment faintly appears from the upper left and turns south across the Stanislaus River through
the center of town. The main channel of the Stanislaus River and its meander belt are well
defined. The river’s flood plain can be easily mapped. Stream cut cliffs on the northern banks are
from 15 to 20 meters in height. Note two or perhaps three distinct levels of bottomland incised by
meander scrolls up river from the town. It is possible to map out riparian vegetation versus
agricultural areas within the river valley.
Conclusions
TEC is now in the process of evaluating the data. As with most remotely sensed data sets, there
are anomalies present within the DEM under certain operating conditions. While there appears to
be much promise in using IFSAR DEMs for terrain modeling and in flood applications; the
limitations need to be better understood in order to exploit the data to its full potential.
The following caveats are provided so that future users and researchers of the IFSAR data
products can better understand some of the operating characteristics that is associated with this
technology.
Radar DEMs do not portray a bald earth. Vegetation and large cultural features are included in
the DEM. In some cases this may be desirable, such as for locating trees or riparian vegetation
along a stream. In other cases, such as dense forest or where buildings are mapped, the ground
surface will be obscured and the DEM height will be that of the features. Various approaches are
being developed and investigated to generate a true ground surface or bald earth DEM.
Flight path orientation is an important factor, in which two problems with the return signal can
occur: 1) no return signal or “dropouts” from terrain shadowing; and, 2) an overly strong reflected
signal, or “blossoming” from manmade structures. Radar has a tendency for shadowing, just as
with optical photography. The flight path should be aligned to minimize terrain shadowing, which
is effected by a combination of sensor altitude, slant range angle and terrain height. In addition,
the flight path should be at a slight angle to the urban pattern, to minimize what can be described
as a blossoming of the return signal from structures facing the radar antenna. If buildings or
structures are perpendicular to the flight path, this will produce a reflected signal that is too strong
for the system to process. Areas with buildings can be “whited out”. This was first noticed in an
image of Burbank, California, and has since been termed, “the Burbank effect.” Both these
problems can be minimized or eliminated with forethought and careful flight line planning.
Flood modeling requires a common vertical datum; it is suggested that orthometric heights be
used. The user’s GIS model and the radar DEM must both have their elevations referenced to the
same datum if they are to be combined. Otherwise there is a danger of a water level in one terrain
model not matching the same elevation in another model.
A set of requirements needs to be developed to define features, dimensions, and the precision and
accuracy to be obtained for flood modeling applications. It is hoped that by providing this data to
users they will be exposed to the utility of radar DEMs and provide feedback to improve future
data collection and applications. A next step will be collecting information from users toward this
effort.
Reference
1. Hansen, John V.E., Johnson, Peter B., Interferometric Synthetic Aperture Radar Applications
to Crisis Support, International Symposium on Spectral Sensing Research ’95 Reports.