Calibration and Validation of Ocean Color Products for Coastal Waters

advertisement
Calibration and Validation of Ocean Color Products for Coastal Waters
March 31-April 1, 2009
CH2M Hill Alumni Center
Oregon State University, Corvallis, Oregon
This workshop addressed the issues of calibration and validation of ocean color products
for coastal waters. It built on the results of an earlier workshop on the cross calibration of
OCM and MERIS data with MODIS and SeaWiFS to extend the useful data sets for
coastal monitoring. This workshop expanded the discussion to include protocols for the
use of above-water and in-situ instrumentation and approaches for validation of products
for the coastal ocean and related issues such as validating atmospheric correction for
coastal locations.
Workshop Objectives:
1. Address the issues of calibration and validation of ocean color products for
coastal waters.
2. Build on the results of an earlier workshop on the cross calibration of OCM and
MERIS data with MODIS and SeaWiFS to extend the useful data sets for coastal
monitoring.
3. Expand the discussion to include protocols for the use of above-water and in-situ
instrumentation and approaches for validation of products for the coastal ocean
and related issues such as validating atmospheric correction for coastal locations.
4. The product of the workshop will be a set of recommendations for algorithms and
approaches for validation of SeaWiFS, MODIS and MERIS and in the future the
Hyperspectral Imager for the Coastal Ocean (HICO) and VIIRS coastal products,
including atmospheric corrections relevant to coastal areas, recommendations for
coastal validation site selection and use, and requirements for new
instrumentation for coastal validation.
5. The long term goal is to develop a set of automated procedures for coastal product
validation for all ocean color sensors.
1
Workshop Agenda and Notes with references to Presentations:
March 31, 2009 Morning
•
•
•
Day 1: coffee and continental breakfast 8:00 AM, meeting begins at 8:30AM
Introductions and meeting overview (Davis, Arnone, Stumpf)
Summary report of the first meeting (Davis, below)
A Workshop on the use of MERIS and OCM by NASA, NOAA, Navy and university
scientists was held at NRLSSC January 29-31, 2008. Current US ocean color sensors are
past their design life and we need to assure continuity of data and products for ocean
applications. Additionally MERIS and OCM are the only sensors designed for ocean
color measurements at higher resolution that meet coastal requirements. The workshop
focused on identifying issues that needed to be addressed to make use of MERIS and
OCM data by US scientists for coastal waters with a goal of transitioning to operations.
The workshop specifically addressed approaches for cross-validating MERIS and OCM-1
with coincident SeaWiFS and MODIS data and in-situ measurements.
A primary result was a plan to assemble a comparison data set for 5 representative US
coastal regions including MERIS, SeaWiFS and MODIS satellite data and in situ data for
the sites for clear days in 2005 and 2006. A plan was developed to process the data to a
standard level for the comparisons at the remote sensing reflectance level and to compare
a standard chlorophyll product and a standard product that estimates particulate scattering
(Rrs 490 nm/ Rrs 560 nm). The data would be analyzed at full resolution for
comparisons to in-situ data, and it would be binned to a 4 km average for comparisons of
large scale averages and temporal changes.
The workshop participants were enthusiastic about the use of MERIS and OCM data for a
host of science and operational applications. The 620 nm and 709 nm channels on
MERIS have proven especially useful to address issues of bathymetry of optically
shallow waters and of harmful algal blooms in the Laurentian Great Lakes and coastal
waters. Also the higher resolution of the MERIS 300 m data and the OCM data should
be particularly valuable in many coastal regions. For MERIS we need to demonstrate the
utility of 300 m for operational coastal applications and the group plans to prepare a Cat 1
proposal to ESA for that work. We recognize that there is no mechanism in place for
operational use of MERIS data at this time but the goal is to demonstrate the utility of the
data and develop products that can be used by US and other scientists and resource
managers.
The group also discussed the fact that OCM-1 may not be operating much longer, but that
OCM-2 is due to be launched in 2009 and that it should provide better coverage and
higher quality data. The group concluded that it was important to continue to work with
OCM-1 to better understand the issues with this type of data and prepare for OCM-2
when it becomes available.
2
Therefore the group had two specific recommendations. First, US agencies (NASA,
NOAA and Navy) should work with ESA for improved access to 1.2 km and 300 m
MERIS data. For MERIS we should request routine near-real time access to 1.2 km data
for the Great Lakes, US coastal water and other areas of interest. The ideal would be to
upgrade this to 300 m data for near coastal regions. Second, the group also expressed
great interest in receiving and working with OCM-1 data and OCM-2 data when it
becomes available. The group considered that work with OCM-1 was essential as a
pathfinder for OCM-2. We look to NOAA and NASA to make an agreement with the
Indian Space Agency (ISRO) for access to that data. In return the US team can share
results and help with issues of calibration and data product validation that can help the
Indian scientists make the best use of the data.
•
•
MERIS, SeaWiFS and MODIS data comparisons: ~3 hours
– Description of the combined MERIS, MODIS and SeaWiFS data set that
was produced in response to the need identified in the first workshop
(Bailey_combined_dataset.pdf, parts 1 and 2)
Presentations on comparison results and on current MERIS processing issues:
– Navy: (Lyon_MERIS_products.pdf)
– NOAA: (Foley_MERIS.pdf, Wang_MERIS.pdf)
– NASA: (Bailey_combined_dataset.pdf, section 3)
Topics discussed by each group included:
– MERIS data used: source, processing methods and products
– Problems and issues
– Comparisons with SeaWiFS and MODIS: techniques and results
– Upgrade to MERIS L1 and L2 processing and MERIS data availability…
General Discussion of MERIS data and the approach for comparing/merging data
from different ocean color sensors. (Stumpf moderator, Davis notes) ~ 1 Hour
MERIS data discussion:
Processing to level 2 is a key issue:
For NOAA and Navy applications MERIS needs to look like SeaWiFS and MODIS they
need consistency because they blend products/image of the day (NAVY, NOAA use high
res MERIS data in specific areas, different than the GlobColour program which is
looking at global averages)
NASA, NAVY and NOAA propose to use L2GEN for processing of MERIS data: They
have demonstrated some success so far, but there are many implications.
Need to sort out L2GEN versions for MERIS processing (merge GSFC and NRLSSC
versions); they are working together to accomplish this.
L2GEN has choices of atmospheric models, etc. for MERIS processing, ultimately there
should be a default processing scheme for MERIS similar to the default set up for
MODIS.
NASA is looking for agreement from ESA before releasing a version of L2GEN that
includes MERIS in SeaDAS.
We need to keep ESA informed of all of our work with MERIS data.
3
In general the community would like to get full resolution MERIS data for US coastal
waters.
How do we get it? We need to improve connection with Canadian stations receiving the
data. Ultimately the 300 m data should come directly from Canada.
A second issue is the archiving the data. (Only 2 week’s worth are maintained on the
ESA website). To address this OSU will look at setting up an archive including data that
has been collected to date. We must talk to ESA but we will start to do this for MERIS
CAT 1 users. (Letelier, Davis work with Paul Lyon, Dave Foley, Sean Bailey, Rick
Stumpf)
Immediately, send Dave Foley a list of west coast sites for collecting 300 m MERIS data.
NRLSSC is collecting other sites; including GOM, MVCO, Chesapeake Bay.
OSU to look at adding all of these to OSU archive?
For coastal cal/val we need to relate the imagery to in situ data. Need to identify selected
data sets for cal/val sites for MERIS, MODIS, SeaWiFS and other sensors.
There was a discussion about the ESA processing for atmospheric correction. Several
users noted differences in BEAM products and operational products. Using L2GEN and
its associated atmospheric correction seems to produce a more consistent product, and
one that is closer to the MODIS and SeaWiFS products for the same areas.
Action items:
 Continue discussion of MERIS data using ocean color forum for MERIS (Bailey),
or if that is not possible at an OSU website (Letelier)
 Dave Foley to collect west coast sites; users should send Dave your site
information.
 Sherwin Ladner will collect a list of east and gulf coast sites.
 OSU will set up an archive for the MERIS 300 m data as it becomes available for
U.S. coastal waters. For now the site will only be open to MERIS CAT 1 team
members.
 Release of L2GEN for MERIS? NASA and Navy will make available to their Cat
1 user community for evaluation and wait for ESA approval before wider release.
 Coast Watch to look into implementing APS including for MERIS processing?
Afternoon March 31, 2009
•
•
Lunch
Coastal validation data sets (how they fit into coastal validation strategy): ~ 3.0
hours
– NASA: SeaBASS/NOMADS other cal/val data
(Bailey_combined_dataset.pdf, sections 5 and 6)
– NOAA: MOBY and Chesapeake Bay (Ondrusek_Cal-Val.pdf)
– OSU HOTS, and data from the coasts off Oregon and Chile
(Letelier_OSU_in_situ_optics.pdf)
– USC selecting a SeaPRISM site for the west coast,
(Jones_SeaPRISM_site_selection.pdf)
4
–
•
•
•
CUNY SeaPRISM site and other measurements
(Ahmed_SeaPRISM_optics.pdf)
Topics discussed by each group:
– Instruments, calibration, protocols
– Water quality and coastal atmospheric conditions at their site
– Data storage and distribution
– In situ vs. satellite data matchups and statistics
– Automation of match ups
Future Cruise programs: Letelier, others?
Discussion of data requirements for validation of coastal products (Davis): ~1
hour
Discussion of data requirements for coastal validation:
Minimum data for validation of normalized water leaving radiances (nlw) or Remote
Sensing Reflectances (Rrs):
A SeaPRISM which operates continuously and unattended and produces both the
aerosols and Rrs is preferred because the satellite makes this measurement (However we
prefer a hyperspectral measurement so that we can match the data to all the satellite
sensors.)
Other above water systems include the MicroSAS (used routinely from a ferry by Barney
Balch), and the HyperSAS which could provide the full spectra data. These systems do
not measure the aerosols and are not designed for long-term continuous operation.
Another system that provides full spectral data from below the water is the HyperPRO
which can be used to collect vertical profile information, needed to propagate Rrs through
the sea surface. A set of HyperPROs at OSU, NRLSSC and other sites should be cross
calibrated for use in the coastal cal val activities.
Other well calibrated in-water radiometers should be considered if the calibration and
operation procedures are appropriate for cal val activities.
A major issue is the spatial temporal variations in coastal waters making it difficult to
relate in situ point measurements to satellite pixels on the order of 0.1 to 1 sq km.
Rick Stumpf suggested we need to study this and how to evaluate it.
It is essential to do this right to get one point that matches with the satellite
In addition, we need to address the aerosol variability for the same time and area.
Sean Bailey admonished us to look at what you have from the satellite; only a 15% return
of good data because of loss due to clouds, glint, etc.
Time variations are also critical: we need to track tidal motions, etc to track the features.
Time series data from moorings and gliders is helpful. A practical approach may be to
use coastal models to track the coastal dynamics.
A related question was how do we distribute the samples in the ocean to best sample the
variability in time and space?
Minimum data set to validate bio-optical algorithms:
The most direct conversion from the satellite measured Rrs is to spectral Inherent Optical
Properties (IOPs). So in situ measurements of spectral a, c and bb at several bands is
essential.
5
Chlorophyll is a widely used product so Chlorophyll (by HPLC) used directly or to
calibrate fluorescence chlorophyll measurements is essential? Need to collect HPLC
samples each time that you take HyperPRO or other in situ optical data. We should also
tap other sources of quality chlorophyll which is widely measured.
Weather information including SST, air temperature, wind speed, air pressure should be
routinely measured as these are essential for atmospheric correction.
Other issues considered:
We should have one or more sites in optically shallow waters.
We also discussed planned future in situ data collections:
NRLSSC; Monterey Bay, CA, and the Alboran Sea
OSU to continue collections on the Oregon coast, Chile-Peru coast and HOTS
NOAA to continue collections on Chesapeake Bay; they can do rapid response sampling
in Chesapeake as needed.
Other investigators (need to make contacts, also look at SeaBASS data sets)
•
Adjourn 5 PM
***********************************************************************
Morning April 1, 2009
•
•
Day 2: coffee and continental breakfast 8:00 AM, meeting begins at 8:30AM
Brief presentations on the status of:
– GOCI (Davis_GOCI.pdf)
– COIS (Davis) the Coastal Ocean Imaging Spectrometer (COIS) was
originally developed for the NEMO program. NRL (Corson) continues to
present it for the DoD Space Experiment Board who prioritizes missions
for the Space Test Program (STP). COIS remains high (5 of 62) on the
STP list and NRL is still looking for a flight opportunity for COIS. One
option is to revive the plan to fly it on NPOESS C1 platform. NRL is
having informal discussions with the IPO on this possibility.
– OCM 2 (DiGiacomo_foreign_satellites.pdf presented by Davis)
OCM 2 on Oceansat-2 satellite is presently scheduled for launch at the end
of April but (perhaps summer?); see
http://www.ioccg.org/news/March2009/news.html
Discussions are currently underway between ISRO, NASA and NOAA as
to potential collaborations. However, it is not clear at this time that there
will be a provision for timely, free & open data access for operational
purposes (especially for high-res data); discussions are continuing in
advance of the upcoming launch. As such NOAA NESDIS/STAR plans
still TBD.
– Sentinel 2 (DiGiacomo_foreign_satellites.pdf presented by Davis)
Apparently both Sentinel 3A and 3B have now been approved; the launch
of Sentinel 3A is anticipated to take place no earlier than Q4 2012.
6
•
•
•
•
•
•
NESDIS-STAR/OSDPD via CoastWatch (Central Ops & Nodes) plans to
build upon the current MERIS operational data stream and associated
developmental efforts (in partnership with R. Stumpf, CIOSS, et al.) and
will work with ESA and partners to similarly develop and/or deliver OLCI
ocean color radiometry data/products.
Presentation on HICO (Corson_HICO.pdf)
Presentation on VIIRS coastal product validation plans (Arnone_VIIRS_CalVal.pdf)
Updates to the QAA algorithms (Lee_QAA_v5.pdf)
Discussions on satellite cross-validation and product validation: (~3.5 hours)
Identify Cross satellite calibration issues (Stumpf discussion leader, Davis notes)
– Strategies (based on first day inter-satellite comparisons)
– Atmospheric correction differences
– Bio-optical differences
Identify satellite validation issues in coastal waters (Arnone) :
– Coastal validation Sites (Martha’s Vineyard, coastal observatories, etc)
– Data flow. (Arnone, et al.)
– Spatial and temporal issues (Letelier, Davis)
– Vicarious calibration methods (Lyon, Bailey, Wang)
– Coastal Atmospheric Correction issues (Wang_Atm_Correction.pdf, Gao)
Afternoon April 1, 2009
•
•
•
•
•
•
•
•
Lunch
Presentation by Burt Jones on low cost optical monitoring system
(Jones_LMSP_bonaire.pdf)
Continue discussions
Review findings and organize into a report.
Consider Presentation to the NASA OCRT meeting May 2009
Consider special session at February 2010 Ocean Sciences meeting in Portland,
OR.
Future plans and directions.
Adjourn 5 PM
7
Discussions and future directions:
Overall goal is to develop methodologies for the cal/val of coastal data products in order
to produce useful validated products from the current and future ocean color sensors.
What are the top 5 things we can do that will have the biggest impact:
1. Greatly increase the number of optical measurements, optimizing for quality and
numbers, in the coastal ocean that can be used for coastal cal/val. First priority is
Rrs, then add aerosols, then IOPs then chlorophyll. (IOPs and chlorophyll should
be measured during routine servicing. Look for opportunities to incorporate these
measurements into drifters, gliders, IOOS, OOI, and other opportunities.)
2. Conduct rigorous error analysis of the in situ data, including measurement and
sampling errors.
3. Establish standard calibration, processing and quality control for all of the data.
4. Establish the scales of temporal and spatial variability and ways to deal with it to
estimate the variability in the in situ data when compared to the satellite data.
5. Establish metrics for skill assessment to evaluate products and algorithms.
How do we test, accept and implement new algorithms:
 Consider using Navy algorithm acceptance procedures as a model?
 Need to demonstrate that the algorithm makes a significant improvement in data
amount or quality or both.
 Need to define better metrics so that we can make a quantitative estimate of the
improvement when we evaluate a new algorithm or product.
 Algorithms may need to be specific for types of water of interest (coastal, high
CDOM, high sediments, etc.), atmospheric conditions, region or season.
 Need to develop protocols for the evaluation process including using synthetic
data sets and the NOMAD data set. Other data sets may be needed for specific
environments (e.g. turbid waters).
Recommendations for MERIS data processing strategies:
 Apply L2gen to process MERIS data including a default process (like SeaWiFS
and MODIS).
 Receive and archive 300 m MERIS data for the North American swaths. Then
subset the data for areas of interest as requested by individual investigators.
 Use MERIS with L2gen to demonstrate the cross satellite comparison approaches
and procedures.
Coastal Validation sites:
 SeaPRISM should receive top priority as coastal validation sites.
 Other Rrs measurements (preferably with aerosol measurements).
 Weather station (wind speed, RH, air temp, SST, other?) information should be
included whenever possible.
 Ancillary data to include IOPs, chl, other as available.
8
 Calibration standards and procedure protocols are needed for all measurements.
E.g. use the same calibration, software and protocols for all of the HyperPROs,
cross calibrate them at Satlantic, etc.
 Consider following the guidance from the NASA working group on standard
processing (Stan Hooker).
 Need procedures and protocols for Satellite to in situ data match ups.
 Need standard approach to establish the scales of temporal and spatial variability
and ways to deal with it.
 Need to establish the error in the in situ data for comparison to the satellite data.
 Need to standardize the data handling and distribution.
Recommendations for algorithms for the unique properties of coastal waters:
 Define high scattering coastal waters in terms of significant normalized water
leaving radiance in 865 nm (>.1)?
 Address the need for site specific Atmospheric correction algorithms for high
scattering waters.
 High absorption coastal waters (CDOM dominates optical properties) may have
issues with negative values in the blue. Use a reference site in high CDOM waters
to check the calibration at 412 nm.
 High chlorophyll waters (chl a >2 mg/m3) may also need unique algorithms and
approaches.
 For all of these waters we may also need unique product (IOP, chl, etc)
algorithms.
 Review aerosol models and evaluate whether we have sufficient models for the
full range of coastal conditions including absorbing aerosols.
 Constrain models using other satellite data e.g. OMI.
 It is likely to require using an ROV with a fiber optic probe or other techniques to
make measurements without shadow in these waters.
Future Plans and Directions:
 Review findings and organize into report on CIOSS website.
 Report to ESA on use of MERIS data, share results and insights (Curt, Bob,






Rick, Ricardo, Sean).
Special Session on Coastal cal/val at Ocean Sciences Meeting February, 2010
in Portland, OR (Curt to organize).
Presentation on meeting findings to the NASA OCRT May 2009 (Rick).
Rick to talk to IOOS office at NOAA explore getting Rrs and other optics as
a routine part of IOOS. Curt to explore the same for OOI.
OSU (Letelier) to look at archiving MERIS 300 m North American swath
data.
Support continuation of the NASA Ocean Biology group algorithm forum for
discussions about MERIS data and coastal cal/val.
Reprocess MERIS data with new L2GEN code and host it on GSFC site as
we did last year (Sean and Paul).
9
Download