--Stuart McMuldroch stuart@jord.jpl.nasa.gov,
MISR Science coordinator, Jet Propulsion Laboratory
The MISR Science Team met on June 4-6, 1997 at the Jet Propulsion Laboratory in Pasadena, CA. The MISR meeting was an opportunity for all members of the team to obtain the latest information regarding the recently completed MISR instrument. Given the proximity to the scheduled launch of the AM-1 platform, emphasis this year was placed on final algorithm refinements, status of science production code, science verification and validation, and initial concepts for early mission science. The following text briefly describes the focus of the meeting and highlights the most important aspects of the proceedings.
The Principal Investigator, Dave Diner, opened the session, bidding welcome to the team and summarizing resolved issues from the previous science team meeting. He then went on to give a brief overview of the meeting's objectives: 1) to update the team on instrument hardware and flight software status, 2) to update the team on ground data processing software status, 3) to develop test data requirements to verify the scientific veracity of the MISR production code, 4) to review MISR's validation strategy, and 5) to hone the MISR science emphasis especially during the early mission phase.
Terrence Reilly gave the instrument and project status report. The instrument, having successfully passed thermal-vacuum testing, was delivered for spacecraft integration at the Lockheed-Martin facility at Valley Forge, PA, on May 26. Several problems that had surfaced during previous thermal-vacuum testing had been solved. Reilly emphasized that the project focus has shifted to science data processing activities and the completion of AirMISRthe airborne version of MISR.
Dave Diner continued describing the instrument performance by detailing the results of several instrument science tests. Preliminary analysis of test images from the Collimator Array Tool (CAT) suggests that the alignment of the cameras on the optical bench is within specifications. Alignment measurements made pre- and post-shake testing are comparable, suggesting that the MISR optical system can tolerate vibration. Additional tests, performed by imaging a target suspended above the instrument, successfully yielded the first true two-dimensional image. Progress on the in-flight radiometric calibration and characterization was reported by the instrument scientist, Carol Bruegge. As reported previously, all thermal, dynamical, radiometric, and spectral testing of the cameras has been successfully completed. Bruegge showed results demonstrating how unacceptable out-of-band errors can be corrected to within reasonable limits.
The science data system status was presented by Graham Bothwell. Activity in the last year has been intense, with staffing reaching maximum levels as we approach launch. Version 1 of the production code is in the process of being delivered to the Langley DAAC. Version 2 of the code, which contains the complete system suitable for launch, is expected to be delivered near the beginning of 1998. Every AM-1 instrument team has drafted backup plans to ensure minimum processing ability immediately after launch. The MISR back-up plan includes the processing of 1 or 2 swaths of data plus several local-mode sites per week at the MISR SCF. Additional functionality, including distribution and archiving, will be achieved by working with the Langley DAAC. The MISR home page on the Web at http://www-misr.jpl.nasa.gov continues to expand, providing increased levels of information to the community.
The MISR test team is building datasets to check both the operability and scientific veracity of the production code. Operability tests range from unit tests which verify the functionality of individual executable components to full system-wide tests. Robert Ando described efforts to build a simulated MISR multi-angle dataset covering a portion of an orbit swath from Minnesota to Mexico. This dataset will soon contain a simulated atmosphere and clouds generated from a Monte Carlo scheme. Bob Vargo led the effort to form science verification teams comprising scientists and software engineers to enhance the effectiveness of the testing efforts. Each team will test a particular MISR product to guarantee product structure and scientific veracity. In addition to simulated data, the test teams plan to use observations from AirMISR, as these will be the closest data to MISR imagery in the pre-launch timeframe.
AirMISR collects multi-angle MISR-like data utilizing a spare MISR camera mounted on a gimbal system aboard an ER-2 aircraft at an altitude of 20 km. AirMISR research objectives include supporting development and validation of MISR algorithms, retrieval algorithms, and products, providing an additional radiometric calibration path to assist in-flight calibration, and enabling scientific research utilizing multi-angle imaging data. Currently, construction of AirMISR has finished, and the instrument is undergoing flight-readiness tests.
Jim Conel gave an overview of the validation strategy and lessons learned from field campaigns during the past year. Pre-launch validation efforts concentrate on algorithm validation and technique development while post-launch efforts focus on MISR product validation and vicarious calibration. The validation team has now perfected their techniques for collecting and analyzing data with subsequent intercomparisons with other teams. Conel continued by discussing planned field campaign agendas and the potential of AirMISR. Peter Muller presented studies comparing varying spatial resolution Bidirectional Reflectance Distribution Function (BRDF) model retrieval data with multi-angle Advanced Solid-State Array Spectro- radiometer (ASAS) data. Results suggest a good agreement confirming MISR's retrieval approach.
Roger Marchand described the MISR cloud validation plan. Pre-launch activities include comparisons of cloud products derived from AirMISR data with those derived from ground instrumentation and other airborne detectors such as MODIS Airborne Simulator (MAS). Post-launch validation efforts will use AirMISR-derived values, ground station climatologies, and comparison with MODIS cloud products.
This section of the meeting focussed on refining the MISR algorithms in preparation for launch and strengthening the link between MISR's standard products and early mission science.
MISR's cloud detection ability relies on both stereoscopic and radiometric techniques. Roger Davies started the session by reviewing MISR's hierarchy of stereo matchers. Simple tests suggest that the combined set of matchers is reliable and robust. Tests using the simulated dataset described above, combined with accurate truth fields, will provide better reliability estimates. Radiometric techniques, developed by Larry Di Girolamo, utilize a dynamic thresholding method to detect clouds. Eugene Clothiaux described his research on cloud detection using texture techniques as an example of possible future algorithms.
Local albedos are calculated from MISR data using a combination of solid angle weighting, deterministic, and stochastic methods. Local albedos are then used to determine coarse resolution restrictive and expansive values. Tests on simulated scenes confirm the validity of these techniques.
MISR aerosol retrieval techniques are dependent on the background surface, i.e., whether the aerosol retrieval is performed over dark water, dense dark vegetation, or heterogeneous land. Ralph Kahn described the on-going work investigating the MISR instrument's sensitivity to aerosols over dark water. This study examines sensitivity to particle shape, optical depth, characteristic radius, and indices of refraction for pure particle types. Future work will examine sensitivity to mixtures of particles and changes in natural conditions. Results from the sensitivity study of aerosol retrievals over land were presented by John Martonchik. Retrievals over dense dark vegetation are well characterized with accuracies which depend on aerosol amount and solar zenith angle. The team agreed to adopt the empirical orthogonal function approach suggested by Martonchik as a superior method for aerosol retrieval over heterogeneous land.
Dave Crisp provided an alternative explanation to the cloud absorption anomaly where GCMs underestimate the atmospheric absorption of sunlight. He suggested that the anomaly arises from incorrect assumptions regarding global annual-average gas and aerosol distributions and illumination conditions, combined with underestimates of water, ozone, and tropospheric aerosol absorption. Calculations performed since the Science Team Meeting show that MISR should be able to determine whether aerosols are distributed just above the clouds as Crisp suggests. Tom Ackerman continued the session by describing differences between models and clear sky diffuse radiation data from the Atmospheric Radiation Measurement (ARM) site in Oklahoma. These differences can be resolved by the inclusion of an additional unidentified absorber at blue wavelengths. Spectral flux observations with 1-2% accuracy are needed to resolve this problem.
Michel Verstraete turned the focus of the meeting to multi-angular measurements of land surfaces. He quickly reviewed the panoply of tools and techniques available for the quantitative interpretation of remote sensing data. The multi-angular data available from MISR will permit the evaluation of existing parametric BRF model assumptions about the anisotropy of the radiance fields and allow for the development of new more-accurate models. Possible future algorithms were presented by Yuri Knyazikhin and Bernard Pinty. Knyazikhin described an empirically based Fraction of Photosynthetically Active Radiation/Leaf Area Index (FPAR/LAI) algorithm. Pinty presented a semi-discrete model for the characterization of land surfaces, which includes soil albedo, leaf characteristics, and canopy structure. This session on land surface studies was concluded by Anne Nolin, who described possible uses of MISR's multi-angle data for accurately determining albedo and grain size over snow-covered areas.