Daniel Wenkert (dwenkert@haleakala.jpl.nasa.gov), jet Propulsion Laboratory
For three days in June (Tuesday June 6 -Thursday June 8), the MISR Science Team, along with many colleagues and project personnel, gathered at JPL to discuss the status of the MISR instrument, scientific questions that need to be answered to develop product generation software, and the results of recent research. After a short welcome and introduction from PI, Dave Diner, Graham Bothwell and Terry Reilly presented recent accomplishments and the coming schedule for software and hardware deliveries, respectively.
Terry Reilly, the MISR Project Manager, described the status of the Engineering Model (EM) and Protoflight Model (PFM) instruments. At the time of the meeting, the MISR PFM was to be delivered to Lockheed Martin in 21 months, and the MISR project had 11 weeks of slack.
Carol Bruegge began the discussions of EM testing and calibration. There are a number of reasons to believe that the goal of absolutely calibrating the MISR instrument radiometrically to 3% is being achieved. Performance specifications for signal-to-noise ratio and locally uniform response are being met. In general, the MISR cameras that have been built and tested are performing very well, with the exception of a larger-than-expected out-of-band spectral response and low-level "halos" around the point-spread functions. Bob Korechoff described the detective work in identifying the causes of these problems. Both arise because the MISR filters are mounted close to the CCDs. Following presentations on the status of the MISR on-board calibrator and the MISR error budget, from Valerie Duval and Nadine Chrien, the Science Team concurred with JPL's proposal to implement corrections for the spectral and point-spread-function response in ground data processing.
A discussion of MISR instruments for later missions, e.g., EOS AM2, was led by Dave Diner. NASA Headquarters insists that any follow-on EOS AM mission must be lightweight and use advanced technology. A number of options were discussed. There was a consensus to maintain the capability to image at nine angles, the capability to cover the Earth in nine days, and to keep the existing calibration requirements. There was no interest in continuous spectroscopy, but the Science Team did want the MISR Project to investigate extending the spectral range to 1.6 micrometers, provided this would not sacrifice the goal of a more-compact instrument.
On Tuesday, Graham Bothwell, the MISR Science Data System Manager, described the plan to ramp up MISR product generation at the DAAC. Before it is possible to retrieve many Level 2 parameters, it is necessary to co-register and ortho-rectify MISR data from all four spectral bands from all nine viewing angles. This requires the use of a global Digital Elevation Model (DEM). Richard Fretz of the Cartographic Group at JPL described their work on processing the Digital Terrain Elevation Data (DTED)-1 data set, to make it self-consistent and put it into a form in which it can be used by MISR. The MISR team plans to use the resulting data in on-orbit geometric calibration at the MISR Science Computing Facility (SCF). A lower-resolution version of these data will be sent to the DAAC for ortho-rectification of MISR data, as the DTED data set itself is not distributable.
Sue Barry discussed the planning of MISR Local Mode (completely high-resolution) observations, then Meemong Lee discussed footprint sharpening of MISR along-track pixels. After some discussion of whether "footprint sharpening" was useful in the presence of clouds, it was concluded that if we are computing-resource limited, then deconvolution of "halos" in the MISR point-spread function is more important than the footprint sharpening; however the Team is still interested in pursuing this option.
The purposes of geo-rectification and registration of MISR Level 1B data, and the techniques that will be used, were described by Veljko Jovanovic. Two sorts of Level 1B2 imagery will be generated: data that have been projected onto an ellipsoid, and data that have been projected onto the surface of the Earth, including the effects of topography. It was proposed and agreed upon that the WGS ellipsoid at sea level would be used for the ellipsoid-projected data, rather than an ellipsoid at 30-km altitude as previously planned. Error analysis for Level 1B2 geo-rectification was later described by Mike Smyth. MISR requirements are being met with the software currently prototyped.
After some discussion of other Level 1B2-related issues, led by Earl Hansen, Scott Lewicki gave a status report on 1B2 and described the work left to be accomplished for the beta delivery. He noted that the most difficult part of the algorithm, image matching for the nine high-resolution red-band images that MISR will continuously acquire, is being prototyped. This has allowed the team to make reasonable estimates of DAAC resources needed for Level 1 processing.
On Wednesday morning, Roger Davies described some recent work by his group on top-of-the-atmosphere (TOA) radiation from variable-thickness and/or broken-cloud fields. Analysis of Earth Radiation Budget Experiment (ERBE) data from such fields does not agree with plane-parallel calculations for TOA radiation. He presented the results of Monte Carlo forward calculations, given realistic variable-thickness clouds. A number of very interesting effects were seen in the Monte Carlo calculations that mimic some effects seen in real data. These effects should go a long way toward explaining the "anomalous shortwave absorption" in clouds that has been discussed in the literature over the past year.
Tom Ackerman presented some work by his group on TOA radiation from clouds. He described the origin of the "anomalous shortwave absorption" issue in analysis of ERBE data and aircraft data. He then described some of his own aircraft data and some Monte Carlo simulations. His feeling is that TOA flux is a fundamentally ambiguous concept; that what one really wants to measure is radiance as a function of illumination and viewing angles. The latter is what MISR will uniquely measure. Both Ackerman and Davies agreed that MISR has a unique role to play in working out issues of TOA radiation in real Earth environments, at moderate and high spatial resolution.
Davies followed these discussions by presenting the current status of developing the algorithms for Level 2 TOA/Cloud (Level 2TC) product generation. He mentioned the major progress made in the spring in building fast and accurate stereo algorithms for retrieving cloud-top height fields. His major concern was the availability of simulated MISR data (from multiple view angles). Peter Muller described the specific stereo algorithms that had been developed and the plan for merging the best parts of these algorithms in the MISR L2TC software.
Later on Wednesday morning, Tamas Varnai and Siegfried Gerstl described progress in developing Azimuthal Models (AZMs). These will be used to retrieve TOA albedos by integrating the observed bi-directional reflectances. Varnai (a graduate student working for Roger Davies) has simulated the upwelling radiation from a variety of broken and continuous cloud fields, while Gerstl and his group have simulated the radiation field above a variety of Earth surfaces under a variety of cloud-free atmospheric conditions.
Most of Wednesday afternoon was devoted to discussions of cloud-screening in MISR data. Larry Di Girolamo (a graduate student working with Davies) and Eugene Clothiaux (a post-doc working with Ackerman) presented the results of their work. Both Di Girolamo and Clothiaux are developing their techniques on 1-km resolution AVHRR data.
Di Girolamo is developing techniques for detecting clouds at any altitude, which have at least a moderate optical thickness, by comparing reflectance-based signatures against thresholds appropriate for the class of surface being observed. This technique will be applied to single pixels in MISR images separately at all nine viewing angles. Another method he has developed is the "Band Difference Angular Signature," which is sensitive to the presence of clouds with low optical depth, especially at high altitudes and high latitudes.
Clothiaux is developing a technique for detecting clouds at any altitude which have at least a moderate optical thickness, over snow, ice, and bright land. This technique relies on measuring the textural properties of neighboring groups of pixels in MISR images (separately for each viewing angle). The exact texture parameters that are calculated are determined by a neural net analysis of a large number of images of sea ice, clouds, snow fields, etc. An alternative technique for cloud-screening over such bright surfaces using MISR data compares the stereoscopically retrieved cloud-top height with the known surface height.
A presentation by John McGuffie on the status of Level 2TC software wrapped up the discussion of TOA radiation and clouds.
Daniel Wenkert presented the current plan for using non-MISR data in processing MISR data at the DAAC. The plan is to use output data from the Goddard Data Assimilation Office (DAO) for as much of the needed atmospheric data as possible. Information on snow and ice cover will come from analysis of passive microwave data (currently Special Sensor Microwave Imager [SSM/I]) at the National Snow and Ice Data Center [NSIDC] (or the MSFC DAAC). As MODIS data products are validated in the post-launch period, Level 2 MODIS data are expected to enter the MISR processing stream.
Larry Thomason, from the SAGE team at Langley Research Center, described the status of the SAGE II and planned SAGE III instruments and their data products, and presented the results of some stratospheric aerosol research using SAM-2, SAGE II, and Lidar In-space Technology Experiment (LITE) data. It is hoped that SAGE instruments capable of measuring stratospheric aerosols at all latitudes will fly throughout the lifetime of MISR, since SAGE data would be used to filter out the signal from stratospheric aerosols.
Ralph Kahn began the discussion of aerosol retrievals late Wednesday afternoon, by describing the aerosol properties that the MISR team will retrieve and how some constraints can be added to the retrieval, by using data on real measured aerosols. He then described the sort of sensitivity studies being done by the MISR team at JPL, with forward radiative-transfer calculations and retrievals.
Bob West presented results using the discrete dipole approximation (DDA) to calculate the scattering phase function of non-spherical particles. This technique is computationally intensive and can be used only for small particles. He recommended using ray tracing to do these calculations for large non-spherical particles. Michael Mishchenko, from the Goddard Institute of Space Studies, presented the results of his work calculating the phase functions of intermediate-size non-spherical particles using the T-matrix (equivalent spheroid) technique, and the ray tracing technique for larger particles. He noted the overlap in the applicable size range of the two approaches.
The presentations on Wednesday ended with Tom Ackerman's description of the work he and his group are doing on the sensitivity of the calculated radiation field to errors in the aerosol retrieval. Thursday morning began with a presentation by Howard Gordon on his group's work on developing aerosol retrieval algorithms for ocean regions. His group is using linear mixing models in their calculations; these models assume that the radiances due to each physical component, e.g., soot, sulfates, etc., of the aerosol can be added in proportion to its abundance. Wedad Abdou of JPL also discussed the linear mixing approximation. It appears to be good under all conditions except for high optical thicknesses of absorbing aerosols (like soot or mineral dust). The Science Team concluded that linear mixing can be used in calculating radiances "on the fly" at the DAAC, using look-up tables of pre-calculated radiances due to "pure" particle types (including specific size distributions). This will simplify those look-up tables.
On Thursday morning, John Martonchik presented his work on developing a vegetation index which is less sensitive to atmospheric conditions and sun and viewing angles. He described an index based on extrapolating the NDVI calculated at MISRs nine viewing angles (for airmasses between one and three) to an airmass of zero. Martonchik also described the results of a sensitivity study in which he showed that differences between hemispherical-directional reflectance factors (HDRFs) and bidirectional reflectance factors (BRFs) are comparable in magnitude to retrieval errors expected from an imperfect characterization of the atmospheric properties.
Michel Verstraete presented the results of his work with the Terrestrial Environment and Atmospheric Modeling (TEAM) group he heads at Ispra. He proposed modifying the MISR aerosol retrieval algorithm over dense, dark vegetation by permitting the surface reflectance to be a free parameter. The approach subsequently adopted by the team fixes the bidirectional reflectance function angular shape, while allowing the absolute surface reflectance to be a free parameter. Peter Muller described the work he is doing with Alan Strahler, developing techniques for retrieving Bidirectional Reflectance Distribution Functions (BRDFs) from combined MODIS and MISR data.
Jim Conel described the plans (including instruments) for validating MISR algorithms on multi-angle imagery before the EOS AM-1 launch and validating MISR Level 2 products after launch. This discussion focussed especially on surface and aerosol parameters.
After Susan Paradise described the current state of development for Level 2AS product generation software, Daniel Wenkert brought up some cloud-screening issues that cut across the Level 2 software system. After much discussion, it was decided that certain cloud-screening procedures would be performed in the Aerosol/Surface algorithms when these are not used in TOA/Cloud processing, and that each subsystem would be responsible for gathering histograms of cloud-screening observables for the purpose of calculating thresholds for cloud screening.
At the end of the day Bob Vargo and Bob Lutz described the overall MISR software system and data quality assurance issues, respectively. Ralph Kahn discussed some issues involving Level 3 products and gridding schemes. Finally, Dave Diner concluded the meeting by summarizing all decisions made.