David Herring (herring@ltpmail.gsfc.nasa.gov), MAST Technical manager, Science Systems & Applications, Inc.
The MODIS Science Team Meeting took place in Greenbelt, Maryland on May 3-5, 1995. It was chaired and called to order by Vince Salomonson, Team Leader. The emphasis of this meeting was to provide a working session for the Science Team.
John Dalton, Deputy Associate Director of the Earth Science Data and Information System (ESDIS) Project, presented a status report on EOSDIS. Dalton reported that the first quarterly review of science software for AM-1 instruments was held recently. Also, EOSDIS hosted a Science Software Integration & Test Workshop, attended by instrument team software developers and representatives from the Distributed Active Archive Centers (DAACs), to discuss and delineate the science software integration procedures and deliverables. At the workshop, the MODIS science software integration and testing procedural steps were reviewed and modified. Also, at that meeting it was determined that the GSFC DAAC will be available to the MODIS Team for pre-acceptance testing. The first draft of the MODIS/GSFC DAAC Science Software Integration and Testing Agreement was due June 1, 1995, and the final draft is due Oct. 1, 1995. EOSDIS is now beyond its Preliminary Design Review (PDR) and is preparing for its Critical Design Review (CDR).
Chris Scolese, EOS AM Project Manager, told attendees that assembly of the hardware has begun for the spacecraft and sensors. Scolese reported that the EOS AM-1 bulkhead structure assembly is underway and will be completed late this summer. Testing of the structure will begin by December 1995. In January 1996, Lockheed-Martin will begin integrating spacecraft flight hardware onto the platform.
Scolese stated that discussions regarding a lunar-view maneuver for calibration are still ongoing. He observed that some lunar-view maneuvers cannot be completed within half an orbit. He explained that the EOS AM Project would like to be able to rotate the spacecraft to look at the moon and deep space, and then back at the Earth again before the spacecraft comes around to the dark side of our planet. There are concerns about the solar impingement on the EOS instruments.
Ricky Rood, Head of the Data Assimilation Office, delivered a presentation on data assimilation. Rood explained that the data currently obtained from all sensors--ground-based, airborne, and satellite-based--are very irregular in spatial and temporal scales. Data assimilation provides added value to these data in that it organizes the data, complements the data and fills in unobserved regions, supplements the data and provides unobserved quantities, provides a means of quality control, and can assist in instrument calibration. In short, data assimilation provides the best estimate of the state of the system in that it allows extraction of maximum information content from data and it allows more-quantitative interpretation of satellite data.
Ed Masuoka, Science Data Support Team (SDST) Leader, told the Science Team that SDST is producing a Validation Plan, the first draft of which will be produced by December 1995. Masuoka told the Team that the Level 2 Beta 3 integrated (swath) science software is due in July 1995; the Level 3 Beta 3 integrated (grid) science software is due in August 1995. He stated that by January 1996, end-to-end system tests will be complete, and the beta releases will be baselined and delivered to the DAAC. By April 1996, the integration and debugging of all MODIS software at the DAAC will be complete.
Al Fleig, of SDST, announced that two primary products are now complete: 1) complete test data sets and 2) tools for making test data on order. The data in the first data set include correct viewing geometry according to the characteristics of an EOS platform orbit. The data include a characteristic MODIS scan pattern with all 36 channels and the bowtie effect, and are processed into MODIS Level 1B format (250 m and 500 m data are replicates). Sun-Earth/orientations are included as a function of time, date, and location.
Steve Ungar, of SDST, presented his work in producing simulated MODIS data scenes of the United States. He showed sample TOA (top of the atmosphere) radiances that are not based on any observations--they are totally simulated. Ungar stated that he is putting together a simulated MODIS data set to include all 36 MODIS bands at a resolution of 1,534 by 1,534 pixels (1.7 km is the average size). Ungar refined the simulated scene over the U.S. to include the MODIS instrument response functions supplied by Ed Knight. The next level of refinement will include an improved representation of the MODIS scan geometry and variable atmospheric path length. The ground truth elements consist of primary surface type (11 categories), secondary surface type (16 categories), mixture ratio, and elevation. The surface-type assignment is based on the classification of eight 1-km AVHRR Normalized Difference Vegetation Index (NDVI) images from March to November 1991. Ungar stated that his simulated data may be accessed via anonymous file transfer protocol (FTP) at highwire.gsfc.nasa.gov in the "/pub/modsim" directory.
Richard Weber, MODIS Project Manager, reported that ambient and thermal vacuum tests of the Engineering Model (EM) are complete at the Santa Barbara Research Center (SBRC). Weber reported that all materials are either on order or have been received for the Prototype Flight Model (PFM) and Flight Model 1. He stated that cost remains a major concern, as is the development schedule in that any slips will negatively impact cost. Weber listed his top five technical concerns currently facing MODIS development: 1) transient response, 2) scan motor lifetime, 3) bandpass filters, 4) radiative cooler, and 5) electronics.
Lee Tessmer, MODIS Project Manager at Hughes SBRC, stated that the MODIS EM optical bench is assembled and the onboard blackbody has been integrated. Regarding the MODIS PFM, Tessmer reported that the engineering documents from the EM are readily transferable to flight status, noting that there are only minor changes to 10 of the 52 drawings. The procurements of hardware for the PFM have already begun. He reported the results of SBRC's EM tests. Overall, the polarization meets most requirements. All bands are within specifications except Band 3. He stated that the thermal vacuum testing includes comprehensive spatial, spectral, and radiative tests, the results of which show excellent co-registration, and low scan-to-scan jitter, as well as optimized gains and offsets.
Tessmer stated that the near-field response tests verify SBRC's test methodology. They found that there is an unacceptable problem with the first dichroic. They have all the instrumentation in house they need to make a new, modified first dichroic and plan to begin immediately. He said that SBRC's ambient test demonstrates good spatial performance‹ambient spectral data were acquired for all MODIS bands. In summary, the EM is fully functional and robust. It performed as expected, demonstrating good linearity, high SNRs, low polarization, and good registration. The instrument meets specifications in terms of size, mass, power, and data rate. MCST Status Report
Bruce Guenther, MODIS Characterization Support Team (MCST) Leader, announced that MCST recently completed the transition to its new support contractor--General Sciences Corporation (GSC). Guenther said his team established a MODIS Test and Analysis Computer (TAC) at GSFC for processing MODIS test data sets. Also, a Risk Management Board was established to identify, track, and control risks. Guenther reported that the MCST Algorithm Theoretical Basis Document (ATBD) was recently revised and will be revised again in early 1996.
MCST will host a vicarious calibration workshop in August at Wallops Flight Facility to focus on its Level 1B data products. Their objective is to identify vicarious calibration data sets, to review the instruments that produce them, and to consider ways to use the data sets in developing Level 1B data products. Guenther noted that some vicarious calibration data sets are produced with varying concepts for error bars. He hopes to establish a common scale for quantifying error and uncertainty.
Guenther reported that MCST is conducting analyses of MODIS' key characteristics, such as scan-angle effects. He explained that in the infrared there are instrument characteristics that pose challenges in determining scan-angle effects. Consequently, there is concern as to how to track these effects once MODIS is in orbit. MCST is considering the following strategies: 1) placing a second onboard blackbody in the scan cavity near the diffuser, 2) using deep space to observe emissive infrared wavelengths, and 3) using the moon to observe reflected solar wavelengths.
MCST is analyzing another key MODIS characteristic including near-field scatter. Guenther stated that at present the MODIS ghosting problem seems to be well understood and well corrected. Guenther stated that, according to test measurements, the filters for Bands 20 - 25 µm need attention. John Barker, of MCST, is conducting sensitivity analyses to understand the impact of the performance of these filters.
The emphasis for this MODIS Science Team Meeting was to provide a working session for the Science Team. Therefore, rather than split into discipline groups, the Team divided into interdisciplinary groups to hold "roundtable" discussions restricted primarily to selected panelists and a moderator. The following sections summarize each of the sixteen roundtable sessions.
Calibration--One session focused on strategies for vicarious calibration. Phil Slater summarized this session, stating that for the first time ever we will have a great deal of calibration information available to us from the moment the EOS sensors are in orbit. If we are to make the most of this information, we must get organized and determine the relative accuracies of the various methods of vicarious calibration being used. Slater suggested that the EOS Calibration Scientist could coordinate this organization activity. Specifically, this person could: a) coordinate vicarious calibration activities between different vicarious calibration groups nationally and internationally, b) centralize the evaluation of different techniques used by different groups, c) arrange the cross-comparison of measurements between various groups to help in evaluating vicarious calibration methods, and d) depending upon a through c above, recommend the role(s) MCST may play in the calibration/characterization of MODIS. Slater pointed out that error budgets between different vicarious calibration teams are not always the same.
Remote Sensing of Aerosol and Atmospheric Correction--Yoram Kaufman proposed holding an international workshop in April of 1996 to broaden the scope of the MODIS Science Team Members and to enhance or facilitate further involvement with other investigators around the world. He reported that some discussion focused on spectral measurements from both ground-based and satellite instruments in determining the effect of single scattering albedo. Kaufman stated that there is a need for strong, continuous interaction among those developing algorithms and acquiring observations and those, in particular, who are modeling aerosols.
Regarding use of AFGL atmospheric models and dynamic models, Kaufman said most groups using models suffer from the fact that they are models of averages, not actual conditions. Kaufman concluded that there is a need to integrate satellite- and ground-based campaigns.
Gridding and Averaging--Alan Strahler stated that gridding deals with the Level 3 products. His panel concluded that it is probably better to work with a fine-resolution grid (250 m) and collapse to coarser grids. An Action Item was assigned to Bob Evans, Robert Wolfe, Dave Diner, and Bruce Barkstrom to pursue this issue with the Science Working Group for the AM Platform (SWAMP).
Strahler recounted that the proposed grid is more or less an equal-area grid, but not exactly. The proposal is for a modified, nested International Satellite Cloud Climatology Project (ISCCP) grid that is defined on the basis of 1.25-degree squares. Steve Ungar persuaded the Panel to move to a fine grid of 270 m so that it will be easy to collapse down to 1.25 degrees. Strahler told the team that the edges of each grid cell will appear ragged like a postage stamp. There is a need for cartographic tools and resamplers so that modelers can go from basic to angled grids to derive map projections.
Resampling and Remapping Procedures--Strahler also summarized Session #3. Strahler stated that there is a need for a MODIS-specific tool for producing interpolated projections with observed physical features or phenomena in the proper place. For the MODIS Land Group (MODLAND), developing this tool is trivial, but for cloud observation it becomes much more difficult.
Strahler explained that trying to fit a MODIS image onto a map projection is going to be difficult, and will require some means for resampling. If we don't resample, then we can take those data and place them into a grid. This logic led to Eric Vermote's idea of a Level 2G data structure for forward binning the data, which would work well for computing surface reflectance and Bidirectional Reflectance Distribution Function (BRDF). Strahler pointed out that the Oceans Group is doing something similar, but with a coarser grid. The panel also examined the MODIS bowtie effect to determine how it affects Level 2 processing. The panel concluded that the bowtie provides interdetector calibration, which is good, but makes resampling with interpolations difficult, which is bad.
The panel concluded that Level 1G and 2G products would be very helpful in some cases, but further work and thought are needed there by Catherine Harnden, Howard Gordon, Paul Menzel, Robert Wolfe, Steve Ungar, and Strahler. The panel recognized that there is a need to define a "day." The grid may represent a unit of time, Strahler rhetorically asked, but what do you do if the dateline falls in the middle of a swath?
Algorithm Integration--Ed Masuoka reported that the Science Team needs to deliver its Level 2 code at the end of July and Level 3 code at the end of August. He noted that the Atmosphere Group may need some help from SDST in integrating the cloud products. Masuoka stated that for MODLAND, integration and testing of algorithms for scientific accuracy will require more scientifically accurate synthetic data. Ancillary data will be necessary for beta and should be consistent with the simulated data. He announced that a second MODLAND/SDST meeting is scheduled for late July 1995, at which metadata for version 1 delivery will be discussed. Masuoka noted that the Ocean Group will use RATFOR programming language for its beta delivery, and FORTRAN 90 for delivery of version 1. He stated that the Ocean Group will provide their own simulated input data that SDST can put into a scan cube.
Remote Sensing in the Infrared--Bill Barnes stated that a full-swath scan of deep space is vital to calibration, so the MODIS Team should continue requesting this capability. Barnes said he is developing a white paper on the subject and is working with the Science Working Group for the AM Platform.
The panel would like the Team to consider the possibility of applying a new coating on the scan mirror in order to reduce polarization in the visible region of the spectrum. Barnes said he will review this possibility and report his findings to the MODIS Technical Team. The panel also urged the Team to continue studying the possible use of a second blackbody in the scan cavity as a "pseudo" blackbody to obtain calibration data at a large angle of incidence to the scan mirror. Barnes pointed out, however, that a second blackbody is no substitute for a deep space view, which will still be required.
Barnes noted that MCST was given an Action Item to evaluate TOA (top of atmosphere) data taken by the MODIS Airborne Simulator (MAS). Jim Smith was asked to locate carbon dioxide polarization data to complement MCST's efforts. Considerable discussion was devoted to the thermal environment of the scan mirror under various conditions and maneuvers. Barnes reported that he expects the scan cavity temperature to change only by a few degrees. The panel asked SBRC to complete and analyze the new scan-mirror temperature measurement design and report at the next Science Team Meeting.
Barnes reported that the latest version of the Level 1B infrared calibration algorithm will soon be forwarded to the Science Team for review. MCST must examine approaches for post-launch validation of infrared radiances.
Simulated Data and Software Verification--In lieu of moderator Wayne Esaias, Al Fleig summarized session # 6. Fleig quoted a point made by Steve Ungar: it is important post launch to have a way to simulate and study any artifacts found by the Team. Fleig reminded the Team that there is a MODIS simulated data set and it is evolving. In its discussion, this panel decided that software verification is similar to quality assurance and validation, so the panel decided to leave further discussion up to those session panels.
MODIS Data Product Browse Capability--Fleig said the purpose of the browse panel was to determine what browse products should be available for MODIS, keeping cost and utility in mind. The purpose of the browse product is to help data users sort through large volumes of data to decide what data to order, what each data product looks like, and whether the requested data segment is a good one. The panel recognized that the nature of browse products can vary for each MODIS product. Fleig stated that we need to be responsive to the Science Team's, as well as the Interdisciplinary Science (IDS) community's, browse needs. SDST will follow up with EOSDIS Core System (ECS) to determine their plans. Consideration will be given to providing browse capability via the WWW.
Ancillary Data and Assimilation--Kendall Carder reported that there are three primary concerns for ancillary data and assimilation: 1) timeliness issues, 2) spatial issues, and 3) pooling ancillary data requests. The panel determined that MODLAND needs precipitation, soil moisture, photosynthetically active radiation (daily), maximum and minimum temperature, and surface pressure data. The Ocean Group's needs have not yet been determined. The Atmosphere Group needs surface emissivity maps over land for its shortwave and longwave IR products. Atmosphere also needs aerosol ancillary data.
The panel concluded that the timeliness of MODIS model outputs will be dependent upon MODIS coverage. The first iteration of Ricky Rood's model will be available within 24 hours using National Meteorological Center (NMC)-like input fields, and the second iteration will be available within 1 month using EOS data fields. Spatially, 1.25 degrees by 1.25 degrees is an adequate grid size to allow modeling--each discipline group will interpolate in space as needed. Regarding quality assurance of ancillary data sets, Carder stated that each algorithm should be tested independently. Ancillary data sets should be sent to the Team Member computing facilities along with MODIS data sets for quality assurance.
SCAR-B Update--Yoram Kaufman told the team that SCAR-B is the third and last of a series of field campaigns focused on the interactions of Smoke, Clouds, and Radiation. SCAR-B will be conducted in Brazil. Kaufman reported that a Memorandum of Understanding (MOU) has yet to signed. [Subsequent to the MODIS Meeting, the MOU was signed by the President of the Brazilian Space Agency (AEB) and the Administrator of NASA.] Otherwise, significant progress is being made in preparation for the campaign. Ground sites for AERONET instruments have been chosen, calibration preparations are being made, a communications infrastructure is being developed, and dates of operation have been determined (August 15 through September 25, 1995). MODLAND plans to participate in the campaign along with the MODIS Atmospheres group.
Cloud Masking and Cloud Products--Vince Salomonson reported that, in short, what is needed to develop a cloud mask algorithm is calibrated, navigated radiance data in fifteen channels, as well as certain ancillary data, such as 1-km land/water maps from the EROS Data Center (EDC) and 1-km topography data. Additionally, snow/ice maps and cloud radiance composite maps are needed. According to Steve Ackerman, University of Wisconsin, the cloud mask will be implemented as a 32-bit word. He said there is some question as to whether cloud shadow can be determined at 250 m resolution--an issue that still needs further research. Ackerman concluded that the cloud shadow efforts should be conducted spectrally initially, and then spatial/geometric algorithms should be added afterward.
Salomonson stated that the 32-bit approach looks good and development should continue. Enhancements will be added as resources and time permit. The data sets currently being used for development are the HIRS/AVHRR cloud mask, AVHRR LAC data, and MODIS Airborne Simulator (MAS) Gulf Experiment data. Salomonson told the team that the main issues remaining are: 1) development of a land/sea cloud flag, 2) confidence flags, and 3) input from the Team stating how complex they want the cloud mask to be. Salomonson concluded that although there was some nervousness last summer during the review of the cloud mask ATBD, it appears now that good progress is being made.
Resources for Product Generation--Masuoka summarized the panel discussion that had been moderated by Wayne Esaias. Masuoka reported that 100 percent of the bandwidth required by MODIS products will be available at launch. In terms of CPU capacity, four times more will be available than MODIS' stated needs; however, this capacity will be phased in. At launch, MODIS storage space allocation will be 400 Gbytes. Capacity models are being developed for the processing of MODIS Level 1 and 2 data products. Masuoka stated that the question of where time slicing will occur still remains. Also, the decision process for determining standard versus research products needs to be established.
DAAC-Team Leader Computing Facility (TLCF)--Science Computing Facility (SCF) Interactions‹Masuoka reported that discussion focused on EDC's beginning software testing early to iron out operational issues before the 1997 delivery deadline. The EDC DAAC is willing to provide time on an SGI system connected to their tape archive but ESDIS will need to increase the capacity of that machine to support MODIS testing. Both the GSFC and EDC DAACs stressed the importance of recruiting highly skilled science and operations staff throughout early software and system integration and later operational processing. This staff would provide better feedback on lessons learned, as well as quicker response time in resolving operational problems. Currently, each DAAC has one full-time person supporting MODIS science software, and both would like to add several more personnel.
MODIS Data Quality Assurance Plan--Bob Evans reported that the MODIS Data Quality Assurance Plan is currently being developed and is in draft form now. He told the team that the EOS Panel on Data Quality is chaired by Mike Freilich. Evans stated that quality assurance (QA) in an algorithm context refers to spectral checks, spatial checks, and temporal checks within one day of obtaining the data. Validation refers to all other tests. The QA process will identify or "flag" pixels to granules which obviously do not conform to expected accuracy. Evans added that QA will also allow the Team to monitor the health of products. Evans said that the beta version of the QA Plan will be submitted by January 1997, and the final version will be completed by mid-1997.
Validation--Chris Justice stated that there are some good models in place as to how the MODIS validation effort can proceed, including the sun photometer network, the SCAR campaigns, land test sites, and the Ocean Color Working Group initiative. He said the Team needs "community" guidelines on the level of required validation. He pointed out that validation planning is a problem in light of the constantly changing budget. For example, he asked, who pays for aircraft for field campaigns?
Justice observed that there are various international coordination mechanisms for validation already in place. He feels that the Land community needs a distributed network of sites to supplement intensive NASA campaigns, such as the First ISLSCP Field Experiment (FIFE), Boreal Ecosystem-Atmosphere Study (BOREAS), and validation field experiments in Amazonia. Dave Starr, EOS validation scientist, encouraged the Team to continue developing its "bottom up" validation initiatives. Justice concluded that to achieve product validation, the Team will need help from the EOS Project with coordination--particularly with interagency and international coordination and funding.
Plans for Flying Aircraft Over Ocean Test Sites--Frank Hoge stated that plans are needed for Case 1 and Case 2 Atlantic Ocean field experiments involving the MODIS Atlantic Test Site (MATS) and the Bermuda Atlantic Time Series. He added that plans are progressing for a joint Marine Optical Buoy (MOBY)-MAS campaign in Hawaii in the spring of 1996 contingent on the launch of ADEOS and SeaWiFs. [Subsequent to the meeting this planned compaign was delayed one year.] He feels there is a need for the Team to identify post-launch airborne field experiments. Hoge concluded that airborne platforms, sites, and institutions do exist to conduct robust experiments--he recommended holding focused planning sessions.
Salomonson concluded the meeting with a discussion on the meeting format. In brief, he liked the roundtable panel discussion format and plans to reuse it. The next meeting, scheduled for Nov. 13-17, 1995, will feature a shorter plenary session, roundtable discussions, discipline group splinter sessions, and a final plenary summary session.