The third U.S. EOS PM-1 AMSR Science Team meeting was held 22-23 October 1996, at Goddard Space Flight Center. The first day was dedicated to team business and Algorithm Theoretical Basis Document (ATBD) status reports from all team members. The EOS PM Project Office updated the team on the hardware status. The second day was taken up by a data types presentation given by G. Cox, EOSDIS, and software discussions led by D. Conway, EOS PM-1 Software Lead Engineer.
<Body text>The meeting started with Michael King, EOS Senior Project Scientist, showing the ATBD status of the other EOS instrument science teams. He gave details of what is expected of the ATBDs and the peer review presentation. The peer review will be scheduled for late February 1997.
Marty Donohoe, EOS PM Project Manager, gave the team an update on the common spacecraft contract. TRW, the spacecraft contractor, will conduct the bus requirements review in December 1996, and the preliminary design review (PDR) in April 1997. The interface control document, between AMSR and the spacecraft, is complete with few TBDs remaining.
Bernie Graf, PM Project AMSR Instrument Manager, reported on the EOS PM-1 AMSR hardware status. The meeting in Japan the last week of October with Mitsubishi Electric Company (MELCO) closed two of the main hardware issues: the thermal problem raised by the accommodations of the control unit, and the antenna tiedown points (two versus three points). He then showed measured EOS PM-1 main reflector and cold sky reflector antenna patterns. The ADEOS II AMSR Critical Design Review (CDR) will be held in December 1996 and the EOS PM-1 AMSR CDR a year later. The scheduled delivery of the instrument to the PM-1 spacecraft is June 1999.
The ATBD status presentations followed. Frank Wentz presented the Level 1C brightness temperature and the ocean suite ATBD status. Chris Kummerow talked about the highlights of the precipitation ATBD. Don Cavalieri and Joey Comiso discussed their work on developing the hybrid sea ice algorithm, and the plans for their joint ATBD. Al Chang and Eni Njoku presented their work on their respective documents: snow water equivalent/depth and the land standard products (surface soil moisture, effective surface temperature, and vegetation water content).
Frank Wentz presented the plan for the Level 1C data record. This record includes a header (time, orbit number, incidence and azimuth angles, location information, and surface classification) and retrieved brightness temperatures with four distinct resolutions: the 6.9, 10.7, 18.7, and 36.5 GHz channel resolutions. Suggestions from the team were to add a fifth resolution: the 89 GHz. Wentz also showed the status of the Level 1C data set simulator: a preliminary simulator exists but it does not have data for all channels. An image created from one of the channels of the preliminary simulator shows the correct orbit and the correct land type.
In the ocean suite ATBD, Frank Wentz, as the leader of the group, presented the complete models that will be part of the theoretical background to the ocean suite algorithm. In the atmospheric model he uses Liebe 1985 for the oxygen, a modified Liebe 1985 for the water vapor, and Rayleigh absorption for the cloud water components. For the specular sea surface model he recalculated the dielectric constant from 14 sources of data and found a 2 K difference from the Klein and Swift dielectric constant value. For rough sea surface, Wentz formulated a model that assumes sea foam to have blackbody characteristics and allows a pixel to have fractional foam coverage. Using all these models in his `AMSR TB model' and the initialization data for sea surface temperatures, wind speeds, cloud layer heights, and radiosonde profiles, he obtains Tb`s for ocean scenes. After adding radiometer noise Wentz uses the standard least squares regression to converge to the ocean standard products: SST, wind speed, columnar water vapor, and columnar cloud water estimates. The retrieval errors for these estimates are well within the EOS AMSR specifications. One of the important components in this model, the effect of wind direction, is yet to be added. In the ATBD, Wentz will show a comparison between the linear statistical regression algorithm and the maximum likelihood estimator. The other group members (Ferraro and Wilheit) will contribute to the calibration/validation and historical perspective sections.
C. Kummerow, leader of the precipitation ATBD group (Ferraro, Kummerow, and Wilheit), presented the highlights of their document. The products generated with the precipitation algorithm are instantaneous rainfall intensity on a pixel-by-pixel basis (Level 2) and estimates of monthly totals on a 5deg.x5deg. basis (Level 3). In addition to rainfall intensity over the ocean, profiles of the hydrometeors will be classified as convective or stratiform (at Level 2). The three components of the algorithm are: ocean component (Level 2), land component (Level 2), and monthly rainfall accumulation (Level 3). In the ocean algorithm component the brightness temperature field is convolved with the AMSR measured antenna gain functions to produce a set of possible cloud profiles. These profiles together with their respective brightness temperatures form the database used in the precipitation retrieval over ocean. The precipitation profile retrieval method is an integral version of the minimum variance solution; it employs a Bayesian inversion methodology, and it is computationally efficient.
Rain over land will be detected by the depression of the brightness temperature caused by precipitation-sized ice particles and/or large rain drops. The magnitude of this depression is proportional to the rain rate. The basis for the retrieval over land comes from N. Grody's work at NOAA. Grody developed a global scattering index at 85 GHz for use with the SSM/I sensor. R. Ferraro built upon Grody's study and developed a more robust set of relationships to be used for detection of rain over land from SSM/I. Ferraro also rederived the relationships to separate rain from snow and deserts and introduced a new screen for semiarid regions. To retrieve instantaneous rain rate, the scattering index has been calibrated with ground-based radar measurements. Cluster analysis and principal components analysis will be used to examine if the lower frequency channels on AMSR are more efficient in separating rain from snow and deserts.
The third component of the precipitation algorithm is the monthly rainfall accumulation. The standard product is monthly rainfall totals for a 5deg.x5deg. area. The embedded Level 2 oceanic rain algorithm is based on the radiative transfer model, where the initial distribution of hydrometeors, the atmospheric temperature structure, the water vapor profile, and the surface reflectivity are specified. The resulting freezing level is used to compute the possible rain rates, valid over a specific dynamic range (depending on the frequencies used). Each of these rain rates is corrected for beam filling and then accumulated into histograms for the area and time period (nominally 5deg.x5deg. by one month). A maximum likelihood estimator is used on a histogram to compute the parameters of the log-normal distribution of rain rates. The Level 3 standard product is the estimate of the monthly rainfall accumulation in this 5deg.x5deg. area.
The sea ice group, D. Cavalieri and J. Comiso, presented a flow diagram for their sea ice concentration standard product algorithm. After ingesting the Level 1C brightness temperatures, the `PR/GR' routine will be used initially to compute concentration. A filter will determine the thin ice and emissivity anomaly areas, and the sea ice concentration will be recomputed (for those areas) with the `19V vs. 37V' routine. Plans for the other two standard products (sea ice temperature and snow depth on sea ice) algorithms were also discussed.
Al Chang summarized the snow water equivalent algorithm, the basis for the snow ATBD. The Nimbus-7 SMMR snow algorithm is the starting point for the current algorithm. The radiative transfer equation includes radiation loss, reradiation, and scattering. The assumptions made (due to lack of snow ground truth information) are: 1) a uniform, spherical grain size, 2) incoherent scattering by snow grains, and 3) a mean snow density. The final algorithm will address issues such as: consideration of a dense media model, a snow particle distribution (instead of assuming uniform particles), snow growth model and different classifications, and corrections for fractional forest cover and elevation. The Microwave Workshop (St. Lary, France, 1993) recommendations for a snow water equivalent global algorithm are the inclusion of the following: snow metamorphism, vegetation cover, snow on warm ground, signal saturation at 150-200 mm, and mixed pixels. Finally, Chang presented estimated versus measured snow water equivalent data; after applying the forest cover correction, the estimates come much closer to the measured values.
E. Njoku presented the retrieval algorithm for the land surface parameters: surface soil moisture, surface temperature, and vegetation water content. The simplified radiative transfer model for surface observations from space has two components: the atmospheric temperature term and the soil/vegetation term; both terms are modified by the atmospheric opacity (a function of water vapor), soil reflectivity (a function of soil moisture) and vegetation canopy opacity (a function of water content). Njoku determines the regression coefficients from a simulated data set, and then uses a nonlinear, iterative retrieval method (minimization of [[chi]]2 merit function) to retrieve each of the land surface parameters.
Software and EOSDIS issues took up the entire second day (from 8:30 until noon). K. Cox from the EOSDIS office presented an overview of the role of ECS metadata, discussed the function and components of Earth Science Data Types (ESDTs), and described the ECS metadata population process. The presentation started with the definition of a granule. The team decided that, for PM-1 AMSR, a granule will consist of data in one full orbit. A question still exists on the definition of an orbit: starting point, overlap scans,.... . M. Schwaller from EOSDIS discussed the availability of the ancillary data needed by the team.
D. Conway wrapped up the meeting with discussions on toolkit implementation, SCF and Team Algorithm Development Plans reviews and inputs, and finally the EOS PM-1 web page. She listed several action items which can be viewed by the team members on the "Team Only" page.
The next AMSR meeting will take place the day after the ATBD peer reviews in mid-March 1997.