Physical Inversion and Data Assimilation Pre-Processing Using the MiRS Variational System

Dr. Sid Boukabara, NOAA/NESDIS/STAR: 
"Physical Inversion and Data Assimilation Pre-Processing Using the MiRS Variational System. Application to the Microwave Sensors Constellation (SNPP, POES, Metop, DMSP, GCOM-W, GPM, M-T and TRMM)"

We present in this seminar the mathematical basis, the technical implementation and the performances assessment of an iterative physical algorithm based on a Bayesian variational approach. This algorithm, called the Microwave Integrated Retrieval System (MiRS), is used operationally in NOAA to generate sounding, surface, hydrometeor and cryospheric parameters from a variety of microwave sensors including AMSU/MHS, SSMIS and ATMS onboard POES/Metop, DMSP and SNPP platforms, respectively. It is also applied routinely in a research mode (non-operationally) to data from AMSR-2, TMI and SAPHIR onboard GCOM-W, TRMM and Megha-Tropiques satellites, respectively.

The algorithm relies on the Community Radiative Transfer Model (CRTM), developed in the Joint Center for Satellite Data Assimilation (JCSDA), to (1) simulate brightness temperatures and (2) generate Jacobi (dxu: weighting function) with respect to all geophysical parameters. These two components, along with the (3) background covariance matrix used, are critical for the physical inversion.  In order to ensure a stable and fast processing, the inversion is undertaken after projecting it into a reduced space using the Empirically Orthogonal Functions (EOF). The state vector parameters are retrieved simultaneously, which ensures that the resulting geophysical solution fits the observations consistently, which is a necessary, although sometimes overlooked, condition for the inversion process.

Obviously, the performances obtained by MiRS when applied to different satellite data, will depend on the information content of those sensors and their configurations (frequencies, polarizations, viewing angles, etc). We will present a snapshot of the performances, both as obtained internally using a locally developed testbed and by independent assessments provided by users and outside teams (such as the International Precipitation Working Group IPWG). We will in particular look at the performances of (1) the total precipitable water over all surfaces including ocean, land, sea ice, snow and coastal surfaces, (2) the surface rainfall rate, (3) the atmospheric temperature and moisture vertical profiles and (4) the surface emissivity and temperature.

In addition to the inversion aspect, we will highlight a few recent JCSDA-led developments that aim at applying the same technology used in MiRS to develop a uniform Quality Control (QC) and pre-processing system called the Multi-Instrument Inversion and Data Assimilation pre-Processing System (MIIDAPS), used as a tool to pre-process all satellite data (to apply to microwave and IR sensors) before they are assimilated into the GSI system. This tool allows to (1) optimize the QC filtering and spatial thinning of the data, as well as (2) provide an estimate of the dynamic surface emissivity as well as (3) provide estimates of sounding profiles in cloudy and rainy conditions, situations where the data is currently rejected. Future directions and collaborative efforts will be presented as part of the presentation.

Location M-Square Building #950 Room # 4102 (Large Conference Room) 5825 University Research Court, College Park, MD 20740
Contact Isaac Moradi, imoradi@umd.edu