The Microwave Integrated Retrieval System (MiRS): Validation Activities for NOAA-20/ATMS Products and New Science Developments

The Microwave Integrated Retrieval System (MiRS) is the official NOAA microwave-only retrieval system which currently processes data from ten different low earth orbit satellites, including the NOAA-20 satellite, which launched in November 2017. MiRS is based on a one-dimensional variational approach. MiRS derived products include vertical profiles of temperature, water vapor, cloud liquid water, liquid water, rain water, ice water, surface temperature, surface emissivity, sea ice and snow related variables and etc. With more than one year of data now available from NOAA-20/ATMS, we have conducted an extensive evaluation of the retrieval products during 2018 and 2019 since this is a key procedure to confirm overall retrieval quality. At the same time, as a check for consistency, SNPP/ATMS retrievals from the same time period are also validated.

The validation of the MiRS products will include temperature and water vapor profiles, total precipitable water (TPW), land surface temperature, land surface emissivity, sea ice concentration, snow cover, snow water equivalent (SWE), rain rate, and cloud liquid water. Reference datasets have been selected for each variable, such as ECMWF and GDAS analysis data, radiosonde profiles (provided by NPROVS), Surface Radiation Budget (SURFRAD), VIIRS, AMSR2 products, Interactive Multisensor Snow and Ice Mapping System (IMS), as well as both Stage-IV and MRMS precipitation analyses. Daily comparison results of the MiRS products compared to ECMWF and GDAS analysis data are available on https://www.star.nesdis.noaa.gov/mirs/geonwp.php and some portions of this validation study are from the daily comparisons. To assess performance quality, validation results are compared with official JPSS program requirements for each retrieval product (JPSS-REQ-1004).

Additionally, brief examples of preliminary science improvements will be provided. Possible examples include the use of a machine learning approach to characterize scene-specific radiometric biases, and a version of MiRS optimized for retrievals in tropical cyclone environments.