P6.17
Testing the CMIS core retrieval algorithm with AMSU data and comparison with NOAA and AMSR products

- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner
Thursday, 2 February 2006
Testing the CMIS core retrieval algorithm with AMSU data and comparison with NOAA and AMSR products
Exhibit Hall A2 (Georgia World Congress Center)
Ryan Aschbrenner, AER, Lexington, MA; and P. Liang, J. L. Moncet, S. A. Boukabara, and F. Weng

The core of the algorithms for the Conical-scanning Microwave Imager/Sounder (CMIS) is a physical inversion algorithm that performs the simultaneous retrieval of water vapor and temperature profiles as well as surface and cloud properties from the combined imaging and sounding CMIS channels. The algorithm uses an inversion method that is based on the maximum likelihood and, in the stand alone mode (i.e. no real-time external data), a priori constraints on atmospheric variables from climatology and on spectral surface emissivity derived from a collection of spectra from model or real data. The forward model is based on the fast and accurate OSS molecular absorption parameterization. The retrieval approach ensures radiometric consistency between the different EDR's and provides a measure (c2) of how well one is able to fit the measurements. Failure to pass the c2 test in the first retrieval attempt with a non-scattering RT model is often indicative of the presence of precipitation/cloud ice and triggers a second trial with a scattering model turned on. Other factors affecting the convergence of the solution include three-dimensional cloud geometry (especially at off-nadir angles) and deficiencies in channel co-registration. The algorithm has been tested in the stand-alone mode with data from NOAA-16 AMSU-A and B (at the AMSU-A resolution). The results are compared with the products of NOAA's algorithms which are run internally at AER and with the EOS/AMSR products. NCEP Global Forecast System model fields provide a first-order verification of temperature/water vapor profiles produced by the core module. This comparison focuses on precipitation detection, precipitable water and cloud liquid water. AVHRR data are used to provide a preliminary assessment of the skills of the different algorithms in the detection of liquid/ice clouds. As part of this exercise, we also investigate the regions of non-convergence and then correlate them with the presence of clouds and areas of precipitation, per the AMSR and NOAA products.