This research seeks to exploit the relationships between surface backscatter cross-section and multi-frequency emissivity in combined radar-radiometer precipitation retrievals. Instead of treating emissivity and σ0 as independent random variables, we constrain the solution space to the observed joint distribution of emissivity and σ0. In addition to reducing the uncertainty in the retrieved precipitation, another useful outcome of this method is that the surface properties can be obtained at higher resolution and with improved accuracy relative to the radiometer algorithms, particularly over water surfaces where a physical model is used to calculate surface emissivity and σ0 as a function of wind speed and incidence angle.
To derive the relationships between σ0 and emissivity over land, the 15-year data record of TRMM's radiometer (TMI) and precipitation radar (PR) data has been analyzed to derive the covariance matrix between multi-channel emissivity and multi-angle σ0 at 0.25 degree spatial and monthly temporal resolution. These covariances indicate strong relationships between emissivity and σ0 in regions with sparse vegetation and high variability in soil moisture content, with correlations (r2) between σ0 and horizontally polarized emissivity at 19 GHz as high as 0.8 over parts of India, Australia, and the African Sahel region. The use of these relationships and impact on rainfall retrievals in the proposed GPM combined algorithm framework will be demonstrated.