Improved combined radar/radiometer rain profiling

The current TRMM combined radar/radiometer profiling algorithm compensates for the known shortcomings of each instrument by exploiting the strengths of the other. It turns out that the strengths/weaknesses of the radiometer measurements are not as they seemed before TRMM. Specifically, the current algorithm presumes that the radiances at the various frequencies are approximately independent (to compute probability weights). It turns out there are strong correlations between the channels. New estimates of the conditional covariance of the radiances given the rain should make the probability weights more realistic. The second problem stems from the representativity of the TRMM cloud database. It turns out that the database is significantly off, especially at higher frequencies. To reduce the rain over-estimation produced by this discrepancy, the database was re-sampled and mean rain-radiances relations were re-derived. The third problem is the lack of any ice estimates. The reason was the large number of unknown variables involved. A principal component analysis has revealed that the frozen hydrometeor profiles can be approximated by a single variable each for ice, snow, and graupel. A straightforward method is currently being implemented to estimate these additional variables, and to include them in the output structure of the algorithm. Linear formulas will enable users to reconstruct the corresponding graupel/snow/ice profiles, and estimate the corresponding latent heating.