Because the relations between precipitation and its electromagnetic signatures are many-to-one, and quite non-linear, Bayesian is a particularly appropriate way to estimate rain and its properties from microwave measurements. When the latter include measurements from different channels (e.g. radar reflectivities as well as brightness temperatures), it becomes very important and rather tricky to give the right weight to each measurement and account for the uncertainty it contributes to the estimates. This proposal will accomplish this by building coarse "constraint" databases along with fine-resolution "profiling" databases. The applications will be to the development of an optimal core reference algorithm, as well as to the improvement of the TRMM combined algorithm so that it can differentiate between liquid and ice water and thereby improve the resulting latent heating estimates.
