Application of the unified retrieval approach to real DMSP sensor data over ocean and land

A physical retrieval algorithm for the simultaneous retrieval of atmospheric temperature, water vapor and cloud liquid water as well as surface skin temperature and emissivity from microwave sensors has been developed by Moncet and Isaacs (1994, 1992). The algorithm uses a nonlinear inversion method similar to the one described by Rodgers (1976) for the inversion of the measured brightness temperatures. Climatology provides the desired inter-correlation between the various elements of the state vector. This information is used to effectively reduce the number of degrees of freedom in the problem, and therefore reduce the dependence of the solution on the first guess. Information from other sources, such as forecast models, is integrated by optimally combining it with the primary background information. Emissivity is treated by retrieving one emissivity value per sensor channel. The degree of correlation between the emissivities in each channel is specified through the first-guess error covariance matrix. This method offers more flexibility than the one proposed by Eyre (1990) and makes it possible to apply the algorithm to combinations of sensors with mixed viewing geometries and polarizations such as the DMSP microwave sensor suite. The algorithm has been tested based on simulated data and was successfully applied to limited sets of real measurements from the combined DMSP microwave sensors (SSM/T-1, T2 and SSM/I) over both ocean and land