A Statistical Inverse Method for Gridding Passive Microwave Data with Mixed Measurements

Abstract

When a passive microwave footprint intersects objects on the ground with different spectral characteristics, the corresponding observation is mixed. The retrieval of geophysical parameters is limited by this mixture. We propose to partition the study region into objects following an object-based image analysis procedure and then to refine this partition into small cells. Then, we introduce a statistical method to estimate the brightness temperature (TB) of each cell. The method assumes that TB of the cells corresponding to the same object is identically distributed and that the TB heterogeneity within each cell can be neglected. The implementation is based on an iterative expectation-maximization algorithm. We evaluated the proposed method using synthetic images and applied it to grid the TBs of sample AMSR -2 real data over a coastal region in Argentina.