IDENTIFICATION OF RAIN CELLS FROM RADAR AND STOCHASTIC MODELING OF SPACE-TIME RAINFALL

We discuss two problems related to the utilization of weather radar: t he automated identification of rain cells and the extension of cluster models to include the random variability of space-time rainfall withi n and between cells. The need for such extension emerges from visual i nspection and formal analysis of radar reflectivity images. The algori thm proposed for the identification of cells is based on statistical t echniques for the estimation of probability density mixtures. The algo rithm does not assign pixels to cells deterministically; rather, it ca lculates the probability with which each pixel belongs to the differen t cells. Through an iterative procedure, the cell parameters and pixel probabilities are updated until the final identification of cells is reached. The second part of the paper deals with a generalization of c luster rainfall models in space and time. The models studied here comb ine an arbitrary birth point process with arbitrary random fields gene rated by the cells. Second-moment properties of these processes are de rived.