Modeling anomalous radar propagation using first-order two-state Markov chains

In this paper, it is shown that radar echoes due to anomalous propagations AP. can be modeled using Markov chains. For this purpose, images obtained i n southwestern France by means of an S-band meteorological radar recorded e very 5 min in 1996 were considered. The daily mean surfaces of AP appearing in these images are sorted into two states and their variations are then r epresented by a binary random variable. The Markov transition matrix, the 1 -day-lag autocorrelation coefficient as well as the long-term probability o f having each of both states are calculated on a monthly basis. The same ki nd of modeling was also applied to the rainfall observed in the radar datas et under study. The first-order two-state Markov chains are then found to f it the daily variations of either AP or rainfall areas very well. For each month of the year, the surfaces filled by both types of echo follow similar stochastic distributions, but their autocorrelation coefficient is differe nt. Hence, it is suggested that this coefficient is a discriminant factor w hich could be used, among other criteria, to improve the identification of AP in radar images. (C) 2000 Published by Elsevier Science Ltd. All rights reserved.