A Genetic Algorithms-Based Technique for the Retrieval of Dry Snow Parameters

Monitoring the quantity, distribution and dry/wet state of snow is important for several applications. Geophysical parameters such as grain size and depth through microwave radiometry can be useful for, water storage estimation, hydrological processes understanding and weather modeling and hazards forecasting. The inversion of relationships between the snow parameters and electromagnetic quantities poses a significant challenge because it is rarely possible to perform this inversion in a strictly analytical way and thus numerical techniques are often used. The inversion problem may be solved by using a linear regression between the snow parameter of interest and a combination of electromagnetic quantities (e.g. Chang's algorithm for the retrieval of SWE) or by using iterative techniques (e.g. the HUT iterative inversion technique). In this paper we propose a technique which uses Genetic algorithms (GAs) to invert the equations of an electromagnetic model based on Dense Media Theory (DMRT) for the retrieval of snow parameters (e.g. snow depth, mean grain size and fractional volume) from microwave brightness temperatures. Genetic algorithms are iterative procedures and are inspired by Darwin's theory of evolution. They have proved to be successful as a search and optimization technique. Different configurations of genetic algorithm parameters are tested by using simulated brightness temperatures, and a sensitivity analysis of their effect on algorithm performance is performed. A configuration of GA parameters is then selected to apply the technique to ground-based and space-borne measured brightness temperatures. In the case of simulated data, the best retrieval accuracy is with respect to the mean particle size while the lowest accuracy is with respect to snow depth. Results obtained with field measurements and satellite data give satisfactory results for all considered parameters.