RETRIEVAL OF SOIL MOISTURE AND SURFACE ROUGHNESS FROM RADAR REMOTE SENSING AND GENETIC ALGORITHM

Abstract

Genetic Algorithms are fast becoming used more often in many electromagnetic applications where a function must be optimized. This is due to their ease of programming, inherent flexibility to adapt to most problems coupled with their simplicity. They are particularly useful when there are many variables to be optimized. Depending on the problem, a global maximum is not always found due to the inherent discretisation of the problem that forms one of the distinct properties of this technique. If a true global maximum is required, then a hybrid scheme can be employed using both traditional techniques and a Genetic Algorithm. The use of these algorithms is particularly useful when the search space is large. It is in these instances where these algorithms flourish and should be used in preference to more traditional techniques. Two key parameters to affect microwave scattering from land surface are the surface roughness and soil wetness. In this thesis an algorithm based on genetic approach is developed for multiparameter i.e. (land surface roughness and soil wetness) retrieval from radar remote sensing data. The bistatic scatterometer system is used for the determination of the angular scattering coefficients for different soil moistures and roughnesses. The experimental data obtained is then used as an input parameter to the GA code. The parameter of wetness and roughness are encoded into genes. Genes are the constituent of chromosomes, which undergoes optimal selection based on the natural evolutionary process in the genetic algorithm. The semi-empirical model is used for the computation of the cost function. The retrieved values of roughnessess and moisture contents with genetic algorithm are in good agreement with observed values of roughnessess and moisture contents. The developed algorithm based on GA has been applied to ERS- I data of Haridwar site (Lat and Long). Satisfactory results are obtained in that case also. This work presents an example of the genetic algorithm for the application of multiparameter retrieval in remote sensing.