Snow depth estimating as one of the consequences of climate change using the combined least squares model approach of support vector machine and genetic algorithm

One of the direct and evident impacts of climate change is the change in the amount of snowfall in different geographical areas. It is worth mentioning that snowfall in mountain basins is always taken into account as the most important source of water supply in dry seasons. Due to some restrictions, data collection, particularly on a large scale, is difficult and at sometimes impossible. Consequently, using indirect methods is recommended. In this study, the efficiency of least squares support vector machine in modeling the depth of snow and the impact of feature reduction with genetic algorithms model in Chelgerd , Iran was investigated. At first, with using the Hypercube model, the locations of 100 points were specified, and the data of snow depth at certain points as well as other 195 points were randomly collected. Afterwards, with using DEM,, 25 Geomorphomety parameters were extracted, and these parameters with six bands, eight Landsat satellite images and the difference index of normalized snow were chosen as the inputs of models. In this study, genetic algorithm is used to increase the speed of support vector machine which is considered as a classifier and make it ready. Also, genetic algorithm is utilized to choose the variants having the most coherence with the snow depth. Since the reduction of ineffective features can increase the accuracy of learning, genetic algorithm was used in this study for the optimization process. The results showed that the least squares method of the S.V.M with the coefficient of determination of 0.36 and the SMSE of 17.8 has modeled the snow depth. However, the genetic algorithm by selecting the effective features was able to model snow depth changes better with a determination coefficient of 0.95 and RMSE equal to 3.97 cm which is more accurate than using all features.