Evaluation of the performance of the support-wavelet vector machine hybrid model in predicting dust storms (Case study: Sistan and Baluchestan province)

The increasing incidence of dust storms indicates the dominance of desert ecosystems in each region. Therefore, in order to properly control and manage dust storms, it is necessary to be aware of the temporal-spatial changes of this phenomenon and the need to predict and model it. In this study, in order to predict the variable frequency of days with dust storm (FDSD), the results of two hybrid methods under the titles of support vector-wavelet (W-SVM) and support vector-artificial plant algorithm (AF-) (SVM) was compared with the individual support vector machine (SVM) model. For this purpose, hourly dust data and codes of the World Meteorological Organization were used on a quarterly scale with a statistical period of 40 years (2018-1980) in five selected synoptic stations of Sistan and Baluchestan province. Explanation coefficient, root mean square error, mean absolute error value and task strain coefficient were used to evaluate and compare the models. The results of goodness-of-fit indices in the training and testing phase showed that the hybrid structures used provide acceptable results in modeling the FDSD index. Support-wavelet car vector hybrid model with correlation coefficient (R2 = 0.911-0.984), root mean square error (RMSE=0.314-0.397), mean absolute error value (MAE=0.236-0.335) And clutch saturation coefficient (NS = 0.927-0.965), had better performance than other models used in predicting the FDSD index. The results of this study can be effective in managing the consequences of dust storms and desertification programs in the study areas.