Evaluation and prediction of droughts in the west and northwest of Iran using artificial neural network

Drought, as a climate threat, has a significant impact on the environment and, consequently, on humans and other living organisms. Therefore, monitoring and predicting this phenomenon is necessary. This study, to examine and evaluate the drought forecast in the west and north -west of Iran, including Hamedan, Kermanshah, Kurdistan, West Azerbaijan, East Azerbaijan, Ardabil, Zanjan, Qazvin, Ilam, Markazi, Gilan, and Lorestan, have been used multivariate standardized drought index (MSDI) and methods based on artificial intelligence. To predict the values of this index in the study area, monthly rainfall, and soil moisture, as the input, and the calculated amount of MSDI, as output, was applied. The grid data on precipitation and soil moisture for a period of 36 years (1980 -2016) were obtained from the MERRA database. The results of monthly drought analysis based on these data showed that the most severe drought in the study area occurred from March to October and the main focus of this phenomenon are Lorestan provinces, especially Ilam and Kermanshah. The findings were following seasonal and annual maps. According to the MSDI index classification, no severe drought was observed in the study area and the droughts were in the middle class. The results of artificial neural network modeling also showed that the use of artificial neural networks, in general, has an appropriate ability to simulate properly. Among the algorithms used to optimize the artificial neural network, the genetic algorithm has the best performance compared to other methods in predicting drought.