A comparative analysis of analytical hierarchy process and machine learning techniques to determine the fractional importance of various moisture sources for Iran’s precipitation

Iran is a semi-arid and arid country in the Middle East region which has faced a water shortage crisis from early times. Hence, various elements of the hydrological cycle should be studied deeply and accurately in this country. Among the various water cycle elements, investigations on the main moisture providing sources of precipitation (as the most important input of water cycle) is so scarce. Moisture for precipitation in Iran is mainly provided by the Red Sea, the Caspian Sea, and the Persian Gulf during the dry period (May to October), while the Arabian Sea, the Persian Gulf, and the Mediterranean Sea provide large amounts of moisture during the wet period (November to April) obtained by FLEXPART v9.0 model. Understanding the role and importance of each moisture source influencing Iran’s precipitation has great application in climatological models to predict droughts across the country. Nowadays, machine learning techniques have improved significantly for simulation and determining the correlation between various parameters in large and complicated data sets due to their high accuracy. In this study, the role and fractional importance of various water bodies providing moisture for Iran’s precipitation has been determined for 35 years (1981-2015) based on the (E-P) values obtained by the FLEXPART model and various machine learning models (artificial neural network (ANN), deep neural network (DNN), decision tree and random forest) and analytical hierarchy process (AHP) and Fuzzy AHP models. The results show that in the wet period, the Arabian Sea in all the developed machine learning and AHP models play the dominant role and its fractional importance varies from 52% to 28.2% of the total importance in decision tree and AHP models, respectively. During the dry period, the Arabian Sea, with 57.8 % and 24.3% of the total importance in AHP and DNN models, respectively, play a dominant role. However, the Mediterranean Sea with 32.83% based on the random forest model, the Black Sea with 24.3% based on the ANN model, and the Indian Ocean with 31.19% of the total importance based on the Fuzzy AHP model’s influence Iran’s precipitation during the dry period.