Short-term and Long-term Drought Forecasts in Iraq Using Neural Networks and GIS

Abstract

Drought is a dangerous phenomenon that affects the general life of the environment. Iraq is one of the countries that is facing drought periodically, especially in the last decades due to the great weather changes in the world including global warming, which resulted in less rainfall below normal levels. Therefore, it must be thought of drought forecasting because it is an important role in the planning and management of the water resources in Iraq. In this Study, Recurrent neural networks (RNN) were used as representing of Artificial Neural Networks (ANNs), which is a non-linear kind of ANNs where the output from it will feedback again as input for the next step. This type of neural network can simulate weather conditions with high precision such as rain, wind, earthquake, drought, and temperature. The model used to forecast droughts is the standardized precipitation index (SPI) series as a drought index in Iraq. The two-time scale which is used in this study, which is SPI 6 which represents short term drought and SPI 24 which represent long term drought. RNN was used to make forecasts for the SPI for the period 2020-2030. The assessment of the work and efficiency of RNN was regressing by (R), mean square error (MSE), and root mean square error (RMSE). Twenty-Four stations were selected to represent all study area (Iraq). Geographic information system (GIS) was used with the aid of Inverse distance weighted (IDW) to represent the forecasted drought for April month from years (2025 and 2030). The results showed that the study region (Iraq) suffered from varied drought levels in different periods ranging from mild to extreme drought, also the study showed improvement by decreasing the drought situation for period 2020-3030, which must be invested well.