ARTIFICIAL NEURAL NETWORK (ANN) BASED METEOROLOGICAL DROUGHT FORECAST: A CASE STUDY

Drought is actually a normal and recurring climate event. It occurs due to decreasing rainfall spread over one or more seasons. However, increasing temperatures and decreasing precipitation in many parts of the world as a result of global climate change increase the frequency and severity of drought events. Drought is different from other events because it is a natural event that starts very slowly, develops over months or even years, and affects very large areas. Therefore, there is a need to create drought prediction models in order to take the necessary precautions. In this study, artificial neural networks (ANN) method meteorological drought analysis and forecasting was performed. For this purpose, firstly, standard rainfall index (SRI) values of the 30-year monthly precipitation data for the years 1992-2022 obtained from the meteorological station of Elazığ province were found and drought analyzes covering drought duration, amplitude and severity (3, 6, 9 and 12 months) were made. Then, drought forecasts for the coming years were made with the ANN method, using the SRI values obtained from the precipitation data of previous years. 80% of the total data was used for training the ANN model and the remaining 20% for testing. R2 , mean absolute error (MAE) and mean absolute percentage error (MAPE) metrics were used to evaluate prediction performance. By comparing the developed model with the known values of 2023, the results obtained were found to be R2 = 0.969, MAE = 2.06 and MAPE = 9.23. It has been observed that the proposed model has high performance in drought prediction.