A Drought Analysis Of Sivas Using The Standardized Precipitation Index (SPI) Method And Drought Estimation With The Artificial Neural Networks
Drought is a natural disaster which has various social and environmental impacts. Given that the exact time, period and intensity of drought are unforeseeable, probabilistic and statistical methods are used while analyzing a drought in a region. While the forecast of a drought in the future is very difficult, the Artificial Neural Networks (ANN) which can be used successfully in the behavior of non-linear systems and the use of which increased in the water resources engineering in recent years can be utilized. In this study, drought analysis was conducted primarily by Standard Precipitation Index (SPI) method for Sivas Province which has a semi-arid climate. Droughts occurring in the region were examined by finding SPI values of monthly precipitation data of 65 years from 1950 and 2014 obtained from Sivas Weather Station and determining drought characteristics (drought duration, amplitude and intensity) in different time periods of the station and their relevance to one another. In the second part of the study, the forecast of the drought in the upcoming years was made by the Artificial Neural Networks (ANN) method while using monthly SPI values. While there are similar studies in the literature, in this study, different network models were developed and their trials were made and the best network model has been tried to be determined as a result of those trials. In order to see how much data can be estimated for upcoming years by the created network model, data of 2007 chosen for being dry year were used as the test data. So, when we use the precipitation data of the current month as input by making the necessary conversions, it has been seen that we can make predictions about the data of the next month.
Key words- Artificial Neural Networks (ANN), Drought, Precipitation, Standard Precipitation Index (SPI).