Analysis and Forecasting Drought Days Using Artificial Neural Networks ‎Model (Case Study: Station Tehran)

Artificial neural networks as a nonlinear techniques in climate and hydrology studies are important to have. Climate change and the global warming of the climate phenomenon known as persistence of drought followed Number of dry days. In this study, the data of daily rainfall during the period (1976-2008) and artificial neural network in MATLAB software is used to predict the number of dry days Tehran station. Feed-forward type of network used by the algorithm reduces the gradient and Levenberg Marquardt is in the process of teaching and learning. Various structures in the input and hidden layers were tested during the training phase. Finally, a network with 4 inputs and 5 neurons in the hidden layer and 1 neuron in the output layer to best structure (4-5-1) with the highest correlation to predict the optimal answer. The results showed that the aforementioned stations, dry days predicted by the network during the period under review increased compared with that by calculating the probability of dry days during the period (2018-2009) using a Markov chain, the above been approved. The correlation coefficient values ​​predicted dry days without a genetic algorithm combined with 86 percent .After teaching network as genetic algorithm combined with 88 percent that able providing algorithm combined to network result passable showing