Developing a Drought Forecasting Model; Case study Abyek City in Ghazvin Province

Optimum utilization of water resources in the country requires improving the accuracy of forecasting and estimation time of drought. One of the most important issues in monitoring and predicting drought is choosing an appropriate index for the area. In the present study¡ besides determining two indices of SPI and CZI in two scales of short-term and medium-term using the precipitation of two rain gauge stations in a period of 43 years (1972-2015) located in Abiyek City¡ the drought forecasting was performed using the Direct Multi-Step Neural Network in six time ahead. The Kappa- Cohen statistic used in order to review the consistency of quality classes between the predicted and observed values. The results of using this network in this study indicated an acceptable performance and capability of this network to estimate the drought using the two scales of SPI and CZI and predict some earlier steps of weather drought. Based on results of the weighted Kappa statistic showed that with increasing the prediction step¡ the similarity between the prediction amounts and the observed amounts in quality classes of drought decrease in two indices of SPI and CZI. So¡ by increasing the time scales (from 3 to 9 month)¡ the similarity increases. The results of the prediction with the two mentioned indices and in different scales showed that Ziaran station seems more appropriate because it is located in the center of the area. Therefore¡ choosing an appropriate station in prediction issues helps improve the models significantly. Finally¡ this research can be useful in predicting the time of drought at least for the next six months¡ and help water planning and water resources managers in macro level in the country.