Monthly precipitation prediction of Ardabil province using ANN and WANN models

Precipitation is one of the most important hydrological events and its prediction can be used as a practical tool for optimum utilization and management of water resources. In the present study, artificial neural network (ANN) and wavelet-artificial neural network (WANN) were used for monthly precipitation perdiction at selected synoptic stations in Ardabil province, including Ardabil, Khalkhal, Meshginshahr and Parsabad during the 225 months for the years 1996-2016. For the short-term forecast of monthly precipitation (one month later), different scenarios were defined based on precipitation delays. Results indicated that the WANN model with the highest determination coefficient (R2) and minimum root mean square error (RMSE) was acceptable for all stations. The values of R2 and RMSE for Ardabil station were equal to 0.88 and 7.13 mm, for Khalkhal station were equal to 0.91 and 6.36 mm, for Meshginshahr station were equal to 0.92 and 6.97 mm and for Parsabad station were equal to 0.86 and 8.51 mm, respectively. In all stations, utilization of  the superior model (WANN model) with the combination scenarios i.e. rainfall delays, the minimum and maximum temperature improved the results of the model, but on the other hand, increased the computational cost of the model. Also, in all stations, the addition of relative humidity and wind speed as input variables somewhat reduced the performance of the model. The general results of this study showed that the WANN model with appropriate rainfall delays on a monthly scale can be utilized to predict monthly precipitation of selected stations in Ardabil province, including Ardabil, Khalkhal, Meshginshahr and Parsabad with acceptable accuracy.