Application of artificial neural networks in simulating the monthly rainfall of Khalkhal station

Mohammad Reza Nikpour¹

1) Associate Professor, Faculty of Agriculture and Natural Resources, University of Mohaghegh Ardabili

IOI Code : AZAA-AZDHED
Publication : 1ST International Congress of Nature-based Ecological Restoration (Emphasizing the Conservation of Bactrian Camels)(neruma.ir)
Abstract :
    Accurate forecasting of monthly rainfall not only plays a significant role in water resources management, but also can significantly improve the accuracy of flood warning systems. In recent decades, the improvement of processor power and the development of intelligent models have led researchers to face new fields in the field of improving simulation methods of hydrological variables. In this research, the artificial neural networks (ANN) model was used to simulate the monthly rainfall of Ardabil synoptic station. The data used in the current research were the monthly rainfall values for the statistical period of 1996-2016. For the short-term simulation of monthly rainfall (one month later), five different patterns were defined based on rainfall delays. To evaluate the results of the model, coefficient of determination (R2) and least root mean square error (RMSE) indicators were used. The results of the research indicated the relatively poor performance of the GEP model in all scenarios. So that the values of R2 and RMSE for the best scenario in the verification stage were 0.34 and 16.99 mm, respectively. It is worth noting that the best performance of the model was related to the last model (11 months lag-times).To improve the results of the current research, it is suggested to use wavelet transform for data preprocessing.

Keywords : khalkhal station monthly rainfall ann model wavelet transform