Congress Resources: Papers, posters and presentations

< Return to abstract list

Rain forecasting using Artificial neural network

Author(s): A case study: Synoptic station of Mashhad (Iran)
Congress: 2008
Author(s): Kamran DAVARI, Najmeh KHALILI
2- Assistant Professor in University of Mashhad, P.O.BOX 91775-1163-Mashhad, Iran 3-Engineer

Keyword(s): rain forecasting, artificial neural networks, tree layers feed,forward perceptron network, black box model.
AbstractRain forecasting and rain estimating is one of the main effective climatology parameters in hydrology problems and has an effective role in optimized usage of the water sources. One way to modeling of the rain behavior is Artificial Neural Networks which is categorized in computational intelligence group. In this group, system dynamic and model outputs can be obtained without considering the complicated nonlinear equations. In this thesis, raining dynamic of Mashhad (in Iran) is modeled using statistical monthly raining data of 53 years ago and daily raining data of March for 23 years ago. Finally these data are used for daily and monthly raining forecasting. Through 636 monthly data sample and 713 daily raining data, 550 data sample is selected as training data and the other data is used for validation of the obtained models. The selected neural network which is used in this thesis is a tree layers feed- forward perceptron neural network. This network is used in different cases and tries to give a black box model for rain prediction. the tools and facilities in the MATLAB software has been used for implementing rain prediction ,thereafter ,the results has been evaluated by validation criteria such as Regression equations between actual and estimated data, Correlation coefficient, Root Mean Square Error (RMSE), Variance Accounted For between two signals(VAF), Non Dimensional Error Index(NDEI), Maximum absolute error and Mean Absolute Error(MAE). It has been shown that the results for daily and monthly rain forecasting are rather acceptable and the validation criteria are satisfied .For example correlation coefficient and root mean square error for monthly rain forecasting are 0.92 and 1.00 and for daily rain forecasting are 0.89 and 0.15 . Keywords: rain forecasting, artificial neural networks, tree layers feed-forward perceptron network, black box model.
2011 IWRA - International Water Resources Association office@iwra.org - http://www.iwra.org - Admin