A Neural Network Based Model for Rainfed Wheat Yield Forecasting Using Climatic Data

In this research we used ANN to predict yield of rain fed wheat. To do this at first we made the input matrix of model included daily rainfall, maximum ,minimum and mean temperature, Evapotranspiration, sunshine hours, number of rainy days, relative humidity, number of days with thermal and cooling stresses(were extracted by assessing daily temperature and threshold temperature of 30 and 0 degree centigrade for thermal and cooling stresses respectively).All of these input meteorological parameters were extracted for all studied meteorological stations and was defined as Input vector in ANN model.The meteorological data was obtained during the different phenological stages of wheat all of them reported by 10 Meteorological stations and by one agricultural meteorological station.The output matrix included wheat yielding reports provided from ministiry of agriculture data bank during the 1999-2005 period. According to the output analysis of the model it can be shown that the most important climatic factor determining wheat yield is the amount of rainfall because omitting this factor from the input matrix, the amount of the models RMSE will increase .The accuracy of model was obtained by d index which is about 77%.