predicting dryland wheat yield from meteorological data using expert system, khorasan province, iran

khorasan province is one of the most important provinces of iran, especially as regards agricultural products. the prediction of crop yield with available data has important effects on socio-economic and political decisions at the regional scale. this study shows the ability of artificial neural network (ann) technology and adaptive neuro-fuzzy inference systems (anfis) for the prediction of dryland wheat (triticum aestivum) yield, based on the available daily weather and yearly agricultural data. the study area is located in khorasan province, north-east of iran which has different climate zones. evapotranspiration, temperature (max, min, and dew temperature), precipitation, net radiation, and daily average relative humidity for twenty-two years at nine synoptic stations were the weather data used. the potential of ann and multi-layered preceptron (mlp) methods were examined to predict wheat yield. anfis and mlp models were compared by statistical test indices. based on these results, anfis model consistently produced more accurate statistical indices (r2= 0.67, rmse= 151.9 kg ha-1, mae= 130.7 kg ha-1), when temperature (max, min, and dew temperature) data were used as independent variables for prediction of dryland wheat yield.