PREDICTION AND MAXIMIZATION OF WHEAT GRAIN YIELD IN SEMIARID ENVIRONMENT BY USING ARTIFICIAL NEURAL NETWORKS.

At field level, prediction of crop yield and determination of appropriate fertilizer application rate based on soil and landscape attributes often require for better management practices finally maximizing crop yield. The present study was designed to evaluate the potential of Artificial Neural Networks (ANNs) model to predict wheat grain yield and to determine the urea fertilizer application rate for maximizing wheat grain yield. The urea fertilizer, %sand, %silt, %clay, elevation, soil nitrogen (N), soil electrical conductivity (EC), soil phosphorus (P) and soil pH were used as input parameters and wheat grain yield was used as output. The ANNs model was trained using 124 data sets collected during growing seasons of 2008-09 to 2010-11 and evaluated using randomly selected 20 data sets of 2010-11 and 48 data sets of201 l-12. The results showed that ANNs model has the potential to predict wheat grain yield under semiarid conditions, as the mean absolute error (MAE) of 6.50% was found for training and 9.48% was observed for testing the model. To determine urea fertilizer application rate for maximizing wheat grain yield, the trained model was run for 14 urea fertilizer levels ranging from O to 400 kgurea ha·1 . It was examined that 210 kg-ureaha·1 produced maximum grain wheat yield (4200 kg ha^-1) and further increase of urea fertilizer rates resulted in decrease of wheat grain yield. These results also showed that ANN s model is a useful tool to estimate the wheat grain yield response to soil and landscape attributes and to determine the optimum urea fertilizer level for maximizing the wheat grain yield in semi-arid conditions of Faisalabad. [ABSTRACT FROM AUTHOR]