Prediction of crop yield mainly strategic plants such as, wheat, corn and rice has since long been an interesting research area to agrometeorologists, as it is important in national and international economic programming. The main purpose of such studies is to estimate the crop production a few days or months before harvesting, using meteorological data. Recently, the application of Artificial Neural Networks (ANNs) has developed into a powerful tool that can compute most complicated equations and numerical analyses to the best approximation. The goal of this study was to apply the ANNs to predict dry farming wheat yield. According to the available data and information from different areas in Iran, this research was accomplished using Sararood station data in Kermanshah Province which has the most complete homogeneous statistics. In this study, the results of climatology for the period (1990-99) for each of the eleven phenological stages of wheat including sowing, germination, emergence, third leaves, tillering, stem formation, heading, flowering, milk maturity, wax maturity, full maturity and also eleven meteorological factors including: mean daily minimum temperature, extreme daily minimum temperature, mean daily maximum temperature, extreme daily maximum temperature, total daily rainfall, number of rainfall days, sum of sun hours, mean daily wind speed, extreme daily wind speed, mean daily relative humidity and sum of water requirement were collected separately for each farming year and arranged in two matrices: A matrix whose rows are repetitions of the statistical years (i) at each phenological stages of wheat (j) and the columns are meteorological factors (k). A matrix whose rows form each of the statistical years (i) and the columns are meteorological factors (k) at each phenological stage (j). In fact, statistical years (i), phonological stages (j) and meteorological factors (k) are the basic elements of 3-D matrix (M ijk) arranged as above and the best answer was derived only from the second matrix.

Finally, different networks were made for each stage and the optimum values of network parameters were obtained by trial and error. It should be reminded that two of the eight- year farming were randomly excluded from network training computations and that the comparison of the estimated data with the real data for these two years were used to test the accuracy of the models. The model that obtained has the following capabilities: 1) Prediction of wheat yield with maximum errors of 45-60 kg/ha at least two months before full maturity stage (end-of-stem formation stage). 2) Determination of the sensitivity of each phenological stage with respect to meteorological factors. 3) Determination of the priority order and importance of each meteorological factor effective in plant growth and crop yield.