Wheat growth simulation and yield prediction with seasonal forecasts and a numerical model

Wheat is a major winter crop in northern Italy. Italian agricultural markets and government agencies would undoubtedly benefit from the early availability of wheat yield forecasts at the regional and national scales as useful support in decision making. In this study we tested the skill of seasonal weather forecasts, in combination with observed weather data, as input to a crop model working in water limited conditions. The observations were used to simulate wheat growth from sowing up to 2 months before harvest, while seasonal forecasts were used afterwards to predict final yields. Observations included climatic variables and water table levels from a location in the Po river plain (Bologna, Italy), while seasonal forecasts came from the Demeter EU project and consisted of a dataset of downscaled multi-model ensemble hindcasts.The Criteria/Wofost simulation model used for this work includes a new numerical scheme for the soil water balance (Criteria) and incorporates a modified version of the Wofost crop growth model. Median wheat yield forecasts were compared with field data collected at the experimental farm of Bologna University during 1977–1987. Forecast yields showed satisfactory agreement with observed ones (MBE 816 kg ha−1, RMSE 1185 kg ha−1, R2 0.65**). In our view, with this result there is a good prospect for extending the proposed methodology to the regional and national scale for the production of operational seasonal forecasts of agricultural yields.