Search

The aim of this study was to examine the ability of a crop model forced by seasonal weather forecasts to anticipate correctly the agrometeorological variables affecting the growth of winter wheat in France. The method is based on wheat growth simulations over 199 sites in France for the period 1981–2005 driven by seasonal weather forecasts, consisting of 6 month long, 45 member hindcasts starting on 1 February. These simulations were challenging a climatologically based wheat growth forecasting system currently used by ARVALIS − Institut du vegetal. Both simulations were evaluated against a reference dataset. The simulations using the seasonal weather forecast present good skills for the dates of occurrence of the main growth stages: the ‘leaf sheaths strongly erect’ stage (also referenced as Zadoks scale Z30), the ‘head visible’ stage (Zadoks scale Z50) and the ‘kernel hard’ stage (Zadoks scale Z91). The simulations forced by the seasonal weather forecasts have lower bias than those forced by the climatological data: respectively −0.2 against 1.4 days for the Z30 stage, 1.1 against 3.2 days for the Z50 stage, 1.3 against 4.2 days for the Z91 stage. It is shown that the seasonal forecasting system provides more hit rates than the climatologically based method does. Thus, for the different events forecasted here, better predictability is found using the seasonal forecasting system. Despite shortcomings highlighted in this study, it is shown that a method based on seasonal weather forecasts could be applied in an operational context over France, in order to improve the method currently used in the agricultural sector.