In crop yield prediction, the unobserved future weather remains the key point of predictions. Since weather forecasts are limited in time, a large amount of information may come from the analysis of past weather data. Mean data over the past years and stochastically generated data are two possible ways to compensate the lack of future data. This research aims to demonstrate that it is possible to predict plant growth in advance using mean climatic data over the last years in combination with actual data.

Field experiments were carried out to measure the crop response (Triticum aestivum L.) under temperate climate (Belgium). A 30-years weather database provided by a meteorological station located 4km from the field allows calculating the daily mean climatic data. For each year of the database, a matrix ensemble was built, in which real data were replaced by the averages values, at a 10-day rate. After calibration, the STICS crop model (Inra-France) was run on the whole climatic matrix ensemble.

A novel form of results presentation allows studying the yields distributions along the season and over the years. This led (i) to highlight the most sensitive periods inducing severe stress(es) on the crop, (ii) to quantify the immediate yield losses due the consecutive stresses, and (iii) to significantly (a=10%) estimate the date of final yield (±10%) prediction.

The study shows that the highest stresses occur since the last-leave stage (Julian day 120), after which two-thirds of the total biomass remains to be produced. The yields could be predicted around Julian day 200 (30 days before harvest). The final yields could range between -45% and +10% of the average predictions (12.2 tons/ha).

The methodology appears to be a powerful diagnostic tool of the potential yield of a defined crop under a set of real climatic conditions, at regional scale. It also highlights the strong interaction between plant growth and climatic conditions.