Crop Yield Forecast for France Based on the CNDVI Technique

The objective of the research presented here is to obtain crop yield forecasts basing on the information of NOAA- AVHRR/NDVI and CORINE land cover data. The methodology described in Genovese et al. (2001) consists of extracting yield indicators from CNDVI (CORINE-NDVI) time series at a regional scale. In Genovese et al. (2001), a preliminary study on Spain for a four year span (1995-1998) has shown that indicators extracted from the CNDVI profiles can be more closely related to crop yield performances than indicators based on simple NDVI profiles. To prove the validity of this approach, a more complete experiment was realised on France for the same period. Linear regressions were calculated using regional CNDVI-based indicators versus regional wheat yield data (EUROSTAT NEW CRONOS database). A French national wheat yield forecast was then derived by aggregation of regional results. The goodness of the results confirms the advantages of such approach. The combination of a CNDVI-based indicator with the linear trend observed on yields between 1975 and 1997 led to very good regression criteria (coefficient of determination higher than 86%) and allowed a satisfying prediction of wheat yields.