Operational yield forecast using AVHRR NDVI data: Reduction of environmental and inter-annual variability

Operational millet yield forecast has been conducted in Senegal using AVHRR data. This was possible by establishing a linear relation between yield collected in the field and NDVI integrated during the reproductive period of millet growth. A bio-physical frame was adapted in order to understand and reduce inter-annual and environmental variability. This was done by limiting potential variation of plant physiological parameters by accounting for vegetation complexity and differences in productivity: (1) Subtracting a pre-growing season NDVI reference level from the NDVI integral improved significantly the level of explained yield variance and was interpreted to reduce the influence of non-crop vegetation within the agricultural domain. (2) Soil and vegetation maps were used to identify areas of homogeneous environmental production conditions. Using this information along with the NDVI integral, accorded a millet yield regression model with a correlation coefficient of r(2) = 0 . 72 and a standard error of estimate of 190 kg ha(-1). Yield data could be available one month before the harvest.