Using time series of SPOT VGT NDVI for crop yield forecasting

In this paper we developed an operational approach using time series Normal Difference Vegetation Index (NDVI) derived from SPOT VGT for crop yield forecasting in China during a five-year span (1988-2002). In order to increase the information content extracted from NDVI profiles, we compose our NDVI profile only in arable area. Thanks to extract the characteristics of the vegetation dynamics, the harmonic analysis of time series algorithm is performed on the NDVI data. We extract our analytical indicators from the time series NDVI profiles, and then remove the trend from the yield series using a linear upward trend function. The difference between the year and last year of residuals and corresponding NDVI indicators were related analysis, we get our predicting indicators by selecting those parameters that had the highest correlation coefficient. At last, we build our yield estimation model based on linear regression analysis between the indicators and residuals just mentioned. These are combined with trend yield, last year residuals yield and satellite based estimation to forecast yields. This year, we used our model in forecasting winter wheat production in China. The resulting productivity could be consistent with other existing data. Our results were well received by the local authorities.