Exploring the potential of crop specific green area index time series to improve yield estimation at regional scale

Crop status, such as the Green Area Index (GAI), can be retrieved from satellite observations by modelling and inverting the radiative transfer within the canopy. Providing such information along the growing season can potentially improve crop growth modelling and yield estimation. However, such approaches have proven difficult to apply on coarse resolution satellite data due to the fragmented land cover in many parts of the World. Advances in operational crop mapping will sooner or later allow the production of crop maps relatively early in the crop growth season, thereby providing an opportunity to sample pixels from medium/coarse spatial resolution data with relatively high cover fraction of a particular crop type to derive crop specific GAI time series. This research explores how to use such time series derived from MODIS to produce indicators of crop yield using two approaches over part of Belgium. The first method consists in looking at metrics of the decreasing part of the GAI curves when senescence occurs. Such metrics, like the position of the inflexion point, have been shown to be significantly correlated to yield. The second approach is to optimize the WOFOST model used in the European Crop Growth Monitoring System (CGMS) based on the GAI time series. Results show that, although the optimized model shows considerably better performance than the model running on the default parameter, the model is sometimes outperformed by the simpler metric approach. In all cases, indicators including remote sensing information provide better estimates that the average yield of previous years.