Regional crop yield assessment with Sentinel-2 and MODIS for the Netherlands

Regional prediction and forecasting of crop yield is important for analyzing food security and mitigating risks of food price volatility. Many countries have methods to come up with annual agricultural statistics (on cropped area, production, and yield), which in most cases rely on in-situ observations. However, such information is often only available at coarse spatial aggregation levels and is not very timely. Crop models fed by gridded meteorological data, but also remotely-sensed indicators of crop performance can be a useful means for estimating crop yields over large areas. For remote sensing, particularly vegetation indices such as the NDVI are often used, which can be translated into anomalies. To capture anomalies effectively, longer time series with coarse spatial resolution, such as from MODIS, are normally used. However, when fields are small, MODIS may not be effective in getting a representative VI profile for a specific crop in a region, due to mixed pixel effects. The main hypothesis to be tested in this study is whether crop-specific Sentinel-2 VI profiles allow for more accurate prediction on crop yield variability in the Netherlands (2016-2023) as compared to MODIS. Annual field boundaries with crop information for the Netherlands (Basisregistratie Gewaspercelen) are available to focus the analysis on specific crops. The suggestion is to include at least maize and wheat, and possibly extend to sugar beet and potato (i.e. the most important crops in the Netherlands).