Assimilation of Sentinel-2 Leaf Area Index Data into a Physically-Based Crop Growth Model for Yield Estimation

The work is based on a previously published study with the aim to further analyse the results obtained. Remote sensing data, crop growth models, and optimization routines constitute a toolset that can be used together to map crop yield over large areas when access to field data is limited. In this study, Leaf Area Index (LAI) data from the Copernicus Sentinel-2 satellite were combined with the Environmental Policy Integrated Climate (EPIC) model to estimate crop yield. The experiment was implemented for a winter wheat crop during two growing seasons (2016 and 2017) under four different fertilization management strategies. A number of field measurements were conducted spanning from LAI to biomass and crop yields.
LAI showed a good correlation between the Sentinel-2 estimates and the ground measurements using non-destructive method. Better RMSE and RRMSE were obtained in 2017 compared to 2016 (RMSE = 0.44 vs. 0.46) (RRMSE = 17% vs. 19%). In 2016 year, a slightly lower R² value was found compared to 2017 (R² = 0.72 vs. 0.89). A correlating fit between satellite LAI curves and EPIC modelled LAI curves was also observed. The work shows that the assimilation of remote sensing data into the crop growth model can help to overtake some structural problems of the model.  The assimilation framework has to be tested under different environmental conditions before being applied on a larger scale with limited field data.