Estimation of Forage Yield Using the Second Derivative of Normalized Difference Vegetation Index Time Series From Sentinel‐2

Accurate estimation of forage yield is essential, as forage plays an indirect yet vital role in the global food supply. Various methods were developed to obtain quantitative forage yield estimates from remotely sensed data. This study's main goal was to estimate forage yield from Sentinel‐2 time series imagery from 2021 and 2022. The novelty of the approach lies in adapting a method originally derived from the AquaCrop model by replacing the crop transpiration coefficient with interpolated values of the normalised difference vegetation index (NDVI). This method, previously implemented with drone imagery, enables rapid, large‐scale estimation of forage production. The experiment was conducted across seven agricultural fields in the province of Salamanca, Spain. NDVI values were extracted from Sentinel‐2A and Sentinel‐2B bottom‐of‐atmosphere reflectance (BOA) at different dates that span the growing period. The sum of NDVI values between selected thresholds was calculated and multiplied by a standard water productivity coefficient. Final yield estimates were validated against direct forage yield measurements collected by farmers at the seven sites. Results demonstrate good performance of the Sentinel‐2 time series in estimating forage yield, with correlation coefficients exceeding 0.75 and errors below 30% when compared to observed yields. This method offers a fast, scalable, and effective alternative to traditional yield measurement techniques.