Estimating grassland biomass from Sentinel 2 - a study on model transferability

Satellite-supported information services support farmers in a variety of ways to make management decisions. However, most satellite-based information services focus on arable land and are not yet available for grassland. An important aspect in the development of such information services is the correct derivation of green leaf area and aboveground dry biomass from satellite data. This derivation forms the technical basis for all subsequent analyses and management recommendations calculated by satellite-based information services. In a previous study, green leaf area and dry biomass were already successfully derived for a grassland site using two alternative modelling approaches – an empirical model based on random forest decision trees and the radiative transfer model SLC. For operational use in information services, transferability of the methods used in different areas is necessary. In this study we calculated DBM with both models and compared the results to DBM data provided by a combine harvester. The spatial heterogeneity within the fields was well captured by both models. Both models calculated significantly lower DBM than was measured with the combine harvester.