From the field to the region - monitoring pre-Alpine grassland characteristics at different spatial scales

Grasslands in their various forms of appearance characterize the pre-Alpine landscape. Despite the economic value and significant role of plants in grassland carbon and nitrogen cycling, spatially explicit information on grassland biomass are rarely available. This study aims to develop routines to monitor grassland traits at different spatial scales. Field sampling campaigns were conducted in April 2018 and at multiple times during the growing seasons of 2019 and 2020 to collect in-situ data of dry aboveground biomass (DM) from differently managed grasslands. The campaigns were partially accompanied by unmanned aircraft system (UAS) flights to acquire very high resolution multispectral imagery at the field-scale. This data was complemented by time series of Sentinel-2 (S2) imagery to address the regional scale. In a first step, we tested different statistical modelling approaches and UAS input datasets to estimate DM for the single-date acquisition in 2018. Promising results were obtained by the machine learning algorithms random forest and gradient boosting machines (cross-validated R 2 of best model = 0.71). A first multi-temporal DM model for S2 imagery was developed and used to create regional maps. Next, we will adapt the algorithms to multi-temporal UAS data and compare the results across different scales.