|Paper title||Upscaling multitemporal UAV data to Sentinel-2 time series enables grassland biomass monitoring|
|Form of presentation||Poster|
Grasslands belong to the most diverse land systems of Europe, covering gradients from intensively managed annual grasslands to natural meadows without management. As detailed information on grassland use intensity in Europe is sparse, spatiotemporally explicit information on the vegetation and its dynamics are needed to develop sustainable management pathways for grasslands. On the one hand, Unmanned Aerial Vehicles (UAV) have great potential for providing high-resolution information from field- to farm-scale on key phenological dates. Sentinel-2 data, on the other hand, allow for frequent, continuous global monitoring delivering data at 10m spatial resolution. Therefore, nested approaches combining UAV with Sentinel-2 offer yet unexplored potential for monitoring grasslands.
Beyond commonly used vegetation indices, time series of biophysical parameters such as biomass, leaf area index, or vegetation height provide physical measures of grassland productivity or vegetation structure for assessing grassland resources over time. Timely information of such parameters directly supports land management decisions of farmers and serves as a basis for designing and evaluating policies. UAV and Sentinel-2 have high suitability for estimating biophysical parameters.
In this study, we therefore present an upscaling approach combining UAV and Sentinel-2 data for improved grassland monitoring based on time series of aboveground biomass. We investigate the potential of combining UAV and Sentinel-2 data in contrasting grasslands including an extensively grazed upland pasture and intensively managed lowland meadows. We use a two-step modelling approach incorporating ground-, UAV-, and satellite-scales. We first derive maps of biomass information from UAV images and ground-based data covering the complete phenological development of grasslands from April to September. Subsequently, we use the high-resolution UAV-based maps from multiple dates in a global machine learning model to estimate intra-annual biomass time series from Sentinel-2 data. First results show that UAV-based maps capture fine-scale spatial patterns of biomass accumulation and removal before and after grazing periods. The Sentinel-2-based time series reproduce vegetation dynamics related to management periods in the two contrasting study areas. Our study demonstrates the potential of combining high-resolution UAV and Sentinel-2 data for establishing monitoring systems for grassland resources. More research is needed to enable multi-scale monitoring of biophysical quantities across different grassland regions in Europe.