Fusing earth observations and modelling to assess the carbon budget of managed grasslands

Abstract:

Managed grasslands are extensive ecosystems that contain climatically significant amounts of carbon (C). Understanding and quantifying their C dynamics is complicated by the spatial and temporal variability of their management (cutting and grazing). This study presents how model-data fusion (MDF) can estimate C dynamics in variably managed grasslands by relying on earth observation (EO) data from newer sensors with resolutions that resolve fields. MDF explores the synergies between observational data and modelling results in order to improve model forecasting skill, characterise model uncertainty and understand underlying ecosystem processes. A parsimonious process-based model of C biogeochemistry in managed grassland was developed and integrated in a probabilistic MDF framework. The MDF algorithm was tested at three sites in England using leaf area index (LAI) data, retrieved from Sentinel-2 images, as observational constraints. The model’s forecasting skill was assessed against in-situ data on harvest dates and yields, and grazed biomass. The results showed that MDF was able to estimate with good accuracy and precision these key management factors. The MDF algorithm was then implemented at 3500 grasslands of Great Britain. The selection of these fields was made on the basis of their representativeness of grassland productivity and management at local-scale (25km²). The MDF-predicted livestock density mirrored that in the most recent livestock census. A key result of this computational experiment is that C partitioning to above and below-ground organs/C-pools is controlled by the ratio of grazed-to-harvested biomass. The C allocation pattern of a grassland determines the amount of C transferred into the soil and, therefore, its C sequestration potential and net C balance. Considering the increasing volume, resolution and availability of airborne and satellite EO data MDF can offer an improved assessment of C dynamics in managed grasslands across space and time.