Updating vegetation information in a land surface model

Vegetation plays a fundamental role in modulating the exchange of water, energy, and carbon fluxes between the land and the atmosphere. These exchanges are modelled with Land Surface Models (LSMs) which are part of numerical weather prediction systems to support the performance of weather forecasts. However, most current LSMs only utilise observed vegetation information in the form of mean seasonal cycles. The potential benefits of additionally including information about shorter-term vegetation anomalies and inter-annual variability are understudied.

In this study, we update vegetation information in the HTESSEL (Hydrology Tiled ECMWF Scheme for Surface Exchanges over Land) model and investigate the resulting effects on the performance of simulated shallow and deep soil moisture as well as latent heat flux. The updated information includes an interactive observation-based leaf area index from Sentinel-3 and THEA GEOV2, and a land use/land cover map from ESA-CCI. The resulting simulations of soil moisture and latent heat flux are validated against global gridded observation-based datasets.

Results show that the updated land surface information deteriorates the overall model performance for both latent heat flux and soil moisture in most regions across the globe. In a second step, we re-calibrate soil and vegetation-related parameters at each grid cell in order to adjust them to the new vegetation information. This leads to improved model performance and illustrates the benefits of updated vegetation information. Morover, we attribute the spatial variations of parameter perturbations resulting from the re-calibration to multiple land surface and climate characteristics. This highlights potential venues in model development to take static ecological and hydroclimatological information into greater consideration.

Furthermore we compare the performances of local model calibration - performed for each grid cell individually - and global model calibration considering a single parameter set for all grid cells globally. We analyse the agreement of parameter calibrations obtained for shallow and deep soil moisture as well as latent heat flux.

In summary, our results highlight that Earth-observation products of vegetation dynamics and land cover changes can improve land surface model performances, which in turn can contribute to more accurate weather forecasts.