Improving Forest Realism in Earth System Models through Satellite Observations

Forests play a critical role in land-atmosphere dynamics, significantly influencing soil-water-climate interactions. A realistic and accurate representation of forest carbon pools in land surface models is essential to understand, monitor, and predict droughts, wildfires, and weather and climate dynamics. Satellites offer detailed and global observations of forest characteristics, such as 10-250m resolution biweekly leaf area index (LAI) from Sentinel and MODIS and 30m resolution canopy height from GEDI-Landsat data products. By integrating these satellite observations with vegetation allometric relationships, we can reconstruct forest carbon biomass pools across roots, trunks, and leaves since the 2000s. These approaches have been pivotal in mapping and reconstructing global above-ground carbon stock at fine spatial scales (10-250m resolution). However, integrating these detailed satellite observations into predictive Earth System Models (ESMs) remains challenging due to the complexity of dynamic vegetation models and the spatiotemporal mismatch between satellite data and the grid size of ESMs.

To bridge this gap and enable a detailed and realistic representation of forest dynamics in ESMs, we introduce an approach to integrate MODIS LAI and GEDI-Landsat canopy height data through the assimilation of carbon biomass pools (roots, trunk, and leaves) into the vegetation dynamics component of the NOAA-GFDL Land Model version 4 (LM4). Leveraging the HydroBlocks sub-grid tiling scheme and LM4 allometric relationships for LAI and canopy height, we assimilate monthly biomass pools at an effective 250m resolution across the continental United States. We assess how improving forest representation through the assimilation of biomass pools impacts transpiration, canopy and soil evaporation, soil moisture, and runoff. By improving the spatiotemporal accuracy of forests-soil-water dynamics at local scales, we can now better map and quantify the role of forests driving ecohydrological hotspots. Such advancements can contribute to improved capabilities to model and predict droughts, wildfires, and deforestation impacts at the spatial scales closer to the scales where conservation and mitigation strategies are implemented (~100s meters).