Global vegetation models (GVMs) simulate CO2, water and energy fluxes at large scales, typically no smaller than 10×10km. GVM simulations are thus expected to simulate the average functioning, but not the local variability. The two main limiting factors in refining this scale are (1) the scale at which the pedo-climatic inputs - temperature, precipitation, soil water reserve, etc. - are available to drive models and (2) the lack of geospatial information on the vegetation type and the age of forest stands. This study assesses how remotely sensed biomass or stand height could help the new generation of GVMs, which explicitly represent forest age structure and management, to better simulate this local variability. For the ORCHIDEE-FM model, we find that a simple assimilation of biomass or height brings down the root mean square error (RMSE) of some simulated carbon fluxes by 30-50%. Current error levels of remote sensing estimates do not impact this improvement for large gross fluxes (e.g. terrestrial ecosystem respiration), but they reduce the improvement of simulated net ecosystem productivity, adding 13.5-21% of RMSE to assimilations using the in situ estimates. The data assimilation under study is more effective to improve the simulation of respiration than the simulation of photosynthesis. The assimilation of height or biomass in ORCHIDEE-FM enables the correct retrieval of variables that are more difficult to measure over large areas, such as stand age. A combined assimilation of biomass and net ecosystem productivity could possibly enable the new generation of GVMs to retrieve other variables that are seldom measured, such as soil carbon content.
- Global vegetation model
- Remote sensing