Despite the need to understand terrestrial biospheric carbon fluxes to account for carbon cycle feedbacks and predict future CO2 concentrations, knowledge of these fluxes at the regional scale remains poor. This is particularly true in mountainous areas, where complex atmospheric flows and relative lack of observations lead to significant uncertainties in carbon fluxes. Many mountainous regions also have significant forest cover and biomass, yet these potential carbon sinks are highly dynamic and vulnerable to disturbance events, such as drought, insect damage, and wildfires. Our Carbon Monitoring System over Mountains (CMS-Mountains) uses a multi-scale approach by examining the site-level relationship between solar-induced fluorescence (SIF), leaf-level physiology, and gross primary productivity (GPP) at an observation intensive field site in Colorado. We incorporate this mechanistic understanding into the latest release of the Community Land Model (CLM 5) and then use this as the basis for initial simulations of biomass and carbon uptake across the Western U.S. Finally, we improve upon these regional simulations by assimilating multiple remotely-sensed observations (LAI, biomass, SIF, atmospheric CO2) within the Data Assimilation Research Testbed (DART).
Results/Conclusions
A ‘free’ simulation of CLM (no data assimilation) led to biospheric carbon fluxes across the Western US that severely underestimated above-ground biomass and carbon uptake. We have improved these simulations by using a bias corrected meteorology product that more accurately represents high elevation precipitation and shortwave radiation. Whereas the simulations were highly sensitive to the type of meteorology forcing used to drive CLM, in contrast, the spatial resolution at which a single meteorology forcing and land surface description was implemented had minimal effect on regional biomass and carbon uptake. When remotely sensed LAI and biomass observations were assimilated, this improved the simulated phenological timing, and reduced the biomass and amplitude of the carbon fluxes. Furthermore, whereas the ‘free’ simulation predicts the Western US was a carbon sink across a decadal simulation (1998-2010), the assimilated run predicted the Western US as carbon neutral. Whereas observations of LAI and biomass greatly improve the magnitude of simulated biomass and carbon uptake, observations of SIF are likely needed to better capture inter-annual variation.