How much and why are global land surface flux estimates uncertain?

Abstract:

So far, most remote sensing studies have focused on quantifying uncertainties in individual components of land surface fluxes such as carbon (photosynthesis, respiration) and energy (latent, sensible) separately, not all together. This raises the question whether there is a trade-off where reduced uncertainty in one flux component can go together with larger uncertainty in another flux component. This has not been tackled properly as most remote sensing models focus only on one flux component. Here, we use Breathing Earth System Simulator (BESS), which couples radiation, energy, and carbon flux modules and is forced mostly with remote sensing datasets, to quantify and attribute uncertainties in land surface energy and carbon fluxes. We quantify uncertainties derived from 1) the choice of forcing data including solar radiation, vapor pressure deficit, and leaf area index, 2) model structure by using different types of Vcmax modules, and 3) other relevant parameters in the Farquhar-von Caemmerer-Berry photosynthesis and Ball-Berry stomata conductance models. Key lessons learned include that 1) the best agreement against FLUXNET in terms of fluxes does not necessarily produce globally consistent flux estimates, and 2) the ranges of uncertainties derived from the aforementioned factors are large such as 10~20 PgC y-1 in global photosynthesis estimates. We discuss how to better constrain coupled flux estimates via functional relationships learned from FLUXNET and multi-objective optimization.