Leaf chlorophyll: a global study employing a process-based model and remote sensing observations

Slowing down climate change calls for a strengthening of natural carbon sinks. Estimating current carbon stocks and the carbon storage potential of natural ecosystems necessitates a good understanding of carbon and nitrogen cycles. As the increase of land carbon sink is likely to be nitrogen-limited in temperate and boreal ecosystems, it is important to constrain the uncertainties related to the carbon and nitrogen processes in the ecosystems. Leaf chlorophyll (chl_leaf) and leaf nitrogen allocated to photosynthetic fractions are closely related, as plants optimise their nitrogen resources between light harvesting and the reactions of the Calvin cycle. chl_leaf is consequently one of the key factors in determining leaf photosynthetic rates and a strong proxy for photosynthetic capacity. The recent advances in remote sensing (RS) provide a novel opportunity for benchmarking the modelled terrestrial nitrogen cycle through leaf chlorophyll content.

In this study, we utilize a terrestrial biosphere model, QUINCY, for simulating the chl_leaf content for different ecosystems in a global scale. QUINCY includes a comprehensive representation of coupled carbon and nitrogen cycles, and also diagnostics for chl_leaf. We use a satellite-based leaf chlorophyll RS product for evaluating how well QUINCY captures spatial and temporal patterns of chl_leaf. The evaluation is conducted for a selection of 400 locations distributed world-wide to represent all major global biomes. In addition, we analyse the accuracy of chlorophyll and productivity (GPP) simulation at 169 sites of the FLUXNET eddy covariance.

Our initial results reveal that on global scale, QUINCY chl_leaf matches well with the RS chl_leaf observations. However, the QUINCY chl_leaf values seem to be constrained to a more narrow numerical range than the RS observations, indicating that not all factors contributing to the observed variation are considered in the modeling framework. For instance, the modeled grassland chl_leaf shows much smaller variation between different locations when compared to RS observations at different sites. For the FLUXNET sites, the mean annual GPP values from QUINCY are slightly underestimated (on average, ~-260 gC m^-2 yr^-1) when compared to flux observations. Nevertheless, the QUINCY mean annual GPP for different sites correlates with the ground station data reasonably well (r=0.67). 

Our study paves way for more versatile use of satellite observations within terrestrial biosphere models. Harnessing satellite products to model evaluation helps to improve model parametrizations related to carbon and nitrogen cycles, which in turn would allow more precise modeling of the terrestrial carbon budget.