Retrieval of seasonal Rubisco-limited photosynthetic capacity at global FLUXNET sites from hyperspectral satellite remote sensing: Impact on carbon modelling

Abstract

Listen

Translate article

Process-based ecophysiological models, which simulate carbon exchange at the land-surface, are powerful and indispensable tools for understanding how vegetation behaves under present and future climate. However, these models are necessarily complex, containing numerous biophysical parameters which are often poorly defined. The current study develops a novel retrieval of Rubisco-limited top-of-canopy photosynthetic capacity (i.e. maximum carboxylation rate, Vcmax25,toc), which is one of the most critical parameters in the calculation of Gross Primary Productivity (GPP). The retrieval combines standard remote sensing satellite products of Leaf Area Index (LAI), from the Moderate Resolution Imaging Spectroradiometer (MODIS), with a hyperspectral index of total canopy chlorophyll concentration from the MEdium Resolution Imaging Spectrometer (MERIS). Monthly values of Vcmax25,toc are determined over a 9 year period for 296 global FLUXNET sites (catalogue made available online) and 8 Plant Functional Types (PFTs). PFT averages agree favourably with compilations of field-based measurements. However, according to a Monte Carlo analysis, our method is still currently subject to large systematic uncertainties (25–30%), much of which arises from the empirical relationship between maximum electron transport and leaf chlorophyll content. For all 8 PFTs, except tropical broadleaf forest, Vcmax25,toc varies considerably across the season (generally a factor of 1.6). Similarly, variability between sites of the same PFT is significant (interquartile range is 40% of the median). This suggests an important additional role for satellites in the spatial and temporal parameterisation of carbon models. Inclusion of this temporal and spatial variability in a process-based ecophysiological model produces, respectively, an impact of 11% and 12% on simulated GPP.