Global comparison of light use efficiency models for simulating terrestrial vegetation gross primary production based on the LaThuile database

Simulating the gross primary productivity (GPP) of terrestrial ecosystems has been a major challenge in quantifying the global carbon cycle. Many different LUE models have been developed recently, but our understanding of the relative merits of different models remains limited. Using carbon dioxide flux measurements from 155 eddy covariance sites, we here assess the performance of seven LUE models (CASA, CFix, CFlux, EC-LUE, MODIS, VPM and VPRM) and compared the major model algorithms. Comparison between simulated GPP and estimated GPP from flux measurements showed that model performance differed substantially among ecosystem types. In general, most models captured well the pattern and magnitude of GPP for deciduous broadleaf forests and mixed forests, but performed less well in evergreen broadleaf forests and shrublands. During cloudy and overcast days, six out of seven LUE models significantly underestimated GPP because the impacts of diffuse radiation on light use efficiency were ignored. Among the seven models, CFlux and EC-LUE exhibited the lowest root mean square error at 80% and 75% sites respectively. Moreover, these two models showed higher performance for simulating the interannual variability of GPP. Two pairwise comparisons revealed the seven models differed substantially due to the differences in the algorithms describing the environmental regulation on GPP, particularly water stress. This analysis highlights the need to improve LUE model algorithms by integrating impacts of diffuse radiation on LUE and reliable water stress equations.JRC.H.7-Climate Risk Managemen