Evolution of light use efficiency models: Improvement, uncertainties, and implications

Light use efficiency (LUE) models have been widely used to estimate terrestrial gross primary production (GPP) at local, regional, and global scales, which is vital for understanding the carbon flux dynamics under climate change. LUE models express GPP as the product of the incoming photosynthetically active radiation (PAR), the fraction of PAR absorbed by plants (FPAR), the maximum LUE, and the environmental stress factors (e.g., temperature, water, and carbon dioxide). Here, we investigate 21 LUE models reported in literatures and conclude their complicated evolutions in the aforementioned four components: 1) the representation of PAR was improved from total PAR to direct and diffuse PARs; 2) the representation of FPAR was improved from one-leaf to two-leaf (i.e., sunlit and shaded leaves) or chlorophyll based strategies; 3) the parameterization of the maximum LUE was improved from a constant value of 0.39 gC/MJ to the C3/C4- and sunlit/shaded leaf-specific values; and 4) the representation of environmental stress factors was improved both in their integration forms (e.g., from the multiplication method to the law of the minimum method) and the proxy optimization for a specific stress factor. For example, the proxy for water stress factor has evolved from atmospheric (e.g., vapor pressure deficit) and soil (e.g., soil moisture) water indicators to the plant (e.g., land surface water index) water indicators. We also identify uncertainties caused by model structures, parameterizations, input data with various resolutions and accuracies, and scale mismatch issues between remote sensing data and flux tower observations. The newly emerged indicators such as the photochemical reflectance index, solar-induced chlorophyll fluorescence, and near-infrared reflectance of vegetation simplify the methods to estimate GPP but fail to disentangle the influences of different environmental factors. These findings on the evolution of LUE models and their uncertainties are expected to contribute to future model improvements.