Understanding moisture stress on light-use efficiency based on MODIS and global flux tower data

Gross primary productivity (GPP) is a key indicator of terrestrial ecosystem functions and global carbon balance. However, accurately estimating GPP is still one of the major challenges in global change study. Compared with other prognostic models, remote-sensing-based light-use efficiency (LUE) modes are considered to have the most potential to characterize the spatial-temporal dynamics of GPP. However, the environmental regulations on LUE, especially from water stress, have relatively large uncertainties, which reversely constrained the applications of LUE models. Here, we used MODIS and global flux tower data to investigate the moisture stress on LUE for different biomes on daily, 8-day and monthly scales. Three groups of moisture stress indicators were adopted in our study, including atmosphere (i.e. precipitation and daytime vapor pressure deficit (VPD)), soil (i.e. soil water content (SWC) and scaled SWC (SWCs) by field capacity and wilting point) , and plant indicators (i.e. land surface wetness index (LSWI) and the ratio of latent heat to the sum of latent and sensible heat (L/(L+H)). We applied a series of steps to eliminate the effects of high/low temperature and diffuse radiation effects on observed LUE. Our analysis showed that there were great variations in moisture stress effects on LUE between and within biomes. Generally, the moisture stress effects on LUE are ranked as plant indicator (i.e. L/(L+H) & LSWI) > atmosphere indicator (i.e. VPD) > soil indicator (i.e. SWC/SWCs). Precipitation has the poorest relationship with observed LUE and doesn't show any significant lag effects. For deep-root biomes (e.g. forest), LUE shows higher sensitivity in VPD than SWC; but for short-root biomes (e.g. grass), LUE is more sensitive to SWC than VPD. Most indicators (except SWC/SWCs) are more effective in affecting LUE at the daily/8-day scale than at the monthly scale probably because the observed LUE becomes more stable as temporal scale increases. SWC do not show close relationship with LUE, suggesting that the current measured SWC in the top-soil layer may not be sufficient to capture the moisture effects on LUE for biomes with different root distributions. Our study highlights the complexity of moisture stress on observed LUE, and provides useful guidance for developing more reliable LUE models to estimate GPP.