MODIS-Based Estimates of Global Terrestrial Ecosystem Respiration

Terrestrial ecosystem respiration (R-eco) represents a large carbon source from land to atmosphere and is highly spatiotemporally heterogeneous across scales. Upscaling of field-measured respiration data using remote sensing information is urgently needed for understanding regional and global patterns of ecosystem respiration. Using Moderate Resolution Imaging Spectroradiometer (MODIS) data with resolutions of 1km and 8days and flux measurements from 171 sites (total of 812 site years) across the world from 2000 to 2014, we developed a semiempirical, yet physiologically based, remote sensing model, which can simulate R-eco observed across most biomes with a small margin of error (R-2=0.55, root-mean-square error=1.67gCm(-2)d(-1), efficiency=0.46, and mean bias error=0.18gCm(-2)d(-1)). The reference respiration at the annual mean nighttime land surface temperature (LST) can be well represented by MODIS enhanced vegetation index and LST. A comprehensive comparison of six respiration-temperature (R-T) models shows that the more physiologically based R-T model (extended Arrhenius model) may be most suitable for estimating the respiration rate at higher latitudes. Integrating an effect of vegetation change on R-eco in different biomes effectively improves estimates of R-eco in almost all of the biomes. Plain Language Summary Here we develop an empirical yet physiologically based remotely sensed R-eco model (Ensemble_all) which is capable of capturing spatiotemporal patterns in R-eco at the global scale. We provide a new data set for global R-eco from 2001 to 2010, which will complement the currently relatively scarce global R-eco data sets and thus help to constrain global R-eco estimates, which are crucial for projecting climate change impacts on terrestrial carbon cycling and future atmospheric CO2 concentrations. This work also offers the opportunity to further understand the drivers behind ecosystem respiration.