Development of a coupled carbon and water model for estimating global gross primary productivity and evapotranspiration based on eddy flux and remote sensing data

Terrestrial gross primary productivity (GPP) and evapotranspiration (ET) are two key ecosystem fluxes in the global carbon and water cycles. As carbon and water fluxes are inherently linked, knowing one provides information for the other. However, tightly coupled and easy to use ecosystem models are rare and there are still large uncertainties in global carbon and water flux estimates. In this study, we developed a new monthly coupled carbon and water (CCW) model. GPP was estimated based on the light-use efficiency (LUE) theory that considered the effect of diffuse radiation, while ET was modeled based on GPP and water-use efficiency (WUE). We evaluated the non-linear effect of single (GPP(OR)) or combined (GPP(AND)) limitations of temperature and vapor pressure deficit on GPP. We further compared the effects of three types of WUE (i.e., WUE, inherent WUE, and underlying WUE) on ET (i.e., ETWUE, E-IWUE and ETUWUE). CCW was calibrated and validated using global eddy covariance measurement from FLUXNET and remote sensing data from Moderate Resolution Imaging Spectroradiometer (MODIS) from 2000 to 2007. Modeled GPP(AND) and GPP(OR) explained 67.3% and 66.8% of variations of tower-derived GPP, respectively, while ETUWUE, E-IWUE and ETWUE explained 65.7%, 59.9% and 58.1% of tower-measured ET, respectively. Consequently, we chose GPPAND and ETuwuE as the best modeling framework for CCW, and estimated global GPP as 134.2 Pg Cyr(-1) and ET as 57.0 x 10(3) km(3) for vegetated areas in 2001. Global ET estimated by CCW compared favorably with MODIS ET (60.5 x 10(3) km(3)) and ET derived from global precipitation (56.5 x 10(3) km(3)). However, global GPP estimated by CCW was about 19% higher than MODIS GPP (109.0 Pg C yr(-1)). The mean global WUE value estimated by CCW (2.35 g C kg(-1) H2O) was close to the mean tower-based WUE (2.60 g C kg(-1) H2O), but was much higher than the WUE derived from MODIS products (1.80 g C kg(-1) H2O). We concluded that the new simple CCW model provided improved estimates of GPP and ET. The biome-specific parameters derived in this study allow CCW to be further linked with land use change models to project human impacts on terrestrial ecosystem functions. (C) 2016 Elsevier B.V. All rights reserved.