Assessment of Land Surface Model's Evapotranspiration Response

Abstract

Biotic factors have been identified as one of the most important controls on evapotranspiration (ET) variation in the scenario of future climate change. Land surface models have developed sophisticated canopy processes to emphasize the importance of vegetation. However, as the vegetation processes become more and more complicated, the relative importance of biotic impact in comparison with abiotic impact on ET has not been well quantified. Failing to understand the relative importance between abiotic and biotic impact may result model bias in water cycle prediction. We collected satellite-based ET dataset (GLEAM, CRv1, P-LSH), climate data, biotic factor estimates, and apply the variance decomposition analysis to quantify the relative importance between biotic and abiotic impacts. Then, we compared with the model counterpart, i.e. the ensemble means of LS3MIP and CMIP6. Variance decomposition analysis on ET dataset show that about 70% of the ET inter-annual variation is contributed from abiotic factors, such as vapor pressure deficit (VPD), net radiation, and precipitation, whereas only 30% of ET variance is explained by biotic factors, such as stomata conductance and leaf area index (LAI). The abiotic contributions of the models show great uncertainties, which range from 36% to 60%. Overall, the abiotic factor contributions of most models are significantly higher than satellite-based ET dataset. ET variation of grassland is mostly explained by abiotic factors, which is consistent between models and ET dataset. VPD and precipitation explained most of the ET variation in ET dataset, especially in high latitude, whereas stomata conductance and LAI explained most of the ET variation in LS3MIP and CMIP6 models in boreal forest. The model overestimates of abiotic contribution indicate more complicated canopy processes require better constraints. Climate change leading to increase in VPD and more frequent extreme precipitation potentially play more important role in future ET changes. More efforts, such as model parameterization, calibration, new process development, still need to be made by modelers to improve model meteorological feedback.