Crucial role of natural processes in detecting human influence on evapotranspiration by multisource data analysis

Abstract

Estimation of human-induced evapotranspiration (ET) changes (e.g., irrigation) is critical for understanding terrestrial water and energy cycle in the Anthropocene. A few studies have detected human influence on ET through comparing natural ET from reanalysis models and actual ET estimated by the water budget analysis with Gravity Recovery and Climate Experiment (GRACE) terrestrial water storage data. However, whether models can reasonably reproduce natural hydrological processes is overlooked. Here we show that uncertainty in model-simulated natural ET can determine whether the human-induced ET change is detectable. The naturalized runoff provided by the Yellow River Conservancy Commission is used to calibrate the Variable Infiltration Capacity (VIC) model to obtain a more reliable estimation of natural ET. The actual ET is estimated by either using the GRACE satellite retrievals with coal mining-induced gravity change removed, or 192 soil moisture station observations combined with surface and ground water resources bulletin data over the Yellow River basin. By comparing with natural ET from VIC simulations, human-induced annual ET changes are 41 ± 23 to 59 ± 31 mm/year during 2003–2010 for GRACE and soil moisture-based estimations respectively, contributing to 11%–16% increase in annual ET over the Yellow River basin. The estimation is comparable to the bulletin data of 47 mm/year. There is an obvious overestimation in natural ET from Global Land Data Assimilation System, prohibiting the detection of anthropogenic influence. This study highlights the importance of reasonably modeling natural hydrological processes as a baseline for detecting anthropogenic influence on the hydrological cycle.