Having recognized the limitations in spatial representativeness and/or temporal coverage of (i) current ground ET_a observations and (ii) land surface model‐ and remote sensing‐based ET_a estimates due to uncertainties in soil and vegetation parameters, a calibration‐free nonlinear complementary relationship (CR) model is employed with inputs of air and dew‐point temperature, wind speed, and net radiation to estimate 0.1°, monthly ET_a over China during 1982–2012. The modeled ET_a rates were first validated against 13 eddy‐covariance measurements, producing Nash‐Sutcliffe efficiency values in the range of 0.72–0.94. On the basin scale, the modeled ET_a values yielded a relative bias of 6%, and a Nash‐Sutcliffe efficiency value of 0.80 in comparison with water‐balance‐derived evapotranspiration rates across 10 major river basins in China, indicating the CR‐simulated ET_a rates reliable over China. Further evaluations suggest that the CR‐based ET_a product is more accurate than seven other mainstream ET_a products. The 31‐year mean annual ET_a value decreases from the southeast to the northwest in China, resulting in a country average of 406 ± 15 mm/year. The country‐representative annual ET_a rates slightly decreased with a rate of −0.5 mm/year (p = 0.86) during 1982–2012. Annual ET_a increased significantly over most parts of western and northeastern China but decreased significantly in many regions of the North China Plain as well as in the eastern and southern coastal regions. The present CR‐based method, with its calibration‐free nature and minimal data requirement, could help future calibrations/verifications of the more complex and more data‐intensive land surface model‐ and remote sensing‐based models.

The 2015 Paris Agreement proposed a more ambitious climate change mitigation target on limiting global warming to 1.5 ∘C instead of 2 ∘C above preindustrial levels. Scientific investigations on environmental risks associated with these warming targets are necessary to inform climate policymaking. Based on the Coupled Model Intercomparison Project phase 5 (CMIP5) climate models, we present the first risk-based assessment of changes in global drought and the impact of severe drought on populations from additional 1.5 and 2 ∘C warming conditions. Our results highlight the risk of drought on a global scale and in several hotspot regions such as the Amazon, northeastern Brazil, southern Africa and Central Europe at both 1.5 and 2 ∘C global warming relative to the historical period, showing increases in drought durations from 2.9 to 3.2 months. Correspondingly, more total and urban populations would be exposed to severe droughts globally (+132.5 ± 216.2 million and +194.5 ± 276.5 million total population and +350.2 ± 158.8 million and +410.7 ± 213.5 million urban populations in 1.5 and 2 ∘C warmer worlds) and regionally (e.g., East Africa, West Africa and South Asia). Less rural populations (−217.7 ± 79.2 million and −216.2 ± 82.4 million rural populations in 1.5 and 2 ∘C warmer worlds) would be exposed to severe drought globally under climate warming, population growth and especially the urbanization-induced population migration. By keeping global warming at 1.5 ∘C above the preindustrial levels instead of 2 ∘C, there is a decrease in drought risks (i.e., less drought duration, less drought intensity and severity but relatively more frequent drought) and the affected total, urban and rural populations would decrease globally and in most regions. While challenging for both East Africa and South Asia, the benefits of limiting warming to below 1.5 ∘C in terms of global drought risk and impact reduction are significant.

A warmer climate is expected to accelerate global hydrological cycle, causing more intense precipitation and floods. Despite recent progress in global flood risk assessment, the accuracy and improvement of global hydrological models (GHMs)-based flood simulation is insufficient for most applications. Here we compared flood simulations from five GHMs under the Inter-Sectoral Impact Model Intercomparison Project 2a (ISIMIP2a) protocol, against those calculated from 1032 gauging stations in the Global Streamflow Indices and Metadata Archive for the historical period 1971–2010. A machine learning approach, namely the long short-term memory units (LSTM) was adopted to improve the GHMs-based flood simulations within a hybrid physics- machine learning approach (using basin-averaged daily mean air temperature, precipitation, wind speed and the simulated daily discharge from GHMs-CaMa-Flood model chain as the inputs of LSTM, and observed daily discharge as the output value). We found that the GHMs perform reasonably well in terms of amplitude of peak discharge but are relatively poor in terms of their timing. The performance indicated great discrepancy under different climate zones. The large difference in performance between GHMs and observations reflected that those simulations require improvements. The LSTM used in combination with those GHMs was then shown to drastically improve the performance of global flood simulations (especially in terms of amplitude of peak discharge), suggesting that the combination of classical flood simulation and machine learning techniques might be a way forward for more robust and confident flood risk assessment.