Understanding the potential drought impacts on agricultural production is critical for ensuring global food security. Instead of providing a deterministic estimate, this study investigates the likelihood of yield loss of wheat, maize, rice and soybeans in response to droughts of various intensities in the 10 largest producing countries. We use crop-country specific standardized precipitation index (SPI) and census yield data for 1961-2016 to build a probabilistic modeling framework for estimating yield loss risk under a moderate (-1.2 < SPI < -0.8), severe (-1.5 < SPI < -1.3), extreme (-1.9 < SPI < -1.6) and exceptional (SPI < -2.0) drought. Results show that there is >80% probability that wheat production will fall below its long-term average when experiencing an exceptional drought, especially in USA and Canada. As for maize, India shows the highest risk of yield reduction under droughts, while rice is the crop that is most vulnerable to droughts in Vietnam and Thailand. Risk of drought-driven soybean yield loss is the highest in USA, Russian and India. Yield loss risk tends to grow faster when experiencing a shift in drought severity from moderate to severe than that from extreme to the exceptional category, demonstrating the non-linear response of yield to the increase in drought severity. Sensitivity analysis shows that temperature plays an important role in determining drought impacts, through reducing or amplifying drought-driven yield loss risk. Compared to present conditions, an ensemble of 11 crop models simulated an increase in yield loss risk by 9%-12%, 5.6%-6.3%, 18.1%-19.4% and 15.1%-16.1 for wheat, maize, rice and soybeans by the end of 21st century, respectively, without considering the benefits of CO2 fertilization and adaptations. This study highlights the non-linear response of yield loss risk to the increase in drought severity. This implies that adaptations should be more targeted, considering not only the crop type and region but also the specific drought severity of interest.

Bias corrected daily climate projections from five global circulation models (GCMs) under the RCP8.5 emission scenarios were fed into a calibrated Variable Infiltration Capacity (VIC) hydrologic model to project future hydrological changes in China. The standardized precipitation index (SPI), standardized runoff index (SRI) and standardized soil moisture index (SSWI) were used to assess the climate change impact on droughts from meteorological, agricultural, and hydrologic perspectives. Changes in drought severity, duration, and frequency suggest that meteorological, hydrological and agricultural droughts will become more severe, prolonged, and frequent for 2020-2049 relative to 1971-2000, except for parts of northern and northeastern China. The frequency of long-term agricultural droughts (with duration larger than 4 months) will increase more than that of short-term droughts (with duration less than 4 months), while the opposite is projected for meteorological and hydrological droughts. In extreme cases, the most prolonged agricultural droughts increased from 6 to 26 months whereas the most prolonged meteorological and hydrological droughts changed little. The most severe hydrological drought intensity was about 3 times the baseline in general whereas the most severe meteorological and agricultural drought intensities were about 2 times and 1.5 times the baseline respectively. For the prescribed local temperature increments up to 3 °C, increase of agricultural drought occurrence is predicted whereas decreases or little changes of meteorological and hydrological drought occurrences are projected for most temperature increments. The largest increase of meteorological and hydrological drought durations and intensities occurred when temperature increased by 1 °C whereas agricultural drought duration and intensity tend to increase consistently with temperature increments. Our results emphasize that specific measures should be taken by specific sectors in order to better mitigate future climate change associated with specific warming amounts. It is, however, important to keep in mind that our results may depend on the emission scenario, GCMs, impact model, time periods and drought indicators selected for analysis

Human water withdrawal has increasingly altered the global water cycle in past decades, yet our understanding of its driving forces and patterns is limited. Reported historical estimates of sectoral water withdrawals are often sparse and incomplete, mainly restricted to water withdrawal estimates available at annual and country scales, due to a lack of observations at seasonal and local scales. In this study, through collecting and consolidating various sources of reported data and developing spatial and temporal statistical downscaling algorithms, we reconstruct a global monthly gridded (0.5 ^∘ ) sectoral water withdrawal dataset for the period 1971–2010, which distinguishes six water use sectors, i.e., irrigation, domestic, electricity generation (cooling of thermal power plants), livestock, mining, and manufacturing. Based on the reconstructed dataset, the spatial and temporal patterns of historical water withdrawal are analyzed. Results show that total global water withdrawal has increased significantly during 1971–2010, mainly driven by the increase in irrigation water withdrawal. Regions with high water withdrawal are those densely populated or with large irrigated cropland production, e.g., the United States (US), eastern China, India, and Europe. Seasonally, irrigation water withdrawal in summer for the major crops contributes a large percentage of total annual irrigation water withdrawal in mid- and high-latitude regions, and the dominant season of irrigation water withdrawal is also different across regions. Domestic water withdrawal is mostly characterized by a summer peak, while water withdrawal for electricity generation has a winter peak in high-latitude regions and a summer peak in low-latitude regions. Despite the overall increasing trend, irrigation in the western US and domestic water withdrawal in western Europe exhibit a decreasing trend. Our results highlight the distinct spatial pattern of human water use by sectors at the seasonal and annual timescales. The reconstructed gridded water withdrawal dataset is open access, and can be used for examining issues related to water withdrawals at fine spatial, temporal, and sectoral scales.

Pervious assessments of crop yield response to climate change are mainly aided with either process-based models or statistical models, with a focus on predicting the changes in average yields, whilst there is growing interest in yield variability and extremes. In this study, we simulate US maize yield using process-based models, traditional regression model and a machine-learning algorithm, and importantly, identify the weakness and strength of each method in simulating the average, variability and extremes of maize yield across the country. We show that both regression and machine learning models can well reproduce the observed pattern of yield averages, while large bias is found for process-based crop models even fed with harmonized parameters. As for the probability distribution of yields, machine learning shows the best skill, followed by regression model and process-based models. For the country as a whole, machine learning can explain 93% of observed yield variability, followed by regression model (51%) and process-based models (42%). Based on the improved capability of the machine learning algorithm, we estimate that US maize yield is projected to decrease by 13.5% under the 2 °C global warming scenario (by ∼2050 s). Yields less than or equal to the 10th percentile in the yield distribution for the baseline period are predicted to occur in 19% and 25% of years in 1.5 °C (by ∼2040 s) and 2 °C global warming scenarios, with potentially significant implications for food supply, prices and trade. The machine learning and regression methods are computationally much more efficient than process-based models, making it feasible to do probabilistic risk analysis of climate impacts on crop production for a wide range of future scenarios.