Land–atmosphere coupling exacerbates the moisture-associated heterogeneous impacts of compound extreme events on maize yield in China

Compound climate events are major threats to crop production under climate change. However, the heterogeneity in the impact of compound events on crop yield and its drivers remain poorly understood. Herein, we used empirical approach to evaluate the impact of compound hot–dry (HD) and cold–wet (CW) events on maize yield in China at the county level from 1990 to 2016, with a special focus on the spatial heterogeneity. Our findings indicate comparable impact of extremely compound CW events (−12.8 ± 3.6%) on maize yield loss to extremely compound HD events (−11.3 ± 2.1%). The spatial pattern of compound HD and CW events impacts on maize yield was dominantly associated with moisture regime, followed by management practices and soil properties. Specifically, drier counties and counties with less fraction of clay soil and organic carbon tend to experience greater yield loss due to compound HD events, and wet condition, excessive fertilizer, clay soil and rich organic carbon aggravate the maize yield loss due to compound CW events. Moreover, the land–atmosphere coupling exacerbated the heterogeneous yield impact through divergent heat transfer. In drier regions, the greater proportion of sensible heat creates a positive feedback between drier land and hotter atmosphere. In contrast, the greater proportion of latent heat in wetter regions results in a positive feedback between wetter land and colder atmosphere. Our results highlighted a critical element to explore in further studies focused on the land–atmosphere coupling in agricultural risk under climate change.


Introduction
Under climate change, extreme weather and climate events are projected to intensify (IPCC 2012) and often co-occur formed by the interweaving of a variety of events.For example, the European heatwave and drought in summer 2018 and 2019 (Bastos et al 2021), as well as the heatwave and drought in July 2012 in the United States (Glotter and Elliott 2016), caused widespread harvest failures and infrastructural damages.The combination of multiple events leading to high social or environmental impacts is often termed compound events or extremes (Zscheischler et al 2020), which may induce larger negative impacts on different systems or sectors than their univariate counterparts (Zscheischler et al 2018), especially agriculture.Understanding compound events and their impacts is critical for yield estimation and adaptation strategy under a changing climate.
Maize is one of the world's most prominent food crops (Zhang et al 2020) and China is the second-largest producer and consumer of maize.From 1981 to 2012, the annual maize yield loss due to meteorological disasters reached 2.31 million tons in China (Jiao et al 2017).Nearly 70% of maize harvested areas in China have experienced compound climatic extreme events in the past three decades (Li et al 2022).There is an increase in the frequency, duration, and spatial extent of compound hot-dry (HD) extremes in China in the past 60 years (Zhang et al 2022).However, studies on the impact of climate extremes have mainly focused on individual climate extremes (Shi et al 2021), neglecting the impact of compound events on crop production in China (Hao 2022).
The occurrence of precipitation deficits is often accompanied by high temperature extremes (or heavy precipitation occurs in cold days) during crop growing season in most of the global croplands due to the physical couplings between temperature and moisture (Lesk et al 2021).The significant impacts of compound drought and heat on crops have already been extensively investigated (Ribeiro et al 2020, Hao et al 2021, Haqiqi et al 2021).For instance, compound hot and dry summer conditions reduced soybean yields by 2 standard deviations in the United States (Hamed et al 2021).Despite existing studies indicating that excessive wetness and cold injuries significantly constrains crop growth and reduces crop production (Farooq et al 2009, Li et al 2019), there has been relatively little focus on the impacts of compound CW events on crops.This lack of attention may hinder our ability to fully understand and assess the climate change impacts on crops.
The impacts of climate extremes on crop yield could vary by environmental factors and management practices.Previous studies have revealed the regional difference of maize yield responses to excessive rainfall in the United States (Li et al 2019) and China (Liu et al 2022) by regional climate conditions, soil properties and management practices.Environmental and managemental factors could also influence yield responses to heat and drought stress, for instance, irrigation can alleviate drought stress and mitigate heat stress via evaporative cooling (Li et al 2020).Additionally, land-atmosphere coupling associated with moisture conditions affects the compound events by controlling near-surface heat and moisture (Fischer et al 2007).Previous studies have shown that dry soil moisture conditions can significantly exacerbate the severity of heatwaves and droughts by positive feedbacks between the soil moisture and the atmosphere (Hsu and Dirmeyer 2023).Such regulating effects of environmental factors and associated land-atmosphere coupling need further investigation in crop responses under complicated compound climatic stresses.
This study aims to explore the dominant factors that can explain spatial heterogeneous impact of compound extremes on maize yield in China.Specifically, this study contributes to literature by addressing the following research questions: (a) to what extent is the impact on maize yield caused by compound HD and CW events?(b) How does such compound-event-induced yield loss differ according to the harvest region in China, and what are the underlying environmental drivers of those differences?The study results highlight the threat of compound events to food security and provide valuable information for formulating effective agricultural measures and enhancing the resilience of food systems under global warming.

Data
To estimate the response of maize yield loss to compound HD and CW events, information on county-level statistical maize yield, climate observations, and maize phenological observations from different regions in China were acquired.
County-level statistical maize yield data from 1990 to 2016 were obtained from the Chinese statistical yearbooks, including 2604 counties and covering all major maize cropping areas in China.To control the uncertainties of maize yield data caused by survey errors and scale issues, the abnormal samples that exceed 3σ of the mean value were removed.Daily temperature and precipitation data for the 1989-2016 period was obtained from the Chinese Meteorological Administration CN05.1 dataset (CMA, http://data.cma.cn) with a spatial resolution of 0.25 • , interpolated from observation from 2400 reference meteorological stations (Wu and Gao 2013).Maize phenological observations by production region were obtained from the Chinese Crop Growth and Development Dataset released by the CMA.Based on the records, the maize growing season was divided into sowing, seedling, ear, and flowering-to-maturity stages (Liu et al 2021a) (table S1).Phenological dates of these periods differ by maize production regions.The six major maize production regions in China are the northern China maize production region (NM), the Huang-Huai-Hai maize production region (HM), the southwestern China maize production region (SW), the southern China maize production region (SM), the northwestern China maize production region (NW), and the Qinghai-Tibet Plateau maize production region (QT) (figure S1).
Maize distribution data were prepared to aggregate gridded climate information into county-level values and calculate the proportion of harvest area irrigated for statistical analysis.Irrigated and rainfed maize harvested areas in years 2000, 2005, and 2010 were generated with a spatial resolution of 10 km by using the spatial production allocation model (SPAM, http://mapspam.info),which adds methodological and data enhancements to available crop downscaling modeling (Yu et al 2020).Since long continuous annual crop data are not available in China, SPAM is currently the best method for obtained ready-to-use crop distribution data.To reduce the uncertainty due to harvested area change as much as possible, the SPAM harvested areas in 2000, 2005, and 2010 were used to aggregate the climate information before 2000, from 2001 to 2005, and after 2006, respectively, to the county level.
To evaluate the role of environmental and management factors, soil moisture, irrigation, soil properties, and fertilizer application rate information was obtained.The soil moisture was obtained from an Earth-system-based daily hydrological dataset (VIC-CN05.1)with a spatial resolution of 0.25 • for China (Miao and Wang 2020).The dataset is developed based on incorporating high-resolution soil parameters and pure station-based atmospheric forcing into the latest variable infiltration capacity model.The fraction of irrigated harvested areas provided by SPAM was used to separate the statistical yield samples with large and small proportions of irrigation.The 1 km soil properties were obtained from the Harmonized World Soil Database (version 1) (Nachtergaele et al 2009).The top soil organic carbon (TOC) and proportion of clay were selected to separate the county samples based on the soil properties.A 10 km nitrogen application rate was obtained from EARTHSTAT, which comprises data for national and subnational units from across the globe (Mueller et al 2012).The nitrogen application rate was used to separate the county-level statistical yield samples with low and high fertilizer regimes.
To illustrate the role of land-atmosphere coupling in the crop response to compound events under different moisture regimes, the monthly averaged surface sensible heat flux and surface latent heat flux from 1990 to 2016 was obtained from a 0.1 • state-of-the-art global reanalysis dataset ERA5-land (Muñoz-Sabater et al 2021).Bowen ratio was obtained from the surface sensible and latent heat flux data and then aggregated to the county level.

Identification of compound events
To explore the impact of the variation in temperature and precipitation on maize yield, their standardized anomalies were calculated.The growing season mean temperature anomaly (TA) and total precipitation anomaly (PA) were calculated as deviations from multiyear averages and were normalized using standard deviations of 1990-2016 (Li et al 2019).TA and PA were calculated as follows: where i represents the year , mean denotes the multi-annual average of each county, total is the sum of each county, and std represents the standard deviation of each county.In this study, TA and PA were used to identify compound HD and CW events.Compound HD events were defined as PA < −0.5σ and TA > 0.5σ, and compound CW events were defined as PA > 0.5σ and TA < −0.5σ.Extremely compound HD and CW conditions were defined as PA < −1.5σ and TA > 1.5σ, PA > 2σ and TA < −2σ, respectively.The different σ thresholds for compound events were chosen to ensure that hot, dry, cold, wet conditions are equally represented (figure S2).The individual hot, dry, cold, and wet conditions counted for 30% of the total county-year TA and PA samples, while the extremely individual conditions accounted for 5% of the total samples.Accordingly, the compound HD and CW events defined this way accounted for roughly 10% of all county-year samples, and extremely compound HD and CW conditions accounted for merely 1% of all county-year samples.

Quantifying impacts of compound events on maize yield
After identifying compound or individual events, their impact on maize yield for corresponding county-specific years was quantified by using yield anomaly (YA).The YA for each county was derived by using the annual maize yield (Yield i ) and the long-term yield trend (Yield trend ).The robust local weighted regression was used to calculate the technological trend of the maize yield to ensure robust performance in the determination of trend under various trend assumptions (Ye et al 2015).The YA was separately calculated for each county, and the YAs can reflect the departure of the yield from the long-term trend owing to climate variability YA = (Yield i − Yield trend ) /Yield trend × 100%. (3) The annual maize YA can be calculated for each county from 1990 to 2016.

Exploring heterogeneous impacts of compound events on maize yield
To explore the main environmental factors affecting the yield impacts under compound events, we compared yield impacts by climatic factors, management practices, and soil properties using Pearson's and partial correlation coefficients, and subsample analysis (Rigden et al 2020).For the correlation analysis, we employed climatic variables (growing season mean temperature T, growing season total precipitation P), soil properties (TOC, proportion of clay Clay) and management practices (nitrogen application rate N, fraction of irrigated area I) (Li et al 2019, Liu et al 2022) to obtain the Pearson's and partial correlation coefficients between environmental factors and YAs at the county level.
For the subsample analysis, we used low and high threshold of environmental factors to subsample the counties and compared the yield impacts among the subsamples.For the fraction of irrigated area, the low and high thresholds are 20% and 80%, respectively.For other environmental factors, the low and high thresholds are the 20th and 80th percentiles, respectively (table S2).

Exploring the effect of land-atmosphere coupling associated with moisture regime
To explore the effect of land-atmosphere coupling, we employed the Bowen ratio to represent the surface energy balance.Bowen ratio is defined as the ratio of sensible to latent heat fluxes at a certain interface, is often used to characterize their partitioning with broad implications for many fundamental physical processes of the Earth system (Bowen 1926).Bowen ratio plays a crucial role in measuring the energy and mass exchange between the land and the atmosphere (Lin et al 2022).A higher Bowen ratio suggests that more available energy is channeled into sensible heat for direct heating, increasing surface temperature.In contrast, a lower Bowen ratio implies that more energy is channeled into latent heat entering the water cycle, which potentially increases precipitation.
We calculated the Bowen ratio anomaly (BA) at the county level from 1990 to 2016 to investigate the changes in energy balance under compound events.A positive BA (>0) represents a higher Bowen ratio than its multiyear average, while a negative BA (<0) means a lower Bowen ratio.We investigated BA during compound events at the county level, and explored the relationship between BA and YA in dry and wet counties.For the division of dry and wet counties, we considered both precipitation P and irrigation I, and their corresponding thresholds of sub-samples as defined in the previous section.

Impacts of compound HD and CW events on maize yield
Over 1990-2016, compound HD and CW events caused more severe yield loss than the other two types of possible temperature-and-precipitation compound events (hot-wet and cold-dry) (figure 1(a)).Maize yield response showed a clear gradient with the changes in TA and PA.Favorable precipitation and temperature conditions with minimal environmental stress resulted in up to 5% positive YA.However, when the precipitation and temperature deviated from normal toward severe conditions, the modest yield gain was replaced by yield loss. Yield loss increased as the conditions intensified, and the average yield loss exceeded 30%.
The extremely compound CW and HD events caused comparable maize yield loss (figure 1(b)).Meanwhile, extremely compound events caused considerably greater yield loss compared to individual counterparts.The average maize YA under extremely extreme compound HD events was −11.3 ± 2.1%, which is statistically significantly larger than that caused by individual dry events (−9.2 ± 0.8%) and hot events (−5.7 ± 0.8%).The average yield loss caused by extremely compound CW events (−12.8 ± 3.6%) was greater than that caused by individual wet events (−3.6 ± 0.7%) and cold events (−1.1 ± 0.8%), both of which are statistically significant.
Among maize growth stages, extremely compound HD events occurred in the sowing stages resulted in the most substantial yield loss (figure 1(c)), reducing maize yield on average by −15.0 ± 2.8%.Compound HD conditions during sowing stage would delay planting (figure S3), due to the damage to young plants of anomalous heat and drought (Sánchez et al 2014, Lu et al 2017, Liu et al 2023).There was no significant difference in the yield loss caused by extremely compound CW events in each growth stage, with relatively greater yield reduction in the flowering-to-maturity stage (figure 1(c)).Severely compound CW events (such as hail) tended to induce greater stress than the earlier stages and lead to more maize yield loss due to physical damage from heavy rainfall and concurrent low temperature (Battaglia et al 2019).

Heterogeneous impacts of compound events on maize yield
The impacts of compound HD and CW events on maize yield in China over 1990-2016 had substantial spatial difference, which was dominantly associated with moisture regime, and regulated by the management measures and soil properties.Under compound HD events, counties with significant maize yield loss were mainly located in, from northeast to the southwest, the Songnen Plain (central Northeast China), the northern agro-pastoral transition zone, and the eastern and southern margins of the loess plateau (figure 2(a)).Under compound CW events, the counties with significant maize yield loss were mainly located in the eastern margins of the northern China maize production region, the south of the Huang-Huai-Hai Plain, and the southern China maize production region (figure 2(b)).The moisture regime (precipitation and irrigation conditions) could explain the spatial difference of yield impact under compound events (figure 3).In the counties with low growing season total precipitation (P < 400 mm) (figure S4(a)), compound HD events caused significantly greater maize yield loss than that in the counties with high precipitation (P > 930 mm), with a difference of 8.3%.Irrigation also mitigated the maize yield loss due to compound HD events, contributing 2.2% on average.Under compound CW events, the counties with a larger fraction of irrigated area (I > 80%) (figure S4(b)) experienced greater yield loss, with a difference of 4.1%.The difference related to irrigation was significant under both compound event types.The Pearson's and partial correlation coefficients showed consistent results (table S3).
Apart from the moisture regime, temperature could also explain the spatial difference of yield impact under compound events.Under compound HD events, counties with cool climate (T < 18 • C) experienced greater yield loss than those with warm climate (T > 24 • C) (with a difference of 2%).Counties with warm climate experienced more yield loss under CW events, with a difference of 2.4%.The correlation coefficients were not significant between temperature and yield impact (table S2).The effect of temperature was related to the precipitation effect to some extent, since the counties with low or high mean temperature are accompanied by low or high total precipitation during growing season, further affecting the yield impact (figure S5(a)).
Management practice strongly regulates the heterogeneous impacts of compound events on maize yield (figure 3).In addition to irrigation conditions, fertilizer applications play an important role in the effects of compound CW events.The use of nitrogen fertilizer induced the greatest difference in maize yield anomalies under compound CW events, with 6% on average between counties applying excessive nitrogen fertilizer (N > 230 kg ha −1 ) and counties using less (N < 150 kg ha −1 ).
Soil properties also affects the spatial pattern of impacts of compound events on maize yield (figure 3).In the counties with higher proportions of clay (Clay > 30%) and richer TOC (>1.25%), compound HD events led to less maize yield loss (with a difference of 4% and 5%, respectively).While under compound CW events, counties with higher soil clay content and richer TOC experienced more severe maize yield loss (with a difference of 2% and 1%, respectively).In the main plots, the compound events impact on BA for any individual county on the map is the BA averaged from compound events years during the period.In the subplots, the bar denotes the average BA grouped by the low and high thresholds of P and I at the county level, and the error bar denotes the ±95% credible intervals.The low and high thresholds of P and I are consistent with figure 3. Asterisks signify the significance level of the difference between individual events and corresponding compound events, and * * * indicates a significance level of 1%.

Land-atmosphere coupling exacerbates the yield loss due to compound events
Land-atmosphere coupling associated with moisture regime exacerbated the heterogeneous impacts of compound HD and CW events on maize yield (figure 4) through the positive feedbacks between land and atmosphere (figure 5).
During compound HD events, 76% counties experienced positive BAs (>0), while 73% counties experienced negative BAs (<0) under compound CW events (figure 4).There were significant differences in BA in drier and wetter counties, 0.6σ for compound HD events and 0.4σ for compound CW events on average.When the drier counties (with limited precipitation and irrigation) experience compound HD events, soils become much drier and water is unable to enter maize roots and supply transpiration, resulting in larger positive BA than that in wetter counties (figure 4(a)).This further intensifies the water potential gradient between the canopy and atmosphere, causing positive feedbacks that damage maize growth and development.When the wetter counties experience compound CW events, the wet soils get wetter and leads to more negative BA (more latent flux heat and less sensible flux heat) than that in drier counties, causing positive feedbacks intensifying the cold and wet conditions.
Those positive feedbacks between land and atmosphere played opposite role in the maize yield impacts under compound events in drier and wetter counties (figure 5).During compound HD events, the positive feedbacks (BA > 0) aggravated the heat and drought stresses, exacerbating maize yield loss in drier counties and mitigating yield loss in wetter counties.During compound CW events, the positive feedbacks (BA < 0) amplified the cold and waterlogged environmental conditions, further damaging maize yield in wetter counties and favoring maize yield in drier counties.

Discussion
In this study, we investigated the heterogeneous impacts of compound HD and CW events on maize yield in China at the county-level during 1990-2016.Our results underline that the compound CW events, which has been largely under-studied previously, can adversely affect Chinese maize yield as much as compound HD events.Our results highlighted the exacerbating effect of land-atmosphere coupling in the heterogeneous impacts associated with moisture regime, and the regulating effect of management practices and soil properties.

Mechanisms of compound events impact on crop yield
Our results showed comparable maize yield impacts of extremely compound CW events with extremely compound HD events in China (figure 1(b)).Among maize growth stages, the yield loss under extremely compound CW events was the most severe during the flowering to maturity stage (figure 1(c)).Compound CW events reduce maize yield through processes associated with crop-physiological interactions: physical damage caused by hail (Rana et al 2022), pollination and fertilization disruption (Thakur et al 2010), delayed planting/harvest (Urban et al 2015), root growth suppression (Ren et al 2016), and pathogens attack (Juurakko et al 2021).The crop-physiological interactions of compound HD events causing the crop yield loss have been widely studied (Leng 2019).High temperatures cause the intensification of water stress resulting in the gradual closure of stomata, reduction in CO 2 uptake, and enhancement of the root growth at the expense of aboveground biomass (Hamed et al 2021).The induced stomata closure subsequently leads to weaker surface cooling enhancing the heat stress experienced by the crops (Siebert et al 2017).The greater impact of compound HD events during sowing stages than other growth stages (figure 1(c)) are divergent from previous findings.Such studies quantified the sensitivity of crop yield to high temperature without considering the compounding effect of drought.When the compound HD events occur, the farmer practices are likely to be influenced because the farmers need to evaluate whether they should plant crop or not (Liu et al 2023).
Some mesoscale-associated weather events are correlated with compound events and are likely to cause physical damages.For compound CW events, excess wetness might hit the maize plants through hail (Liu et al 2022), particularly in those colder regions in northern maize-harvested areas.While in coastal regions, the compound CW events are associated with windstorms, such hourly strong wind and rainfall are detrimental to maize growth (Lesk et al 2020).For compound HD events, the regional heat dome is the pre-condition of occurrence of compound HD.The heat dome-like atmospheric circulation can inhibit the cross-ventilation (Zhang et al 2023).

Exacerbating effect of land-atmosphere coupling associated with moisture regime
Our results showed the substantial heterogeneous impacts of compound extreme events on maize yield, and the exacerbating effect of land-atmosphere coupling associated with moisture regime.Local moisture regime could strongly explain the regional differences of the yield impacts. Dryer and wetter counties exhibited a statistically significant difference of 2%-8% in yield impact under compound HD events, while the yield impact difference was 1%-4% under compound CW events.
The land-atmosphere coupling is the feedback between land-surface responses, associated with moisture regime and posing a greater threat to crop production.Changes in the partitioning of energy at the land surface affect the surface temperatures (Seneviratne et al 2010).A portion of the incoming radiation leads to warmer surface temperatures (sensible heating), whereas most fuels evapotranspiration (latent heating).During compound HD extremes, as the land surface becomes moisture-limited, surface temperatures rise more quickly due to partitioned energy to sensible heating.Hotter conditions then boost the soil drying, causing positive feedbacks (Lesk et al 2021).Such feedbacks occur most strongly in zones that are semi-humid to semi-arid (that is, transitional zones) (Holmes et al 2017), including important breadbaskets in Eurasia and the North American Great Plains.During compound CW events, as the land surface becomes energy-limited (Hsu and Dirmeyer 2023), air temperatures tend to decrease as the energy partitioned to latent heating and the Bowen ratio decreased.Colder conditions then intensify the waterlogging stress, resulting in positive feedbacks causing more damage to crop yields.
Not only does land-atmosphere coupling exacerbate yield responses, but atmospheric circulation coupling also generates or amplifies the compound events and crop impacts.Our results showed that stronger atmospheric circulation coupling led to more occurrence of compound HD and CW events, among different growth stages (figure S6) and different regions (figures S7 and S8).The strengthening temperature-moisture couplings (Lesk et al 2021) have the potential to exacerbate the impact of climate warming on global crop yields (Liu et al 2021b).The exacerbating effect of land-atmosphere coupling in yield impacts of compound events indicate the increasing threat of compound events under climate change.Although the occurrence of extreme cold events has decreased, the occurrence of extreme heat events has increased over the past three decades (figure S9) and will increase under climate change (high confidence).Robust adaptation and mitigation of cropping systems need to consider this underappreciated risk to food security from climate change.

Regulating effect of management practices and soil properties
Management practice, i.e. fertilization application, contribute to the regional differences in the yield loss due to compound events.Excessive nitrogen application tends to exacerbate the maize yield loss by 6% due to compound CW events, which is associated with the softening effect of overfertilization on soil and nitrogen leaching (Lu et al 2021).The adverse effects of excessive fertilization and irrigation on maize yield under excess moisture stress would be exacerbated when compounding with cold stress.
Soil properties have a regulating effect on yield impact of compound events, which has rarely been investigated in previous studies.The spatial distribution of clay content and TOC is related to precipitation (figure S5).A high clay content in soil signifies low saturated hydraulic conductivity, i.e. high water retention and poor drainage would mitigate the yield loss under extreme dryness and aggravate the yield loss under extreme wetness (Sala et al 2015, Li et al 2019).Soil organic carbon can mitigate heat and drought impact and exacerbate waterlogging impact (Droste et al 2020, Liu et al 2022).
The heterogeneous impacts of compound events on maize yield across management practices and soil properties indicate that more attention should be paid to localized and event targeted adaptation and mitigation strategies under climate change.As different management practices might have opposite regulating effects on the yield impacts of compound HD and CW events, leveraging the advantages and disadvantages of adaptation strategies leave a critical issue in agricultural risk management.

Conclusion
Our results highlighted the comparable impacts of compound CW and HD events on maize yield in China, and the substantial regional difference.The heterogeneous yield impacts of compound events were dominated by moisture regime, and regulated by management practices and soil properties.The positive feedbacks induced by land-atmosphere coupling associated with moisture regime exacerbated the heterogeneous yield impacts. Drier counties tended to experience greater yield loss under compound HD events, while the yield loss in wetter counties was greater under compound CW events.The counties with well nitrogen fertilizer application experienced greater yield loss due to compound CW events.The abundant clay soil and organic carbon mitigated the yield loss due to compound HD events, and exacerbated the yield loss due to compound CW events.
Our results suggest that heterogeneous yield impact of compound events associated with moisture regime, management practices and soil properties.Our findings highlight the need for improved understanding and modeling of soil-plant-atmosphere dynamics in agricultural risk management and spatially targeted adaptation and mitigation measures under climate change.

Figure 1 .
Figure 1.Impacts of compound hot-dry and cold-wet events during growing season (a), (b) and growth stages (c) on maize yield in China from 1990 to 2016.(a) Average maize yield anomaly (YA) to precipitation anomaly (PA) and temperature anomaly (TA).(b) Maize yield response to extremely compound hot-dry (HD) and cold-wet (CW) events, and individual hot (H), dry (D), cold (C), and wet (W) events.(c) Maize yield anomaly caused by extremely compound hot-dry (HD) and cold-wet (CW) events during the sowing (SO), seedling (SE), ear (EA), and flowering-to-maturity (FL) stages.Error bars denote the 95% confidence interval of the mean.Asterisks denote the significance level of the difference between individual events and corresponding compound events.* * * indicates a significance level of 1%, * * indicates a significance level of 5%, and * indicates a significance level of 10%.

Figure 2 .
Figure 2. The impacts of compound hot-dry (a) and cold-wet (b) events on maize yield in China from 1990 to 2016.The compound events impact on maize yield for any individual county on the map is the yield anomaly (YA) averaged from compound events years during the period.Counties in white color have not been affected by compound events or have no maize grown.

Figure 3 .
Figure 3. Spatial dependence of county-level maize yield anomaly (YA) due to environmental factors under compound hot-dry (a) and cold-wet (b) events.The bar denotes the average anomaly grouped by the low and high thresholds of environmental factors (growing season mean temperature T, growing season total precipitation P, top soil organic carbon TOC, proportion of clay Clay, nitrogen application rate N, fraction of irrigated area I), and the error bar denotes the ±95% credible intervals.Asterisks signify the significance level of the difference between individual events and corresponding compound events.* * * indicates a significance level of 1%, * * indicates a significance level of 5%, and * indicates a significance level of 10%.

Figure 4 .
Figure 4.The Bowen ratio anomaly (BA) under compound hot-dry (a) and cold-wet (b) events in China from 1990 to 2016.In the main plots, the compound events impact on BA for any individual county on the map is the BA averaged from compound events years during the period.In the subplots, the bar denotes the average BA grouped by the low and high thresholds of P and I at the county level, and the error bar denotes the ±95% credible intervals.The low and high thresholds of P and I are consistent with figure3.Asterisks signify the significance level of the difference between individual events and corresponding compound events, and * * * indicates a significance level of 1%.

Figure 5 .
Figure 5. Relationship of Bowen ratio anomaly (BA) and maize yield anomaly (YA) under compound hot-dry (a), (b) and cold-wet (c), (d) events in counties of dry (a), (c) and wet (b), (d) regime.The bar denotes the average YA grouped by BA, and the error bar denotes the ±95% credible intervals.The low and high thresholds of P and I are consistent with figure 3.