Spatiotemporal Heterogeneity and Zonal Adaptation Strategies for Agricultural Risks of Compound Dry and Hot Events in China’s Middle Yangtze River Basin

Abstract

Compound dry and hot events or extremes (CDHEs) have emerged as major climatic threats to agricultural production and food security in the middle reaches of the Yangtze River Basin (MRYRB), a critical grain-producing region in China. However, agricultural risks associated with CDHEs, incorporating both natural and socio-economic factors, remain poorly understood in this area. Using a Hazard-Exposure-Vulnerability (HEV) framework integrated with a weighting quantification method and supported by remote sensing technology and integrated geographic data, we systematically assessed the spatiotemporal dynamics of agricultural CDHE risks and corresponding crop responses in the MRYRB from 2000 to 2019. Results indicated an increasing trend in agricultural risks across the region, particularly in the Poyang Lake Plain (by 21.9%) and Jianghan Plain (by 9.9%), whereas a decreasing trend was observed in the Dongting Lake Plain (by 15.2%). Spatial autocorrelation analysis further demonstrated a significant negative relationship between gross primary production (GPP) and high agricultural risks of CDHEs, with a spatial concordance rate of 52.6%. These findings underscore the importance of incorporating CDHE risk assessments into agricultural management. To mitigate future risks, we suggest targeted adaptation strategies, including strengthening water resource management and developing multi-source irrigation systems in the Poyang Lake Plain, Dongting Lake, and the Jianghan Plain, improving hydraulic infrastructure and water source conservation capacity in northern and southwestern Hunan Province, and prioritizing regional risk-based adaptive planning to reduce agricultural losses. Our findings rectify the longstanding assumption that hydrological abundance inherently confers robust resistance to compound drought and heatwave stresses in lacustrine plains.

1. Introduction

With ongoing global warming, the frequency and intensity of extreme weather events, such as heatwaves, floods, and droughts, have significantly increased. Particularly, compound extreme events—in which two or more hazards occur simultaneously or successively—have become more frequent since the 1990s, further amplifying future climate risks [1,2,3,4,5]. Among these extreme events, droughts and heat are especially destructive. Heatwaves are characterized by prolonged periods during which temperatures exceed specific thresholds, while droughts denote sustained water shortages resulting from imbalanced water budgets. When drought and heat events coincide to form compound dry and hot events (CDHEs), their synergistic impacts often result in severe water scarcity, substantial crop yield losses, food insecurity, and even threats to human health and safety [6,7,8,9,10,11].

The latest IPCC (AR6) report indicates that the frequency, duration, and intensity of CDHEs have significantly increased in many regions globally [2]. For instance, severe CDHEs occurred in North America, Western Europe, and China’s Yangtze River Basin during 2021–2022 [12,13,14]. Such compound extremes have led to disproportionately catastrophic consequences for water resources, food and energy security, human health, and ecosystem stability [15,16,17,18,19]. Examples include cumulative global grain crop losses of approximately 3 billion tonnes, economic damages totaling €3.3 billion in Europe in 2018, around 5500 fatalities in Russia in 2010, and water shortages affecting over 52 million people in southern China during the summer of 2022 [20,21,22,23,24]. Compared to single extreme events, CDHEs impose significantly greater impacts on both the natural environment and human society. Furthermore, measuring, predicting, and defining CDHEs remain considerable challenges [25,26]. Therefore, a comprehensive risk assessment of CDHEs and a deeper understanding of their influencing factors are crucial steps toward mitigating the negative impacts associated with these disasters.

Climate risk assessment is crucial for understanding climate-change-induced risks and implementing effective mitigation and zonal adaptation strategies in advance [27]. The IPCC (AR5 and AR6) proposes a comprehensive risk-assessment framework, integrating three core components: hazards, exposure, and vulnerability (HEV) [28]. Unlike reactive approaches, this proactive method emphasizes anticipating and managing risks before they materialize, which is particularly important for compound dry and hot events [29]. The IPCC framework defines risk as the probability of adverse impacts on humans and ecosystems, resulting from interactions among hazards, exposure, and vulnerability [30]. Within this framework, hazards represent events caused by CDHEs that may lead to substantial losses to life, property, or ecosystems [31,32]. However, hazards alone do not constitute risk; rather, it is the system’s exposure—the extent to which populations, livelihoods, and assets exist within hazard-prone areas—that determines the emergence of risk [33]. Vulnerability further reflects the susceptibility and adaptive capacity of people or ecosystems, capturing both physical and socio-economic dimensions, and influences their ability to resist, withstand, or recover from impacts [34]. Combining these elements enables a holistic understanding and assessment of climate risk, overcoming the limitations of traditional single-perspective approaches, thus enhancing regional resilience to CDHEs.

The IPCC risk framework has been extensively applied in drought risk assessments worldwide, including regional studies in the Middle East (Iran), Africa, and the Indian Himalayas [27,34,35,36,37]; agricultural drought evaluations in Thailand’s Mongkhon River Basin and northern New South Wales [38]; global drought risk mapping and regional studies in the Caribbean [39]; and drought and water resource assessments in China’s Beijing–Tianjin–Hebei region, the Yellow River Basin, and the Huaihai Plain [22,40]. Additionally, assessments projecting future drought risks based on this framework have been widely conducted [35,41,42,43,44,45]. However, these studies primarily focus on single drought events, with limited attention given to CDHEs. Current CDHE research mostly emphasizes event monitoring, causality analyses, and adverse impacts [46,47,48,49,50]. Few scholars have systematically assessed CDHE risks from an integrated perspective. For instance, Tabari and Willems. evaluated global future CDHE impacts using the IPCC framework but did not consider agricultural adaptive capacities [42]. Zhang et al. [51] assessed historical and future CDHE risk in China but overlooked contributions from individual risk components. Consequently, assessments neglecting sector-specific user demands and the relative contributions of hazard, exposure, and vulnerability may result in inadequate adaptation planning. Therefore, conducting comprehensive CDHE risk assessments that fully incorporate all risk components is critically important.

This study focuses on the MRYRB in southern China, aiming to comprehensively assess the agricultural risks of CDHEs based on the HEV framework. The MRYRB features diverse topography, a humid climate, and abundant water resources with highly heterogeneous spatial and temporal distributions. Agricultural production in the MRYRB primarily centers on rice cultivation. As one of China’s five primary grain-producing regions [52], it encompasses three major agricultural plains: the Poyang Lake Plain, Jianghan Plain, and Dongting Lake Plain. Rice cultivation is particularly intensive in this region, contributing about one-third of China’s total rice production and accounting for approximately 12% of national agricultural output [53]. However, the frequency of CDHEs in the Yangtze River Basin has significantly increased in recent years [54], occurring approximately every 4–5 years [55]. In 2022, the region experienced severe CDHEs, resulting in record drought conditions that left 6.09 million hectares of farmland unirrigated and caused economic losses totaling 51.28 billion yuan [55]. Notably, the agricultural systems within the MRYRB’s three major plains are highly vulnerable, given their heavy reliance on rainfall and natural water bodies for irrigation [56]. CDHEs dramatically reduce river and lake water resources, thereby limiting irrigation availability and significantly lowering crop yields. Thus, accurately assessing the risks posed by CDHEs is critical for developing effective regional agricultural adaptation strategies.

This research focuses on CDHEs, comprehensively considering the interactions between various risk components and their respective contributions to assess the agricultural risks of CDHEs based on remote sensing technology and integrated geographic data. Specifically, the main objectives of this study are as follows: (i) to assess the spatial distributions of CDHE hazard, exposure, vulnerability, and overall agricultural risks; (ii) to analyze crop responses to CDHE risks; and (iii) to recommend targeted adaptation strategies to mitigate agricultural risks associated with CDHEs.

2. Study Area and Data

2.1. Study Area

The MRYRB refers to the inter-regional watershed that originates at Yichang in Hubei and terminates at Hukou in Jiangxi, encompassing six provinces: Hubei, Hunan, Jiangxi, Shaanxi, Henan, and Guizhou. This region is an important part of the Yangtze River Economic Belt in China. It encompasses several sub-basins including the Han River Basin, the Yuan River Basin, the Xiang River Basin, and the Gan River Basin (Figure 1). The MRYRB are located within the subtropical humid monsoon climate zone, where summer temperatures can reach 40 °C [57]. Due to the abundant rainfall and high temperature in this region, it has become one of the major agricultural production areas and water sources in China. Rice is the primary food crop in this region, predominantly cultivated in the three major plains of Jianghan, Dongting Lake, and Poyang Lake (see areas 1, 2, and 3 in Figure 1).

2.2. Data

The data utilized in this study are divided into three distinct categories: meteorological data, satellite remote sensing data, and socioeconomic data, with the latter including the Gross Domestic Product (GDP), population, cropland land, and the proportion of irrigated cropland (

Table 1. Data catalog.

Data Name	Data Type	Spatial Resolution	Data Source
Meteorological data	netcdf	10 km	European Centre for Medium-Range Weather Forecasts
GPP	Tiff	500 m	MODIS (https://modis.gsfc.nasa.gov/, accessed on 12 December 2024)
Population	Tiff	1 km	Chinese Academy of Sciences
GDP	Tiff	1 km	Chinese Academy of Sciences
Cultivated land	Tiff	30 m	[60]
Irrigation data	Tiff	500 m	[11]

). Meteorological data employs precipitation and 2 m temperature (T2m) variables from ERA5_Land. ERA5_Land constitutes an iteration of ECMWF reanalysis 5th generation (ERA5). It is based on the ERA5 dataset and provides longer and higher-resolution time series data compared to ERA5. The data has a time resolution of one month and a spatial resolution of 10 km. ERA5_Land data have been widely utilized in the meteorological field [58,59]. The vegetation Gross Primary Productivity (GPP) is derived from the MODIS version 6 total primary productivity product, with a time frame selected of 2001–2019. Cultivated land data were adopted from the 30 m spatial resolution dataset of China 1986–2021, which was published by Tu [60]. Irrigation data were obtained from the 500 m resolution dataset of China from 2000–2019, as reported by Zhang [11]. To maintain the same spatial resolution as the other data, we set the size of each fishing grid to 5 km. The proportion of irrigated cropland within this grid was subsequently calculated. Population and Gross Domestic Product (GDP) density are adopted from the 1 km resolution data published by the web of the Chinese Academy of Sciences. Exposure and vulnerability were calculated using 2005 and 2015 data as representative of the 2000–2009 and 2010–2019 periods, respectively. All of the above raster data were uniformly resampled to 5 km and are preprocessed by using the maximum–minimum standardization method.

3. Methodology

This research adopted the framework proposed by the IPCC (AR5), which consists of the relationship between the magnitude of risk and the occurrence of hazardous events, the target vulnerability, and the exposure range of the affected object [28]. CDHEs are natural hazards that occur simultaneously with drought and heatwave hazards, which are much more catastrophic. In this paper, we define the agricultural risks of CDHEs including three components of the CDHE hazard, environmental vulnerability, and exposure. Among these, the CDHE hazard is defined as the frequency of composite extreme events occurring during the corresponding calculation period; environmental vulnerability is defined as the ability to adapt to disasters, composed of GDP and the arable land irrigation rate; and exposure is defined as the degree to which the arable land and population are exposed to CDHEs (Figure 2). The risk formula is expressed as follows, Equation (1):

$$R=\alpha H·\beta V·\gamma E$$

(1)

In the formula, R refers to the agricultural risks of CDHEs. $H$ refers to the frequency of CDHE hazards in a certain period. V denotes vulnerability, which represents a region’s capacity to withstand hazards. E refers to the exposure exposed to CDHEs. $\alpha$, $\beta ,$ and $\gamma$ are their weights of risk, respectively. We classify risk levels according to the natural breaks method.

3.1. Hazard

Hazards are defined as the frequency of occurrence of a CDHE during the corresponding calculation period. In this study, a dry event is defined as monthly precipitation below the 25th percentile threshold, and a heat event is defined as a monthly average temperature exceeding the 75th percentile threshold. The thresholds were calculated based on the 1961–1990 normal climate period, as recommended by the World Meteorological Organization [61]. When both drought and heat occur in the month, it is defined as one dry–hot compound hazard event. In this article, the research period from 2000 to 2019 is divided into two calculation periods, namely 2000 to 2009 (T1) and 2010 to 2019 (T2).

3.2. Vulnerability

Vulnerability is commonly defined as the degree to which the structure and function of a system are impaired when subjected to internal or external disturbances. Many current studies on drought risk have included the regional GDP and cropland irrigation rate as vulnerability indicators to characterize the adaptive capacity of cropland to disasters, Equation (2) [44]. GDP is a key indicator of a region’s economic society, and larger values indicate that the region has invested more in disaster resilience and is more resilient to disasters [36]. The cultivated land irrigation rate indicates the ratio of cultivated land that can be irrigated in each grid, and a larger value indicates a stronger resilience to the hazard, and all of the above indicators are negatively correlated with vulnerability. In this study, the two indicators are considered to be equally weighted.

$$V={\displaystyle \frac{({\mathrm{V}}_{GDP}+{\mathrm{V}}_{\mathrm{i}\mathrm{r}\mathrm{r}})}{2}}$$

(2)

where $V$ represents vulnerability; ${\mathrm{V}}_{GDP}$ and ${\mathrm{V}}_{\mathrm{i}\mathrm{r}\mathrm{r}}$ represent the normalized result of GDP and cropland land irrigation rate, respectively.

3.3. Exposure

Exposure refers to the degree to which a subject is exposed to a hazardous event. In this study, we focus on agricultural risks and select cropland areas and population density exposed to CDHEs as the indicators of exposure. Previous studies, such as that by Dong et al. [62] and Babel et al. [38], have extensively explored the relative weights of cropland and population under climate-related hazards. In this study, a weighted approach is used to calculate overall exposure levels, with the following formula Equation (3):

$$E={\epsilon P}_c+\theta P_p$$

(3)

where $E$ represents the exposure level, $P_c$ denotes the proportion of cropland in each grid, and $P_p$ represents the normalized population density. $\epsilon$ and $\theta$ are the corresponding risk factor weights under drought climate exposure. Based on previous research, we found that both population and cultivated land are significant for exposure [38], so we assigned equal weight to both.

3.4. Trend Analysis and Testing

The Theil–Sen Median (TS) method is a robust non-parametric statistical approach for trend calculation, often employed in the trend analysis of long-term time series data [63]. The Mann–Kendall test (MK) is a non-parametric hypothesis-testing method used to determine the significance of trends in time series data [63]. It identifies significant monotonic trends at p < 0.05 when Zs < 1.96. In this analysis, the TS and the MK were employed to investigate the interannual trends of precipitation and T2m from 2000 to 2019.

3.5. Analytic Hierarchy Process (AHP)

The AHP is a decision analysis method that combines qualitative and quantitative approaches. It is a widely utilized method in the context of multi-objective, multi-factor, unstructured, and complex decision-making problems, as well as in the field of risk assessment [38]. We established a hierarchical structure with CDHE risk as the goal layer, hazard, exposure, and vulnerability as the criterion layer, and the next level of indicators as the alternative layer. Based on this, we constructed a judgment matrix using the scaling method, calculated the weight vector, and performed a consistency test. The results are shown in

Table 2. AHP hierarchy analysis results.

Term	Characteristic Vector	Weight Value (%)	Largest Eigenvalue	CI Value
Hazard	1.366	45.547
Exposure	0.700	23.346	3.001	0
Vulnerability	0.933	31.107

below. The weights for hazard, exposure, and vulnerability are 0.46, 0.23, and 0.31, respectively.

3.6. Spatial Autocorrelation and Hotspot Analysis

The primary objective of spatial autocorrelation is to ascertain whether the study variables manifest a tendency towards spatial clustering across the designated study area. This is achieved by calculating the relationship between the variance observed at a specific location and the variance observed at neighboring locations, to assess the presence of spatial dependence [64]. In this study, Moran’s $I$ is employed as a means of reflecting the spatial variation of compound dry–hot risk and vegetation GPP. The formula is as follows, Equation (4):

$$I={\displaystyle \frac{\sum _{i=1}^m\sum _{j=1}^m{W}_{ij}(x_i-\overline{x})(x_j-\overline{x})}{B^2\sum _{i=1}^m\sum _{j=1}^m{W}_{ij}}}$$

(4)

where $I$ is the global autocorrelation index; $m$ is the number of grid cells within the study area; $x_i$, and $x_j$ are the attribute values of the elements at locations i and j (with I ≠ j); $W_{ij}$ is the spatial weight matrix; and B is the variance of the elements. Values of $I$ > 0, $I$ < 0, and $I$ = 0, respectively, indicate positive and negative spatial correlations and no spatial correlation.

Hotspot analysis explores local spatial clustering characteristics by calculating each feature’s Getis-Ord G* statistic to determine whether there are clusters of high or low values and to identify their locations, Equations (5) and (6). This analysis reveals the spatial distribution patterns of hot spots or cold spots [65]. The calculation formula is as follows:

$$G_i^{*}={\displaystyle \frac{\sum _{j=1}^m{W}_{ij}{M}_j}{\sum _{i=1}^m{M}_i}}$$

(5)

$$Z\left(G_i^{*}\right)={\displaystyle \frac{G_i^{*}-E\left(G_i^{*}\right)}{\sqrt{\mathrm{v}\mathrm{a}\mathrm{r}\left(G_i^{*}\right)}}}$$

(6)

where Gi* is the local autocorrelation index, $W_{ij}$ is the spatial weight matrix, and $M_i$ and $M_j$ are the values of the research target i and j, respectively. E(Gi*) represents the expected value of Gi*, var(Gi*) represents E(Gi*) and is the mathematical expectation of Gi*, and var(Gi*) is the variance of Gi*. If Z(Gi*) is significantly positive, it means that there is a high-value aggregation of elements.

4. Result

4.1. Hazard of CDHEs

The spatial and temporal variations in precipitation and T2m in the MRYRB from 2000 to 2019 were significant. T2m showed a significant increasing trend in general (Figure 3a), and spatially, the warming of the three plains regions was slightly lower than that of the surrounding areas. Precipitation showed a significant decreasing trend, with the northwest and south-central regions experiencing greater decreases than other regions, reaching 1.23–1.72 mm/a and 1.75–2.2 mm/a, respectively (Figure 3b). The larger warming trend is consistent with the regions with larger decreases in precipitation (Figure 3b), which indicates that heat and drought often accompany each other. The precipitation also showed a significant decreasing trend in the three major plains, with a larger decrease in the Dongting Lake and Poyang Lake areas relative to the Jianghan Plain. Precipitation decreased at an annual rate of 0.085 mm/a and T2m increased at a rate of 0.023 °C/a in the three plains areas (Figure 4).

In this study, the frequency of CDHEs was counted for two historical periods, 2000–2009 and 2010–2019. Figure 5a presents the frequency distribution of CDHEs in China during 2000–2009, indicating that the frequency of hazard of CDHEs was higher in the south-central, southwestern, and northeastern parts of the study area. The spatial distribution pattern of CDHEs in the 2010–2019 period was found to be similar to that of the 2000–2009 period (Figure 5b). During the 2010–2019 period, the frequency of CDHEs exhibited a marginal increase compared to the 2000–2009 period. Approximately 51.4% of the region demonstrated an increase in CDHE frequency, while 40.9% exhibited a decrease. The frequency of CDHEs increased from 3.7% to 3.9% over the study area. The increase was found to be significant in central and eastern Jiangxi Province (118–139%), northeastern Hubei Province (93–207%), and southwestern Hunan Province (33–58%), while the increase was not significant in other regions. It is also noteworthy that CDHEs were found to be equally prevalent in the historically drought-prone regions of Poyang Lake and Dongting Lake.

4.2. Vulnerability of CDHEs

Based on the cropland irrigated rate and regional GDP, the magnitude of vulnerability in the MRYRB was calculated for the periods 2000–2009 and 2010–2019. The results show that the overall spatial pattern of vulnerability in 2000–2009 and 2010–2019 (Figure 6b) is similar but with significant regional differences. The vulnerability around Dongting Lake and Poyang Lake areas and in the northeast of Hubei Province is lower, with values distributed from 0.55 to 0.75, which is attributed to the high proportion of irrigation district construction and GDP in these areas. The higher vulnerability in the northwestern, southwestern, and southern parts of the study area may result in poorer resistance to CDHEs. Figure 6c shows that 54% of the regional vulnerability increased from 2000–2009. The vulnerability of the Jianghan Plain, Dongting Lake Plain, and Poyang Lake Plain, is decreasing, decreasing to 0.67%, 2.55%, and 0.52%, respectively, and the construction of irrigation districts within the three major plains may be a potential factor for the decrease in vulnerability. The significant spatial difference in vulnerability indicates the imbalance in the development of regional economic level and irrigation district construction in the past 20 years.

4.3. Exposure of CDHEs

Exposure under CDHEs was calculated based on the distribution of cropland and population density. Areas with high exposure in the 2000–2009 and 2010–2019 periods were concentrated in the northern, central, southern, and eastern zones of the study area, with a spatial pattern consistent with the distribution of cropland (Figure 1). A comparison of exposures in these two periods showed a decrease in exposure in 2010–2019, with 61% of the zones experiencing a decrease in exposure. The northwestern, southwestern, southern, and northeastern parts of the study area experienced different degrees of exposure reduction, ranging from 5 to 27%, 10 to 36%, 9 to 67%, and 16 to 24%, respectively. Exposure in the Poyang Lake plain areas also decreased, with reductions ranging from 5% to 8% (Figure 7c).

4.4. Risks of CDHEs

To quantify the different weights of each risk component based on the CDHE risk, this study calculated the weights of hazard, exposure, and vulnerability factors based on the AHP method, which were 0.46, 0.23, and 0.31 respectively. The overall spatial pattern of agricultural risks of CDHEs in the 2000–2009 and 2010–2019 periods is consistent, but the spatial differentiation is evident. Higher risks occur in the middle reaches of the Xiangjiang River, the Dongting Lake Plain, and the Poyang Lake Plain in Hunan Province. Comparing the changes between the two periods, it was found that the agricultural risks of CDHEs increased in the southern part of Hunan Province, the Poyang Lake area, the middle reaches of the Gan River, and the Hanjiang River basin (Figure 8c). In particular, the middle reaches of the Gan River had the largest increase in CDHE risk, which increased by 40–170% compared to the period 2000–2009. Among the three major plains, the CDHE risk increased by 9.9%, −15.2%, and 21.9% in the Jianghan Plain, Dongting Lake Plain, and Poyang Lake Plain, respectively.

We also mapped the risk of agricultural risks of CDHEs for T1, T2 (Supplementary Figure S1), and the period of 2000–2019 (Figure 9). The risk level of CDHEs for the period 2000–2019 shows a relatively high risk for the three major plains regions and the southern and southwestern parts of Hunan Province, while the risk for the other regions will be relatively low. This suggests that the three major plains regions face more severe climate hazards than other regions (e.g., northeastern Hubei).

4.5. GPP Response to the Agricultural Risk of CDHEs

By analyzing the spatial response of GPP to the CDHEs in the three major plains, we found that crop GPP showed a significant increasing trend over the period 2000–2019 (Figure 10c), fixing plant carbon at an average rate of 6.07g/C/m² per year. Figure 10a shows that crop GPP was negatively correlated with precipitation and positively correlated with T2m, with atmospheric temperature having a greater effect on GPP than precipitation ($p$= 0.09 < $p$ = 0.32) (Figure 10b). Neither T2m nor precipitation passed the 95% significance test, suggesting that natural precipitation and atmospheric temperature are not the only sources of the hydrothermal conditions required for crop production, especially precipitation. Artificial irrigation may be one of the factors contributing to the non-significance of crops to natural precipitation.

Previous studies have indicated that a single CDHE does not necessarily impact vegetation GPP. However, CDHE risk that integrates environmental and socioeconomic factors can quantify the adverse effects of hazards on GPP, thereby enabling the direct estimation of vegetation GPP losses through quantified risk. We further explored the spatial pattern of vegetation multi-year average GPP and CDHE risk. Spatially, the risk of CDHEs showed a significant positive correlation, with Moran’s index reaching 0.89. Large areas of high-value aggregation were presented in southeastern Hunan Province, northeastern Jiangxi Province, and the three major plain areas, and localized areas of high-value aggregation were presented in southwestern Hunan Province and central and eastern Jiangxi Province, suggesting that the spatial correlation of CDHE risk in the above areas is high and occurs in clusters (Figure 11). In addition, we found that the spatial distribution of vegetation GPP growth also showed a significant positive spatial correlation, with a Moran’s index of 0.53, and spatial clustering of high values (Figure 12c). In the three major plains regions, GPP showed highly significant low–low aggregation areas, which spatially coincided with the high aggregation of CDHE risk, with 52.6% perfectly coinciding and 34.5% of the regions with an insignificant CDHE risk and GPP.

5. Discussion

5.1. Temporal and Spatial Variations in the Risk of CDHEs

The results on CDHE hazards show that it has shown an increasing trend in the past 20 years, and the lake area is a high-incidence area of CDHE hazards. Previous research has also found that temperature is a more direct factor contributing to drought events. Both single drought events and CDHEs have exhibited an increasing trend since 2000 [50,54,66]. The spatial pattern of these CDHEs is unlikely to shift significantly under future climate warming, as it is predominantly driven by atmospheric circulation and the global sea surface temperature. Compared with the Dongting Lake Basin and the Hanjiang River Basin, the Poyang Lake Basin has experienced a significant increase in disaster frequency in the last decade, which means that the probability of compound dry and hot hazards in the Poyang Lake Basin will be much higher in the future than in other basins.

Changes in vulnerability and exposure are concentrated primarily in the capital cities Wuhan, Changsha, and Nanchang, as well as their adjacent regions, underscoring the critical role of socioeconomic development in hazard adaptation and exposure management. This is because socioeconomic factors indirectly influence regional investments in drought mitigation and the trajectory of cropland development. In the northwestern, southwestern, southern, and northeastern parts of the MRYRB, the exposure decreased significantly, and the reduction of cropland was the main cause of the exposure reduction. However, according to the field study in the plain area of Poyang Lake [67], the abandonment of cropland or the non-fertilization of cropland was the main reason for the reduction of cropland in the last decade. Analyzing all factors together, the risk of CDHEs has been increasing over the past 20 years, especially in the Poyang Lake Basin and the Jianghan Basin area. This emphasizes the need to develop appropriate measures to address the risk of CDHEs.

5.2. Impact of CDHEs on GPP

In this study, identifying CDHEs essentially involves recognizing precipitation and T2m that exceed certain thresholds. Therefore, precipitation and T2m are the primary factors influencing GPP. Our research finds that in the three major plains regions, GPP is negatively correlated with precipitation and positively correlated with air temperature, which is consistent with the results obtained by Chen et al. [68] in their study of the Mekong River Basin. This difference may be attributed to the varying responses of vegetation to water and heat. The three major plains are primarily devoted to rice cultivation, with irrigation and precipitation serving as the primary sources of water for rice. Rice has different water requirements at different stages of its growth cycle. When the soil reaches its optimal field water-holding capacity, excess precipitation will instead inhibit crop growth [69,70]. The irrigation areas in the region are well constructed to ensure that rice can be adequately irrigated. Conversely, in conditions of adequate soil moisture, an increase in ambient temperature will enhance the efficiency of crop carbon sinks, whereas precipitation may not necessarily be conducive to crop carbon sequestration. Therefore, in the humid zone, among the natural factors affecting rice growth, temperature is a more critical influence than precipitation, and drainage construction should be emphasized.

Establishing the connection among the CDHE hazard, GPP, and CDHE risk is the core objective of this paper. In the three major plains with intense human activities, CDHE hazards do not necessarily impact GPP. Therefore, it is difficult to directly explain changes in GPP using CDHEs. However, the CDHE risk transcends this uncertainty and can directly account for changes in GPP under hazard events. The results of the hotspot analysis showed that the three major plains and the Xiangjiang River Basin were hotspot regions for CDHE risk, while GPP showed significant cold spot regions, indicating a strong spatial correlation between CDHE risk and GPP. A higher CDHE risk corresponded to lower GPP in most regions, suggesting that the level of risk can to some extent characterize vegetation growth. In the three major plains, the ability of risk to characterize GPP showed differences. In the Dongting Lake and Poyang Lake plains, the results of the cold hotspot analyses could be matched in 52.6% of the regions, and only 30% of the regions of the Jianghan Plain (the non-lake plains) could be matched, suggesting that the risk of CDHEs is stronger in characterizing the change in GPP in the lake plains than in the non-lake plains. Furthermore, the correlation between agricultural risk and GPP in the MRYRB has not been established before. This study clarifies the spatial coupling relationship between agricultural risks of CDHEs and vegetation productivity by establishing the correlation between the two, which provides a basis for differentiated agricultural CDHE risk-management measures in different regions in the future.

5.3. Agricultural Adaptation Strategies Under CDHEs

The results show that the three major plains in MRYRB, where intensive crops are grown, are exposed to high agricultural CDHE risks, especially the Dongting Lake and Poyang Lake plains around the lakes, and that the agricultural sector should take different adaptation measures according to the risk levels and trends in different regions. For example, in the three major plains, existing irrigation and drainage capacities need to be re-evaluated, and more effective water-resource-management measures need to be developed to cope with the complex climate in the future. While the Xiangjiang and Yuanjiang river basins face a higher risk, most of the areas are located in mountainous areas with high forest cover, and their nature-based solutions (e.g., maintaining forest ecosystems stable and high water conservation conditions) are feasible responses. Non-structural adaptation strategies such as drought or heat-tolerant crops and varieties, improved agricultural management, and insurance could be considered in areas such as Guizhou Province and southwestern Hunan Province [71]. Our findings can inform the development of disaster-reduction or adaptation strategies based on either structural measures (e.g., effective water resource management infrastructure, drought warning systems) or non-structural measures (e.g., adjusting crop varieties, purchasing climate insurance for crops) [72,73,74].

5.4. Limitations and Future Work

This study provides the first comprehensive assessment of the agricultural risks of CDHEs in China’s humid zone based on remote sensing technology and integrated geographic data, which is important for agricultural drought management in similar climatic zones; however, this paper still has the following limitations: (1) in analyzing the spatial and temporal characteristics of the risk of CDHEs, although China has introduced active policies for farmland protection and the population to minimize their variations, there is vulnerability and exposure on time scales; (2) crop growing seasons and varieties (e.g., response of rice and maize to CDHEs, etc.) were not taken into account when selecting risks for assessment, and since different crops respond to CDHEs at different times, the differentiated treatment of the crop response to CDHEs will be a very worthwhile element to explore in the future; and (3) the study only analyzed the relationship between the risk of CDHEs and vegetation GPP in spatial correlation and aggregation and did not quantify the impact of the risk on vegetation GPP, which limits the application value of the risk assessment to some extent. The risk assessment of CDHEs mainly relies on the selection of indicators, and the selection of indicators will have a certain impact on the results of the risk assessment. The six indicators selected in this paper do not cover all aspects of risk assessment, but they can embody the key components of agricultural risk formation. In the future, under the premise of data availability, environmental, social, and economic assessment indicators should be selected from different perspectives to more comprehensively assess the changes in CDHE risk, future trends, and risk metrics, which will be topics worthy of in-depth discussion in the future.

6. Conclusions

This study assessed agricultural risk of CDHEs in the MRYRB based on remote sensing technology and integrated geographic data, enhancing our understanding of CDHEs in humid climate regions. An analysis of the spatiotemporal pattern of agricultural risk of CDHEs during 2000–2019 revealed an increasing trend in the MRYRB. The main conclusions are as follows:

(1)

During the period 2000–2019, the agricultural risk level of CDHEs was high in the Poyang Lake Plain and the Dongting Lake Plain, moderate in the Jianghan Plain, and extremely high in southern Hunan Province. Over the last decade of the study period, agricultural risk increased by 21.9% in the Poyang Lake Plain and 9.9% in the Jianghan Plain, while it decreased by 15.2% in the Dongting Lake Plain.

(2)

Both the agricultural risk of CDHEs and GPP exhibited significant spatial aggregation. In the MRYRB, high-risk areas overlapped spatially with low GPP areas at a rate of 52.6%, indicating a clear spatial association between the two. This was particularly evident within the three major plains, where vegetation productivity was significantly suppressed under these conditions.

(3)

To cope with the agricultural risk of CDHEs, it is necessary to re-evaluate the existing irrigation and drainage capacities and develop a multi-source irrigation system in the three major plains areas. In southern Hunan province, where the forest coverage rate is high, nature-based solutions should be adopted after strengthening the construction of hydraulic infrastructure, such as maintaining the stability of forest ecosystems and high water retention conditions. In addition, building a targeted hazard early-warning system is also very important to withstand CDHE risk based on structural measures.