Spatiotemporal Evolution of Compound Dry–Hot Events and Their Impacts on Vegetation Net Primary Productivity in the Yangtze River Basin

Abstract

Compound dry–hot events increasingly threaten ecosystem productivity under global warming. Using ERA5-Land and MODIS NPP (2002–2024) for the Yangtze River Basin, we built climate indices and developed a Copula-based standardized compound dry–hot index (SCDHI) to detect events and examine spatiotemporal patterns. Trend and correlation analyses quantified NPP sensitivity and lag, and an NPP–SCDHI coupling framework assessed resistance and resilience across major vegetation types. Basin-wide monthly NPP increased slightly, while SCDHI decreased, indicating a warmer and drier tendency. Under dry–hot conditions, NPP was mainly negatively related to event intensity in the upper basin but positively related across much of the middle–lower plains. The mean NPP response time was approximately 2 months, with forests and croplands typically lagging 2–3 months. Under extreme stress, forests showed high resistance but limited recovery, whereas shrublands showed moderate resistance and low resilience. Cultivated vegetation exhibited the lowest resistance and weak resilience, grasslands had low resistance but relatively rapid recovery, and alpine vegetation showed moderate resistance and the highest resilience. Cultivated vegetation and grasslands may therefore represent high-risk types for ecological management.

1. Introduction

Under global warming, extreme climate events exhibit significant increases in frequency, intensity, and duration, and drought and heat events not only occur in isolation but more often co-occur or occur in succession in a compound form, forming so-called compound dry–hot events [1,2]. Such events are often characterized by “high temperatures superimposed on drought” and “short-duration, high-intensity, concentrated outbreaks”, and can markedly weaken vegetation photosynthesis and carbon sequestration capacity over short timescales, triggering concurrent degradation of ecosystem structure and function at the regional scale and posing severe challenges to agricultural production, water-resource security, and ecological security [3,4]. Therefore, under the context of global change and increasingly frequent extreme events, systematically characterizing the spatiotemporal evolution of compound dry–hot events and quantitatively assessing their impacts on terrestrial ecosystem productivity have become research frontiers in the fields of climate change and ecohydrology [3,5].

Net primary productivity (NPP) is a key indicator of the balance between vegetation photosynthesis and respiration, directly reflecting both the capacity of ecosystems to fix atmospheric CO₂ and their level of energy input, and constituting a fundamental variable for assessing regional carbon budgets and ecosystem functioning [6,7]. Numerous studies have demonstrated that climatic factors such as precipitation, temperature, and radiation are among the dominant drivers of the spatiotemporal variability of NPP, whereas drought and high-temperature extremes can substantially suppress NPP through mechanisms including water stress, stomatal closure, and leaf scorching, and may even trigger large-scale yield losses and vegetation degradation [8,9,10]. Most existing studies have examined the effects of either drought alone or heatwaves alone on vegetation growth and NPP, often relying on traditional drought indices and temperature anomaly metrics for correlation analyses. Although these efforts have yielded abundant insights, relatively limited attention has been paid to the synergistic mechanisms of “heat–drought” interactions, making it difficult to fully capture the actual impacts of compound dry–hot events on vegetation productivity [11,12,13].

From the perspective of monitoring and quantitative assessment, traditional drought and heat indices are primarily constructed from single variables and thus have limited ability to simultaneously characterize the joint distribution of moisture deficits and temperature anomalies [14]. In recent years, multivariate drought indices based on Copula theory and the standardized composite dry–hot index (SCDHI) have been progressively developed and applied, enabling the integration of precipitation, potential evapotranspiration, temperature, and other information on a unified probabilistic scale while accounting for the dependence structure among variables, thereby providing a new technical pathway for identifying and quantifying compound dry–hot events [15,16]. However, most existing studies have focused on issues such as drought risk assessment, the evolution of hydrological drought, or agricultural yield losses, and research that couples SCDHI with ecosystem productivity metrics such as NPP to systematically evaluate vegetation responses and recovery capacity to compound dry–hot events remains relatively limited, particularly with respect to large-basin scales and systematic comparisons across multiple vegetation types [15,17,18].

The Yangtze River Basin is a densely populated and economically important region, and it is also critical for China’s ecological conservation and energy, food, and water security [19,20]. Influenced by both the East Asian monsoon and the Tibetan Plateau, the basin has a highly variable climate. Extreme droughts and high temperatures occur frequently. In recent years, droughts and heatwaves have increasingly co-occurred, leading to more compound dry–hot events [20,21,22]. However, existing studies on the Yangtze River Basin have largely focused on precipitation–runoff changes, the spatiotemporal evolution of single drought indices, and overall trends in vegetation productivity, and there is still a lack of systematic understanding of the occurrence patterns and intensity changes in compound dry–hot events and their impact mechanisms on NPP, particularly with respect to the quantitative assessment of vegetation resistance and resilience [23,24,25].

Accordingly, this study aims to examine the spatiotemporal evolution of compound dry–hot events and their impacts on vegetation net primary productivity in the Yangtze River Basin during 2002–2024. Three objectives are addressed in this study. The first objective is to characterize the spatiotemporal patterns and long-term changes in compound dry–hot conditions and to identify regions with intensified dry–hot stress. The second objective is to quantify the spatial variation in the direction, magnitude, and timing of vegetation NPP responses to compound dry–hot variability. The third objective is to assess differences in resistance and resilience among major vegetation types under extreme compound dry–hot events, thereby evaluating ecosystem stability and vulnerability. Together, these analyses provide an integrated assessment of compound dry–hot dynamics and ecosystem productivity responses, with implications for ecological security assessment and adaptive management in the Yangtze River Basin.

2. Materials and Methods

2.1. Study Area

The Yangtze River Basin is located between 90°33′–122°25′ E and 24°30′–35°45′ N, with a drainage area of about 1.8 × 10⁶ km², accounting for roughly one-fifth of China’s land area. Its main stream is approximately 6300 km long, with a total elevation drop of more than 5400 m from the source to the estuary, and flows eastward across the Tibetan Plateau, the Hengduan Mountains, the Sichuan Basin, and the middle–lower reaches plains, making it one of China’s most water-abundant and ecologically crucial river basins.

Controlled by China’s three-step topographic pattern, the basin exhibits an overall west–east elevation gradient, with higher terrain in the west and lower terrain in the east. The upper reaches are dominated by plateaus and deeply incised valleys, the middle reaches by mountains, hills, and basins, and the lower reaches by extensive alluvial plains and lake–wetland systems. This topographic transition produces pronounced relief and strong spatial heterogeneity across the basin.

Climatically, the Yangtze River Basin is generally situated in the subtropical monsoon zone, with a multi-year mean temperature of about 12.9 °C and a mean annual precipitation of approximately 1126.7 mm, exhibiting pronounced monsoonal and seasonal characteristics. Precipitation decreases roughly from the southeast to the northwest, and shows a “coincidence of rain and heat” in time, with flood-season rainfall (May–September) accounting for more than 60% of the annual total.

Marked interannual and intra-seasonal climatic variability, together with frequent extreme high-temperature and precipitation anomaly events, readily triggers regional droughts, floods, and dry–hot anomalies that may occur concurrently or in close succession at the monthly scale throughout the year, providing a favorable background for the occurrence of compound dry–hot events.

Ecologically, the basin hosts diverse vegetation types. Alpine meadows, shrublands, and coniferous forests dominate the upper reaches and northwestern areas, whereas evergreen broadleaved forests, mixed conifer–broadleaf forests, and shrublands prevail in the middle reaches and parts of the southern margins. while croplands, economic forests, and peri-urban artificial vegetation systems are predominant in the lower plains and riparian zones. Forest resources and cropland ecosystems jointly constitute the main body of terrestrial ecosystem productivity in the basin, playing a crucial role in maintaining regional carbon sink functions, soil and water conservation, and biodiversity. However, under the background of global warming, drought and high-temperature events in the Yangtze River Basin show increasing frequency, intensifying severity, and expanding spatial extent, and the potential impacts of compound dry–hot events on vegetation net primary productivity (NPP) and ecosystem stability are becoming increasingly prominent. Therefore, taking the Yangtze River Basin as the study area and systematically analyzing the spatiotemporal evolution of compound dry–hot events during 2002–2024 and their impacts on vegetation NPP is of considerable scientific and practical significance for elucidating the mechanisms of ecosystem responses to extreme climate in typical monsoonal basins in China and for assessing regional ecological security (see Figure 1).

2.2. Data Sources

The primary datasets used in this study include 276 months (2002–2024) of meteorological variables such as potential evapotranspiration, temperature, and precipitation, together with NPP and GPP data. The meteorological data are all derived from the ERA5-Land Monthly Aggregates land reanalysis dataset (dataset ID: ECMWF/ERA5_LAND/MONTHLY_AGGR) provided by the European Centre for Medium-Range Weather Forecasts (ECMWF), with a spatial resolution of 0.1° and a monthly temporal resolution. The GPP data are obtained from the MOD17A3HGF V6.1 dataset (https://appeears.earthdatacloud.nasa.gov/ (accessed on 2 September 2025)), which has a spatial resolution of 500 m and a temporal resolution of 8 days. The NPP data are obtained from the MOD17A2HGF V6.1 dataset (https://appeears.earthdatacloud.nasa.gov/ (accessed on 5 September 2025)), which has a spatial resolution of 500 m and an annual temporal resolution (1 y). In this study, meteorological, GPP, and NPP data were processed using Python 3.10.6 for spatial resampling, and their spatial resolution was uniformly adjusted to 0.25° by applying a mean-resampling method. given that NPP is available only at an annual scale, whereas GPP has an 8-day temporal resolution, the annual NPP was allocated to the monthly scale according to the intra-annual relative contribution of GPP. Specifically, 8-day GPP values were first aggregated to each month, the proportion of monthly GPP to the annual total GPP was then calculated, and monthly NPP was obtained by scaling annual NPP according to these proportions [26,27], which has been widely applied to represent the intra-annual variability of vegetation productivity when monthly NPP products are unavailable.

The vegetation data used in this study are derived from the “1:1,000,000 Vegetation Atlas of China”, downloaded from the Resource and Environment Science and Data Center of the Chinese Academy of Sciences (https://www.resdc.cn/ (accessed on 2 October 2025)). During preprocessing in ArcGIS 10.8, the main steps included projection, alignment, mosaicking, resampling, and clipping of the maps to ensure seamless spatial matching and consistency with the study area. Subsequently, vegetation types were reclassified according to the research objectives. The original vegetation classes were grouped into seven major categories, including cultivated vegetation, grassland, coniferous forest, broadleaved forest, shrubland, mixed conifer–broadleaf forest, and alpine vegetation. In addition, the source shapefile includes an “Others” class, which represents non-vegetated areas (i.e., areas without vegetation cover). This class is displayed in Figure 2 as background and is not counted as a vegetation type.

2.3. Methods

2.3.1. Standardized Temperature Index

The standardized temperature index (STI) is an indicator used to quantify and compare temperature conditions across different regions [28]. STI is commonly used to investigate the effects of climate change on temperature, the frequency and intensity of extreme temperature events, and to evaluate the impacts of temperature anomalies on ecosystems and human activities [29]. The index is first derived by computing the cumulative probability of monthly temperature using a Gamma distribution function, followed by a normal standardization of the results to obtain the final STI values [30]. The STI is calculated as follows

$$STI={\mathsf{\phi }}^{-1}(F_{G\mathrm{a}mma}(x))$$

(1)

where $x$ is the temperature value, $F_{G\mathrm{a}mma}$ denotes the cumulative distribution function of the Gamma distribution, and ${\mathsf{\phi }}^{-1}$ is the inverse standard normal distribution function.

2.3.2. Standardized Precipitation Evapotranspiration Index

The standardized precipitation evapotranspiration index (SPEI) is a drought monitoring index that jointly considers precipitation and potential evapotranspiration. By accounting for these two components, SPEI provides a more comprehensive assessment of drought conditions [31]. The SPEI is calculated as follows

$$D_i=P_i-E_{p,i}$$

(2)

$$F(D_i-\mathsf{\gamma })={\displaystyle \frac{1}{1+{\left(\frac{\mathsf{\alpha }}{D_i-\mathsf{\gamma }}\right)}^{\mathsf{\beta }}}}$$

(3)

$$SPEI={\mathsf{\varphi }}^{-1}(F(D_i))$$

(4)

where $D_i$ is the difference between precipitation and potential evapotranspiration, P_i is precipitation (mm), $E_{p,i}$ is potential evapotranspiration (mm), $\mathsf{\alpha }$ is the scale parameter, $\mathsf{\beta }$ is the shape parameter, $\mathsf{\gamma }$ is the location parameter, $F(D_i)$ is the normalized cumulative probability, and SPEI denotes the standardized precipitation evapotranspiration index. SPEI values greater than −0.5 indicate normal conditions, whereas SPEI values less than −0.5 indicate drought conditions [32].

2.3.3. Standardized Compound Dry–Hot Index

In this study, a pixel-wise standardized compound dry–hot index (SCDHI) is constructed within a Copula-based framework using the multitemporal gridded indices SPEI and STI [15]. Considering that moisture processes exhibit pronounced cumulative effects, whereas high-temperature processes are typically manifested as short-term anomalies at the monthly scale, this study adopts 3-month SPEI as the drought indicator and 1-month STI as the heat indicator. Compared with the 1-month scale, 3-month SPEI better integrates antecedent precipitation and evapotranspiration conditions, characterizes root-zone soil moisture and persistent moisture deficits, and is therefore more representative of the substantive stress on vegetation productivity [28,33,34]. A Copula function is a multidimensional joint distribution function defined on the 0–1 domain, which integrates the marginal distributions of multiple random variables and captures their dependence structure without relying on the specific form of the marginal distributions [35]. The joint probability of compound dry–hot events, P, can be expressed as

$$P(X\le x,Y\ge y)=u-c(u,v)$$

(5)

where X is the climatic water balance (i.e., precipitation minus potential evapotranspiration) and Y is the maximum temperature. The values x and y are the corresponding drought and heatwave thresholds. A compound dry–hot condition is recorded when $X\le x$ and $Y\ge y$. To characterize the joint behavior of drought and heat components on a unified probabilistic scale while allowing flexible dependence structures, copula-based multivariate standardized indices have been widely used in compound-event studies [15,35]. Accordingly, we adopted an Archimedean copula framework for pixel-wise implementation. Considering the large number of pixels and the need for robust parameter estimation, we selected the one-parameter Frank copula, which provides a parsimonious yet flexible representation of dependence and is computationally efficient for large-scale applications [35,36,37]. For each pixel, the copula parameter $\mathsf{\theta }$ was estimated using maximum likelihood, and the copula-based joint cumulative distribution function was then computed to obtain the joint probability P. Uncertainty in P was quantified via a nonparametric bootstrap on 200 spatially stratified representative pixels (B = 200 resamples per pixel). For each pixel, P was recomputed after refitting the copula to resampled paired monthly (NPP, SCDHI) observations; the 95% CI was defined by the 2.5–97.5% percentiles of the bootstrap distribution, yielding a median CI width of 0.0025. Subsequently, P is transformed to a standard normal distribution using the inverse standard normal distribution function, yielding SCDHI as the indicator of compound dry–hot events, which is calculated as follows

$$SCDHI={\mathsf{\phi }}^{-1}(P)$$

(6)

where ${\mathsf{\phi }}^{-1}$ is the inverse standard normal distribution function. Similarly to STI and SPEI, SCDHI can also be classified into different categories, with lower SCDHI values indicating more severe compound dry–hot conditions. When SCDHI < −0.5 denotes a compound dry–hot event, whereas SCDHI < −1.3 denotes an extreme compound dry–hot event [38].

2.3.4. Correlation Analysis and Response Timescale

To examine ecosystem responses to compound dry–hot events in the Yangtze River Basin, we computed pixel-wise Pearson correlation coefficients between monthly NPP and SCDHI over 2002–2024, as follows

$$R={\displaystyle \frac{{\displaystyle {\sum }_{i=1}^n(x_i-\overline{x})(y_i-\overline{y})}}{\sqrt{{\displaystyle {\sum }_{i=1}^n{(x_i-\overline{x})}^2{\displaystyle {\sum }_{i=1}^n{(y_i-\overline{y})}^2}}}}}$$

(7)

where R is the Pearson correlation coefficient between NPP and SCDHI. x_i and y_i are the NPP and SCDHI values in month i. $\overline{x}$ and $\overline{y}$ are the corresponding means of NPP and SCDHI over 2002–2024.

Vegetation growth typically responds to climatic factors with a time lag, and the response timescale varies among regions and vegetation types [39]. Previous studies have shown that the dominant lagged response of vegetation to climate variability is mostly within 3 months [39,40]. Accordingly, this study computes, at the pixel scale, the Pearson correlation coefficients between NPP and SCDHI in the current month ($R_0$), and with 1-month ($R_1$), 2-month ($R_2$), and 3-month ($R_3$) lags. For each pixel, the maximum absolute value among these four correlation coefficients is defined as the maximum correlation coefficient $R_{\max }$ between NPP and SCDHI, and the corresponding lag month is defined as the response timescale of NPP to dry–hot variability.

$$R_{\max }=\max (\left|R_i\right|),i=0,1,2,3$$

(8)

$$T=t(R_{\max })$$

(9)

where $R_i$ is the correlation coefficient between NPP and SCDHI with an $\mathrm{i}$-month lag. $R_{\max }$ represents the maximum correlation strength of NPP with dry–hot variability. T is the corresponding lag in months (0–3 months), indicating the dominant response timescale of vegetation to dry–hot variability.

2.3.5. Theil–Sen Median Slope Estimator and Mann–Kendall Trend Analysis

The Theil–Sen estimator is a nonparametric method for calculating slopes that can effectively characterize the magnitude of trends in a time series and serves as a useful complement to the Mann–Kendall trend test [41]. When the trend value is greater than 0, it indicates an increasing trend. Conversely, a negative value indicates a decreasing trend, and a value equal to 0 suggests no obvious trend in the region. The calculation formula is as follows

$$\mathsf{\beta }=Median({\displaystyle \frac{x_i-x_j}{i-j}}),\forall i>j$$

(10)

where $x_i$ and $x_j$ are the sample values at times $i$ and $j$, respectively. Median denotes the median operator. $\mathsf{\beta }>0$ indicates an increasing trend in the SCDHI and NPP time series, whereas $\mathsf{\beta }<0$ indicates a decreasing trend.

The Mann–Kendall test is a nonparametric method used to detect whether a trend in a time series is statistically significant [42]. When the standardized test statistic |Z| ≥ 1.96, it indicates passing the significance trend test at a 95% confidence level [41].

2.3.6. Assessment of Vegetation Resistance and Resilience Based on the NPP–SCDHI Framework

Building on the “resistance–resilience” framework proposed by Lloret et al. [40], this study quantifies vegetation resistance and resilience to extreme compound dry–hot events using the ratio of NPP during or after dry–hot conditions to NPP before dry–hot conditions (normal period), and extends this approach to the NPP–SCDHI framework at the pixel scale [43]. To quantitatively assess the response capacity of vegetation NPP to extreme compound dry–hot events in the Yangtze River Basin, vegetation responses to dry–hot events are divided into two dimensions, namely resistance during the event and recovery after the event. Based on SCDHI, extreme dry–hot months are identified using a threshold of −1.3. Months with SCDHI < −1.3 are defined as extreme dry–hot months, whereas other months are treated as non-extreme dry–hot months. At the pixel scale, the multi-year mean NPP of all non-extreme dry–hot months is calculated as the baseline productivity under normal climatic conditions and denoted as G.

$$\overline{G}={\displaystyle \frac{1}{N_{nor}}}{\displaystyle \sum _{t\in \Omega nor}{G}_t}$$

(11)

where $G_t$ is the NPP in month $t$. ${\Omega }_{nor}$ is the set of non-extreme dry–hot months. $N_{nor}$ is the number of non-extreme dry–hot months, and G represents the multi-year mean NPP under normal conditions.

Considering that extreme dry–hot processes often persist for several consecutive months [44,45], all months with SCDHI < −1.3 are merged into individual extreme dry–hot events according to temporal continuity. Let the $k$-th extreme dry–hot event start in month $s_k$ and end in month ($e_k(s_k\le e_k)$). The resistance of this event is defined as the ratio of the minimum NPP during the entire event to the baseline level.

$$R_k={\displaystyle \frac{G_{d_k}}{\overline{G}}}$$

(12)

For each pixel, the mean resistance over all extreme dry–hot events during the study period is then calculated as

$$R={\displaystyle \frac{1}{N_{ext}}}{\displaystyle \sum _{k=1}^{N_{ext}}{R}_k}$$

(13)

Here, $N_{ext}$ is the number of extreme dry–hot events experienced by the pixel. If $R\approx 1$, NPP under extreme dry–hot conditions is close to the normal level, indicating strong resistance. $R<1$ means that NPP during extreme dry–hot periods is lower than normal, and smaller R values indicate stronger suppression by dry–hot stress. $R>1$ suggests that NPP during extreme dry–hot periods exceeds the normal level, reflecting strong maintenance capacity or pronounced effects of human management.

To characterize post-event recovery, this study adopts a one-year window after the end of each extreme compound dry–hot event as the recovery timescale. On the one hand, productivity at the annual (or full growing-season) scale can comprehensively reflect the overall recovery status of vegetation after a complete growing season [46]. On the other hand, many tree-ring- and remote-sensing-based studies commonly construct resistance–recovery–resilience metrics using “pre-drought year–drought year–first post-drought year (or the first 1–2 years)” and have found that the negative legacy effects of extreme drought are most pronounced in the first year after drought cessation [47,48,49]. For the k-th extreme dry–hot event, let $G_{\min ,k}$ denote the minimum NPP during the event, $\overline{\mathrm{G}}$ the multi-year mean NPP in non-event (normal) months, and $G_{e_k+12}$ the NPP in the 12th month after the event ends. The resilience (relative recovery) of this event is defined as

$${\mathrm{Re}}_k={\displaystyle \frac{G_{e_k+12}-G_{\min ,k}}{\overline{G}-G_{\min ,k}}}$$

(14)

Similarly, by averaging ${\mathrm{Re}}_k$ over all extreme dry–hot events at each pixel, we obtain the pixel-scale mean resilience.

$$\mathrm{Re}={\displaystyle \frac{1}{N_{ext}}}{\displaystyle \sum _{k=1}^{N_{ext}}{\mathrm{Re}}_k}$$

(15)

This index measures the fraction of the drought-induced productivity loss that has been regained one year after the event. When $\mathrm{Re}\approx 1$, NPP has essentially recovered to the normal level, indicating strong resilience. When $\mathrm{Re}<1$ implies that only part of the loss has been compensated, indicating incomplete or delayed recovery. When $\mathrm{Re}>1$ indicates that NPP one year after the event exceeds the normal level, suggesting an over-recovery or compensatory growth effect.

3. Results

3.1. Spatiotemporal Variations in NPP

3.1.1. Temporal Variations in NPP

From 2002 to 2024, the annual mean of monthly NPP in the Yangtze River Basin showed a significant increasing trend (Figure 3), with a multi-year mean of approximately 46.12 g·m⁻²·month⁻¹ and a Sen’s slope of 0.255 g·m⁻²·month⁻¹·yr⁻¹ (p < 0.001). The Mann–Kendall test statistic (Kendall’s tau) was 0.670. Basin-wide mean NPP ranged from 43.01 to 50.04 g·m⁻²·month⁻¹, reaching a maximum in 2023 and a minimum in 2005.

3.1.2. Spatial Variations in NPP

The annual mean of monthly NPP in the Yangtze River Basin exhibits an overall spatial gradient, increasing from the northwest to the southeast, with low NPP values mainly occurring in the northwestern basin and high-elevation upstream areas, and high values concentrated in the middle–lower reaches plains and southeastern regions (Figure 4). This is because the middle–lower reaches and southeastern parts of the basin are characterized by low elevations, warm and humid climates, abundant precipitation, and favorable thermal conditions. These areas are dominated by forest and crops, thereby supporting relatively high NPP levels. In contrast, the upstream and northwestern basin are mostly high-cold or relatively arid areas with higher elevations, lower temperatures, and insufficient water supply in some regions, leading to shorter growing seasons, lower vegetation cover, and consequently generally lower NPP values.

The spatial patterns of NPP trends and their significance levels in the Yangtze River Basin are shown in Figure 5 and Figure 6. During 2002–2024, NPP in most parts of the Yangtze River Basin shows an increasing trend, with approximately 85.8% of the area exhibiting NPP growth and about 16.7% identified as significantly increasing regions, mainly concentrated in the central basin and the middle–lower reaches plains. In contrast, NPP in the western basin and parts of the upper reaches is relatively stable, showing largely unchanged or only slight fluctuations. Meanwhile, about 2.5% of the area shows a decreasing NPP trend, sporadically distributed along the southern margin of the basin and in some hilly and mountainous areas in the east. Overall, during the study period, NPP changes in the Yangtze River Basin are characterized by “a generally slow increase, more pronounced growth in the central basin, relative stability in the west, and slight declines in some local areas,” indicating an overall improvement in vegetation productivity across the basin, albeit with pronounced spatial heterogeneity.

3.2. Spatiotemporal Variations in SCDHI

3.2.1. Temporal Variations in SCDHI

From 2002 to 2024, the mean SCDHI in the Yangtze River Basin showed a significant decreasing trend (Figure 7), with a Sen’s slope of −0.029 yr⁻¹ (p < 0.005). The Mann–Kendall test statistic (Kendall’s tau) was −0.447. During 2002–2024, the annual mean SCDHI fluctuated between approximately −2.03 and −0.73, with the highest value in 2012 (SCDHI = −0.73), indicating the weakest dry–hot conditions within the study period, and the lowest value in 2023 (SCDHI = −2.03), suggesting that dry–hot conditions in the Yangtze River Basin were most severe in 2023. Notably, although a severe drought occurred in the Yangtze River Basin in 2022, the basin-wide annual-mean SCDHI does not identify 2022 as the year with the lowest value. Overall, despite pronounced interannual fluctuations, SCDHI has shown a persistent decline over the past two decades, indicating an intensification of the compound drought background in the basin.

3.2.2. Spatial Variations in SCDHI

The spatial patterns of SCDHI trends and their significance levels in the Yangtze River Basin (Figure 8 and Figure 9) show that during 2002–2024, SCDHI exhibits a decreasing trend over approximately 99.6% of the basin, with 70% of the area characterized by a significant decline, mainly distributed in the middle–lower reaches plains and the upper reaches around Sichuan–Chongqing and Yunnan–Guizhou, indicating that dry–hot conditions in these regions have been continuously intensifying.

Only about 0.4% of the area shows an increasing SCDHI trend, with significantly increasing regions sporadically distributed in the headwaters of the upper reaches and along the northwestern margin of the basin.

Overall, from 2002 to 2024, SCDHI trends in the Yangtze River Basin are dominated by declines, with a spatial pattern characterized by “slight local alleviation but a generally warmer and drier tendency,” indicating a marked strengthening of the dry–hot background across the basin during the study period.

3.3. Response of NPP to Compound Dry–Hot Events

3.3.1. Relationship Between NPP and Compound Dry–Hot Events

During 2002–2024, NPP in approximately 47.00% of the Yangtze River Basin is positively correlated with compound dry–hot events (SCDHI < −0.5), among which about 39.72% of the area shows a significant positive correlation and 7.28% a non-significant positive correlation. These regions are mainly distributed in the middle–lower reaches plains of the Yangtze River and in the eastern part of the basin. In about 53.00% of the basin, NPP is negatively correlated with compound dry–hot events, with approximately 42.47% of the area exhibiting a significant negative correlation and 10.52% a non-significant negative correlation, mainly concentrated in the upper Yangtze and some hilly and mountainous areas of the middle reaches (Figure 10 and Figure 11). Overall, NPP in the upper Yangtze tends to decline significantly under dry–hot conditions, whereas ecosystems in the middle–lower reaches are more likely to maintain or even increase NPP during compound dry–hot events, indicating marked differences among basin segments in both the direction and sensitivity of vegetation responses to such events.

3.3.2. Response Time of NPP to Compound Dry–Hot Events

From 2002 to 2024, the mean response time of NPP to compound dry–hot events in the Yangtze River Basin is about 2.03 months (Figure 12). Regions with a response time within the current month account for 21.50% of the basin and are mainly located on the Tibetan Plateau and in the alpine canyon areas of western Sichuan in the upper reaches, as well as in some headwater valleys. In these areas, vegetation is dominated by grasslands, shrublands, and alpine vegetation, with relatively shallow root systems and weak soil water-storage capacity, so that once a compound dry–hot event occurs, NPP responds rapidly within the same season. Same-season responses also occur in some tributary and main-stem valleys of the middle–lower reaches and around the Yangtze River Delta, but these areas are relatively small and mostly correspond to urban land, irrigated farmland, and zones with dense hydraulic projects, where water and energy balance processes have been substantially modified, thereby amplifying and rapidly manifesting the impacts of compound dry–hot events on NPP. Areas with response times of 1 and 2 months account for 9.35% and 13.99%, respectively, and are patchily distributed in the transitional zones from the upper to the middle reaches and from the middle to the lower reaches, exhibiting a certain degree of lagged response to compound dry–hot variability. Regions with a response time of 3 months account for the largest proportion, reaching 55.15%, and are mainly distributed in the Jianghan Plain, the Dongting–Poyang Lake plains, and the surrounding hilly and mountainous areas in the middle–lower reaches. A similar 3-month lag-dominated response pattern also appears along the margins of the Sichuan Basin, in the mountains of western Sichuan, and in parts of the Yunnan–Guizhou Plateau in the southwestern basin. In these regions, vegetation is dominated by evergreen broadleaved forests, mixed conifer–broadleaf forests, and some cropland ecosystems, characterized by high vegetation cover, well-developed root systems, and generally deep soil layers that allow the use of deep soil water and groundwater to buffer short-term dry–hot stress. Together with the widespread lakes, wetlands, and irrigated agriculture in the middle–lower plains, and the strong soil water-storage and vegetation water-holding capacities in the southwestern mountains, these factors collectively enhance the “buffering–lag” effect of ecosystems in response to moisture fluctuations. Consequently, in these areas, anomalies in surface and shallow soil moisture require a certain time to propagate downward to deep soil layers and the root-active zone after the onset of a compound dry–hot event, so that the pronounced response of NPP to dry–hot stress often lags by more than one season.

3.4. Resistance and Resilience of Vegetation in the Yangtze River Basin to Extreme Compound Dry–Hot Events

At the basin scale, the mean resistance and resilience values of all vegetation types are less than 1 (

Table 1. Mean resistance and resilience of different vegetation types to extreme compound dry–hot events.

Vegetation Type	Mean Resistance	Mean Resilience
Coniferous forest	0.65	0.21
Mixed conifer–broadleaf forest	0.53	0.31
Broadleaved forest	0.65	0.37
Shrubland	0.57	0.18
Grassland	0.43	0.32
Alpine vegetation	0.60	0.46
Cultivated vegetation	0.25	0.17

). This indicates that during extreme compound dry–hot events, net primary productivity (NPP) across all vegetation types declines below the baseline under normal climatic conditions, and within one year after event termination only a fraction of the drought-induced productivity loss is recovered, precluding a full return to normal levels. Overall, extreme dry–hot events impose a pronounced and persistent suppressive effect on vegetation productivity in the Yangtze River Basin, although resistance during events and resilience after events differ substantially among vegetation types.

Among vegetation types, coniferous and broadleaved forests both exhibit relatively high resistance, with mean values of approximately 0.65, indicating that NPP during extreme compound dry–hot events remains around 65% of the normal level. However, their resilience is generally low and differs markedly. Resilience is only 0.21 in coniferous forests and 0.37 in broadleaved forests, suggesting that forests—especially coniferous forests—can buffer immediate impacts but recover only a limited proportion of lost productivity within one year after the events. Mixed conifer–broadleaf forests show slightly lower resistance but relatively faster recovery, with resistance of 0.53 and resilience of 0.31, implying larger instantaneous losses yet somewhat quicker post-event rebound. Shrublands have moderate resistance (0.57) but low resilience (0.18), indicating pronounced NPP declines during extreme dry–hot events and recovery of less than one-fifth of the productivity loss within one year. Cultivated vegetation exhibits the strongest sensitivity to these events, with the lowest resistance among all vegetation types (0.25), reflecting the largest productivity reductions during extreme compound dry–hot events. Its resilience is also low (0.17), indicating that within one year after event termination, productivity losses are only partially compensated and productivity remains far below normal levels. Grasslands are characterized by relatively weak resistance (0.43) but comparatively rapid recovery (0.32), indicating substantial productivity losses during events yet recovery of nearly one-third of the drought-induced loss within one year. Alpine vegetation shows moderate resistance (0.60) and the highest resilience (0.46), recovering nearly half of the lost productivity within one year and demonstrating strong post-event recovery capacity. Taken together, these results reveal pronounced divergence in vegetation responses to extreme compound dry–hot events across the Yangtze River Basin. Forests exhibit higher resistance but limited recovery on annual timescales. Shrublands have moderate resistance but low resilience. Cultivated vegetation lies in a low-resistance/low-resilience regime. Grasslands have weak resistance but relatively fast recovery. Alpine vegetation combines moderate resistance with the highest resilience.

4. Discussion

The results of this study indicate that from 2002 to 2024, NPP in the Yangtze River Basin exhibited an overall slow increasing trend, whereas the compound dry–hot background (SCDHI) evolved markedly towards warmer and drier conditions. Consequently, the two variables display a clear pattern of divergent trends and regional differentiation in both time and space. Specifically, NPP increased in approximately 85.8% of the basin, while SCDHI decreased in roughly 99.6% of the area. This implies that despite ongoing global warming and intensifying compound dry–hot events (CDHEs), vegetation productivity in the basin remains, on the whole, on an upward trajectory. This spatial predominance of NPP increases is consistent with MODIS-based assessments reporting widespread productivity gains across the Yangtze River Basin since the early 2000s and attributing them to ecological restoration and land-use management in addition to climate effects [50]. At broader scales, satellite syntheses have similarly highlighted a strong greening signal since around 2000, in which land management and CO₂-driven physiological effects contribute alongside climate variability [51,52]. In contrast, the pronounced decline in SCDHI accords with the broader literature showing that compound drought–heat extremes have intensified in many regions under warming, and that their ecological impacts often exceed those of drought or heat alone because soil moisture deficits and high atmospheric demand co-occur [1,15]. Taken together, the divergence (increasing NPP but decreasing SCDHI) suggests that greening processes and the relaxation of energy limitation can partially offset the rising dry–hot risk at the basin scale, even though negative impacts remain evident in specific sub-regions.

Notably, in the basin-wide annual-mean SCDHI series, the widely reported 2022 Yangtze River Basin drought is not identified as the year with the lowest value. This is not inconsistent with evidence that 2022 constituted an exceptional drought–heat episode with extremely rare occurrence when both meteorological anomalies and impacts are considered [53]. Rather, the discrepancy likely reflects scale and metric effects. Short-lived extremes can be diluted by annual and spatial averaging, and the 3-month SPEI component embedded in SCDHI further smooths short-duration signals. Consistent with this interpretation, event-focused studies using sub-seasonal remote sensing indicators show that the 2022 summer event caused substantial suppression of vegetation activity and carbon uptake, including marked declines in photosynthesis-related proxies and GPP during July–August 2022 [54,55]. These comparisons highlight that basin-mean annual indices are well suited for characterizing background conditions, whereas event-scale diagnostics are necessary to capture short, intense compound extremes and their immediate ecosystem impacts.

Meanwhile, the NPP–SCDHI relationship is spatially heterogeneous. NPP is negatively correlated with SCDHI over approximately 53% of the basin and positively correlated over approximately 47%. This contrast likely reflects a shift in dominant controls along hydroclimatic gradients, from energy limitation in cold/high regions to water limitation in warm/humid lowlands [51]. In the upper reaches and some hilly middle-reach areas, NPP tends to increase or remain stable despite intensifying dry–hot conditions, suggesting that warming-induced relaxation of temperature/radiation constraints and ecological restoration can offset dry–hot stress [50,51,52]. In the humid middle–lower reaches and plains, however, stronger dry–hot conditions exacerbate soil moisture deficits and atmospheric demand (higher VPD), suppressing stomatal conductance and carbon assimilation and leading to more pronounced NPP declines [56]. Although water regulation, irrigation, improved management, and forestry practices can buffer impacts, they mainly reduce losses and accelerate recovery rather than reversing downward trajectories under moderate to severe dry–hot stress. In cropland-dominated areas of the mid–lower basin, irrigation and agricultural management may partially buffer dry–hot stress and thereby alter the apparent NPP response to dry–hot conditions. Similar management effects on CASA-based estimates of cropland NPP and carbon sequestration have been reported for maize in the Fen River Basin [57]. However, because spatially explicit irrigation intensity and management indicators are not included in our analysis, we discuss these anthropogenic influences as plausible mechanisms rather than providing a quantitative causal attribution. Future work incorporating irrigation intensity and management datasets would help quantify the contribution of anthropogenic regulation. Overall, these contrasting mechanisms jointly produce the coexisting positive and negative NPP responses across the basin.

The distribution of response times reveals buffering and lag characteristics in hydro–ecological processes. The mean response time of NPP to CDHEs in the basin is approximately 2.03 months. Rapid responses (in the current month or with a lag of less than one month) dominate on the Tibetan Plateau and in the alpine canyons of western Sichuan. In these regions, shallow soils, limited water-storage capacity, and alpine vegetation types mean that soil moisture is rapidly depleted during dry–hot episodes, causing immediate vegetation stress. Conversely, lagged responses (2–3 months) prevail in the middle–lower plains. Here, thicker soil layers and vegetation with deeper root systems (e.g., forests) benefit from the groundwater, lakes, and wetlands. This buffers short-term drought, delaying the transmission of surface moisture anomalies to the root zone. Similar multi-month lags have been reported in satellite-based drought-impact studies, with response times modulated by rooting depth, soil water storage, and local hydroclimatic conditions [49]. Our results quantify, at the basin scale, the contrast between rapid responses in cold, high-elevation regions and delayed responses in humid regions under compound dry–hot stress.

From the perspective of ecosystem stability, differences in resistance and resilience among vegetation types delineate clear vulnerability patterns across the Yangtze River Basin. Vegetation productivity is broadly suppressed during extreme compound dry–hot events, and full recovery remains difficult within one year after event termination. Forest ecosystems generally exhibit strong resistance, but their post-event recovery is relatively constrained. This resistance–resilience trade-off is consistent with flux-based evidence showing that forests generally exhibit stronger resistance but weaker resilience than grasslands under extreme drought conditions [58]. Shrublands typically show moderate resistance yet limited resilience, whereas cultivated vegetation is characterized by both low resistance and low resilience, making it particularly susceptible to persistent productivity deficits. Grasslands often experience larger productivity reductions during events but tend to recover more rapidly afterward, while alpine vegetation shows comparatively strong recovery capacity. Global syntheses further indicate that regions with higher forest fractions are generally more resistant to extreme drought than croplands, and that irrigation can substantially enhance cropland drought resistance [59]. This helps explain why cultivated vegetation in the Yangtze River Basin remains vulnerable despite management buffering. Overall, these response characteristics reveal pronounced vulnerability contrasts among vegetation types and suggest that cultivated vegetation, together with some grassland systems, may represent especially sensitive components of basin ecosystems under intensifying compound dry–hot stress.

It should be noted that although this study provides a relatively systematic depiction of NPP responses to compound dry–hot events and their type-specific differences at the basin scale, it remains constrained by data and methodological limitations. Uncertainties exist in the spatial resolution and retrieval accuracy of ERA5-Land reanalysis data and MODIS NPP products, and after harmonizing multiple datasets to a 0.25° spatial grid, it is difficult to fully capture fine-scale differences in topography, soils, and land use at hillslope and small-basin scales. In addition, this study did not perform an independent accuracy evaluation or quantitative uncertainty propagation for the input datasets; therefore, the uncertainty of our results largely inherits the limitations documented for ERA5-Land and MODIS products as well as the uncertainties introduced by the harmonization workflow. SCDHI mainly integrates information on temperature and water balance and does not yet explicitly incorporate key factors such as soil moisture, radiation, wind speed, vegetation structural characteristics, and the intensity of human activities. Thus, its representation of compound dry–hot processes and their ecological effects remains relatively simplified, and it is not yet possible to quantitatively disentangle the relative contributions of climate change versus land-use/management changes to NPP. In the future, it will be necessary, under the support of higher-resolution meteorological and ecological datasets, to incorporate multi-source soil-moisture and vegetation-structure observations and to integrate ecosystem process models in order to assess, from the perspectives of individual event processes and different climate scenarios, the long-term impacts of compound dry–hot events on ecosystem structure and services in the Yangtze River Basin, thereby providing more refined and differentiated scientific evidence for basin-scale adaptive management and ecological restoration strategies.

5. Conclusions

This study used ERA5-Land reanalysis and MODIS NPP/GPP data (2002–2024) to construct a pixel-scale standardized compound dry–hot index (SCDHI) for the Yangtze River Basin and to quantify NPP responses (magnitude, direction, and lag), as well as vegetation resistance and resilience under compound dry–hot events. The main conclusions are:

(1)

Basin-wide annual mean of monthly NPP increased slightly from 2002 to 2024 (43.01–50.04 g m⁻² month⁻¹), peaking in 2023 and reaching a minimum in 2005. NPP increased from northwest to southeast and 85.8% of the basin exhibited an upward trend (16.7% significant), whereas only limited areas showed weak declines or remained stable.

(2)

Annual mean SCDHI decreased significantly (p < 0.005) at approximately −0.029 yr⁻¹, indicating an overall drying and warming trend. The weakest dry–hot conditions occurred in 2012 and the most severe in 2023. Spatially, 99.6% of the basin showed decreasing SCDHI, with about 70% declining significantly.

(3)

Under dry–hot conditions (SCDHI < −0.5), NPP was positively correlated with dry–hot intensity over approximately 47% of the basin (mainly the middle and lower reaches and eastern areas) and negatively correlated over 53% (primarily the upper reaches and parts of the middle-reach hills). The mean response time was about 2 months, with faster responses in the upper and high-elevation western regions (0–1 month) and longer lags in the humid middle–lower reaches (2–3 months).

(4)

Extreme compound dry–hot events affected vegetation stability, with clear differences among vegetation types. Forests showed relatively high resistance but limited recovery, shrublands exhibited moderate resistance but low resilience, and cultivated vegetation had the lowest resistance and resilience and was therefore the most sensitive. Grasslands showed weak resistance but comparatively faster recovery, whereas alpine vegetation exhibited the highest resilience. Overall, these patterns suggest that cultivated vegetation and grasslands may represent high-risk vegetation types under extreme dry–hot stress.