Response of Vegetation Net Primary Productivity to Extreme Climate in a Climate Transition Zone: Evidence from the Qinling Mountains

Abstract

The Qinling Mountains, situated in the climatic transition zone between northern and southern China, represent a critical region for climate and ecological studies due to their unique transitional characteristics and the rising frequency of extreme climate events. As net primary productivity (NPP) is a key indicator of ecosystem stability, clarifying its response to extreme climate events is essential for understanding ecological resilience in this region. In this study, daily observational data from 123 meteorological stations (1960–2023) were used to derive eight extreme temperature and precipitation indices. Combined with MODIS NPP data (2001–2023), we applied Theil–Sen slope estimation, Mann–Kendall significance testing, ridge regression, Pearson correlation analysis, and Moran’s I spatial autocorrelation to systematically investigate the spatiotemporal dynamics and driving mechanisms of NPP. The main findings are as follows: (1) From 2001 to 2023, the mean annual NPP in the Qinling region was 558.43 ± 134.27 gC·m⁻²·year⁻¹, showing a significant increasing trend of 5.44 gC·m⁻²·year⁻¹ (p < 0.05). (2) Extreme temperature indices exhibited significant changes, whereas among the precipitation indices, only the number of days with daily precipitation ≥ 20 mm (R20) showed a significant trend, suggesting that extreme temperatures exert a stronger influence in the region. (3) Correlation analysis indicated that temperature-related indices were generally positively correlated, precipitation-related indices displayed even stronger associations, and covariation existed among extreme precipitation events of varying intensities. Moreover, precipitation indices demonstrated relatively stable spatial distributions, while temperature indices fluctuated considerably. (4) Absolute contribution analysis further revealed that the number of days with daily minimum temperature below the 10th percentile (TN10p) contributed up to 3.53 gC·m⁻²·year⁻¹ to annual NPP variation in the Henan subregion, whereas maximum rainfall over five consecutive days (Rx5day) exerted an overall negative effect on NPP (−0.77 gC·m⁻²·year⁻¹). By integrating long-term meteorological observations with remote sensing products, this study quantitatively evaluates the differential impacts of extreme climate events on vegetation within a climatic transition zone, offering important implications for ecological conservation and adaptive management in the Qinling Mountains.

1. Introduction

Under the background of global climate change, the frequency and intensity of extreme climate events have significantly increased, posing a serious threat to terrestrial ecosystems and the carbon cycle [1,2]. A substantial body of research evidence indicates that extreme events such as heatwaves and droughts are progressively weakening vegetation productivity and the carbon sequestration function of terrestrial ecosystems [3,4]. For instance, global-scale comprehensive analyses have found that abnormal droughts, high temperatures, and extreme precipitation events reduce the carbon cycling processes of ecosystems and lower the ability of vegetation to absorb atmospheric CO₂ [5]. In 2022, extreme heat and drought in the Europe region led to a significant reduction in the forest net carbon sink [6]; similarly, high temperatures and droughts in the Yangtze River Basin of China have led to a significant decrease in the total land water storage and other hydrological variables [7]. Therefore, extreme climate events are becoming the new norm that impacts ecosystem carbon balance, significantly affecting global and regional vegetation changes [8,9].

The 27 extreme climate indices recommended by the Expert Team on Climate Change Detection and Indices (ETCCDI) have become the standard tool for monitoring extreme temperature and precipitation events and provide a unified scale for assessing extreme climate events [10]. Applying these extreme climate indices to vegetation ecology research helps reveal the impact mechanisms of different types of extreme events on ecosystem functions [11]. These indices, based on daily meteorological data, effectively represent the frequency, intensity, and duration of extreme events [12]. In vegetation ecology research, extreme temperature and precipitation indices offer critical pathways for understanding the impact of climate on net primary productivity: extreme high-temperature events generally cause a reduction in photosynthesis and an increase in respiration due to heat stress and enhanced transpiration, thus suppressing NPP [13]; extreme low temperatures (e.g., frosts, cold waves) may damage plant tissues and shorten the growing season, also reducing vegetation productivity [14]. Regarding precipitation, prolonged drought (extreme low rainfall) leads to severe soil moisture deficits, resulting in a rapid decline in NPP, whereas extreme short-term rainfall events may cause soil erosion or waterlogging, negatively affecting vegetation growth [15]. Empirical studies show that the decline in vegetation NPP is most significant during severe drought events, and the more extreme the drought, the more sensitive NPP is to it [16]. In contrast, the impact mechanism of temperature-related extreme events on NPP is more complex: in high-altitude or high-latitude areas, moderate–extreme high-temperature events may temporarily increase NPP by extending the effective growing season, while in low-altitude, warm and humid regions, extreme high temperatures generally exacerbate drought stress, suppressing plant growth [17]. Overall, both temperature and precipitation extremes can significantly alter the carbon balance of vegetation, with their mechanisms involving imbalances in photosynthesis and respiration, changes in water use efficiency, and physiological damage [18,19].

The Qinling Mountains hold a critical position in China’s ecological geography [20]. Stretching across central China, with distinct climatic differences on its northern and southern slopes, it forms a key climatic transition zone and biodiversity boundary [21]. As a climatic transition zone, the Qinling Mountains are highly sensitive to climate change, making them an important area for testing the ecological impacts of climate variability and extreme events. Studies have found that in recent decades, ecosystems in the Qinling region have exhibited high sensitivity and vulnerability to climate change [22]. Changes in temperature and the spatiotemporal patterns of precipitation are causing significant alterations to the structure and function of forest ecosystems, with some ecosystem services being weakened [23]. Therefore, understanding the impact of extreme climate events on vegetation NPP in the Qinling region is of great importance for regional ecological management.

Although the potential harm of extreme climate events to ecosystems has gained increasing attention, previous studies still have some limitations and gaps. Many studies on vegetation dynamics in the Qinling region have focused on the effects of average climate factors and environmental variables, with limited quantitative assessments of the impact of extreme climate events on NPP. For example, previous research has investigated the response of NPP to mean temperature and precipitation, as well as the spatial distribution patterns of NPP in Qinling forests under different altitudes, slopes, and other topographical factors [24,25]. While these studies provide ample theoretical and practical foundations, reflecting the diversity and complexity of the factors influencing NPP in the Qinling region, they primarily focus on the impact of average climate conditions and environmental variables, with little systematic research on the potential effects of extreme climate events on NPP. In fact, extreme climate events are more likely to cause short-term, irreversible damage to ecosystems, and their impacts are often overlooked in conventional climate studies [26].

Given the above research background, this study takes the Qinling region as a case study, using daily meteorological data from 123 stations over 1960–2023 to calculate multiple extreme climate indices, combined with MODIS NPP data from 2001–2023. By applying trend analysis, ridge regression, spatial autocorrelation, and other methods, this study comprehensively analyzes the spatiotemporal variation in vegetation productivity in the Qinling region and its relationship with extreme climate events. The aim is to reveal the sensitivity of NPP to extreme climate in climatic transition zones and identify its dominant driving factors, thus providing valuable references for assessing the response of ecosystems in transition zones to global climate change.

2. Materials and Methods

2.1. Study Area

The Qinling Mountains are situated in central China (102–114° E, 32–36° N), stretching over 1000 km from east to west across the provinces of Gansu, Shaanxi, and Henan, covering an area of approximately 160,500 square kilometres. The terrain is characterised by a notable elevation gradient, with higher land in the west and lower terrain in the east, ranging from 40 to 4831 m. During the study period, the region’s annual average temperature varied between −3.64 and 16.25 °C, while the annual average precipitation ranged from 421.42 to 1101.97 mm (Figure 1). The landforms are predominantly hills, interspersed with basins, river valleys, and plains. Vegetation is mainly composed of warm temperate deciduous broad-leaved forests and north subtropical evergreen–deciduous broad-leaved mixed forests, with additional distributions of shrubs, grasslands, and cultivated crops [27,28]. The region serves not only as a transition zone for China’s culture, geography, climate, and biology but also as a sensitive area for climate change and an ecologically vulnerable region [29].

2.2. Data Sources

This study utilizes the basic meteorological observation data from ground-based stations in China, covering the period from 1960 to 2023, with a total of 123 daily meteorological stations distributed across the Qinling Mountains and surrounding areas. Data sources are presented in

Table 1. The basic data information of this study.

Data	Data Source	Time Resolution	Spatial Resolution
Meteorological	https://data.cma.cn/ URL (accessed on 2 August 2024)	day	/
NPP	https://earthdata.nasa.gov/ URL (accessed on 17 August 2024)	year	500 m
DEM	http://www.gscloud.cn/ URL (accessed on 17 August 2024)	/	30 m

. In this study, daily meteorological data were quality-controlled using the RClimDex model, and 27 extreme climate indices recommended by ETCCDI. Eight indices reflecting long-term changes in temperature and precipitation events were selected for study [30,31]. These indices not only capture extreme shifts in the climate system but also effectively represent the influence of factors such as temperature and precipitation on vegetation growth and distribution. The specific definitions of these indices are provided in

Table 2. The eight extreme climate index information selected in this study.

ID/Units	Indicator Name	Definitions
CSDI/d	Cold-spell duration indicator	Annual count of days with at least 6 consecutive days when TN < 10th percentile
ID/d	Ice days	Annual count when TX (daily maximum) < 0 °C
TN10p/d	Cool nights	Percentage of days when TN < 10th percentile
TR/d	Tropical nights	Annual count when TN (daily minimum) > 20 °C
CDD/d	Consecutive dry days	Maximum number of consecutive days with RR < 1 mm
R10/d	Number of heavy precipitation days	Annual count of days when precipitation ≥ 10 mm
R20/d	Number of very heavy precipitation days	Annual count of days when precipitation ≥ 20 mm
Rx5day/mm	Max 5-day precipitation amount	Monthly maximum consecutive 5-day precipitation

. To ensure data consistency, the NPP dataset spans from 2001 to 2023, aligning with the availability of MODIS data. Meanwhile, to mitigate the randomness in extreme climate events due to a short time scale, the study of extreme climate indices covers the period from 1960 to 2023, offering long-term climate trends. Finally, spatial interpolation was carried out using Anusplin 4.36 software, and all data were standardised to a spatial resolution of 500 m [32,33]. This method, based on the thin plate spline function theory, incorporates multiple influencing factors as covariates for the spatial interpolation of meteorological variables, significantly improving interpolation accuracy. It also enables the simultaneous spatial interpolation of multiple surfaces, making it particularly suitable for time-series meteorological data.

2.3. Study Methods

2.3.1. Trend Analysis

Sen’s ($S$) trend analysis was used to detect long–term changes in climate variables in the QL [34], defined as follows:

$$S=Median\left({\displaystyle \frac{x_j-x_i}{j-i}}\right),\left(\forall i\le j\right)$$

(1)

Among them, $Median\left(\right)$ represents the slope of time series change involved in the calculation; the $i$ and $j$ represent the years.

The Mann–Kendall (MK) non–parametric test was used to test the significance of the change trend of NPP and extreme climate index [35,36]. The MK statistic ($K$) is calculated as:

$$K=\sum _{j=1}^{n-1}\sum _{i=j+1}^{n}sgn\left(x_j-x_i\right)$$

(2)

where $x_i$ and $x_j$ represent the values of time series with length $n$, and $sgn\left(x_j-x_i\right)$ is estimated as:

$$sgn\left(x_j-x_i\right)=\left\{\begin{array}{c}1,if(x_j>x_i)\\ 0,if(x_j=x_i)\\ -1,if(x_j<x_i)\end{array}\right.$$

(3)

The $K$ statistic represents the positive and negative variance of the dataset. Under the null hypothesis of $E\left(K\right)=0$, the average value of the statistic is 0, and the variance can be expressed as:

$$V={\displaystyle \frac{n\left(n-1\right)\left(2n+5\right)-{\sum }_{j=1}^n\left(t_j-1\right)\left(2t_j+5\right)}{18}}$$

(4)

Because of the approximate normal distribution, it obeys the $Z$ transform:

$$Z=\left\{\begin{array}{c}{\displaystyle \frac{K-1}{\sqrt{V}}},if\left(M>0\right)\\ 0,if\left(M=0\right)\\ {\displaystyle \frac{K+1}{\sqrt{V}}},if\left(M<0\right)\end{array}\right.$$

(5)

When the absolute value of $Z$ is greater than 1.65, 1.95 and 2.58, respectively, the trend is considered significant, and the confidence levels are 90%, 95% and 99%, respectively.

2.3.2. Attribution Analysis

In order to effectively alleviate the multicollinearity problem [37,38], the coefficients b are estimated by adding a perturbation λI, that is, using ridge regression method to analyze the impact of various extreme climate indices on NPP changes from 2001 to 2023:

$${\widehat{\beta }}_{ridge}={\left(X^{T}X+\lambda I\right)}^{-1}{X}^{T}y$$

(6)

where ${\widehat{\beta }}_{ridge}$ is the estimated value of the ridge regression coefficient; $\lambda$ is the regularization parameter, used to control the strength of the regularization term., use cross-validation to select the optimal λ value. When $\lambda =0$, ridge regression degenerates into ordinary least squares regression.; $I$ is the identity matrix; $X$ represents the matrix of independent variables composed of extreme climate indexes; $y$ represents the vector of the dependent variable constituted by the NPP.

Normalize each set of data:

$$X_m={\displaystyle \frac{x-min\left(x\right)}{max\left(x\right)-min\left(x\right)}}$$

(7)

where $X_m$ represents the normalized results of each extreme climate index.

The NPP and extreme climate index data were input into the ridge regression model:

$$Y_m=\underset{i=1}{\sum ^n}{a}_i{x}_{im}+b$$

(8)

where $Y_m$ represents the normalized NPP data; $x_{im}$ is the normalized extreme climate index; $a_i$ is the regression coefficient.

According to the ridge regression coefficient and the normalized trend of extreme climate index, the contribution of extreme climate index to NPP change trend is calculated:

$${\eta }_{c1}=a_1\times X_{1s\_trend}$$

(9)

$${\eta }_{rc1}={\displaystyle \frac{\left|{\eta }_{c1}\right|}{\left|{\eta }_{c1}\right|+\left|{\eta }_{c2}\right|+\left|{\eta }_{c3}\right|+\cdots }}$$

(10)

$${\eta }_{ac}={\displaystyle \frac{{\eta }_{c1}}{Y_{n\_trend}}}\times Y_{trend}$$

(11)

Among them, ${\eta }_{rc1}$ represents the relative contribution of extreme climate indices; ${\eta }_{ac}$ represents the absolute contribution of extreme climate indices; $X_{1s\_trend}$ represents the trend of extreme climate index after normalization. $Y_{n\_trend}$ represents the trend of changes in NPP after normalization. $Y_{trend}$ represents the trend of NPP predicted by ridge regression.

2.3.3. Pearson Correlation Analysis

Analyze the correlation between extreme climate indices from 1960 to 2023 using Pearson correlation analysis [39]. The specific formula is:

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

(12)

where $x_i$ and $y_i$ are two extreme climate indices values, respectively; $\overline{x}$ and $\overline{y}$ are the respective mean values. A two-sided hypothesis test was performed on the $r$ value of each pair of indexes, and the significance level was set to α = 0.05 to determine whether the correlation was significant.

2.3.4. Spatial Autocorrelation Analysis

Spatial autocorrelation analysis is primarily used to describe the degree of similarity between spatial units and their surrounding areas within the study region [40]. Construction of spatial weight matrix:

$$W=\left(w_{ij}\right)$$

(13)

Global Moran’s I:

$$I={\displaystyle \frac{n}{S_0}}\cdot {\displaystyle \frac{{\sum }_{i=1}^n{\sum }_{j=1}^n{w}_{ij}\left(x_i-\bar{x}\right)\left(x_j-\bar{x}\right)}{{\sum }_{i=1}^n{\left(x_i-\bar{x}\right)}^2}}, S_0=\sum _{i=1}^n\sum _{j=1}^n{w}_{ij}$$

(14)

It is used to measure the spatial aggregation degree of attribute values in the whole study area: $I>0$ is positive autocorrelation, $I<0$ is negative autocorrelation, and $I\approx 0$ is approximately random.

Local Moran’s I:

$$I_i=\left(x_i-\bar{x}\right)\sum _{j=1}^n{w}_{ij}\left(x_j-\bar{x}\right)$$

(15)

The local autocorrelation index is calculated for each spatial unit $I$, and the significance is determined based on the permutation test.

Getis–Ord G_i*:

$$G_i^{\mathrm{*}}={\displaystyle \frac{{\sum }_{j=1}^n{w}_{ij}{x}_j}{{\sum }_{j=1}^n{x}_j}}$$

(16)

In the detection of local hot spots (significant high-value aggregation) and cold spots (significant low-value aggregation), $G_i^{\mathrm{*}}$ is usually standardized as a z-score for significance judgment.

3. Results

3.1. Spatial and Temporal Variation in NPP

The multi-year average NPP in the Qinling region from 2001 to 2023 exhibited significant regional variation, with higher values in the central region and lower values in the eastern and western regions. Specifically, the multi-year average NPP in the subregion QL_Gs showed an increasing trend from northwest to southeast, while the subregions QL_Sx and QL_Hn exhibited a decreasing trend from west to east. The overall multi-year average NPP for the entire QL region was 558.43 ± 134.27 gC·m⁻²·year⁻¹ (Figure 2). Among the three subregions, QL_Sx had the highest average NPP (609.57 ± 107.86 gC·m⁻²·year⁻¹), followed by QL_Gs (565.14 ± 143.23 gC·m⁻²·year⁻¹), and QL_Hn had the lowest (469.70 ± 108.50 gC·m⁻²·year⁻¹). The overall NPP trend in the study area was positive, with multi-year NPP change rates ranging from –32.07 to 30.52 gC·m⁻²·year⁻¹. The annual average NPP increased significantly by 5.44 gC·m⁻²·year⁻¹, and NPP increases were significant (p < 0.05) in most areas.

3.2. Spatial and Temporal Variation in Extreme Climate Indexes

The annual mean values of the interpolated raster data were calculated, and the temporal variations and statistical results of the extreme climate indices in the Qinling region from 1960 to 2023 are presented in Figure 3 and

Table 3. Interannual variation and change significance of extreme climate index.

Indices	Units	Average Regional Trends	Units	p
CSDI	d	−0.56	days/decade	p < 0.05
ID	d	−0.86	days/decade	p < 0.05
TR	d	−0.49	days/decade	p < 0.05
TN10p	d	−1.19	days/decade	p < 0.05
CDD	d	−0.28	days/decade	p = 0.724
R10	d	0.32	days/decade	p = 0.179
R20	d	0.28	days/decade	p < 0.05
Rx5day	mm	−0.04	mm/decades	p = 0.946

. Notably, CSDI, ID, TR, TN10p, and R20 exhibit significant interannual trends. Among them, R20 shows a significant increasing trend of 0.28 days per decade (p < 0.05), while CSDI (−0.56 days/decade, p < 0.05), ID (−0.86 days/decade, p < 0.05), TR (−0.49 days/decade, p < 0.05), and TN10p (−1.19 days/decade, p < 0.05) all show varying degrees of decline. Although CDD, R10, and Rx5day did not pass the Mann–Kendall significance test, the trends of CDD (−0.28 days/decade, p = 0.724) and R10 (−0.32 days/decade, p = 0.179) remain visually apparent, whereas Rx5day (−0.04 days/decade, p = 0.946) shows no clear trend.

Figure 4 and Figure 5 illustrate the spatial patterns of long-term changes in the QL extreme climate indices from 1960 to 2023, as well as the proportion of the area with significant changes. In general, the trend of change in the extreme climate indices in the study area shows clear differences between the east and west, with the changes in the extreme temperature indices being more pronounced than those in the extreme precipitation indices. Over the past 60 years, CSDI (93.30%) and TN10p (99.97%) have significantly decreased in more than 90% of the region. ID predominantly exhibited a downward trend, with 78.82% of the area showing significant decreases. TR showed a significant downward trend in the northwest of QL_Gs and Taibai Mountain in QL_Sx, with a significant decrease in 59.70% of the area, but a significant increase in QL_Hn. In the extreme precipitation indices, there is a strong correlation between CDD and Rx5day. The increase or decrease in CDD corresponds to the opposite trend in Rx5day, though the changes in CDD and Rx5day are generally not significant. The spatial distribution of R10 and R20 is largely the same, with significant changes primarily concentrated in QL_Sx. In QL_Sx and QL_Hn, an upward trend predominates, whereas in QL_Gs, a downward trend is more prevalent.

3.3. Spatio-Temporal Correlation and Hot Spot Analysis of Extreme Climate Index

Figure 6 shows the correlation significance between extreme climate indices. Pearson correlation analysis reveals that extreme temperature indices generally exhibit significant positive correlations. In particular, the correlation between TR and TN10p is the highest (r = 0.636, p < 0.001), and there are also highly significant positive correlations between CSDI and TN10p (r = 0.505) as well as CSWI and TR (r = 0.495). On the other hand, ID shows a weaker correlation with other temperature indices, with only moderate positive correlations with TR (r = 0.215) and TN10p (r = 0.387). In contrast, the correlations between extreme precipitation indices are more closely related. The correlation coefficient between R10 and R20 is as high as 0.880 (p < 0.001), and there are also significant positive correlations between R20 and Rx5day, as well as between R10 and Rx5day (0.655 and 0.571, respectively), indicating the covariation of different intensity extreme precipitation events. Additionally, CDD shows negative correlations with R10, R20, and Rx5day.

The Moran’s I index distribution of extreme temperature indices in the Qinling region is shown in Figure 7. The Moran’s I values for CSDI fluctuate significantly across the three periods (0.653, 0.168, 0.366), indicating that its spatial pattern changes considerably over time. ID maintains moderate to strong spatial autocorrelation (0.556–0.617). TR exhibits the weakest spatial autocorrelation (0.082–0.170), with a relatively random point distribution. TN10p shows large fluctuations across different periods (0.774, 0.395, 0.654).

The Moran’s I index distribution of extreme precipitation indices in the Qinling region is shown in Figure 8. The Moran’s I values for CDD remain stable between 0.631 and 0.681 across different periods, indicating a stable positive spatial autocorrelation. R10 has relatively high Moran’s I values (0.771–0.801), exhibiting clear spatial clustering characteristics. R20 has the highest Moran’s I values (0.818–0.866), with spatial autocorrelation gradually increasing over time. The Moran’s I values for Rx5day are relatively lower (0.600–0.653), but still show significant spatial autocorrelation. Overall, the extreme precipitation indices exhibit a stable spatial distribution pattern in the study area, primarily characterized by the clustering of low-value areas (LL) and high-value areas (HH). In contrast, the spatial distribution patterns of extreme temperature indices show more pronounced temporal variability, especially with CSDI and TN10p, where spatial autocorrelation significantly weakened during the period from 1981 to 2000.

Figure 9 shows the spatial distribution characteristics of four extreme temperature indices in three periods. The hotspot areas for CSDI are mainly concentrated in the central part of the study area, with cold spot areas in the east and west. This distribution pattern weakened during the second period but showed an increasing trend in the third period. The distribution of ID is more stable, with the northwest and central high-altitude areas being the hotspots, and the southeast and south being the cold spots, with little change in the distribution pattern across the three periods. TR shows a more complex distribution, with hotspot areas mainly in the northern and eastern parts, and the cold spot area in the center gradually expanded over time. TN10p shows larger temporal variability, with hotspots mainly in the west and central parts, and cold spots in the east. The spatial distribution pattern changed significantly during the second period.

Figure 10 shows the spatial distribution characteristics of four extreme precipitation indices. CDD presents a significant hotspot in the west, with cold spots in the central and eastern parts, indicating more significant drought characteristics in the western region of the study area. R10 and R20 show similar spatial patterns, both exhibiting a “hot in the south, cold in the north” distribution. The hotspot area of Rx5day is concentrated in the southeast, with the cold spot area in the northwest, and the distribution pattern remains relatively stable across the three periods. A comparison of the three periods reveals that the spatial distribution patterns of extreme precipitation indices are more stable than those of extreme temperature indices.

3.4. Relative Contribution of Extreme Climate Indices

Figure 11 shows that NPP in different regions of the Qinling Mountains is influenced by various extreme climate indices. QL_Gs NPP is primarily affected by TN10p, CDD, R10, and R20; QL_Sx NPP is mainly influenced by TR, TN10p, R10, and R20; QL_Hn NPP is predominantly shaped by TR and TN10p. According to

Table 4. Relative Contributions of Extreme Climate Indices to NPP Changes.

Regions	CSDI (%)	ID (%)	TR (%)	TN10p (%)	CDD (%)	R10 (%)	R20 (%)	Rx5day (%)
QL	7.98	2.19	11.71	29.43	10.41	17.64	13.12	7.52
QL_Gs	6.70	2.33	5.10	23.56	15.62	21.89	16.23	8.57
QL_Sx	7.41	1.36	15.48	28.50	7.63	20.39	13.06	6.17
QL_Hn	10.96	3.26	16.85	40.61	6.07	6.30	8.05	7.90

, with the exception of QL_Gs, where NPP is primarily influenced by extreme precipitation, NPP variations in the other regions are mainly governed by extreme temperature indices. Notably, TN10p (29.43%) contributes the most to NPP changes in QL. This significant contribution can be attributed to its considerable impact on plant physiological processes and seasonal temperature fluctuations. The contributions of the other indices are ranked as follows: R10 (17.64%), R20 (13.12%), TR (11.71%), CDD (10.41%), CSDI (7.98%), Rx5day (7.52%), and ID (2.19%).

Figure 12 illustrates the spatial distribution of the dominant factors driving NPP change trends across individual pixels. Overall, the influencing factors exhibit significant spatial heterogeneity. TN10p (44.02%) maintains an overwhelmingly dominant position across the entire study area and its three subregions. Specifically, R10 (20.97%) predominantly influences the northwestern and southeastern parts of QL_Gs, as well as the western region of QL_Sx; R20 (10.58%) dominates the northern areas of QL_Gs and QL_Sx; TR (11.96%) controls the northwestern and northeastern parts of QL_Sx, along with the northern region of QL_Hn; CDD (5.44%) primarily affects the central and northeastern parts of QL_Gs; CSDI (4.60%) and Rx5day (2.02%) exert notable influences on the western QL_Sx and central QL_Hn, while ID (0.39%) plays a negligible role in NPP change trends. Moreover, TN10p’s dominant area in QL_Hn exceeds 60%, indicating that the NPP change trends in this subregion are predominantly attributed to TN10p.

3.5. Absolute Contribution of Extreme Climate Indices

Analysis of the absolute contributions of the various extreme climate indices to NPP changes (Figure 13) demonstrates that NPP’s response to extreme climate events in the Qinling region exhibits marked spatial heterogeneity. Overall, TN10p exerts the strongest positive effect on NPP, most prominently in the QL_Hn sub-region (3.53 gC·m⁻²·year⁻¹). Although the extreme precipitation index R10 makes substantial positive contributions in the QL_Gs and QL_Sx sub-regions (2.27 and 2.44 gC·m⁻²·year⁻¹, respectively), its impact in QL_Hn is comparatively limited. Moreover, CDD displays the greatest disparity between positive and negative contributions, whereas Rx5day contributes predominantly negative effects on NPP throughout the Qinling region (–0.77 gC·m⁻²·year⁻¹).

4. Discussion

4.1. Changes in NPP and Driving Background

This study reveals that, between 2001 and 2023, the Qinling region experienced a significant overall increase in Net Primary Productivity (NPP), with an average annual growth of approximately 5.44 gC·m⁻²·year⁻¹. In most parts of the region, the rise in NPP was statistically significant. Spatially, the highest NPP values were concentrated in the central Qinling Mountains, while comparatively lower values were observed at the eastern and western margins. This spatial pattern is closely associated with the region’s north–south climatic gradient and corresponding vegetation distribution [41]. The sustained growth in NPP over the past two decades reflects an improvement in regional climate conditions and an overall enhancement of vegetation productivity [25].

On the one hand, the warming trend has extended the growing season in temperate and high-altitude areas, thereby reducing the limitations imposed by low temperatures [42]. On the other hand, the regional precipitation regime has remained relatively stable, with no prolonged drought events observed [43]. In addition, rising atmospheric CO₂ concentrations may have enhanced photosynthetic efficiency [44]. Together, these factors have driven the increase in NPP. This phenomenon is consistent with the global “greening” trend—for instance, several studies have shown that moderate increases in precipitation and temperature can significantly enhance vegetation productivity [45,46]. It is worth noting that the Qinling Mountains, situated within China’s transitional climate zone between north and south, are particularly sensitive to climate change. In this study, we observed notable changes in extreme climate indices, including a marked decline in cold-night frequency (TN10p) and a significant increase in extreme precipitation days (R20), both of which have contributed, to some extent, to favourable conditions for vegetation growth [47]. However, such positive drivers are not strictly linear, and the overall trajectory may still be disrupted by the increasing frequency and intensity of extreme climate events [2,48].

4.2. Mechanisms of the Influence of Extreme Temperature Indices on NPP

Extreme temperature events exert a complex, dual influence on vegetation productivity, with strong regional dependence [49]. The attribution analysis in this study shows that indices related to extreme low temperatures make particularly significant contributions to NPP. Among all factors, the decline in cold-night frequency (TN10p) accounted for the largest relative contribution to NPP growth, reaching 29.4%. This suggests that the reduction in cold nights has alleviated frost damage and low-temperature constraints on plant growth [50]. The effect is especially pronounced in the northern, high-altitude, and higher-latitude areas of the Qinling Mountains, where fewer cold nights correspond to a longer growing season, increased photosynthetic accumulation, and consequently, a substantial improvement in NPP.

In contrast, extreme high-temperature events often have a negative impact on NPP. Heatwaves can reduce photosynthetic rates, increase plant respiration, and intensify water stress, thereby suppressing vegetation productivity [51]. In our findings, the overall contribution of tropical nights (TR) to NPP was relatively low (approximately 11.7%), and its effects were highly dependent on local hydrothermal conditions. In high-altitude areas with sufficient moisture, occasional elevated temperatures may even enhance seasonal productivity [52]. However, in lowland humid regions or during periods of drought, high temperatures tend to significantly inhibit plant growth [53]. Moreover, increases in TR may offset daytime carbon gains by intensifying night-time respiratory carbon losses in vegetation [54]. Critically, when extreme heat events exceed vegetation tolerance thresholds, plants may close their stomata to reduce transpiration, leading to a marked decline in carbon assimilation rates [55]. This non-linear response mechanism suggests that under ongoing climate warming, extreme temperatures in the Qinling region pose both potential productivity benefits and significant ecological risks to NPP.

4.3. Regional Differential Responses to Extreme Precipitation Indices

Extreme precipitation events exhibit marked regional differences in their impact on NPP. In the western Qinling region—particularly the high-altitude semi-arid areas along the Gansu–Shaanxi border—NPP is most sensitive to extreme precipitation indices. Attribution analysis shows that in the QL_Gs subregion, the combined relative contributions of CDD, R10, and R20 exceed 50%, highlighting the significant constraints imposed by drought stress on vegetation in semi-arid zones [56]. In contrast, increased precipitation events help to alleviate water deficits and enhance NPP in these areas. In the eastern Qinling region (QL_Hn subregion), extreme precipitation plays a comparatively minor role in driving NPP changes, with temperature factors being more dominant. This is primarily because the eastern region receives relatively abundant rainfall, and short-term droughts have limited impact on plant growth [41]. Instead, variations in temperature—particularly those influencing the length of the growing season and the intensity of photosynthesis—are the key determinants of productivity [57]. It is also important to note that different types of extreme precipitation exert divergent effects on NPP [58]. Moderate increases in heavy precipitation days generally show a positive influence. In this study, both R10 and R20 contributed positively to NPP change, as increased water availability in the central and western Qinling regions supported greater vegetation productivity and carbon accumulation [59].

In contrast, CDD and Rx5day generally exert negative effects on NPP. Prolonged dry spells (CDD) suppress plant growth through water deficit, while excessive short-term precipitation events (Rx5day) may damage vegetation due to oversaturation. In our study, Rx5day showed a negative absolute contribution to overall NPP (–0.77 gC·m⁻²·year⁻¹), indicating that overly concentrated rainfall events can diminish net primary productivity by disrupting the water balance and increasing physical stress on plants [60]. Regional comparisons reveal that vegetation in the humid southern slopes of the Qinling Mountains demonstrates greater resilience to fluctuations in extreme precipitation, whereas the northern slopes and western subregions, characterised by lower average annual rainfall, are more vulnerable to the impacts of drought. This finding is consistent with previous studies, which have shown that in the southern part of China’s monsoon zone, variability in precipitation is the primary factor influencing NPP, and a warming–drying trend tends to exacerbate the negative effects of drought on productivity [61]. Therefore, in relatively arid ecosystems, increased water availability from extreme precipitation can elicit a strong positive response in NPP [62]. However, in more humid regions, where baseline moisture levels are already high, vegetation is less responsive to additional rainfall, and excessive precipitation may even result in adverse ecological outcomes [53].

4.4. The Spatial Autocorrelation Characteristics of Extreme Climate Events

Spatial autocorrelation analysis revealed pronounced structural differences in the spatial distribution of extreme climate indices across the Qinling region, highlighting the fundamental distinction between precipitation and temperature-related events in terms of geographic control and climatic response mechanisms [63]. Precipitation extremes exhibited strong and consistent spatial clustering, with the global Moran’s I values for R20 remaining between 0.82 and 0.87 across all periods. This indicates that intense rainfall events were primarily concentrated along the humid southern slopes of the Qinling Mountains. Such a pattern reflects the topographic asymmetry shaped by precipitation gradients between the northern and southern slopes [64], and aligns closely with the hotspot distributions of other indices such as R10 and CDD. The persistent CDD hotspots in the western subregion further indicate sustained drought stress in that area, with a clear and stable gradient between wet and dry zones. This spatial consistency suggests that extreme precipitation events are strongly influenced by geographic factors and represent a coupled interaction between climate and terrain [65].

In contrast, temperature-related extreme events exhibit greater spatiotemporal variability. For instance, the Moran’s I value for TN10p dropped sharply from 0.774 to 0.395 during the period 1981–2000, indicating a temporary weakening in the spatial clustering of cold-night events under the backdrop of global warming [66]. This trend reversed in the early 21st century, consistent with the globally observed rise in minimum night-time temperatures [54]. CSDI and ID were predominantly clustered in high-altitude areas, suggesting a strong topographic control, while TR displayed the most scattered spatial pattern, with Moran’s I values remaining around 0.1 over time. This indicates that tropical nights are more randomly distributed at the regional scale, possibly influenced by non-climatic factors such as surface albedo and urban heat island effects [67,68]. By integrating hotspot analysis with the spatial distribution of NPP drivers, it becomes evident that climate indices with strong spatial clustering tend to exert more stable regulatory effects on ecosystems [69]. For example, areas with significant R20 hotspots coincide with regions where extreme precipitation positively drives NPP, suggesting that heavy rainfall—when within a tolerable threshold—can enhance vegetation carbon sequestration [70]. In contrast, overlapping hotspots of CDD and Rx5day identify regions where NPP is strongly suppressed, highlighting the compound stress effect of drought and intense short-duration rainfall [71]. Notably, TN10p hotspots cover over 60% of the QL_Hn subregion, and this index showed the highest absolute contribution to NPP change (3.53 gC·m⁻²·year⁻¹), indicating that the spatial redistribution of extreme temperature events has had a substantial impact on regional carbon cycling patterns [9].

4.5. Limitation and Prospect

Firstly, the NPP dataset used in this study only spans the early 21st century, which, although effective in capturing recent trends, limits our ability to assess longer-term fluctuations and the full impact of extreme climate events [72]. Moreover, while our attribution analysis primarily focused on the influence of extreme climate indices, it did not account for other important drivers such as land-use change and anthropogenic disturbances, which may also significantly affect NPP dynamics [73,74]. Methodologically, the ridge regression model employed assumes a linear relationship between extreme climate events and NPP variation, which may not fully capture the lag effects or compound impacts associated with such extremes [75,76]. Therefore, future research should incorporate a broader range of influencing factors and adopt integrated, multi-factor analytical approaches to better uncover the complex interactions between extreme climate and ecosystem processes, and to enhance our understanding of the regional impacts of climatic extremes.

5. Conclusions

Based on ETCCDI extreme climate indices derived from daily meteorological station data spanning 1960–2023 and MODIS NPP data from 2001–2023, this study conducted a systematic investigation into the spatiotemporal dynamics of Net Primary Productivity (NPP) in the Qinling Mountains and its response mechanisms to extreme climate events. By integrating trend analysis, ridge regression, and spatial autocorrelation techniques, the main conclusions are as follows:

(1)

From 2001 to 2023, the multi-year average NPP in the Qinling region was 558.43 ± 134.27 gC·m⁻²·year⁻¹, with a statistically significant upward trend at a rate of 5.44 gC·m⁻²·year⁻¹. Among the subregions, the order of NPP performance was QL_Sx > QL_Gs > QL_Hn, indicating notable differences in vegetation productivity trends across the region.

(2)

During the period 1960–2023, all temperature-related extreme indices (ID, TN10p, TR, and CSDI) showed significant downward trends (ranging from −0.49 to −1.19 days/decade, p < 0.05), while the extreme precipitation index R20 exhibited a significant upward trend (+0.28 days/decade, p < 0.05). Other precipitation indices showed no statistically significant trends, suggesting that changes in temperature extremes have been more pronounced than those in precipitation.

(3)

Pearson correlation analysis indicated strong positive correlations among temperature-related indices (e.g., TR and TN10p, r = 0.636, p < 0.001), with even stronger correlations among precipitation indices (e.g., R10 and R20, r = 0.880, p < 0.001). CDD showed negative correlations with precipitation intensity indices. Spatial autocorrelation analysis revealed that the Moran’s I values for precipitation indices generally remained above 0.60, reflecting consistent spatial clustering. In contrast, the spatial patterns of temperature indices exhibited greater temporal variability.

(4)

The relative contributions of individual extreme climate indices to NPP variation across the region were ranked as follows: TN10p (29.43%) > R10 (17.64%) > R20 (13.12%) > TR (11.71%) > CDD (10.41%) > CSDI (7.98%) > Rx5day (7.52%) > ID (2.19%). Spatial analysis showed that TN10p had the highest average positive absolute contribution to NPP (3.53 gC·m⁻²·year⁻¹), while R10 contributed notably in the QL_Gs and QL_Sx subregions (2.27 and 2.44 gC·m⁻²·year⁻¹, respectively). In contrast, Rx5day exhibited a negative contribution to NPP (–0.77 gC·m⁻²·year⁻¹). Temperature-related indices played a more dominant role in NPP variation in QL_Sx and QL_Hn, whereas precipitation-related indices were the primary drivers in QL_Gs.