Climate–Vegetation Coverage Interactions in the Hengduan Mountains Area, Southeastern Tibetan Plateau, and Their Downstream Effects

Known as the third pole of the world, the Tibetan Plateau (TP) is the highest and largest topographic feature on Earth, with a very fragile ecological environment (Duan et al., 2006). Environmental changes on the TP significantly modulate local atmospheric dynamic and thermal conditions, which also affect climate in downstream regions and even globally (Wang et al., 2017; Wu et al., 2018; Ma et al., 2021). As a lasting hotspot of global changes, the accelerated warming rate over the TP has been double that of the global average in recent decades (Dorji et al., 2022). The evident warming could impact the biosphere over the TP, such as increasing vegetation coverage (Zou et al., 2020), which in return changes local air temperature and moisture content as well as East Asian monsoon circulation (Shen et al., 2015; Hua et al., 2019). The climate–vegetation interaction on the TP is vital for predicting regional and global climate and eco-environmental changes.

Located in the southeast of the TP, the Hengduan Mountains Area (HMA) has particular geographical and climatic significance. There are many habitable zones in the HMA’s river valleys, leading to much higher population density compared with other regions of the TP. However, meteorological and geological hazards are also frequent in this region, limiting the scope for economic development (Peng et al., 2020). In addition, environmental changes in the HMA could greatly affect the climate in its downstream regions, such as the Sichuan Basin (SCB), a vital and densely populated area adjacent to the eastern side of the HMA (Dong et al., 2019; Tao et al., 2021). The weather systems and anomalous climatic signals that generate over the HMA generally move or expand over the SCB along the westerly flow, resulting in weather and climate extremes there (Fu et al., 2019; Xu et al., 2019). Therefore, it is crucial for sustainable regional development to understand the interactions between changes in climate and the eco-environment in the HMA, as well as consider their possible downstream climatic effects.

Progress has been made on the climate–vegetation coverage interaction in the HMA. For instance, in terms of the influence of climate changes on vegetation coverage changes, it has been noted that temperature and precipitation variations have jointly contributed to the vegetation coverage variation in the HMA in recent decades, whereas the main climate factors that regulate changes in vegetation coverage vary seasonally and regionally; in general, the temperature (precipitation) variation makes a higher contribution in a relatively cold (dry) environment (Tao et al., 2018; Zhang et al., 2019; Piao et al., 2020). Moreover, changes in vegetation coverage can thereby modulate water and heat fluxes between the land and atmosphere and eventually affect climate variability, demonstrating the process of vegetation-driven climate feedback (Fredlund and Xing, 1994; Green et al., 2017). Vegetation greening reduces the surface albedo so that the amount of incoming solar radiation absorbed by the surface increases. According to Kirchhoff’s law, the release of longwave radiation from the surface to the atmosphere also increases; thus, there is a warming effect on the atmosphere (Yu et al., 2016; Winckler et al., 2019). On the other hand, the increase in vegetation coverage also enhances surface evapotranspiration, reduces soil moisture, and increases atmospheric water vapor content, leading to an increase in cloud cover and thus a decrease in solar radiation; therefore, there is a cooling and humidification effect on the atmosphere (Unger, 2014; Lian et al., 2020; Xue et al., 2021). So far, the effects of changing vegetation coverage in the HMA on the local water and heat flux and on climate changes in downstream regions remain inconclusive. Some studies have suggested that the increase in vegetation coverage of the southeastern TP has led to an increase in heat flux from the surface to the atmosphere, thus strengthening the South Asian high and the East Asian monsoon (Lai and Gong, 2017; Hua et al., 2019), whereas other studies have concluded that the TP greening has reduced the surface temperature and ultimately weakened the East Asian monsoon (Zuo et al., 2011; Shen et al., 2015). In brief, previous studies have revealed the characteristics of vegetation variability and its relationship with climate variability in the HMA. However, there is little reference to the nature of the feedback that vegetation exerts on climate in the region, especially regarding the coupled effect between climate and vegetation coverage and its impacts on the climate of downstream regions, such as the SCB.

Accordingly, this study focused on the interaction between changes in climate factors (i.e., temperature and precipitation) and vegetation coverage in the HMA, with an emphasis on the most significant features of these changes and a demonstration of how they affect the SCB climate. We first analyzed the changes in climate and vegetation coverage in the HMA, then explored the mutual feedback mechanisms between them, and finally investigated the influences that coupled climate–vegetation coverage changes in the HMA have had on the SCB climate. Our study of the climate–vegetation interaction in this climate-sensitive region is of significance for evaluating the climatic and vegetation growth responses under foreseeable warming scenarios.

The HMA (94°–104°N, 24°–34°E) is located in the southeastern part of the TP, Southwest China (Fig. 1). The downstream region, the SCB, is on the east side of the HMA. The elevation of the HMA is high in the northwestern part and low in the southeastern part, ranging from 500 m to more than 7000 m (Fig. 1). Compared with the HMA, the SCB has a relatively low and flat topography with an average elevation of less than 500 m.

Figure 1.  Topographic map of the HMA and SCB. The ranges of the HMA and SCB are outlined by the black border and green box, respectively, with the same indication in the following figures. The thick (thin) purple lines plot the national (provincial) boundaries. The subgraph in the left bottom corner displays the position of the TP (yellow border), HMA, and SCB in China.

The distributions of annual-mean temperature, precipitation, and relative humidity in the HMA are opposite to those of altitude, with lower values in the northwest and higher values in the southeast (Figs. 2a–c). Annual precipitation increases from less than 500 mm (semiarid environment) to more than 1000 mm (humid environment) from the northwest to the southeast of the HMA. Sunshine is abundant in the HMA. The average annual sunshine duration is generally above 2000 h, and only the areas near the SCB have less sunshine, with a sunshine duration of about 1400 h (Fig. 2d). Temperature, precipitation, and relative humidity in the HMA are high in summer and low in winter (Fig. 2e), showing a monsoonal climate characteristic. However, the seasonal variation in sunshine duration is opposite to the other three variables, implying that the increase in cloud cover decreases the sunshine duration in the rainy season.

Figure 2.  Spatial distribution of climatological-mean annual (a) temperature (units: °C), (b) precipitation (units: mm), (c) relative humidity (%), and (d) sunshine duration (units: h) during 1982–2015 in the HMA and SCB. Panels (e) and (f) are the climatological-mean seasonal precipitation (bars), temperature (red line), relative humidity (blue line), and sunshine duration (yellow line) averaged across the HMA and SCB, respectively, where the bars indicate ±1 standard deviation. The meteorological data used were the interpolated data from meteorological stations.

The SCB is relatively warm and humid, and there is less sunshine than in the HMA. The average annual temperature is above 16°C, annual precipitation is above 800 mm, annual relative humidity is above 75%, and annual sunshine is below 1400 h (Figs. 2a–d). The seasonal changes in meteorological variables in the SCB also present a monsoonal climate characteristic (Fig. 2f).

The normalized difference vegetation index (NDVI), a widely used parameter, was employed to indicate vegetation changes (Piao et al., 2003; Geerken et al., 2005; Sun et al., 2020). The NDVI data were obtained from the Global Inventory Monitoring and Modeling System (GIMMS) 3g.v1 product from the National Oceanic and Atmospheric Administration. This product was derived from remote sensing observations gathered by the Advanced Very High Resolution Radiometer, with radiometric calibration and geometric correction and thus high accuracy (Tucker et al., 2005; Pinzon and Tucker, 2014). The NDVI data have global coverage from July 1981 to December 2015 with a spatial resolution of (1/12)° × (1/12)° (≈ 8 km) and a temporal resolution of bimonthly. We averaged the bimonthly NDVI data into seasonal means for analysis. The NDVI varies from −1 to 1, with higher values indicating higher vegetation coverage.

Monthly meteorological observation data (including temperature, precipitation, relative humidity, and sunshine duration) were obtained from the China Meteorological Data Service Center (http://data.cma.cn/), which covers 1142 meteorological stations all over the HMA, SCB, and nearby areas. The station data were interpolated onto the grid of NDVI data with the iterative improvement objective analysis method (Barnes, 1964) using the NCAR Command Language application, where scanning radii of 4°, 3°, 2° and 1° were employed in sequence when performing the interpolation. Data for geopotential height at multiple pressure levels and the thermal flux between the Earth’s surface and atmosphere (including the net solar radiation absorbed by the surface and the net thermal radiation, sensible heat flux, and latent heat flux from the surface to the atmosphere) were obtained from the fifth generation European Centre for Medium-Range Weather Forecasts reanalysis (ERA5) atmospheric and land subsets, respectively, with spatial resolutions of 0.25° × 0.25° and 0.1° × 0.1°, respectively (Hersbach et al. 2023). Before analysis, these data were interpolated onto the NDVI grid via bilinear interpolation.

In this study, spring, summer, autumn, and winter refer to March to May, June to August, September to November, and December to February of the following year, respectively. Considering the temporal coverage of the NDVI dataset, the study period for all investigations was confined to 1982–2015.

The trends of the meteorological variables and NDVI were calculated and tested. The trend coefficient (s) of a variable X during a given period was calculated by least squares linear regression and can be expressed as

where n is the total number of years and X_i indicates the value of X in the ith year. Then, we examined whether the time series of variable X increases/decreases significantly within a given period via the nonparametric Mann–Kendall (M–K) trend test (Mann, 2004), which examines the ranks (r) of time series. In this method, a normally distributed statistic Z was defined as follows:

where the meaning of n is the same as in Eq. (1), r_i is the rank of series X at time i and represents the number for X_j > X_i, where j = i, i+1, …, n, indicating the time after time i; thus, r_i ranges from 0 to n−i−1. When | Z | > Z_α, the trend of the time series is determined to be significant at the significance level of α.

We employed linear regression to examine the relationship among the variables. For variable Y regressed on variable X, the regression model is written as

where a, b, and ε denote the intercept, regression coefficient, and error, respectively. The magnitude of the regression coefficient represents the degree of the effect that X has on Y. The significance level (p) of the regression coefficient is determined by the t-test.

Multivariate linear regression was applied to investigate the contribution of climate changes to NDVI changes. Temperature (T) and precipitation (P) were used as predictors for NDVI regression. The regression equation is written as follows:

where β₁ and β₂ are partial regression coefficients for T and P, respectively. NDVI, T, and P series are normalized before regression so that the respective contributions of T and P to NDVI can be determined according to values of β₁ and β₂. The explained variance (R²) for the regression equation was also calculated to determine the total contribution of climate factors to NDVI. The significance of partial regression coefficients is determined by the t-test.

The trends of variation in temperature and precipitation in the HMA during 1982–2015 were analyzed. The temperature in the HMA (Figs. 3a–d) shows increasing trends over the four seasons, and the areas with significant (p < 0.05) warming trends are widespread throughout almost the entire HMA. In terms of the regional-mean seasonal temperature in the HMA, the warming trends are highest in winter (0.05°C yr⁻¹), then spring (0.039°C yr⁻¹), then autumn (0.035°C yr⁻¹), and lowest in summer (0.028°C yr⁻¹) (Figs. 4a–d). The trends of precipitation in the HMA (Figs. 3e–h) generally exhibit an increase in spring and summer and a decrease in autumn and winter, but only the increased precipitation in spring in the northern HMA and the decreased precipitation in autumn in the southwestern HMA exceed the 0.05 significance level. For the regional-mean precipitation in the HMA, the increasing trend in spring and decreasing trend in the other seasons are all insignificant (Figs. 4a–d), and the mismatches between the temperature and precipitation series indicate they vary independently on the interannual time scale. These results are consistent with those of previous research (Li et al., 2010; Hu et al., 2019; Wang et al., 2021).

Figure 3.  Spatial distribution of trend coefficients of (a) spring-, (b) summer-, (c) autumn-, and (d) winter-mean temperature (units: °C yr⁻¹) during 1982–2015 in the HMA and SCB. Panels (e–h) are similar to (a–d) but for the precipitation anomaly percentage (units: % yr⁻¹). Areas with trends exceeding the 0.05 significance level according to the M–K trend test are hatched. The interpolated meteorological station observed data were used.

Figure 4.  Time series of (a, e) spring-, (b, f) summer-, (c, g) autumn-, and (d, h) winter-mean temperature (orange lines) and precipitation (blue lines) averaged within the (a–d) HMA and (e–h) SCB during 1982–2015. The thick solid lines plot their linear trends. The trend coefficients (s) are given in the upper right corner with the same color as the lines and are set in bold when exceeding the 0.05 significance level according to the M–K trend test.

Next, we analyzed the seasonal vegetation coverages and their trends in the HMA during 1982–2015. The climatological-mean NDVI (Fig. 5) in the northern HMA presents noticeable seasonal changes, with the lowest vegetation coverage in winter and the highest vegetation coverage in summer, which is consistent with the seasonal changes in local temperature and precipitation (Fig. 2e). However, the seasonal vegetation variability in the southern HMA is minimal (Fig. 5) since the vegetation coverage is consistently high throughout the year (NDVI > 0.5). Relatively low altitude (Fig. 1) and adequate rainfall, heat, and sunshine conditions (Figs. 2a–d) may be the main reasons for the weak seasonal variation in vegetation coverage in these regions.

Figure 5.  Spatial distribution of climatological-mean (a) spring, (b) summer, (c) autumn, and (d) winter NDVI during 1982–2015 in the HMA and SCB.

Figure 6 displays the trends of NDVI in the HMA during 1982–2015. The NDVI all over the HMA has an increasing trend in each season except summer, when the NDVI decreases in the southern HMA. Areas where the NDVI significantly (p < 0.05) increases expand across the whole HMA during spring and winter. In autumn, areas with significant increasing trends are only distributed in the southwestern HMA. The regional-mean NDVI in the HMA (Figs. 7a–d; black lines) shows a decreasing trend in summer and does not reach the 0.05 significance level (−1.6 × 10⁻⁴ yr⁻¹), while the regional-mean NDVI in other seasons increases significantly. The seasons with increasing trends are, from largest to smallest, as follows: spring (9.3 × 10⁻⁴ yr⁻¹), winter (8.5 × 10⁻⁴ yr⁻¹), and autumn (5.6 × 10⁻⁴ yr⁻¹). The results regarding the vegetation coverage variability agree with the results from previous studies (Hu et al., 2019; Yin et al., 2020; Wang et al., 2021).

Figure 6.  Spatial distribution of the trend coefficients of (a) spring, (b) summer, (c) autumn, and (d) winter NDVI (units: 10⁻² NDVI yr⁻¹) during 1982–2015 in the HMA and SCB. Areas with trends exceeding the 0.05 significance level according to the M–K trend test are hatched.

Figure 7.  Time series of the (a) spring-, (b) summer-, (c) autumn-, and (d) winter-mean NDVI (black lines) and temperature (orange lines) averaged across the HMA during 1982–2015. Panels (e–h) are similar to (a–d) but for the SCB. The thick solid lines plot their linear trend. The letters s and R indicate the trend coefficient of NDVI and the correlation coefficient between the NDVI and temperature, respectively, with bold font used when exceeding the 0.05 significance level.

The effect of changing climates on vegetation coverage was analyzed. We conducted point-to-point multivariate regression analysis on NDVI with temperature and precipitation as predictors. It was found that the NDVI in the HMA is mainly positive in response to temperature changes (Figs. 8a–d). In other words, the vegetation coverage increases when the temperature increases. Such an effect is more significant in spring and winter than in summer and autumn and more significant in the northern HMA than in its south. Referring to the spatial and seasonal distributions of air temperature (Figs. 2a and e), it can be seen that under relatively cold conditions, air temperature is the main factor limiting vegetation growth, so the vegetation coverage is more sensitive to temperature changes (Prevéy et al., 2017; Tao et al., 2018; Wang et al., 2021). However, the summer NDVI in the southwestern HMA responds negatively to temperature changes (Fig. 8b). This may be because the water and heat conditions in this region are suitable for vegetation growth year-round. When the temperature further increases, plant respiration is enhanced, but photosynthesis is no longer enhanced, which is not conducive to plant growth (Michaletz et al., 2014). This situation has also been found in other warm and moist regions of China (Wang et al., 2010; Xu et al., 2014). For the regional-mean NDVI and temperature in the HMA (Figs. 7a–d), the positive correlations (R) between them are significant (p < 0.05) in all seasons except summer, with the highest correlation in winter (R = 0.7).

Figure 8.  The results of in-situ temperature and precipitation regressed on NDVI. Spatial distribution of partial regression coefficients for (a–d) temperature (β₁) and for (e–h) precipitation (β₂) and (i–l) the explained variance (R²). Panels (a, e, i) are for spring, (b, f, j) for summer, (c, g, k) for autumn, and (d, h, l) for winter. In (a–h), areas with partial regression coefficients exceeding the 0.05 significance level according to the t-test are hatched.

The responses of NDVI to precipitation changes are negative in most of the HMA (Figs. 8f–h); thus, the vegetation coverage decreases when precipitation increases. However, this effect exceeds the 0.05 significance level only in the southern HMA in summer and autumn and the central HMA in winter. Compared with the northwestern HMA, precipitation is abundant in the central and southern HMA (Fig. 2b). Precipitation in this region is not the main factor that limits vegetation growth. Therefore, on the one hand the negative response of the NDVI to precipitation may be due to soil erosion caused by excessive rain, which destroys the vegetation growth environment (Liang et al., 2015; Qu et al., 2018); while on the other hand, it may be due to the greater cloud cover when rainfall is stronger, which inhibits photosynthesis (Piao et al., 2003; Wang et al., 2021). The correlation coefficients between the regional-mean NDVI and precipitation in the HMA are insignificant (p > 0.05) for all the seasons (not shown).

The explained variance (R²) was used to determine the total contribution of climate factors (i.e., temperature and precipitation) to NDVI changes. The contribution of climate factors to NDVI in winter (Fig. 8l) is the highest, with R² greater than 0.3 in most of the HMA, and the contributions of climate factors in spring, autumn, and summer decrease gradually (Figs. 8i–k). In particular, the R² in summer does not exceed 0.3 in most regions, indicating that the contribution of climate factors to NDVI changes in the plant growing season is small. Previous studies have attributed this situation to the influence of human activities, such as changes in land use and carbon dioxide emissions (Qu et al., 2018; Yin et al., 2020; Wang et al., 2021).

According to the above results, we can explain the trends of vegetation coverage induced by HMA climate changes in recent decades. The increasing trends of vegetation coverage in the spring and winter HMA (Figs. 6a and d) have been mainly due to the significant warming trend (Figs. 3a and d). The decreasing trend of precipitation in the central HMA in winter (Fig. 3h) also makes a certain contribution to the increase in local vegetation coverage. In autumn, the warming trend and precipitation decreasing trend (Figs. 3c and g) in the southern HMA have contributed to the increase in vegetation coverage there (Fig. 6c). The warming trend in the southwestern HMA in summer (Fig. 3b) has contributed partly to the decline in vegetation coverage in this region (Fig. 6b).

Additionally, we investigated the feedback of vegetation coverage changes to the heat flux between the surface and atmosphere in the HMA. Figure 9 illustrates the trend of heat flux between the surface and atmosphere in the HMA during 1982–2015. Figure 10 presents the effect of NDVI changes on local heat flux changes in the HMA via point-to-point regression. It was found that net solar radiation absorbed by the surface generally follows an increasing trend in all seasons except summer (Figs. 9a–d). The areas with significant increasing trends are distributed in the northwestern HMA in spring, the southwestern HMA in autumn, and the central HMA in winter. For all seasons, there is a positive correlation between the NDVI and net solar radiation absorbed by the surface in the HMA, and the areas with significant correlations are distributed throughout the western HMA in spring and autumn, the eastern HMA in summer, and the entire HMA in winter (Figs. 10a–d). Combined with the distribution of the NDVI variability, we can infer that the significant increase in vegetation coverage in the northwestern, southwestern, and central HMA during spring, autumn, and winter, respectively (Figs. 6a, c and d), reduces the local surface albedo and thus significantly increases the absorption of solar radiation locally, and the decrease in vegetation coverage in the eastern HMA in summer (Fig. 6b) may lead to a decrease in net solar radiation absorbed by the surface. The trend of surface net thermal radiation toward the atmosphere in the HMA and its correlation with the NDVI (Figs. 9e–h and 10e–h) are generally consistent with the trend of net solar radiation absorbed by the surface and its correlation with the NDVI (Figs. 9a–d and 10a–d). This result is consistent with Kirchhoff’s law of thermal radiation. Thus, our inference regarding net solar radiation absorbed by the surface can be applied to surface net thermal radiation toward the atmosphere. There are increasing trends of surface-to-atmosphere sensible heat flux (Figs. 9i–l) and latent heat flux (Figs. 9m–p) in most of the HMA, but the increasing trends are insignificant in most areas and much weaker than the rising trends of net thermal radiation. Generally, the increase in evapotranspiration caused by vegetation greening can increase the surface latent heat flux, but the accompanying decrease in land surface temperature may reduce the surface sensible heat flux (Unger, 2014; Chen et al., 2020; Xue et al., 2021). The positive correlation between latent heat flux and vegetation coverage in the HMA conforms to this principle (Figs. 10m–p), but the sensible heat flux in our case does not present a negative correlation with vegetation coverage (Figs. 10i–l). This may be due to the complex surface conditions in the HMA, which complicate the relationship between vegetation coverage and sensible heat flux. For example, in some areas of the TP, the increase in the NDVI caused by snow melting can also significantly increase the surface sensible heat flux (Dai et al., 2016; Xie et al., 2018; Wang et al., 2019).

Figure 9.  Spatial distribution of trend coefficients of net solar radiation absorbed by the surface in (a) spring, (b) summer, (c) autumn, and (d) winter during 1982–2015 in the HMA and SCB. Panels (e–h), (i–l), and (m–p) are similar to (a–d) but for surface-to-atmosphere net thermal radiation, sensible heat flux, and latent heat flux, respectively. The units are W m⁻² yr⁻¹. Areas with trends exceeding the 0.05 significance level according to the M–K trend test are hatched. The flux data were obtained from ERA5.

Figure 10.  Spatial distribution of regression coefficients for net solar radiation absorbed by the surface regressed on the in-situ NDVI in (a) spring, (b) summer, (c) autumn, and (d) winter during 1982–2015 in the HMA and SCB. Panels (e–h), (i–l), and (m–p) are similar to (a–d) but for surface-to-atmosphere net thermal radiation, sensible heat flux, and latent heat flux, respectively. The units are W m⁻² yr⁻¹. Areas with regression coefficients exceeding the 0.05 significance level according to the t-test are hatched. The flux data were obtained from ERA5.

The time series of the regional-mean surface-to-atmosphere net heat flux (NHF) in the HMA during 1982–2015 are shown in Figs. 11a–d (orange lines). The NHF is obtained by summing the surface-to-atmosphere net thermal radiation, sensible heat flux, and latent heat flux. The surface-to-atmosphere NHF in the HMA presents an increasing trend during all four seasons, which can be ranked from largest to smallest as follows: spring (0.41 W m⁻² yr⁻¹), autumn (0.37 W m⁻² yr⁻¹), winter (0.35 W m⁻² yr⁻¹), and summer (0.09 W m⁻² yr⁻¹). Only the summer NHF increase does not reach the 0.05 significance level. The correlation coefficients between the NHF and NDVI in the HMA are significantly positive in all seasons except autumn. Therefore, for the entire HMA, the increasing trend of vegetation coverage in spring and winter contributes significantly to the increasing trend of the surface-to-atmosphere NHF.

Figure 11.  Time series of the (a) spring-, (b) summer-, (c) autumn-, and (d) winter-mean NDVI (black lines) and NHF (orange lines) averaged across the HMA during 1982–2015. Panels (e–h) are similar to (a–d) but for the SCB. The thick solid lines plot their linear trend. The letters R and s indicate the correlation coefficient between the NDVI and NHF and the trend coefficient of NHF, respectively, with bold font used when exceeding the 0.05 significance level.

In summary, the increase in vegetation coverage during all seasons except summer in the HMA in recent decades has had a warming effect on the atmosphere, mainly driven by the increase in surface-to-atmosphere net thermal radiation. Moreover, the increases in sensible (latent) heat flux have also made a certain contribution to the vegetation-driven warming in part of the HMA during spring and winter (autumn). The decreasing trend of summer vegetation coverage has also reduced the surface-to-atmosphere NHF in part of the HMA, but these changes are all insignificant and would have been unable to have a significant impact on regional climate changes. It should be noted that vegetation coverage change also modulates hydrological factors, such as surface evapotranspiration, runoff, and soil moisture infiltration (Liang et al., 1994; Bernacchi and Vanloocke, 2015; Li et al., 2018), and then acts on the variability of precipitation and the NHF between the surface and atmosphere. However, our study did not consider changes in hydrological factors, and so this remains to be studied in future work.

Combining the above results regarding the mutual effects between changes in climate and vegetation coverage in the HMA, we can see that the temperature and vegetation coverage changes in winter and spring constitute a process of mutual facilitation. Therefore, it is inferred that the warming and greening trend in the HMA during winter and spring may continue with positive temperature–vegetation coverage coupling. However, this process of mutual facilitation cannot proceed indefinitely because the increase in vegetation coverage can also lead to an increase in carbon dioxide absorption in the atmosphere and an increase in cloud cover, both of which produce a cooling effect (Winckler et al., 2019; Piao et al., 2020). In summer and autumn, the coupled temperature–vegetation coverage effect in the HMA is insignificant. Nevertheless, if the precipitation variation is involved, the coupled climate–vegetation coverage effect may have a certain significance in the southern HMA, where the significant precipitation decreasing trend contributes partly to the NDVI increase and the vegetation greening enhances the latent heat flux from the surface to the atmosphere.

Before investigating the impacts of coupled changes in climate and vegetation coverage in the HMA on climate variability in the SCB, we briefly analyze the interactions of changes in climate and vegetation coverage in the SCB during 1982–2015. Similar to the HMA, the SCB temperature (Figs. 3a–d) exhibits an increasing trend in all four seasons. From large to small, the warming trends for the four seasons are spring (0.49°C yr⁻¹), summer (0.3°C yr⁻¹), autumn (0.27°C yr⁻¹), and winter (0.25°C yr⁻¹), and only the increasing trend in winter does not reach the 0.05 significance level (Figs. 4e–h). The precipitation in the SCB has a decreasing trend in summer and winter and an increasing trend in spring and autumn, but these trends are all insignificant (Figs. 3e–h and 4e–h). The regional-mean NDVI in the SCB (Figs. 7e–h) increases in all four seasons, and the rates of increase from large to small are spring (27.7 × 10⁻⁴ yr⁻¹), autumn (19 × 10⁻⁴ yr⁻¹), winter (11.2 × 10⁻⁴ yr⁻¹), and summer (10.1 × 10⁻⁴ yr⁻¹), which are all greater than those in the HMA (Figs. 7a–d). In terms of the effect of climate on vegetation coverage, spring temperature strongly influences the NDVI variability in the SCB, while the temperature in the other seasons has only a mild effect on NDVI variabilities, and precipitation in all seasons has little effect on vegetation coverage changes (Figs. 7e–h and 8). Therefore, the rise in spring temperature in the SCB (Fig. 4e) has largely driven the increase in vegetation coverage (Fig. 7e). In terms of vegetation feedback on climate, the regional-mean surface-to-atmospheric NHF in the SCB increases in all seasons and is positively correlated with the NDVI, whereas only the increasing trend of the surface-to-atmospheric NHF in spring (0.47 W m⁻² yr⁻¹) reaches the 0.05 significance level (Figs. 11e–h). Therefore, the significant increase in spring vegetation coverage in the SCB also has a significant warming effect on the atmosphere, which is mainly accomplished via the significant increase in surface-to-atmosphere net thermal radiation and latent heat flux (Figs. 9e and m; Figs. 10e and m). In summary, the coupled effect between temperature and vegetation coverage changes in the SCB is only significant during spring, which is different from in the HMA, where the coupled effect during winter is the strongest.

To analyze the influence that the coupled effect of temperature and vegetation coverage changes in the HMA has had on the SCB, the temperature anomalies in the HMA and SCB were regressed on the regional-mean surface-to-atmosphere NHF series of the HMA during the corresponding season (Figs. 12a–d). Except in autumn, the surface-to-atmosphere NHF over the HMA has a significant (p < 0.05) positive correlation with the temperature in the SCB. However, we conducted the same regression analysis on the surface-to-atmosphere NHF series of the SCB and found that the areas with significant (p < 0.05) positive temperature anomalies are only located within the SCB and along its eastern side, but not within the HMA. This situation indicates that the NHF change over the HMA may have affected the temperature of the SCB, but the reverse of this effect is not apparent.

Figure 12.  Spatial distribution of regression coefficients for temperature (units: °C) in (a) spring, (b) summer, (c) autumn, and (d) winter regressed on the surface-to-atmosphere NHF series of the HMA in the same season during 1982–2015. Panels (e–h) are similar to (a–d) but for those regressed on the NHF series in the SCB. Areas with regression coefficients exceeding the 0.05 significance level according to the t-test are hatched. The flux data were obtained from ERA5; the temperature data were from our interpolated station data.

To explain the associated atmospheric dynamic mechanism, we regressed upper-air (200 hPa) geopotential height anomalies on the NHF series of the HMA during the corresponding season. There is a high-pressure anomaly center along the eastern side of the HMA in spring and winter (Figs. 13a and d). However, after the same regression analysis was conducted on the NHF of the SCB, it was found that the significant geopotential height anomalies are mainly distributed at the middle and high latitudes and follow a pattern of atmospheric wave trains (Figs. 13e–g), which cannot be attributed to the influence of SCB NHF variability. Therefore, the mechanism through which the coupled temperature–vegetation coverage effects in the HMA influence the climate variability in the SCB can be identified. In winter and spring, the increase in temperature and vegetation coverage within the HMA heats the atmosphere by enhancing the surface-to-atmosphere NHF. Atmospheric heating thickens the atmospheric column and increases the upper-air pressure. The high-pressure anomaly disperses downstream along the westerly flow and covers the upper air of the SCB, leading to a positive temperature anomaly in the SCB.

Figure 13.  Similar to Fig. 12 but for geopotential height at the 200-hPa level (units: gpm) regressed on the NHF series. The flux and geopotential height data were obtained from ERA5.

We also considered the cross-seasonal effects of NHF changes over the HMA on SCB climate variability. Figure 14 displays the regressions of spring temperature and the 200-hPa geopotential height on the HMA NHF series in the winter of the previous year. The surface-to-atmosphere NHF anomaly over the HMA in winter also has a significant impact on upper-air pressure and near-surface air temperature anomalies in the SCB during the following spring. Therefore, the influence that atmospheric heating anomalies over the HMA in winter (Fig. 11d) have on atmospheric circulation and air temperature over the SCB can persist until the following spring, which may be a reason why the most significant coupled effect between temperature and vegetation coverage changes occurs during winter in the HMA, while it is during spring in the SCB.

Figure 14.  Spatial distribution of regression coefficients for spring (a) temperature and (b) geopotential height at 200 hPa regressed on the NHF series of the HMA in the winter of the previous year during 1982–2015. Areas with regression coefficients exceeding the 0.05 significance level according to the t-test are hatched. The flux and geopotential data were obtained from ERA5; the temperature data were from our interpolated station data.

This study investigated the interactions between climate change and vegetation coverage change in the HMA and their downstream impacts on the SCB. Mutual facilitation between rising air temperature and increasing vegetation coverage was found in the HMA in recent decades. This effect is very significant in winter, followed by spring. The local vegetation coverage increases with rising temperature, and the greening of vegetation reduces the surface albedo and in turn heats the atmosphere. Anomalous atmospheric heating over the HMA thickens the atmospheric column and increases the upper-air pressure. The upper-air high-pressure anomaly is dispersed downstream over the SCB via the westerly flow and eventually enhances the SCB temperature. This effect lasts from winter until the following spring, contributing to the maximum increases in springtime temperature and vegetation coverage in the SCB.

Previous studies have separately focused on the effects of climate changes on vegetation coverage changes, the feedback of vegetation coverage changes to climate changes, and the downstream effects of vegetation changes. In this study, the three processes were integrated into a coherent map. Our findings can be applied to predict regional eco-environmental and climate changes in the HMA and SCB under warming scenarios and assess simulations regarding regional climate and vegetation variabilities and their mutual feedback. Our results imply that the warming and greening trends are likely to continue in the future in the HMA and SCB during the non-growing season or early growing season, while this trend could be reversed if hydrological factors play a controlling role. If the HMA vegetation and air temperature exhibit positive anomalies in winter, the SCB air temperature in the following spring may be higher and the climate will be conducive to vegetation growth.

Note that there are limitations in our study. The analysis period was limited to 1982–2015 because of the availability of the GIMMS NDVI data; therefore, our results do not reveal the characteristics in recent years. The MODIS NDVI data cover recent years but start from 2001 with a length of less than 30 years (i.e., a complete climatological period), so the analysis of climate change during such a short period may be unconvincing. Our future work will compare the results from multisource NDVI data and explore the coupled effect between climate and vegetation coverage changes on longer time scales. Another limitation is related to the analysis methods. Our analyses were based on linear approaches, but the climate system is nonlinear. Therefore, the results may fail to capture the nonlinear changes in climate–vegetation coverage coupling. For example, the mutual facilitation between air temperature and vegetation coverage may be terminated if the increases in air temperature and vegetation coverage exceed a certain threshold. In addition, variabilities in climate and vegetation coverage are also not stationary, and they may abruptly change at certain times. These issues will also be studied in future work.

Acknowledgements. This study was jointly supported by the National Natural Science Foundation of China (Grant Nos. 42205059 and 42005075), the Strategic Priority Research Program of the Chinese Academy of Sciences (Grant Nos. XDA23090303 and XDB40010302), the State Key Laboratory of Cryospheric Science (Grant No. SKLCS-ZZ-2024 and SKLCS-ZZ-2023), and the Key Laboratory of Mountain Hazards and Earth Surface Processes.