4.1 Comparison of GC and CCM for causal methods
The causality results for GC were generally lower than those for CCM, and GC showed no causality in the causality test between T and NPP, whereas this relationship showed clearer results in CCM (Fig. 7).
As a widely recognized causality test (Ahsan et al. 2020; He et al. 2020; Németh-Durkó 2021), GC method is designed to identify causal relationships within time series based on predictive power (Granger 1988). A fundamental assumption of the GC approach is the separability of variables, meaning that the cause operates independently of the effect, such that the removal of a cause variable from the system does not impact the remaining elements (Granger 1988; Sugihara et al. 2012). This assumption does not align with real-world scenarios, where there is often a weak coupling between P and T as causal variables. Consequently, the GC method has clear limitations for the current study. To overcome these limitations, CCM method, initially introduced by Sugihara et al. (2012), was employed. This method better compensates for the deficiencies of the GC method, and its accuracy improves with longer time series, resulting in smaller methodological errors.
Ye et al. (2015) discovered that CCM method could not differentiate between bidirectional causality and strong unidirectional causality. Consequently, this study exclusively utilizes the outcomes of the unidirectional causality test and refrains from delving into the role of NPP in modulating climate factors. Furthermore, Bartsev (2021) concurs, emphasizing that CCM constitutes a necessary but not sufficient condition for the presence of causality. They clarify that if X exerts no influence on Y, the outcome is definitive. Nevertheless, should the CCM outcomes suggest that X impacts Y, additional validation of the causal relationship is warranted.
Thus, the findings suggest that the direct causal relationship between T and NPP is not distinctly delineated, a result that holds some merit. Nevertheless, the causal relationship between P and NPP still warrants further examination.
In Fig. 7, there is no direct causal effect of T on the change in NPP, suggesting that T does not directly drive the variation in NPP. This finding is further substantiated by the correlation tests presented in Figs. 8—10. T may function more as an indirect driver, influencing vegetation growth by affecting potential evapotranspiration (Chu et al. 2016; Li et al. 2022). Figures 4 and 8 illustrate that, statistically, NPP tends to be marginally lower in sunny (eastern and southern) vegetation compared to shady (western and northern) vegetation, yet there is no significant correlation with T. This could be attributed to sunny slopes receiving direct sunlight and being warmer, leading to higher evapotranspiration rates and, consequently, slightly reduced NPP compared to shady slopes. In summary, T may only influence NPP changes indirectly.
4.2 Analysis of NPP trends in spatial and temporal changes
Utilizing the Theil-Sen estimator and MK significance test, the results depicted in Figs. 2 and 3 reveal that the eastern part of SHMB exhibits a more pronounced degradation trend, whereas the western part demonstrates a more distinct increase trend. Over the past two decades, the SHMB region has witnessed a notable rise in the NDVI and in the expanse of forest and grassland areas, influenced by policy initiatives such as the "Returning Cultivated Land to Forestry Project." Stand structure plays a pivotal role in forest productivity, influencing resource availability and the intensity of competition (Kweon and Comeau 2019). Complex stand structures can enhance forest carbon storage and production by improving light harvesting capabilities (Forrester et al. 2018). Conversely, the eastern part of the SHMB, which includes the economic hinterland of the Pearl River Delta, experiences intense human-land interactions. High-intensity human activities and crude land management practices have resulted in severe soil erosion, substantial depletion of soil nutrients, and extreme ecological degradation. Despite the replenishment of newly planted plantation forests in terms of area, the forest stand structure remains simplistic, and the carbon sequestration capacity significantly diverges from that of the original forests. Consequently, the eastern region manifests a significant degradation trend.
Among them, Pinus massoniana is a quintessential silvicultural species in subtropical China, constituting approximately 20% of the afforested areas in southern China. However, inefficiently managed Pinus massoniana plantations often suffer from poor growth, severe degradation, and diminished ecological services (Zeqing et al. 2020). The biomass and carbon storage capacity of this species increase over time, reaching peak values at around 17 and 23 years of age, respectively (Ali et al. 2019). The northward shift in the distribution of Sargasso pine at lower latitudes is anticipated to be an inevitable consequence of future climate warming. The southern region, being warmer and more humid than the northern region, exhibits an inverse latitudinal trend in NPP. Furthermore, T emerges as the principal climatic factor contributing to the spatial heterogeneity of NPP for this tree species (Li et al. 2023).
In summary, NPP in the eastern and central regions of the SHMB exhibits a significant degradation trend. This trend is primarily influenced by the simplistic stand structure of plantation forests, the suboptimal realization of ecological benefits, and the dampening effects of climate warming.
4.3 Analysis of response mechanisms between NPP to meteorological factors
The causality analysis, as depicted in Figs. 8 to 10, revealed a more pronounced causal relationship between P and NPP across all the varying topographic factors. Conversely, the influence of the warming trend on the control of vegetation growth demonstrated a weaker causal link, attributable to T surpassing the optimal T threshold for NPP (Huang et al. 2019; Zhang et al. 2022b).
Spatially heterogeneous units under varying gradients of topographic factors exhibit both commonalities and unique characteristics. Notably, in line with findings from other studies, P exerts a significantly more substantial influence on regional NPP than T (Wang et al. 2021a). Zhang et al. (2022a) proposed that in tropical and subtropical climates of low and middle latitudes, humidity serves as a more effective regulator of vegetation activity than T. Nevertheless, T plays an essential role in the synergistic modulation of both factors. Typically, an increase in T can enhance vegetation respiration by expanding the leaf respiratory surface area. However, SHMB is situated in a subtropical zone where T does not constitute a limiting climatic condition for regional vegetation, as the T often surpass the optimal levels for vegetation growth. In recent years, the rise in T has also impacted vegetation growth (Cao et al. 2020). Warming intensifies evapotranspiration from surface waters, enabling T to influence vegetation through the regulation of evapotranspiration. Consequently, T has demonstrated a more pronounced capacity to regulate NPP in a synergistic fashion. Across different topographic gradients, the explanatory power of air T on NPP in the SHMB exhibits a trend of diminishing with varying degrees of gradient. This phenomenon is primarily attributed to the fact that these regions are progressively less affected by human activities, featuring a higher prevalence of deep-rooted woody perennials and complex stand structures, which confer greater resilience to climatic fluctuations (Wang et al. 2003).
The SHMB exhibits a decreasing trend in both Ttrend and Ptrend under both horizontal gradients (from east to west) and vertical gradients (from low to high altitude), as depicted in Fig. 11 and Table S1. This outcome aligns with the findings of Ge et al (2021). Despite the SHMB encompassing small areas of plains, hills, plateaus, and other complex terrains across two thermal belts, and despite a certain degree of spatial heterogeneity, there remains a discernible pattern in the distribution of the various spatially heterogeneous units from east to west within the SHMB. As illustrated in Fig. 11, consistent with previous studies, human activities primarily regulate the region's vegetation productivity by altering the forest stand structure(Chen et al. 2021).
Firstly, the area below 650 meters, represented by the eastern region, is marked by intense human activities, a simplified stand structure, a significant increase in temperature, and enhanced transpiration. Consequently, the interplay of T and P in the eastern region, particularly in the JNM, exerts a substantial influence on NPP. Secondly, the Ttrend and Ptrend of the GP are largely similar to those of JNM. However, GP's main altitude distribution ranges from 650 to 1000 meters, with fewer human activities and a more complex stand structure. This results in no significant correlation on slopes ascending upwards, while the lower altitude areas display a notable positive correlation. Thirdly, the YP is situated at a higher elevation relative to GP and features a more intricate stand structure. As shown in Fig. 2, NPP exhibits a more favorable trend, primarily due to reduced anthropogenic impacts and the absence of more extreme climatic constraints (Liu et al. 2020; Tao et al. 2022). Lastly, the southwestern part of the SHMB is characterized by highland terrain, particularly the SYVH region, where most areas are above 1000 meters in elevation, exhibit high surface roughness, and are subject to climate factors strongly influenced by the synergistic regulation of topography (slope and elevation). Water vapor in this region is generally channeled through the southwest towards the valley areas. Some higher and lower elevation areas remain inaccessible to water vapor due to the surrounding topography, leading to a diverse array of regional microclimates. The heightened volatility and reduced regularity of the SYVH are reflected in the results.