Precipitation Dominates Forest Net Primary Productivity Variations With Distinct Regional Differences in Yunnan Province, China

TL;DR

This study shows that precipitation is the main driver of forest productivity in Yunnan, China, with regional differences and nonlinear effects.

Contribution

The study introduces a comparative framework for analyzing stable and changing forests using random forest and SHAP methods in Yunnan.

Findings

Changing forests show greater NPP variation and spatial heterogeneity than stable forests.

Precipitation has strong synergistic effects on NPP, especially in changing forests.

Key factors like temperature and elevation show nonlinear threshold responses for NPP.

Abstract

In the topographically and climatically diverse region of Yunnan, clarifying the driving mechanisms and threshold effects of factors influencing forest net primary productivity (NPP) is crucial for managing forest carbon sinks. In this study, we established a comparative analysis framework for the stable forest (SF) and the changing forest (CF). Yunnan was then divided into five subregions based on topography and climate. Using a random forest model and the SHAP method, we systematically analyzed drivers of spatiotemporal NPP variations. The results show that CF exhibits greater NPP variation and spatial heterogeneity than SF and maintains stronger trend persistence. At the provincial scale, elevation and precipitation are the main drivers of NPP in SF and CF, respectively, while at the subregional scale, dominant factors differ and include solar radiation, temperature, and forest age, indicating clear scale‐dependent effects. The explanatory power of multiple‐factor interactions for NPP variations is generally higher than that of single factors, with precipitation showing particularly strong synergistic effects. Human activities also have a pronounced impact on NPP in CF. Key driving factors exhibit nonlinear threshold responses, with NPP in SF most suitable under temperature (14°C–20°C), precipitation (100–125 mm), elevation (~2000 m), and forest age (50–70 years), whereas CF shows broader response ranges, reflecting greater adaptability. This study highlights the nonlinear responses and threshold characteristics of forest NPP to multiple driving factors in complex mountainous environments. It also emphasizes the importance of interactions and scale effects in ecological modeling.

We first classified the forests in Yunnan Province into stable forest and change forest regions, and then further divided the province into five sub‐regions based on topography and climate. Based on this classification, we systematically analyzed the spatiotemporal variation of forest net primary productivity (NPP) across these regions. Moreover, by employing the Random Forest model and SHAP analysis, we identified the key driving factors behind the NPP changes in different sub‐regions.

Figures