Disentangling gross primary productivity drivers of forested areas in China and its climate zones from 1990 to 2018

Gross primary productivity plays a critical role in global carbon balance. However, quantifying the effects of different drivers still constitutes a challenge due to the different modeling techniques and data used. This study employs spatiotemporal analysis, machine learning, and statistical approaches to measure the significance of forest gross primary productivity drivers in China and its climate zones from 1990 to 2018. The results show that the annual average forest gross primary productivity in China was 914.74 gC m⁻² y⁻¹ during the study period and showed a significantly increasing trend (p < 0.01) at a rate of 4.09 gC m⁻² y⁻¹. Forest gross primary productivity had a southeast-northwest downward spatiotemporal trend with significantly different distributions within the six climate zones, except in the arid and semi-arid zones. A Random Forest model did better than an eXtreme Gradient Boosting model when 10 explanatory variables were used. These variables included the novel forest fragmentation index and climate zones, which helped explain the effects of forest structure and climate characteristics of the climate zones better. The most important forest gross primary productivity drivers in China were mean annual temperature (26.2 %), mean annual precipitation (18.6 %), solar radiation (11 %), forest fragmentation index (8.8 %), and elevation (8.1 %). Using Chatterjee's correlation coefficient, this study provides, for each climate zone, its unique signature regarding the order and importance of the drivers of forest gross primary productivity. This study helps us understand what factors affect forest gross primary productivity in China and its climate zones better by showing how they work using machine learning. These findings may help China reach its carbon neutrality goals.