Multi-time scale impact analysis of GPP-SIF based on random forest model

Global climate change is a hot issue in the world today, and the terrestrial carbon cycle system is an important part of climate change. Gross primary productivity (GPP) is one of the important indicators to study the carbon cycle of ecosystem. Remote sensing data of solar-induced chlorophyll fluorescence(SIF) provide a method for spatial scale expansion of GPP. Exploring the relationship between GPP-SIF in DingHu Mountain region from different time scales can clarify the utility of SIF in estimating carbon assimilation, comprehensively understand the environmental factors controlling forest productivity in DingHu mountain region, and have great significance for predicting the future carbon cycle feedback of the ecosystem. Environmental factors have a great influence on plant photosynthesis, which reduces the accuracy of GPP-SIF linera relationship model, and the cause of nonlinear part needs to be further studied. Taking DingHu mountain evergreen ecosystem as an example, the linear relationship of GPP-SIF on monthly, seasonal and half-year timespan was fitted to explore the dynamic changes of the two timespan. On this basis, the Random Forest Regression model was used to analyze the importance of different environmental factors to GPP-SIF. The results show that the linear correlation between GPP and remote sensing products of SIF decreases significantly at higher time frequency. The environmental factors that have great influence on SIF and GPP at different time scales are air temperature and relative humidity, and the differences become smaller as the temporal scale increses, providing evidence for the changes in GPP-SIF.