Revisiting and attributing the global controls over terrestrial ecosystemfunctions of climate and plant traits at FLUXNET sites via causal graphicalmodels

Using statistical methods that do not directly represent the causality between variables to attribute climate and plant traits as controlling ecosystem functions may lead to biased perceptions. We revisited this issue using a causal graphical model, the Bayesian network (BN), capable of quantifying causality by conditional probability tables. Based on expert knowledge and climate, vegetation, and ecosystem function data from the FLUXNET flux stations, we constructed a BN representing the causal relationship of climate-plant-trait-ecosystem functions. Based on the sensitivity analysis function of the BN, we attributed the control of climate and plant traits over ecosystem functions and compared the results with those based on random forests and correlation analysis. The main conclusions of this study include the following: BN can be used for the quantification of causal relationships between complex ecosystems in response to climate change and enables the analysis of indirect effects among variables. The causality reflected in the BN is as good as the expert knowledge of the causal links. Compared to BN, the feature importance difference between "mean vapor pressure deficit and cumulative soil water index" and "maximum leaf area index and maximum vegetation height" reported by random forests is higher and can be overestimated. With the causality relation between correlated variables constructed, a BN-based sensitivity analysis can reduce the uncertainty in quantifying the importance of correlated variables. The understanding of the mechanism of indirect effects of climate variables on ecosystem functions through plant traits can be deepened by the chain casuality quantification in BNs.