A Bayesian network model to disentangle the effects of stand and climate factors on tree mortality of Chinese fir plantations

Tree mortality is a complex process that not only be affected by the various factors, such as stand and climate factors, but also the various long-term effects of the factors to each other. In this study, based on the long-term spacing trials of Chinese fir in four regions of southern China, a Bayesian network was used to model tree mortality in response to stand and climate factors, as well as comparing this approach with logistic regression and random forest method. The results showed that the Bayesian network method had the highest accuracy in predicting tree mortality. In addition, the Bayesian network approach could find the dependency in the relationship between data and provide a theoretical framework for modeling uncertainty by using probabilistic calculus and underlying graph structure. Sensitivity analysis showed relative diameter was the most important factor, and temperature was the most important climate factor. Furthermore, climate factors not only directly affected tree mortality, but also indirectly affected tree mortality through affecting relative diameter, stand density and Gini coefficient. We also found that stand competition, structural heterogeneity and age affected tree mortality under climate change, and a moderate level of competition condition and stand structure heterogeneity weakened the negative impact of climate factors on tree mortality. Old trees were more sensitive to climate change than young trees, especially under extreme climate conditions. Besides, we found that tree mortality was negatively correlated with moderate annual precipitation, winter mean minimum temperature, and stand structure (Gini), and low age, but positively correlated with low relative diameter, high density and age. The results will provide adaptive options for effective forest management of Chinese fir plantations under the backdrop of global climate change in the future.