A climate sensitive mixed-effects diameter class mortality model for Prince Rupprecht larch (Larix gmelinii var. principis-rupprechtii) in northern China

Forest mortality is an important variable commonly included as one of the predictors in the growth and yield models that are the fundamental decision-making tools in forest management. The existing forest mortality models for Larch (Larix gmelinii var. principis-rupprechtii) forest, which plays key roles in maintaining forest ecosystem functions and reducing atmospheric carbon concentration through the sequestration, are based on the traditional modeling methods, and these models do not account for the effects of climate on the forest mortality. Developing climate sensitive mortality models are useful for formulating effective forest management strategies in the context of climate change. We developed the diameter class mortality models using the data of 102 temporary sample plots distributed across the Guandi Mountain National Forest Park and Wutai Mountain Boqiang State-owned Forest Farm in the Shanxi Province of northern China. Four commonly used mortality functions were used to fit the data. Among many climatic and dendrometric variables evaluated, number of trees per diameter class (NC), mean annual precipitation (MAP); median diameter class (MDC), and mean temperature difference (DT) contributed significantly highly to the mortality variations. Then, these variables were selected as predictors to develop the two-level nonlinear mixed-effects mortality models applicable to the diameter class levels. The random effects at the levels of both the sample plots and stands with different site quality (blocks) were taken into account to build the models. Compared to other three models derived from the Poisson; Zero-inflated Poisson; Hurdle Poisson functions, the two-level Logistic mixed-effects mortality model better explained the effects of climate variables on the forest mortality of Larch. Tree mortality increased with increasing NC and MAP, however, MDC and DT had the opposite effects with diameter class increasing. Modeling the random effects at the block-level alone led to significantly high correlations among the residuals, and correlations were significantly reduced when the random effects were modeled at both the block- and sample plot-levels. Excessive rain during the growing season decreased the rate of tree mortality, especially for diameter classes with high number of trees. Findings from this study can be combined with the knowledge of the adaptive management to reduce the risks and uncertainties associated with forest management decision.