Tree spatial pattern and mortality prediction in burned patches of Dahurian larch (Larix gmelinii Rupr.) forest that experienced a mixed-severity wildfire

Background. Fire-caused tree mortality and spatial pattern are crucial for evaluating forest dynamics and developing management prescriptions. Aims. We investigated direct fire effects on spatial distribution and mortality of Dahurian larch (Larix gmelinii Rupr.) and assessed the Ryan and Amman (R-A) model performance and the importance of immediate mortality predictors. Methods. We analysed spatial patterns of fire-killed and surviving trees of three size classes in plots that burned at low- to high-severity using pair-correlation functions and tree mortality with the R-A model and generalised linear mixed models. Key results. The mixed-severity fire caused strong density-dependent mortality and more aggregated surviving tree patterns at short distances. The R-A model generally performed acceptably, and crown scorch and bole char height were critical predictors determining post-fire tree mortality. Conclusions. Fire-caused tree mortality and spatial patterns are controlled primarily by spatial variation in tree size and biological and structural characteristics. The prediction biases of the R-A model arose primarily from the intrinsic traits of Dahurian larch and the imbalanced dataset. Fine-scale neighbourhood density might be a fundamental priority for fire management and restoration. Implications. This study could possibly improve mechanistic understanding of spatial pattern development and tree mortality in similar fire-prone conifer forests.