The density and identity of tree neighbourhood is a key factor to explain tree mortality in forests, especially during the stem exclusion phase. To understand this process, we built a logistic model for mortality in a spatially explicit context, including tree and neighbourhood predictors. Additionally, we used this model to build mortality risk frequency distributions. Finally, we tested this model against a random mortality model to predict the spatial pattern of the forest. Annual mortality rate was high for pedunculate oak (Quercus robur, 6.99%), moderate for birch (Betula celtiberica, 2.19%) and Pyrenean oak (Q. pyrenaica, 1.58%) and low for beech (Fagus sylvatica, 0.26%). Mortality risk models for pedunculate oak and birch included stem diameter, tree height, canopy position and neighbourhood. Mortality was affected by the specific nature of the neighbourhood showing a clear competitive hierarchy: beech > pedunculate oak > birch. Models based on random mortality and logistic regression model were able to predict the spatial pattern of survivors although logistic regression predictions were more accurate. Our study highlights how simple models such as the random mortality one may obscure much more complex spatial interactions.
Mendeley saves you time finding and organizing research
Choose a citation style from the tabs below