Globally, alpine treeline positions are defined by temperature which has led to the assumption that treelines would rise in response to climate change. But recruitment above treeline is also controlled by local drivers which modulate the effect of climate. This study aims to measure the relative effect of abiotic (temperature, snow, wind) and biotic interactions to predict seedling survival in future conditions. A natural experiment was conducted in the Canadian Appalachian Mountains where white spruce is the dominant treeline species. We transplanted seedlings and used the directional effect of adult trees on microclimate (cast shadow, wind shelter, snow redistribution) to create a set of different abiotic conditions. At each “microclimate”, we removed neighboring vegetation on one out of two seedlings to assess the biotic interactions effect. A first cohort was transplanted in fall 2010 and survival was assessed in fall 2011 (calibration dataset). A second cohort was transplanted in fall 2011 and monitored in fall 2012 (validation dataset). We used a Bayesian Generalized Linear Mixed Model to estimate the effects of explanatory variables and simultaneously predict survival of the second cohort. We analysed the confusion matrix based on predicted versus observed survival to assess the robustness of our predictions.
Results/Conclusions
A year after transplantation, the first cohort had a higher survival ratio (77.5%) than the second cohort (47.5%). The model presented a reasonable fit with the calibration dataset (Bayesian p-value =0.75). Results showed that even though the number of growing degree days did increase probability of seedling survival, the direct effect of temperature was not important. Snow depth had a strong positive effect on survival but presented an optimum beyond which probability of survival decreased, possibly due to snow fungus and waterlogged conditions. Neighboring vegetation effect was positive in 94% of the simulations, illustrating positive interactions through neighbor sheltering frequently found in harsh environments. The analyses of the confusion matrix revealed that our model is accurate (rate of correct predictions 0.51) although with a high fraction of false positives (0.83), meaning that our model predicts survival efficiently but is underestimating mortality. Our study shows that in the Canadian Appalachians Mountains, seedling survival at treeline is mainly controlled by snow conditions and not by temperature per se, illustrating the importance of including precipitation when making predictions of treeline ascent in light of climate change scenarios.