Towards Modelling Future Trends of Quebec’s Boreal Birds’ Species Distribution under Climate Change

Abstract

Adaptation to climate change requires prediction of its impacts, especially on ecosystems. In this work we simulated the change in bird species richness in the boreal forest of Quebec, Canada, under climate change scenarios. To do so, we first analyzed which geographical and bioclimatic variables were the strongest predictors for the spatial distribution of the current resident bird species. Based on canonical redundancy analysis and analysis of variance, we found that annual temperature range, average temperature of the cold season, seasonality of precipitation, precipitation in the wettest season, elevation, and local percentage of wet area had the strongest influence on the species’ distributions. We used these variables with Random Forests, Multivariate Adaptive Regression Splines and Maximum Entropy models to explain spatial variations in species abundance. Future species distributions were calculated by replacing present climatic variables with projections under different climate change pathways. Subsequently, maps of species richness change were produced. The results showed a northward expansion of areas of highest species richness towards the center of the province. Species are also likely to appear near James Bay and Ungava Bay, where rapid climate change is expected.

An Assessment of the Climate Change Impacts on the Distribution of the Glacial Relict Woodpecker Three-Toed Woodpecker Picoides tridactylus

“…Because bioclimatic variables are derived from the same temperature and precipitation data, they are highly correlated [83], which can lead to incorrect conclusions about their importance [84], reduce the predictive ability of a model, and increase the uncertainty of the results even with a low degree of collinearity [85]. Although certain modelling methods, such as Maxent, are less sensitive to the correlation between variables [85,86], to avoid possible issues, environmental data were first extracted from all 105 occurrence points for each bioclimatic layer separately using the Sample Raster Values function in QGIS.…”

Computational biogeographic distribution of the fall armyworm (Spodoptera frugiperda J.E. Smith) moth in eastern Africa

“…To include less correlated variables in species distribution models (SDM), an elimination procedure of some of the correlated variables should be conducted to reduce the chances of model overfitting and variable inflation [ 42 ]. Commonly, bioclimatic variables such as the ones obtained from the WorldClim [ 43 ] are highly correlated because they are all derived from similar data [ 44 ]. These bioclimatic variables are widely used in habitat suitability models like the one employed in this study.…”