Species distribution predictions are a common way of understanding the effects of climate change on forest ecosystems, by predicting vulnerability of individual species we can make infrerences about broader scale change. Until recently, much of the focus of researchers and forest managers in the Pacific Northwest has been on economically valuable conifers (Hagar, 2007). However, it is becoming more common to consider other species in forest management strategies, as broadleaf trees such as A. macrophyllum have been shown to be important indicators of overall ecosystem function (Hagar, 2007). This project considers the distribution of A. macrophyllum in western North America, and aims to model how it will change throughout this century as a result of climate change. Currently, the range mapped by the US Forest Service and Department of Agriculture includes the Pacific coast west of the coastal and cascade mountain ranges, from BC’s north coast to California (Fryer, 2011). Previous studies have estimated that its range may increase by over 50% before 2100 (Hamdan & Schmidt, 2012), but the spatial characteristics of the increase have not been thoroughly demonstrated.
There are a variety of available methods for modelling species distribution, machine learning methods are widely used in ecology due to their versatility and ability to make predictions with a large set of input parameters. More common methods include MAXENT and regression-based models, but in recent years many studies have found that the random forest classification algorithm has a higher accuracy in predicting species distribution than most other modelling approaches (Cutler et al., 2007; Garzan, et al., 2006; Heikinnen, Marmion & Luoto, 2012; Mi et al., 2017). This method does not rely on assumptions about the distribution of data, is capable of modelling non-linear relationships, and is efficient in finding the best predictors out of a large set of parameters (Pawley, 2018).