Linking remote sensing and various site factors for predicting the spatial distribution of eastern hemlock occurrence and relative basal area in Maine, USA

Abstract

Introduced invasive pests are perhaps the most important and persistent catalyst for changes in forest composition. Infestation and outbreak of the hemlock woolly adelgid (Adelges tsugae; HWA) along the eastern coast of the USA, has led to widespread loss of hemlock (Tsuga canadensis (L.) Carr.), and a shift in tree species composition toward hardwood stands.Developing an understanding of the geographic distribution of individual species can inform conservation practices that seek to maintain functional capabilities of ecosystems. Modeling is necessary for understanding changes in forest composition, and subsequent changes in biodiversity, and one that can be implemented at the species level. By integrating the use of remote sensing, modeling, and Geographic Information Systems (GIS) coupled with expert knowledge in forest ecology and disturbance, we can advance the methodologies currently available in the literature on predictive modeling.This paper describes an approach to modeling the spatial distribution of the less common but foundational tree species eastern hemlock throughout the state of Maine (∼84,000km2) at a high resolution. There are currently no published accuracy assessments on predictive models for high resolution continuous distribution of eastern hemlock relative basal area that span the geographic extent covered by our model, which is at the northern limit of the species’ range. A two stage mapping approach was used where presence/absence was predicted with an overall accuracy of 85% and the continuous distribution (percent basal area) was predicted with an accuracy of 84%. Overall, these findings are quite good despite high variability in the training dataset and the general minor component that eastern hemlock represents in the primary forest types in Maine.Eastern hemlock occurs along the southern half of the state stretching the east-west span with little to no occurrence in the northern regions. Several environmental and site characteristics, particularly average yearly maximum and minimum temperatures, were found to be positively correlated with hemlock occurrence. Eastern hemlock dominated stands appeared predominantly in the southwest corner of the state where HWA monitoring efforts can be focused. Given the importance of climate variables in predicting eastern hemlock, forecasts of future range shifts should be possible using data generated from climate scenarios.