Article ID | Journal | Published Year | Pages | File Type |
---|---|---|---|---|
4379056 | Ecological Modelling | 2006 | 11 Pages |
Quaking aspen (Populus tremuloides Michx.) is the most widely distributed tree species in North America, but its presence is declining across much of the Western United States. Aspen decline is complex, but results largely from two factors widely divergent in temporal scale: (1) Holocene climatic drying of the region has led to water limitation of aspen seedling recruitment, and (2) anthropogenic fire suppression during the 20th century has allowed shading of aspen clones by fire-intolerant conifers. These processes interact variously and often diffusely, but traditional, binary habitat mapping approaches can only resolve their net effect after complete loss of aspen patches. To provide information for preemptive land management in the Book Cliffs – a biogeographic link between the Utah and Colorado Rocky Mountains and a location experiencing aspen decline typical of the region – we developed a regression-based generalization of niche/habitat analyses that is usable in GIS, is capable of detecting anomalies in cover before complete patch conversion, and can be incorporated smoothly into the decision-making process. We estimated the realized Hutchinsonian niche of quaking aspen to potential evapotranspiration (PET) by nonparametric, likelihood-based regression techniques, projected (continuous) values of aspen's niche expectation and uncertainty geographically, and correlated differences between observed and expected aspen cover to remotely sensed conifer cover. Results confirm the strong constraint of site water demand on aspen cover and suggest that conifer cover decreases aspen cover beneath its expectation given the PET environment. Compared to sites without quaking aspen, our aspen sites have lower PET in every month of the growing season, but the difference increases over the growing season as drought effects become more extreme. Superimposed on this broader environmental constraint, conifer cover displaces aspen cover and shows a positive correlation with model deviance (r = 0.344). Ultimately, the thematic information conserved by our approach allowed us to resolve detailed rasters of management potential and map a modest potential increase of aspen cover – 14.63 ha (0.14%) of the study area, or +2.46% of current aspen cover – within one management cycle.