Predictability and Change Over Fifty Years in Southern Wisconsin Forests

Document Type


Publication Date


Find this in a Library

Catalog Record


To assess predictability in ecosystem change we use a fifty- five year dataset of 121 forested sites across southern Wisconsin, first sampled in the early 1950s by J.T. Curtis. We first look at pattern of species distribution and abundance in the context of environmental and physical distances in both time periods. We then measure rates of change among the variables between the two time periods and use a combination of landscape, environmental and species attribute data to ask whether site, landscape and functional traits can be used to predict species loss and homogenization. We hypothesize that the correlation between the physical environment and community composition will have changed over time, specifically that the environmental conditions of a site have less predictive power now than it did during the 1950s. Results indicate a decrease in the predictive power of environmental variables over time, with soil conditions explaining over 30% of the variance on the primary axis of an NMDS ordination in 1950 while explaining only 21% for the same sites in 2003. Mantel tests of correlation between tree and herb similarities showed a marked reduction in the strength of the relationship (from correlation of 0.445 in the 1950s to 0.299 today), again suggesting loss of predictability over time. In terms of predicting change, landscape factors such as fragmentation and urbanization are significant predictors of species loss and homogenization, while growth form and specific leaf area are good functional trait predictors of the types of species being lost or gained. In general, however, the predictability of rates of species loss and homogenization are relatively low (about 35% in the best models). While we can make very general predictions about average changes that are taking place, the changes at any individual sites are difficult to predict, reflecting the importance of individual stand histories and landscape context. This points to the need for managers and researchers that have intimate knowledge of regional ecological diversity and social systems if we hope to understand and maintain diversity at landscape scales.


Presented at the 90th Ecological Society of America Annual Meeting, Montréal, Quebec, Canada.

Organized Oral Session 51: Predictability of Vegetation Dynamics: From Quadrats to Landscapes, From Years to Decades.

Catalog Record