USING AN ECOLOGICAL LAND HIERARCHY TO PREDICT SEASONAL-WETLAND ABUNDANCE IN UPLAND FORESTS

Abstract

Hierarchy theory, when applied to landscapes, predicts that broader-scale ecosystems constrain the development of finer-scale, nested ecosystems. This prediction finds application in hierarchical land classifications. Such classifications typically apply to physiognomically similar ecosystems, or ecological land units, e.g., a set of multi-scale forest ecosystems. We speculated that hierarchical constraint also controls the development of small, nested ecosystems that are structurally distinct from the constraining matrix. We tested this hypothesis using seasonal wetlands in upland forest. Specifically, we related seasonal-wetland abundance in upland forest stands to multi-scale terrestrial ecological units, as defined by hierarchical combinations of regional physiography, glacial landform, soils, and forest cover. Moreover, we determined the spatial scale of terrestrial ecological unit that is the best predictor of seasonal-wetland abundance. Our study area is mapped into a set of nested terrestrial ecological units, including two subsections (differing in physiography), four land-type associations (glacial landforms), and 11 land types (forest vegetation, soil). We used a geographic information system to determine seasonal wetland densities in 16-ha plots located within the nested terrestrial ecological units. Cumulative plot frequency distributions of wetland density did not differ between subsections; 50% of plots contained no wetlands, 38% contained 1–3 wetlands, and 12% contained ≥4 wetlands. Frequency distributions differed among land-type associations (LTA). Ninety percent of plots on a glacial lake plain contained no wetlands, compared to 63%, 42%, and 38% for outwash, end moraine, and ground moraine LTAs, respectively. Ten percent of plots on the lake plain contained 1–3 wetlands, compared to 32%, 48%, and 43% for the outwash, end moraine, and ground moraine, respectively. The remaining plots on the latter three LTAs contained >3 wetlands. Frequency distributions rarely differed among land types. Compared to occurrence, identity and scale of the ecological unit were poorer predictors of actual wetland densities. Regression tree analysis, while significant, explained only 11.6% of variation in wetland density among plots. Still, the leaves of the regression tree differed in densities primarily based on LTA. Our results demonstrate that identity of constraining upland forest ecosystem explains significant amounts of variation in seasonal wetland abundance. We identify glacial landform as the scale of ecological unit having the greatest control over seasonal wetland abundance. We focus on seasonal wetlands in forests, yet our approach should apply to other small, nested ecosystems. This approach may facilitate conservation management of small, nested ecosystems by providing likelihood estimates of occurrence within mapped terrestrial ecological units.