Modeling Spatial and Temporal Dynamics of Montane Meadows and Biodiversity in the Greater Yellowstone Ecosystem

Abstract

We used remotely sensed imagery to identify and map montane meadow types (M-types) along a moisture gradient, and to develop spectrally-based spatially-explicit models for predicting species diversity patterns in two regions of the Greater Yellowstone Ecosystem: 1) Grand Teton National Park and Bridger Teton National Forest and 2) the Gallatin National Forest and Yellowstone National Park. We investigated the potential to predict species assemblages associated with these meadow types and we also tested our ability to classify specific subsets of montane meadow types such as wetland and sagebrush communities. We also compared our results to the predictability of communities when our sampling sites were classified by GAP analysis. We classified wetlands into 2 categories that differed by the percentage of obligate wetland plant species. Accuracy of wetland classification based upon remotely sensed data was 70%. We classified 4 sagebrush communities [low sagebrush (Artemesia arbuscula), big sagebrush (Artemesia tridentata ssp. vaseyana), mixed low sagebrush/big sagebrush, and bitterbrush (Purshia tridentata)/big sagebrush]. Overall accuracy of our sagebrush community classification based upon remotely sensed data was 65%, and was highest for the mixed big sagebrush/low sagebrush community at 86%. We also investigated the association of plant, bird and butterfly species with each of the meadow types. Because of the rarity of many of the species, our analyses focused on species for which we had minimum standards of data. These standards varied among taxonomic groups, but species only observed infrequently were not included in the analyses. The abundances of 6 of the 11 most abundant bird species were significantly correlated with meadow type as defined by satellite imagery. However, 10 out of 11 bird species showed a significant correlation when both remotely sensed data and landscape variables (e.g., shrub biomass, percent cover of willow or sagebrush, and meadow area) were added to the models. Butterfly species showed even stronger associations with particular meadow types, especially in the Teton region. We used regression tree analyses to separate meadow types by their associated species of butterflies. Fourteen of 67 butterfly species distribution patterns could be used to classify sampling sites into one of five different meadow types with 92% accuracy in 1997 and 96% accuracy in 1998. From the perspective of global climate change indicators, mesic meadows showed the greatest seasonal and interannual variability in spectral response and highest species diversity of plants. Given the rich biodiversity of mesic montane meadows and their sensitivity to variations in temperature and moisture, they may be important to monitor in the context of environmental change. Finally, we were able to do some additional related studies funded by a grant from the Nature Conservancy's David H. Smith Fellowship program (to Su and Debinski). We compared the scale of mapping of biotic communities in the EPA-funded project with the GAP analysis approaches in Montana and Wyoming. We found that both the 1 ha MMU M-Type map and 100 ha MMU Wyoming GAP map were significantly associated with bird, butterfly and plant community similarity. However, the 2 ha Montana GAP map was not associated with bird and plant community similarity and only discriminated differences in butterfly species composition between one map class and the others. This difference is probably explained by the fact that Montana GAP had a coarser categorical resolution of meadow types which was not sensitive to community variation associated with the hydrology of the meadows.