MAPPING WETLAND PLANT COMMUNITIES USING MULTISPECTRAL IMAGERY: PRELIMINARY RESULTS FROM THE UINTA MOUNTAINS

A significant variety of groundwater-dependent ecosystems make up the montane wetlands of the Uinta Mounatins, northeastern Utah. Unlike their counterparts in other ranges of the Rocky Mountains, these systems are relatively unstudied. The drainage basin of Reader Creek, located near the crest of the range on the south slope, was selected for detailed study because of its variety of wetland plant communities, minimal human impact, and homogenous bedrock geology. One objective of this multidisciplinary study is to classify the wetland types and constituent plant communities within the drainage basin. In addition to traditional field-based botanical and geochemical studies, multispectral image-classification techniques on 2006 NAIP imagery are being used to map the distribution of specific wetland types within the basin.

The wetlands within the Reader Creek drainage basin are classic patterned fens; wetland hydrology is dominated by groundwater discharge, average pH is circum-neutral, peat is accumulating, and the surface of the wetlands is dominated by distinctive string-and-flark morphology. The valley floor of the drainage basin is comprised of three stair-stepping wetlands bounded by glacial till at their lower end. These wetlands contain six (6) distinctive wetland-obligate plant communities dominated by members of the Cyperaceae Family. These communities are composed of one or more of the following species: Carex aquatilis, Carex limosa, Carex saxatilis, Eriophorum polystachion, and Eleocharis quinqueflora.

The multispectral images were processed using a supervised classification. Twenty-two locations were used as training sites, which determined 16 spectral classes based on vegetative cover. To the untrained eye the wetland-obligate plants are very similar in morphology. However, our preliminary classification shows that most of the communities are spectrally unique. On the other hand, there are some vegetation classes with obvious differences in plant morphology that have very similar spectral signatures in the multispectral bands (e.g. the overlap on non-wetland conifer areas and C. aquatilis/E. polystachion fens). Future work will attempt to identify criteria to isolate these overlapping land-cover classes. Ultimately, this research may lead to an efficient method to classify and inventory wetlands within the mountains of Utah.