Landscape analysis of plant diversity

Abstract

Studies to identify gaps in the protection of habitat for speciesof concern have been inconclusive and hampered by single-scale orpoor multi-scale sampling methods, large minimum mapping units(MMU's of 2 ha to 100 ha), limited and subjectively selected fieldobservations, and poor mathematical and ecological models. Weovercome these obstacles with improved multi-scale samplingtechniques, smaller MMU's (< 0.02 ha), an unbiased sampling designbased on double sampling, improved mathematical models includingspecies-area curves corrected for habitat heterogeneity, andgeographic information system-based ecological models. We applythis landscape analysis approach to address resource issues inRocky Mountain National Park, Colorado. Specifically, we quantifythe effects of elk grazing on plant diversity, identify areas ofhigh or unique plant diversity needing increased protection, andevaluate the patterns of non-native plant species on thelandscape. Double sampling techniques use satellite imagery,aerial photography, and field data to stratify homogeneous andheterogeneous units and “keystone ecosystems” (ecosystems thatcontain or support a high number of species or have distinctivespecies compositions). We show how a multi-scale vegetationsampling design, species-area curves, analyses of within- andbetween-vegetation type species overlap, and geographic informationsystem (GIS) models can be used to quantify landscape-scalepatterns of vascular plant diversity in the Park. The new multi-scale vegetation plot techniques quickly differentiated plantspecies differences in paired study sites. Three plots in the OuzelBurn area (burned in 1978) contained 75 plant species, while only17 plant species were found in paired plots outside the burn.Riparian areas contained 109 plant species, compared to just 55species in paired plots in adjacent forests. However, plant speciesrichness patterns inside and outside elk exclosures were morecomplex. One elk exclosure contained more species than its adjacentopen range (52 species inside and 48 species outside). Two elkexclosures contained fewer species inside than outside (105 and 41species inside and 112 and 74 species outside, respectively).However, there was only 26% to 48% overlap (using Jaccard'sCoefficient) of plant species composition inside and outside theexclosures. One elk exclosure had 13% cover of non-indigenousspecies inside the exclosure compared to 4% outside, butnon-indigenous species cover varied by location. We compared plantdiversity patterns from vegetation maps made with 100 ha, 50 ha, 2ha, and 0.02 ha MMU's in the 754 ha Beaver Meadows study area usingfour 0.025 ha and twenty-one 0.1 ha multi-scale vegetation plots.Preliminary data suggested that the 2 ha MMU provided an accurateestimate of the number of plant species (–14%) for a study area,but the number of habitats (polygons) was reduced by 67%, andaspen, a unique and important habitat type, was missed entirely. Wedescribe a hypothesis-driven approach to the design andimplementation of geospatial databases for local resourcemonitoring and ecosystem management.