Mapping vegetation communities in Ozark National Scenic Riverways: final technical report to the National Park Service

Abstract

Vegetation communities were mapped at two levels in Ozark National Scenic Riverways (ONSR) using a hybrid combination of statistical methods and photointerpretation. The primary map includes 49 cover classes, including 24 classes that relate to vegetation associations currently described by the United States National Vegetation Classification Standard (USNVC; The Nature Conservancy, 1994a). The remaining types include cultural features, ruderal communities on abandoned agricultural lands, and non-vegetated classes. Overall map classification accuracy is 63 percent. The secondary mapping level aggregates communities with similar appearance and ecologically related associations into Community Types. The resultant 33-class Community Type map has an overall classification accuracy of 77 percent and identifies groups of communities based on resource management goals within the park. Important additional products include 1) a general probability map for all vegetation associations, which can be used to assess final classification certainty, and 2) individual probability maps for each association, which can be used to identify areas that have a high likelihood of supporting a given type, beyond where that type was identified in the final map products. Other secondary map products include data layers derived from primary color-infrared imagery, secondary imagery data and digital elevation models. A field key and photo guide to associations and complete community descriptions were produced, along with a photo guide of fuel conditions. Wildland fuels data were used to generate a fuels map based upon Anderson's fuels models (1982). Executive Summary v * Full 49-class USNVC vegetation association map is available in Appendix 12 Executive Summary ix *Full 33-class Community Type map is available in Appendix 13.

Mapping Vegetation Communities Using Statistical Data Fusion in the Ozark National Scenic Riverways, Missouri, USA

“…Comparison of the results of the limited pilot classifications with the final classification of the entire ONSR mapping region indicates that a hybrid combination of statistical methods and photointerpretation is needed to obtain overall and class-wise accuracy levels adequate for USGS-NPS Vegetation Mapping Program standards. While the discriminant analysis approach presented here was applied to map 84 percent of the ONSR mapping area (Chastain et al, 2006), complementary methods proved valuable to accurately map additional land-cover classes. The search for repeated patterns that drives a statistical classifier tends to be undermined by the anomalous land-use patterns that produce highly altered vegetation communities and humandominated land-cover types.…”

“…During the second season, a preliminary classification map was used to stratify sampling and identify sites with high provisional vegetation community diversity. Data collection at each point followed NVCS guidelines (The Nature Conservancy, 1994) and are described in the final technical report to the National Park Services (Chastain et al, 2006). Geospatial training information collected at each sample location included identification of the NVCS association type within which the sample point was located, a ranking of its quality from 1 to 5 to describe the degree to which that community matched the description in the current NVCS, and an estimation of the extent in meters of the NVCS type in each cardinal direction from the sampling point.…”

“…where M and B are regression parameters (reported in Chastain et al, 2006), R o is the original reflectance, and R(hat) is the mean reflectance (Meyer et al, 1993).…”

“…Twenty-four map classes relate to NVCS vegetation associations, while the 25 remaining classes include cultural and non-vegetated features, and ruderal communities on abandoned agricultural lands. The discriminant analysis statistical method described here (Figure 4) used a large number of the overall set of input variables to map NVCS vegetation communities, which cover 84 percent of the total spatial extent of the ONSR mapping area (Chastain et al, 2006). Photointerpretation methods were applied to map the cultural and non-vegetated landcover classes.…”