PARENT SESSION
Poster Session #38: Landscape Ecology.
Thursday, August 9, 2001. Presentation from 10:30 AM to 12:00 PM. Exhibition Hall

Assessing the scale of usefulness of regional-scale vegetation maps with ground-based data.

Bouldin, Jim ¹, Parks, Sean¹, ¹

ABSTRACT- The analysis of vegetation attributes over very large areas has historically been a problem. This is especially manifest in areas characterized by vegetational variability at all spatial and temporal scales, such as mountainous areas. In California's Sierra Nevada, vegetation type, physiognomy and structure are mapped by the US Forest Service using Landsat TM source data and classification algorithms and models that relate expected vegetation type to physiographic predictors. The accuracy of the resulting maps can only be assessed using a sample of ground based data. These data can also be used to assess other maps, such as maps of old-growth forest. We assessed the quality of both standard USFS vegetation maps and a much coarser old-growth map. For each map we tested its ability to discriminate particular attributes, and did so at several spatial scales. We developed a GIS-based sampling program that randomly chose pairs of landscape points located a specified maximum distance (8km) from each other, noted the polygon labels for each point, and then randomly chose ground plots sampled in polygons having the chosen labels to "represent" the vegetation structure at those points. This procedure created a statistical estimate of the expected vegetational variation as a function of separation distance, national forest and topographic position. Results showed that both maps vary widely in their discriminatory ability, but that finer-scale maps are more useful in discriminating a wider variety of attributes at a wider range of spatial scales. This is to be expected in a complex mountainous region.

KEY WORDS: vegetation, mapping, accuracy, scale