2008 BES Annual Meeting Presentation and Poster Abstracts
Agreement Assessment of Visual Interpretation and Digital Classification for Mapping Urban Landscape Heterogeneity
Co-Authors: Kirsten Schwarz, Mary Cadenasso
Abstract: Visual interpretation has long been used for patch mapping and patch composition (i.e., within-patch percent cover) estimation. However, relatively few studies have investigated human variation in percent cover estimation, particularly in urban settings. This paper presents the methods and results from the comparison of within-patch percent cover provided by visual interpretation with that from digital classification. The study site is the Gwynns Fall watershed, Maryland. Specifically, we aimed to answer the following three questions: 1) What is the relative agreement of percent cover estimation between the two methods? 2) What are the spatial patterns of the patches with high degree of disagreement? and 3) How does patch heterogeneity affect the degree of disagreement? The comparison was conducted separately for each of the five HERCULES landscape elements -- building, coarse textured vegetation, fine textured vegetation, pavement, and bare soil. Using a modified agreement assessment, we found strong agreement between the estimates from the two methods for all of the five landscape elements. Except fine textured vegetation, patches with high degree of disagreement tended to cluster spatially. Patch heterogeneity measured by indices of patch composition and shape complexity affected the degree of disagreement, but the effects varied by landscape elements.