Forward change detection 2000 - 2004: urban sprawl and imperviousness in Lexington, KY

Abstract

Over time, changes in land cover are likely to alter the distribution of imperviousness in landscapes, bringing the possibility of negative effects on watersheds due to degradation of water quantity and quality parameters (Schueler, 1994). Imperviousness is generally correlated with and utilized as a strong indicator of urban and suburban growth, serving as a proxy for the “developed” Anderson Level II classes (low, medium, high, open). This classification scheme serves as the working paradigm upon which the National Landcover Dataset 2001 (NLCD01) and the Kentucky Landcover Dataset 2001 (KLCD01) were built (http://kls.ky.gov; http://landcover.usgs.gov/index.asp). The current phase of the Kentucky Landscape Snapshot project necessitates the development of a change detection component. At this stage, work is revolving around utilizing medium resolution imagery, from contemporary (2004; 20 m SPOT®) - and past (late 90’s - early 00’s; 30 m Landsat 7) – epochs for that purpose (Figures 1 and 2). The latter imagery is the basis of the NLCD01 imperviousness and canopy closure datasets (http://kls.ky.gov) (Yang et al., 2002). The goal of this paper is to present preliminary results of on-going, forward-change detection efforts, quantifying temporal change patterns in imperviousness in the vicinity of Lexington, KY. A subset of imagery where change in imperviousness was assessed to be minimal between ca. 2000 and 2004 was used as training for the classification and regression tree procedure (CART tools for ERDAS® Imagine® 8.7 and RuleQuest®’s Cubist®) applied to the 20 m, 4 band SPOT® imagery, resampled to 30 m. The result was a 2004, 30 m imperviousness dataset, which was then subtracted from and made relative to the earlier data (Figure 3). This change “mask” is used next to locate areas of potentially sizable imperviousness change. References.