This research examines the feasibility and accuracy of using digital imagery and image-object analysis with object-oriented software for large-area landscape quantification and classification, resulting in faster, more accurate, and less expensive automated land use and sinkhole inventory. The study is undertaken in a highly-urbanizing karst subwatershed of the Potomac River, the Opequon Creek Watershed in Berkeley and Jefferson Counties, West Virginia. Five LandSat scenes between 1984-2003 and numerous high-resolution aerial photographs from 1987-2007 are merged with Definiens Professional Software for supervised classification and spatial analysis using object-based algorithms, rather than traditional spectral analysis.
Temporal assessment of land use and land cover change shows significant conversion of forested (-39%) and agricultural lands (-35%) to urban (+177%) and impervious surface (162%) over the twenty-five year period. Meanwhile, sinkholes numbers increased 16% from 802 to 932. Comparison of traditional to object-oriented method shows high accuracy of image-object oriented analysis. High producers (91%), users (94%) and overall (92%) accuracies achieved by the Definiens Analysis verify and validate this method’s ability to catalogue and analyze land use and land cover. Spatial analysis of new sinkhole development using various statistical tools shows dense clustering northwest of Martinsburg over the last 25 years, while dispersion takes place with increasing distance from town. Regression analysis shows higher R² values associated with urban class of land use, suggesting a positive correlation between development in the watershed and the frequency and occurrences of sinkholes. Lastly, land use planning watershed management principles are suggested for Opequon Creek Watershed to evaluate their ability to mitigate changing land use and land cover, resulting from development pressures.