Forest fragmentation mapping, change detection, and gas well site identification in remotely sensed imagery: the role of spatial resolution.

摘要

This research focuses on identifying optimal data and methods for detecting and quantifying land cover/land use change associated with natural gas drilling in Appalachia. Airborne and satellite remote sensing provides a tool for examining land cover changes. Although change detection has been used in many applications related to natural and anthropogenic land cover change, little research has investigated the spatial pattern of land cover change associated with the ongoing expansion of natural gas drilling, particularly within the Marcellus Shale formation found across much of the northern Appalachian Region. Mapping land cover/land use change facilitates research of other relevant topics that require an understanding of methodological and scaling issues, particularly concerning environmental impacts of gas well drilling. I found that object-based classification is not significantly more accurate than pixel-based classification and accuracy does not vary with spatial resolution. However, object-based classifications were qualitatively more suitable for identifying land cover change related to well clearings at finer (1 m) resolutions, whereas the pixel-based classifications had a higher percentage of correctly identified well clearings at the coarsest resolution (30 m). In addition to investigating classification accuracy, scaling relations of landscape metrics at fine resolutions were compared to previous research of Wu et al. (2002) and Wu (2004), and were found to be consistent with previous research. Landscape metrics were also used to test for statistically significant changes in the forested class between the 2004 and 2010 classifications. Landscape metrics derived from random quadrats sampled across the entire county produced results that differed from quadrats sampled from high and low well density areas. Changes in metrics for sample areas with low densities of new wells indicate forest area and mean patch size increased significantly. One metric—edge density (ED)—was found to be sensitive to clearings and other development (access roads) related to natural gas development in areas with higher densities of new well clearings. This research provides a foundation for future investigation into natural gas development and suggests significant changes to forested ecosystems must be observed on a finer scale than the county level to assess the ecological significance of natural gas development.