Validating, analyzing, and predicting lawn maps: Application of GIScience and spatial analysis in the northern Boston suburbs

The residential lawn is one of the most recognizable land-cover features of the suburban landscape in the United States, accounting for more planted acres than major irrigated crops such as barley, cotton, rice, or corn. The lawn landscape in the United States is closely associated with suburbanization and urban sprawl, and its maintenance and management represents one of the most prominent anthropogenic environmental challenges in US urban and suburban areas today. As a first step in addressing these environmental challenges and understanding the social processes responsible for the creation and management of this landscape feature, it is necessary to produce a geospatial database of fine-resolution lawn land-cover maps. While there have been a few attempts at mapping lawns, understanding the social processes that create and maintain them. and producing national-scale estimates of lawns, there are a number of analytical and methodological challenges that need to be addressed, particularly related to the time, cost, and resource intensive nature of mapping this fine-scaled and spatially heterogeneous land-cover type. Furthermore, there is a lack of consensus among experts about the best method for assessing the accuracy of fine-resolution land-cover maps, which is vitally important when using the land-cover maps in other research applications. This dissertation attempts to address these challenges by fulfilling three main research goals. The first goal is to produce and evaluate a method for validating the accuracy of fine-spatial resolution object-based image analysis (OBIA)-derived land-cover maps of a spatially heterogeneous suburban landscape. The second goal is to gain an understanding of the social processes that explain the patterns of land cover in United States suburbs, particularly residential lawns. The third and final goal is to apply our understanding of these social processes and their relationships with lawns to spatially predict maps of the lawn landscape as an alternative method to expensive, time-consuming, and resource intensive mapping efforts.