This dissertation addresses one of the fundamental issues in landscape architecture, environmental planning as well as the general field of environment and behavior: the relationship between vegetation settings and neighborhood safety. Being built upon a critical debate on whether vegetation is actually helpful or harmful to crime prevention, this piece of research seeks to answer how different types of vegetation affect various categories of crime in neighborhoods with varied socio-economic and environmental characteristics. It has a firm grounding in a group of interdisciplinary theories, including environmental psychology, crime prevention through environmental design (CPTED), and neighborhood criminology. Through synthesizing these theories, an innovative theoretical framework that integrates vegetation factors with traditional crime factors identified by major schools of thought of neighborhood criminology is constructed to comprehensively explain the incidence of neighborhood crime. More specifically, the study seeks to answer how the existence of various types of vegetation settings is related the occurrence of neighborhood crime after removing the effects of major crime factors as identified by neighborhood criminology literature. To achieve these objectives, a data rich environment is built for the investigation of the issue through collecting a rich set of geospatial and attribute data on all relevant factors, including vegetation settings and traditional crime factors, for the study area, i.e. Oakland, California.
With the huge raw dataset at hand, the primary task of research is to transform them into a uniform format through remote sensing image processing and GIS modeling. The data transformation process is therefore in two phases. The first one was to extract detailed, individual-tree level vegetation cover and vegetation height from very high resolution remote sensing images. A comprehensive approach, which consists of image registration, segmentation, object-based image classification, ground truth sampling, accuracy evaluation, and geospatial modeling, is developed for this purpose. A shadow modeling method comprised of geometric modeling, spatial decomposition, geo-processing and other GIS methods is then applied to extract vegetation height from detail image objects of vegetation and shadow. Next, the study further models the view-blocking effects of vegetation settings through a rule-based hierarchical evaluation system. Lastly the accuracy of the approaches is checked with in situ data collected with GPS instruments. Accuracy assessment indicates that the approaches are successful in measuring detailed vegetation distribution and their view-blocking effects.
The second phase of data processing focuses on spatial data integration and modeling. In this phase we transform data from multiple sources and in multiple formats to data in uniform format and consistent quality with state of the art geospatial methods and automated GIS procedures. GIS methods applied include geo-coding, spatial sampling, areal interpolation, network routing, geospatial statistics, etc. As outcome, a new geo-database containing uniform data for over two hundred variables is generated for each of the sampled neighborhoods in Oakland.
To analyze the entire dataset, the research applies multiple regressions and multivariate analyses to answer the proposed research questions, i.e. whether there is a significant relationship between vegetation settings and neighborhood crime, after having controlled the effects of other crime factors. Results of this research prove that the integration of geospatial modeling and advanced statistical analyses is of central importance to ensure the explanation of up to 97% of Part I neighborhood crime. In light of their effects on crime, vegetation settings are found to have a positive relationship with property crime, while a negative relationship with most violent crime. More high view-blocking vegetation settings, especially in public space, are significantly correlated with more violent crime but less property crime. In summary, the relationships between vegetation settings and crime are more complex than that were reported in the literature. How vegetation settings affect neighborhood safety is not only a matter of demographic and socioeconomic status of people, planning and management of urban space, and the construction of landscape settings, but also determined by the mechanisms, through which different types of crime occur in varied social and physical context. Therefore, for various categories of crime, the impacts of landscape settings could be totally different. (Abstract shortened by UMI.)