Analysis of urban-rural gradients using satellite data

Abstract

Growing urban populations has sparked concerns over the urbanization impacts to ecosystems. Gradient analysis has been presented as a method of assessing human impacts on the environment, but few studies have used satellite data to provide information for these analyses. In this dissertation techniques to interpret satellite data were used to investigate urban-rural gradients. The combination of a spectral mixing model and V-I-S (vegetation, impervious, soils) model provided information about the land cover in and around several cities. Based on accuracy assessments using ground data, the model output was determined to be an acceptable predictor of land cover values. An analysis of the Seattle Washington urban-rural region revealed a land cover gradient does indeed exist along a distance gradient from the city center. Traditional land use classes, human population and distance to city center were poor predictors of land cover for a specific location, but acceptable for determining large scale trends. Techniques using V-I-S data to provide landscape level analysis were compared with traditional thematic data. In Seattle Washington no relationship was observed between forest patch size or patch shape along the urban rural gradient. Traditional landscape ecology analysis for urban-rural areas is problematic. A data smoothing technique is presented that allows a landscape context to be determined from spatially continuous raster data. Using this approach, small urban forest patches can be distinguished from small rural forest patches. Lastly, the land cover along the urban-rural gradient of eight cities was compared. Within these eight cities the urban land cover was relatively similar even though the rural land cover was dissimilar. Rural vegetation cover was correlated with precipitation and changes in percent vegetation along the gradient were greater for wet environments than dry. The population of a city is correlated with both the extent of the built environment and the rate of change in impervious cover along the gradient. The techniques presented in this dissertation can be used to compare the land cover along the urban rural gradient of different cities, changes over time, and provide information useful in understanding changes in ecosystem processes from urbanization.