Spatial approaches to understanding current and future landscape development patterns and environmental impacts in a northern Minnesota lake district

Increasing human development in amenity rich areas affects the same natural systems that draw people to those areas. Quantifying trends of development spatially allows major drivers of landscape change to be assessed, which can then be used to create a series of projected scenarios to help visualize potential future landscape changes across a variety of spatiotemporal scales. The ability to accurately track and predict landscape changes can substantially increase our fundamental understanding of how to balance ecological health with increasing development.;The research reported in this work represents an integrated approach for creating, examining and depicting these landscape changes in a northern Minnesota landscape experiencing extensive lakeshore development. My primary goal was to quantitatively identify the driving factors of past development patterns and use these to predict and examine changes likely to occur in future landscapes.;To classify current land cover, I utilized a variety of methods. Random Forests (RF) classification methods outperformed supervised maximum likelihood for classifying land cover in the current landscape, and the resulting classification was then used to determine spatially explicit estimates of Effective Impervious Area (EIA) through modeling techniques using variables derived from Light Detection And Ranging (LiDAR), multi-spectral passive remote sensing, and ancillary data. While the amount of EIA at the study area level does not exceed thresholds for water quality degradation, these thresholds were exceeded in several minor watersheds.;This study also identified factors that have driven development trends in the past, and determined that these factors are global, meaning that they hold true across the entire study area. By understanding these factors, potential directions of future development can be modeled within the study area. A RF regression model outperformed both generalized linear models and geographically weighted regression approaches, and when the resulting Development Probability surface was compared to actual development five years later, the RF model was 89.4% accurate.;The RF modeling approach was then used to guide the development of nine additional potential future landscape scenarios, based on current planning and zoning regulations. Changes in landscape patch composition and configuration were quantified using landscape metrics. Analysis of the landscape metrics in these future scenarios clearly indicates substantial levels of change. There was a steady increase in impervious surfaces throughout the future landscapes, particularly in minor watersheds with private ownership as compared to those with primarily public ownership. The increase in impervious surface comes at the expense of forests and grasslands. In addition to overall loss of forest and grassland, both classes experience a corresponding increase in the amount of edges, and an overall decrease in the core area index. This indicates a fundamental shift in the nature of the remaining forest and grassland habitats with a lower total amount of core habitat and a higher total amount of edge habitat, and increased total amount of impervious surfaces throughout the landscape.