Landslide riskscapes in the Colorado Front Range: a quantitative geospatial approach for modeling human-environment interactions
Date
2021
Authors
Hicks, Heather Brainerd, author
Laituri, Melinda, advisor
Fassnacht, Steven, committee member
Grigg, Neil, committee member
Rathburn, Sara, committee member
Journal Title
Journal ISSN
Volume Title
Abstract
This research investigated the application of riskscapes to landslides in the context of geospatial inquiry. Riskscapes are framed as a landscape of risk to represent risk spatially. Geospatial models for landslide riskscapes were developed to improve our understanding of the spatial context for landslides and their risks as part of the system of human-environment interactions. Spatial analysis using Geographic Information Systems (GIS) leveraged modeling methods and the distributed properties of riskscapes to identify and preserve these spatial relationships. This dissertation is comprised of four separate manuscripts. These projects defined riskscapes in the context of landslides, applied geospatial analyses to create a novel riskscape model to introduce spatial autocorrelation methods to the riskscape framework, compared geostatistical analysis methods in these landslide riskscape assessments, and described limitations of spatial science identified in the riskscape development process. The first project addressed the current literature for riskscapes and introduced landslides as a measurable feature for riskscapes. Riskscapes are founded in social constructivist theory and landslide studies are frequently based on quantitative risk assessment practices. The uniqueness of a riskscape is the inclusion of human geography and environmental factors, which are not consistently incorporated in geologic or natural hazard studies. I proposed the addition of spatial theory constructs and methods to create spatially measurable products. I developed a conceptual framework for a landslide riskscape by describing the current riskscape applications as compared to existing landslide and GIS risk model processes. A spatial modeling formula to create a weighted sum landslide riskscape was presented as a modification to a natural hazard risk equation to incorporate the spatial dimension of risk factors. The second project created a novel method for three geospatial riskscapes as an approach to model landslide susceptibility areas in Boulder and Larimer Counties, Colorado. This study synthesized physical and human geography to create multiple landslide riskscape models using GIS methods. These analysis methods used a process model interface in GIS. Binary, ranked, and human factor weighted sum riskscapes were created, using frequency ratio as the basis for developing a weighting scheme. Further, spatial autocorrelation was introduced as a recommended practice to quantify the spatial relationships in landslide riskscape development. Results demonstrated that riskscapes, particularly those for ranked and human factor riskscapes, were highly autocorrelated, non-random, and exhibited clustering. These findings indicated that a riskscape model can support improvements to response modeling, based on the identification of spatially significant clustering of hazardous areas. The third project extended landslide riskscapes to measurable geostatistical comparisons using geostatistical tools within a GIS platform. Logistic regression, weights of evidence, and probabilistic neural networks methods were used to analyze the weighted sum landslide riskscape models using ArcGIS and Spatial Data Modeler (ArcSDM). Results showed weights of evidence models performed better than both logistic regression and neural networks methods. Receiver Operator Characteristic (ROC) curves and Area Under the Curve validation tests were performed and found the weights of evidence model performed best in both posterior probability prediction and AUC validation. A fourth project was developed based on the limitations discovered during the analytical process evaluations from the riskscape model development and geostatistical analysis. This project reviewed the issues with data quality, the variations in results predicated on the input parameters within the analytical toolsets, and the issues surrounding open-source application tools. These limitations stress the importance of parameter selection in a geospatial analytical environment. These projects collectively determined methods for riskscape development related to landslide features. The models presented demonstrate the importance and influence of spatial distributions on landslide riskscapes. Based on the proposed conceptual framework of a spatial riskscape for landslides, weighted sum riskscapes can provide a basis for prioritization of resources for landslides. Ranked and human factor riskscapes indicate the need to provide planning and protection for areas at increased risk for landslides. These studies provide a context for riskscapes to further our understanding of the benefits and limitations of a quantitative riskscape approach. The development of a methodological framework for quantitative riskscape models provides an approach that can be applied to other hazards or study areas to identify areas of increased human-environment interaction. Riskscape models can then be evaluated to inform mitigation and land-use planning activities to reduce impacts of natural hazards in the anthropogenic environment.
Description
Rights Access
Subject
geostatistics
riskscape
landslide
geographic information system