Urban Growth Modeling for GLISA

Starting in Summer 2021, I have been working with Dr. Derek Van Berkel to develop means of predicting and visualizing urban growth for use in knowledge co-production efforts by Great Lakes Integrated Sciences and Assesments (GLISA). My primary role in this project has been to implement the FUTure Urban-Regional Environment Simulation (FUTURES) model to predict future urban development, but I have also been an active participant in stakeholder meetings and contributed to the development of an interactive web app for presenting model results.

Skills and software:

GRASS GIS
R (packages: raster, sf, tidycensus, ROCR, MuMIn, lme4, shiny)
Model accuracy assesment and hindcasting
Shiny web app development

Investigating Drivers of Urban Development

Much of my involvement has centered around testing various potential predictors of urban development, such as distance to amenities, ease of building, and climate-related risk factors.

Predictors are tested by running general linear models in R and then comparing relative effect sizes and significance of each variable. Measurements of goodness-of-fit are computed using the ROCR package.

Site suitability for Colorado Springs and Pueblo, CO

Visualizing Future Urbanization

An ongoing component of the project has been to create animations and map layouts with which to visualize the projected impact of future urban growth.

Hindcasting

I used the grass.jupyter package for Python to automate hindcasting runs and calibrate patch compactness and distribution variables. Models were run during time periods where development is known via NLCD datasets, with hindcasting predictions compared to actual development to calculate error rates and modified kappa statistics.

Comparing Climate Risk Scenarios

FUTURES can be used to investigate the role climate-related risks play in urban planning and to anticipate future challenges for growing cities.

The animations on the left depict the area around Colorado Springs and Pueblo, CO. In the top model prediction, the USDA Forest Service's Risk to Potential Structures data layer was used in calculating site suitability. On bottom, it was not. The risk layer is used for both maps as a basemap.

Animation of predicted development for Colorado Springs area

Predicted development for Grand Rapids, MI overlaid with FEMA floodplain map

Animation of predicted development for Grand Rapids area

Participatory GIS App

In addition to running models and investigating drivers of urban change, I also contributed to a participatory GIS app that can be used to display and contextualize FUTURES model results. The app was used in several GLISA workshops to prompt dialogue between great-lakes area urban planners and University of Michigan Researchers. This app was later developed into an open-source R package, with more details here.

Page updated

Google Sites

Report abuse