WUI Mapping Tool

A Python tool for mapping the Wildland-Urban Interface (WUI) anywhere in the contiguous United States. It combines building density from HISDAC-US historical built-up property records with vegetation data from the National Land Cover Database (NLCD), and classifies WUI intermix and interface zones following the Radeloff/SILVIS methodology. Results can optionally be validated against SILVIS Lab WUI maps.

The initial tool was developed by Max Warnock, Quin Browder, and Victoria Madden for GEOG 4303 at CU Boulder (Spring 2025). See the project paper for methods and results as of May 2025. I (Max) have since added additional features and improvements, such as automatic downloads for anywhere in the country, and fixes for some of the WUI mapping calculations and SILVIS map comparisons.

WUI map of Colorado's Northern Front Range, from Fort Collins to Castle Rock. Intermix (red) is housing mixed into wildland vegetation, and interface (orange) is housing next to large vegetated areas.

Features

• Downloads NLCD land cover and HISDAC-US built-up property locations (BUPL) for any state and year, plus SILVIS Lab WUI maps (optional)
• Adjustable WUI criteria (moving window radius, housing density threshold, vegetation thresholds, and buffer distance), with standard values suggested at each prompt when running the script
• Intermix and interface classification
• Optional vegetation WUI map representing WUI areas of predominantly forest or shrubland cover
• Comparison to SILVIS Lab maps with accuracy, precision, recall, F1, and a confusion matrix
• ArcGIS Pro layer files (.lyrx) for map symbology

Workflow

1. NLCD is reprojected and resampled to the BUPL grid (about 250 m), and BUPL is clipped to the NLCD study area. We do lose resolution from this, and the final WUI output cell size is 250m. However, the BUPL data provides trusted historical records of building data which is valuable for this tool.
2. A circular moving window calculates the share of wildland vegetation and the building density around each cell. Wildland vegetation includes the NLCD forest, shrubland, grassland, and wetland classes. Agriculture is left out, as in SILVIS methodolgy.
3. Cells below the minimum building density are removed. The standard value is 6.17 buildings per km^2, or about 1 housing unit per 40 acres.
4. Cells with at least 50% wildland vegetation (the standard value) are classified as intermix.
5. Cells with less vegetation are classified as interface if they fall within 2.4 km of a large vegetated area. By default, a large vegetated area is at least 5 km^2 with at least 75% vegetation.

Requirements

• ArcGIS Pro with the Spatial Analyst extension, for arcpy
• scikit-learn in the ArcGIS Pro Python environment, for the SILVIS comparison statistics
• A separate conda environment with geopandas, rasterio, and requests for the data download script (environment.yml). These are kept separate from the ArcGIS Pro environment because they can conflict with the package versions ArcGIS Pro uses.

Installation

git clone https://github.com/maxwarnock/wui_mapping.git
cd wui_mapping

ArcGIS Pro environment

This environment runs the mapping tool. ArcGIS Pro's default environment can't be modified, so you need to clone it first. You can do this in ArcGIS Pro's Package Manager (under Settings), or from the Python Command Prompt:

conda create --clone arcgispro-py3 --name wui_mapping
activate wui_mapping
conda install scikit-learn

Data download environment

This environment runs download_data.py:

conda env create -f environment.yml

Usage

1. Download data

In the wui_data_download environment:

python download_data.py --state CO --year 2020 --silvis

This saves data/nlcd_CO_2019.tif, data/BUPL_CO_2020.tif, and the SILVIS Lab WUI map for Colorado in silvis_data/. NLCD is only published for certain years (2001, 2004, 2006, 2008, 2011, 2013, 2016, 2019, and 2021), so the closest available year is used. If you run the script without any arguments, it will ask you for the state and year.

Option	Description
--state	State name or two-letter abbreviation
--year	Year to download
--resolution	NLCD resolution in meters (default: 90). Only use 30 for small states, since downloading a whole state at 30 m can run out of memory.
--silvis	Also download the state's SILVIS Lab WUI map. This is off by default because the files are large.
--skip-nlcd, --skip-bupl	Skip downloading NLCD or BUPL, for example if you only need the SILVIS map

2. Run the mapping tool

In the wui_mapping (ArcGIS Pro) environment:

python wui_mapping_tool.py

The tool uses the most recently downloaded nlcd_*.tif and BUPL_*.tif files in data/, then asks a series of questions. Press Enter at any prompt to use the standard value.

#	Question	Standard value	Affects
1	Also produce a WUI map split by vegetation class?		Extra output
2	Moving window radius (1 = 250 m)	2	Intermix and interface
3	Minimum building density (buildings per km^2)	6.17	Intermix and interface
4	Minimum wildland vegetation for intermix (%)	50	Intermix
5	Minimum size of a large vegetated area (km^2)	5	Interface
6	Minimum vegetation cover in large vegetated areas (%)	75	Interface
7	Buffer distance around large vegetated areas (km)	2.4	Interface
8	Compare the result to a SILVIS Lab WUI map?		Validation
9	Recompute statistics and a confusion matrix with scikit-learn?		Validation

To use specific files instead of the most recent downloads, set these environment variables:

Variable	Purpose
WUI_NLCD_PATH	Path to a specific NLCD raster
WUI_BUPL_PATH	Path to a specific BUPL raster
WUI_SILVIS_PATH	Path to a specific SILVIS shapefile
WUI_MAPPING_WORKSPACE	Workspace folder (default: the repository folder)

3. View the results

Outputs are saved to results/ and named with the state and year of the input data:

• wui_<STATE>_<YEAR>.tif is the WUI map (intermix and interface).
• wui_veg_<STATE>_<YEAR>.tif is the WUI map by vegetation class, if you answered yes to question 1.

Intermediate rasters and shapefiles are saved to steps/.

To apply the symbology in ArcGIS Pro, add the raster to a map and open its Symbology pane. Open the menu in the top right corner, choose Import from layer file, and select wui_symbology.lyrx (or wui_veg_symbology.lyrx for the vegetation map).

Comparing to SILVIS Lab WUI maps

Answer yes to question 8 to compare your map with the SILVIS Lab WUI map for the same state. You'll need to download the SILVIS map first with --silvis (see above). The tool finds the file for your state and compares against the SILVIS year (1990, 2000, 2010, or 2020) closest to your input data.

The comparison uses SILVIS as the reference and scores each cell as WUI or non-WUI, then reports accuracy, precision, recall, and F1. Answer yes to question 9 to also see a confusion matrix. Most cells in a study area are usually non-WUI, so precision, recall, and F1 are more useful than overall accuracy.

Project structure

Path	Contents
wui_mapping_tool.py	Main script: data preparation, WUI classification, and prompts
wui_mapping_module.py	Moving windows, vegetation mapping, and SILVIS comparison
download_data.py	Downloads NLCD, HISDAC-US BUPL, and SILVIS data for a state
environment.yml	Conda environment for download_data.py
wui_symbology.lyrx	ArcGIS Pro symbology for the WUI map
wui_veg_symbology.lyrx	ArcGIS Pro symbology for the vegetation WUI map
docs/	Project paper and README images
data/	Input rasters (created by download_data.py)
silvis_data/	SILVIS Lab shapefiles (created by download_data.py --silvis)
steps/	Intermediate outputs (created when the tool runs)
results/	Final WUI maps (created when the tool runs)

Data sources

• NLCD: U.S. Geological Survey, Multi-Resolution Land Characteristics Consortium
• HISDAC-US BUPL: Uhl, J. H., and Leyk, S. Historical Settlement Data Compilation for the United States, Harvard Dataverse. https://doi.org/10.7910/DVN/U2P66Z
• SILVIS Lab WUI maps: University of Wisconsin-Madison SILVIS Lab. https://silvis.forest.wisc.edu/data/wui-change/
• State boundaries: U.S. Census Bureau TIGER/Line

References

• Bar-Massada, A., Stewart, S. I., Hammer, R. B., Mockrin, M. H., and Radeloff, V. C. (2013). Using structure locations as a basis for mapping the wildland urban interface. Journal of Environmental Management, 128, 540-547. https://www.fs.usda.gov/rm/pubs_other/rmrs_2013_bar_masada_a001.pdf
• Radeloff, V. C., Helmers, D. P., Kramer, H. A., et al. (2018). Rapid growth of the US wildland-urban interface raises wildfire risk. Proceedings of the National Academy of Sciences, 115(13), 3314-3319. https://doi.org/10.1073/pnas.1718850115

License

This project is licensed under the MIT License.