Brooks Groves
Featured Project View on GitHub →

Wildfire · GIS · R · Haversine · NIFC · OpenStreetMap

🔥 Wildfire Response Gaps

"It's already on the ground."

When a wildfire ignites in eastern Oregon, help isn't minutes away — it's hours. Sometimes it isn't coming at all, not from a mapped fire station. This project measures that gap across every wildfire perimeter on the West Coast, names the stations carrying the load, and tracks the 2026 season live.

Haversine distances from 3,635 fire perimeters to the nearest mapped fire station across California, Oregon, and Washington. The Oregon finding was unexpected: a median fire-to-station distance of 26.7 km — three times California's — with 30% of fires beyond 50km from any station. In Malheur County, the median is 139 km. Every single fire is beyond 50km.

The pipeline runs automatically: Python fetches fresh perimeters from the NIFC WFIGS API at 5am Pacific, recomputes distances, commits GeoJSON. R rebuilds the NatGeo-style static map at 6am. No manual intervention required.

R Python Leaflet NIFC WFIGS OpenStreetMap US Census TIGER GitHub Actions ggplot2
View Live Map → GitHub →
Oregon median distance
26.7 km
vs 8.6 km Washington, 10.8 km California
Oregon fires beyond 50km
30.1%
vs 1.9% WA · 1.5% CA
Malheur County median
139 km
larger than New Hampshire · 9,930 sq mi
Fire perimeters analyzed
3,635
CA · OR · WA · historical + 2026 YTD
Frenchglen coverage zone
516,867 ac
148 fires · 114 min mean response · pop. ~12
State Comparison
State Fires Median Distance % Beyond 50km Pattern
Washington 1,388 8.6 km 1.9% Dense station coverage · western model works
California 1,051 10.8 km 1.5% Populated corridors well-covered
Oregon 1,196 26.7 km ⚠ 30.1% ⚠ Eastern rangeland structural gap — RFPAs fill the void
Context

Why Oregon Is Different

Eastern Oregon is enormous, sparsely populated rangeland. Malheur County alone is 9,930 square miles — larger than New Hampshire — with a median fire-to-station distance of 139.1 km. Every single fire in the county is beyond 50km from any mapped station.

Oregon's answer is Rangeland Fire Protection Associations (RFPAs) — legally recognized volunteer networks of ranchers and landowners organized to protect over 17.5 million acres that have no existing state or local fire protection. Codified in Oregon statute. Oregon is the only state in the country with this formal structure.

The infrastructure gap is real. The people filling it are remarkable.

Frenchglen Fire Guard Station

Frenchglen Fire Guard Station sits in Harney County — population approximately 12. It is the nearest mapped station to 148 wildfires covering 516,867 acres, with a mean estimated response time of 114 minutes.

But Frenchglen isn't alone. It's one node in a network that includes the Frenchglen RFPA, BLM resources, and US Fish & Wildlife. The ranchers in these associations are often on scene hours before federal resources arrive — with fire-equipped trucks, slip-on water tanks, and generations of knowledge of this land.

The largest fire in Frenchglen's coverage zone: the Falls Fire — 151,683 acres, 2024.

More Geospatial Work All on GitHub →

WUI · Search & Rescue · Phoenix FD · Clustering

🔥 RIDGELINE — WUI Call Volume

86K+ Phoenix Fire Department dispatch incidents. Four behavioral clusters extracted from real nature codes. Interactive maps, CI/CD pipeline.

View project →

Hydrology · USGS · Leaflet · Sierra Nevada

💧 Sierra Streamflow Monitor

8 USGS stream gages mapped across Sierra Nevada watersheds. Historical spaghetti charts, Leaflet map, weekly GitHub Actions cron.

View dashboard →
Project Kiva Flag

LiDAR · Archaeology · USGS 3DEP · Chaco Canyon

🏜 Project Kiva

Southwest archaeology remote sensing pipeline using real USGS 3DEP LiDAR DEMs. Chaco Canyon, Mesa Verde, Hohokam Phoenix, Canyon de Chelly. Sky-View Factor, Local Relief Model, automated kiva detection, live Leaflet map.

View live map →
New — Remote Sensing Intelligence GitHub →

Sentinel-2 · ESA · Planetary Computer · pixi · Columbia River

🛰️ Sentinel Stockpile

Commodity intelligence pipeline using free ESA Sentinel-2 satellite imagery. Monitors industrial stockpile levels at fixed locations over time — currently tracking the lumber cycle at the Port of Longview, Washington, the largest log export terminal on the Columbia River. 12-month time series: 47 hectares in July, 5 hectares in January. The $40 billion Pacific Northwest timber supply chain, visible from 500 miles up.

Spectral classification using NDVI, BSI, and brightness indices. Every 10m pixel classified as water, vegetation, stockpile, or ground. Extensible to container terminals, oil tank farms, grain elevators, mining stockpiles. Free data, open tools, reproducible pipeline.

Read the Build Post → GitHub →