Ridgeline

Search and rescue at the wildland–urban edge, from public fire-department dispatch data.

Ridgeline collects every mountain, technical and water rescue call that three fire departments publish, puts each one on a map, and counts them by year, month, hour and trailhead. The goal is to show when people get into trouble near cities built against steep country, and where it happens.

Live site: brooksgroves.com/ridgeline · Detailed Phoenix map (trails, trailheads, preserve boundaries): brooksgroves.com/ridgeline/map.html

City	Agency	Coverage	Mountain & technical rescues located
Phoenix, AZ	Phoenix Fire Department	2019–2025	1,489 of 1,619 (92%)
Scottsdale, AZ	Scottsdale Fire Department	Dec 2022 – present	360 (466 of 522 rescue calls located, 89%)
Boulder, CO	Boulder Fire-Rescue	2015 – present	581 (published points, 100%)

Plus a national parks section: all 20,058 search-and-rescue incidents the National Park Service logged from 2013 to August 2021, on a map of every park's boundary with one circle per park, and each park's calendar and incidents per million visits.

The city pipeline reruns every Monday, the national parks data monthly, and both whenever the code changes (see Automation). The counts in this README reflect the run of September 29, 2026; the live site always shows the current numbers.

What the data shows

These are descriptive patterns in dispatch counts. None of them is a causal claim.

A few trailheads account for most of the calls. In Phoenix, Echo Canyon on Camelback Mountain (dispatch address 49XX E MCDONALD DR) accounts for 310 of the 1,489 located mountain rescues, about a fifth. Piestewa Peak, Cholla on Camelback, Pima Canyon on South Mountain and the South Mountain Park entrance make up most of the rest. In Boulder, 243 of 581 technical rescues sit at Chautauqua Park, the main entry to the Flatirons.

Phoenix mountain rescues have fallen by about a fifth. Counting every call, located or not, there were 255 in 2019 and about 203 a year in 2023–2025. (An earlier version said "about a third", from located calls only; that overstated the drop, because which calls could be located changed over time.) Those years overlap Phoenix's heat-triggered trail closures, but the closures can't be most of the explanation (next paragraph). Other changes over the same years (visitation, dispatch coding, the pandemic) could produce the same shape.

Heat closures: a signal on closure days, too small to explain the long drop. Phoenix closes the Echo Canyon, Cholla and Piestewa Peak trails on National Weather Service heat days, and added South Mountain in October 2024. The rules changed several times, and src/analysis/heat.py scores every date under the rules in force that day:

From	Closed	Trigger	Rules
Jul 16, 2021	11 a.m.–5 p.m.	Excessive Heat Watch	Pilot through Sep 30: Echo Canyon and Piestewa Peak (Cholla closed for renovation)
May 1, 2022	11 a.m.–5 p.m.	Warning	May–September
May 1, 2023	11 a.m.–5 p.m.	Warning	May–September rules
Aug 31, 2023	9 a.m.–5 p.m.	Warning	Year-round
Oct 25, 2024	8 a.m.–5 p.m.	Warning	South Mountain added
Mar 27, 2025	8 a.m.–5 p.m.	Warning	South Mountain limited to Holbert, Mormon, Hau'pal Loop and Pima Canyon access to the National Trail

The timeline was shared by Yun-Peng (Liz) Lu (University of Maryland) and checked against 2021 and 2023 coverage and the city's 2024 release.

The analysis lines up all 1,619 mountain-rescue calls (trail assigned from the dispatch address, so unlocated calls count) with every NWS heat warning and watch for zone AZZ543, Central Phoenix, from the IEM VTEC archive, and Open-Meteo daily highs at Sky Harbor. Counting every date a warning touched gives 20, 18, 42 and 45 days for 2021–2024, exactly the city's 2024 program review. Closure days are stricter: the trigger must be in effect during that day's closed hours, inside the program's season (8, 17, 42 and 45 days; 2021 is the pilot only, and 2022 drops July 17, when the warning expired at 2 a.m.). As an independent check, an ABC15 report at the end of the 2021 pilot, found by Yun-Peng Lu, said the trails closed on eight days with no mountain rescues during closure hours: the same 8 days and zero closed-hour rescues this analysis reconstructs. Before is January 2019 to July 15, 2021, where a heat day is one with a warning in effect from 11 a.m. to 5 p.m.

Rescues per 100 days, May–Sep	Before, heat days	Before, ordinary	After, closure days	After, ordinary	Net change (95% range)
Closure trails	32	34	11	24	0.49 (0.25–0.98)
South Mountain (closed from Oct 2024)	13	10	8	7	0.83 (0.31–2.22)
Other Phoenix trails	34	32	30	26	1.08 (0.62–1.87)

Closure days got relatively quieter only at the closed trails, about a 51% relative drop, and the interval only just excludes no effect. The counts are small (25 and 15 calls), so it's a signal, not a measurement. The closure trails' long decline (119 rescues in 2019, about 74 a year in 2023–2025) is year-round, and with about 31 warning days a year the closures could account for about 6 rescues a year at most. Since the pilot began, 17 rescues at closed trails happened during that day's closed hours (none in 2021–22); the site lists them.

Per hiker, not per day. Phoenix publishes daily trail-counter data (infrared counters, 2019 on, CC BY; src/ingest/fetch_trail_counts.py, weekly). Two counters sit on the closure trails for the whole period, Echo Canyon and the Piestewa Summit Trail, so rescues at those two trailheads can be divided by the counters' passes on the same days (2,296 days with both reporting). A pass is one person crossing the counter, so the rates are for comparison, not "1 in N hikers."

	Passes a day	Rescues per 100,000 passes
Daily high below 80°F	~2,170	8.9
Daily high 105°F and up	~740	26.3
May–Sep, before the program: ordinary days / heat days	1,355 / 918	21.5 / 50.5
May–Sep, after: ordinary days / closure days	1,060 / 602	20.6 / 10.5
2019 / 2024, all days	1,588 / 1,518	14.3 / 9.6
2019 / 2024, November–April	1,952 / 2,042	10.2 / 7.8

Three things follow. Heat roughly triples the risk per hiker: rescues per day look flat with temperature only because the crowd shrinks. Closure days are safer per hiker than ordinary days, even with about 600 passes a day still counted outside the closed hours (only 8 rescues, so treat it as a signal). And the long decline isn't fewer hikers: traffic in 2024 was about 2019's, including winter, but rescues per pass fell by about a third.

Did rescues move to another call type? fetch_phoenix_fire.py tallies every call type at the closure trailhead addresses from the full dispatch data (data/external/trailhead_call_types.csv). Mountain rescues went 119, 83, 108, 92, 75, 71, 77 (2019–2025); every other published call type combined, including calls whose type the city withholds, went 51, 37, 39, 49, 48, 35, 42, with no rise. Medical calls left out of the public data entirely can't be checked.

The pandemic and trail closures. Two checks, both rerun weekly (sensitivity and counter_checks in heat_report.json):

Closure-trail net change	Ratio	95% range
All years	0.49	0.25–0.98
Without 2020	0.42	0.19–0.90
Without 2020–21	0.49	0.20–1.20

Leaving out the pandemic years doesn't make the closure effect go away. Using the 17 city trail counters that reported every year, city-wide traffic ran +19% in 2020 and +12% in 2021 against 2019, then fell back below 2019 from 2022 (−13% in 2024). An earlier sum over all counters overstated the boom, because the city added counters over time. The long decline in rescues runs from 2019 (before the pandemic) to 2023–25 (after it), so the pandemic can't explain it, though it can explain the 2021 bump.

The counters also show trails closed for other reasons. Echo Canyon's counter reads zero in June–July 2020 (16 passes a day in August), which is most of why 2020 had so few rescues at the closure trails; I found no city notice for it. Echo Canyon closed again in October 2025 after storm damage (KJZZ), which lowers late-2025 counts, and Cholla was closed for about a year and a half after a 2020 boulder accident. Per-hiker rates skip zero-count days, so they aren't distorted. What counters can't show is who hikes: a post-pandemic crowd with more experienced locals and fewer visitors could make each hike safer on its own.

Watches. A watch later upgraded to a warning is stored in the VTEC archive with its end before its start, so it has no in-effect time of its own. In this zone every 2021 pilot watch was upgraded to a warning covering the same days, so watches add no closure days unless a closure began on the day a watch was issued (the product time). That reading isn't used here.

The city's "rescues on closed trails" figures. Phoenix's October 2024 release reports 57, 47, 30 and 35 rescues on closed trails for 2021–2024, without a definition. Counting every mountain-rescue call at the closure trails from May through October, on any day and at any hour, gives 55, 50, 33 and 36, within 3 a year. So the figures most likely count trails that close, not rescues while closed (those are 0, 0, 3 and 9 under the rules in force each day). Over the same years the November–April count at those trails fell just as much (53 to 35), with no closures in effect.

National parks: a few parks carry most of the load. From 2016 to 2020 the Park Service logged about 3,050 search-and-rescue incidents a year. Yosemite (~280 a year), Grand Canyon (~260) and Lake Mead (~240) account for 26% of them. Per visitor, Sequoia & Kings Canyon is highest at 90 incidents per million visits, then Yosemite (70) and Grand Canyon (47). Great Smoky Mountains, the most-visited park, has 5.8. July is the busiest month nationally. At Yosemite, June through September is the season, and January is almost silent.

Each city has its own season:

	Busiest month	Quietest month	Share on weekends
Phoenix	March	December	41.2%
Scottsdale	February	June	41.4%
Boulder	July	January	41.5%

Boulder follows the classic mountain calendar. Scottsdale is the reverse: winter and spring are busy, and summer goes quiet as people avoid the heat. Phoenix barely has a season at all; March and May are as busy as July. Across all three cities about 40% of calls fall on Saturday or Sunday, which make up 29% of the week. Calls peak late morning in Phoenix and around midday in Boulder. Scottsdale publishes dates without times.

Data sources

All three sources are public and need no key. Each is pulled fresh on every run.

	Phoenix	Scottsdale	Boulder
Dataset	Fire calls for service, one CSV per year	Fire Department calls for service (ArcGIS layer)	Response times for Boulder Fire-Rescue (ArcGIS layer)
License	CC BY	City of Scottsdale open-data terms	CC0 1.0
Unit	One row per call	One row per call (deduplicated on incident number)	One row per incident
Location	Address masked to the hundred block (`49XX E MCDONALD DR`) or an intersection (`N 7TH ST/E DUNLAP AV`)	Full street address	Point geometry, no address
Time	Date and time	Date only	Date and time
What counts as mountain / technical	Nature text Mountain Rescue, plus tree, confined-space and rescue-service calls	Department code `MTNRES`, plus High-angle rescue	Call types Rescue, Auto-Aid Tech Rescue, Auto-Aid Rescue; NFIRS High-angle rescue and land searches
What counts as water	Water Rescue, Check Flooding Condition, swift-water calls	Code `WATER`, Swift water rescue	Water Rescue call types; NFIRS swift-water, ice and in-water search codes

Scottsdale's land searches (Search for person on land outside the MTNRES code) are kept as a separate land search category. Many of them are urban missing-person calls rather than trail work. Boulder's auto-aid calls are Boulder units sent outside the city, often into county open space.

National parks. The NPS FOIA reading room publishes two search-and-rescue incident lists, NPS-SAR-Incidents-List-2013-2018.xlsx and SAR-Incidents-List-2019-2020.xlsx. The second actually runs to August 12, 2021. Each row is an incident number, date, incident type, park code and region; there is no location inside the park, no time of day and no outcome. src/ingest/fetch_nps.py combines them (20,058 incidents after dropping 420 rows with no date, most of which also have no park, and 2 duplicate IDs), maps a few spelled-out park codes (GRANDCANYON → GRCA), and treats Sequoia & Kings Canyon as one park (SEKI) as the incident data does. Park outlines come from the NPS Land Resources Division boundary service, and annual recreation visits from the NPS IRMA visitor use statistics service. These are periodic FOIA releases, so nps.yml refreshes them monthly rather than weekly, and adds any new SAR list that appears on the FOIA page. To run it locally: pip install httpx openpyxl, then python src/ingest/fetch_nps.py.

On the map, every park with incidents gets its boundary. Zoomed out, the page draws a light copy snapped to about 2 km (site/data/nps_overview.geojson, built from the committed boundaries); from zoom 7 each park in view swaps to its detailed outline (site/data/nps/<CODE>.geojson). Ten very small units have no outline at national scale and show as circles only. Clicking a park's circle or boundary opens a popup (incidents a year, rank, rate per million visits, busiest month, weekend share, date range) and loads its charts below the map.

Reporting ramps up over the first years: 570 incidents are dated 2013, 725 in 2014 and 1,230 in 2015, then about 3,000 a year. Yosemite logged 9, 11 and 47 in 2013–2015 and 372 in 2016. That's the reporting system coming into use, not safer parks, so averages and rates use the five complete years, 2016–2020.

Other cities were checked and set aside. Santa Barbara (city and county) doesn't publish incident-level fire calls. San Diego publishes a generic RESCUE category with only a ZIP code for location, which is too coarse to map. The probe that tested these sources is in tools/probe_sources.py, and its last results are in data/external/source_probe.json.

Method

Phoenix CSVs ─┐                                        ┌─ phoenix_sar_incidents.geojson
Scottsdale ───┼─ filter rescue call types ─ locate ─────┼─ scottsdale_sar_incidents.geojson ─┐
Boulder ──────┘                                        └─ boulder_sar_incidents.geojson     ├─ site/index.html
                         geocode_cache.csv ◄──► Maricopa County geocoder                    │
                         *_report.json  (match rates, counts)  ─────────────────────────────┘

1. Filter

Phoenix: calls are kept when the nature text matches a rescue, flood or heat term, or when the address names a preserve or trailhead. Preserve matching uses patterns that mean the place (Echo Canyon, Piestewa Peak, McDowell Sonoran), never an arterial that shares its name. The patterns are in src/ingest/addresses.py.
Scottsdale and Boulder: rescue call types are selected on the server with an ArcGIS query.

2. Locate

Boulder publishes points, so nothing there is geocoded. Phoenix and Scottsdale addresses go to the Maricopa County geocoder in these steps:

Normalize the address. Hundred blocks become the block midpoint (49XX → 4950). Intersections become A & B. Phoenix Fire's abbreviations are expanded (PW → PKWY, AV → AVE). A renamed street is updated (Squaw Peak Dr → Piestewa Peak Dr).
Accept a match scoring 80 or higher.
Accept a near miss (score 70–80) only when it lands on the same street. Hundred-block midpoints often fall outside a street's address range, which lowers the score even when the street is right. So a near miss is kept only if it is an address, street or intersection match, the geocoder's matched address contains the street name that was asked for, and the direction (E/W, N/S) and street type (Rd/Pl) agree.
Try the U.S. Census Bureau geocoder. It interpolates along TIGER address ranges, which cover park roads and hundred blocks the county file doesn't (North Mountain on 7th Street, Lookout Mountain on 16th Street, the south side of South Mountain). Its results are public domain, so they can be cached. The same street, direction and type checks apply; they rejected, for example, E Northern Ave landing on W Northern and S 19th Ave landing on N 19th.
Fall back to a preserve. If the geocoder fails and the address clearly names a preserve, the call is placed at a representative point for that preserve.
Otherwise the call is not located. It stays in the counts that note it, but it isn't on the map.

Every result is written to data/external/geocode_cache.csv, so the next run only queries addresses it hasn't seen. The cache records the query, score, matched address, match type, method and a logic version. Misses from an older logic version are retried; hits are kept.

For Phoenix mountain rescues, the 1,619 calls resolve like this:

Outcome	Calls
Geocoder match, score ≥ 80	275
Same-street near miss, score 70–80 (same direction and street type)	683
Census Bureau geocoder (same checks)	198
Preserve representative point	333
Not located	130
Total (1,619 unique incident IDs)	1,619

3. Publish

One GeoJSON file per city is written to data/external/, together with a JSON report of counts and match rates. src/viz/build_site.py computes every number on the site directly from those files and renders the page. Its charts are inline SVG, and the map is Leaflet reading the same GeoJSON. Nothing on the site is modeled or estimated.

Automation

Three workflows keep everything current. None needs a key or secret.

Workflow	When	What it does
`deploy.yml`	Every Monday 06:00 UTC, every push, on demand	Fetches the three cities' dispatch data, filters, geocodes (cache first), exports GeoJSON; refreshes heat warnings and daily highs and reruns the heat analysis; commits the cache, reports and data back; builds the site and deploys it to GitHub Pages
`nps.yml`	1st of each month 07:00 UTC, on demand, and when the fetch code changes	Re-downloads the NPS search-and-rescue lists, park boundaries and visits; picks up any new SAR list the FOIA page adds; if anything changed, commits it and starts `deploy.yml`
`probe.yml`	On demand	Tests candidate cities' open data and records what they contain in `source_probe.json`

To refresh by hand: Actions → workflow → Run workflow on GitHub.

How to read the site

A circle is a place, not a call. Calls that share a location are drawn as one circle, and a bigger circle means more calls. Click a circle to see its call count, date range, and how its location was determined.
Map layers. The map opens showing mountain and technical rescues only. Use the buttons under the map to add flood and water rescues or Scottsdale's land searches.

Location precision. Each circle is labeled with how precisely it is placed:

Label	Meaning	Typical accuracy
Middle of the hundred block	Phoenix address masked to the block	within ~50 m along the street
Intersection named in the call	The call gave two streets	the intersection itself
Street address	Scottsdale full address	the parcel frontage
Point published by the city	Boulder's own location	as published; many calls share one point
Preserve (representative point)	Only the preserve was known	anywhere in that preserve

Counts are dispatches, not people or outcomes. A call can involve one hiker or a whole group. Some calls were cancelled en route or found no one.
Compare shapes, not totals, across cities. Each department codes rescues differently, and county sheriff's teams and volunteer groups handle many backcountry rescues that never appear in city fire data.
Faded bars are partial years. Totals for those years aren't comparable to full years.

Output files

Everything the site needs is committed, so the site can be rebuilt without re-downloading anything.

File	Contents
`data/external/{city}_sar_incidents.geojson`	One point per located call
`data/external/geocode_report.json`	Phoenix counts, match rates by method and precision, top unlocated addresses
`data/external/nps_sar_incidents.csv`	NPS SAR incidents: id, date, type, park, region
`data/external/nps_parks.geojson`, `nps_visitation.csv`	Park outlines (with a label point) and annual recreation visits
`data/external/nps_report.json`	NPS fetch report: rows kept, dropped and why; boundary and visitation matches
`data/external/heat_report.json`	Heat analysis: warning days per year, rescue rates by temperature and by warning day, before/after, closed-hour calls
`data/external/phoenix_mountain_calls.csv`	All Phoenix mountain-rescue calls: date, hour, dispatch address
`data/external/phoenix_heat_warnings.csv`, `phoenix_daily_weather.csv`	NWS heat warnings (zone AZZ543) and daily weather at Sky Harbor
`data/external/scottsdale_report.json`, `boulder_report.json`	Per-city counts, categories and date ranges
`data/external/geocode_cache.csv`	Every address ever looked up, and how it resolved
`data/external/source_probe.json`	Schema and vocabulary check of candidate sources

GeoJSON feature properties:

Property	Description
`incident_type`	The call type or description as published
`category`	`mountain`, `water` or `search` (Scottsdale and Boulder; for Phoenix the site derives it from `incident_type`)
`date`, `year`, `hour`	Local date and time of the call (`hour` is empty for Scottsdale)
`is_weekend`	Saturday or Sunday
`location_name`	The dispatch address as published (empty for Boulder)
`precision`	`hundred_block`, `intersection`, `address`, `preserve_centroid` or `published_point`

Limitations

Unlocated calls. 28% of Phoenix mountain rescues aren't on the map. The largest unmatched addresses are 106XX N 7TH ST (North Mountain), 51XX N INVERGORDON RD and N GALVIN PW (Papago Park). The county geocoder doesn't resolve them.
Near-miss matches are street-accurate, not door-accurate. They are checked against the street name, but the point is the geocoder's position along that street.
Preserve points are area-level. 180 Phoenix mountain rescues sit at a preserve's representative point.
Boulder shares points. 99 calls share a single point near Bear Peak, which is probably a stand-in location for backcountry calls. Treat Boulder's map as area-level outside Chautauqua.
Coverage differs by city. Phoenix is 2019–2025, Scottsdale starts in December 2022, and Boulder starts in 2015. Scottsdale has no time of day.
Fire-department calls only. Sheriff's offices, volunteer teams and land agencies run many backcountry rescues, and none of those appear here.

Corrections

September 2026. An earlier version reported 2,263 Phoenix incidents, led by Crisis / Behavioral Health (1,120), which it described as concentrated "along the preserve edges," plus a Wildland Fire (186) cluster. Both were artifacts of two bugs:

Street names matched as preserves. The preserve-name fallback matched street names: McDowell, Camelback and South Mountain also matched McDowell Road, Camelback Road and South Mountain Avenue. That pulled ordinary city calls on those arterials into the dataset and pinned them to preserve centroids up to 20 miles away. About 1,560 of the 2,263 old points sat on four such centroids.
Hundred-block addresses didn't geocode. The geocoder couldn't score hundred-block addresses or A/B intersections, so most real mountain rescues dropped out without any error. Only 474 of 1,619 reached the map.

Both are fixed. Mapped Phoenix mountain rescues rose from 474 to 1,159 (1,197 after the follow-up fixes below). The crisis-care and wildland-fire clusters disappeared: 5 crisis calls remain, all at genuine preserve addresses. Every point now records its precision, and match rates are published with each run in geocode_report.json. The weekly workflow had also been disabled by GitHub for inactivity, so the map had been a hand-exported snapshot. Geocoding now runs in CI.

Running it yourself

Requires pixi. Everything else, including Python, comes from pixi.

git clone https://github.com/bdgroves/ridgeline.git
cd ridgeline
pixi install

pixi run phoenix      # download Phoenix Fire CSVs and filter rescue calls
pixi run geocode      # geocode Phoenix (uses and updates the committed cache)
pixi run export       # write phoenix_sar_incidents.geojson
pixi run scottsdale   # fetch + geocode Scottsdale
pixi run boulder      # fetch Boulder
pixi run build        # build site/ from the committed GeoJSON
pixi run serve        # http://localhost:8080

pixi run build works on a fresh clone with no downloads, because it reads only the committed files. The first full geocoding run takes a few minutes; later runs only query new addresses.

Repository layout

ridgeline/
├── .github/workflows/
│   ├── deploy.yml             weekly pipeline → GitHub Pages
│   └── probe.yml              one-off source probe
├── src/
│   ├── ingest/
│   │   ├── fetch_phoenix_fire.py   Phoenix CSVs → filtered calls
│   │   ├── geocode_phoenix.py      geocoder, cache, match report
│   │   ├── addresses.py            normalization + preserve patterns
│   │   ├── fetch_scottsdale.py     Scottsdale ArcGIS → geocoded GeoJSON
│   │   ├── fetch_boulder.py        Boulder ArcGIS → GeoJSON
│   │   ├── arcgis.py               ArcGIS paging + GeoJSON writer
│   │   └── fetch_gis.py            trails, trailheads, parks for the Phoenix map
│   ├── analysis/                   exploratory plots, weather join, model (not on the site)
│   └── viz/
│       ├── build_site.py           site/index.html
│       └── build_map.py            site/map.html (detailed Phoenix map)
├── export_geojson.py               Phoenix geocoded calls → GeoJSON
├── tools/probe_sources.py          checks candidate open-data sources
└── data/external/                  committed outputs (see above)

Follow-up, late September 2026.

Opposite-direction matches. A Scottsdale call on East Indian School Road had been matched with a high score to West Indian School Road, 15 miles away. Matches on the right street with the opposite direction are now rejected (7 of about 1,300).
Cholla's new address. Since the Cholla Trail reopened in 2022, Phoenix Fire logs its calls at 51XX N INVERGORDON RD, the trailhead corner in Paradise Valley, instead of 62XX E CHOLLA LN. Those 59 rescues had failed to geocode, and the first run of the heat analysis counted them as "other trails", which overstated the closure-trail decline. Both are fixed.
Trend from all calls. The "down about a third" headline used located calls only. Because the Cholla calls became unlocatable from 2022, that overstated the decline. Trends now use all 1,619 calls: down about a fifth.
Closure policy timeline. The first heat analysis assumed one rule (9 a.m.–5 p.m. on a warning, from May 2021). The program actually began as a July 16, 2021 pilot at 11 a.m.–5 p.m. on a watch, and moved to 9 a.m. only on August 31, 2023. With each date scored under its own rules, the closure-trail net change went from 0.68 (0.34–1.33) to 0.49 (0.25–0.98), and closed-hour rescues from 21 to 17. Thanks to Yun-Peng (Liz) Lu for the timeline and for flagging the watch records.
"Rescues don't rise with temperature." True of rescues per day, but misleading: trail-counter data shows the crowd shrinks in the heat, and per hiker the risk roughly triples above 105°F. The heat section now leads with the per-hiker rate.
Wrong street type. Near-miss matches must now also have the same street type. This rejected six wrong matches, including 7XX E DESERT FOOTHILLS PW → Desert Flower Ln (21 South Mountain rescues, now unplaced rather than misplaced) and Invergordon Rd → Invergordon Pl, three miles north.

Roadmap

Credits and licensing

Data belongs to its publishers: the City of Phoenix (CC BY), the City of Scottsdale (open-data terms), and the City of Boulder (CC0). Geocoding is by Maricopa County GIS; basemap tiles are © Esri. Built by Brooks Groves. Questions or corrections: contact@brooksgroves.com.

Name		Name	Last commit message	Last commit date
.github/workflows		.github/workflows
data		data
src		src
tools		tools
.gitignore		.gitignore
README.md		README.md
export_geojson.py		export_geojson.py
pixi.lock		pixi.lock
pixi.toml		pixi.toml

Navigation Menu

Repository navigation

About

Repository files navigation

Ridgeline

Contents

What the data shows

Data sources

Method

1. Filter

2. Locate

3. Publish

Automation

How to read the site

Output files

Limitations

Corrections

Running it yourself

Repository layout

Roadmap

Credits and licensing

About

Releases

Packages

Contributors

Languages

Latest commit

History

Folders and files

Repository files navigation

Ridgeline

Contents

What the data shows

Data sources

Method

1. Filter

2. Locate

3. Publish

Automation

How to read the site

Output files

Limitations

Corrections

Running it yourself

Repository layout

Roadmap

Credits and licensing

About

Topics

Resources

Stars

Watchers

Forks

Releases

Packages

Contributors

Languages