Crosswhite was first reported in Wheeler County, Oregon, at 5:51 p.m. PDT on July 15. In the seven-day forecast sequence for that same target period, guidance at its reported origin began at Very high and the final update was High.
The sequence shows changing forecast guidance, not observed fire growth, ignition prediction, or increasing forecast accuracy. The earlier day-ahead analysis is retained separately below.
FLARE makes that change visible location by location. For an insurer reviewing properties or a utility team monitoring infrastructure, the useful question is where danger is rising around the places they protect. This story follows that question from Wheeler County to the land around Crosswhite.
01 / SEVEN-DAY REPLAY · WHEELER COUNTY
Seven forecasts for July 15
Seven evening updates, issued at 8 p.m. PDT from July 8–14, each target the same period: July 15 at 5 a.m. to July 16 at 5 a.m. PDT. The issue date advances; the valid period stays fixed.
Experimental: Days 2–7 use NBM; Day 1 uses HRRR. This is changing forecast guidance, not observed fire growth or a prediction of ignition or spread.
Loading the seven verified forecast updates…
Forecast issue: July 8, 8 p.m. PDT · Day 7 · NBM · target period fixed at July 15–16, 5 a.m. PDTDownload this issue map (PNG)
Reported-origin markers: 1 Crosswhite · 2 Red Rock · 3 Twickenham · 4 Camel Hump · 5 Cove Creek. Colored outlines are later observed reference perimeters, not forecast fire extent. USGS Topo supplies geographic context beneath the native forecast cells.
County area rated High or aboveArea-weighted; denominator includes all county pixel centers.Compare all five reported fire origins
Stored category at each reported origin across the seven issue dates. Reference points are not forecast fire boundaries.
This earlier analysis uses a different category scale and reports Very high-or-above area. By July 14, it rated about 2,659 km² of Wheeler County Very high or Extreme.
Gray: unavailable or special land/status codes July 16 Crosswhite footprint, shown as a reference
JULY 14
59.8% of the county rated Very high or Extreme / 2,659 km²
97.8% valid FLARE coverage
The solid outline follows Wheeler County’s administrative boundary, with forecast colors shown inside it. The smaller dashed outline marks Crosswhite’s early fire perimeter. Both maps cover the same area at each product’s original grid resolution, with separate scales and units below.
THE LONGER VIEW
The wider July pattern.
Elevated danger was already widespread earlier in July. The four days around Crosswhite sit within a series of rises and falls. By July 20, the county’s share rated Very high or Extreme had fallen to 4.1%.
County area rated Very high or Extreme % of the entire area within the county boundary
July 13–16Total county area is the reference for every day.
From the county to a location.
A typical FLARE cell here covers 0.122 km², compared with 1.002 km² for the USGS products. FLARE shows the landscape in cells covering roughly one eighth of that area.
For teams reviewing a property or utility corridor, this reveals how mapped danger varies across the surrounding land. The Crosswhite view below follows that local pattern.
FLARE0.122 km²
USGS1.002 km²
Squares show the relative area of a typical cell, using median values. The map cells follow the shapes of the original grids.
03 / EARLIER DAY-AHEAD · CROSSWHITE
Very high danger near the fire’s starting point.
This earlier FireOutlook used a different category scale: it rated the reported origin Very high for July 15, 5 a.m.–July 16, 5 a.m. PDT. In the seven-day product, the final index is about 0.7086, categorized High on its stored scale. Keep the two methods separate.
FLARE brings the local pattern into view: how fire danger changes across the land around Crosswhite, as well as across the county.
04 / FROM OUTLOOK TO ACTION
Anticipate. Analyze. Act.
FLARE connects fire danger guidance with vegetation changes and observed thermal activity. For an insurer or utility team, that means a single place to set monitoring priorities, examine a location, and prepare a briefing.
1 / 3 · Anticipate
Plan for tomorrow’s conditions.
FireOutlook puts the next day’s fire danger beside the recent pattern. For a utility team or insurer, that local view helps focus the next day’s monitoring and exposure review.
Where should we focus tomorrow?
Wheeler County · July 15 FireOutlookThe July 15 day-ahead forecast, alongside the July 13–14 outlooks.
At 7:16 p.m., the fire came into view.
Heavy clouds obscured FLARE’s view when Crosswhite was first reported. At 7:16 p.m. PDT, FLARE recorded its first clear-sky detection, with a FireScore of 73 at the selected location.
Fire first reported · Heavy cloud cover
First clear-sky FLARE detection · FireScore 73
Location
Thermal history
Crosswhite area · July 15, 2026 · 7:16 p.m. PDT
The landscape around the fire.
The thermal record follows activity through the day. Vegetation greenness and moisture trends provide the wider context, while FireOutlook shows the day’s forecast danger.
Regional thermal activity
Wheeler County · 24-hour window through July 15, 7:56 p.m. PDT · thermal footprint (km²).
Vegetation and outlook
Wheeler County · NDVI/NDMI daily composites ending July 15 at 03:00 UTC · FireOutlook for July 15–16.
Give the team a clear next step.
A location-based watchlist brings recent activity, observation times, and supporting records together. Teams can investigate a location and share an assessment to guide site checks, exposure reviews, and briefings.
Which locations need follow-up?
Wheeler County · July 15, 7:56 p.m. PDTThe watchlist connects observed activity with a place to investigate. Teams use this evidence alongside official incident guidance.
These are original captures from the FLARE dashboard’s July 2026 historical replay, taken on October 16, 2026. They retain their original numerical scales and illustrate the workflow alongside this retrospective case study.
The dashboard’s percentages count eligible, scored burnable pixels rated Very high or Extreme. The county analysis above measures land area in those categories across the whole county. Each uses its own stated denominator and boundary treatment.
NDVI and NDMI here are daily composites ending at 03:00 UTC, derived from FLARE’s five-minute observations. FireTrack outlines represent observed thermal activity; NIFC/WFIGS perimeters represent the mapped fire extent. The watchlist retains the available reference-evidence status for each track. FireScore describes thermal activity and has its own scale.
FLARE brings local fire danger and changing conditions into a decision workflow. The USGS indicators provide complementary context for the same landscape.
FLARE: a connected local view
Day-ahead FireOutlook identifies areas of elevated danger. Vegetation greenness and moisture trends add context, while five-minute thermal observations help teams follow activity through the day.
For the team: focus monitoring, investigate a location, and prepare follow-up.
USGS: complementary indicators
WFPI describes vegetation flammability. WFSP and WLFP describe fire-growth and large-fire probabilities. Their published grids provide an additional view of conditions across the study area.
For the comparison: examine the same dates and geography using each product’s own measure.
The public FLARE repository includes forecast data and notebooks for exploring maps and time series. Open a location, follow its daily outlook, and use the examples for your own analysis.
The original figures label discovery by its UTC date. July 16 at 00:51 UTC falls on July 15 at 5:51 p.m. in Oregon. Incident times in this story are given in PDT.
THE RESEARCH RECORD
Sources and methods.
The maps and numbers come from the published Wheeler County analysis. The data tables and source records are included with the story.
A day-ahead fire-danger index on a 0–1 scale, presented as six categories from Minimal to Extreme. The underlying index preserves local variation, and the dashboard connects the outlook with vegetation trends and observed thermal activity.
USGS WFPI
Vegetation flammability on a 0–150 scale, combining vegetation and weather information. USGS product definition
USGS WFSP
The percentage probability that an existing one-acre fire grows beyond 500 acres. USGS product definition
USGS WLFP
The probability of a new fire that eventually grows beyond 500 acres, expressed per million reference pixel-days for 1 km² cells. USGS product definition
Each product keeps its own units and color scale across all four dates. WFPI uses 0–150. WFSP and WLFP use the range of values found across the county over those dates.
What was analyzed?
Nineteen FLARE FireOutlook v11 files cover July 3–21, 2026. The map comparison focuses on July 13–16 and uses the Census BAS 2026 boundary for Wheeler County, Oregon (FIPS 41069). The source is the published analysis at commit 5e8e208.
The FLARE files were produced after the July study period in forecast mode. Their issue timestamps describe the schedule used in the historical forecast replay. This story examines the resulting forecasts and their day-to-day changes.
How were the county percentages calculated?
Each grid cell contributes in proportion to the area it covers inside the county. Areas are measured using the EPSG:5070 equal-area projection. The percentage uses the entire 4,445.717 km² county boundary as its denominator and counts stored category codes 5 and 6: Very high or Extreme, beginning at an index of 0.70.
The downloadable table also preserves the earlier analysis fields named high_area, calculated with a strict > 0.70 threshold. The story uses the very_high_extreme_area fields, calculated directly from categories. The rounded county percentages for July 13–16 remain the same.
Valid FLARE values cover 97.762% of the county on these dates. The remaining 99.477 km² is classified as unavailable and remains part of the county total. USGS statistics use numerical measurements, with special land/status codes kept separate.
Each location has a fire-danger category. The categories follow the October 15 analysis revision and use these exact index ranges.
FLARE FireOutlook codes, categories and index ranges
Code
Category
Index range
0
Unavailable
Missing/invalid
1
Minimal
0–<0.40
2
Low
0.40–<0.50
3
Moderate
0.50–<0.60
4
High
0.60–<0.70
5
Very high
0.70–<0.80
6
Extreme
0.80–1
Each range includes its lower boundary. Extreme also includes 1. Category code 0 means unavailable; an index value of 0 belongs to Minimal. Original index values remain available in the optional Index view and source files.
Which dates and hours do the maps cover?
FLARE files carry a 03:00 UTC issue time and a 24-hour valid period from 12:00 UTC to 12:00 UTC the following day. In July, that is an 8 p.m. PDT issue on the preceding local evening and a 5 a.m.–5 a.m. PDT valid period.
The retained USGS Day 1 metadata identifies the archive and ground-condition dates. This comparison follows those daily dates. A same-hour comparison would require additional USGS issue-time and forecast-lead metadata.
Where do the origin and perimeter come from?
The reported starting point comes from NIFC/WFIGS, at latitude 44.803333 and longitude −120.199167. The discovery timestamp is July 16, 00:51:54 UTC, or July 15, 5:51:54 p.m. PDT.
The early footprint comes from the archived perimeter record (object 58164, BurnPeriod2), timestamped July 16, 16:25 UTC. That single recorded footprint provides a fixed geographic reference across all four dates. Daily point values come from the nearest native FLARE cell center, approximately 216 m from the reported starting point.
How are the maps and cell sizes drawn?
Both datasets are drawn from their original cell corners in EPSG:5070. FLARE corners are estimated from adjacent center coordinates using the source notebook’s method. USGS corners come from the retained raster geometry. Each displayed cell keeps its original numerical value and grid footprint.
The geographic background comes from USGS Topo, exported in the same EPSG:5070 projection and displayed in light grayscale. Roads, rivers, place names, and terrain provide present-day geographic context. The forecast colors begin at 72% opacity; the opacity control also offers a full-color view and a geography-only view. The early fire boundary and origin remain visible throughout.
The 0.122 and 1.002 km² figures are median full-cell areas for valid cells intersecting the county on July 13. The 8.2 ratio expresses the difference in area covered by a typical cell. Zoom enlarges those existing cells.
What does this comparison establish?
This case study shows how mapped fire-danger values vary across Wheeler County and change from day to day. It also measures the difference in spatial detail between the FLARE and USGS grids.
Measuring comparative forecast accuracy would require a larger set of fires and control locations, matched issue and valid times, and a shared measure of performance. Those tests would establish how the additional spatial detail affects predictive skill.
Credits, files, and reuse
FLARE research and data: Manju Smriti; BRIDGE – Research and Policy Foundation; ARTH GEO Intelligence. Comparison products: U.S. Geological Survey. Incident origin and perimeter: NIFC/WFIGS. County boundary: U.S. Census Bureau.
The seven-day replay uses FireOutlook2.0-hrrr1-nbm2to7-1: NBM Days 2–7 (experimental) and HRRR Day 1. Stored categories use the product's native thresholds; county percentages are area-weighted over all county pixel centers, including unavailable pixels. Scored coverage is 99.07%. The earlier day-ahead analysis uses a different category scale and reports Very high-or-above area; do not combine its labels or percentages with this replay.
The seven-day maps draw native longitude/latitude cells using the source notebook's corner estimation, without interpolating category values. Present-day USGS Topo context is exported in EPSG:4326 and placed at its geographic bounds; forecast colors use 72% opacity. The earlier comparison maps retain their original EPSG:5070 geometry.
The retained audit checked weather publication and vegetation timing; historical receipt times for the NIFC-proximity prior were not fully verified. Quality flags remain available in the source files, including the prior flag where applied. The July 16–17 fire outlines are later observed reference geography, not forecast perimeters. This replay does not predict ignition or spread or demonstrate increased accuracy.
Seven-day source data, manifest, and NetCDF: GitHub data bundle; executable source notebook. Summary CSV and reference geography are downloadable above. FLARE source data carries CC BY-NC-ND 4.0. The provenance file retains source URLs, hashes, and original-figure hashes. See the included README for reproducibility and local use.
Map images need their accompanying files. Keep the assets and data folders beside this HTML file, then reload.