Developmental wildfire morphology

Fires have developmental histories.

Final size tells us how big a fire became. Fire VASE captures how it got there.

Fire VASE turns ordered, dated growth observations into comparable developmental trajectories. It reconstructs the observed history first, then tests weather and other possible explanations against that common response.

See how it works Explore the findings Reproduce the research

278,569 FIRED events in the reanalyzed source archive

10,246 histories with ≥3 consecutive daily observations in the primary morphospace

87,944 exact next-calendar-day transitions used in the recent-state analysis

The problem

The endpoint hides the pathway

Two fires can finish at a similar size after a similar span of time while allocating growth very differently—early, steadily, late, or in repeated pulses. Endpoint summaries collapse those sequences. Fire VASE keeps the order visible so that the developmental pathway itself becomes something scientists can compare and explain.

What to notice. The top row records when observed growth was added; the lower row translates that cumulative sequence into a VASE. Gray bands are unobserved dates, not zero-growth days. The primary comparative analysis therefore uses only consecutive histories. N = 10,246 for the primary morphospace; the four fires shown here are examples.

Read the technical figure caption

The idea

Reconstruct first. Explain second.

Weather, final area, duration, and observation count do not define the primary shape coordinates. Fire VASE first standardizes the allocation of observed growth through relative developmental time. External variables are added only afterward, making representation and explanation separate, testable steps.

01Observed polygonsDated FIRED growth

→

02Ordered growthHistory reconstructed

→

03Standardized trajectoryComparable relative time

→

04Fire VASEDevelopment made visible

The findings

What the evidence supports

01 · Representation

Fires follow different developmental pathways.

Broad gradients distinguish earlier from later allocation and concentrated from more distributed growth. They are continuous coordinates, not established fire types.

5 axes · 89.4% of standardized shape variance 02 · Ordering

The order of growth contains information.

Shuffling each fire’s same observed increments changes front-loading, detected pulses, reactivations, and morphospace coverage.

0.541 vs 0.500 mean first-half allocation 03 · Prediction

Recent state explains much more than weather alone.

Weather matters, but in these held-out analyses it adds little next-day predictive skill beyond recent observed fire state.

0.448 R² state baseline · +0.005 weather 04 · Open questions

Similar weather can accompany different histories.

Matched mismatches are study candidates for fuels, terrain, active-edge conditions, ignition, suppression, and observation uncertainty—not proof of a missing causal mechanism.

49.7% vs 50.4% observed vs conditional-null mismatch

Confidence

We tried to break the result.

The analysis was challenged with stricter observation thresholds, shuffled and synthetic null histories, alternative compositional geometry, blocked prediction, day-specific climate attribution, subgroup tests, independent data checks, deliberate corruptions, and a full software test suite. Broad developmental gradients and informative ordering persist; exact neighborhoods, dimensionality, weather associations, and mechanistic interpretation remain qualified.

6 / 6real-data validation modules pass

152repository tests pass; 2 intentionally skipped

0.969distance-rank agreement at ≥7 observations

See how Fire VASE was challenged →

Choose your evidence route

Go from the claim to the record

See the evidenceWalk through every main figureRead the panels, comparisons, claim boundary, and evidence trail. See how we challenged itStart with the failure testsSee what would have weakened each conclusion—and what happened. See how AI was usedInspect the accountability recordSeparate AI assistance, executable outputs, and human scientific responsibility.

Meaning

A common response for the next layer of wildfire science

Fire VASE does not replace weather, fire-behavior physics, or endpoint summaries. It supplies a transparent response against which those explanations can be tested. The next scientific step is to align dated growth with process-specific measurements—especially fuel continuity, terrain, active-edge conditions, ignition context, and suppression—while improving observation-depth and burn-date uncertainty.

01

How can similar fires develop differently?See the idea →

02

We reconstructed their developmental histories.See the approach →

03

Development occupies broad, reproducible gradients.See what we found →

04

Measured weather adds little beyond recent state.See the evidence →

05

We tested where the result holds—and where it does not.See the validation →

Open evidence

Inspect every layer

Data, schemas, four Jupyter notebooks, analysis and figure scripts, validation outputs, tests, provenance records, manuscript sources, and historical corrections remain available beneath this narrative.

Reproduce this research → Read the paper Read the AI accountability record