Benchmarking Existing Fire Models
This page focuses on how scaling diagnostics can be used to evaluate operational or physics-based fire models.
Why benchmarking matters
Many fire models are assessed mainly through local spread rates, final extent, or behavior under specific conditions. This project adds another question:
Do the simulated fires exhibit the same large-scale perimeter geometry as observed fires once definition scale and measurement scale are made explicit?
Candidate benchmark outputs
- perimeter growth exponent
- area-perimeter scaling
- boundary roughness or fractal dimension
- presence or absence of a regime-specific scaling interval
L_d(ε)curves for observed and simulated perimeters under aligned boundary definitions- sensitivity of inferred geometry to sensor resolution, temporal aggregation, and smoothing choices
What this contributes
This gives wildfire modeling a new evaluation language based on emergent geometry, not only on local behavior rules. It also reframes validation as a curve-matching problem across scales rather than a point comparison at one chosen perimeter definition or map resolution.