Tecnobosque Geospatial Methods · Research note · Prepared 5 November 2026 · Reviewed 6 November 2026 · Reproducible analytical workflow
Research objective
Given two polygon datasets describing the same named wildfire, measure where their boundaries agree and disagree. The result should remain reproducible after the source maps, timestamps and analytical choices are independently checked. This protocol is designed for QGIS and applies conceptually to Global Mapper and other GIS environments.
Do not start with GIS operations. First confirm that the datasets represent comparable events and products. An MTBS burn scar is not automatically equivalent to a NIFC operational perimeter: MTBS expressly documents why those geometries can differ. MTBS
Required inputs and metadata
For each source layer, retain: original source URL, incident identifier, source organisation, record/version identifier, observation date(s), release or revision date, retrieval timestamp (UTC), geometry type, data CRS, target CRS, units, and uncertainty notes. Preserve the original downloaded file and store a checksum if possible. For an incident complex, state how component incidents have been combined or excluded.
Step 1 — Prepare both polygons
- Add the downloaded polygons to QGIS. Inspect CRS information; never assume the layer’s declared CRS is correct merely because an overlay looks plausible.
- Filter to the same incident and check for duplicate or overlapping features within a source. Preserve IDs and a traceable copy of the raw dataset.
- Reproject each layer to a suitable equal-area projected CRS in metres. Explain the regional choice. For European analysis, ETRS89 / LAEA Europe (EPSG:3035) may be suitable; for other regions choose an appropriate local or regional equal-area system.
- Run geometry validity checks. If fixing invalid polygons is necessary, keep the original, record the correction and confirm whether the footprint or area changed.
- Dissolve features within each defined comparison set (only after verifying which records belong together). Do not dissolve unrelated fires or independent dates.
Step 2 — Construct the overlay products
Use Processing → Vector overlay in QGIS on the dissolved polygons:
Intersection(A,B)produces common mapped area, I.Difference(A,B)produces the portion present only in A, A-only.Difference(B,A)produces the portion present only in B, B-only.Union(A,B)provides the area covered by at least one product, U. For sum-of-area use a dissolved nonoverlapping union to avoid double counting.Symmetrical difference(A,B)provides the disagreement footprint, S.
Recheck geometry validity after overlays. Boundary slivers may be legitimate small differences or technical artifacts; do not erase them without documenting a tolerance and a reason.
Step 3 — Calculate seven indicators
All area fields below are in hectares, computed from planar metre-based geometry after correct reprojection (area($geometry) / 10000 for per-feature area in QGIS). Where an output contains multiple features, sum their areas. Avoid confusing QGIS’s ellipsoidal project-dependent $area with the planimetric area($geometry) function when reproducing a calculation.
| Metric | Formula | Research meaning |
|---|---|---|
| Area A | Area(A) |
First product’s footprint |
| Area B | Area(B) |
Second product’s footprint |
| Common area I | Area(A ∩ B) |
Agreement footprint |
| A-only | Area(A \ B) |
Area exclusively assigned by A |
| B-only | Area(B \ A) |
Area exclusively assigned by B |
| Symmetric difference S | A-only + B-only |
Total mismatch footprint |
| IoU | I / (I + A-only + B-only) |
Scale-independent overlap score, if union > 0 |
Also report signed net difference Area(A) − Area(B) and relative difference using a transparently chosen denominator. These optional summary metrics must never replace spatial overlay results.
Independently checkable synthetic calculation
This is an artificial geometry exercise, unrelated to any historical fire. Consider two overlapping rectangles measured in an equal-area metre coordinate system. Rectangle A spans x=0–4000 m and y=0–3000 m. Rectangle B spans x=466.6667–4233.3333 m and y=0–3000 m. Rounding of the printed coordinates can introduce tiny floating-point deviations; the underlying construction uses x=1400/3 m and x=12700/3 m.
| Calculated quantity | Expected result |
|---|---|
| Area A | 1,200 ha |
| Area B | 1,130 ha |
| Intersection | 1,060 ha |
| A-only | 140 ha |
| B-only | 70 ha |
| Symmetric difference | 210 ha |
| Union | 1,270 ha |
| IoU | 0.8346456693 (83.46%) |
The example demonstrates the diagnostic limitation of headline acreage: a 70 ha net difference coexists with 210 ha of nonoverlapping footprint. If GIS outputs differ substantially from these reference values, revisit projection units, overlay processing, and shape repair before applying the method to real data.
Step 4 — Attribute, explain and archive
Give each discrepancy a possible cause code: T (time), D (definition), R (resolution), G (geometry), V (source version), I (incident identity), U (unresolved). Preserve source links and screenshots or downloaded data licenses as appropriate. Annotate the comparison with both observation and publication times. If only the publication time is known, do not infer exactly when the mapped pixels were observed.
Save a GeoPackage with at least source_A, source_B, intersection, a_only, b_only, symmetric_difference, and a nonspatial metadata table. Save the QGIS project file and analysis instructions. Export a one-row CSV containing the seven metrics and the date/source IDs. Check that A ≈ I + A-only, B ≈ I + B-only, and U ≈ I + A-only + B-only within a disclosed numeric tolerance.
Common failure modes
- Comparing a current incident perimeter against a final burned-area polygon and interpreting the difference as new growth.
- Mixing degrees and metres, or applying an unsuitable area calculation.
- Pairing multiple incidents by a similar name without validating IDs and locations.
- Losing polygon holes, unburned islands or multipart segments while repairing geometries.
- Inferring causation or mapped burn severity purely from overlapping polygons.
- Publishing metrics with excessive decimal precision inconsistent with input resolution.
The protocol is a research analysis, not official incident mapping or an emergency operations product.
Related Tecnobosque research: Why perimeters disagree; evidence timestamps; burn severity; Fire archive.