Managed fire in the NT, 2000–2025

Savanna projects earn credits two ways. The emissions-avoidance methods (2013, 2015, and a 2018 update) credit the greenhouse gases avoided by shifting fire early. The separate 2018 “Sequestration + Emissions Avoidance” method also credits carbon stored in dead organic matter — and, unlike avoidance, carries a permanence obligation (25 or 100 years). Existing projects were never required to switch, so uptake is marginal: of 37 NT savanna projects, 36 are avoidance-only and exactly one uses the sequestration method. So the permanence and dead-matter questions that dog the sequestration method barely touch the NT — the estate is a fire-timing scheme, which loops back to the area/season charts above.

Source: Clean Energy Regulator ACCU project register (the same layer that defines the “managed” boundaries above), fields METHOD and Perm_Per. Note: a “Cook et al. (2016)” citation describing a 2018 update is chronologically off — pin the exact source before quoting it.

reproducible from the public NAFI fire-scar archive and the ACCU carbon-project register.

1. Fire data. NAFI monthly fire-scar shapefiles (250 m), one per year 2000–2025, from firenorth.org.au. Each polygon is a burnt patch tagged with the month it burnt.
2. NT land selection. Each year's fire is clipped to a longitude/latitude box (129°E–138°E, south to ~26°S, open to the coast in the north). NAFI's own region attribute is unreliable (absent before 2013, compound strings like “(a) NT (b) WA” in 2015–21, mapping-block codes in 2025) whereas the NT's borders are meridians and a parallel and there is no fire at sea, so the box selects NT land exactly and identically every year.
3. Projects. ACCU project-area polygons where the method type begins “Savanna” and status is Declared. For each year only projects registered on or before that year are counted (a time-varying footprint), so managed area is never credited before a project existed.
4. Managed fire. The NT fire layer intersected with the union of that year's active project boundaries; everything outside is “unmanaged”.
5. Season. Month ≤ July = early dry season; August onward = late dry season — the damaging wildfire window that early burning is meant to pre-empt.
6. Measurement. All areas are computed in an equal-area projection (GDA94 Australian Albers, EPSG:3577) and reported in km²; geometries are repaired with a zero-width buffer first.
7. The two ratios. Managed share = managed ÷ all NT fire. Late-season managed share = managed-late ÷ all late-season NT fire. Note these have different denominators: the two lines in chart 1 aren't slices of one whole.

These are lifted directly from the analysis script's own header. They decide how the figures should and shouldn't be read.

• Denominator is all NT land fire (a lon/lat box: 129°E–138°E, north of ~26°S). That's the "all-NT" figure: a floor. The method-eligible savanna-zone denominator (a rainfall mask) would be smaller, so the true managed fraction of eligible country is higher.
• NAFI 250 m under-detects small, patchy early-season fires. That's significant, cos it describes exactly the management fires. So managed area, and every managed fraction, is a lower bound. TODO: Validate against Sentinel-2 on sample project-years before quoting hard numbers.
• Project footprint is "current" boundaries (STATUS = Declared, registered ≤ that year). The register lacks historical revocation dates, so earlier years are slightly over-counted.
• Early/late boundary is month ≤ July = early. The real EDS/LDS cut-off (~31 Jul) is method- and zone-specific; this is an approximation based on opinion.
• Pre-2012 zeros are real, not missing data. No ACCU savanna projects were registered before 2012 (earliest registration 2012; 1 project that year, 3 in 2013, 85 by 2025).

This is a personal look at the physical fire picture in the NT: how much country burns, when in the season, and how much falls inside registered savanna carbon projects. A conversation-starter for people who care about fire and Country; not a formal evaluation, and certainly not peer-reviewed. I'm a greenie coder, not a scientist.

It measures area and season, not emissions. The savanna carbon method is credited in tonnes of CO₂-e from a seasonal emissions model. A project can burn the same area and still abate by shifting fire early. So this deliberately looks at a different axis than the one the credits are counted on; it is not a verdict on whether the method works on its own terms.

My view — and it is purely personal opinion: fire's environmental ledger — biodiversity, the age-structure of long-unburnt habitat, severity, cultural burning, standing carbon — doesn't collapse into one tradeable number. A scheme optimised for emission credits is optimised for a single axis, and whether that serves the others is genuinely contested: the “bioperversity” concern points one way, co-benefit studies the other. I think the single-number frame hides more than it shows, that's why this page looks at the fire itself.

There's nothing “natural” about fire in the NT these days. Displaced traditional fire managers and invasive African grasses have transformed the regime. Gamba grass in the north swaps a 1–3 t/ha native fuel bed for up to 30 t/ha of 4 m grass, and buffel grass (an EPBC-listed Key Threatening Process) drives a hotter, self-reinforcing fire cycle across the arid centre and beyond. So an NT-wide view of harm isn't a category error. The whole Territory's fire is altered, not just the savanna zone.

Scope is deliberate: all-NT fire (the question is NT-wide harm, not project performance), and NAFI's 250 m mapping under-detects small early fires, so managed-early here is a floor.