“Leaving early means leaving the area before a fire starts, not when you can see flames or smell smoke.” That sentence belongs to CFA Chief Officer Jason Heffernan, and it is the most exact description available of what Australia’s central bushfire safety message asks a household to do. It does not ask anyone to leave when things deteriorate. It asks them to leave on a quiet morning, from a house under no visible threat, on the strength of a forecast.

Put that way, the difficulty is obvious. Every cue a person would ordinarily use to judge danger arrives after the moment the advice nominates, and by then it has stopped applying to them.

Victoria’s post-season community survey, reported at the start of September, looked like it had put a number on that gap. CFA found that 88 per cent of respondents agreed or strongly agreed that leaving early was the safest option, while 81 per cent “indicated they were unlikely to leave until fire was threatening the landscape”. Twelve per cent said they would stay and defend. CFA headed the release “More than 80 per cent would wait too long to leave during bushfire”.

That pairing is the kind of finding a preparedness campaign gets built on, and if it were what had been measured, the constraint would have moved from information to something far harder to shift. It is a good argument, and CFA’s own numbers do not support it.

The same annual survey reported 30.9 per cent in 2023 who would wait for authorities to tell them what to do, about 30 per cent delaying in 2024, and 31 per cent in 2025 who “would delay leaving until threatened by a bushfire”. A jump from 31 per cent to 81 per cent in twelve months is not a change in the Victorian public, because those earlier figures come from a forced-choice item whose options sum to roughly a hundred. This year’s figure reads instead as a likelihood scale, which is perfectly compatible with intending to go: a person can believe leaving early is safest, plan to go on a Catastrophic forecast, and still rate themselves unlikely to move before there is fire in the landscape. The two numbers answer structurally different questions.

The internal arithmetic points the same way. If 88 per cent believe leaving early is safest, 81 per cent will not move before fire is in the landscape, and 12 per cent would stay and defend, then at most about 7 per cent both hold the belief and would act on it. That residual is not believable. The planning figure moved just as far: CFA reported a year earlier that 40 per cent of Victorians in bushfire-prone areas had no plan, so about 60 per cent did, and this year’s release says 93 per cent. Populations do not move thirty-three points in a year. Questions do.

The instrument would settle it, and the instrument is not published. CFA’s site carries no sample size for the 2026 wave and no fieldwork dates, and it has released no report. Earlier waves ran to about 600 respondents, which tells a reader nothing about whether this year’s questions match last year’s. The honest description of the 81 per cent is that it is a quoted attitude from an unpublished instrument, reported by the agency that commissioned it, and discontinuous with that agency’s own series. “Wait too long” is the headline writer’s gloss on that attitude.

Here is why the discontinuity matters. The claim the number is used to make happens to be true, and it has been sitting in peer-reviewed Australian research for five years.

Whittaker and colleagues interviewed 113 people and surveyed 549 households after the 2017 New South Wales fires. Most people found the warnings they received easy to understand and useful, and many still did not respond in the way fire services intended (Whittaker et al., 2020). They waited until they were threatened before evacuating. The mechanism those authors identify is confirmation: people look for proof that a threat is real before acting, most often to avoid the cost of an evacuation that proves unnecessary. Three messages in bushfire warnings are not personally meaningful for many people, and one of them is leave early.

Strahan and Gilbert then reviewed the international literature twice in 2021, first on delayed evacuation and later on leaving early (2021a, 2021b). At least a quarter of people delay the decision, and the weight of the evidence puts it nearer 45 per cent. People who intend to leave early are frequently unable to name a specific event that would start them moving, so they fall back on proximate cues, which in practice means smoke and flame. Both arrive at the worst possible time. Their recommendation follows directly from that finding: agencies should help households identify a trigger and write it into the plan.

What is new this spring is a statistic, and it is the weakest thing in the room.

The season’s framing document is AFAC’s Seasonal Bushfire Outlook for Spring 2026, published in August. It reports “a heightened risk of fire across northern and southeastern New South Wales, south-east Queensland, northeastern Tasmania, parts of Western Australia (WA), South Australia and Victoria, and large parts of the Northern Territory”. AFAC chief executive Rob Webb points to “the need for communities to understand their local conditions and prepare early”, and the Outlook tells householders to maintain their properties and update their bushfire survival plans.

That is the trigger message, near enough. It is also precisely what Strahan and Gilbert found people cannot convert into action without a named trigger, and what Whittaker showed people understand perfectly well without acting on. The agencies are saying the right thing. The research has held since 2020 that saying it is not sufficient.

The Outlook and its media release carry no guidance on when to leave, a curious silence in the document that frames the season.

A second signal is being read as more formal than it is. People still say the Bureau declared an El Niño, and it no longer has a mechanism for declaring one. The ENSO Outlook page that housed the Watch and Alert dial is now headed “The ENSO Outlook is no longer available”, above a notice reading “We are streamlining some climate pages including removal of the ENSO Outlook.”

What the Bureau publishes now is a monitoring statement, and the current one is less comfortable than a dial position ever was. El Niño is “firmly established”, with indices “consistent with a strong El Niño, with further intensification likely during spring” (BoM, 2026). Relative Niño3.4 sat at +2.45°C in the week to 30 August, against a threshold of +0.80°C. That signal does not mean a dry spring everywhere, since the Bureau’s own October to December outlook has above-median rainfall for most of Western Australia and western South Australia.

Both cases run the same way: a number keeps the authority of a formal process after that process has quietly changed. In the survey case, the question behind the number was rewritten. In the climate case, the dial behind the number was switched off. Neither change was concealed, and neither travelled with the number it produced, so readers go on treating both figures as though the machinery behind them still stood.

The question this season actually poses is not whether Victorians believe in leaving early. Eighty-eight per cent of them have already told somebody they do. The question is what would have to happen, and on which day, for you to put the dog in the car and drive to your sister’s place in town. A plan can name that day in advance. Smoke on the ridge cannot.

References