From Early Warning to Early Action: Why Anticipatory Finance Belongs at the Heart of COP31

*We can see most climate disasters coming. The question COP31 has to answer is whether the money can move before they arrive, and whether the trigger fires for the communities the forecast keeps missing.*

In the 2024 winterisation season in Afghanistan, my team produced a set of maps that, on their face, looked unremarkable: snow cover, snow depth, precipitation and temperature, each compared against prior years. To a casual reader they were just shaded rasters. To the clusters preparing for a hard winter, they were a decision: where to pre-position, before the first households were cut off. That is the entire logic of anticipatory action in one product. You act on the forecast, not on the funeral. The hazard had not happened yet. The point was to move while it still hadn't.

As attention turns to Antalya, much of the climate-finance conversation understandably centres on the *volume* of finance. I would gently add a second question that tends to get less airtime: its *timing*. In disasters, when money arrives can matter as much as how much, because acting early often changes the entire course of a crisis.

The case for acting early is, by now, almost a cliché in the resilience world: every dollar spent before a shock is worth several spent after. The evidence base for anticipatory action (pre-positioning supplies, releasing cash, evacuating, reinforcing) consistently shows it reduces suffering and cost relative to waiting for the disaster to certify itself. I have written before that anticipatory cash is one of the highest-leverage instruments we have. Predictable, dignified, and fast.

And yet the dominant model of climate and humanitarian finance remains reactive. Money flows after the flood crests, after the harvest fails, after the displacement. We have built an entire architecture optimised to respond to disasters we could have seen coming weeks out. The reason is not ignorance of the economics. It is that pre-arranged, trigger-based finance requires three things that are genuinely hard to assemble at once: a credible forecast, a pre-agreed trigger, and money that is already committed to move when the trigger fires. Get any one of them wrong and the system either fails to fire or fires into the void.

COP31 is where this stops being a humanitarian side-conversation and becomes central to the finance debate, because the NCQG, the Loss and Damage Fund, and the Early Warnings for All (EW4All) initiative all converge on the same operational question. Can the system act in advance?

Here is the part that most finance discussions skip, and the part I cannot stop thinking about.

A trigger, the pre-agreed threshold that releases anticipatory finance, is built from historical data. You set it by looking at the record: how often a hazard of a given magnitude has occurred, and what it did when it did. Rainfall thresholds, river-discharge levels, vegetation-stress indices like the NDVI and VHI products I ran monthly in Afghanistan, SPI drought indicators, all of them are calibrated against what the data says is normal and what counts as alarming.

Which means the trigger inherits every bias in the underlying record. And as I have argued repeatedly, the underlying record systematically under-counts the small, recurrent, extensive-risk events that grind down the poorest communities. When those events are missing from the historical baseline, the model underestimates how often the hazard actually occurs. The trigger gets calibrated too high. And the pre-arranged finance, designed precisely to protect vulnerable communities, fails to fire for the very people it was meant to reach, because on paper their disaster never quite qualifies.

I have watched the converse work, too. In Ethiopia, the cash programming I supported leaned on climate-risk-informed targeting precisely because we had built the analytical layer to see the shock developing, combining minimum-expenditure-basket data, market assessments, and hazard indicators so the Cash Working Group could position ahead of the crisis rather than chase it. One delivered CERF multi-purpose cash operation reached roughly 185,000 people, with the overwhelming majority reporting livelihood improvements. That worked because the data was good enough to justify acting early. The lesson generalises and it is uncomfortable: anticipatory action is only as equitable as the data its triggers are built on. Improve the forecast without improving the loss record, and you simply automate the existing blind spots faster.

The Early Warnings for All initiative, the UN's drive to cover every person on Earth with multi-hazard early warning, is one of the most important things happening in my field, and it will feature prominently on the road to Antalya. But I want to push against how it is often framed, because the framing shapes the funding.

EW4All is usually presented as a technology and infrastructure programme: more sensors, better forecasts, faster dissemination. All necessary. But a warning that no one can act on is not protection. It is information. The pillar that consistently lags is the one that turns the warning into a response: the pre-arranged finance and the institutional readiness to spend it. I saw this directly: in Afghanistan, the forecasts and the bilingual Dari/Pashto warning products were the *easy* part. The hard part was wiring them to clusters and to money, so that a warning triggered a pre-agreed action rather than a meeting.

So my argument for COP31 is that early warning and anticipatory finance are the same system, and they have to be funded as one. A warning without committed, fast-moving money is a smoke alarm in a building with no exits. The disaster funds being operationalised, including the FRLD's faster-disbursing windows, should be explicitly designed to support pre-arranged, trigger-based release, not only post-event reconstruction. The "last mile" of EW4All is not a transmission tower. It is a budget line that moves before the hazard does.

There is a particular category that exposes the failure most sharply: slow-onset hazards. Drought. Heat. Creeping food insecurity. These are the disasters we can predict furthest in advance, sometimes months, and they are precisely the ones our reactive systems serve worst, because there is no single dramatic moment that triggers the cameras and the cheques.

My clearest memory of this is the 2020 desert locust upsurge in the Horn of Africa, the worst in a generation. Working IM for the FAO-led Agriculture Cluster in Ethiopia during the second swarm wave, the analytical task was to integrate swarm-tracking, vegetation and precipitation data with food-security projections so the response could get ahead of the damage to agro-pastoral livelihoods. The hazard was forecastable. The breeding areas were mappable. The window to act before the swarms matured was real. Whether that window got used came down, every time, to whether finance could move on a forecast rather than a body count.

Slow-onset disasters are the strongest possible argument for anticipatory finance: long lead times, clear indicators, enormous avoidable losses. And they are chronically under-served, because they do not fit the reactive funding cycle. If COP31 wants a concrete test of whether it is serious about anticipatory action, it is this. Does the finance architecture reward acting on a drought forecast in month one, or does it still wait for the famine to be declared in month nine?

I am not asking COP31 to discover anticipatory action. The Anticipation Hub, the forecast-based financing community, and a decade of pilots have already proven it works. I am asking the finance negotiations to stop treating it as a niche humanitarian technique and start treating it as a design principle for climate money.

Concretely, three things. First, the climate and loss-and-damage funds should build in pre-arranged, trigger-based disbursement windows: money that is committed in advance to release on a forecast, with the speed that anticipatory action requires. Second, EW4All financing should be evaluated on the whole chain, all the way to the committed money and institutional readiness at the end of it, not just the forecasting hardware at the front. And third, the one closest to my own work: investment in the historical loss data that triggers are calibrated against, because a trigger built on a biased record protects the already-visible and abandons the already-overlooked.

We can see most of these disasters coming, and have been able to for some time. The unfinished work is not prediction. It is the plumbing that lets money move at the speed of a forecast rather than the speed of the disaster itself. On the road to Antalya, I will be watching the anticipatory-action agenda for one thing. When the warning fires, does support actually move, and does it reach the people the data most often leaves out?

A forecast tells you what is coming. Anticipatory finance decides whether knowing makes any difference. The whole point of seeing the disaster early is to spend before it arrives. Everything else is just a very well-documented response.