By Jesse Schrage, WP1 Researcher

Fragmented present, forecasting futures

In April 2025, Francisco Orozco-Meléndez (aka “Paco”) and I spent 10 days in Madagascar’s capital city, Antananarivo with the mission to better understand how anticipatory action (AA) is coordinated in the country. And while we came with the intent to map the temporal dimensions involved in coordinating disaster response, we found, quite quickly, that it became impossible to neglect the political dimensions involved in this work.

Forecast-based anticipatory action promises to intervene before the disaster strikes—delivering cash, relocating vulnerable households, even staging emergency supplies days ahead of impact. It is a powerful idea, and in a country like Madagascar, where tropical cyclones have caused more than $2 billion in damages over the past decades, it carries real, life-saving potential.

But as with many things in disaster governance, the devil is not just in the data—it’s in the definitions. Over the past five years, AA has taken root in Madagascar, seeded by international agencies, NGOs, and the Malagasy government itself. The Bureau National de Gestion des Risques et des Catastrophes (BNGRC), the country’s national disaster authority, has sought to provide some central coordination. But the implementation landscape remains a work in progress.

The Malagasy Red Cross, Save the Children, OCHA, and a growing host of others each use their own forecast models, threshold criteria, and institutional mandates to determine when early action should be triggered. One actor’s “imminent risk” is another’s “still monitoring”. One agency may release funds when winds reach 118km/h; another waits for 166km/h. While the science behind each trigger may be rigorous, their divergence creates operational noise.

The BNGRC makes efforts to unify these fragmented systems, convening coordination platforms and technical working groups. Yet aligning these triggers has proven tricky. This is not merely a bureaucratic hiccup. It is a reflection of deeper institutional dynamics—of mandates that don’t align, accountability structures that point in different directions, and conceptions of risk that aren’t just technical, but profoundly political.

A trigger is not just a switch—it’s a story

Madagascar’s cyclone season doesn’t arrive quietly. It begins with a humid hush, a clench in the sky, and the familiar throb of anxiety in coastal towns. Radios chatter warnings; house roofs are checked. And somewhere, in a brightly lit coordination room, a map glows, a model runs, and a threshold is crossed. Someone presses send.

To the untrained eye, a trigger is a technical object: a threshold in forecast probability that, once breached, releases a pre-agreed action. But scratch the surface and you’ll find that triggers are best understood as what science studies scholars call boundary objects—entities that coordinate across institutional divides while meaning different things to different actors.

For an international humanitarian agency, the trigger is an actuarial decision rule, designed to maximise cost-efficiency and justify donor expenditure. For a national government, it may signify a moment of sovereign accountability. For a community, it is a signal—perhaps welcome, perhaps premature—that help is on the way.

And here’s where things get messy. Each actor operates under a distinct institutional logic. NGOs are often under pressure to demonstrate “value for money” and act swiftly. National agencies may be concerned with reputational risk, political implications, or resource constraints. Triggers are negotiated outcomes, not neutral tools. They reflect assumptions about what counts as risk, who gets to define it, and when it becomes urgent enough to act.

Social scientists have long argued for a notion of risk that is not a singular, technical condition, but a situated social construct, entangled with history, identity, and power. Climate risk, in particular, is shaped not just by exposure or probability, but by how different actors interpret the world. This insight is particularly relevant in Madagascar: in some communities, risk is read not in millimetres of forecast rainfall, but in the smell of the soil, the shape of clouds, the compound effects of small storms, or the memory of 2004. These lived knowledges often sit uneasily alongside the probabilistic models used by international AA systems.

Too often, AA initiatives arrive with pre-set thresholds derived from models built elsewhere. Local knowledge is, at best, consulted; at worst, sidelined. Even well-intentioned efforts to include local actors can reproduce extractive dynamics—where data is taken but control remains firmly with the forecasting centres and funding agencies. The result is anticipatory action that may be technically precise but socially tone-deaf. Triggers are activated, but trust is not. Aid arrives early—but not always where or how it’s most needed.

When coordination means compromise

Studying attempts to harmonise triggers is, in many ways, a case study in the politics of coordination. Harmonisation requires actors with different mandates, funding streams, and operating logics to agree not only on what constitutes a credible forecast, but on what kinds of uncertainty are tolerable and what forms of action are appropriate.

This is not a matter of science alone, but rather, it is a negotiation over authority, and legitimacy. Should the trigger be based on the European Centre for Medium-Range Weather Forecasts? Should it incorporate local indicators of vulnerability? Who validates the model, and who takes responsibility if the action is launched, and the cyclone veers off?

As Forsyth (2023) reminds us, risk governance is not about controlling uncertainty; it is about navigating difference. So, what would a more inclusive, critical approach to anticipatory action look like?

First, it would treat triggers not as static thresholds, but as co-produced artefacts—emerging from dialogue between scientists, practitioners, and communities. This means involving local actors not merely as data sources, but as co-authors of the trigger logic itself.

Second, it would recognise the plurality of risk knowledges, and create space for them to be translated—not flattened—into operational frameworks. This could involve integrating traditional weather indicators, community risk mapping, and local institutional capacities into trigger design.

Third, it would rethink the very purpose of harmonisation. Rather than seeking a single, universal trigger, coordination might aim for interoperability—a modular approach where actors share data, understand each other’s thresholds, and commit to aligning responses when appropriate, without erasing their differences.

Looking forward: anticipation with accountability

The promise of anticipatory action lies not only in its speed, but in its ability to shift the politics of disaster response—from reactive crisis management to proactive, dignified protection. But to fulfill that promise, we must move beyond technocratic fixes and confront the political questions AA inevitably raises.

Who defines risk? Who decides when enough is enough to act? Who is accountable when anticipation misfires? In Madagascar, these questions are being asked—not always answered, but asked—in meetings, in communities, and, yes, in those coordination rooms with the glowing maps. Perhaps, then, the most important trigger is not a threshold in a spreadsheet, but a shift in mindset: from anticipation as a technical intervention to anticipation as a social contract.

Reference:

Forsyth, T. (2023). Climate Change Adaptation and Resilience: sociotechnical and knowledge dimensions. In Z. Baker, T. Law, M. Vardy, & S. Zehr, Climate, Science and Society (1st ed., pp. 198–206). Routledge. https://doi.org/10.4324/9781003409748-32