During the cyclone season in early 2025, ACACIA ran a series of weekly forecasting testbeds bringing together Madagascar’s national meteorological service (DGM), ACACIA scientists, and humanitarian stakeholders, including Croix-Rouge Malagasy (CRM). These hour-long online sessions created a regular space for joint discussion of current tropical cyclone forecasts, interpretation of different forecast products, emerging needs from users, and reflection on how earlier forecasts had performed in practice.

The testbeds were deliberately designed as a “safe-to-fail” environment, where forecast producers, researchers, and users could ask questions, explore uncertainty, and challenge assumptions without the pressure of operational decision-making. Discussions focused in particular on sub-seasonal forecasts (two to four weeks ahead), a timescale that is increasingly relevant for preparedness actions but still difficult to interpret and use. Over the course of the season, participants compared DGM’s operational products with regional and global guidance (including ECMWF products), explored how cyclone risk is assessed, and identified where communication, timing, or framing could be improved for humanitarian decision-making.

We have now synthesised the main lessons from these sessions into a Testbed Handbook for Madagascar, which is available on our Resources page. The handbook documents how the testbeds were organised, the co-production principles that guided them, the forecast products discussed, and key concepts needed to interpret sub-seasonal cyclone information. It is intended both as a practical reference for participants and as a foundation for future testbeds, supporting more consistent discussions and a shared understanding across institutions.

This is only a first iteration. The handbook will be updated following the next cyclone season, when the testbed meetings resume in January. Future updates will reflect new forecast products developed within ACACIA, deeper integration with DGM’s operational procedures, and continued feedback from humanitarian partners as the testbed approach evolves.

Illustration by ChatGPT