National Anticipatory Action Framework for Tropical Cyclones in Vanuatu
The framework was endorsed by the Emergency Relief Coordinator on 15 October 2026 for the 2026/27 and 2027/28 cyclone seasons. This page shows the historical activations of its trigger from 2006 to 2025, as listed in Annex 2 of the framework document, and the analysis behind the trigger (section III.2).
The trigger: at least 15,000 people exposed to winds of 64 knots (about 119 km/h) or more, in the RSMC Nadi forecast 72 to 120 hours ahead (readiness) or the VMHS forecast 24 to 72 hours ahead (action), counted in Shefa, Sanma and Tafea; or in the VMHS observed track, anywhere in Vanuatu (observational). Full definition.
Storms near Vanuatu from 2006 to 2025, as listed in Annex 2 of the framework. Red cells are above the threshold: at least 15,000 people exposed to 64-knot winds in the forecasts 1 to 3 days ahead (Shefa, Sanma and Tafea) or in the observed track (all of Vanuatu). There are no forecast records before 2012, or for Tam 2025 and an unnamed 2024 system. Gene 2008 and Atu 2011 (pale red) are counted together as one assumed activation. Seasons are named by the year they end: Lola occurred in October 2023, in the 2023/24 season.
Forecast exposure: JTWC official forecasts (RSMC Nadi and VMHS forecasts are not archived with wind radii), 24–72 h track segment, WorldPop population inside the forecast 64 kt swath in Shefa, Sanma and Tafea. Observed exposure: IBTrACS best track, all of Vanuatu. People affected: EM-DAT. Gene 2008 and Atu 2011 both caused damage that EM-DAT does not record for Vanuatu. Storms without a forecast record show 0 observed exposure, as in the framework; observed exposure was calculated for all 20 seasons and none of these storms reached 15,000. The table includes all storms with any 64 kt exposure, people affected or a CERF allocation, plus the next closest storms.
As in section III.2 of the framework, counted by season. The readiness trigger is not included (too few archived long-range forecasts), so the overall return period is an upper bound.
Select a storm to see each of its forecasts as issued. Blue: forecast 64-knot wind area, 1 to 3 days ahead. Orange: observed 64-knot wind area. Green outline: Shefa, Sanma and Tafea.
The boundaries and names shown and the designations used on this map do not imply official endorsement or acceptance by the United Nations.
The forecasts are JTWC official forecasts, used because RSMC Nadi and VMHS forecasts are not archived with wind radii. Forecast exposure is the WorldPop population of Shefa, Sanma and Tafea inside the forecast 64-knot area. All storms and forecasts, and the alternatives considered during the design, are in the technical analysis.
The framework aims to capture most cyclones that required international humanitarian assistance, using past CERF allocations for cyclones in Vanuatu as a proxy, with a return period of around 4 years.
Storms with any exposure to 64-knot winds. Upper bar: highest forecast exposure 1 to 3 days ahead (Shefa, Sanma and Tafea). Lower bar: observed exposure (all of Vanuatu). The vertical line is the threshold of 15,000 people.
The working group considered adding a rainfall condition and decided against it. In the historical analysis, adding a rainfall threshold made little or no difference to which storms would have activated the trigger, and a single indicator is simpler to monitor and communicate.
The charts show the correlation of each indicator with the number of people affected and with CERF allocations, for all storms from 2006 to 2025. Exposure to 64-knot winds is much more strongly correlated with both than rainfall is.
Pearson correlation (0 = none, 1 = perfect). “Exp” = people exposed at that wind speed in Shefa, Sanma and Tafea (observed track).
All three triggers use the same threshold: at least 15,000 people exposed to maximum sustained winds of 64 knots (about 119 km/h) or more. They differ in lead time, area and data source.
| Trigger | Lead time | Wind speed | Number of people exposed | Provinces | Data source |
|---|---|---|---|---|---|
| Readiness | 72 to 120 hours | ≥ 64 knots | ≥ 15,000 people | Shefa, Sanma, Tafea | RSMC Nadi extended track forecast |
| Action | 24 to 72 hours | VMHS track forecast | |||
| Observational | n/a | All of Vanuatu | VMHS observed track |
Lead time is the time before landfall. Wind speeds are 10-minute sustained winds, as used by RSMC Nadi and VMHS. No anticipatory activities begin in the 24 hours before landfall (NDMO's operational cut-off); activities under the observational trigger start after landfall and NDMO's all-clear. Triggers are monitored from 1 October to 31 May. CERF has pre-arranged up to $1.5 million, disbursed once, on the first trigger reached: a readiness activation authorises the readiness share of the funds, an action or observational activation all of them. The Regional Humanitarian Pooled Fund for Asia and the Pacific adds $1.2 million for NGOs and the Vanuatu Red Cross.
Forecast data. The trigger is monitored with RSMC Nadi and VMHS forecasts. Testing it on past storms requires the forecasts as issued, including the forecast extent of 64-knot winds, and neither source archives this. VMGD shared its historical archive (9 GB, sixteen seasons); its Forecast Track Maps give position and intensity category, with wind extents only drawn on the map images. The official forecasts of the US Joint Typhoon Warning Center (JTWC) include wind radii (four quadrants, 34/50/64 kt) and are kept in the VMGD archive for 2012–2024 and in a public research archive for 2025. JTWC forecasts the same storms at the same times, with skill comparable to the operational sources, so its forecasts are used here.
Exposure. For each forecast (534 forecasts for 47 storms), only the 24–72 hour segment of the forecast track is used: the action window. The forecast 64-knot wind radii along that segment give the wind area, and the population of Shefa, Sanma and Tafea inside it is counted with WorldPop 1 km data. Observed exposure is calculated the same way from IBTrACS best-track data. JTWC reports 1-minute average winds and RSMC Nadi and VMHS 10-minute averages, so 64 kt is evaluated at its 1-minute equivalent (≈73 kt, factor 0.88), with the radii interpolated to that speed. The method was checked by calculating one storm's wind area in two independent ways (96% area agreement) and by comparing each false alarm with damage reports from the time.
Gaps in the record. There are no forecast records for 2006–2011 (six seasons), so the action trigger cannot be assessed for those years. Two storms in that period (Gene 2008, Atu 2011) brought 64-knot winds to or near the trigger provinces and caused documented damage, so the statistics include one assumed activation for that period. Season 2005 is excluded: it is the only year between 2003 and 2011 with forecast files, and their source is unclear. The readiness trigger (3–5 days) is not assessed: archived long-range forecasts with wind radii are too few before 2017. For these reasons the overall return period is given as an upper bound (≤).
The data is the same as in the technical analysis, which shows each storm and forecast in detail. The code is in the ds-aa-vut-cyclones repository.