About this replay
- What is replayed. Everything below the bar is the monitoring page's own code, fed with the pipeline's own evaluation and chart of the forecast issue selected.
- Forecasts. The GloFAS reforecast archive the trigger was tested on: 11 ensemble members, two issues a week. The live pipeline reads the full operational ensemble every day, so the first day on which a trigger is reached is only known to within the gap between two issues. GloFAS was already on version 4 in October 2023, the version today's thresholds are fitted on.
- Result. The readiness trigger is first reached on the 21 Oct issue (Juba). The activation trigger is first reached on the 21 Oct issue (Juba).
- The Somalia Humanitarian Fund allocation. The Fund announced a $15 million allocation for early action and response on 12 Oct 2023, on the seasonal El Niño outlook (OCHA, 12 Oct 2023). This trigger did not exist then and was not the basis for it.
- Thresholds are today's. They are fitted on the GloFAS version 4 record for 1999 to 2023, so the 2023 season is part of the record it is judged against.
- The Shabelle. Neither trigger is reached on the Shabelle on any issue replayed. Two Shabelle gauges had reached their own 1-in-3 level by 6 Nov. The highest forecast in the activation band stays between 50% and 70% of its threshold. The GloFAS version switch page describes this gap in version 4.
- The whole season on one chart (model, gauges and flood exposure): Timing of activations, Deyr 2023.
The forecast against the trigger thresholds
Forecast peak at each station
For each window of the season being monitored, the forecast peak at each station within the leg's lead band and inside the window months, against the station's threshold. A ratio of 1.00× or more means the peak reaches the threshold and the station votes.
How to read this page
- Four windows, one source each. Gu (window February to June) runs on Google Flood Hub, Deyr (window September to January) on GloFAS. A window's trigger is reached when, on a single forecast day, enough of the river's stations are forecast at or above their own return-period threshold: Juba needs 3 of 4 stations, Shabelle 2 of 3. Any one window reaching its trigger releases the allocation.
- Activation reads leads 1 to 7 days on the window's source. Readiness reads GloFAS at leads 8 to 12 days for every window, against discharge return level thresholds fitted on the same reanalysis record as the action leg, and releases only the mobilisation share.
- SWALIM. The framework also allows the readiness trigger to be reached on a SWALIM moderate flood risk alert for either river. Bulletins are not read by this pipeline. Check the SWALIM river levels alongside it.
- A window is open by calendar month: Deyr from September to January, Gu from February to June. Every forecast day inside the window counts, so monitoring for Deyr starts in September and for Gu in February. The thresholds and return periods are fitted on the three season months only (Gu March to May, Deyr October to December); the months either side are a surveillance buffer.
- Full design and evidence: the trigger summary and the analysis page. Which GloFAS version is live, and what it does to the thresholds: GloFAS version switch.
Contact
Reach out to OCHA Data Science at ocha-datascience@un.org with any questions.