Why ENACTS didn’t trigger — and what the gauges actually say

01 · THE OFFICIAL NUMBERWhat the Maproom actually shows

The operational trigger surface is the IRI FBF Maproom (fbfmaproom2/niger). Its observational indicator, enacts-mon-spi-jj, is the ENACTS monitoring product: DMN station data merged onto a CHIRP (satellite-only) background, averaged as per-pixel SPI over the national zone south of 17°N. Queried today through the Maproom’s export API:

Two of the Maproom’s corroborating gauge datasets are simply not current: the pure DMN station-spi-jj series ends in 2022 and the final ENACTS product ends in 2021. Only the MON product is live, and the gridded ENACTS datasets behind it are access-restricted on the Data Library — the export API is the public window.

02 · WHY THE PROXIES DISAGREEOne satellite background, two station stories

CHIRP and CHIRPS share the same infrared satellite estimate; CHIRPS adds the Climate Hazards Center’s gauge blend. In a typical recent year that blend adds 15–30 mm to the Niger Jun–Jul box mean (+55 mm in 2024). In 2026 it subtracted 6 mm — a swing of 25–35 mm downward, entirely attributable to what the gauges reported. CHC’s published station lists confirm 13–14 Niger gauges were ingested in June and July 2026, the same synoptic set as in 2024–25. CHIRPS’s dry 2026 signal is gauge-driven, not a satellite artifact.

ERA5 is the loudest witness (2nd driest of 36) but also the least reliable for this zone: it assimilates no rain gauges, and against the Maproom’s own DMN station SPI (1991–2022) it correlates at just r ≈ 0.59 — the weakest of everything we tested.

The historical benchmark cuts against dismissing the proxy: over 1991–2022, CHIRPS tracked Niger’s own station SPI as well as the official ENACTS MON product did. CHIRPS is not a poor cousin for this zone.

03 · THE GROUND TRUTH PROBLEMWhat actual rain gauges say about 2026

First, the uncomfortable finding: no curated public gauge archive covers Jun–Jul 2026. GHCN-Daily has complete Niger precipitation through 2025 and then goes silent; GSOD and ISD end at 2025; Meteostat’s feed dies in March 2026. What remains is the GTS stream itself, which reaches the public in two forms — the daily synop reports, and the monthly CLIMAT summaries DMN compiles and transmits after each month (archived by OGIMET back to ~2008).

These two routes disagree, and the CLIMAT monthlies win. The first version of this note summed the daily synop stream, which only captured ~70% of days — and the missing days carried rain, so the eastern totals came out far too low (Zinder read 82 mm against a CLIMAT-complete 211; Magaria 97 vs 186; Mainé-Soroa 33 vs 90). The complete monthly record, below, is the better gauge truth — with its own defects: Diffa’s July 2026 CLIMAT transmits 447 mm from 4 rain days (garbage), and Maradi’s June transmits exactly 0.0 mm (doubtful — the daily stream shows June rain there).

04 · THE REMAINING PUZZLEIf the gauges say “pockets”, who is the outlier?

The corrected gauge record changes the shape of the question. A wet west offsetting eastern and July-only deficits is consistent with ENACTS MON’s near-normal zone mean (−0.12) — the official indicator no longer looks like it ignored its own inputs. What now needs explaining is the other side: why does CHIRPS place the same season at the 15% line, and ERA5 at the 6th percentile? A plausible mechanism for CHIRPS is that CHC’s blend ingests the GTS daily stream — the same incomplete feed that biased the first version of this note dry — rather than the complete monthly CLIMATs. ERA5, with no gauge input and the weakest historical correlation, is most simply read as an outlier. The MON-vs-CHIRP station pull is still worth watching, though: 2026’s pull (−0.11) is small next to recent years, and station counts in the MON run remain unpublished.

In 2022 the station merge pulled the index down by three-quarters of an SPI unit — that pull is what activated the framework. In 2026 the pull was 0.11 — which the corrected gauge record suggests may simply be the truth (wet west, dry pockets), though two residual checks remain worthwhile:

1. Station input to the 2026 run. The endorsed 2024 framework specifies falling back to satellite-only CHIRP if ≥20% of stations are missing by 7 August. Asking IRI/DMN for the station count costs nothing and settles whether the small pull reflects real conditions or thin input.
2. Dilution by design. The indicator is an equal-weight average of per-pixel SPI over the whole zone south of 17°N. That is exactly why a pocket drought — Zinder/Magaria/Mainé-Soroa’s failed July, a failed June around Maradi/Tahoua — cannot move it much. Working as designed is not the same as measuring what matters for a departmental-scale food-security shock.

05 · THE COUNTEREXAMPLE2022 cuts the other way

Before concluding ENACTS is broken, note that the exact mirror image happened in 2022 — the framework’s one real observational activation. ENACTS MON read −0.79 (4th driest on record) and DMN stations agreed (−0.58); our CHIRPS and ERA5 zone means put 2022 mid-pack (16th–17th of 35). If the argument this year is “CHIRPS says 2026 should have triggered,” the symmetric argument is “CHIRPS says 2022 shouldn’t have.” The 2022 drought was concentrated in the southwest; 2026’s is in the east. A single zone-wide average is fragile to regional droughts, and which dataset “sees” a given drought depends on where its information sources sit. That — not any one product being wrong — is the structural finding.

06 · POSITIONINGWhat to say under scrutiny

1. The process was followed and the call wasn’t close on its own terms. The designated indicator, computed by the designated operator on the designated platform, read the 44th percentile against a 15% threshold. It would not have activated even at the endorsed 35% threshold. There is no version of “it just missed.”
2. Niger’s own monthly gauge reports support a “pockets” reading, not a zone-wide drought. Most stations’ Jun–Jul totals were normal to above normal (the west clearly wet); severe deficits were real but localized — a failed July around Zinder, Magaria and Mainé-Soroa, a failed June around Maradi and Tahoua — matching FEWS NET’s “deficits in central-southern Niger.” A zone-average indicator reading near-normal over that pattern is arithmetic, not malfunction.
3. Dataset disagreement is structural and symmetric. CHIRPS puts the zone at the 15% line and ERA5 lower still, while ENACTS, CHIRP and the CLIMAT gauges read near-normal; in 2022 the disagreement ran the other way (ENACTS triggered, CHIRPS/ERA5 mid-pack). Which product “sees” a given drought depends on where its information sits — which is why no single zone-average product should carry the whole decision.
4. Two technical checks are still open, and we are pursuing them: the station count behind the 2026 MON run (the framework’s own ≥20%-missing CHIRP fallback rule makes this a fair question), and why CHIRPS’s gauge blend diverges from the complete CLIMAT record — plausibly because it ingests the incomplete GTS daily stream.
5. For the 2027 redesign this argues for sub-zonal resilience, not dataset loyalty: 2026 (an eastern pocket drought the zone mean averaged away) and 2022 (a southwestern one it caught) together show the binding constraint is spatial aggregation more than dataset choice. Departmental return periods — the approach on the pockets page — plus a multi-dataset check (CHIRPS tracks DMN gauges historically as well as ENACTS MON does, r ≈ 0.90 vs 0.87) would catch what either misses alone, at the cost of a more complex trigger to govern.

07 · OPEN QUESTIONS FOR IRI / DMNThe checklist

• How many DMN stations were merged into the 2026 Jun–Jul ENACTS MON run, and with what cutoff date? Was the ≥20%-missing CHIRP fallback rule evaluated?
• Can the station-spi-jj series (ends 2022) and final ENACTS (ends 2021) be brought current on the Maproom, so gauge-vs-product checks stay possible?
• What is the revision policy for the MON SPI series? (We observe historical values shifting between April and August 2026 exports.)
• Can IRI publish the MON gridded anomaly map for Jun–Jul 2026, to confirm whether the eastern deficit appears spatially but averages out?

Method notes. Maproom values from the fbfmaproom2/niger/export API (national region, 1991–2026, include_upcoming). CHIRP/CHIRPS/ARC2 series computed on the IRI Data Library over 0–16°E, 11–17°N; zone-masked CHIRPS/ERA5 from this repo’s docs/rainfall pipeline (Niger south of 17°N). Gauge figures are DMN monthly CLIMAT reports decoded from OGIMET (2008–2026, 8–12 complete Jun+Jul years per station; the repo’s exploration/pockets_fetch_gauges.py pipeline); per-station July ranks computed within each station’s own CLIMAT Julys. Known bad values: Diffa Jul 2026 (447 mm), Maradi Jun 2026 (0.0 mm, doubtful). Revision 1 Sep 2026: the first version of this note summed GTS daily synop reports (~70% day coverage) and overstated the eastern seasonal deficits; the CLIMAT monthlies supersede those figures. Historical correlations computed against the Maproom’s DMN station-spi-jj series, 1991–2022. FEWS NET: West Africa seasonal monitor, July 2026. Companion view: 2026 drought-pockets page (per-department RPs, HNRP severity overlay).