Chad drought AA — Window 3 (biomasse), season so far
Trend-adjusted cumulative biomasse anomaly over the 7-province area of interest (Bahr-El-Gazal, Batha, Kanem, Lac, Ouaddaï, Sila, Wadi Fira) · data through dekad 21 (21–31 July 2026)
2026 anomaly at dekad 21
62.5 %
of the 1999–2024 linear-trend expectation
Trigger threshold (at dekad 24)
< 84.5 %
2026 sits 22 pts below it at dekad 21
Rank at this dekad
2nd lowest
since 1999 — only 2015 (51.6 %) was lower
Provinces below 84.5 %
7 / 7
Kanem lowest at 31.7 %
Anomaly at dekad 21 (21–31 July), 1999–2026
above 84.5 % at dekad 21 (no signal)
below 84.5 % at dekad 21 (⚠️ drought signal)
season actually finished below 84.5 % at dekad 24 (official backtest)
Did the seasons that looked bad at dekad 21 actually trigger?
Of the 8 historical seasons below 84.5 % at dekad 21, 5 actually finished below the threshold at dekad 24 and would have triggered Window 3 per the framework's backtest (2006, 2009, 2011, 2015, 2017). 2014 is on the line: it finishes at 84.2 % in this dataset but at 90 % in the framework's calibration data — officially not a trigger year. Two seasons — 2002 and 2013 — recovered on August rains, and both had been borderline at dekad 21, not deep deficits. No season that was above the line at dekad 21 has ever finished below it. 2026's dekad-21 value is lower than any of the recovery years by 10+ points.
at dekad 21 (21–31 July)
finish at dekad 24: triggered
finish: marginal (2014)
finish: recovered
Data table (year, cumulative biomasse, trend fit, anomaly at d21 and d24, outcome)
| Year | Biomasse at d21 (kg/ha, AOI sum) | Trend fit | Anomaly d21 | Anomaly d24 | Triggered at d24? |
|---|---|---|---|---|---|
| 1999 | 3,981 | 3,069 | 129.7 % | 135.5 % | no |
| 2000 | 3,771 | 3,095 | 121.8 % | 107.6 % | no |
| 2001 | 3,040 | 3,121 | 97.4 % | 101.6 % | no |
| 2002 | 2,334 | 3,147 | 74.2 % | 85.2 % | no |
| 2003 | 4,207 | 3,173 | 132.6 % | 112.4 % | no |
| 2004 | 3,343 | 3,199 | 104.5 % | 92.0 % | no |
| 2005 | 4,110 | 3,225 | 127.5 % | 116.7 % | no |
| 2006 | 2,368 | 3,251 | 72.9 % | 77.7 % | ✗ would have triggered |
| 2007 | 3,139 | 3,277 | 95.8 % | 108.8 % | no |
| 2008 | 3,540 | 3,303 | 107.2 % | 99.4 % | no |
| 2009 | 2,608 | 3,329 | 78.4 % | 79.8 % | ✗ would have triggered |
| 2010 | 2,875 | 3,355 | 85.7 % | 99.0 % | no |
| 2011 | 2,262 | 3,381 | 66.9 % | 74.7 % | ✗ would have triggered |
| 2012 | 4,538 | 3,407 | 133.2 % | 126.0 % | no |
| 2013 | 2,896 | 3,433 | 84.4 % | 96.0 % | no |
| 2014 | 2,273 | 3,459 | 65.7 % | 84.2 % | marginal (source-dependent) |
| 2015 | 1,798 | 3,484 | 51.6 % | 68.4 % | ✗ would have triggered |
| 2016 | 3,860 | 3,510 | 110.0 % | 98.3 % | no |
| 2017 | 2,886 | 3,536 | 81.6 % | 76.5 % | ✗ would have triggered |
| 2018 | 3,846 | 3,562 | 108.0 % | 103.0 % | no |
| 2019 | 4,142 | 3,588 | 115.4 % | 105.8 % | no |
| 2020 | 3,433 | 3,614 | 95.0 % | 105.4 % | no |
| 2021 | 3,863 | 3,640 | 106.1 % | 104.4 % | no |
| 2022 | 4,218 | 3,666 | 115.1 % | 122.8 % | no |
| 2023 | 5,093 | 3,692 | 137.9 % | 102.4 % | no |
| 2024 | 3,807 | 3,718 | 102.4 % | 117.0 % | no |
| 2025 | 4,500 | 3,744 | 120.2 % | 116.9 % | no |
| 2026 | 2,356 | 3,770 | 62.5 % | — | ? (early Sept) |
2026 by province (same construction, per province)
| Province | Anomaly at dekad 21 | Status |
|---|---|---|
| Kanem | 31.7 % | below threshold |
| Wadi Fira | 49.2 % | below threshold |
| Batha | 57.2 % | below threshold |
| Lac | 62.6 % | below threshold |
| Ouaddaï | 62.8 % | below threshold |
| Barh El Gazel | 64.4 % | below threshold |
| Sila | 67.0 % | below threshold |
How to read this — caveats
- The season can still recover. Dekad-21 and dekad-24 anomalies correlate historically (r ≈ 0.84), but two seasons (2002, 2013) crossed back above the threshold on August rains — see the season-outcome panel above. Both were borderline at dekad 21 (74 % and 84 %); the only year lower than 2026 at this dekad (2015, 51.6 %) went on to trigger.
- Substitute data source. The dekadal feed used by the repo pipeline
(
WA_DMP_ADM1_ef_v1) has not updated since dekad 31 of 2025, so this check uses the dekadal DMP columns of GeoSahel'sWA_BIO_ADM1_v4layer (observed data through dekad 21 of 2026, no climatological fill). Validation: at dekad 24 that source reproduces the framework's five backtest activation years (2006, 2009, 2011, 2015, 2017), but also puts 2014 marginally below threshold (84.2 %), and year-level anomalies differ from the operational feed by a few percentage points — treat all values here as approximate. - The threshold does not formally apply at dekad 21. 84.5 % was calibrated on dekad-24 values for a ~5-year return period. At dekad 21 the same cutoff would have flagged 8 of 27 historical years — a more frequent event.
- Consistent with the forecast windows. Windows 1 and 2 already triggered on 30 April and 9 May 2026 on SEAS5 forecasts of a poor July–September season; this observational reading points the same way.
What actually matters: the official Window 3 check runs in early September on dekad 24 (21–31 August) data, verified with DREM before any notification.
Exploratory analysis generated 12 August 2026 · data accessed 12 August 2026 from
GeoSahel (ACF BioGenerator, VITO/Copernicus DMP) ·
methodology: ds-aa-tcd-drought,
Cadre de l'action anticipatoire — sécheresse au Tchad (v. 2025-03-03) ·
CH/IPC evolution report