Ethiopia rainfall forecast verification — FMAM 2026, January initialization
January 2026 initialization | FMAM 2026 season and its months | ECMWF seasonal forecasts | CHIRPS v3.0 verification | Reference period 1993–2025
Verified targets: Feb, Mar, Apr, May, FMAM. FMAM is assessed directly against its complete seasonal observations; overlapping monthly and seasonal scores are not pooled.
This is a retrospective assessment of the reconstructed forecasts. It does not establish an actual January 2026 issuance, independent prospective validation, or an official EMI/ICPAC forecast. All results below concern the submitted verification evidence.
Decision and principal findings
Retain frozen shared-blend forecasts. Regional annotations are post-event verification findings, not forecast-time confidence labels or new masks. No automatic model switch.
Feb: corrected rainfall CRPSS -20.9%; final probability RPSS -3.7%; amount-mean error +2.64 mm. Probability results cover 85.9% of country grid area.
Mar: corrected rainfall CRPSS +10.1%; final probability RPSS +4.9%; amount-mean error -48.21 mm. Probability results cover 98.5% of country grid area.
Apr: corrected rainfall CRPSS +1.4%; final probability RPSS +1.0%; amount-mean error +5.95 mm. Probability results cover 100.0% of country grid area.
May: corrected rainfall CRPSS +3.2%; final probability RPSS +2.8%; amount-mean error +4.91 mm. Probability results cover 100.0% of country grid area.
FMAM: corrected rainfall CRPSS -6.6%; final probability RPSS -2.6%; amount-mean error -35.98 mm. Probability results cover 100.0% of country grid area.
Positive skill means a lower score than the stated climatological reference on the same support. It is not percentage forecast accuracy. Positive mean error means the forecast was wetter than observed, which can occur even when the forecast anomaly is below normal.
Country results
| Target | Raw CRPS mm | Corrected CRPS mm | Climatology CRPS mm | Corrected CRPSS | Final RPSS |
|---|---|---|---|---|---|
| Feb | 9.20 | 5.70 | 4.71 | -20.9% | -3.7% |
| Mar | 68.90 | 41.55 | 46.21 | +10.1% | +4.9% |
| Apr | 27.68 | 24.95 | 25.29 | +1.4% | +1.0% |
| May | 26.11 | 17.58 | 18.16 | +3.2% | +2.8% |
| FMAM | 105.99 | 47.84 | 44.87 | -6.6% | -2.6% |
Main FMAM rainfall domain: FMAM rainfall contribution domain
This fixed pre-2026 domain contains 1120 cells and 75.7% of country grid area. Definition: Fixed 1993-2025 CHIRPS v3.0 domain: cells where mean FMAM rainfall (actual-calendar monthly totals, February 29 included) is >= 100 mm and FMAM brings >= 20% of the mean annual rainfall (ratio of climatological means); complete baseline required; no patch removal or smoothing. Selected in assessment fmam_chirps_domain_review_v1. The same domain is used for the season and each of its months. The included share of Ethiopia describes this rule; it is not an independently established official share of FMAM-dependent area.. Monthly rows evaluate each month inside that domain; they are not a complete-FMAM assessment.
| Target | Amount CRPSS | Smoothed RPSS | Final RPSS | Mean error mm | Probability coverage |
|---|---|---|---|---|---|
| Feb | -18.7% | -33.7% | -3.6% | 2.24 | 90.8% |
| Mar | +10.1% | +23.8% | +4.9% | -59.03 | 99.7% |
| Apr | +2.4% | +2.4% | +1.5% | 17.65 | 100.0% |
| May | +3.1% | +4.0% | +3.3% | 3.06 | 100.0% |
| FMAM | -9.7% | -9.5% | -5.4% | -38.12 | 100.0% |
Regional performance notes
Feb 2026
| Regime | Amount CRPSS | Smoothed RPSS | Final RPSS | Probability coverage | Blend effect |
|---|---|---|---|---|---|
| R0: arid / marginal | -35.9% | -32.6% | -5.7% | 100.0% | Helped |
| R1: western unimodal | -11.5% | -18.9% | -1.5% | 74.8% | Helped |
| R2: Belg–Kiremt rule | -48.8% | -41.4% | -7.5% | 99.1% | Helped |
| R3: Gu–Deyr rule | -14.6% | -35.4% | -1.4% | 82.8% | Helped |
Mar 2026
| Regime | Amount CRPSS | Smoothed RPSS | Final RPSS | Probability coverage | Blend effect |
|---|---|---|---|---|---|
| R0: arid / marginal | -9.0% | -4.6% | +1.9% | 100.0% | Helped |
| R1: western unimodal | +7.8% | +16.7% | +3.4% | 95.3% | Weakened |
| R2: Belg–Kiremt rule | +8.5% | +16.6% | +3.1% | 100.0% | Weakened |
| R3: Gu–Deyr rule | +13.4% | +35.8% | +7.4% | 99.5% | Weakened |
Apr 2026
| Regime | Amount CRPSS | Smoothed RPSS | Final RPSS | Probability coverage | Blend effect |
|---|---|---|---|---|---|
| R0: arid / marginal | -11.4% | -27.8% | -2.7% | 100.0% | Helped |
| R1: western unimodal | +3.2% | +5.6% | +1.9% | 100.0% | Weakened |
| R2: Belg–Kiremt rule | -0.6% | -5.4% | -0.1% | 100.0% | Helped |
| R3: Gu–Deyr rule | +1.3% | +2.9% | +1.4% | 100.0% | Weakened |
May 2026
| Regime | Amount CRPSS | Smoothed RPSS | Final RPSS | Probability coverage | Blend effect |
|---|---|---|---|---|---|
| R0: arid / marginal | +12.9% | +16.4% | +10.4% | 100.0% | Weakened |
| R1: western unimodal | +1.8% | -0.6% | +0.7% | 100.0% | Helped |
| R2: Belg–Kiremt rule | +8.3% | +7.9% | +4.8% | 100.0% | Weakened |
| R3: Gu–Deyr rule | +0.8% | +1.7% | +2.2% | 100.0% | Helped |
FMAM 2026
| Regime | Amount CRPSS | Smoothed RPSS | Final RPSS | Probability coverage | Blend effect |
|---|---|---|---|---|---|
| R0: arid / marginal | -11.5% | -19.6% | -12.1% | 100.0% | Helped |
| R1: western unimodal | +3.6% | +8.3% | +6.3% | 100.0% | Weakened |
| R2: Belg–Kiremt rule | -3.0% | -3.1% | -1.6% | 100.0% | Helped |
| R3: Gu–Deyr rule | -18.0% | -17.1% | -10.2% | 100.0% | Helped |
These labels describe the observed performance of these targets. They do not justify hiding weak regions, declaring confidence at issuance, applying a new mask, or replacing forecast values after observing outcomes.
Historical context
Historical review was not supplied to this build. No historical ranking is inferred.
Methods and interpretation
Rainfall amounts: equal-year, cell-specific mean–variance bias correction. All 51 operational members are retained. The ensemble and climatological reference are evaluated using empirical CRPS; their finite ensemble sizes are 51 and 33, respectively, without an iid/fair-ensemble adjustment.
Probabilities: observed training-period terciles, member counts, alpha = 0.5 additive smoothing, then the selected shared climatology blend. The final product does not use the experimental Dirichlet, local or regime-specific probability mappings. Probability blending does not change the corrected rainfall members or amount CRPS.
RPS is the sum of the first two cumulative-category squared errors. Brier scores are category-specific; log loss uses natural logarithms and the existing probability floor. Country and domain scores use grid-cell area weights on identical support across methods within each metric family.
Amount and probability masks differ; unscored cells are not near-normal conditions. R0–R3 are the reconciled GitHub-derived climatological rules, not administrative boundaries or independently validated official EMI zones. No onset-detection gate is used.
One verification year cannot establish multi-year reliability or statistical significance. Spatial cells and overlapping months/FMAM are not independent evaluation samples. No pooled overall score is produced. The 2026 outcomes are now inspected evidence and must not be reused as an untouched test set.
Verification maps
Completion and future work
Retain the frozen forecasts and archive this report as a verification addendum. A future calibration revision needs a separate historical evaluation with all preprocessing and model selection inside year-based validation. This report does not fit or adopt a revised model.
Evidence and reproducibility
Built on the project computer. Frozen forecast hashes checked before and after report construction; Step 30 report, field and observation hashes matched to Step 31. Summary consistency and disjoint-regime reconstruction checked. No new calibration or scoring performed.