Seasonal Rainfall Outlook Platform
Interpretation · Bias correction · February initialization · FMAM and its months

Interpretation · February initialization · FMAM and its months

Interpretation — national aggregate (All Ethiopia)

  1. All tested corrections improve national FMAM rainfall-amount CRPSS over raw ECMWF in 1993–2016 (raw -0.585; corrected -0.059 to -0.005). Performance varies for monthly tercile probabilities: 0 method–month combinations score below the raw model on RPSS.
  2. Accounting for the whole search. With Holm adjustment across all 60 method–target comparisons, the amount (CRPS) screening passes are none, and the tercile (RPS) passes are Square-root-space Mean-Variance Scaling (Feb, Mar), Power Transformation (Feb, Mar) and Empirical Quantile Mapping (Feb). Within-method passes that do not survive this check: Empirical Quantile Mapping Mar, Gamma Quantile Mapping Feb, Gamma Quantile Mapping Mar, Power Transformation May, Square-root-space Mean-Variance Scaling Feb, Square-root-space Mean-Variance Scaling Mar and Variance Scaling (unbounded) Feb.
  3. Empirical Quantile Mapping reproduces the CHIRPS distribution and, with the dry policy, its dry share (totals ≤ 1 mm: Feb 12.9% vs CHIRPS 12.9%); it passes the screening for Feb and Mar, but its FMAM amount skill is lower than Mean-Variance Scaling (-0.018 vs -0.007). A possible explanation is sampling noise in an empirical quantile function built from 23 CHIRPS values per fold; the results do not isolate it.
  4. Scaled Distribution Mapping follows the published algorithm (Switanek et al. 2017, steps 1–7; checked numerically against an independent implementation of the steps). Where the model is drier than CHIRPS it cannot add rain: in Feb 58% of model members exceed 1 mm against 87% of CHIRPS years, and SDM, by design, keeps the model’s number of rain members, so its corrected dry share is 44% (CHIRPS 13%).
  5. The methods that use the forecast’s own distribution (Detrended Quantile Mapping, Quantile Delta Mapping, Scaled Distribution Mapping, CDF-transform) score lower on FMAM terciles (-0.045 to -0.023 vs +0.005) and none passes the season screening. A possible explanation is that the distribution of a single forecast’s 25–51 members is a noisy estimate of the “future” distribution these methods were designed for; the results do not isolate the cause. Where model quantiles are dry, QDM falls back to the CHIRPS quantile (members affected: Feb 42% and Mar 18%), so its quantile changes are only partly preserved.
  6. Published probabilities: confirmed better than Mean-Variance Scaling: Square-root-space Mean-Variance Scaling (Feb). Choosing the method each year from training-only scores, the probability procedure (SEL-P) scores FMAM blended RPSS +0.020 against +0.021, confirmed for Feb. Amounts: the amount procedure (SEL-A) scores CRPSS -0.011 against -0.007, confirmed for Feb and Mar; fixed methods confirmed against the family of all declared methods and targets: none.
  7. The climatology blend narrows the differences. Across the methods, FMAM RPSS (1993–2016) spans 0.010 after the blend against 0.051 for the unblended screening counts. A better amount correction is trusted more by the blend: in Mar the fitted weight on climatology is 0.47 with Square-root-space Mean-Variance Scaling against 0.71 with Mean-Variance Scaling.
  8. Choosing by training scores can backfire. For Apr, SEL-P chose Linear Scaling from the 1993–2016 scores; over 2017–2025 it has a worse mean RPS (Δ +0.000, 95% interval -0.004 to +0.004; better than Mean-Variance Scaling in 4 of 9 years, with the largest losses in 2021 and 2020); for May, SEL-P chose Square-root-space Mean-Variance Scaling from the 1993–2016 scores; over 2017–2025 it has a worse mean RPS (Δ +0.001, 95% interval -0.006 to +0.009; better than Mean-Variance Scaling in 5 of 9 years, with the largest losses in 2021 and 2025). These choices fail the no-harm rule; where the interval includes 0 they do not establish a significant deterioration. A method that ranks first on the training years is not, on that alone, a replacement.
  9. No cell of All Ethiopia has exactly 0 mm in every training year, for any target, at 0.25° with CHIRPS. The all-zero guard (a separate MVS variant, outside the declared family) therefore never acts and equals Mean-Variance Scaling: the non-zero totals that Mean-Variance Scaling gives where every training year is 0 mm (its spread ratio cannot fall below 0.5) do not arise in these data. The guard remains a documented option; a finite history of zero rainfall does not make rain impossible.

5 of the 14 interpretation statements do not hold for this cycle’s results and are not shown (each is checked against the numbers before it is displayed).

Caveats

  • Screening skill is from 1993–2016 leave-one-year-out (clean) and 2017–2025 (exploratory: these years were seen while the operational method was chosen).
  • Screening tercile scores use smoothed member counts without the climatology blend; the confirmation scores the blended product.
  • Two multiplicity families are shown: within one method across its targets (✓) and across all methods and targets (✓✓). The confirmation’s selection-aware procedures include the method choice in the evaluation itself.
  • Scores are area-weighted over many correlated cells but tested year by year, so each year counts once.
  • The ordinary CRPS scores the delivered ensemble; the fair CRPS is a diagnostic under a sampling interpretation that DC’s rescaled years and methods tied to the forecast mean or ranks only approximate.
  • Results are national aggregates unless a table says otherwise; a method can rank differently in a region (see the regional tables).

← SummaryData and methods →