Historical skill · May 2026
How forecasts made with the same method scored in past years against climatology (the observed frequency of each tercile in the training years, about one-third each). The main number is the cross-validated 1993–2016 score: each year is forecast with parameters fitted without that year. Later years are exploratory: they were seen while the method was chosen.
Takeaway — May 2026
Historical skill is uncertain: cross-validated 1993–2016 RPSS is +0.007, and its interval (−0.051 to +0.050) includes no improvement over climatology. The exploratory 2017–2025 result (+0.017) is encouraging but those years were inspected during method selection. The 2026 verification (+0.002) is one season and does not override the historical uncertainty.
Stronger historical evidence (interval above zero): Feb +0.137.
| Target | Cross-validated 1993–2016 (main) | In words | Exploratory 2017–2025 |
|---|---|---|---|
| May | +0.007 (−0.051 to +0.050) | Small estimated improvement; skill uncertain | +0.017 (−0.033 to +0.057) |
| FMAM | +0.017 (−0.038 to +0.067) | Small estimated improvement; skill uncertain | +0.061 (+0.029 to +0.095) |
| Feb | +0.137 (+0.050 to +0.234) | Moderate improvement over climatology (interval above zero) | +0.049 (−0.017 to +0.128) |
| Mar | +0.011 (−0.020 to +0.042) | Small estimated improvement; skill uncertain | +0.041 (+0.007 to +0.080) |
| Apr | +0.011 (−0.032 to +0.048) | Small estimated improvement; skill uncertain | +0.015 (−0.009 to +0.042) |
RPSS +0.05 means a 5% lower ranked probability score than climatology. An interval that includes zero means the gain is not established.
Technical detail: years better, p-values and blend weights
| Target | Years better (cross-validated) | p (one target / Holm) | Years better (exploratory) | Blend weight λ |
|---|---|---|---|---|
| FMAM | 15 of 24 | 0.276 / 1.000 | 8 of 9 | 0.40 |
| Feb | 16 of 24 | 0.002 / 0.012 | 5 of 9 | 0.50 |
| Mar | 15 of 24 | 0.251 / 1.000 | 7 of 9 | 0.73 |
| Apr | 16 of 24 | 0.305 / 1.000 | 5 of 9 | 0.58 |
| May | 17 of 24 | 0.434 / 1.000 | 8 of 9 | 0.55 |
Each interval and one-target p-value describes that target alone. Because 5 targets are examined, the Holm-adjusted p is the stricter test: Feb remains significant after adjustment. λ is the weight given to climatology in the blend (higher = closer to climatology).
Historical verification explorer
The evidence behind the scores above, for May 2026 in this area. Choose the evaluation period; all diagnostics use the blended probabilities against climatology.
Performance by year
Probability skill (RPSS) of each year; better than climatology in 17 of 24 years. Weakest years: 1999 (−0.468), 1996 (−0.119), 1995 (−0.073).
Reliability
Points on the dashed diagonal are reliable: events forecast with probability p happen p of the time. Points flatter than the diagonal mean over-confident probabilities. Open points have fewer than 134 cell-years (0.5% of the sample, at least 50) and are not joined. Hover a point for its count.
Reliability table (forecast bins, observed frequency, counts)
| Category | Bin | Mean forecast | Observed frequency | Cell-years |
|---|---|---|---|---|
| below normal | 10%–20% | 18% | 52% | 853 |
| below normal | 20%–30% | 26% | 29% | 6,440 |
| below normal | 30%–40% | 35% | 33% | 15,294 |
| below normal | 40%–50% | 43% | 47% | 4,248 |
| below normal | 50%–60% | 51% | 38% | 45 |
| near normal | 10%–20% | 19% | 19% | 207 |
| near normal | 20%–30% | 27% | 25% | 10,187 |
| near normal | 30%–40% | 34% | 33% | 13,313 |
| near normal | 40%–50% | 43% | 38% | 2,371 |
| near normal | 50%–60% | 53% | 40% | 787 |
| near normal | 60%–70% | 63% | 0% | 15 |
| above normal | 10%–20% | 19% | 89% | 19 |
| above normal | 20%–30% | 27% | 27% | 7,613 |
| above normal | 30%–40% | 34% | 37% | 15,382 |
| above normal | 40%–50% | 43% | 47% | 3,413 |
| above normal | 50%–60% | 53% | 7% | 426 |
| above normal | 60%–70% | 63% | 7% | 27 |
How strong are the signals?
Share of cell-years by the leading tercile probability.
| Category | Forecast | Climatology | Brier skill |
|---|---|---|---|
| below normal | 0.2262 | 0.2265 | +0.002 |
| near normal | 0.2109 | 0.2122 | +0.006 |
| above normal | 0.2239 | 0.2268 | +0.013 |
Rainfall amounts
Historical rainfall-amount diagnostics (CRPS, CRPSS and amount errors) have not been generated for May 2026 in this area and period; this needs a separate cross-validated amount evaluation. Single-season amount scores are in the Verification section.
Performance by year
Probability skill (RPSS) of each year; better than climatology in 8 of 9 years. Weakest years: 2021 (−0.143), 2025 (+0.002), 2020 (+0.013).
Reliability
Points on the dashed diagonal are reliable: events forecast with probability p happen p of the time. Points flatter than the diagonal mean over-confident probabilities. Open points have fewer than 50 cell-years (0.5% of the sample, at least 50) and are not joined. Hover a point for its count.
Reliability table (forecast bins, observed frequency, counts)
| Category | Bin | Mean forecast | Observed frequency | Cell-years |
|---|---|---|---|---|
| below normal | 20%–30% | 26% | 9% | 2,413 |
| below normal | 30%–40% | 34% | 27% | 6,766 |
| below normal | 40%–50% | 42% | 36% | 897 |
| below normal | 50%–60% | 51% | 0% | 4 |
| near normal | 20%–30% | 28% | 31% | 2,072 |
| near normal | 30%–40% | 34% | 42% | 6,825 |
| near normal | 40%–50% | 43% | 36% | 861 |
| near normal | 50%–60% | 53% | 39% | 322 |
| above normal | 20%–30% | 27% | 35% | 2,794 |
| above normal | 30%–40% | 34% | 36% | 6,626 |
| above normal | 40%–50% | 43% | 54% | 660 |
How strong are the signals?
Share of cell-years by the leading tercile probability.
| Category | Forecast | Climatology | Brier skill |
|---|---|---|---|
| below normal | 0.1835 | 0.1900 | +0.034 |
| near normal | 0.2422 | 0.2421 | 0.000 |
| above normal | 0.2337 | 0.2346 | +0.004 |
Rainfall amounts
Historical rainfall-amount diagnostics (CRPS, CRPSS and amount errors) have not been generated for May 2026 in this area and period; this needs a separate cross-validated amount evaluation. Single-season amount scores are in the Verification section.
Historical performance by rainfall region
Probability skill by rainfall region (full regional matrix)
| Region | FMAM | Feb | Mar | Apr | May |
|---|---|---|---|---|---|
| All Ethiopia | +0.021 (−0.031 to +0.069) 15/24 yrs | +0.126 (+0.043 to +0.217) 18/24 yrs | +0.011 (−0.017 to +0.038) 15/24 yrs | +0.009 (−0.034 to +0.044) 15/24 yrs | +0.008 (−0.042 to +0.047) 16/24 yrs |
| R0 arid / marginal | +0.023 (−0.072 to +0.103) 15/24 yrs | +0.057 (−0.081 to +0.206) 13/24 yrs | −0.005 (−0.058 to +0.044) 15/24 yrs | −0.028 (−0.102 to +0.038) 13/24 yrs | −0.015 (−0.079 to +0.040) 13/24 yrs |
| R1 western unimodal | +0.019 (−0.035 to +0.067) 18/24 yrs | +0.156 (+0.068 to +0.248) 18/24 yrs | +0.009 (−0.018 to +0.034) 15/24 yrs | +0.016 (−0.024 to +0.047) 17/24 yrs | +0.009 (−0.033 to +0.046) 15/24 yrs |
| R2 Belg–Kiremt | +0.037 (−0.050 to +0.108) 17/24 yrs | +0.166 (+0.076 to +0.264) 20/24 yrs | +0.023 (−0.017 to +0.058) 16/24 yrs | −0.001 (−0.063 to +0.052) 14/24 yrs | +0.009 (−0.051 to +0.059) 16/24 yrs |
| R3 Gu–Deyr | +0.012 (−0.042 to +0.064) 15/24 yrs | +0.082 (−0.006 to +0.180) 14/24 yrs | +0.006 (−0.030 to +0.041) 14/24 yrs | +0.017 (−0.034 to +0.064) 16/24 yrs | +0.009 (−0.066 to +0.062) 19/24 yrs |
| FMAM rainfall contribution domain | +0.017 (−0.038 to +0.067) 15/24 yrs | +0.137 (+0.050 to +0.234) 16/24 yrs | +0.011 (−0.020 to +0.042) 15/24 yrs | +0.011 (−0.032 to +0.048) 16/24 yrs | +0.007 (−0.051 to +0.050) 17/24 yrs |
| FMAM main-season rainfall domain (Belg / Gu / Ganna, >=40% of annual) | +0.013 (−0.040 to +0.066) 16/24 yrs | +0.080 (−0.005 to +0.176) 14/24 yrs | +0.007 (−0.029 to +0.042) 15/24 yrs | +0.013 (−0.041 to +0.060) 17/24 yrs | +0.009 (−0.066 to +0.062) 18/24 yrs |
Positive values mean lower RPS than climatology. Intervals resample whole years; regions are the fixed 1993–2025 rainfall regimes, and the last rows are this cycle's rainfall domains. Regional results are noisier than national ones.