September DMI gave no clear added dry-season probability skill in eight East Africa cells in 2000-2025

We tested whether September Indian Ocean Dipole data improve dry-season probability forecasts in eight GPCP cells at 5 S-5 N, 35 E-40 E. A dry season has October-December rain at or below the earlier-year one-third quantile in each cell. Each logistic model used earlier years from 1979. The baseline used September local rain, September Nino 3.4 and a linear trend. Adding September DMI reduced area-weighted Brier error by 1.90% in 2000-2025. The conditional 95% three-year block interval was -2.78% to +6.04%. Removing the trend gave 0.04% added skill. Four cell point estimates were positive and four were negative. A prior campaign study found a larger gain for mean rain in the same box, but that score has a different target. This test gives no clear added probability skill under the specified models. It does not rule out useful ocean information. Only 26 years, eight related cells, revised products, simple models and conditional uncertainty limit inference. The package gives source hashes, monthly data, all forecasts, code, a figure and a second-solver audit.

East Africa · Indian Ocean Dipole · dry season · Brier score · probability forecast · GPCP · null result · chronological validation