September Nino 3.4 cut inland southern Africa rain forecast error in 2000-2024; adding tropical DMI raised error

We tested September ocean indices for December-February mean rain in an inland box at 20-30 S, 20-30 E. Each forecast used earlier seasons only from 1979. In 25 tests for December years 2000-2024, Nino 3.4 reduced mean squared error by 23.2% against the earlier-year mean. RMSE fell from 0.748 to 0.656 mm/day. The conditional 95% three-year block interval for error reduction was 7.6% to 38.1%. Adding tropical Indian Ocean DMI raised error by 8.7%, with a conditional interval for the increase of 1.7% to 15.0%. August inputs, trend terms, omission of two major training events and period checks kept both directions. These results give a saved forecast benchmark, not a new ocean-rain mechanism. The test has only 25 seasons, uses revised observations and does not establish real-time skill. It does not test the subtropical Indian Ocean dipole, local extremes or water storage. The package includes cell and monthly evidence, all forecasts, source hashes, code, uncertainty draws, a numerical audit and repeat instructions.

southern Africa · rainfall · seasonal prediction · ENSO · Indian Ocean Dipole · GPCP · chronological forecasts · added forecast information