Linear autumn ENSO forecasts gave no clear gain for a California-minus-Washington winter precipitation index in 1981-2026

We tested whether autumn Nino 3.4 improves forecasts of a California-minus-Washington January-March precipitation index. We used all 17 NOAA climate divisions in the two states, with fixed 1951-1980 scales and equal division weights. Every forecast used earlier target years only. In 46 test winters, 1981-2026, linear model mean squared error skill was -1.36% against the earlier-year mean. The conditional 95% five-year block interval was -27.97% to 20.44%. The skill difference against the common component was 0.07 percentage points, with an interval of -23.68 to 19.92 points. California and Washington fitted ENSO slopes had opposite signs at every issue. This did not establish a forecast gain. Planned model, region, season, and block-length checks left the main comparison uncertain. The result does not prove that ENSO has no useful information. Revised data, fixed index weights, a short record, simple models, and conditional intervals limit inference. This study does not measure water supply or identify a physical cause. The package gives exact source rows and hashes, code, all forecasts and resamples, two figures, a scalar audit, and an exact offline repeat.

ENSO · California · Washington · winter precipitation · regional contrast · chronological forecast · null result · nClimDiv · Nino 3.4