An ENSO change term raised spring rain forecast error across 344 US divisions in a 1981-2025 linear test

We tested whether the change in Nino 3.4 from autumn to winter improves forecasts of March-May precipitation. We used fixed NOAA ERSSTv5 and nClimDiv records for 344 contiguous US climate divisions. Each linear fit used earlier years only. The baseline used winter rainfall, winter Nino 3.4, its absolute value and a time trend. Adding the change term raised mean scaled squared error by 2.23%. The conditional 95% five-year block interval for the increase was 0.19% to 4.90%. Both fitted models had more error than the earlier-year mean forecast. The three-year block interval and later-period interval included zero. This is a retrospective benchmark with 45 test years, not an operational forecast or a test of physical cause. One data pair, simple models, shared training sets and conditional uncertainty limit inference. The package gives monthly evidence, all predictions, code, source hashes, a figure and a numerical audit.

ENSO · spring precipitation · forecast validation · Nino 3.4 · negative result · United States