Winter rain gave no clear summer rainfall forecast gain in three British countries in 1961-2025

We tested whether December-February rainfall improved forecasts of June-August rainfall in England, Wales and Scotland. We used fixed Met Office HadUK-Grid country records and earlier years only for each fit. There were 65 test years, 1961-2025, and 195 country forecasts. A linear winter term raised mean squared error by an equal-country mean of 6.12% against a mean of the same 30 training summers. The conditional 95% five-year block interval for error reduction was -13.75% to +0.17%. Each country had negative point skill and an interval that included zero. All planned checks also gave negative mean point skill. A quadratic winter model raised error by 15.66%, with conditional limits of 4.40% to 32.79%. The primary result gives no clear forecast gain under this simple model. It does not prove seasonal independence or measure water storage. Country averages, revised source data, model choice and intervals conditional on fitted forecasts limit inference. The package includes exact raw snapshots, source hashes, monthly inputs, all forecasts, uncertainty draws, code, an arithmetic audit and a figure.

rainfall · seasonal prediction · Britain · chronological forecast · HadUK-Grid · null result