Winter totals gave no clear summer precipitation forecast gain in seven Arizona divisions in 1951-2025

We tested whether November-March precipitation improves a linear forecast of July-September precipitation in seven Arizona climate divisions. We used a fixed NOAA nClimDiv snapshot and earlier years only for each fit. There were 75 test years, 1951-2025, and 525 forecasts. Adding winter totals reduced mean squared error by an equal-division mean of 2.66% against the earlier summer mean. The conditional 95% five-year block interval was -1.86% to 7.17%. All seven division intervals included zero. Planned checks with other seasons, a recent training window, a year term, other blocks and shorter test periods also left the gain uncertain. Full-record winter-summer correlations were negative, from -0.220 to -0.113. The association thus did not establish a clear forecast gain under this model and score. The result does not show independence or rule out useful forecasts with other predictors. One retrospective product, spatial aggregation, simple model form and conditional intervals limit inference. The package includes exact source rows, hashes, forecasts, uncertainty draws, figures, code, an independent arithmetic audit and an exact offline repeat.

Arizona · precipitation · monsoon · seasonal compensation · chronological forecasts · nClimDiv · null result