A fixed 27.5 C SST hinge lowered eastern Pacific monthly rain test error in 2002-2025

We tested a fixed sea surface temperature (SST) hinge for concurrent monthly rain in 5 S-5 N, 150 W-90 W. We used fixed GPCP V2.3 and ERSSTv5 products. Both models were fit on 276 months in 1979-2001 and tested on 288 months in 2002-2025. The baseline had a separate intercept and SST slope for each calendar month, plus a time trend. One added slope above 27.5 C reduced test mean squared error by 71.6%. RMSE fell from 1.191 to 0.634 mm/day. The conditional 95% paired 36-month block interval for error reduction was 59.9% to 79.7%. Fixed checks with nearby thresholds, no trend, and omitted warm-event test years retained positive gains. A quadratic term also gave a gain, so the result does not uniquely support a hinge. The test uses concurrent SST and revised products. It gives no forecast skill or proof of a physical threshold. Short records, product changes, unequal native grids and conditional uncertainty limit inference. Code, selected cells, source hashes, estimates, bootstrap scores, a figure and an independent audit permit an offline repeat.

SST · eastern Pacific · rainfall · GPCP · ERSSTv5 · threshold · chronological validation