A 72-month negative PDO run had tail fractions of 0.029 and 0.114 under pre-2020 and full-record models

We tested the longest negative monthly PDO run in a fixed NOAA ERSSTv5 record for 1950-2025. The 72-month run from January 2020 through December 2025 reaches the record boundary. We fitted monthly persistence and lagged Nino 3.4 models and simulated 19,999 records per check, holding ENSO history fixed. Each simulation included a search for the longest run over the full record. The full-record AR1 and ENSO fit gave a tail fraction of 0.11425. The same model fitted through 2019 gave 0.02880. Three- and twelve-month residual blocks gave similar values within each fitting period. Other full-record checks gave 0.09080 to 0.19845. Thus evidence for an unusual run depends on the fitting period. These are conditional fitted-model tail fractions, not exact p values or proof of a regime shift. Fixed coefficients, one revised index, limited residual memory and an unfinished run limit inference. The package includes raw tables, hashes, code, all simulation maxima, a numerical audit and a figure.

PDO · ENSO · negative runs · persistence · model sensitivity · ocean conditions