Thirteen Years Below the Median: a Persistent Surface-Salinity Departure of the Subpolar Gyre in a Cached Ocean Reanalysis, Read Against Its History and Two Observational Comparisons
Year: 2026
Type: article
Abstract: A monthly record of surface salinity over the North Atlantic subpolar gyre, 1958 to 2024, cached from the ORAS5 ocean reanalysis for an earlier evaluation and reused in a monitoring experiment, holds a departure that the experiment surfaced and this paper describes. Every one of the 156 months from January 2012 to December 2024 lies below the record's own median over January 1979 to May 2004, in every calendar month of the year, and the 127 consecutive 36-month windows ending June 2014 to December 2024 each have all 36 months below it: the longest such run in the record, right-censored at its end. The record reached the same state once before, in the 22 windows ending June 2001 to March 2003, and never in the salty 1960s. The departure is present in 2010 and 2011 and in every month of 2012 to 2014, before the reanalysis changes product in January 2015, so the product boundary cannot explain its presence; what the boundary does to the series after 2015 is left open. The statistic is deliberately plain: a month is below the median or not, a window's fraction is the share of its months below, and the counts are counts of overlapping windows, with no probability attached. Two observation-based datasets were read against the record under comparison rules fixed before either was scored, each series against its own median over its own reference span. A ship-track surface product on two ship-of-opportunity routes crossing the gyre has 10 of twelve months below its own median in 2010, 11 in 2011, 2 in 2014 and all twelve in 2013 and in 2016 to 2018: its below-median months overlap the record's in timing, with a 2014 the record does not show; of its windows ending in the record's later run, 12 of 55 have all 36 months below, and of those ending in the earlier run, none of 22. The EN4 objective analysis over the same box, at the surface and averaged over the upper 500 m, has no all-below window in either run, although its upper-500 m series has 57 of the 60 months of 2015 to 2019 below its own median; the correlation of its monthly anomalies with the record's is 0.236 at the surface and 0.304 at depth over the whole record, and EN4 shared its profiles with the reanalysis until May 2015, so agreement before that date is not independent. Published observations document a large freshening of the eastern subpolar gyre from 2012 and of the Irminger Sea from 2015, overlapping the record in timing on their own supports; the literature search made for this paper found none documenting a gyre-wide surface freshening in 2000 to 2003, so the earlier run stands as a feature of the cached record, uncorroborated at the surface. The paper claims the cached pattern and its limited reproduction across these comparisons: a persistent departure in one record, overlapping in timing with direct observations and not matched in the persistence of its all-below ordering by either comparison dataset. It claims no physical cause, no onset date, no amplitude, nothing about the Atlantic meridional overturning circulation, and no probability. The monitoring experiment that surfaced the pattern, a standard feature-drift monitor on the gyre's temperature and salinity state read against a simple overturning forecast, gave two limited nulls on this record, no forecast improvement from the climate features and no validated alert rate, and that outcome is reported as the ordinary result of its kind. Every quantity quoted from the analysis is emitted from the recorded results by a tabulation step; the comparison rules, the missingness rules and the construction of each series are stated in full.
Author: Merrill, Christopher K.
Language: English
Subfield: Oceanography
Field: Earth and Planetary Sciences
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W7225346215