Freshening of the Upper Pycnocline in the North Pacific Subtropical Gyre Associated With Decadal Changes of Rainfall

The Hawaii Ocean Time-series (HOT) project has observed full-depth water mass variability at 22° 45’N, 158° W since October 1988. A pronounced freshening of the upper pycnocline started between 1991 and 1994, continuing into 1998 when the ENSO-related drought near Hawaii resulted in a marked increase of salinity. The freshening was most pronounced (~0.15 psu) near 25 s_q just below the salinity maximum; its phase appears progressively later on deeper isopycnals. The first multivariate EOF of salinity, dissolved oxygen and potential vorticity over the potential density range 24.1–27.3 s_q explains 36% of the non-seasonal variance, and it includes the trend-like freshening since 1991. The freshening was accompanied by a decrease in dissolved oxygen and an increase in potential vorticity over 24–26.5 s_q . The time scale of the salinity signal is decadal.

An EOF analysis of North Pacific rainfall (using the CMAP rainfall dataset) shows that the first two EOFs explain 26% and 17% of the variance respectively. The first EOF is dominated by ENSO events, and the second corresponds to the Pacific Decadal Oscillation. Both involve modulations of rainfall in the midlatitude Pacific storm track region.

One hypothesis for the pycnocline freshening is that ventilation of the upper pycnocline has carried the signature of midlatitude North Pacific rainfall anomalies into the pycnocline. This is consistent with observed rainfall signals and with the delay in the arrival of the signal at the HOT site with depth, because advective pathways are slower with depth. Another hypothesis is that gyre "wobble" associated with long, low-frequency Rossby waves results in meridional displacement of the mean salinity, oxygen and potential vorticity fields. This hypothesis is consistent with the multivariate EOF structure of the freshening signal, but it cannot account for the delay of the signal with depth. Because the two hypotheses are not mutually exclusive, it is necessary to combine skillful ocean models with high quality surface forcing information to estimate the relative contributions of the hypothesized processes to the observed signal.