Sea level fluctuations have many causes.3 Daily tides are associated with a divergence in horizontal volume flux with no attendant significant change in density. Direct atmospheric heating and cooling of the water column and exchanges of fresh water are associated with expansion and contraction on time scales from days to millennia. Local changes in temperature and salinity, and hence of density, are also associated with lateral shifts due to ocean currents. Changes with no immediate density signature, as with the tides, are “barotroprc” and are not directly relevant to inferences about stored heat; otherwise, changes are “baroclinic.” For example, the observed secular rise in sea level4 is a combination of the melting of glaciers (barotropic) and thermal expansion. Determining the relative contributions of barotropic and baroclinic processes on the myriad time scales of climate change is complex.
Theoretical studies provide some clues as to the relative importance of barotropic and baroclinic fluctuations. A recent model study5 suggested that wind-driven changes in ocean circulation are largely baroclinic in the tropics but barotropic at higher latitudes. Observations with sparse current meter moorings in the North Pacific6 show that on a time scale of 100 days the relative contribution of barotropic processes varied between 10 and 70%, depending upon location. Here we examine the evidence on time scales from months to years.
OBSERVATIONS AND MODEL
Acoustic Component. Ocean acoustic tomography7 has the ability to sample and average the large-scale oceanic thermal structure, synoptically, along several sections at regular intervals. In late October 1995, the ATOC program deployed an acoustic source at a depth of 939m on Pioneer Seamount, 100km west of San Francisco, Califomia.8 Transmissions began in December 1995, and the transmitted signals have been received on various arrays consisting of those mounted on the sea floor (for example, k, l, n, and o in
Fig. 1) and of two 40-hydrophone vertical line arrays (v1 and v2). Vertical arrays permit the detailed study of the received acoustic signals. The transmission schedule typically consists of 4-day periods, two to four times a month.9 Transmissions are spaced 4 hours apart during transmission periods. A total of 772 transmissions in 43 groups was made between December 1995 and March 1997. A recognizable one-to-one correspondence exists between the observed and predicted ray arrivals10 (Fig. 2). This and other work11 show that at these ranges, ray arrivals are resolvable, identifiable, and stable. Ray arrivals were tracked and then used to infer range-averaged profiles of sound speed and temperature along each section.12 Despite the presence of mesoscale eddies and internal waves, the arrival times vary smoothly through the months as a result of the spatial integration.
The vertical resolving power of the acoustic data is determined by the ray structure. For sections k and l, all identified rays are steep and surface reflecting, and they have lower turning depths between 2000 and 3500m. For sections n and o, identifiable rays begin as surface reflecting near the source and change to near-surface refracting as they approach the receivers. Section v1 has both surface-reflecting and purely refracting rays.
The travel times for section k Fig. 2), and with one exception all other sections, decreased in the summer and increased in the winter, consistent with the expected seasonal heating and cooling of the surface layer. For section v1, the situation is different: maximum heat content was recorded in March 1996. From a comparison of sections o and v1 Fig. 3), we infer that the winter surface layer cooling near the source, where the two sections overlap, was more than offset by a subsurface warming near receiver v1, where the rays do not sample the surface layer.
