A.1. A XBT data
The principal thermal data set used in this study is the historical XBT compilation of Donoso et al. (1994), which will be referred to here as the "AOML XBT data". These profiles originated from earlier compilations produced principally by the National Ocean Data Center, the Joint Environmental Data Center at Scripps, and the Global Subsurface Data Center in Brest, France. A total of 36,185 profiles were available within 15°S–25°N, and from 120°W to the American coast; as is typical of volunteer observing ship data, the profiles were primarily organized along the major shipping routes. The main concentration of profiles were taken during 1979–1992, with smaller numbers after that date. Donoso et al. (1994) describe the objective and subjective quality control procedures used to flag questionable observations and produce a research-quality data set. Only data points they found to be acceptable were used in this study.
The irregularly distributed profiles were mapped to a regular 1° latitude by 1° longitude by 10 m by climatological-month grid in two steps. First, the temperature data in each individual profile were linearly interpolated to 10 m depth intervals from the surface to 450 m (on average, each profile had about 41 samples in the upper 450 m). Second, in a separate calculation at each 10 m level z, the irregularly spaced temperatures were mapped onto the (x, y, z, t) grid using a Gaussian-weighted three-dimensional mapping (Kessler and McCreary, 1993), with scales of 2° longitude, 1° latitude and 2 months. The region of influence of each observation was limited to where its weighting was greater than e, that is, to twice the scale distance from the center of each gridbox. Only the month was considered in the gridding, not the year, to produce a climatological average annual cycle. These data were previously used in a study of the mean circulation (Kessler, 2002), and the same gridding procedure was used here.
Although some earlier XBT-based climatologies of the entire tropical Pacific used coarser grid spacings [e.g., Kessler, 1990, who mapped onto a 5° longitude by 2° latitude grid], the short spatial scales of the phenomena of interest in this region required higher resolution, and the profile distribution in the new compilation was dense enough to justify a 1° × 1° spacing. Each 1° × 1° gridbox contained, on average, more than 24 profiles during the year, spread over an average of 7.4 months. The gridded temperature climatology was checked by comparison against other sources, including the Reynolds SST (Reynolds and Smith, 1994) and time series from the TAO mooring array (McPhaden et al., 1998).
Dynamic height relative to 300 or 450 m reference levels was found from the gridded temperatures via a mean T-S relation constructed from the Levitus et al. (1994) atlas, and geostrophic currents were found from centered differences of the dynamic heights.
An additional meridional section at 125°W, shown in Fig. 6, was constructed from the XBT data compilation of Kessler (1990). These data partly overlap the AOML XBT compilation, but extend further west. Quality control and gridding is described by Kessler (1990); of note is that the lesser data density required a coarser gridding as mentioned above. Along the 125°W section shown, 50–150 profiles were found in each 2° latitude by 5° longitude box, roughly half as dense than the AOML data.
Return to previous section or go to next section
PMEL Outstanding Papers
PMEL Publications Search