SCALES OF VARIABILITY IN THE EQUATORIAL PACIFIC INFERRED FROM THE TROPICAL ATMOSPHERE-OCEAN BUOY ARRAY

The highly temporally resolved time series from the Tropical Atmospher e-Ocean moored buoy array are used to evaluate the scales of thermal v ariability in the upper equatorial Pacific. The TAO array consists of nearly 70 deep-ocean moorings arranged nominally 15 degrees longitude and 2 degrees-3 degrees latitude apart across the equatorial Pacific. The bulk of the data from the array consists of daily averages telemet ered in real time, with some records up to 15 years long. However, at several sites more finely resolved data exist, in some cases with reso lution of 1 minute. These data form the basis for spectral decompositi on spanning virtually all scales of variability from the Brunt-Vaisala frequency to the El Nino-Southern Oscillation timescale. The spectra are used to define the signal to noise ratio as a function of sample r ate and frequency, and to investigate the effects of aliasing that res ults from sparser sampling, such as ship-based observational technique s. The results show that the signal to noise ratio is larger in the ea st, mostly because the low-frequency signals are larger there. The noi se level for SST varies by as much as a factor of 10 among the locatio ns studied, while noise in thermocline depth is relatively more homoge neous over the region. In general, noise due to aliased high-frequency variability increases by roughly a factor of 10 as the sample rate de creases from daily to 100-day sampling. The highly resolved spectra su ggest a somewhat more optimistic estimate of overall signal-to-noise r atios for typical ship of opportunity (VOS) XBT sampling (generally ab out 2) than had been found in previous studies using sparser data. Tim e scales were estimated for various filtered versions of the time seri es by integration of the autocorrelation functions. For high-passed da ta (periods longer than about 150 days removed), the timescale is abou t 5 days for both surface and subsurface temperatures everywhere in th e region. Conversely, for low-passed data (the annual cycle and period s shorter than 150 days removed), the timescale is roughly 100 days. H orizontal space scales were estimated from cross-correlations among th e buoys. Zonal scales of low-frequency SST variations along the equato r were half the width of the Pacific, larger than those of thermocline depth (about 30 degrees-40 degrees longitude). In the east, meridiona l scales of low-frequency SST were large (greater than about 15 degree s latitude), associated with the coherent waxing and waning of the equ atorial cold tongue, whereas in the west these scales were shorter. Th ermocline depth variations had meridional scales associated with the e quatorial waves, particularly in the east. Spatial scale estimates rep orted here are generally consistent with those found from the VOS data sets when the ENSO signals in the records of each dataset are taken in to account. However, if signals with periods of 1 to 2 months are to b e properly sampled, then sampling scales of 1 degrees-2 degrees latitu de by 8 degrees-10 degrees longitude, with a 5-day timescale, are need ed.