Figure 2 Hovmo1ler diagrams of upper ocean heat content anomaly (change in the 400m depth-averaged temperature) at the equator in the Pacific Ocean for (a) the total data base, denoted ALL DATA, (b) VOS XBT data alone, (c) TAO data alone, and (d) as (a) but with 70% of the stations removed (see text for details). Light regions denote cool conditions, dark regions warm (ENSO-like) conditions. The 1972, 1982-3, 1986-87 and 1991-92 warm events are readily distinguished. in the full data case. The VOS XBT and TAO alone analyses are performed only for the period 1985-1994, and the 70% reduced data set analysis for 1980-1994.
the amplitude of the low frequency signal is well determined. VOS on the other hand have generally better spatial sampling (though not adequate) but repeat only infrequently in time, so neither the amplitude nor phase of the adjusting waves is well determined. Such a sampling pattern beckons for true space-time interpolation as offered by an adjoint data assimilation system.
The 70% reduced data set still resolves a considerable portion of the seasonal-to-interannual variability, though the resolved signal is quite weak through the early and mid