1980's. From these results alone one might conclude that any period prior to 1983 is unlikely to have sufficient data to significantly impact the initialisation of a coupled model. However, recalling that the coupled model assimilation does in fact fit a space-time curve over 12 months, then perhaps there is some case for cautious optimism.
Smith and Meyers [1996] also examined specific periods to see whether either or both of the networks were critical for following the spatial pattems associated with the evolution of significant events. As noted above, it was clear that TAO was superior for delineating high and low-frequency variations in the equatorial region; VOS XBT assisted in filling out some of the zonal detail but it was not obvious this information was critical for monitoring or model initialisation. However some of the important annual and interannual adjustment processes also have important off-equatorial signatures and it is in this respect that VOS XBT data can play a significant role. Smith and Meyers [1996] showed that the XBT network data was critical for following the westward traveling equatorial Rossby wave that apparently led to the termination of the 199 1-92 warm phase [Kessler and McPhaden 1995]. Similar conclusions follow from the 400m depth-averaged temperature analyses (not shown). Smith and Meyers [1996] concluded that this difference suggested TAO and the VOS XBT were complementary, not competitive, with respect to their sampling.
Figure 3 shows the effective number of observations in the equatorial and tropical regions corresponding to the time periods and data sets used in constructing Figure 2. For the TAO and VOS XBT systems the conclusions, though not the actual numbers, are similar to that of Smith and Meyers [1996]. The XBT sampling peaked around 1987 and has been in gradual decline since, though this has not strongly impacted the net information. TAO took over as the primary subsurface information source in 1991, providing the bulk of data for the equatorial region and the largest proportion for the tropical region as a whole (for the 10 day 20℃ isotherm analyses, the tropical contributions are more comparable). The implication is that, taken over the 20°S-20°N tropical region, both of these networks are important sources of information, TAO principally in the equatorial region, VOS principally off-equatorial.
Figure 3 (a) Plot of the number of effective observations in the equatorial Pacific between 505 and 50N for the ALL DATA, vos XBT, TAO and "-70%" reduced data set experiments, expressed as a count per month over the period 1970-94. Only profiles that reach to 400m are included in the data sets. Note also that in order to ensure that, as far as was possible, no mooring data were inadvertently retained in the VOS XBT data set, some further non-TAO data were excluded. See text and Smith and Meyers [1996] and the text for an explanation of the effective observation count. (b) As (a) but for the region 20。? - 20。?.