space and time, and potentially from a variety of platforms, to produce regularly gridded temperature analyses which represent a temporal mean for some period. Smith and Meyers [1996] discuss results from such a study and we will revisit a selection of those results in this study. In the second approach (b) we use the same analysis system to produce an estimate of the upper ocean heat storage for each of the 12 months prior to the initialisation time of the model; a space-time trajectory of the model, forced by observed winds, is then sought such that it gives the best fit to the observed space-time pattern of heat storage, in effect assimilating the temperature data, and this trajectory is then used as the initial state for the coupled model forecast.
Potentially, we have a variety of data which might be used for producing temperature analyses and/or model forecast initial states:
surface winds;
SST;
subsurface temperature;
sea level (in situ and altimeter);
ocean currents;
sea surface salinity, subsurface salinity.
The various subjective assessments done during and after TOGA have ranked these sources of information in roughly the order presented above [OOSDP 1995; NRC 1994]. In this paper we are going to focus on the third source, subsurface temperature data, and (i) show we can quantify the amount and quality of information provided by the TAO and vos XBT networks, and (ii) demonstrate the sensitivity of the Kleeman et al. [1995] model to changes in these networks. The more general problem awaits future studies.
Three different sampling strategies are considered for the analysis evaluation [Smith and Meyers 1996],
(1)TAO data alone;
(2)VOS XBT data alone; or
(3)all data.
For the experiments with the forecast model we also consider the additional cases,
(4)No subsurface data at all; and
(5)70% degradation of the total network.
Since the forecast model also uses winds through the initialisation phase, case (3) versus case (4) is in effect presenting the impact of subsurface data. Prior to TOGA, and particularly through the early to mid-1970's, the subsurface network sampling was relatively sparse; the first elements of what was to become TAO did not arrive till 1980, and the VOS program did not start in earnest in the tropical Pacific until the late 1970's.
