Using the ensemble technique, several practice at various institutions including National Centers for Environmental Prediction (NCEP/NOAA) demonstrated that ensemble numerical prediction is capable of providing useful forecast up to a seasons or more.
All the above atmospheric model integrations assume that the sea surface temperature is specified and known, in other word, this shows the upper limit of the climate predictability. For practical seasonal prediction, future sst is not known and must be predicted. This requires numerical ocean model. Since ocean model requires forcing by atmosphere at the surface to predict sst, and atmospheric model requires sst to predict the forcing, it is necessary to deal with the two models at once, i.e., the two models need to be coupled. Since the coupling allows sst to vary freely, the simulation becomes more easily drift away from reality, and small deficiency in both models tend to magnify. The coupling of the two models is now the most important activities in the climate prediction activities.
The next event that advanced the climate forecast was the deployment of TAO array of moored buoys in the equatorial Pacific under Tropical Ocean and Atmosphere (TOGA) program. These buoyes provided crucial information on the spatial and temporal structure of El Nino and La Nina events. The use of buoy observation in the initial analysis of ocean was shown to improve the ocean prediction significantly (Ji and Leetmaa, 1997).
The atmospheric prediction as well as the ocean prediction require initial conditions of winds, temperature, pressure and moisture for the former and temperature, ocean flow and salinity for the latter. These variables need to be specified over the three dimensional space. The recent development of 'four dimensional variational analysis' allows more accurate analysis when the data is available in four dimensions, i.e. also in time. The specification of initial condition for ocean is crucial in determining how accurately ocean forecast is made. In fact, current ocean prediction is severely affected by the inaccuracies in the initial condition of the ocean, thus, it is necessary to provide accurate three dimensional structure of ocean, including temperature, salinity and motion in high spatial resolution covering the entire global ocean. This is the place where ARGO program comes in. The Argo is currently the only realizable program that makes it possible to observe three dimensional structure of the ocean with high spatial resolution that can be used in real time climate prediction. This observation program is indispensable in making ocean prediction more accurate and eventually making climate prediction of atmosphere more dependable. Such program requires close international coordination and support, which in return, provide truly useful climate prediction to the international community.