For seasonal forecasts a good ocean initial state is essential. This is obtained by combining observations of the ocean with the dynamics of an ocean model. The observations are generally sea surface temperature measurements, in-situ subsurface temperature measurements in some regions, and satellite altimetry measurements representing a measure of integrated heat content. To optimally combine the observations with an OGCM a 4DVAR data-assimilation scheme is implemented that varies the momentum, heat and freshwater forcings of the ocean model over 16-week periods.
To gain insight in how the method adjusts the ocean analysis, several identical twin experiments are carried out in which a model run serves as the `truth'. Either the initial state or the forcing is perturbed with respect to this `truth'. It is then approximated again by assimilation of sea surface temperature data, subsurface temperature data and/or sea surface height data generated by the truth run but with the same properties as real observations.
The method provides an ocean analysis in which the temperature structure of the ocean is improved. The assimilation of either altimetry data or subsurface temperature data gives comparable reductions in all cost functions. However, the analysis does depend on which dataset is assimilated. Subsurface temperature data leave relatively small residuals near the equator at the cost of larger residuals at higher latitudes. Altimetry data result in a more uniform adjustment, in which the ocean analysis at depth is also improved.
Different fluxes alter the state of the ocean in different regions. The effect of changes in the zonal wind stress is most obvious in a band up to 30\dg\ from the equator and at thermocline depth. Adjusting the heat flux also has effects at higher latitudes. It has impact through the ocean mixed layer. The effect of freshwater flux variations is negligible. To make improvements below the thermocline and in chaotic regions, the initial state will also have to be adjusted.
G Appeldoorn, GJ van Oldenborgh. 4DVAR assimilation of subsurface and altimetry observations in the HOPE OGCM adjusting surface fluxes