Predictability of SST in a Stochastic Climate Model and Its Application to the Kuroshio Extension Region

Abstract

The influence of deterministic forcing on SST predictability is investigated in a zero-dimensional, stochastic, coupled atmosphere?ocean climate model. The SST anomaly predictability time is found to be very sensitive to the properties of the deterministic forcing. Comparison of the amplitudes of the deterministic and stochastic forcing terms, for example, as determined from linear regression analysis, may give a misleading impression of their relative importance. The importance of instead comparing the time-integrated forcing terms is emphasized. The conditions under which the model exhibits preferred timescales and the conditions under which the model power spectrum approaches that of a univariate Markov process (red noise) are also determined. The idealized model results are complemented with an analysis of climate observations for the Kuroshio Extension region. Observational errors and unresolved components of the enthalpy budget limited the maximum timescale considered to about 4 yr. This analysis revealed that the advection of anomalous geostrophic currents is a minor source of SST variability and not the limiting factor in determining SST predictability in that region, at least for the timescales considered.