This paper analyzes the ability of three conceptual stochastic models (one-box, two-box, and diffusion models) to reproduce essential features of sea surface temperature variability on intra-annual time scales. The variability of sea surface temperature, which is particularly influenced by feedback mechanisms in ocean surface–atmosphere coupling processes, is characterized by power spectral density, commonly used to analyze the response of dynamical systems to random forcing. The models are aimed at studying local effects of ocean–atmosphere interactions. Comparing observed and theoretical power spectra shows that in dynamically inactive ocean regions (e.g., north-eastern part of the Pacific Ocean), sea surface temperature variability can be described by linear stochastic models such as one-box and two-box models. In regions of the world ocean (e.g., north-western Pacific Ocean, subtropics of the North Atlantic, the Southern Ocean), in which the observed sea surface temperature spectra on the intra-annual time scales do not obey the ν−2 law (where ν is a regular frequency), the formation mechanisms of sea surface anomalies are mainly determined by ocean circulation rather than by local ocean–atmosphere interactions. The diffusion model can be used for simulating sea surface temperature anomalies in such areas of the global ocean. The models examined are not able to reproduce the variability of sea surface temperature over the entire frequency range for two primary reasons; first, because the object of study, the ocean surface mixed layer, changes during the year, and second, due to the difference in the physics of processes involved at different time scales.
Read full abstract