Upper mixed layer temperature anomalies at the North Atlantic storm-track zone

Abstract. Synoptic sea surface temperature anomalies (SSTAs) were determined as a result of separation of time scales smaller than 183 days. The SSTAs were investigated using daily data of ocean weather station "C" (52.75°N; 35.5°W) from 1 January 1976 to 31 December 1980 (1827 days). There were 47 positive and 50 negative significant SSTAs (lifetime longer than 3 days, absolute value greater than 0.10 °C) with four main intervals of the lifetime repetitions: 1. 4–7 days (45% of all cases), 2. 9–13 days (20–25%), 3. 14–18 days (10–15%), and 4. 21–30 days (10–15%) and with a magnitude 1.5–2.0 °C. An upper layer balance model based on equations for temperature, salinity, mechanical energy (with advanced parametrization), state (density), and drift currents was used to simulate SSTA. The original method of modelling taking into account the mean observed temperature profiles proved to be very stable. The model SSTAs are in a good agreement with the observed amplitudes and phases of synoptic SSTAs during all 5 years. Surface heat flux anomalies are the main source of SSTAs. The influence of anomalous drift heat advection is about 30–50% of the SSTA, and the influence of salinity anomalies is about 10–25% and less. The influence of a large-scale ocean front was isolated only once in February-April 1978 during all 5 years. Synoptic SSTAs develop just in the upper half of the homogeneous layer at each winter. We suggest that there are two main causes of such active sublayer formation: 1. surface heat flux in the warm sectors of cyclones and 2. predominant heat transport by ocean currents from the south. All frequency functions of the ocean temperature synoptic response to heat and momentum surface fluxes are of integral character (red noise), though there is strong resonance with 20-days period of wind-driven horizontal heat advection with mixed layer temperature; there are some other peculiarities on the time scales from 5.5 to 13 days. Observed and modelled frequency functions seem to be in good agreement.