Abstract: Much of the heat transported poleward by the oceans is carried in the midlatitude western boundary currents in the northern hemisphere. As these currents separate from the coastal boundaries and extend eastward into the ocean interior, they flux some of their heat to the atmosphere and store some of their heat in the recirculation gyres south of the current core; the heat content anomalies are negatively correlated with changes in the volume of an isothermal layer known as the "subtropical mode water". An analysis of upper ocean heat content observations (1955-2001) shows that there are substantial interannual variations in the amount of heat stored in the upper 400 m of the water column. About 26% of the variations in heat content in the North Atlantic and North Pacific Oceans (corresponding to the first principal component and with maxima in the western boundary current extension regions) are in phase and slightly lag the atmospheric Northern hemisphere Annular Mode (NAM or Arctic Oscillation). The simplest explanation, that changes in the westerlies cause corresponding changes in the air-sea fluxes and therefore in heat content, can be ruled out by by the sign of the correlation: strong westerlies (strong AO) are correlated with positive heat content anomalies. This conclusion is supported by previous analyses of the upper ocean heat budget, which show that the heat content anomalies are primarily caused by variations in ocean advection. The heat content anomalies, rather than being caused by changes in air-sea fluxes, instead appear to be the source of interannual variations in those fluxes. The magnitude of the flux anomalies, their association with advection and heat storage in the mode water, and the coherence between the two oceans suggest a role for ocean circulation in interannual to decadal variations in climate variability through local air-sea interaction.

Plate 1. Sea surface height maps from the TOPEX/POSEIDON altimeter. (left) Gulf Stream region in the North Atlantic and (right) Kuroshio Extension region in the North Pacific for years 1993, 1996, 1999, and 2001. Units are meters. More positive SSH indicates more heat stored in the ocean.

Figure 1. Heat storage rate for the western boundary current regions. (a) Gulf Stream and (b) Kuroshio Extension. The heat storage rate (bold line) is the sum of the surface heating (gray), the advection and diffusion (dashed), and the vertical motion of isotherms (dash-dot). Advection plus diffusion account for 70% of the heat storage rate variance in the Gulf Stream. The KE budget is more complicated, but advection-diffusion dominates after 1997. After Dong and Kelly [2004]; Vivier et al. [2002].

Figure 2. Heat content and mode water in the Gulf Stream region. The mean SSH (dash), heat content (solid), and the thickness of the 18\deg C layer (dash-dot) south of the Gulf Stream. The SSH is from the TOPEX/POSEIDON altimeter, and heat content and the layer thickness are from the GTSPP archive. After Dong [2004].

Figure 3. Heat budget of the western North Pacific: 1970-2000. The heat storage rate (bold line) is the sum of the surface heating (dash-dot), lateral (geostrophic) fluxes (dashed), and the sum of Ekman advection and a surface flux correction (thin line). Lateral fluxes are more highly correlated with the HSR than are surface fluxes. After Kelly [2004].

Plate 2. Empirical orthogonal functions of SSH. The (a) first and (b) second EOFs of SSH from the altimeter and (c) and (d) their respective time series. The first EOF describes SSH anomalies that are negative in the Pacific and weakly positive in the Atlantic. The second EOF describes SSH anomalies that are in phase in the two oceans and have their maxima in the western boundary currents. Indices of the Pacific Decadal Oscillation and the Arctic Oscillation are shown for comparison.

Plate 3. Empirical orthogonal functions of heat content. The (a) first and (b) second EOFs of JEDAC heat content and (c) and (d) their respective time series. The first EOF describes heat content anomalies that are in phase in the two oceans and have their maxima in the western boundary currents. The second EOF describes SSH anomalies that are negative in the Pacific and positive in the Atlantic. Indices of the Pacific Decadal Oscillation and the Arctic Oscillation are shown for comparison.

Figure 4. The Northern hemisphere Annular Mode and ocean heat content. (a) First EOF of sea level pressure (NAM). (b) Principal components of heat content (dash) and sea surface height (dash-dot). The sea level pressure (solid) time series is also known as the Arctic Oscillation Index. The correlation between the AO index and the heat content is 0.49, significant at better than 95%, with the AO leading heat content by 13 months.

Plate 4. NAM and heat content for low and high index periods. Maps of sea level pressure (contours) and ocean heat content (color) for (a) a low index period, 1985-1987, and for (b) a high index period, 1989-1991.

Figure 5. Wind stress curl and its relationship to heat content. (a) The negative wind stress curl in the North Atlantic [20-50N, 10-80W] (dash-dot), in the North Pacific [20-50N, 140E-130W] (dash), and the Arctic Oscillation Index (solid). Heat content (dash), negative wind stress curl (solid), and SSH anomaly (dash-dot) in (b) the North Pacific and in (c) the North Atlantic. Heat content and SSH are domain averages over 25-45N and 140-180E in the Pacific and 40-80W in the Atlantic. A second estimate for the North Atlantic heat content is shown (thin line) in (c).

Figure 6. Relationship of net surface flux to heat content. The surface flux anomaly for the year 2000, based on a regression with the time series of the heat content in Plate 3c. Negative values indicate flux of heat from the ocean to the atmosphere. There were positive heat content anomalies in the boundary currents in 2000.