-
Although no increased CO2 forcing is added in the CTRL and 0.1A experiments, AA is still achieved in the 0.1A experiment compared with the results in the CTRL experiment. Due to artificially decreased surface albedo, additional solar radiation is absorbed, which leads to net radiation addition at the top of atmosphere (Fig. 2a). Arctic surface warming (Fig. 2c) is achieved by additional solar radiation absorption induced by sea-ice loss (Fig. 1a in Dai, 2021). Influenced by ocean dynamics at different timescales, the sea surface temperature (SST) anomaly exhibits different spatial patterns in the first 20 years (positive SST anomaly in subpolar in NH), 21−300 years (positive SST anomaly equatorward expansion in mid-high latitude of SH) and 301−500 years (no significant change in SSTA Fig. 2b), which are defined as the early-transient stage, late-transient stage and equilibrium stage, respectively. Although SST in the 2xCO2 experiment evolves on different timescales (the early transient stage in 2xCO2 ends by the 80th year, while the equilibrium stage of 2xCO2 here is actually during the middle period of stage I in Yang et al., 2018; Fig. A1 in the Appendix), we still use the definition above for a better comparison. We also found that the stage definition is consistent with the features of AMOC index (rapid decrease, slow decrease and no obvious trend) in the 0.1A experiments (blue line in Fig. 1b in Dai et al., 2021). In this study, the mechanisms of surface warming in the early-transient stage and equilibrium state are revealed in section 3.1 and section 3.2, respectively. The corresponding anomalies in the Hadley cell and possible consequences are also studied, together with the potential influence of multiple-scale ocean dynamics.
Figure 2. Temporal changes in (a) globally integrated annual net radiation flux (black), net downward solar radiation (royal blue), and net outgoing longwave radiation (red) at the TOA (positive for downward anomaly; W m−2), zonally averaged annual (b) sea surface temperature anomaly (units: °C) and (c) surface air temperature anomaly (units: °C) in the 0.1A experiment.
-
In the 0.1A experiment, surface albedo over the ice-covered region is decreased artificially, and additional solar radiation is absorbed at the surface, especially in summer (approximately 67.4 W m-2 in the first 20 years). The additional solar radiation absorption is stored as seasonal heat storage in the subsurface ocean during summer and released to the atmosphere in autumn-winter (not shown), which leads to Arctic warming (66°−90°N) reaching its minimum (0.8°C) and maximum (2.3°C) in the warm season and cold season, respectively. Sea-ice melt (Fig. 3a, black line; different from current climate studies (Dai, 2022; Deng and Dai, 2022) in which annual sea-ice changes little) induced by Arctic warming produces freshwater (Fig. 3a, blue line) and decreases the water density in the North Atlantic (Fig. 3a, red line; Fig. 1a in Dai et al., 2021b), which inhibits deep-water formations and decreases AMOC from 22 Sv to 18 Sv in the first 20 years (nearly 20%; Fig. 1b in Dai et al., 2021). Following Yang et al., 2015, the meridional overturning circulation (MOC) and the associated heat transport are partitioned into Euler mean circulation, circulation induced by mesoscale eddies (Bolus term) and submesoscale eddies (submeso term) as well as a diffusion term. Although the results show that the MOC decreases significantly in the first 20 years, the meridional oceanic heat transport (OHT; Fig. 3b blue line) anomaly is mainly contributed by mesoscale processes (Fig. 3c short dash line) both in the Northern Hemisphere (NH) and Southern Hemisphere (SH), which is much larger than the contribution from mean circulation (Fig. 3c long dash line). Similar local forcing and feedback also occurred in the Antarctic region and resulted in surface warming in the high latitudes of the SH (Fig. 2c), which decreased the meridional temperature gradient and eastward wind.
Figure 3. (a) Zonally integrated ice-covered area anomaly (black line, units: 105 km2), zonally averaged sea surface salinity anomaly (blue line, units: PSU), and sea surface density anomaly (red line, units: kg m−3) in the 0.1A experiment. (b) Oceanic heat transport (OHT) anomaly (blue line, unit: PW, 1PW=1012 W), atmospheric heat transport anomaly (red line, units: PW), total meridional heat transport anomaly (black line, units: PW) and zonal mean surface air temperature anomaly (units: °C) in 0.1A are shown in (b). (c) OHT anomaly partitioned into Eulerian (long dash), Bolus (short dash), submesoscale and diffusion terms, two of which are shown, PW). The results are annual anomalies over the first 20 years in the experiments with a confidence level of 95% (Student’s t-test).
Both the equatorward wind-driven Ekman transport (Fig. 4c) and poleward eddy-induced circulation decrease in the Deacon cell, although the former dominates the Deacon cell anomaly and pushes it equatorward (Fig. 4b shading). In turn, the weakened Deacon cell decreases the subtropical cell, especially in the Pacific-Indian Ocean (Fig. 4c). The circulation anomaly mainly occurs in the lower layers, and mesoscale processes dominate the heat transport anomaly (0.5 PW at 30°S; 1PW=1012 W; Fig. 3b blue line). Arctic warming expands equatorward (as far as 50°N, Fig. 3b green line), and less poleward transport also warms the lower layers (approximately 1.0°C in the upper 500 m; Fig. 4a, shading). In this stage, the temperature profile (Fig. 4a shading) and upper-layer current (Fig. 4b shading) remain nearly unchanged at mid-low latitudes. As a result, there is no significant warming at the surface at mid-low latitudes (Fig. 3b green line). Since there is strong (weak) warming in polar regions (mid-low latitude), AA appears.
Figure 4. (a) Vertical profile of the temperature anomaly (shading, °C) in the Global Ocean and the climate mean temperature (contour). (b) Streamfunction of the global meridional overturning circulation anomaly in the 0.1A experiment is shown and the climate mean in the CTRL (contours, Sv.) (c) is the same as (b) but for the Eulerian part in the Pacific-Indian Ocean. All anomalies are shown at the 95% confidence level (Student’s t-test).
On the other hand, the meridional temperature gradient remains unchanged at low latitudes, as do the easterlies (via the thermal wind) and the subtropical cell (STC; anomaly is smaller than 1%) it drives (via Ekman transport) in the NH. The NH Hadley cell increases slightly (1%−2.4%; Figs. 5a, c) in spring and autumn, while it decreases slightly (1%−2%; Figs. 5e, g) in summer and winter. As a result, the strength of the annual mean Hadley cell in the NH remains nearly unchanged in the first 20 years (red line in Fig. 1b in Dai et al. 2021; seasonal Hadley cell anomalies in the early transient stage in 2xCO2 are also shown for comparison in Figs. 5b, d, f, h). This result is similar to the meridional atmospheric heat transport (AHT) at low latitudes (Fig. 3b, red line). Strong (weak) warming at high (mid) latitudes decreases the meridional temperature gradient (−), which decreases the zonal wind via the thermal wind. In turn, weaker westerly (jet shear) decreases the eddy activity (via barotropic instability) and meridional AHT (approximately 0.17 PW and 0.2 PW in the NH and SH, respectively; Fig. 3b red line). The strong warming in the Arctic also inhibits local subsidence and decreases the Polar cell, which in turn decreases the Ferrel cell at mid-high latitudes. As a result, the Hadley cell moves northward, except in winter (Table 1). In addition, the SH Hadley cell also strengthens slightly (0%−3% in each season, Table 1). The width of the NH Hadley cell can be explained via the Hadley cell expansion estimate (Held and Hou, 1980; Lu et al., 2007)
Figure 5. The seasonal atmospheric meridional mass streamfunction in the CTRL (contour) and differences between the sensitivity experiments and the CTRL (shading), taken from the first 20-year numerical results (109 kg s−1)
Experiment Component MAM JJA SON DJF CTRL stren-S −86.6 −222.47 −114.75 −34.97 stren-N 122.31 53.37 83.91 211.20 edge-S −30.98 −27.20 −28.5 −33.17 mid −5.39 14.54 8.1 −13.03 edge-N 23.25 27.42 30.91 26.95 0.1A stren-S −89.3 −223.52 −115.33 −34.82 stren-N 125.2 52.63 84.72 208.64 edge-S −30.4 −27.04 −28.12 −33.08 mid −4.8 14.93 8.3 −13.02 edge-N 23.21 27.73 30.57 26.74 2xCO2 stren-S −82.8 −220.15 −110.6 −36.93 stren-N 127.16 51.33 82.91 211.26 edge-S −31.4 −27.21 −28.6 −33.67 mid −5.26 15.09 8.5 −13.68 edge-N 23.43 28.71 31.02 26.93 Table 1. The seasonal variation in the Hadley cell in the first 20 years in each experiment is presented in this table. N and S indicate the north and south branches of the Hadley, respectively, and stren and mid indicate the strength and the ITCZ at 500 hPa, respectively. For the Hadley cell strength, a positive (negative) value means mass transport northward (southward), in 109 kg s−1. For the Hadley cell location, a positive (negative) value means that the location is in the Northern (Southern) Hemisphere.
where
$ {\phi _{\text{H}}} $ is the Hadley cell width, g is gravitational acceleration,$ {H_{\text{t}}} $ is the height of the tropopause,$ \Omega $ is the rotational angular speed of the Earth, a is the radius of the Earth,$ {\Delta _{\text{h}}} $ is the meridional surface temperature gradient, and$ {\theta _0} $ is the global mean temperature. In this case,$ {\phi _{\text{H}}} $ decreases due to a decreased$ {\Delta _{\text{h}}} $ between subtropics-tropics and an increased$ {\theta _0} $ .This anomalous Hadley cell pattern is also achieved in a slab ocean experiment, which displays the result of ocean dynamics as a surface forcing (Shin and Kang, 2021), although the polar warming here is much larger than that due to the extremely high amount of solar radiation absorption. The results suggest that even though the AMOC decreases due to continuing sea-ice melt, the surface temperature at low latitudes does not respond immediately, while the Hadley cell moves northward due to Arctic warming.
-
In the following two stages, we find that the vertical profile of the global oceanic temperature anomaly is consistent with that of the temperature anomaly in the Atlantic Ocean (Fig. 6a, shading). Response timescale analysis is applied to the vertical profile of oceanic temperature (Fig. 6a) to determine the sequence of ocean dynamics. Here, we define the response timescale as the beginning of a period in which the temperature anomaly always reaches 70% of that in the equilibrium state [following Yang and Zhu (2011), in the 2xCO2 experiment, we took the last two centuries as equilibrium for comparison, but it is during the middle period of stage I in Yang et al. (2018)]. The result of the response timescale analysis (Fig. 6a) suggests that the first region to reach equilibrium is the polar region [90°−60°S, 60°−90°N], followed by the thermocline in mid-low latitudes. The sea surface temperature in mid-high latitudes [30°−60°N] reached equilibrium slightly earlier than the sea surface temperature in the tropics. Based on the timescale analysis above, the implication in the last two stages is described as follows: warming at high latitudes decreases the meridional temperature gradient, which in turn decreases the westerly at the surface via the thermal wind. As a result, cyclonic wind stress decreases (Fig. 6c) and induces a weaker southward gyre via the Sverdrup relation. Due to mass conservation, the western boundary current decreases, especially in the region of 30°−40°N. Without warm water being transported poleward and sinking at high latitudes, there is significant cooling (approximately 4°C) in the region of approximately 60°N (Fig. 6b; Figs. 8a, c, e, g). The cooling (warming) at mid (low) latitudes in the NH enhances the meridional temperature gradient (−) and decreases the westerlies at the surface via the thermal wind. As a result, more heat is trapped at low latitudes, resulting in warming in the tropics, as well as weaker surface westerly (Fig. 6c). Then, the STC as well as the Walker circulation are weakened, and more heat is stored in lower layers and in the eastern part of the oceans (Fig. 6a shading). On the other hand, with surface warming in the Antarctic region, sea ice melts and freshens the surface ocean, which inhibits bottom water formation. As a result, there is significant warming at high latitudes in the SH (Fig. 6b; Figs. 8a, c, e, g).
Figure 6. (a) Vertical profile of the temperature anomaly (shading) in the Atlantic Ocean during the equilibrium state (°C) and the temperature response timescale (contour, year). (b) Sea surface temperature anomaly (shading, °C). (c) Surface wind stress anomaly (vectors, dyn cm−2) and barotropic streamfunction anomaly (shading, Sv, 1 Sv=106 m3 s−1). The anomalies presented in (b) and (c) are during the equilibrium state with a confidence level of 95% (Student’s t-test).
Figure 8. The seasonal surface air temperature anomalies in the 0.1A experiment (left column) and 2xCO2 experiment (right column) are shown, which are taken from the numerical results over the last two centuries (℃).
During the equilibrium state, the OHT decreases mainly due to a weaker mesoscale process, which is larger than the mean circulation contribution. Thus, we conclude that OHT anomalies occur mainly due to mesoscale processes. Strong cooling at 30°−40°N (Fig. 6b; Fig. 8a, c, e, g shading) largely enhances the subsidence branch there (as well as the Ferrel cell; Figs. 9a, c, e, g); thus, the edge of Hadley cell moves equatorward (Table 2). On the other hand, the enhanced warming in the tropics increases the convection in the Hadley cell. As a result, the strength of the Hadley cell grows, although the expansion of the Hadley cell is eliminated (nearly no cold color in NH; Fig. 9a, c, e, g shading). The strengthened Hadley cell increases the AHT everywhere, except in the region of 40°N-90°N, where the eddy activity decreases due to a smaller meridional temperature gradient.
Experiment Component MAM JJA SON DJF CTRL stren-S −87.55 −222.38 −112.34 −36.01 stren-N 127.65 52.2 83.33 210.05 edge-S −30.94 −27.20 −28.52 −33.41 mid −5.25 14.71 8.33 −13.42 edge-N 23.29 27.99 30.79 26.91 0.1A stren-S −78.41 −202.84 −105.18 −35.6 stren-N 133.04 48.54 84.5 224.63 edge-S −31.4 −27.1 −28.17 −33.49 mid −6.52 13.23 6.79 −13.52 edge-N 23.44 26.64 30.09 26.72 2xCO2 stren-S −76.18 −201.6 −102.8 −35.55 stren-N 131.52 47.99 82.35 217.18 edge-S −32.53 −27.73 −29.08 −34.4 mid −6.34 14.17 7.22 −14.03 edge-N 23.91 28.12 31.04 27.24 Table 2. Same as in Table 1 but for the Hadley cell in the equilibrium state.
Without strong Arctic warming (Figs. 7a, b; Figs. 8b, d, f, h shading) induced by artificially decreased surface albedo, there is little meridional temperature gradient between 30°N and 50°N in the 2xCO2 experiment (Fig. 7b Figs. 8b, d, f, h shading). Thus, the wind curl (Fig. 7c vector) and horizontal gyre remain strong (Fig. 7c shading). On the other hand, a decreased AMOC prevents warmer water vapor transport poleward at high latitudes in the NH. Thus, cooling (warming) occurs at approximately 60°N (at 30°−40°N). On the other hand, an increased CO2 forcing also induces warming globally due to the greenhouse effect (Figs. 8 b, d, f, h). In the 2xCO2 experiment, the sea surface warming in the SH (Fig. 7b shading) exhibits a similar pattern as that in the 0.1A experiment (Fig. 6b, shading), which suggests that the sea surface warming in the SH is mainly a result of the AMOC decreasing instead of a local forcing. Without a slow adjustment (temperature-wind-gyre-temperature) between 30°N and 40°N, the temperature reaches equilibrium much faster (Fig. 7a, contour), as does the sea surface temperature in the tropics. On the other hand, enhanced warming in the tropics (due to feedback of water vapor and clouds; Figs. 8b, d, f, h) suggests stronger convection and Hadley cell, while warming at midlatitudes and weaker warming at high latitudes in the NH suggest that the subsidence branch may move poleward (blue color is in the NH; Figs. 9a, h, shading; Table 2). As a result, the Hadley cell strengthened and expanded poleward in the 2xCO2 experiment (Figs. 9b, d, f, h; Table 2).
Experiment | Component | MAM | JJA | SON | DJF |
CTRL | stren-S | −86.6 | −222.47 | −114.75 | −34.97 |
stren-N | 122.31 | 53.37 | 83.91 | 211.20 | |
edge-S | −30.98 | −27.20 | −28.5 | −33.17 | |
mid | −5.39 | 14.54 | 8.1 | −13.03 | |
edge-N | 23.25 | 27.42 | 30.91 | 26.95 | |
0.1A | stren-S | −89.3 | −223.52 | −115.33 | −34.82 |
stren-N | 125.2 | 52.63 | 84.72 | 208.64 | |
edge-S | −30.4 | −27.04 | −28.12 | −33.08 | |
mid | −4.8 | 14.93 | 8.3 | −13.02 | |
edge-N | 23.21 | 27.73 | 30.57 | 26.74 | |
2xCO2 | stren-S | −82.8 | −220.15 | −110.6 | −36.93 |
stren-N | 127.16 | 51.33 | 82.91 | 211.26 | |
edge-S | −31.4 | −27.21 | −28.6 | −33.67 | |
mid | −5.26 | 15.09 | 8.5 | −13.68 | |
edge-N | 23.43 | 28.71 | 31.02 | 26.93 |