Historical seasonal cycle of Q pen and its impacts on SST
Historically, the seasonal cycle of the Qpen has exhibited complex regional dependence, with considerable changes in the mid-latitude oceans, especially in the mode water formation region, with marked changes in the seasonal Hm (Fig. 1a). Interestingly, the minimum Hm corresponds well to the seasonal variation in the Qpen (Fig. 1a, contour lines), suggesting that the summertime minimum Hm plays a dominant role in determining the seasonal cycle of the Qpen. Furthermore, in the mid-latitude ocean, the spatial pattern of the seasonal variation in Qpen resembles the structure of the subtropical gyre circulation (Fig. 1a), with a maximum value of 20 W m− 2. In contrast, the seasonal variation in Qpen is relatively small due to the weak seasonal change in Hm in tropical oceans and the extremely deep ML in high-latitude oceans (Figure S1). This case can be found in the Southern and North Atlantic Oceans, where the extremely deep annual mean ML prevents SW from any penetration out of the base of the ML. Biologically, the seasonal variation in Hp was more considerable in the oligotrophic region and in the North Atlantic Ocean (Figure S1), which can efficiently modulate the seasonal cycle of the Qpen.
Due to the significant change in Qpen in the mid-latitude ocean (Fig. 1a), we define two latitudinal bands (40°S-20°S and 20°N-45°N) in the northern (NH) and southern (SH) hemispheres as the regions of focus. The time series shows that the seasonal cycle of Qpen peaks in January (July) in the SH (NH) at more than 40 W/m2, which corresponds to approximately half of the hflx (Fig. 1b and 1c). The minimum value of Qpen appears in July (January) in the SH (NH), which is accompanied by the deepest ML. Therefore, the effect of Qpen is mainly in summer, i.e., during December-January-February (DJF) in the SH and June-July-August (JJA) in the NH. This significant change in the seasonal Qpen can affect the seasonal SST evolution, which can be further quantified by a simplified mixed-layer heat budget. In particular, hflx-induced heating (Qnet) controls the mixed-layer temperature on a seasonal timescale, while ocean dynamical processes tend to partially offset hflx-induced heating (Figure S2). Furthermore, Qpen reduces SST warming by directly decreasing the absorbed shortwave radiation in the mixed layer (Figure S2), and this cooling effect is comparable to hflx-induced warming during summer
Quantifying the factors contributing to Q pen change
Overall, due to the combined influences of decreased Hm and increased Hp, the change in Qpen under the SSP5-8.5 scenario exhibits a consistent seasonal increase, with the most pronounced change occurring in summer, reaching 4–6 W m− 2 in the mid-latitude ocean (Fig. 1d and f). Quantitatively, the summertime Qpen in the multimodel ensemble mean increased by 3.9 (3.7) ± 1.9 (1.6) W m− 2 in the northern (southern) mid-latitude oceans during the twenty-first century. The increase in Qpen is most pronounced in the subtropical Pacific and the North Atlantic Ocean, particularly in the northeastern subtropical Pacific, which differs from the region where climatological Qpen is large in the present climate (Figs. 2a-b), reflecting the significant change in this region under a warmer climate. Moreover, in the mid-latitude SH oceans, the increase in Qpen is mainly concentrated in the sub-Antarctic front, with the largest increase reaching 10 W m− 2 (Fig. 2a).
Under the SSP5-8.5 scenario, the three factors that determine Qpen exhibit great diversity in the mid-latitude ocean. For the SW, the mid-latitude oceans show negligible changes, with the most significant change occurring in the boreal spring, with an increase of 1.6%, corresponding to 4 W m− 2 (Figure S3). Therefore, this contribution to the change in Qpen can be ignored in the following analysis. Hm tends to decrease in the mid-latitude ocean, with the most pronounced decrease occurring in February in the NH and September in the SH, albeit with greater intermodel uncertainty. Note that the change in Hm during summer is relatively small and is accompanied by a shallow climatological mean ML. Therefore, the shallowing of the ML in summer can lead to a significant increase in Qpen by permitting much shortwave radiation to enter the subsurface layer. Hp exhibits relatively uniform changes during all seasons, increasing by ~ 1 m. Nevertheless, the change in Qpen was sensitive to the change in Hp due to the nonlinear relationship between Qpen and Hp (Eq. 1 in the methods section), especially in summer.
The contribution to Qpen can be further quantified in Fig. 2, with a focus on summer in the Northern Hemisphere (JJA) and Southern Hemisphere (DJF). In the mid-latitude ocean, the SW has a small contribution to the change in Qpen (Figs. 2c and 2d). Most of the change in Qpen is attributed to ML shoaling, accounting for approximately 75% of the variability, while Hp contributes 25% of the variability during the summer (Figs. 2e, 2f, and S4). In the mid-latitude SH oceans, the change in Hp also contributes less to Qpen than to Hm by 5 W m− 2 (Figs. 2g, 2h, and S4). Although the Hm and Hp exhibites considerable intermodel uncertainty (Figure S3), Qpen exhibites a relatively small spread due to the mutual cancellation of the three factors.
By analyzing the contributions of Hm and Hp to Qpen under the SSP5-8.5 scenario, we found that the change in Hp plays a significant role in influencing Qpen in the mid-latitude ocean during summer, albeit with relatively small changes (Fig. 2g and 2h). Therefore, a sensitivity analysis is performed by a Taylor expansion for Eq. 1 according to 38, which can be linearly decomposed into contributing terms as follows:
$$\:\varDelta\:{Q}_{pen}=\gamma\:{Q}_{sr}exp(-\frac{{H}_{m}}{{H}_{p}})\frac{1}{{H}_{p}}\left[\varDelta\:{H}_{m}-{H}_{m}\varDelta\:{H}_{p}\right]$$
;
where Qsr, Hm, and Hp represent the historical climatological mean SW, Hm, and Hp fields, respectively. ΔQpen, ΔHm, and ΔHp denote their changes under the SSP5-8.5 scenario. This equation illustrates that the change in ΔQpen is determined by the relative magnitudes of ΔHm and HmΔHp. Thus, the relatively small change in ΔHp can cause a substantial change in Qpen because its effect can be amplified by a multiple of Hm. Physically, the climatological Hm is extremely shallow in summer due to surface warming, which indirectly amplifies the effect of penetrative shortwave radiation induced by the increased Hp. The diagram of the effects of Hm-Hp on the Qpen shown in Fig. 3 further confirms this theoretical analysis.
The Hm-Hp diagram plot shows that the change in Qpen is sensitive to the changes in Hp and Hm, which is due to the climatological shoaling of the ML (Fig. 3). The CMIP6 ESMs can well capture the observed Qpen during summertime and wintertime (Fig. 3a and 3b). More importantly, as shown in Eq. 1 derived from the Taylor expansion, the change in Qpen is more sensitive to that in Hp under the SSP5-8.5 scenario than to that in Hm, which is well illustrated in Fig. 3c-d. In Fig. 3c, when there is a change in Qpen, less change in Hp can achieve a change in Qpen than in Hm during the austral summer, and similar conditions can be found in the NH during the boreal summer (Fig. 3d). In other words, under the global warming scenario, a subtle change in Hp can efficiently modulate the change in Qpen during summer, contributing to the greater sensitivity of Hp to Qpen under a relatively shallow ML. The high sensitivity of Qpen to Hp indicates that a change in phytoplankton (e.g., a decrease in chlorophyll or a phytoplankton bloom) can significantly influence the Qpen and the thermodynamics of the subsurface layer or even affect the evolution of SST by changing the timing of blooms14. In contrast, during winter, the deep ML can lead to a minimum value of Qpen under the SSP5-8.5 scenario, indicating that the climate effect of Qpen is negligible.
Impacts of increased Q pen on the seasonal cycle of SST
The contributions of the increased Qpen to the seasonal upper ocean temperature are quantified by a mixed-layer heat budget analysis. Due to the different vertical mixing processes among CMIP6 models and the lack of data in some selected CMIP6 simulations, oceanic processes (advection and diffusion) are considered residual terms in the Methods section. It should be acknowledged that the impact of increased Qpen on specific ocean dynamic processes (e.g., vertical advection and circulation change) is relatively small compared to the changes in hflx and Hm at the ocean-basin scale. Therefore, we only emphasize the potential effect of Qpen on the mixed layer temperature, and the contribution from an ocean dynamic process is considered to compensate for the heat flux change, as adopted in a previous study 18,19.
The heat budget analysis reveals that the SST tends to increase under the SSP5-8.5 scenario, with an increase in SST emerging during the summertime, i.e., JJA in the NH and DJF in the SH (Fig. 4a and b). Moreover, relative cooling occurs during the winter in both the SH and NH. The mixed layer warming due to the increased hflx to the mixed layer (Qnet) plays a dominant role in determining the seasonal tendency of the SST. In contrast, ocean dynamic processes act to partially compensate for the changes induced by changes in hflx. The effect of Qpen shows a relatively constant cooling effect due to its enhanced effect in a warmer climate, as shown in Figs. 3a-b. Similar to the marked change in Qpen compared to Qnet in the annual mean 25, the enhanced Qpen causes a cooling effect that is comparable to the change in Qnet on the seasonal timescale, which is predominantly in summer and reaches 0.3°C month− 1. The relative relationship between Qnet and Qpen is shown in Figs. 4c and 4d, suggesting that the comparable contribution of Qpen occurs mainly in the low- to mid-latitude ocean, especially in the northern Atlantic Ocean, the northeastern subtropical Pacific, and the entire mid-latitude Southern Ocean.
To highlight the sensitivity of Qpen to radiation forcing, an analysis was performed using four emission scenarios (SSP1-2.6, SSP2-4.5, SSP3-7.0 and SSP5-8.5) from CESM2, which is considered the best model for reproducing the observed spatial pattern of Qpen (Figures S5-S6). The results suggest that with increasing greenhouse gas emissions, Qpen gradually increases in the Southern Hemisphere during the DJF period (Figure S7). Note that the magnitude of Qpen even exceeds the change in hflx under SSP2-4.5, which is a relatively reasonable emissions pathway for the future. In contrast, the change in Qpen during summer can reach half of the change in hflx in the NH during JJA. Nevertheless, the changes in Qpen and hflx are not linear, with a decrease occurring in SSP3-7.0 due to the decrease from the SW in the northeastern Pacific Ocean.
Based on previous analysis, we identified three factors that affect Qnet under the SSP5-8.5 scenario in Eq. 3. These factors are hflx, Hm, and Qpen. Each factor is separately analyzed to better understand its effects by using linear decomposition. The computational error due to linear decomposition is negligible (Figure S8). In the 30°N-60°N region during boreal summer (JJA), the change in Hm has the largest effect on surface warming, while hflx has a secondary effect (Fig. 4f). In the 0°-30°N region, the change in hflx dominates the SST warming, and the contribution of Hm is minimal (Fig. 4f), especially in the northeastern tropical Pacific (Figure S9). In contrast, an increase in Qpen induces a cooling effect that gradually increases toward the north. In the SH, the contributions of these three factors exhibit a complex distribution (Fig. 4e). The change in SST due to hflx is relatively small north of 60°S. Moreover, the contributions of Hm and Qpen tend to cancel each other out, especially in the subantarctic front (Figure S10), resulting in a slightly weak SST change in the SH during the austral summer. These results indicate that the comparable change in Qpen with hflx can induce a remarkable decrease in the SST in the mid-latitude ocean during summer, albeit with the dominant role of Hm.
Tight relationship between Q pen and SST changes in the mid-latitude ocean
Due to the remarkable cooling effect of Qpen on the SST on the annual mean timescale under the global warming scenario 25, the potential feedback between increased Qpen and SST may also exist on a seasonal timescale. Figure 5 shows that the significant correlations are mainly located in the mid-latitude ocean, with R = 0.59 (P = 0.07) and R = 0.70 (P = 0.02) in the NH and SH summers (Figs. 5a-b), respectively. In contrast, the correlation between Qpen and SST in the global and tropical oceans is insignificant in the annual mean field (Fig. 5c-d), where the seasonal cycle is relatively weak. As a significant positive correlation between SST and Qpen changes in the mid-latitude ocean, an increase in SST corresponds to an increase in Qpen. As discussed in the previous sections, increased Qpen can result in surface cooling through more shortwave penetration into the subsurface layer (less shortwave absorption within the surface layer); the strong correlation between Qpen and SST suggests that Qpen may exert negative feedback on surface warming in mid-latitude oceans under a global warming scenario. Note that the observed correlation does not necessarily imply causation, which requires further numerical experiments to demonstrate it.