Enhanced Asian warming increases Arctic amplification

The Arctic has been experiencing prominent warming amplification. However, despite anthropogenic emissions and oceanic variability, whether Arctic amplification has a connection with land in the lower latitudes remains unknown. Here, we newly identify enhanced Asian warming as a factor underlying Arctic amplification. The simulations demonstrate that enhanced Asian warming contributes 22% of the wintertime amplified warming over the Barents–Kara Seas (BKS). We demonstrate that Asian warming remotely affects the Arctic by affecting poleward atmospheric heat and moisture transport. The external anomalous heat and moisture further trigger local feedbacks concerning sea ice-albedo feedback and changes in longwave radiation and evaporation, thus facilitating BKS warming amplification. The capacitor effect of the Arctic Ocean further modulates the seasonality of BKS warming via turbulent heat flux exchange between the atmosphere and ocean. Moreover, anomalous Rossby wave trains are responsible for the anomalous atmospheric circulations favoring the atmospheric heat and moisture transport into BKS. Our findings illuminate a new factor from remote lower latitudes affecting Arctic climate change.


Introduction
The Arctic, especially the Barents-Kara Seas (BKS), has been experiencing the strongest warming on Earth, which is often called Arctic amplification (Screen and Simmonds 2010, Stouffer and Manabe 2017, IPCC 2021. This warming amplification originates from anthropogenic emissions (Gillett et al 2008, Pithan and Mauritsen 2014, Stouffer and Manabe 2017 and is regulated by local feedbacks (Screen and Simmonds 2010, Pithan and Mauritsen 2014, Dai et al 2019, Gao et al 2019, Xie et al 2019 and external heat and moisture transport (Cai 2005, Lu and Cai 2009, Delworth et al 2016, Graversen and Burtu 2016, You et al 2021.
The local feedbacks influence the Arctic warming by modifying the atmospheric and oceanic energy balance. For example, the sea ice-albedo feedback in summer favors the Arctic warming in winter via absorbing energy in summer and releasing energy in winter (Boeke and Taylor 2018, Xie et al 2022a). The enhanced downward longwave (LW) radiation associated with atmospheric moisture is also suggested to contribute to Arctic warming (Lu and Cai 2009, Gao et al 2019, Lesins et al 2012. The atmospheric moisture is further influenced by the external moisture transport from lower latitudes (Cai 2005, Graversen and Burtu 2016, Luo et al 2017, Cai et al 2022. In addition, external heat transport to the Arctic can directly increase Arctic warming (Delworth et al 2016, You et al 2021. On interdecadal time scales, Arctic climate change is also modulated by the internal climate variability associated with oceans (Delworth et al 2016, Ding et al 2018, Svendsen et al 2018. The influence of Arctic climate change on the midlatitudes has been widely studied (Kretschmer et al 2016, Blackport et al 2019, Cohen et al 2019, Xie et al 2020. Meanwhile, studies have also suggested an influence of mid-latitudes on the Arctic (Luo et al 2017, Cai et al 2022, Xie et al 2022b. However, the influence from land, especially the remote regions in lower latitudes, has not been investigated. Enhanced Asian warming has been demonstrated to be prominent over the land (Huang et al 2012, Xie et al 2019, Yao et al 2019. This enhanced warming mainly includes two parts, enhanced warming in the Asian dryland (Huang et al 2016a(Huang et al , 2017 and enhanced warming over the Tibetan Plateau Xiao 2015, You et al 2020). The warming magnitude over the global dryland during the last century is 20%-40% larger than that over the global humid regions (Huang et al 2017). The warming over the Tibetan Plateau was 1.8 times the global mean warming during the period from 1979 to 2020 (You et al 2021). However, the Asian dryland and elevation higher than 500 m domains overlap vastly (figure S1). Hence, in this study, the effect of enhanced Asian warming was simulated by halving the sensible heat (SH) (vertical diffusive heat) over the Asian land higher than 500 m (AL500-sh0.5).
The influence of the warming over the Asian dryland and Tibetan Plateau on the remote Arctic climate has not been investigated, although the climatic effect of Tibetan Plateau heating on mid-latitude regions across the Northern Hemisphere has been widely recognized (Yanai et al 1992, Wu et al 2012, Lu et al 2018, Yang et al 2020. Therefore, it is important to explore whether the Tibetan Plateau heating can exert an additional influence on the Arctic, aside from the influence on mid-latitudes. Moreover, the climatic effect of dryland warming needs to be investigated urgently given the threat of persistently expanding and warming dryland in the future (Feng and Fu 2013, Huang et al 2016b).

Data
The observational temperature was based on the ERA5 reanalysis data provided by ECMWF (Hersbach et al 2020) in 1 • × 1 • grids from 1950 to the present. The observational precipitation and potential evapotranspiration datasets were based on the Climatic Research Unit (CRU) TS version 4.05 data (Harris et al 2020) in 0.5 • × 0.5 • grids from 1901 to 2020. The dryland classification was based on the aridity index calculated as annual precipitation divided by potential evapotranspiration (Feng andFu 2013, Huang et al 2016a). In addition, the aridity index data provided by Feng and Fu (2013) were used. This aridity index was calculated from the NOAA's Precipitation Reconstruction over Land (PREC/L) (Chen et al 2002) and Global Land Data Assimilation System specific humidity, solar radiations, and wind speed (Rodell et al 2004) in 0.5 • × 0.5 • grids from 1948 to 2008. The potential evapotranspiration for both datasets was calculated using the Penman-Monteith algorithm (Allen et al 1998). Although the subtypes of the dryland were different in some regions, such as hyper-arid and arid regions, the overall Asian dryland domains agreed well between the two datasets (figure S1). Therefore, the dryland classification was reliable.

Numerical simulations and models
We performed two series of numerical experiments using the Community Earth System Model (CESM) (table S1; Danabasoglu et al 2020): the halving-SH (sh0.5) and the non-SH (nosh) experiments, which were developed from the benchmark experiments of the Global Monsoons Model Intercomparison Project (GMMIP) (Zhou et al 2016) of the Coupled MIP Phase 6 (CMIP6) (Eyring et al 2016). The sh0.5 and nosh experiments were fully coupled and atmosphere-only simulations (amips), respectively. In this study, the sh0.5 experiment was mainly examined, and the nosh experiment was a supportive experiment that was used to verify the reliability of the CESM model and the reasonableness of the experimental designs. The control runs of the sh0.5 and nosh experiments were historical (HIST) simulations of CMIP6 labeled as HIST and amip-HIST, respectively, where amip indicates atmosphere-only simulation. The sensitivity runs of the sh0.5 and nosh experiments were designed as halving or removing the SH over Asian land above 500 m (AL500), labeled as AL500-sh0.5 and amip-AL500-nosh, respectively. In this study, SH indicates the vertical diffusive heat from the surface to the top of the atmosphere (Wu et al 2007, Xie et al 2023 rather than just SH flux at the surface. Because the nosh experiment was the benchmark experiment of GMMIP (table S1), nosh experiments were also available for two models participating in GMMIP: FGOALS-f3-L (He et al 2020) and FIO-ESM-2-0 (Bao et al 2020). Therefore, a comparison of the results from CESM with those from CMMIP can verify the performance of CESM in SH-modification experiments. The details of the models are introduced in table S1. Although the amip-HIST of GMMIP has three ensemble integrations, the amip-AL500-nosh of GMMIP has only one integration, so only one integration was examined. The integration periods for the nosh experiments of CESM and GMMIP were from 1979/01 to 2014/12 and 1870/01 to 2014/12, respectively. Therefore, the period from 1979/01 to 2014/12 was analyzed for nosh experiments. In addition, the period from 1850/01 to 2014/12 was analyzed for sh0.5 experiments.
The performance of CESM in simulating the basic climate elements was addressed as follows: the control run could reproduce the climate in the observations (figure S2), ensuring the basis for further analyses of the sensitivity run. Specifically, the control run well simulated the climatological near-surface air temperature distributions and seasonality from the subtropics to the Arctic, including our target AL500 and BKS domains (figures S2(a)-(d)); some observed smallscale features were lacking in simulations because the horizontal resolution of CESM was lower than The stripes indicate that the linear trend is significant at the 99% confidence level based on the two-tailed Student's t-test and FDR-adjusted P-values. (b) Difference in annual mean near-surface air temperature for 1850-2014 determined by subtracting the AL500-sh0.5 (halving the sensible heat over the Asian land above 500 m) simulation from the HIST (historical) simulation. The stripe indicates that the difference between the two simulations is significant at the 99% confidence level based on the two-tailed Student's t-test. The thick gray contours indicate the regions of the Barents-Kara Seas (BKS) and the Asian land above 500 m (AL500), respectively. the resolution of ERA5. In addition, the atmospheric circulations were well simulated in the control run, as indicated by the geopotential height at 500 hPa (figures S2(e)-(h)). For example, the extratropical troughs and ridges were identical between simulations and observations. The geopotential height was higher in the tropics in simulations than in observations, which probably occurred because the vertical resolution of CESM was lower than that of ERA5, specifically, 32 and 137 vertical levels for CESM and ERA5, respectively (Danabasoglu et al 2020, Hersbach et al 2020. The considerations underlying the SHmodification experiments were addressed as follows: (i) Two typical methods can be used to modify temperature over land, namely modifications in surface albedo (Wang et al 2008, Lu et al 2018 and SH (Wu et al 2007, Zhou et al 2016. (ii) SH is the major component of total heating over the Tibetan Plateau (Yanai et al 1992, Wu et al 2007 and Asian dryland (Huang et al 2017, Xie et al 2019. Consequently, modifying SH can directly modify the major parts of total heating. Because modifying albedo indirectly changes total heating via multiple feedbacks, total heating changes induced by modifying albedo are complicated. Therefore, we followed the GMMIP method to modify SH. (iii) The benchmark nosh experiments generated strong warming, approximately 4.5 • C averaged over AL500 ( figure  S3). Therefore, we redesigned the sensitivity experiment by halving, instead of removing, the SH. The sh0.5 experiment had a 1.5 • C warming (figure 1(b)), which was less than the observed 1.9 • C accumulated warming magnitude of AL500 from 1950 to 2021 (figure 1(a)). In addition, for a fair comparison with observation, the results based on sensitivity experiments were further scaled to observation, as introduced in section 2.3.
The technical details of the SH-modification experiments were addressed as follows: (i) The SH was not prescribed to a predetermined value but was modified in each time step of model integration; thus, all feedbacks were completely active. (ii) For attribution of the influence of sea surface temperature variation, pacemaker experiments performed by restoring the sea surface temperature to observed values have been widely used (Kosaka andXie 2013, Zhou et al 2016). However, the temperature of the atmosphere is a predict and of the atmospheric circulation model, instead of the variables, such as sea surface temperature, that are communicated among the components of the model through the coupler. Therefore, the restoring method used for sea surface temperature was not suitable for the temperature of the atmosphere. (iii) The modification was not limited to the single-level SH flux exchange between the land surface and atmosphere but was on the entire column of the atmosphere; that is, the halving or removing of the SH (vertical diffusive heat) was done in all model levels of the atmosphere. The practice of modifying the entire atmosphere was implemented to assure the vertical continuity of total heating of the atmosphere. (iv) In coding, the SH (variable 'cam_in%shf ' in CESM codes) in all model levels of the atmosphere over the AL500 regions was multiplied by 0.5 for the sh0.5 experiment or zero for the nosh experiment, in each time step of the model integration. Refer to the Data availability statement for our codes regarding CESM experiments. (v) Because the feedbacks are completely active in the simulations, the energy conservation in the model after halving SH was established by the instant adjustment in each time step of the model integration, such as the decrease in LW cooling which was directly indicated by the temperature changes ( figure S4(a)). In addition, because the diffusive SH generally decays with height, halving SH mainly changes the SH in the lower troposphere over the AL500 (Wu et al 2007).
The effects of halving SH on basic variables of local climate were addressed as follows: The warming induced by halving SH occurred from the surface through the entire troposphere over AL500 (figure S4(a)), which was consistent with the observations ( figure S4(b)). In addition, the greenhouse gas-induced HIST warming from 1861 to 2004 over the dryland shows a prominent increase in SH at the surface (Xie et al 2019). Despite decadal variability due to near-surface wind variation, the surface SH over the Tibetan Plateau was also suggested to increase with long-term warming . Therefore, halving SH simulation can reasonably represent the warming over AL500 shown in observations. Furthermore, the high consistency in the differences in temperature, precipitation, and geopotential height at 500 hPa of the nosh experiment between CESM and GMMIP suggests that the SHmodification simulations of CESM model were reliable (figure S3). The warming over AL500 in the nosh experiment (figures S3(a) and (b)) showed similar patterns but stronger magnitudes than the warming in the sh0.5 experiment ( figure 1(b)). The differences in the atmosphere-only nosh experiment were not statistically significant in the Arctic, which suggests that air-sea interactions were crucial for the Asia and Arctic connections in the sh0.5 experiment. Although only SH was modified, the latent heat (LH) due to precipitation (figures S3(c) and (d)) also changed in the SH-modification experiments because of feedback . Therefore, the climate impact in the SH-modification experiments was induced not only by SH but also by the pertinent feedbacks.
In the analysis, the differences determined using HIST minus AL500-sh0.5 indicate the influence of Asian warming. Because the AL500-sh0.5 experiment was not a pacemaker experiment, the 1850-2014 mean differences between control and sensitivity runs were analyzed as the general sensitivity experiments, instead of analyzing the trends as with pacemaker experiments (Kosaka and Xie 2013). The 165 year average was sufficient to eliminate the influence of internal climate variability. For example, the Pacific Decadal Oscillation and Atlantic Multidecadal Oscillation have long periods of only 50-70 and 50-80 years (Mantua andHare 2002, Lin et al 2019), respectively. The benchmark time-slice experiments of the Polar Amplification Model Intercomparison Project (PAMIP), which aims to investigate Arctic climate change, adopted 100 ensemble integrations, where the time-slice experiments perform a one-year integration each time and repeat many times (Smith et al 2019). Consequently, the 100 ensemble integrations of PAMIP were equal to the 100 year integrations of our experiments. Therefore, although only one integration was performed in our experiment, the 165 year integration assures the reliability of the simulations. Although the influence of internal climate variability on sensitivity experiments was mostly eliminated by the experimental designs in this study, internal climate variability was crucial for the observed trends over a decadal timescale (Kosaka andXie 2013, Trenberth 2015).

Diagnostic methods
To compare the results of the sensitivity experiment with observations, the temperature differences in the simulations should be scaled to the warming trend based on the sensitivity of response to forcing, namely, the BKS temperature change (T BKS ) in response to the AL500 temperature change (T AL500 ) in the sh0.5 experiment. Specifically, the temperature difference in the simulations (T S ) was scaled to the warming trend in the observation (T O ), according to the formulation T Scaled = indicates the response-to-forcing ratio in the simulations. The formulation physically means that the scaled BKS warming was the product of the simulated response-toforcing ratio and the observed changes in forcing. The forcing of AL500 warming had a cross-season influence on the BKS temperature response (section 3.2) because of the cross-season energy cycles in the BKS (Boeke andTaylor 2018, Xie et al 2022a). Therefore, T S AL500 and T O AL500 used annual mean values, while T S BKS used monthly mean values. The values of T S AL500 and T O AL500 were 1.51 • C and 0.26 • C decade −1 , respectively (figure 1). To quantify the amplified warming of AL500 and BKS relative to the global mean, the global mean warming values (T O Global ) were further removed when calculating the contribution rate of enhanced Asian warming to BKS warming amplification; namely, the formulation of the contribution rate was × 100. The local energy budget at the sea surface of the Arctic Ocean was formulated as where E ↓ net is the net energy uptake due to local absorption by seawater and external horizontal oceanic heat transport and SW ↓ net and LW ↓ net indicate the net shortwave (SW) and LW radiation at the sea surface, respectively. TH ↑ indicates the upward turbulent heat (TH) flux, which is the sum of SH and LH. In an alternative form, the budget of upward LW radiation is In addition, the net SW solar radiation change can be decomposed as the sum of contributions by albedo feedback and downward SW solar radiation (SW ↓ ) change as follows: where the operator δ indicates the HIST minus AL500-sh0.5 difference and α = SW ↑ SW ↓ is the surface albedo. The formulations SW ↑ = αSW ↓ and SW ↓ net = (1 − α)SW ↓ were introduced.
The horizontal heat flux was expressed as ⃗ H = ρ ⃗ Vθ, where ρ, ⃗ V, and θ indicate the density, horizontal velocity, and potential temperature of the atmosphere, respectively. The vertical integral of ⃗ H is calculated according to formulation{ ⃗ where g, p, and p s indicate the acceleration of gravity, pressure, and surface pressure, respectively. The horizontal moisture Because the fluxes are nonlinear, we calculated the heat and moisture fluxes using the three-hourly output for the period from 2000 to 2014; only 15 years of high-frequency outputs were available because high-frequency outputs cost large storage. Furthermore, the eddy stream function and Plumb wave activity flux were used to examine the stationary Rossby wave trains. The eddy stream function means that the zonal mean of the stream function was removed. The Plumb wave activity flux was calculated based on the formula (7.1) in Plumb (1985).

Statistical analysis
The linear trend for the period from 1950 to 2021 was calculated using linear regression based on the least square method, and its statistical significance was estimated using a two-tailed Student's t-test and the false discovery rate (FDR)-adjusted P-values (P adj ), namely, P adj < 0.01 for the 99% confidence level. The FDR method was used to enhance the reliability of statistical significance when the linear trends were analyzed on a spatial map (Benjamini and Hochberg 1995). FDR-adjusted P-values were formulated as P adj = P × [N/Rank(P)], where N indicates the total number of grids on the map and Rank(P) indicates the rank of P starting from 1 when all P-values were sorted in ascending order. In addition, the statistical significance (99% confidence level, P < 0.01) of the difference between the control run and sensitivity run was estimated using a two-tailed Student's t-test.

Contribution of enhanced Asian warming to Arctic warming
The linear trend in annual mean near-surface air temperature during 1950-2021 demonstrates that Arctic warming was the strongest over the BKS (figure 1(a)). This BKS warming amplification has been widely suggested by observations and simulations (Gillett et  In this study, the difference between the control run (HIST) and the sensitivity run (AL500-sh0.5; halving the SH over Asian land above 500 m), i.e. HIST minus AL500-sh0.5, indicates the influence of Asian warming (section 2.2). As in the observations, the Asian warming in the simulations was not only at the surface but also in the entire troposphere ( figure S4). The simulations demonstrate that Asian warming significantly (99% confidence level) contributed to both the overall Arctic warming and the BKS warming amplification ( figure 1(b)). Given that the influence on BKS amplification was the most prominent in the simulations, the influence of Asian warming on BKS warming was focused on.
Arctic amplification is characterized by strong seasonality (Lu and Cai 2009, Pithan and Mauritsen 2014, Stouffer and Manabe 2017, Dai et al 2019, Xie et al 2019. The seasonality of the BKS warming in the simulations after scaling (section 2.3; red line) was consistent with the observation (black line) (figure 2(a)). BKS warming was much stronger in winter than in summer, so BKS warming amplification was observed in the cold season from late autumn to early spring but not observed in the summer Cai 2009, Pithan andMauritsen 2014). The amplified BKS warming relative to the global mean was further quantified by removing the global mean warming trend from BKS warming (section 2.3).
The enhanced Asian warming-induced BKS warming in the simulations accounted for a large portion of the observed BKS warming amplification ( figure 2(b)). Specifically, the results demonstrate that enhanced Asian warming contributed 25% of the annual mean BKS warming amplification in the observation ( figure 2(b)). The contribution rate was higher in summer because the observed BKS warming was weak and almost not amplified relative to the global mean. Thus, the wintertime BKS warming amplification was focused on. In the cold season, when BKS warming amplification was most prominent, the contribution of enhanced Asian warming was 22%. Therefore, Asian warming exerted a crucial influence on BKS warming amplification.

Mechanisms
To determine the mechanisms, the physical processes and large-scale atmospheric circulations that contribute to BKS warming attributable to Asian warming were investigated. The budget of upward LW radiation change was examined because it is another manifestation of the sea surface temperature change according to the Stefan-Boltzmann law (Lu and Cai 2009, Pithan and Mauritsen 2014, Gao et al 2019, Xie et al 2019, Lesins et al 2021. Downward LW radiation made the largest contribution to the upward LW radiation increase for the entire year ( figure 3(a)). However, other processes also had substantial contributions, which varied with the season.
The net energy uptake (local absorption by seawater and external horizontal oceanic heat transport) had a positive contribution in the cold season, accounting for 51% (59%/115%) of the downward LW radiation (figure 3(a)). The TH flux (sensible plus latent) had a negative contribution, and solar radiation was negligible in the cold season due to the long polar nights. In contrast, net energy uptake negatively contributed to upward LW radiation increase in the warm season. Net SW solar radiation and TH flux also favored upward LW radiation increase in the warm season. Thus, for BKS sea surface warming, downward LW radiation and net energy uptake had positive contributions in the cold season, in a relative proportion of 7/3. With the relative proportions of 6/3/1, downward LW radiation, net SW solar radiation, and TH flux had positive contributions in the warm season.
Downward LW radiation increase was favored by higher humidity (figures 3(b) and S5) because of the greenhouse effect of water vapor (Gao et al 2019, Cai et al 2022. Additionally, dense cloud coverage in BKS (figure S6) was another effective factor favoring downward LW radiation in the winter with long polar nights (Xie et al 2019). Net energy uptake enhanced the seasonal difference of BKS sea surface warming by suppressing and strengthening warming in the warm and cold seasons, respectively (Boeke and Taylor 2018). This implies a cross-season energy cycle, which is analogous to the capacitor effect of the ocean (Xie et al 2009), charging energy in summer and discharging energy in winter (Xie et al 2022a). Net SW solar radiation increase was attributable to the sea icealbedo feedback (figures 3(b) and S7), which overwhelms the negative contribution of downward solar radiation change due to cloud (He et al 2019, Xie et al 2022b. Inputting of external heat or moisture to BKS was required to trigger the downward LW radiation changes and sea ice-albedo feedback. Enhanced heat was transported from lower latitudes into west BKS via the anticyclonic circulation occupying the region to the south of BKS in the cold season (figures 4(b) Figure 5. Large-scale atmospheric circulations and schematically summarized mechanisms associated with BKS warming. (a), (b) Difference in the eddy stream function (shading) and Plumb wave activity flux (vector; section 2.3) at 500 hPa, using HIST minus AL500-sh0.5 for 1850-2013, in the warm (MJJAS) and cold (NDJFM) seasons, respectively. The white stripe indicates that the difference between the two simulations is significant at the 99% confidence level based on the two-tailed Student's t-test. The thick arrows mark the wave trains directing to the anomalous circulations associated with heat and moisture transport into BKS. (c), (d) Schematic views of the mechanisms linking enhanced Asian warming with BKS warming in warm and cold seasons. and S8). This anticyclonic circulation also transported additional moisture from the south into west BKS (figure 4(d)). The enhanced moisture affected the sea surface temperature via increasing downward LW radiation. Both the enhanced heat and moisture transport occupied the west BKS; however, the enhanced moisture transport extended farther eastward (figures 4(b) and (d)). In particular, the moisture influenced more regions of the strongest warming and sea ice decline (indicated by white mesh) (figures 4(b) and (d)). In addition, the external inputting heat directly affected the air, but not the sea surface, in the cold season because the TH flux was from the sea to the atmosphere (Blackport et al 2019). Therefore, the enhanced moisture transport was more important than the transport of heat in favoring BKS warming and sea ice decline in winter.
In contrast to the cold season, cyclonic circulation occurred in the warm season (figures 4(a), (c) and S8). Nonetheless, this cyclonic circulation also favored the heat and moisture input into east BKS from the south (figures 4(a) and (c), further triggering sea ice-albedo feedback (regions with white mesh) in the warm season. The enhanced heat transport extended more westward than the enhanced moisture transport (figures 4(a) and (c)); thus, the enhanced heat transport was more important than the enhanced moisture transport in favoring BKS warming and sea ice-albedo feedback in summer. Notably, the enhanced moisture transport was dominant in the BKS moisture increase in the warm season (figures 3(b) and S9(a)) because the local evaporation change in the BKS was weak ( figure  S9(c)). The local evaporation increase (figure S9(d)), especially in the regions of the strongest warming and sea ice decline (figures 1(b) and 4(d)), was also important for the BKS moisture increase in the cold season (figures 3(b) and S9(b)). In addition, anomalous Rossby wave trains (figures 5(a) and (b)) were responsible for the anomalous atmospheric circulations associated with the heat and moisture flux. Specifically, the two wave trains from downstream and north of AL500 stimulated the anomalous cyclonic circulation in the warm season ( figure 5(a)). The wave train downstream AL500 also occurred in the cold season, but it was located more southwestward ( figure 5(b)). Furthermore, a wave train from the west AL500 facilitated anomalous anticyclonic circulation in the cold season ( figure 5(b)).
Although the local energy budget is applicable to the sea surface rather than near-surface air, the TH flux and LW radiation connect the sea surface temperature with the near-surface air temperature (figures 5(c) and (d) (Blackport et al 2019), the negative contribution of TH to sea surface warming in the cold season ( figure 3(a)) indicates its positive contribution to near-surface air warming. Therefore, oceanic heat was transported upward to the atmosphere via TH flux in the cold season ( figure 5(d)). In contrast, TH suppressed the near-surface air warming (figure 2(a)), but it facilitated oceanic warming in the warm season (figure 5(c)) (Xie et al 2022b). Consequently, although the external atmospheric heat input (figures 4(a), (b)) and moisture increase (figure 3(b)) were stronger in the warm season than in the cold season, the near-surface air warming was weak in the warm season (figure 2(a)). Moreover, the capacitor effect of the ocean (Xie et al 2022a), i.e. net energy uptake in the warm season and net energy release in the cold season (figures 3(a) and 5(c), (d)), relied on the TH flux.

Conclusions and discussion
This study revealed a new factor underlying the Arctic amplification, as well as a new climatic influence of the enhanced Asian warming. The enhanced Asian warming includes the warming over the Asian dryland (Huang et al 2016a, Xie et al 2019 and Tibetan Plateau (Duan and Xiao 2015, Yao et al 2019, You et al 2020. The Tibetan Plateau is recognized as the third pole (Qiu 2008, Barry andHall-McKim 2018) on Earth (in addition to the North Pole and South Pole). An influence of the Arctic on the Tibetan Plateau has recently been proposed (Li et al 2020, Duan et al 2022. Here, we further demonstrated an influence of the Tibetan Plateau on the Arctic. These results imply potential pole-to-pole interactions between the Arctic and the third pole. Therefore, our findings prompt future research on pole-to-pole interactions. In addition to the local environmental influence, we argue that the influences of dryland climate change on remote regions should be considered carefully, especially under accelerated dryland expansion and warming (Feng and Fu 2013, Huang et al 2016b, Xie et al 2019.
The mechanisms underlying the connections between Asian warming and BKS warming were schematically summarized in figures 5(c) and (d). Asian warming exerted its influence on BKS by inducing anomalous circulations that transported external heat and moisture into BKS. As triggers, the external heat and moisture stimulated local feedbacks; the heat directly influences temperature, while the moisture influences temperature by its greenhouse effect associated with LW radiation. The LW radiation increase made the largest contributions to the BKS warming in both summer and winter. The sea ice-albedo feedback also made a crucial contribution to BKS warming in summer. Because of the capacitor effect of the ocean, the ocean acquired energy in summer and released the energy in winter. Therefore, the cross-season energy cycles due to the oceanic capacitor effect suppressed the summer warming but promoted the winter warming in the BKS, in which the TH flux was the main pathway exchanging heat between ocean and atmosphere. In addition, as two components of enhanced Asian warming, the Tibetan Plateau and dryland (averaged over the AL500 regions higher than 2 km or not, respectively), had warming magnitudes of 1.7 • C and 1.3 • C in simulations ( figure 1(b)). Therefore, the Tibetan Plateau warming may have a larger influence on the BKS warming than dryland, as inferred from the perspective of linear superposition. However, the individual effects of the Tibetan Plateau and dryland and their possible interactions will be investigated in future studies. In addition to SH, other factors, such as changes in surface albedo over the Tibetan Plateau and downward LW radiation over the dryland, have also influenced the HIST warming over the AL500 (Ma et al 2017, Gao et al 2019, Xie et al 2019, You et al 2021. Therefore, more experiments were expected to verify our results based on the SH-modification method in future studies.