Potential remote forcing of North Atlantic SST tripole anomalies on the seesaw haze intensity between late winter months in the North China plain: A case study

This study identified a prominent temporal seesaw haze intensity case that occurred between the late winter months of 2010 in the North China Plain (NCP), featuring considerably suppressed haze intensity in January and enhanced haze intensity in the adjacent month of February in 2011. We suggest that dramatic alternations of atmospheric and oceanic anomalies played fundamental roles in forming this seesaw haze intensity case, rather than changes in manmade emission anomalies. The suppressed haze intensity in January 2011 was tied to an equivalent barotropic cyclonic anomaly that dominated the NCP and its surroundings, which generated in situ haze‐suppressed meteorology characterized by strengthened lower‐level northerly anomalies with cold and dry conditions, as well as elevated boundary layer height and destabilized atmospheric stratification. In stark contrast, the enhanced haze intensity in February 2011 was connected to an equivalent barotropic anticyclonic anomaly, linking a haze‐favourable meteorology opposite to that in January 2011. The pronounced North Atlantic sea surface temperature (SST) tripole anomalies, with positive anomalies in the tropical and mid‐latitudinal North Atlantic and negative anomalies in the subtropical North Atlantic, made a significant contribution to the above‐mentioned seesaw haze intensity case. Diagnostic analyses suggested that the January North Atlantic SST tripole anomalies were linked to a significant negative North Atlantic Oscillation (NAO)‐like pattern, which acted as the source of the Rossby wave train to generate concurrent haze‐suppressed meteorology over the NCP. In February, although the NAO‐like pattern was drastically dampened, the enhanced barotropic cyclonic anomaly centred southeast of the Yamal Peninsula played a critical role in relaying the impact of January tripole SST anomalies, thus inducing concurrent haze‐favourable meteorology. Consequently, January North Atlantic SST tripole anomalies could exert an effective modulation effect on the generation of seesaw haze intensity. The proposed mechanism was further verified using the Community Earth System Model Large Ensemble Numerical Simulation (CESM‐LENS) datasets.


| INTRODUCTION
Haze pollution with high PM 2.5 (particulate matter (PM) with an aerodynamic diameter of 2.5 μm or less) concentration and low visibility has deleterious effects on health issues , transportation systems (Wu et al., 2005) and crop production (Tie et al., 2016). Based on the observed and modelled surface PM 2.5 , it was found that China has experienced more severe haze intensity and more frequent haze occurrence since the beginning of the 21st century Ding et al., 2017;Ding & Liu, 2014;Mao et al., 2019;Wang et al., 2014;Yang et al., 2016Yang et al., , 2018Zhang et al., 2013). As such, haze prevention and mitigation have garnered widespread attention in China because of their potential hazard to socioeconomic development and human health (Mu & Zhang, 2014; The State Council of the People's Republic of China, 2013China, , 2018Xia et al., 2018;Zhang, 2017).
Previous studies have identified that the North China Plain (NCP) is the most polluted region, with the highest haze occurrence frequency climatologically in boreal winter Chen & Wang, 2015;Li et al., 2022;Mao et al., 2019;Wang et al., 2020a;Yin et al., 2021). Haze pollution over the NCP is a typical response to anthropogenic emissions and overlying meteorological conditions Liao et al., 2015;Wang et al., 2020a;Yang et al., 2018). When concentrating on meteorological impacts on temporal variations in winter haze frequency/intensity in the NCP, the previous literature documented that haze variation at multiple timescales, including intraseasonal changes (e.g., Zhang et al., 2019), year-to-year changes (e.g., Wang et al., 2020b) and interdecadal variations (e.g., Wang et al., 2020c), can be modulated by regional-scale circulation anomalies and remote climate systems. For example, the modulator of El Niño-Southern Oscillation (ENSO) in boreal winter may drive the interannual variability of the concurrent Northeast Asian anticyclonic anomaly (NEAACA)/ Northeast Asian cyclonic anomaly (NEACA) through inducing teleconnections, which is indicative of the weakened/intensified East Asian winter monsoon intensity (Chen et al., 2020a;Wang et al., 2020aWang et al., , 2020b. As a response, haze-favourable/haze-unfavourable meteorology around the NCP can be induced, causing more/less frequent haze events in situ. It is worth noting that a recent study reported a distinct temporal seesaw haze pollution case in the NCP between December 2015 and adjacent January 2016, occurring under a sudden regime shift tied to the weakening of a super El Niño . This case was characterized by abnormally high PM 2.5 concentration anomalies in December 2015 and abrupt low PM 2.5 concentration anomalies in January 2016. Furthermore, inspired by the work of Yang et al. (2021), who explored the monthly evolution of the frequency of severe haze days in Beijing, we identified another type of temporal seesaw haze pollution case that occurred between the late winter months of 2010. This case featured low PM 2.5 concentrations with notable negative anomalies in January 2011 (repressed haze intensity) and high PM 2.5 concentrations with notable positive anomalies in the adjacent February 2011 (strengthened haze intensity) (see Figures 1 and S1, S2), experiencing a profound intraseasonal enhancement of haze intensity. This pattern can be viewed as a reversed temporal seesaw haze intensity pattern compared with that reported by Zhang et al. (2019). It is important to point out that, from the perspective of monthly severe haze frequency in Beijing, our identified seesaw haze intensity case may be the most striking between the adjacent late winter months for the recent period 2009-2019, with no severe haze days in January and seven severe haze days in February 2011 their Figure 1a).
Furthermore, in comparison with the decadal and interannual winter haze variations in the NCP that have been broadly investigated, the modulators responsible for such profound sharp intraseasonal haze enhancement cases are much less understood. As revealed by Zhang et al. (2019), we can infer that climate variability may play a predominant modulatory role in the generation of the identified seesaw haze intensity case, which is the main focus of the present study. Recent research has suggested that as the major predictability source of global climate variability on the interannual timescale (e.g., Trenberth, 1997;Wang et al., 2000); ENSO exerts highly weakened impacts on late winter haze pollution variations in the NCP (Zhao et al., 2022). Note that basin-wide North Atlantic sea surface temperature (SST) tripole anomalies have been proven to have intimate connections with the interannual variations in autumn and winter haze frequency in the NCP through exciting zonal eastward propagation of Rossby wave trains Xiao et al., 2015). Naturally, the question arises here as to whether there exists a close relationship between our identified seesaw haze intensity pattern and the North Atlantic tripole (NAT) SST anomaly pattern. If this is the case, what is the underlying physical mechanism behind this linkage? Such knowledge could be of high scientific and societal significance for policymakers to cope with the mitigation of regional haze pollution via adjustment of manmade aerosol emission planning according to climate predictions.
The remainder of this article is organized as follows. Section 2 describes the data and the methods used in this study. Section 3 shows the characteristics of our identified seesaw haze intensity case, investigates the respective roles of anthropogenic emissions and climate anomalies in the generation of this case and further explores the underlying mechanism. The discussion and conclusions are presented in Sections 4 and 5, respectively.

| Data
The datasets used in this study consist of the following: 1. Two near-surface PM 2.5 mass concentration datasets.
First, the long-term daily gridded PM 2.5 dataset across China, with a horizontal resolution of 1 Â 1 . Based on the space-time random forest model with atmospheric visibility observations and other auxiliary data, this PM 2.5 product is in excellent agreement with the ground measurements, archived at https://zenodo.org/record/4293239. Readers can  Huang et al., 2017), with a horizontal resolution of 2 Â 2 , covering the range of 1979-2019. 7. Global monthly precipitation data from NOAA's precipitation reconstruction (Chen et al., 2002), with a horizontal resolution of 2.5 Â 2.5 , covering the range of 1980-2019.

The monthly Community Earth System Model Large
Ensemble Numerical Simulation (CESM-LENS) datasets (Kay et al., 2015), from the National Centre for Atmospheric Research (NCAR) Climate Data Gateway (https://www.earthsystemgrid.org/dataset/ucar.cgd. ccsm4.cesmLE.html). Following previous studies (e.g., Piao et al., 2022;Zheng et al., 2018), this study focused on the ensemble mean results among 35 ensemble members of historical  CESM-LENS simulations. All members were forced under the same radiative conditions but initiated from small perturbations of the atmospheric states. The variables used here included atmospheric variables of geopotential height and aerosol optical depth (AOD) at 550 nm (horizontal resolution: 0.9 Â 1.25 ), as well as the oceanic variable SST, with 384 points Â 320 points in the horizontal direction (latitude by longitude). The pattern of NAT SST anomalies (SSTAs) is defined as the first leading empirical orthogonal function (EOF1) mode of SST in the North Atlantic sector (0-60 N, 0-80 W) (Chen et al., 2020b). The NAT SST pattern features same-sign SSTAs in the tropical and mid-latitudinal North Atlantic and opposite-sign SSTAs in the subtropical North Atlantic (Chen et al., 2020b). In this study, following Chen et al. (2020b), a positive NAT SST pattern corresponded to negative SSTAs in the tropical and midlatitude North Atlantic and positive SSTAs in the subtropical North Atlantic and vice versa for a negative NAT SST pattern. In addition, the NAT SST pattern index was obtained as the first principal component (PC1) associated with the EOF1 mode of SST in the North Atlantic (Chen et al., 2020b).

| Methods
Wave activity flux (WAF) was used to examine the propagation of stationary Rossby waves (Takaya & Nakamura, 2001) and was calculated using the following formula: where ψ 0 is the perturbation stream function, U is the horizontal wind speed, u(v) is the zonal (meridional) component of the basic flow and P is the pressure divided by 1000 hPa.
To exclude the potential impacts of long-term trends in the variables, all data were linearly detrended before the analyses. The two-tailed Student's t-test was used to evaluate statistical significance.

| RESULTS
3.1 | Characteristics of the temporal seesaw haze intensity case between the late winter months of 2010 over the NCP Figure 1a, b depicts anomalies of monthly mean surface PM 2.5 concentrations in January and February 2011, respectively. We identified a prominent temporal seesaw haze intensity case occurring between the late winter months of 2010 in the NCP, featuring suppressed in situ haze pollution intensity in January 2011 with noticeable negative PM 2.5 concentration anomalies ( Figure 1a) and an abrupt haze intensity enhancement in the adjacent month of February 2011, with noticeable positive PM 2.5 concentration anomalies ( Figure 1b). Note that the sharp intraseasonal alternation of PM 2.5 concentration anomalies was exceptionally conspicuous in the corresponding difference field between February and January 2011 ( Figure 1c). Furthermore, because the national-scale ground measurements of PM 2.5 concentrations in China commenced in late 2013 Li et al., 2022), the observational PM 2.5 concentration data in 2011 were unavailable. To qualitatively verify the accuracy of the reconstructed gridded PM 2.5 concentration data used in this study , the observed monthly mean visibility anomalies for three stations in Beijing (BJ), Shijiazhuang (SJZ) and Jinan (JN) from north to south within the NCP are shown (Figure 1a, b). Corresponding to noticeable negative PM 2.5 concentration anomalies, there were prominent positive visibility anomalies ( Figure 1a), with values of 6.3 km for BJ [exceeding 2 standard deviations (SDs)] and 2.6 km for SJZ and 4.9 km for JN (exceeding 1 SD). However, corresponding to noticeable positive PM 2.5 concentration anomalies, obvious negative visibility anomalies were observed ( Figure 1b), with values of À1.9 km for BJ (exceeding 1 SD), À0.6 km for SJZ and À 1.5 km for JN (exceeding 1 SD). The abrupt intraseasonal changes in visibility anomalies still denote the dramatic alternation of the seesaw haze intensity from more suppressed to more enhanced haze pollution between the late winter months of 2010, from the perspective of the dry extinction coefficient calculated based on observed visibility (Zhao et al., 2022).
Observational evidence of daily PM 2.5 concentrations in Beijing (a representative site for the NCP region; Yang et al., 2021) ( Figure S1) could provide a good interpretation of the temporal seesaw haze intensity. Beijing was slightly polluted in January 2011 ( Figure S1a), when Beijing had 11 healthy air quality days defined by a daily PM 2.5 concentration below 25 μg m À3  and only one severe haze day defined by daily PM 2.5 concentration above 150 μg m À3 (Cai et al., 2017), accounting for 36.7% and 3.3% of the total days, respectively. In stark contrast, the haze pollution intensity was considerably more severe in February 2011 (Figure S1b), when Beijing had only one healthy air quality day and nine severe haze days, accounting for 3.6% and 32.1% of the total days, respectively. In addition, the maximum hourly PM 2.5 concentration in February 2011 reached 595 μg m À3 (on 21 February), which was twice as high as that in January 2011 (286 μg m À3 on 3 January). Note that the intraseasonal enhancement in the frequency of severe haze days in Beijing was also the most prominent between the late winter months in the period 2009-2019 after removing the influence of year-to-year emission changes their Figure 1a). As a result, under the above-mentioned abnormal environmental conditions, a turnabout in the haze intensity from a notably suppressed state to a notably strengthened state can be formed between the consecutive months of January and February 2011 ( Figures S2a, b), generating the corresponding largest difference in PM 2.5 concentration anomalies during the last four decades ( Figure S2c).
Existing studies have established that anthropogenic emissions and climate anomalies are two fundamental factors that determine haze pollution variability (e.g., An et al., 2019;Dang & Liao, 2019;Wang et al., 2020a;Yang et al., 2016Yang et al., , 2018Yin et al., 2021). In the following sections, we will examine their respective roles in linking the temporal seesaw haze intensity case occurring between the late winter months of 2010 over the NCP.

| Role of emissions
In this study, we first examined the role of intraseasonal variations in emission anomalies. Regarding anthropogenic emissions, CO 2 emissions are a substitute proxy representing the amount of total fossil energy consumption (Zhao et al., 2020), which is associated with temporal haze pollution variations in China (Xu et al., 2015). From the spatial distribution of anomalous FFCO 2 emissions in central eastern China (Figure 2), we can discern that the estimated major anomalies of anthropogenic CO 2 emissions were concentrated in economically developed subregions of China with frequent haze occurrences (i.e., NCP, the Huang-Huai basin, Yangtze River Delta [YRD] and Pearl River Delta [PRD] [Yin et al., 2015;Mao et al., 2019;Chang et al., 2020]) and CO 2 emission anomalies intensified around the NCP and the YRD. However, noticeable FFCO 2 emission anomalies persisted from January 2011 (Figure 2a) to February 2011 (Figure 2b), exhibiting almost no changes in the anomalies of CO 2 emissions. Therefore, based on the aforementioned relatively invariant anthropogenic emission anomalies, it is plausible to speculate that abrupt sharp intraseasonal changes in climatic conditions could be major factors giving rise to the formation of our identified seesaw haze intensity case, rather than manmade emissions. In the following sections, we demonstrate the active roles of atmospheric and oceanic anomalies in driving the seesaw haze intensity case.

| Related atmospheric anomalies
First, we scrutinized the localized dynamic and thermodynamic fields associated with the seesaw haze intensity pattern shown in Figure 1. Figure 3 shows the anomalies of the monthly mean boundary-layer meteorological parameters, which are intimately tied to the formation/dissipation of winter haze pollution over the NCP (Wang et al., 2020b;Zhao et al., 2022). In January 2011, the near-surface NCP was dominated by negative relative humidity (RH) and temperature anomalies (  (Figures 3d2, 4h). As such, stable atmospheric stratification was established with a decreased PBLH in February 2011 (Figure 3e2-f2). Note that the aforementioned warm, humid and stabilized environmental conditions, together with decreased wind speeds, SLP and PBLH were highly conspicuous in the corresponding difference fields between February and January 2011 (Figure 3a3-f3).
It is essential to elucidate the processes of the above anomalous meteorological variables in forming the  identified temporal seesaw haze intensity case. In February 2011 (Figure 3c2-d2), low-level southerly anomalies with decreased speeds can be observed. In general, the low-level southerly anomalies were linked to an anticyclonic anomaly over Northeast Asia ; Figure 4h), which denotes a weakened East Asian winter monsoon (EAWM) with suppressed atmospheric ventilation (Wang et al., 2020b). Note that low-level southerly anomalies can induce decreased speeds over the NCP through reducing the mean wintertime northerlies locally. These meteorological conditions facilitated the transportation of PM 2.5 and water vapour from other polluted areas south of the NCP (e.g., YRD and PRD) to areas surrounding the NCP, sparking dense accumulation of PM 2.5 and warm and moist water vapour in front of the Yanshan and Taihang Mountains Yang et al., 2021). In such a scenario, stagnant meteorological conditions with repressed boundary heights can be set up, favouring the self-amplification and trapping of near-surface PM 2.5 via a positive feedback loop mechanism involving interactions among PM, PBL and water vapour Zhang et al., 2018). Thus, pronounced positive PM 2.5 concentration anomalies can be observed over the NCP (Figure 1b). In January 2011 (Figure 3c1-d1), however, northerly anomalies with positive wind velocity anomalies correspond to a cyclonic anomaly over Northeast Asia (Wang et al., 2020b; Figure 4g), causing increased wind speeds via enhancing the mean northerlies locally and thus favouring the intrusion of high latitude cold air with quite dry and pristine conditions into North China (Wang et al., 2020c). Accordingly, haze-conducive meteorology was substantially suppressed in January 2011 (Figure 3a1-f1), greatly reducing the haze pollution intensity with notable negative PM 2.5 concentration anomalies ( Figure 1a). As such, a seesaw haze intensity case can be generated ( Figure 1c). Furthermore, sharp intraseasonal changes in meteorological conditions (Figure 3) were driven by regionalscale circulation anomalies over the Northeast Asian-Western North Pacific (WNP) sector (Figure 4). In January 2011, a marked mid-to-upper tropospheric lowpressure anomaly (i.e., the NEACA) with an equivalent barotropic structure dominated the NCP and its surroundings south of 50 N (approximately 30-50 N), and a pronounced equivalent barotropic high-pressure anomaly (i.e., the NEAACA) dominated areas north of $50 N (Figure 4a, d). This north-south dipole pattern may facilitate the southward intrusion of lower-level northerly wind anomalies into the NCP (Figure 4g), thereby inducing localised enhanced atmospheric ventilation under cold and dry conditions. In sharp contrast, reversed patterns with opposite signs can be discerned in February 2011 (Figure 4b, e, h). An equivalent barotropic NEAACA controlled mid-latitudinal Asia (Figure 4b, e), thus inhibiting the transport of cold advection from midto-high latitudes toward the NCP. As such, intensified southerly anomalies dominated the NCP and its neighbouring areas (Figure 4h), exciting warm and moist conditions with suppressed atmospheric ventilation. Regional-scale circulation anomalies conducive to enhanced haze were exceptionally pronounced in the corresponding difference fields (Figure 4c, f, i). The atmospheric anomalies shown in Figure 4 are highly similar to those during below-normal and above-normal winter haze frequency years in Beijing (Wang et al., 2020b).

| Related oceanic anomalies
Existing studies suggest that regional-scale atmospheric anomalies responsible for wintertime haze pollution over the NCP are strongly linked to remote oceanic forcing from the North Atlantic Ocean (e.g., An et al., 2019;Wang et al., 2020a;Xiao et al., 2015). Figure 5 shows the monthly evolution of the SST anomaly pattern in the  (Chen et al., 2020b). For a better comparison, we carried out an EOF analysis of the SSTs in the North Atlantic domain for December and January ( Figure S3). The meridional patterns of the North Atlantic SST tripole anomalies (Figure 5b, c) were quite similar to the reversed EOF1 modes of the North Atlantic SST in December and January, which explained 29.4% and 32.1% of the total variance, respectively. Therefore, the patterns of the North Atlantic SST tripole anomalies shown in Figure 5b, c correspond to negative NAT-like SST anomaly patterns (Chen et al., 2020b). Additionally, large loadings in the dominant reversed EOF1 modes of the North Atlantic SSTs in December and January ( Figure S3) matched well with the vigorous centres shown in Figure 5b, c. As such, we can formulate a new NAT-like SST index (I NAT ) related to the North Atlantic SST tripole anomalies, which is defined as: where SST A , SST B and SST C denote the normalized areal mean SSTAs over regions A-C, respectively ( Figure 5). Our newly defined I NAT time series in January and February ( Figure 6) were highly correlated with the reversed PC1s of the North Atlantic SSTAs in January and February (not shown), with a temporal correlation coefficient (TCC) of 0.90 and 0.96, respectively. Therefore, in subsequent sections, we utilised the I NAT index to quantitatively characterize the corresponding North Atlantic SST tripole anomalies. The positive (negative) phase of I NAT corresponds to the negative (positive) NATlike SST anomaly pattern proposed by Chen et al. (2020b). The highest I NAT during 1980-2019 occurred in 2011 ( Figure 6). The above analyses allowed us to surmise that our identified seesaw haze intensity case may be strongly linked with the peak North Atlantic SST tripole anomalies in January. To quantify their correlation, we calculated the TCC between the difference in PM 2.5 concentration anomalies between January and February and December and January I NAT for the period 1980-2019 ( Figure 6). The corresponding TCCs were 0.34 (p < 0.04) and 0.39 ( p < 0.02), respectively. Note that the TCC between the difference in PM 2.5 concentration anomalies and the prior November I NAT was only 0.16 (p > 0.3). Based on the above analyses, we suggest that the temporal seesaw haze intensity pattern in the NCP between the late winter months was more significantly positively correlated with January I NAT . For the period 1980-2019, I NAT was the highest in 2011 (Figure 6), which could have prominent implications for our identified seesaw haze intensity case. Next, we focus on exploring the possible mechanism by which the January North Atlantic SST tripole anomalies link the generation of the seesaw haze intensity pattern in the NCP between the late winter months from the perspectives of teleconnection, regional-scale circulation anomalies and hazeassociated meteorology.
3.4 | Possible mechanism of how January North Atlantic SST tripole anomalies link the generation of the seesaw haze intensity case Previous research suggests that the NAT-like SST pattern is the dominant EOF mode of SSTAs over the North Atlantic sector, featuring strong interannual variations (Chen & Wu, 2017). It can exert a profound influence on far-reaching Eurasian regions by inducing teleconnections (e.g., Chen et al., 2016;Zuo et al., 2013). To add robustness to the modulation effect of the January North Atlantic SST tripole anomalies, we examined the composited difference results based on the five highest January I NAT years (1982, 1998, 2002, 2006 and 2011) and the five lowest January I NAT years (1983, 1985, 1987, 2015 and 2016), as shown in Figure 6. Figure 7 shows the composited difference of the 500 hPa geopotential height (Z500) for January and February. A clear significant North Atlantic Oscillation (NAO)-like meridional dipole pattern at 500 hPa with a negative phase (Ambaum & Hoskins, 2002;Czaja & Frankignoul, 2002;van Loon & Rogers, 1978) was detected in January (Figure 7a), which is closely coupled with the negative NAT-like SST anomaly pattern (Chen et al., 2020b). Corresponding to the North Atlantic SST tripole anomalies, we observed a significant simultaneous meridional 'positive-negative-positive-negative' quadrupole precipitation anomaly pattern over the North Atlantic Ocean (Figure S4a), suggesting a vigorous atmospheric response to local SSTAs. In this circumstance, a pronounced equivalent barotropic planetary-scale zonal Rossby wave train can be forced (Figures 7a and 8a), which emanates from the central North Atlantic with highly notable WAF anomalies (Figure 8a), acting as the source of the Rossby wave train. Alternating cyclonic and anticyclonic anomalies were embedded within the Rossby wave train, thus transmitting the influence of remote North Atlantic SST tripole anomalies downstream through a far-reaching teleconnection to form a mid-level NEACA around the NCP in January (Figure 7a). Consequently, haze-suppressed meteorology was established (Figure 9a1-f1), leading to significant negative PM 2.5 anomalies over the NCP in January ( Figure S5a).
In stark contrast, the quadrupole precipitation anomaly pattern was greatly dampened the following February ( Figure S4b), suggesting an insignificant atmospheric response to January North Atlantic SST tripole anomalies. As such, the negative NAO-like pattern attenuated drastically in February (Figures 7b and 8b) with decreased WAF anomalies over the North Atlantic sector (Figure 8b). However, the equivalent barotropic cyclonic anomaly centred southeast of the Yamal Peninsula, which was significantly enhanced in February (Figures 7b and 8b), may play a critical role in relaying F I G U R E 8 Regressed anomalies of 300-hPa stream function (shadings; 10 6 m 2 s À1 ) and horizontal WAF (vectors; m 2 s À2 ) in (a) January and (b) February onto the January I NAT during the period 1980-2019. Areas with significant values of stream function anomalies exceeding 90% confidence level are stippled.
F I G U R E 9 Composite differences of January (a1-f1) and February (a2-f2) meteorological parameters between five highest I NAT years and five lowest I NAT years in January, including 925-hPa RH (shadings; %) (a1-a2), 925-hPa air temperature (shadings; C) (b1-b2), 925-hPa wind speed (shadings; m s À1 ) (c1-c2), SLP (shadings; hPa) (d1-d2), PBLH (shadings; m) (e1-e2) and air temperature between 850 and 1000 hPa (shadings; C) (f1-f2). (a3-f3) The corresponding difference fields between anomalies of parameters in February (a2-f2) and January (a1-f1). Anomalies that are significant at the 90% confidence level in (a1-f1) and (a2-f2) are stippled, respectively. the impact of the January North Atlantic SST tripole anomalies, acting as an intermediate atmospheric role. Prominent WAF anomalies were observed there, propagating southward to northeastern Asia, thus forming a significant NEAACA in February (Figure 8b) to induce haze-favourable meteorology (Figure 9a2-f2). As such, significant positive PM 2.5 anomalies over the NCP can be formed in February ( Figure S5b). Note that hazefavourable meteorology and positive PM 2.5 anomalies were quite obvious in the corresponding difference field (Figures 9a3-f3 and S5c). Furthermore, from the February-minus-January Z500 difference field over the Northeast Asian-WNP sector, we observed highly noticeable positive anomalies around the NCP (Figure 10b), which suggests a notable mid-level NEAACA. This NEAACA could make a positive contribution to the observed critical system of the observed NEAACA (Figure 10a vs. Figure 10b), exerting an effective modulation effect on the generation of our identified seesaw haze intensity case.

| DISCUSSION
The following five points are worthy of further discussions. First, the physical motivation for building correlations of the difference in PM 2.5 concentration anomalies with North Atlantic SST tripole anomalies shown in Figure 6 is to examine the possible connection between the dramatic alternation of seesaw haze intensity over the NCP between late winter months and the tripolar North Atlantic SST pattern. Because the difference of PM 2.5 concentration anomalies could largely contributes to the dramatic seesaw haze intensity alternation according to previous studies (e.g., , we simply employ the difference in PM 2.5 concentration anomalies. To verify this assertion, it is essential to construct an index to quantitatively measure the dramatic seesaw haze intensity alternation over the NCP (DSHA NCP ). In light of the work of , the DSHA NCP index (DSHAI NCP ) is formulated as follows: where P 1 and P 2 are the normalised January and February PM 2.5 concentration in the NCP. P 1 j jþ P 2 j j ð Þ indicates the magnitude of the seesaw haze pollution, representing the abnormally high/low and low/high monthly PM 2.5 concentrations between late winter months. The weight coefficient term 2 À P 1 þP 2 j j acts as a balance role, reducing the weight of the suppressed or intensified haze pollution through the entire late winter months. The TCC between the detrended difference in PM 2.5 concentration anomalies (i.e., P 2 -P 1 ) and detrended DSHAI NCP for the period 1980-2019 is 0.88, explaining 77.4% of the total variance of DSHAI NCP . The above analysis do established the considerably large contribution of (P 2 -P 1 ).
Second, to validate the aforementioned remote forcing role of the January North Atlantic SST tripole anomalies obtained from observational analyses, following previous studies (e.g., Yang & Yuan, 2020;Yin et al., 2019), the results from CESM-LENS historical simulations were further utilized to confirm this modulation effect. Composite difference analysis suggested that the January North Atlantic SST tripole anomalies could be well reproduced by the ensemble mean of the 35 members of the CESM-LENS historical simulations ( Figure S6). Therefore, we can explore the composite differences of Z500 in January and February based on the ensemble mean results (Figure 11). From Figure 11a, we observe a significantly negative NAO-like pattern in January. However, compared with the negative NAO-like pattern shown in Figure 7a, the simulated pattern shifted more southward by approximately 10 . The simulated Z500 anomalies over the Northeast Asian-WNP sector also exhibited an obvious southward trend, and the NCP was controlled by weak negative Z500 anomalies, suggesting relatively cold conditions. For February (Figure 11b), the negative NAO-like pattern obviously dampened and shifted more southward. The simulated positive Z500 anomalies over the Northeast Asian-WNP sector were more extensive, extending from southern Japan to Lake Balkhash, which may be linked to the enhanced negative Z500 anomalies around the Yamal Peninsula. Compared to the observations (Figure 10b vs. Figure 10c), the February-minus-January Z500 difference based on the ensemble mean results exhibits highly extensive positive Z500 anomalies around the northeast region, confirming a positive contribution of the January North Atlantic SST tripole anomalies. Overall, although some biases existed in the simulated atmospheric circulations, CESM-LENS simulations could reproduce critical processes such as Rossby wave-like patterns over the region from the North Atlantic to Northeast Asia, the dampened negative NAOlike pattern and enhanced negative Z500 anomalies around the Yamal Peninsula from January to February, as well as the NEACA-like and NEAACA-like patterns. Moreover, AOD at 550 nm is an effective metric for measuring the severity of haze pollution (e.g., Tao et al., 2016). In response to the alternation of atmospheric circulations shown in Figure 11, we observed remarkable changes in simulated AOD anomalies over the NCP, shifting from negative in January ( Figure S7a) to positive in February ( Figure S7b), although positive AOD anomalies were observed in the northern portion of the NCP in January. Positive AOD anomalies were exceptional in the corresponding difference fields ( Figure S7c). Such simulated changes in AOD anomalies may also confirm the effective faraway modulation effect of January North Atlantic SST tripole anomalies. However, some intrinsic systematic biases existed regarding the simulated positions of the atmospheric anomalies over the North Atlantic and Northeast Asian sectors. The pathway by which the North Atlantic SST tripole anomalies influenced the Northeast Asian atmospheric circulations proposed in this study requires justification for future research. Furthermore, the mechanism behind the notable connection between the January North Atlantic SST tripole anomalies and the enhanced negative Z500 anomalies around the Yamal Peninsula in the following February deserves further investigation, which is beyond the scope of this study. Third, with the aim of lending further credence to the linkage between North Atlantic SST tripole anomalies and the seesaw haze intensity, we examined another notable abrupt subseasonal haze intensity enhancement case that occurred between the late winter months of 2008 ( Figure 6). We can also discern a similar negative NAT-like SST pattern persisting from November to January, which was obviously dampened in February 2009 ( Figure S8). Correspondingly, the abrupt intraseasonal alternation of circulation anomalies from cyclonic anomalies to anticyclonic anomalies between the late winter months of 2008 ( Figure S9), which are responsible for the formation of the seesaw haze intensity case, can be observed. In this circumstance, an abrupt change from local haze-suppressed meteorology to local hazeconducive meteorology was triggered ( Figure S10). Note that for different cases, there existed noticeable diversities regarding the positions and intensities of cyclonic and anticyclonic anomalies, as shown in Figure 4 and Figure S9. Therefore, the anomalous airflows over the NCP exhibit notable differences. For example, the airflow over the NCP box is easterly in Figure 4e whilst westerly in Figure S9e. However, in contrast to northwesterly ( Figure 4d) and northerly ( Figure S9d) in the context of cyclonic anomaly with cold conditions, both the easterly and westerly are tied to haze-conducive meteorology in the context of anticyclonic anomaly with warm conditions (Figure 3 and Figure S10), thus inducing subseasonal haze intensity enhancement cases between late winter months of 2008 and 2010. Additionally, we also showed a pronounced abrupt subseasonal haze intensity abatement case that occurred between the late winter months of 2011 ( Figure 6) to further examine the modulation role of North Atlantic SST tripole anomalies. As shown in Figure S11, there existed a positive NAT-like SST pattern. In this circumstance, an alternation of anticyclonic anomalies to cyclonic anomalies can be observed ( Figure S12), exciting an abrupt intraseasonal alternation of relatively haze-conducive meteorology to relatively haze-suppressed meteorology, which is quite evident in the corresponding difference field ( Figure S13). To conclude, we suggested that the North Atlantic SST tripole anomalies could make a positive contribution to the seesaw haze intensity case between the late winter months in the NCP, serving as a useful prediction source.
Fourth, in addition to North Atlantic SSTAs, numerous previous findings have identified various land surfaceocean-cryosphere systems that play pivotal modulation roles on wintertime hazy conditions over the NCP on the intraseasonal timescales (e.g., Li & Yin, 2020;Wang et al., 2020a;Yang & Yuan, 2020;Yin et al., 2021Yin et al., , 2022Yin & Wang, 2017Zhang et al., 2022). For example, SST cooling over the Mid-Atlantic Ridge can exert significant impacts on the aggravated haze pollution over north China from December to January through linking a prominent anticyclonic anomaly difference pattern over north China . Moreover, enhanced October-November mid-latitude Eurasian snow cover anomalies can drive a higher December haze frequency in the NCP via triggering a large-scale teleconnected wave train (Yin & Wang, 2018). Zhang et al. (2022) reported that the joint effects of warming signal in Arctic sea ice and the cooling signal in Eurasian soil temperatures can induce January 'warm Arctic-cold Eurasia' pattern, which is responsible for suppressed February hazy conditions in the NCP. Whether the aforementioned land surfaceocean-cryosphere systems have connections especially combined effects with our identified seesaw haze intensity case? If have, how to identify the predominant systems and how to understand the underlying physical mechanisms? Exploring these intriguing questions can deepen our understanding of the generation of such temporal seesaw haze intensity case, which merits more systematic analysis in the future work.
Fifth, one may ask whether the February North Atlantic SST tripole anomalies made a significant contribution to the seesaw haze intensity pattern over the NCP between the late winter months. In fact, however, the TCC between the detrended February PM 2.5 concentrations in the NCP and the simultaneous I NAT was extremely low, with a value of À0.01 (figure not shown). Therefore, the February North Atlantic SST tripole anomalies may exert negligible impact on the temporal seesaw haze intensity pattern. The haze-conducive meteorology in February 2011 was possibly attributed to other oceanic factors, which merits further exploration.

| CONCLUSIONS
This study identified a highly notable temporal seesaw haze intensity pattern occurring between the late winter months of 2010 over the NCP, with a considerably repressed haze intensity in January 2011 and substantially enhanced haze intensity in the adjacent month of February 2011. The major factor responsible for this temporal seesaw haze intensity pattern was the sudden dramatic intraseasonal alternations of atmospheric and oceanic conditions rather than changes in anthropogenic emission anomalies. The anomalous regional-scale circulation tied to the suppressed haze intensity in January 2011 was the equivalent barotropic NEACA, which dominated the NCP and its surroundings and induced the intrusion of lower-level mid-to-high latitude northerly anomalies into the NCP. In such environments, in situ enhanced atmospheric ventilation was established, featuring cold and dry conditions with elevated PBLH and destabilized atmospheric stratification. Therefore, hazesuppressed meteorology was generated in the NCP. In contrast, in February 2011, the NEAACA dominated the NCP with localised southerly anomalies, leading to the establishment of haze-favourable meteorology opposite to that in January 2011. Such a turnabout in local atmospheric anomalies directly contributed to the seesaw haze intensity.
Further diagnoses and numerical model simulations suggested that the negative January NAT-like SST pattern contributed positively to the aforementioned seesaw haze intensity case. In January, a significantly negative NAOlike pattern was connected to the negative NAT-like SST pattern, transmitting the influence of North Atlantic SST tripole anomalies downstream through a far-reaching teleconnection to form concurrent haze-suppressed meteorology. In the following February, although the NAO-like pattern was greatly dampened, the cyclonic anomaly centred southeast of the Yamal Peninsula could relay the impact of the January North Atlantic SST tripole anomalies, forming a significant NEAACA to induce simultaneous haze-favourable meteorology. As a result, a temporal seesaw haze intensity pattern between the late winter months of 2010 in the NCP can be induced. Nevertheless, we should acknowledge the limitation of the role played by the January North Atlantic SST tripole anomalies in the formation of the seesaw haze intensity case. RH anomalies over the NCP in January or February show obvious differences (Figure 3 vs. Figure 9). Higher (lower) I NAT years were linked to, by and large, positive (negative) RH anomalies with insignificant correlations (Figures 9a1, a2), which were opposite to the observations (Figure 3a1, a2). Therefore, North Atlantic SST tripole anomalies may not exert appreciable impacts on RH anomalies, which is key to the suppressed/enhanced haze intensity Tie et al., 2017).