The synergistic effect of the North Pacific and the North Tropical Atlantic sea surface temperature on winter–spring cold events in Eastern North America

Abstract

Cold events occur in North America (NA) frequently, which pose a significant threat to extensive damage to traffic and agriculture. Research on cold events in NA has mainly focused on the individual impacts of the sea surface temperature (SST) variabilities in the North Pacific and the North Atlantic, with less analysis of their synergistic effect. Here we utilize observational data and numerical experiments to show that the positive North Pacific Victoria mode (VM) and the negative North Tropical Atlantic (NTA) SST exert significant synergistic effects on the winter–spring cold events in eastern NA via Rossby waves propagation. Specifically, the VM-related negative western North Pacific (WNP) SST anomalies are linked to upper-tropospheric wind convergence anomalies in situ, which excite a Rossby wave train from the WNP to the North Atlantic and thus strengthen the high-level negative geopotential height anomalies (GHAs) over eastern NA and the positive air-ocean feedback over the North Atlantic. As a result, the negative GHAs strengthened by the positive VM and negative NTA SST cause increased extreme cold events. Further results reveal that the negative GHAs and cold events in eastern NA during the co-occurrence of the positive VM and negative NTA SST events are more than three times stronger than those for the two kinds of single events. This study suggests that the synergistic effect of the positive VM and negative NTA SST anomalies could advance our comprehensive understanding of the combined effect of the inter-basin SST anomalies on the mid-to-high-latitude climate.

Introduction

Variations in surface air temperature (SAT) have received increasing attention in recent decades due to their potential links to extreme weather and climate events^1,2,3,4. Particularly, the increased cold events in North America (NA) exert pronounced influences on agriculture, ecosystems, and human health^1,4,5,6. For instance, the record-breaking cold winters of 2013/14 and 2020/2021 in NA resulted in extensive damage to traffic and agriculture and affected people’s lives, with significantly high socioeconomic ramifications^1,4. Therefore, understanding SAT variability, especially the occurrence of cold events in NA is of great importance for making scientific organizational strategies and reducing the risk.

Multiple processes have been suggested to contribute to the cold events variation in NA, including local atmospheric circulation and lower boundary forcings^{7,8,9,10,11,12}. Specifically, the Arctic Oscillation (AO) or the North Atlantic Oscillation (NAO), as the dominant atmospheric mode over the mid-to-high-latitude North Atlantic, could control the variability of the SAT in NA^7,8,11,12. In addition, lower boundary forcings, like sea surface temperature (SST) in remote areas, may exert substantial influences on the SAT anomalies in NA^13,14,15. It is well established that SST anomalies over the tropical Pacific, particularly El Niño–Southern Oscillation (ENSO), could induce the Pacific–North American (PNA) atmospheric teleconnection^13,15, resulting in the anomalous climate conditions in NA¹⁴.

Beyond the tropical Pacific, SST anomalies in the North Atlantic have been extensively documented to play a crucial role in the SAT variability in NA^12,16,17. For instance, the North Tropical Atlantic (NTA) SST anomalies, coupled with the NAO during boreal winter and spring seasons^18,19, are closely linked to the SAT variability in NA^12,16. The winter SAT anomaly pattern over NA associated with NTA SST anomalies is characterized by a north-south dipole-like feature with strong positive anomalies over the northeastern NA and relatively weak negative anomalies over the southeastern US¹⁶. Besides, the Victoria mode (VM)^20,21,22, defined as the second dominant mode of SST anomalies in the North Pacific, is also an important factor influencing the climate in NA^22,23. Ji and Ding²² found that the VM could induce anomalous atmospheric circulation propagating from the western North Pacific (WNP) to eastern NA, leading to significant negative SAT anomalies in eastern NA^22,24. Thus, the VM has a significant impact on the winter–spring SAT variability in eastern NA, specifically, the positive VM favors the increase in cold events in this region.

Since both the winter–spring NTA SST and VM could affect the SAT variability in eastern NA, one may argue whether they have a combined effect on the variation in winter–spring cold events in eastern NA. To address this knowledge gap, we apply a new method put forward by Li, et al.²⁵ to determine the climatic synergistic effect of multiple factors and potential mechanisms based on observational datasets and numerical experiments. Our results highlight the joint importance of SST anomalies in the North Pacific and the North Atlantic on the occurrence of cold events in NA, thus providing insights into the combined role of the inter-basin SST anomalies in modulating the mid-to-high-latitude climate.

Results

Linkage of the VM and NTA SST with cold events in North America

Before investigating the relationships between the winter–spring SAT in NA and the VM and NTA SST, it is necessary to identify the spatial features and temporal variabilities of winter–spring (December–May; DJFMAM) SAT in NA. We extracted the first empirical orthogonal function (EOF1) mode of the SAT in NA based on NCEPI data (Fig. 1a). The EOF1 account for 26.5% of the total variance, featuring a meridional dipole pattern with out-of-phase SAT variations over the northern and southern NA continent, which is roughly consistent with the previous findings^10,26. We observed a similar pattern when using the ERA5 data (Fig. 1b) and the corresponding first principal components (PC1) between the two different datasets are highly correlated (R = 0.96; significant at the 99% level), indicating that the NA SAT variability is not dataset dependent. It is noted that the negative SAT anomalies dominate most regions of the mid-to-high-latitude over the eastern NA continent. Therefore, we defined a simple SAT index to describe the SAT variability in eastern NA as the domain-averaged northeast NA (95°–35°W, 45°–75°N) SAT anomalies based on the NCEPI data. As expected, the inverted domain-averaged northeast NA (95°–35°W, 45°–75°N) SAT index can well represent the PC1, supported by a high correlation coefficient of 0.98 (significant at the 99% level; Fig. 1c), suggesting that the SAT anomalies in eastern NA can effectively reflect the SAT variation in NA during winter–spring seasons. Hereafter, the SAT index was used to analyze the SAT variability in eastern NA, which is the focus region in the following.

Fig. 1: Spatial and temporal characteristics of the winter–spring surface air temperature (SAT) anomalies over North America (NA).

Spatial patterns of the first empirical orthogonal function (EOF1) mode of the NA SAT anomalies obtained by correlating the SAT anomalies with the normalized principal component (PC) time series of the EOF1 between 1950 and 2021 based on (a) NCEPI data and (b) ERA5 data. The numbers in the upper-left corners indicate the percentage of variance explained by EOF1. The red box represents the eastern North America region (95°–35°W, 45°–75°N) where the domain-averaged North America SAT anomaly index is defined using NCEPI data. Dots indicate the 95% confidence level based on a two-tailed Student’s t-test. c Normalized PC1 of the NA SAT anomalies derived from NCEPI data (green line) and ERA5 data (black line). The inverted SAT index is indicated by colored bars. The correlation coefficient between the inverted SAT index and the PC1 from NCEPI data is shown in the upper-right corner of (c).

To investigate the relationship between the NTA index and the SAT variation over NA during winter–spring seasons, we show the correlation map of the DJFMAM SAT anomalies with the NTA index (Fig. 2a), which is characterized by anomalous positive SAT over eastern Canada and Greenland. Specifically, the strong positive correlation indicates that negative SST anomalies in the NTA are usually accompanied by cold events in eastern NA. However, it is noted that not every nNTA year is accompanied by cold event (defined as a year when the winter–spring SAT index was less than −0.5 SD) in eastern NA. Some nNTA years correspond to the strong cold events in eastern NA, while the weak cold and even heat events occur in some nNTA years (Fig. 2b). This result implies that some factors might strengthen the effect of nNTA on cold events in eastern NA.

Fig. 2: Relationship between the NA SAT variability and SST anomalies.

a Correlation map between the winter–spring SAT anomalies and the NTA indices. b Time series of the SAT index (colored bars). The black dots represent the occurrences of nNTA events. c Regressed anomalous patterns of winter–spring SST (shading; units: °C) with respect to the SAT index. Dotted areas in (a, c) indicate significant values at the 95% confidence level. Boxes in (a, c) indicate the domain to represent eastern NA.

Next, we checked the anomalous SST associated with the variation in eastern NA SAT obtained by regressing SST anomalies on the inverted SAT index (Fig. 2c). As expected, the regression map exhibits a significantly negative SST anomaly pattern in the NTA region. In addition, there is a traditional positive VM-like structure in the North Pacific featuring a distinct dipole pattern oriented in the northeast–southwest direction with positive SST anomalies in the eastern North Pacific and a band of negative SST anomalies extending from the central North Pacific to the coast of Asia. It implies that the pVM may occur in the nNTA years and the pVM and nNTA could exert a synergistic effect on cold events in eastern NA. Thus, we further investigated the combined effect of these two factors on winter–spring cold events in eastern NA. Compared with the single pVM events and single nNTA events, the composite maps of SAT anomalies for the joint events pVM & nNTA have a large same-sign area in eastern NA (Fig. 3). Based on Li, et al. ²⁵, the synergistic effect of the pVM and nNTA SST anomalies on winter–spring SAT anomalies is significant in eastern NA (Fig. 3a), especially the region where the high correlations between the SAT anomalies and the NTA index are located (Fig. 2a).

Fig. 3: Composite maps of the winter–spring SAT anomalies.

Composite maps of the winter–spring SAT anomalies (shading; units: °C) for (a) the joint events of pVM and nNTA (pVM & nNTA). b As in (a) but for the single pVM events (pVM\nNTA). (c) As in (a), but for the single nNTA events (nNTA\pVM). Numbers at the top-right of (a–c) indicate the number of the three categories. The purple cross-hatched areas indicate the synergistic effect of pVM and nNTA. The white dotted areas indicate significant values at the 95% confidence level. Boxes indicate the domain to represent eastern NA.

To quantify winter–spring cold events from different categories, we show the time series of the SAT index with the marked different categories (Fig. S1a). Most of the joint events pVM & nNTA (10 out of 12) are corresponding to cold events. In contrast, for the single pVM events and single nNTA events, only one-sixth (2 out of 12) and a half (6 out of 12) show cold events, respectively, which are significantly less than those for the joint events pVM & nNTA. This indicates that there is a greater probability of cold events in eastern NA for the cooccurrence of the pVM and nNTA events than for the occurrence of only one factor. As expected, the strength of anomalous SAT in the joint events (−1.14) is nearly 3 times that of the single events (0.05 for the single pVM events and −0.37 for single nNTA events, respectively; see Fig. S1b). These results are consistent with the spatial distribution shown in Fig. 3, indicating the strong synergistic effect of the pVM and nNTA on winter–spring cold events in eastern NA, in other words, the VM plays a key role in strengthening the impact of the NTA SST on the variation in SAT over eastern NA.

Synergistic effect of the VM and NTA SST on cold events in eastern North America

To explore the mechanism responsible for the synergistic influence of the VM and NTA SST on winter–spring cold events in eastern NA, we first show the anomalous upper-level atmospheric circulation associated with the SAT in eastern NA obtained by regressing the 200-hPa geopotential height (Z200) anomalies on the SAT index (Fig. 4a). Corresponding to the spatial pattern of the SAT anomalies, anomalous positive geopotential heights are observed in eastern NA. To focus on atmospheric circulation variability related to the SAT anomalies over eastern NA, we defined a simple index, referred to as the Z200 index, as the normalized winter–spring Z200 anomalies averaged over the region (95°–35°W, 45°–75°N). Robust correlation between the Z200 and SAT indices (R = 0.81; significant at the 99% confidence level) suggests that the anomalous cyclone may lead to strengthening of cold events in eastern NA (Fig. 4b). It is noted that the Z200 index is highly correlated with the VM and NTA indices (R = –0.51 and 0.50, respectively; significant at the 99% confidence level). The results indicate that anomalous atmospheric circulation is key to the variation in SAT over eastern NA and may play an important role in the linkage between the SAT anomalies and VM and NTA SST. Moreover, the composite Z200 index for pVM & nNTA events (−1.08) is significantly greater than the values of the single pVM events (−0.27) and single nNTA events (−0.26; Fig. 4c). Thus, the two factors show a significant synergistic effect on the local atmospheric circulation over eastern NA, which may serve as a crucial bridge linking the VM and NTA SST and the SAT variability in this region.

Fig. 4: The atmospheric circulation associated with the SST anomalies in the North Pacific and the North Tropical Atlantic.

a Regressed anomalous patterns of winter–spring 200hPa geopotential height (Z200; units: m) with respect to the SAT index. Areas with regression significant at the 95% level are indicated by white dots. b Correlations of the Z200 index with the SAT (red bar), VM (yellow bar), and NTA (green bar) indices. The horizontal dashed lines indicate the 95% confidence level. c Composite mean of the Z200 index for the joint events winter-spring pVM & nNTA (red bar), single pVM events (yellow bar), and single nNTA events (blue bar), respectively. Error bars indicate 95% confidence intervals based on a t-test. d As in (a), but for the winter–spring tropospheric (1000–300hPa) atmosphere thickness (AT; units: °C). e As in (b), but for AT index. f As in c, but for AT.

Previous studies have suggested that upper-level atmospheric circulation over mid-to-high-latitude land regions is relatively insensitive to local surface warming or cooling^27,28,29. Thus, upper-level atmospheric circulation can influence the variations in surface temperature through adiabatic expansion or compression. Specifically, an anomalous high (low), corresponding to higher (lower) geopotential height, leads to warmer (cooler) local SAT^28,30. The anomalous upper-level circulation is conducive to an increase in tropospheric (1000–300hPa) atmosphere thickness. The increased atmosphere thickness anomalies could increase SAT by modulating adiabatic expansion/compression, indicating that SAT anomalies may be impacted by local circulation over eastern NA (Fig. 4d). Furthermore, there is a significant correlation coefficient of −0.45 and 0.56 between the atmosphere thickness anomalies over eastern NA and VM and NTA indices (significant at the 99% confidence level), respectively (Fig. 4e). As expected, the composite atmosphere thickness for pVM & nNTA events (−1.07) is significantly greater than the values of the single pVM events (−0.10) and single nNTA events (−0.27; Fig. 4f). The above results show that the pVM and nNTA have a synergistic effect on the atmospheric circulation over eastern NA, and the resultant decreased atmosphere thickness contributes to the strengthened cold events.

Figure 5 shows the composite maps of anomalous geopotential heights at 200-hPa and 850-hPa for different events. For pVM & nNTA events, significant negative geopotential heights are observed over eastern NA, corresponding to the anomalous SAT pattern. Moreover, the anomalous geopotential heights at 200-hPa (Fig. 5a) and 850-hPa (Fig. 5b) exhibit a similar wave pattern over the mid-to-high-latitude Northern Hemisphere, indicating an equivalent barotropic vertical structure. Particularly, there is a distinct wave pattern over the North Pacific extending to eastern NA, which could be induced by the VM (Fig. 5a)²². In addition, composite maps of anomalous geopotential heights over the extratropical North Atlantic feature a dipole pattern (Fig. 5a, b), which is reminiscent of the NAO, with a lobe of negative anomalies over Greenland and broad positive anomalies spanning the middle latitudes of the North Atlantic. Previous studies have suggested that the NTA SST anomalies could excite anomalous heating over the North Tropical Atlantic, inducing a Rossby wave train propagating northeastward, which leads to an NAO-like atmospheric pattern via transient-eddy feedback in the Atlantic jet exit^31,32. These suggest that the wave pattern of atmospheric teleconnections serves as a crucial mechanism linking the SST anomalies over the North Pacific and North Atlantic and the variation in the SAT over eastern NA, and the VM-induced atmospheric teleconnection may be of great importance to strengthen the impact of the NTA SST on the variation in SAT over eastern NA. Compared with pVM & nNTA events, the negative geopotential height anomalies in eastern NA and the wave pattern of atmospheric teleconnection are weaker for the single pVM events (Fig. 5c, d) and single nNTA events (Fig. 5e, f).

Fig. 5: Composite maps of the winter–spring anomalous atmospheric circulation.

As in Fig. 3, but for (a, c, e) 200hPa geopotential height (shading; units: m) and (b, d, f) 850hPa geopotential height (shading; units: m).

All the above analyses elucidate that the pVM and nNTA could exert a synergistic effect on the atmospheric circulation over eastern NA, contributing to the strengthened cold events. However, why the atmospheric circulation over eastern NA for pVM & nNTA events is clearly different from those for the two kinds of single events and the role of the VM in strengthening the impact of the NTA SST on the cold events in eastern NA remains unclear. To address the questions, we compared the Rossby wave propagation for different events obtained by the composite map of anomalous winter−spring 500-hPa Rossby wave activity fluxes. As shown in Fig. 6a, strong Rossby wave activity fluxes originate from the central North Pacific, propagating eastward across North America, and then turn southward to the mid-latitude North Atlantic for pVM & nNTA events. As such, a strong anomalous cyclone and NAO-like atmospheric teleconnection could occur over eastern NA and the North Atlantic, respectively. In contrast, the northward Rossby wave propagation from the North Tropical Atlantic for the single pNTA events is quite weak, corresponding to the weak NAO-like atmospheric circulation response over eastern NA (Fig. 6b).

Fig. 6: Composite wave activity flux (WAF) and 200hPa geopotential height (Z200) anomalies.

Composite WAF (vectors; unit: m² s^–2) and Z200 (shading; unit: m) anomalies for (a) the joint events of pVM and nNTA (pVM & nNTA). b As in (a) but for the single nNTA events (nNTA\pVM). c Regressed anomalous WAF (vectors; unit: m² s^–2) and Z200 (shading; unit: m) with respect to the VM index. In (a–c), contours and vectors indicate the values significant at the 95% confidence level. d As in (c), but for Rossby wave source (RWS; shading; unit: 3 × 10^–10 s^–2). The climatological 200-hPa zonal winds (contours; unit: m s^–1) obtained by the DJFMAM-averaged 200-hPa zonal winds during 1981–2010 are also presented.

How do the SST anomalies associated with the pVM events influence the strength of NAO? We show that the negative SST anomalies associated with the VM over the west-central North Pacific could induce significant positive RWS anomalies near the East Asian westerly jet core, which act as disturbances that propagate eastward, exciting a wave train-like pattern (Figs. 6c, d, 7a). This result is roughly consistent with the findings of Ji and Ding²². The resultant negative and positive geopotential height anomalies of the Rossby wave dominate eastern NA and the mid-latitude North Atlantic, respectively, leading to the strengthening of the NAO, especially the north pole of the NAO. Furthermore, given that the NAO could affect the SST anomalies in the North Tropical Atlantic via modulating the variation in turbulent heat flux^13,18, the enhanced NAO for the pVM & nNTA events may be accompanied by the strengthened NTA SST anomalies during winter−spring seasons (Fig. 7). It is noted that the amplitude of the NTA index for the pVM & nNTA events is obviously stronger compared with the single nNTA events (Fig. 7d). This enhanced amplitude is statistically significant above the 95% confidence level based on a bootstrap test (Fig. S2), implying that the VM may play an important role in the strengthened NTA SST anomalies. The strengthened NTA SST anomalies, coupled with the enhanced NAO, strengthen the positive air−ocean feedback between the NAO and the NTA SST for the pVM & nNTA events. This result could be supported by the CEOF analysis (Fig. S3), which indicated that the positive feedback for the pVM & nNTA events is significantly greater than that for the single pVM events and single nNTA events. The enhanced NAO−NTA SST positive feedback results in a stronger NAO. Thus, the pVM SST anomalies could lead to a strengthened NAO via the easterly propagating Rossby wave from the west-central North Pacific. The resultant strengthened positive air−ocean feedback between the NAO and the NTA SST for the pVM & nNTA events may lead to a stronger NAO, which is conducive to the cold events in eastern NA.

Fig. 7: Composite maps of the winter–spring SST anomalies.

Composite maps of the winter–spring SST anomalies (shading; units: °C) for (a) the joint events of pVM and nNTA (pVM & nNTA). (b) As in (a) but for the single pVM events (pVM\nNTA). c As in (a), but for the single nNTA events (nNTA\pVM). The purple cross-hatched areas indicate the synergistic effect of pVM and nNTA. The white dotted areas indicate significant values at the 95% confidence level. d Composite mean of the VM (red bars) and NTA (blue bars) indices for the pVM\nNTA, nNTA\pVM and pVM & nNTA events.

The synergistic effect simulated by numerical experiments

The observational results shed light on a synergistic effect of the pVM and nNTA on winter–spring cold events in eastern NA and its underlying mechanism. To further support our analysis, we examined this synergistic effect based on the output of the 35 members of the CESM Version 1 Large Ensemble (CESM1-LE) from 1950 to 2005. Most of the members of the CESM1-LE are able to reproduce reasonably well the spatial patterns of the VM and NTA SST. Specifically, the pattern correlations of the VM and NTA SST between most of the members and observations range from 0.5 to 0.9 (Figs. S4, S5). Thus, the VM and NTA indices in the CESM1-LE follow the same definitions as used for the observations. The composites mean of the simulated winter–spring SAT anomalies in the pVM & nNTA events, the single pVM events and single nNTA events for the 35 members of the CESM1-LE are shown in Fig. 8. We find that the composite SAT indices for the pVM & nNTA events are higher than those for the single pVM events and single nNTA events in most of the members of the CESM1-LE (30 of 35 members). The multi-member ensemble mean of the climate model shows that the composite mean of the SAT index for the pVM & nNTA events has a stronger strength than that for the single pVM events and single nNTA events, which provide further evidence that the pVM and nNTA have a significant synergistic effect on cold events in eastern NA.

Fig. 8: Synergistic effect of the VM and the NTA SST on winter–spring cold events in eastern North America based on the CESM1-LE.

Composite mean of the SAT index for the (a) pVM & nNTA, (b) pVM\nNTA, and (c) nNTA\pVM events from the 35 members of the CESM1-LENS. The horizontal lines indicate the multi-member ensemble means of the SAT index. (d) Multi-member ensemble means of the SAT index for the pVM & nNTA (red bar), pVM\nNTA (yellow bar), and nNTA\pVM (blue bar) events, respectively. Error bars indicate 95% confidence intervals based on a t test.

Furthermore, the ensemble-mean of composite Z200 anomalies for the pVM & nNTA events, the single pVM events, and single nNTA events are also evaluated (Fig. S6). The CESM1-LE results show clearly that the significant wave train-like pattern extends from the west North Pacific to the North Atlantic, leading to a strong negative geopotential height over eastern NA for the pVM & nNTA events, which is favorable to cold events in the region. Besides, a significant NAO-like pattern also could be observed over the North Atlantic, although the simulated center of the north pole of the NAO is shifted southward compared with the observations. In contrast, the atmospheric teleconnection over the North Pacific and the North Atlantic is relatively weak for the two kinds of single events. These results from the CESM1-LE model are very similar to those from observations, supporting the idea that the pVM and nNTA have a synergistic effect on the atmospheric circulation and thus cold events over eastern NA.

To make a stronger case that the SST anomalies of the pVM and nNTA exert a significant influence on the SAT anomalies in eastern NA, rather than merely being associated with it, we performed numerical experiments with Community atmosphere model version 5 (CAM5). Figure 9 shows the simulated anomalous SAT responses to the different SST anomaly combinations of the pVM and nNTA. It is found that distinct negative SAT anomalies appear in eastern NA for the VMNTA experiment (Fig. 9a). However, no obvious negative SAT anomalies can be simulated in the SVM and SNTA experiments (Fig. 9b, c). Furthermore, simulated anomalous atmospheric wave train responses to the SST anomaly forcing are well corresponding to the SAT anomalies (Fig. S7). These findings provide further support for the proposed hypothesis that the pVM and nNTA exert a substantial effect on winter–spring cold events in eastern NA via the wave-like atmospheric teleconnection. The consistency between observations and model simulations increases our confidence in the aforementioned synergistic effect of the pVM and nNTA on winter–spring cold events in eastern NA.

Fig. 9: Synergistic effect of the VM and the NTA SST on the winter–spring SAT based on the CAM5.

a Differences in SAT anomalies (shading; units: °C) between the VMNTA and CTRL experiments. b As (a) but for SVM experiment. c As (a) but for SNTA experiment. Boxes indicate the domain to represent eastern North America. The purple cross-hatched areas indicate the synergistic effect of pVM and nNTA. Boxes indicate the domain to represent eastern North America.

Discussion

Previous studies have identified that the North Pacific SST anomalies associated with the VM and the North Tropical Atlantic SST anomalies could exert a substantial influence on the variation in SAT anomalies in eastern NA. However, a systematic investigation of how the VM and NTA SST impact jointly the SAT anomalies in eastern NA is still lacking. In this study, we focused primarily on the combined effect of the VM and NTA SST on winter–spring cold events in eastern NA and the physical mechanisms responsible for this effect. We found that winter–spring cold events in eastern NA would be stronger when both factors occur in tandem. Specifically, anomalous SAT in eastern NA for the joint events pVM & nNTA was more than 3 times stronger than that for the two kinds of single events, indicating that the pVM and nNTA have a significant synergistic effect on the increased cold events in eastern NA.

Figure 10 shows the process of the synergistic effect of the pVM and nNTA on winter–spring cold events in eastern NA. The negative SST anomalies associated with the VM over the west-central North Pacific are linked to upper-tropospheric wind convergence anomalies in situ, which induce significant positive RWS anomalies around the East Asian westerly jet core. These RWS anomalies, act as disturbances, excite a wave train-like pattern propagating eastward from the west-central North Pacific to eastern NA. Hence, the high-level negative geopotential height anomalies of the Rossby wave dominate eastern NA. In addition, the pVM may increase the strength of the NAO via the easterly propagating Rossby wave from the west-central North Pacific, and the resultant strengthened positive feedback between the NAO and the NTA SST anomalies for the pVM & nNTA events may lead to a stronger NAO, which is favorable to the negative geopotential height anomalies over eastern NA. Therefore, the pVM and nNTA have a synergistic effect on the atmospheric circulation over eastern NA, and the significantly decreased atmosphere thickness induces a decrease in the air temperature of the atmospheric layer, which contributes to the strengthened cold events.

Fig. 10: Schematic diagram of the synergistic effect of the VM and the NTA SST on the winter–spring cold events in eastern NA.

Shadings over the North Pacific and the North Tropical Atlantic denote the patterns of the positive VM and the negative NTA, respectively. The blue box filled with blue oblique lines indicates the negative SAT anomalies in eastern NA. The blue vectors over eastern NA indicate the anomalous cyclone. The yellow vectors over the North Atlantic represent the positive feedback between the NAO and the NTA SST. The red (blue) circles with “A” (“C”) represent the positive (negative) the geopotential height anomalies at 200 hPa, the indicator of the atmospheric bridge. The yellow oval indicates the anomalous Rossby wave source related to the VM.

North America frequently suffered from cold events in recent decades, which seriously affected agriculture, ecosystems, and human health. Our findings reveal that the pVM and nNTA have a significant synergistic effect on the increase in cold events in eastern NA, which may provide implications for understanding the SAT variability and thus variation of extreme climate events. The results highlight that when considering the climatic influence of two factors, it is of great importance for better understanding its synergistic effect and the physical mechanism responsible for the synergistic effect on the regional climatic variability^25,33,34. In addition, it is also worth noting that although we discuss the synergistic effect of the winter–spring pVM and nNTA in this study, there might be other factors having synergistic effect with the North Tropical Atlantic SST. Besides the VM, there are SST anomalies over other ocean basins showing significant correlation with the variation in SAT in eastern NA (Fig. 2c), such as the North Indian Ocean and the Mediterranean Sea. These SST anomalies may also have a potential impact on the cold events in eastern NA synergistically. Given the complexity of the inter-basin interactions, combined effect of multiple factors is worth investigating in future research, which may help to understand the variation in winter–spring extremes in eastern NA.

Recent study indicated that the anthropogenic global warming would enhance the variability of the winter–spring VM³⁵. This suggests a potential enhancement of the impact of VM on winter–spring SAT in eastern NA in a warming climate. Previous studies have revealed a robust intensification of NTA variability under greenhouse warming, which results from strengthening of ENSO-forced PNA teleconnection and tropospheric temperature anomalies³⁶. Given the significant synergistic effect of the VM and NTA SST variabilities on cold events in eastern NA, investigating the combined impact and relative significance of the two factors in eastern NA SAT anomalies in a greenhouse warming scenario is imperative for improving future predictability of extremes.

Methods

Reanalysis data

The monthly SST dataset is provided by the Met Office Hadley Centre Global Sea Ice and Sea Surface Temperature dataset with a horizontal resolution of 1° × 1°³⁷. The monthly mean atmospheric variables used in this study include the air temperature, sea level pressure (SLP), geopotential height, which are obtained from the National Centers for Environmental Prediction-National Center for Atmospheric Research reanalysis 1 (NCEPI) with a 2.5° latitude × 2.5° longitude grid³⁸. To verify the robustness of the results, the SAT dataset from the European Centre for Medium-Range Weather Forecasts (ECMWF) Reanalysis v5 (ERA5) was also employed³⁹. All datasets cover the period from 1950 to 2021.

The VM index is defined as the second principal component (PC2) associated with the second empirical orthogonal function (EOF2) mode of SST anomalies in the North Pacific poleward of 20°N²¹. It is noteworthy that the VM is partly forced by the North Pacific Oscillation (NPO) atmospheric variability^21,40. To remove the effect of the NPO and isolate the internal variability of the VM, linear regression was conducted with respect to the November–April (NDJFMA) averaged NPO index prior to analysis. The NPO index is defined as the PC2 associated with the EOF2 of SLP anomalies over 20°–60°N, 120°E–80°W⁴⁰. The NTA index is defined as the averaged SST anomalies in the NTA region (0°–20°N, 80°–20°W). The Niño3.4 index, which is used to represent ENSO variability, is defined by SST anomalies averaged over the 5°S–5°N, 170°–120°W region. To isolate the climatic influence of the VM and NTA SST independent of ENSO, the effect of the simultaneous winter’s ENSO has been removed from the VM and NTA indices, respectively, using linear regression with respect to the Niño3.4 index. The NAO index is obtained from the National Oceanic and Atmospheric Administration (NOAA) Climate Prediction Center website (http://www.cpc.ncep.noaa.gov/data/indices). These indices from 1950 to 2021 were all linearly detrended and standardized using population standard deviation (SD).

CESM Large Ensemble data archive

To examine the synergistic effect of the VM and the NTA SST on winter–spring cold events in NA, we used output of the 35 members of the Community Earth System Model Version 1 Large Ensemble (CESM1-LE) simulations forced with historical radiative fluxes from 1920–2005 available at https://www2.cesm.ucar.edu/models/experiments/LENS⁴¹. The simulation uses a 1-degree latitude/longitude version of CESM1.

Definition of synergistic effect

To examine whether there is a synergistic effect of two factors, the VM index and NTA index on cold events, we used the method of Li, et al. ²⁵ to statistically diagnose the combined effect, which is widely used to determine the existence of synergistic effect^33,34. We first defined the positive VM (pVM) and negative NTA (nNTA) events as years when the winter–spring VM index was greater than 0.5 SD and the winter–spring NTA index was less than −0.5 SD, respectively. Further, the three categories for the pVM and nNTA are defined as the following. The pVM & nNTA denotes the cases with the cooccurrence of pVM and nNTA events. The pVM\nNTA indicates the cases when pVM events occur without nNTA, while the nNTA\pVM represents the cases when nNTA events occurs without pVM (Table 1). If the |SAT| of pVM & nNTA is greater than the maximum of the |SAT| of pVM\nNTA and nNTA\pVM, it denotes a synergistic effect of the pVM & nNTA on the SAT. However, if the |SAT| of pVM & nNTA is smaller than the minimum of the |SAT| of pVM\nNTA and nNTA\pVM, it represents an antagonistic effect of pVM & nNTA. In addition, if the |SAT| of pVM & nNTA is between the |SAT| of pVM\nNTA and nNTA\pVM, it represents that there is no combined effect of pVM & nNTA. As a result, we can compare the differences between different categories to determine whether there is a synergistic effect.

Table 1 Numbers and years of the joint events of winter–spring pVM and nNTA events (pVM & nNTA), single pVM events (pVM\nNTA) and single nNTA events (nNTA\pVM)

Significance tests

The statistical significance of the correlations, regressions, and composite in observations and the models was determined using a two-tailed Student’s t-test. Bootstrap test⁴² was conducted to examine whether the changes in the VM and NTA SST indices are statistically significant under different categories. A total of 10,000 realizations were conducted to obtain the composites based on pVM & nNTA, single pVM, and single nNTA events. Each realization was independently and randomly resampled. The SD of the 10,000 realizations was calculated for each category of the events. If the mean value difference between the joint events and single events was greater than the sum of the two separate 10,000-realization SD values, then the difference was deemed statistically significant above the 95% confidence level.

Dynamics diagnosis methods

A coupled EOF (CEOF) analysis⁴³ was applied over the 0°–20°N and 80°–20°W, and 20°N–70°N and 80°W–20°E domains to identify the temporally covarying spatial patterns of the NTA SST and the North Atlantic SLP, respectively, which explain most of the covariance between the two fields⁴⁴. The temporal evolution of these patterns is demonstrated by coupled PC (CPC) index for SST and SLP anomalies.

The perturbation hypsometric equation was applied to explore the effect of the anomalous atmospheric circulation on the variation in the SAT^30,45:

$${\left\langle T\right\rangle }^{{\prime} }\,=\,{{{g}_{0}\atop}\!\!}/R{\left({ln}{{{p}_{1}\atop}\!\!}/{p}_{2}\right)}^{-1}\varDelta Z$$

(1)

where $\left\langle T\right\rangle^{\prime}$ is the anomaly of the mean temperature of the layer, $\varDelta Z$ is the atmospheric thickness between two pressure surfaces ${p}_{1}$ and ${p}_{2}$, ${g}_{0}$ is the gravitational acceleration, and $R$ is the gas constant of dry air. It can be seen that the perturbation mean temperature of the atmospheric layer is proportional to the perturbation atmospheric thickness bounded by isobaric surfaces. Thus, the atmospheric thickness can represent the perturbation mean air temperature of the atmospheric layer. We used this perturbation hypsometric equation of the layer mean temperature to discuss the influences of upper-level atmospheric circulation on SAT anomalies.

The wave activity flux⁴⁶ was used to investigate the stationary Rossby wave source (RWS) and propagation induced by the VM in this study. It is a phase-independent flux and is parallel to the local group velocity of a stationary Rossby wave train in the Wentzel–Kramers–Brillouin (WKB) approximation. The horizontal flux is calculated according to the following equation:

$$W=P/2\vec{|U|}\left\{\begin{array}{c}U\left({{v}^{{\prime} }}^{2}-{\psi }^{{\prime} }{{v}^{{\prime} }}_{x}\right)+V(-{u}^{{\prime} }{v}^{{\prime} }+{\psi }^{{\prime} }{{u}^{{\prime} }}_{x})\\ U\left(-{u}^{{\prime} }{v}^{{\prime} }+{\psi }^{{\prime} }{{u}^{{\prime} }}_{x}\right)+V({{u}^{{\prime} }}^{2}+{\psi }^{{\prime} }{{u}^{{\prime} }}_{y})\end{array}\right.$$

(2)

where P, $\vec{U}$ = (U, V), ψ ′ and $\vec{V}$′ = (u′, v′) denote relative pressure (pressure/1000 hPa), basic state wind velocity, perturbation geostrophic stream-function and perturbed wind velocity, respectively.

To investigate the source of the Rossby waves, following Sardeshmukh and Hoskins (1988)⁴⁷, the RWS can be described by the barotropic vorticity equation:

$$S=\,-\nabla \left(\overrightarrow{{V}_{\chi }}{\xi }_{a}\right)=\,-{\xi }_{a}D-\overrightarrow{{V}_{\chi }}\cdot \nabla {\xi }_{a}$$

(3)

where S is the source/sink of the Rossby wave; ${\xi }_{a}$ is the absolute vorticity, $\vec{{V}_{\chi }}$ denotes the divergent wind vector, and D is the horizontal divergence. This equation indicates that the divergence of vorticity flux contributes to the generation of the RWS²⁴.

Numerical simulation

To further verify the synergistic effect of the VM and the NTA SST on winter–spring cold events in NA, we conducted several modeling experiments based on the CAM5⁴⁸, which has 30 hybrid sigma-pressure vertical levels with a horizontal resolution of 0.937° × 1.25° (latitude × longitude). The climatological monthly mean SST, as a reference state, was used to force the atmospheric model (CTRL). In addition, a group of sensitivity experiments (VMNTA, SVM and SNTA) was designed. To inspect the synergistic influence of the VM and NTA SST, in the VMNTA experiment we added the composite DJFMAM SST anomalies for the pVM & nNTA case on the climatological SST in the extratropical North Pacific (20°N–65°N, 120°E–100°W) and North tropical Atlantic (0°–20°N, 80°W–20°W) regions. The other two experiments (SVM and SNTA) are the same as the VMNTA experiment, except that the SST anomalies are the composites for single nNTA (pVM\nNTA) and single pVM (nNTA\pVM) cases, respectively. Each experiment was integrated for 40 years, and only the last 30 years of the integrations were taken for analyses to avoid any influence of the initial conditions.