Why could ENSO directly affect the occurrence frequency of Arctic daily warming events after the late 1970s?

The National Centers for Environmental Prediction/National Center for Atmospheric Research (NCEP/NCAR) Reanalysis data is used in this study, which includes the daily and monthly air temperatures, horizontal winds, vertical wind velocities, and geopotential height with a spatial resolution of 2.5° latitude-longitude at the tropospheric level. Monthly sea surface temperature (SST) data with a horizontal resolution of 1° latitude-longitude is obtained from the Hadley Center sea ice and sea surface temperature dataset. We also use daily air temperature, horizontal winds, and monthly SST from eight CMIP6 models under historical simulation from 1950 to 2014 and three shared socio-economic pathway (SSP) based scenarios (SSP1-2.6, SSP2-4.5 and SSP5-8.5; Eyring et al 2016) from 2015 to 2100. Here, only one ensemble member ('rli1p1f1') run is used for each model. The basic information for the CMIP6 models is shown in supplementary table S1.

To better describe Arctic daily warming, we choose Pacific-RTDW events (Wang et al 2019) that satisfy the following two criteria, as follows: (1) over the Arctic region (66.5°–90° N), the daily increment anomaly of air temperature at 1000 hPa is greater than a standard deviation in the winters of 1950–2018; and (2) the daily warming event is triggered by the 1000 hPa southerly anomaly from the Pacific. In Wang et al's (2019) paper, the criterion of three standard deviations presents 13 °C daily warming and the number of extreme events is very small. In the present paper, using the criterion of one standard deviation, not only can these events catch the characteristics of the rapid warming over the Arctic but also enough events are identified to explore the relationship between the Pacific-RTDW and ENSO.

The daily 'increment anomaly' of air temperature ${{\Delta }}T_{{\text{anomaly}}}^i$ in the definition is obtained by a two-step calculation. First, the daily air temperature 'increment' ${{\Delta }}{T^i}$ is calculated as ${T^i} - {T^{i - 1}}$, where T is the daily air temperature, the superscript 'i' is the day of interest and 'i-1' is the previous day. Then, the daily air temperature 'increment anomaly' ${{\Delta }}T_{{\text{anomaly}}}^i$ is expressed as ${{\Delta }}{T^i} - {{\Delta }}T_{{\text{mean}}}^i$, where the subscript 'mean' is the climatological mean of daily air temperature increments from 1950 to 2018 on that day of the year. For example, in order to calculate the air temperature 'increment anomaly' on 31 December 2000, the daily air temperature 'increment' on 31 December 2000 is obtained first by subtracting the daily air temperature on 30 December 2000 from the daily air temperature on 31 December 2000. Then, the 'increment anomaly' is calculated as the difference between the air temperature increment on 31 December 2000 and the average of the air temperature increments on every 31 December from 1950 to 2018. For convenience, Pacific-RTDW day −3 and day 0 denote 3 and 0 days before the occurrence of the Pacific-RTDW event, respectively.

The Pacific Ocean and the Atlantic Ocean are the two primary energy sources for Arctic warming (Overland and Wang 2016, Graham et al 2017). In addition to Pacific-RTDW events, the other Arctic daily warming events that meet the first criterion above are Atlantic pattern Arctic rapid tropospheric daily warming (Atlantic-RTDW) events. Events are classified as Pacific-RTDW events according to the second criterion by using the fuzzy c-means method (FCM; Fujibe 1989, 1999), which is a classification method and does well in classifying representative patterns. Moreover, the number of clusters is equal to 2 and other parameters are defaulted when the FCM is used. Given that the Atlantic Ocean is the other primary energy source for Arctic warming (Graham et al 2017, Wang et al 2019), the Atlantic-RTDW event is also important for Arctic daily warming events and will be examined in future work. In the present paper, we focus on Pacific-RTDW events.

There were 437 Pacific-RTDW events identified in the winters from 1950 to 2018, with an average of 6.3 events each winter. However, the frequency of Pacific-RTDW events varies greatly between winters, with the Arctic experiencing 16 such daily warming events in the winter of 1988 but only one in 2009. The Pacific-RTDW index is defined as the number of Pacific-RTDW events that occur each winter (December–February; also called the occurrence frequency of Pacific-RTDW events). The successive warming days are counted separately; that is, each event is weighted by its duration time. There was almost no significant linear trend of the Pacific-RTDW index during 1950–2018. One possible explanation is as follows: the Pacific-RTDW event is based on the temperature difference between the day of interest and the prior day. As a result, the trend impact contained in these 2 days is greatly offset.

The Niño3.4 index is defined as the winter (December–February) SST anomalies averaged over 5° S–5° N, 170° W–120° W. The El Niño event is identified by the 3-month sliding mean of the Niño 3.4 index above 0.5 and persisting for at least 5 months (https://origin.cpc.ncep.noaa.gov/products/analysis_monitoring/ensostuff/ONI_v5.php). The El Niño years (supplementary table S2) selected for composite analysis are 1950–1977 and 1978–2018.

Wave flux analysis (Takaya and Nakamura 2001) can reveal quasi-stationary Rossby wave energy propagations. It is also a suitable diagnostic tool for detecting propagating quasi-stationary planetary waves.

The composite anomalies for the Pacific-RTDW event are calculated by averaging the geophysical field anomalies over Pacific-RTDW day −3 and day 0. For the P1 and P2 periods, the composite anomalies for the ENSO events are calculated as the winter geophysical field of the El Niño years minus the average of the corresponding period, and the differences between the two periods are calculated as the winter geophysical field of the El Niño years in the P2 period minus those in the P1 period. The two-tailed Student's t-test is used to test the significance of the composite anomalies.

3.1.1. Synoptic atmospheric conditions for the Pacific-RTDW event

3.1.2. Large-scale atmospheric conditions associated with the frequency of Pacific-RTDW events

To explore the winter large-scale atmospheric conditions associated with the frequency of Pacific-RTDW events, anomalies of winter geopotential height and horizonal wind speeds are regressed on the Pacific-RTDW index. In the positive phase of the Pacific-RTDW index, there is a meridional tripolar system from the tropics to the Arctic in the mid–high troposphere (figures 1(a) and (b)). Two cyclonic anomalies appear over the Arctic and south of Hawaii, and the anti-cyclonic anomaly is over the Aleutian region. We refer to this pattern as the typical Pacific-RTDW pattern (figures 1(a) and (b)). For the typical Pacific-RTDW pattern, there are notable southerly anomalies between the trough (cyclone) and ridge (anticyclone) in the middle troposphere from mid latitudes to the Arctic, which is a necessary condition to produce a guiding flow that permits the cyclones with warm and humid air masses to enter the polar region (figure 1; e.g. Murray and Simmonds 1995, Qiangen et al 2000).

Zoom In Zoom Out Reset image size

At the near-surface, the Pacific-RTDW related cyclone and anti-cyclone are, respectively, located over the Arctic and the Aleutian (figure 1(c)), consistent with mid–high latitude anomaly centers in the mid–upper troposphere, which show equivalent barotropic structures. However, there is no significant anomaly center over the south of Hawaii, which displays a baroclinic property. These findings suggest that local convective heating may be responsible for the baroclinic anomaly centered over low latitudes and the barotropic anomalies centered over the mid–high latitudes may result from teleconnection in the middle troposphere.

Furthermore, the meridional tripolar system originates from the East Central Tropical Pacific and may be induced by ENSO. This motivates us to address the question whether the occurrence frequency of Pacific-RTDW events is affected by ENSO. Next, their relationships will be investigated in greater detail.

Prior to the late 1970s, the Pacific-RTDW index and the Niño3.4 index showed an uncertain phase relationship; however, after the late 1970s, the two indices were frequently out of phase (figure 2(a)). Specifically, the 31-year moving correlation shifts from insignificant to significant in 1993 and maintains a relatively stable significance level of p < 0.05 over 1993–2003 (the central years of 1978–2018). These results suggest that ENSO has had an influence on the frequency of Pacific-RTDW events since the late 1970s. As a result, the entire study period is divided into two subperiods: the P1 period (1950–1977) and the P2 period (1978–2018). Note that there is a lack of independence of the time periods in the blue line in figure 2(a) and CMIP6 models will be investigated in section 3.4 to add more samples.

Zoom In Zoom Out Reset image size

To further demonstrate the strengthened effect of ENSO on the frequency of Pacific-RTDW events, figures 2(b) and (c) show the correlation distributions of SSTA with the Pacific-RTDW index over two periods. There is no significant correlation region of the Pacific-RTDW index with the SSTA in the East Central Tropical Pacific in the P1 period, while statistically significant negative correlations occupy the East Central Tropical Pacific in the P2 period (figure 2(b) and (c)). During the P1 period, the correlation coefficient between the Pacific-RTDW index and the winter Niño3.4 index is only 0.11 (insignificant), while it achieves a statistically significant level (p < 0.02) in the P2 period, which is −0.39 (figures 2(d) and (e)). This finding reminds us that ENSO may have an enhanced effect on the frequency of Pacific-RTDW events after the late 1970s and motivates us to explore the possible mechanism.

In addition, there was an interdecadal change in ENSO intensity in the late 1970s (Lee and McPhaden 2010, McPhaden et al 2011). Therefore, we hypothesize that the change in the relationship between the frequency of Pacific-RTDW events and ENSO is attributed to the ENSO intensity change. This will be confirmed in the following subsection.

The ENSO-related SSTA, which could determine the remote impacts on the global atmosphere (e.g. Li et al 2021), is examined. In the warm phase of ENSO, the ENSO-related spatial manifestation of SSTA is characterized by warm anomalies ranging from the dateline to South America in both periods. The spatial distributions of SSTA are similar, whereas the intensity of SSTA in the P2 period is greater than that in the P1 period (figures 3(a)–(c)). Therefore, more details about the atmospheric response to the enhanced SSTA are needed.

Zoom In Zoom Out Reset image size

A stronger significantly anomalous ascending motion centered over the equatorial region is seen in the P2 period (supplementary figure S2(b)), corresponding to stronger diabatic forcing. In contrast, the magnitude of such ascending motion is much smaller in the P1 period (supplementary figure S2(a)). These configurations reflect the tropical local atmospheric circulation's response to changes in ENSO-related SSTA intensity between the P1 and P2 periods. We will now gain further insight into the role of different diabatic forcings between the two periods in impacting the mid–high latitudes of the Northern Hemisphere (NH).

The influence of tropical variability could be brought into mid–high latitudes by poleward propagating Rossby waves triggered by ENSO (e.g. Hu et al 2016). The ENSO-related Rossby wave activity fluxes and quasi-geostrophic stream functions for the two periods are compared in figures 3(d)–(e). During the P2 period, the Rossby wave exhibits an obvious northwards propagation trajectory in the warm phase of ENSO, which emanates from the East Central Tropical Pacific toward the Arctic (figure 3(e)). The corresponding quasi-geostrophic stream function shows a strong low over the Aleutian region. The Aleutian Low is characterized by the North Pacific (NP) index, which is defined as the area-weighted sea level pressure over the region (30° N–65° N, 160° E–140° W; Trenberth and Hurrell 1994). The correlation between the NP index and the Niño3.4 index is −0.61 in the P2 period at the significance level of 0.01. However, in the P1 period, the Rossby wave propagating northwards hardly exists, and the quasi-geostrophic stream function anomalies over the eastern part of the NP are also much weaker (figures 3(d)–(f)); the correlation between the NP index and the Niño3.4 index is −0.25, which is not statistically significant.

In the warm phase of ENSO, the intensities of cyclonic anomalies over the NP in the two periods are obviously different (figures 3(g)–(i)). In the P1 period, the cyclonic anomaly center over the NP is weak, while it becomes much stronger in the P2 period (figures 3(g)–(i)), indicating a greater deepening of the Aleutian Low in El Niño in the second period. Correspondingly, in the warm phase of ENSO, the resulting consistent northerly anomalies to the west of the Aleutian Low dominate over the NP in the P2 period, which could inhibit typical Pacific-RTDW event occurrences (figures 1(a) and 3(h)). Similarly, in the cold phase of ENSO, there are obvious southerlies over the NP in the P2 period, which could increase the Pacific-RTDW event occurrences. Thus, in the P2 period, ENSO could influence the occurrence frequency of Arctic daily warming events. However, no obvious meridional wind anomalies can be observed over the same region for the P1 period (figure 3(g)), implying that ENSO did not impact the occurrence frequency of Pacific-RTDW events.

In the following, the Rossby wave activity fluxes and quasi-geostrophic stream functions associated with the Pacific-RTDW index for the two periods are examined in figure 4. During the P1 period, there is almost no significant wave flux activity associated with the Pacific-RTDW propagating northwards from the East Central Tropical Pacific (figure 4(a)). The corresponding cyclonic and anticyclonic anomaly centers are only seen over the NP and the west of the Bering Strait (figures 4(a) and (c)), respectively, which is different from the ENSO-related circulation anomalies. These phenomena mean that the connection between the Pacific-RTDW and ENSO is weak in the P1 period. The Pacific-RTDW may be influenced by other factors, which will be studied in the future.

Zoom In Zoom Out Reset image size

In contrast, in the P2 period, the wave activity flux associated with the Pacific-RTDW exhibits an obvious northwards propagation trajectory (figure 4(b)); the resulting stream function shows that anticyclonic and cyclonic anomaly centers appear over the East Central Tropical Pacific and the NP (figures 4(b) and (d)), respectively. These results are consistent with the ENSO-related wave flux activity and circulation anomalies (figures 3(e) and (h)), implying that ENSO could have an impact on the Pacific-RTDW during the P2 period.

Since the late 1970s, the potential physical mechanism of the ENSO's strengthening effect on the occurrence frequency of Pacific-RTDW events could be interpreted as follows. During the P2 period (1978–2018), enhanced ENSO excites and reinforces a Rossby wave propagating toward the pole by stronger diabatic heating, which causes a deepened Aleutian Low accompanied by northerly anomalies over the NP to decrease the occurrence frequency of Pacific-RTDW events in the warm phase of ENSO. In the cold phase of ENSO, it increases the occurrence frequency of Pacific-RTDW events. Therefore, in the P2 period, ENSO could directly affect the occurrence frequency of Pacific-RTDW events. However, due to the weak ENSO intensity, there are no such impacts in the P1 period. Under a global warming scenario, ENSO intensity will most likely determine the connection between the tropics and the Arctic.

By using a set of eight CMIP6 models under historical simulation and three future scenarios, the correlation coefficient of the Pacific-RTDW index with the Niño3.4 index (shortened as 'Corr') and the standard deviation of the Niño3.4 index (shortened as 'STD') are calculated. 'Corr' denotes the relationship between the frequency of Pacific-RTDW events and ENSO, and the 'STD' denotes ENSO intensity. Furthermore, figure 5(a) depicts the relationship between 'Corr' and 'STD'. The inter-model correlation coefficient is −0.43 (p < 0.02), indicating that 'Corr' is clearly related to 'STD'. This is consistent with the above findings that ENSO intensity could affect the relationship between ENSO and the frequency of Pacific-RTDW events. Moreover, we divided the models into high Niño3.4 STD models with the standardized 'STD' greater than 'X' and low Niño3.4 STD models with the standardized 'STD' less than 'X' multiplied by −1 (table S3). We set 'X' to 1.5 to obtain the extreme Niño3.4 STD models. Figures 5(b) and (c) show that there are more extreme ENSO events in the high Niño3.4 STD models than those in the low Niño3.4 STD models. Furthermore, in the high Niño3.4 STD models, the correlation coefficient between the Pacific-RTDW index and the Niño3.4 index is −0.42 (significant; figure 5(b)); however, it is insignificant in low Niño3.4 STD models (figure 5(c)). We also set 'X' to 0.5 and 1.0 to define the high and low Niño3.4 STD models, and the results support the above findings (figure S3). These results further confirm that the ENSO intensity could influence the relationship between ENSO and the frequency of Pacific-RTDW events.

Zoom In Zoom Out Reset image size

Furthermore, for the same model, as the radiative forcing increases from SSP1-2.6 to SSP5-8.5, so does the ENSO intensity (figure 5(a)). As a result, ENSO will continue to affect the Arctic with global warming in the future. Previous studies have found a wide range of historical simulated ENSO intensities among the CMIP5 and CMIP6 models (Wengel et al 2018, Brown et al 2020, Planton et al 2020). In terms of ENSO intensity under future scenarios, different CMIP5 models behave very differently, with some increasing and others decreasing or remaining unchanged (Guilyardi et al 2012, Stevenson 2012). Most CMIP6 models show an increase in future ENSO intensity (Fredriksen et al 2020). In the present study, we simply diagnose the impact of ENSO intensity on the relationship between ENSO and Arctic daily warming events by using CMIP6 models. The evaluation of the CMIP6 model, e.g. whether the models reproduce the ENSO statistical data reasonably, the relationship between the ENSO and Pacific-RTDW, and many aspects of ENSO in the future merit further investigation.