A new approach to cold surge classification in East Asia

Evidence showing a strengthening of intense cold surge event (CSE) in East Asia, e.g. CSE of Jan 2016 and Jan–Feb 2008, is focusing attention towards the science of CSE onset prediction. Predicting the onset of such strong CSEs remains elusive as the extent of these surges varies over spatial and temporal scales. Changes in radiative cooling over Siberia in winter as potentially affected by changes in the Arctic are further expected to influence CSE occurrences in East Asia. Moreover, unprecedented and long lasting CSEs in East Asia have a very distinct Jet Stream pattern via their shifts from the climatological mean, influencing the lower troposphere. Here, using modelling framework we propose a new relationship between Jet Stream and Aleutian Low for identifying and characterizing atmospheric process that leads to CSEs in East Asia. Our results reveal new insight into the mechanisms of CSEs occurrences, the absence of which may lead to major constraints on reducing CSE onset prediction error.

www.nature.com/scientificreports/ 80° E-120° E) from 1979-2016 for Dec, Jan, and Feb. The highest MSLP of 1040 hPa is attained in Jan (2011) followed by 1037 hPa in Dec (2005) and 1034 hPa in Feb (1988). During this 38-year period, there is an overall increase in the MSLP strength for Jan whereas no significant trend is observed for Dec and Feb. This observation is consistent with the recent research that shows Jan is the most critical month for the outbreak of cold air mass 17 and that the SH has been intensifying gradually 22 . This strengthening is further attributed as a response to Arctic amplification as related to albedo changes from loss of snow and sea ice, and the presence of heat-trapping clouds and water vapor 23 . Figure 1a further illustrates the SH intensification is well marked during the intense cold air outbreak as evident during CSEs of Dec 2005, Jan 2011, Feb 1988. However, during the onset of unusual CSEs that occurred during Jan 2016, Jan-Feb 2008, Dec 2009 the SH intensification was found to be comparatively lower. Recent research shows that the intensity of atmospheric circulation over SH to the surface environment over the AL, as characterized by a pressure difference (PD) between the two leads to a more robust indicator of cold air progression in East Asia 17 . Here the AL domain is defined as 25° N-75° N and 160° E-130° W for determining the monthly average AL MSLP used in the computation of the PD. Our analysis shows that the amplified period of PD is more evident for the latter period of 1999-2016 (Fig. 1b) as compared to the earlier period from 1979 to 1998, with exception of Jan 1981, which was characterized by a deep polar vortex and a strong planetary wave 24 . This suggests that there is a recent increase in PD amplification. Thus e.g. Jan 2016 had the strongest PD of 33 hPa in the last 38 years followed by Jan 1981. For Dec, the highest PD of 31 hPa was observed in 2005, while for Feb the highest PD of 26 hPa was observed in 1980. This further shows that Jan is the most crucial month for the progression of cold air outbreak followed by Dec and Feb. Moreover, this amplifying period is also coincident with the warmer decadal period, 2001-2010 which was the warmest on record, with five recent years, viz., 2016 (1.2 °C above preindustrial baseline), 2015, 2017, 2019 (all three years are 1.1 °C above pre-industrial baseline), and 2018 (1.0 °C above pre-industrial baseline) being the warmest as reported by World Meteorological Organization 25,26 . However, regional variation of PD could be a complex process, resulting in varying intensities of CSEs at different locations due to changes experienced in the atmosphere-ocean circulation systems. Thus, the ability to accurately predict the progression of CSEs necessarily needs an understanding of the dynamical processes that significantly influence the synoptic patterns.

Connection between JS and AL
The physical process of the upper troposphere needs to be considered as they are likely related when explaining the mechanisms for cold air outbreak and CSE progression. Our analysis further shows that for the month of Jan, a jet streak within JS defined as extending from 125° E to 165° E (Fig. 2a, pinkish color) reaches a monthlyaveraged maximum speed of 70 m/s at its center. As the air masses enter the JS it accelerates into the upstream side of the streak which we term here as an 'entrance' . The accelerating air masses reach their maximum speed at the core of the JS i.e. center of the jet streak and start decelerating which we term here as an 'exit' . These accelerating air masses as it enters jet streak shows two entrance regions around the jet streak, referred to here as the left entrance and right entrance regions, as seen in the vertical cross section along with AB. There is convergence at the left entrance and divergence at the right entrance (Fig. 2b). As the air masses exit jet streak, they decelerate and create divergence at the left exit and convergence at the right exit (Fig. 2c), as shown in the cross section along CD. These regions indicate warm air rising in the right entrance region and cold air sinking in the left entrance region (Fig. 2b). Whilst cold air is sinking in the right exit region, and warm air rising in the left exit region (Fig. 2c). Such changes in the upper air thus significantly influence the pressure systems at the lower troposphere. For example, during the CSE of Jan 2016, the jet streak attained a maximum speed of 100 m/s which is well above its climatological Jan mean of 70 m/s as the JS progressed over the western Pacific (Fig. 3a). The cross sectional plots along AB and CD show the rising and sinking air masses (Fig. 3b,c) as resulting from the progression of the diverging air masses at the entrance and exit inducing low level convergence, acceleration of lower winds, and enhanced rising motion towards the upper level (Fig. 3d). This further resulted in the lowering of the surface pressure beneath the atmospheric column, leading to the formation of a low pressure system (Fig. 3e). Once formed the low pressure system progresses eastward in the direction as inherited from the diverging air masses of the jet streak i.e. in the JS progression direction. It later merges with AL, and consequently intensifying the AL system (Fig. 3c). Therefore, the intensification of jet streak contributes to the genesis of the low pressure system which significantly influences the progression of CSEs in terms of their intensity. The evolution of the synoptic weather experienced during the surge month allows an overview of the synoptic scale pressure systems that trigger CSE conditions. The synoptic composites of the 31 days average of strengthening of jet streak wind speed (where the jet streak anomaly is 6.67 m/s above the 1981-2010 average) and intensification of AL (where AL MSLP anomaly is (−) 6.14 hPa below 1981-2010 average) leading to the onset of CSE 2016 are shown in supplementary Fig. 1. The cross sectional variations in jet streak wind speed and MSLP presented in the figure show the synoptic structure associated with the CSE that leads to its intensifying and weakening stages (represented in Fig. 5 on the mechanism). The cross sections, analyzed from the reanalysis dataset, are centered at 35° N for the jet streak and 36° N for MSLP based on the climatological mean with the longitudinal extent of 29 and 17° respectively. The air as it enters the jet streak (extreme left) starts accelerating and reaches the maximum speed and as the air leaves the jet streak it decelerates (extreme right). During this period within the same longitudinal domain (marked within the red box) sudden intensification of AL MSLP is observed. This shows that the influence of the progression of easterly jet stream brings unstable conditions over the AL domain, as observed by the sudden decrease in AL MSLP in contrast to the gradual increase in AL MSLP for the domain that is not under the direct influence of jet streak wind speed.
As noted above, we analyzed the formation of low pressure systems due to the progression of the jet stream during the reported CSEs period (please see Table 1 for a list of reported CSEs). For all these events, jet streak wind speed has a positive anomaly while low pressure system has negative anomaly relative to the climate normal www.nature.com/scientificreports/ (Fig. 4a). This shows the intensification of low pressure systems with the strengthening of jet streak wind speed. Moreover, we also observed the formation of series of low pressure systems before the onset of the CSEs. These low pressure systems are seen to be most intense and more frequent during the month of Jan, followed by Dec, and Feb depending on the strength of the jet streak wind speed (Fig. 4b). This further confirms that jet stream progression results in the formation of low pressure systems which progress eastward following the direction of the jet stream to intensify AL. Intense AL well establishes PD leading to the progression of cold air masses at lower latitudes as CSEs (Fig. 4c). Therefore, with the outbreak of cold air mass and under the influence of JS, the formation of low pressure systems marks the developing stage of CSEs. Whilst the continued formation of the low pressure systems intensifies AL, leading to well established PD resulting in CSE onset. The weakening of jet streak wind speed marks the decaying stage of CSEs due to no PD establishment. To support the development of appropriate diagnostics for detecting the onset of CSEs, a generalized description of the key mechanisms that cause CSE in East Asia is shown schematically in Fig. 5a,b. First, this is based on the analysis of 10 such CSEs (Table 1)  We have seen that jet streak plays a key role in the formation of the low pressure system. Apart from the intensification of the jet streak, the pattern of JS also significantly influences the progression of CSEs in terms of their stability. Jet streak strength is also dependent on its longitudinal extent in manifesting the role of diverging air masses in inhibiting the intensification of CSEs at the surface. For example, the larger longitudinal extent of jet streak relative to its climatology leads to a fully strengthen jet streak resulting in the formation and development of intense low pressure systems. For example, under the influence of a fully developed jet streak with longitudinal extent > 15°, intense low pressure systems resulted in the strongest AL anomaly (supplementary Table 1 Table 1). We, therefore, categorize the Jet Stream into two subcategories, one with a fully developed Jet Streak i.e. spanning over 15° in longitude, and the other with a partially developed jet streak i.e. spanning less than 15° along the longitude. Moreover, unusual long lasting CSEs are observed to be associated with partially developed jet streak embedded within a slower progressing jet stream relative to the climate normal. For instance, during the CSE progression over Jan-Feb 2008, a distinct pattern of zonal JS over three different stages is observed that remained stable for a period lasting 20 days. First, the embedded jet streak reaches a stall-like situation, slowing to 50 m/s, well below the climatological mean of 70 m/s. For comparison, the WMO's adopted minimum JS speed is 30 m/ s 27 . The jet streak then starts to strengthen gradually from 60 m/s on 26th Jan 00 UTC (Fig. 6a) and continually increases till 17th Feb 00 UTC reaching a maximum of 100 m/s (Fig. 6b). Second, this pattern of JS has an elongated continuous stretch extending from 70° E up to 180° E. This distinctive JS pattern with an embedded jet streak (within longitude 130° E to 180° E; Fig. 6b), extended beyond its climatological longitudinal extent of 125° E to 165° E (Fig. 2a). Lastly, the JS later started to split from 17th Feb at 12 UTC (Fig. 6c) and further weakened as it moved eastward (Fig. 6d). Over this duration, the regions of Vietnam, South China, Hong Kong, and Taiwan were under the influence of consistent slow moving JS with its speed ranging from 35 to 55 m/s. The slow progression of JS shows no significant difference in wind speeds at the different stages of JS over these regions. Under the continued influence of JS, Hong Kong Observatory recorded the minimum temperatures below 12 °C for 24 consecutive days, making it the longest cold spell in Hong Kong since 1968 28 . In contrast, the regions of South Korea (Busan), and Japan (Fukuoka) experiences an increased wind speed ranging from 40 to 100 m/s due to the influence of jet streak embedded within the JS and also explicitly shows the changes in wind speed at the four different stages of JS (Fig. 7). However, it's still unclear at this stage what determines the nature of JS patterns. It has been linked that changes in the Arctic are intensifying SH gradually, influencing synoptic pattern 29 . Therefore, one possibility could be the Arctic amplification influencing the upper air and JS patterns, but this requires further investigation. www.nature.com/scientificreports/ (as based on its PD, Fig. 1b), the jet streak intensity was observed to increase rapidly and persistently for 8 consecutive days over 19th to 27th Jan (Fig. 9a, black box). The jet streak wind speed attained a maximum value of 68 m/s (corresponding to the maximum value attained during that month) on 27th Jan 18 UTC, starting from a value of 52 m/s on 19th Jan 00 UTC. During the same 8 day period, the AL continually dipped from 1011 hPa on 20th Jan 06 UTC to a minimum value of 996 hPa on 26th Jan 06 UTC (corresponding to the minimum value attained during the month) (Fig. 9b, black box). The strength of the jet streak contributes to AL intensification. It is also observed that AL intensifies significantly as a function of jet streak duration, such that the longer duration exhibits higher AL intensification. For example, as the jet streak peaks for 8 days, from 19 to 27th Jan 2016, a consistent intensification in AL MSLP is observed. In contrast, as the jet streak peaks for 5 days from 1st to 5th Jan 2016, the AL MSLP dips initially from 1 to 3 days but then rises for one day and then again dips from 4 to 5th day. Comparatively, with a shorter duration of strengthening the jet streak, an inconsistent drop in AL MSLP is observed. The longer duration of the jet streak allows a continuous progression of diverging air masses at upper air that leads to the formation of intense and frequent low-pressure systems at the surface, resulting in enhanced AL intensification (Fig. 3e). Therefore, under the influence of the diverging air masses of the strengthened jet streak, the continual intensification of AL along with corresponding increased PD led to a severe weather situation that lasted a week over East Asia. Furthermore, AL MSLP becomes more variable as the embedded jet streak strengthens and weakens within the jet stream. For example, for the most stable (longlived) CSE over Jan-Feb 2008, the jet streak increased rapidly for 7 consecutive days over 1st Jan-9th Jan UTC (Fig. 9c, black box). The jet streak wind speed attained a maximum value of 60 m/s on 9th Jan 06 UTC starting from 49 m/s on 1st Jan 18 UTC. During the same period, a sudden and drastic drop of MSLP over the AL was observed. The MSLP suddenly dipped 12 hPa from 1014 hPa on 1st Jan at 00 UTC to 1002 hPa on 3rd Jan 18 www.nature.com/scientificreports/ UTC (Fig. 9d, black box). Thereafter it started to rise gradually till 9th Jan at 18th UTC but remained below 1012 hPa i.e. below the climatological normal 1013 hPa pressure. However, when jet streak weakens from 9 to 13th Jan and 17th to 25th Jan, AL MSLP shows only a modest response that seems unlikely to further increase its intensity and remains concentrated primarily with jet streak strength. As seen, three repeated episodes of jet streak strengthening coupled with AL intensification illustrate the casual effect amongst the two that resulted in the longest ever reported CSE in January 2008 over East Asia. Our further analysis based on long term records  shows that the mechanism driving this response is linked to the embedded jet streak wind speed within the jet stream that influences AL variability. We have seen that the vertical transverse circulations around jet streaks are the significant link that affects the regions within and underneath, leading to the formation of low pressure systems that intensifies AL at the surface (Figs. 2, 3). Scatter plots between JS anomaly and AL anomaly show that that strengthen JS plays a key role in AL intensification (Fig. 10). Strengthen jet streak with a positive anomaly intensifies AL. These cases are shown within the black box in Fig. 10. However, we also observe that strengthened JS with positive anomaly may not intensify AL under certain conditions. There are additional factors that also plays role in intensifying AL apart from jet streak wind speed. This includes pattern of JS in terms of its longitudinal shifts from the climatological mean position. For instance, the westward shift in JS mean position (e.g. Jan 2012 JS anomaly Supplementary Fig. 4a) weakly intensifies AL as the low short lived low pressure systems formed may de dissipated before progressing towards AL. In addition, the presence of a relatively smaller longitudinal spread of JS allows weaker diverging air masses due to lower momentum gained by the air masses within the jet streak and warrants no formation of low pressure systems. Therefore, a Larger westward shift > 3.50 E or the presence of relatively smaller longitudinal spread of JS compared to its climatology (e.g. Jan 2009 JS anomaly Supplementary Fig. 4b) does not intensify AL. These cases are marked with the red box in Fig. 10.

Proposed CSEs classification.
Our results here demonstrate a proximate explanation of CSE onset with the synoptic weather patterns prevailing at that time. We further analyzed 10 strong CSEs that have resulted in severe socio-economic disasters in East Asia as reported in Emergency Events Database (EM-DAT) https:// www. cred. be/ proje cts/ EM-DAT (Table 1). These CSEs were analyzed and ranked based on three identified classes of synoptic pattern drivers. This comprises the strength of cold air outbreak over SH (Class I), intense progression of CSEs via atmospheric circulations over SH and AL as characterized by PD (Class II), and long lasting CSEs (Class III) as resulting from the directional shifts of JS and its speed from their climatological mean values. Such Class III CSEs have durations extending over a week and up to a month as compared to Class I and Class II CSEs that last for a few days. These climatology means cover over a period of 38 years . For Class I, the monthly mean SH MSLP exceeding 1029 hPa, 1031 hPa, and 1028 hPa for Dec, Jan, and Feb, respectively are taken to trigger cold air outbreaks over SH. For Class II, the establishment of intense SH-AL PD exceeding 18 hPa, 16 hPa, 21 hPa for Dec, Jan, and Feb, respectively either from SH or AL intensification are taken as the criteria for intense progression of CSE. Class III CSEs are ranked by durations and described as long and short durations based on their reported prevailing periods 30-32 . Long duration is those extending over two weeks, while average comprise 1-2 weeks and short less than one week. The duration of the Class III CSEs are further enhanced with the weakening of the jet streak speed.
One advantage of the classification shown in Table 1 is that it allows an inter-comparison of CSEs dominance under the three different synoptic weather drivers. For example, the most severe outbreak of cold air occurred for www.nature.com/scientificreports/ Jan 2011 CSE under Class I as arising from the strengthening of the SH system above its threshold values. Class II considers an established PD as the SH-AL coupled systems with the strongest PD established during the Jan 2016 CSE. As the PD usually controls the atmospheric circulation and influences the movement of moisture, such Class II CSEs can alter the regional weather situation including a sudden drop in temperature and rise in wind speed. Long-lasting Class III CSEs could be explained by considering both the lower and upper troposphere as a system, specifically the deviation of the central jet streak position and jet streak speed from the climatological mean. This significantly influences the AL intensification arising from the formation of intense low pressure systems. This approach allows an approximate characterization of long lasting CSEs based on jet streak shift, its speed, and AL intensification together influencing Class III CSEs duration as shown in Table 2. In Table 2, the position shift is the primary criteria in identifying the duration of CSEs followed by the jet streak speed as directly correlated with the AL intensification (Fig. 5). Also, the CSE in 2008 extended over two months period severely impacting two regions comprising central China in Jan 2008 2 , and regions of southern China and Hong Kong in Feb 2008 24 . Hence this is considered as two CSEs in EM-DAT and in Table 1. Going down the Class III CSEs in Table 1, the longest duration Class III CSE in Jan 2008 has a westward shift (southwest shift) of the jet streak from its climatological mean and the smallest Jan jet streak speed along with slower eastward progression (Tables 1 and  2, Supplementary Fig. 5a,b). Such a slower eastward progression is via weaker jet streak velocities being lower than the climatological mean of 62 m/s, 70 m/s, and 68 m/s for Dec, Jan, and Feb, respectively. Most importantly, the westward shift (southwest shift) of a slower moving jet streak allowed an abundance of cold air from the polar region to intrude at lower latitude. In contrast, the next four Class III CSE (rank 2 to 5) have eastward shift www.nature.com/scientificreports/ towards north (northeast shift) of the jet streak (Tables 1 and 2) for the CSE in Feb 2008 (a lower jet streak speed case), and for CSEs in Jan 2016, Feb 2012, and Jan 1981 (higher jet streak speed cases, Supplementary Fig. 5a,c). This slower to faster eastward progression in jet streak results in an increasing (lower to higher) AL intensification, and in decreasing (long to short) duration CSE (Tables 1 and 2   www.nature.com/scientificreports/ considered as the first indicator, Class II CSEs are the likelihood of Class I CSEs to further intensify and progress based on PD establishment, and finally, Class III CSEs that provide more information on the duration of Class II CSEs from the influence of JS progression and positioning. Moreover, it is also noted that not all strong cold air outbreaks (Class I) lead to Class II or III CSEs. Since Class I CSEs occur due to the cold air outbreak over the Siberia region, the progression of the cold air masses is dominant usually up to Mongolia or may extend up to northern China region depending on the intensity of the outbreak. For example, a CSE that was reported in Jan 2000 was confined over the Mongolia region and remained as a Class I CSE ( Supplementary Fig. 6a, and listed in Table 1 under NA). This CSE could not further progress as a Class II CSE due to the weakening of PD.
In contrast, the CSE reported in Jan 2006 over Japan region intensified from Class I to Class II CSE due to the strengthening of PD, allowing an abundance of cold air to progress towards the Pacific (Supplementary Fig. 6b; Table 1 under NA). However, this CSE failed to further intensify to be Class III due to only weak intensification of AL under the influence of the jet streak. These classifications of CSEs are further extended for a long term period (1979-2016) based on the above criteria. All Class I CSEs had an outbreak of cold air masses with SH MSLP anomaly being positive for Dec, Jan, and Feb respectively (Fig. 11). Once CSEs qualify for Class I, the next step is the establishment of the pressure gradient between SH and AL marked with a positive PD anomaly. The PD anomaly then drives the CSEs either as Class II or Class III based on SH or AL dominance respectively. For long lasting Class III CSEs, the mean position of JS is shifted westward within 0-3.5° E from its climatology. Our analysis shows that AL intensification is well correlated with the strengthening of jet streak wind speed for Jan as compared to Dec and Feb. This is likely because the jet streak wind speed climatology reaches its highest value in Jan (63 m/s) followed by Dec (55 m/s) and Feb (60 m/s). Higher jet streak wind speed during Jan results in the formation of more intense low pressure systems intensifying AL. Intense SH marks the outbreak of cold air masses. A strong SH even leads to CSEs for the regions in the vicinity of the outbreak of cold air masses. However, for the progression of these cold air masses at lower latitudes as CSEs, they are dependent on the establishment of PD. AL significantly contributes to the establishment of PD depending on JS wind speed and its longitudinal extent. During the month of Dec, Jan, and Feb, JS progression continually intensifies AL due to the frequent formation of low pressure systems leading to Class II and Class III CSEs, whereas the outbreak of strong cold air masses primarily leads to Class I CSEs alone.

The way forward
The recent rise in the intensity of SH would make the strong cold air outbreak even stronger (e.g. the Jan 2011, Class I CSE in Table 1) and the larger amplification in PD would probably further intensify CSEs (e.g. Jan 2016, Class-II CSE in Table1). Whilst intensification of AL under the direct influence from JS would lead to more stable long lasting surges (e.g. Jan 2008, Class III CSE in Table 1)). In the recent past, the use of temperature related indicators for interpreting strong CSEs have led to major forecast failures (e.g. 2008, 2016). Therefore, exploring the underlying physics that leads to the onset of CSEs is not only of scientific interest but also of particular use towards reliable predictions.
The role of atmospheric circulation has strong implications on how it influences CSE onset. Our analysis shows that the tropospheric disturbances associated with the progression of JS and its longitudinal extent of the embedded jet streak, result in the genesis of the synoptic disturbances at lower troposphere as evident by the formation of low pressure systems. These together lead to strong CSEs onset in East Asia 17 . Moreover, JS positioning from its climatological mean further play a key role in the onset of long lasting CSEs. For example, the unusual extent of JS as extending over southwest China to northern Pacific during the Jan 2008 CSE (Fig. 6) indicated significant progression of the cold air masses at the upper atmosphere. This unusual extent lasted for 14 days continuously leading to colder than average weather over south central China during the period. A schematic showing the pathways that influence the East Asia weather system and CSE onset during the north east monsoon is provided in Fig. 12. Arctic amplification strengthens SH system 17,22 and alters JS in the high and mid latitudes 33 . The three major systems i.e. SH amplification, coupled SH-AL intensification via the PD, and the position and structure of the jet streak within the JS alter the progression of CSEs. Together these cause changes in CSEs intensity and its propagation direction under the three classes of CSEs.
This research thus shows a new relationship between JS and AL for identifying and characterizing atmospheric process that leads to CSEs classification in East Asia. The classification allows an inter-comparison of CSEs dominance and provides insight into the mechanism leading to unusual CSEs. Besides requiring a fuller understanding of this relationship, possible further steps are to investigate and characterize CSEs based on possible contributions of from the mid-latitude or high-latitude processes such as the impact of Arctic amplification on temperature extremes to further enhance deciphering of casual linkages that enable CSEs predictions for impact and risk based forecasting.

Methods
Dataset (year 1979-2016) obtained from the European Centre for Medium Range Weather Forecasts (ECWMF) (https:// www. ecmwf. int/ en/ forec asts/ datas ets/ reana lysis-datas ets/ era-inter im). Analysis was performed using the analysis fields from ERA-Interim global data archived at ECMWF (archival since 1979). The dataset was post processed to use a 4-dimensional data environment using the Grid Analysis and Display System (GrADS) software (http:// www. dmc. fmph. uniba. sk/ public_ html/ doc/ grads/ head. html). Standard statistical analysis was performed on the data sets using Mann-Kendall (MK) test to determine the monotonic decreasing trend on the time series of the meteorological variables. For measuring the statistical relationship between AL and JS system, Pearson's correlation coefficient was computed based on the method of covariance. Pathways that summarize potential mechanisms that contribute to amplified, intensified, and more persistent weather patterns in East Asia leading to CSE onset. The pathway highlighted by blue shaded boxes is presented in the manuscript.