Revisiting the impact of mid-latitude cold air outbreaks on the Maritime Continent weather

The East Asian winter exhibits frequent cold air outbreaks (CAOs), or cold surges, which often cause severe cold waves in the extratropics. CAOs are also of importance for forecasters in the Maritime Continent because the associated equatorward outflows can influence tropical weather. Various CAO definitions have been documented, but a definition based on a quantitative approach was only recently proposed. Studies of the CAO impact often linked a surge index in the subtropics instead of an index in the midlatitude where CAOs basically occur. Here, we investigate the impact by using a quantitative CAO index, which is defined by integrating cold air mass flux below a threshold potential temperature over midlatitude East Asia (45°N, 90°-135°E). From a climatological analysis, tropical convections are significantly observed two-to-four days following CAO event (i.e., the peak of CAO index), indicating CAO impacts. However, case studies show that the impacts vary among independent CAO cases due to influences of associated synoptic conditions, which affect the pathways of northerlies propagation and consequently hinder or amplify the impacts. Several impact patterns and their possible causes are discussed. A better understanding of East Asian CAO variability can improve the predictability of weather over the Maritime Continent.


Introduction
Cold air outbreak (CAO) is a wintertime phenomenon characterized by the outflow of cold air mass from high latitudes to lower latitudes.CAOs, often referred to as cold surges, basically occur at a synoptic time scale and greatly influence extratropical weather [1,2,3].In East Asia, the CAOs are typically strong-its associated northerly flow sometimes penetrates far to tropical regions and plays a role on the weather over the Maritime Continent [1,4].Previous studies investigated CAOs by using various CAO definitions [4,5,6,7].However, the studies may be inaccurate in defining CAO magnitude because the mass of cold air was not quantitatively measured [3].Some proxies, such as surface temperature or wind, are used.Moreover, studies focused on the tropical impact of CAOs tend to rely on surge indices over subtropical/tropical regions [8], not in the midlatitudes where CAOs occur.
Here, we aim to revisit the impact of East Asian CAOs by using a quantitative CAO definition.This paper follows a recent isentropic method as documented in [3,9,10], which used CAO index defined by equatorward cold air mass fluxes below a threshold potential temperature of 280 K at 45°N.Section 2 explains the methodology in details.Section 3 discusses the mean impact of CAO from lagged 2. Data and method 2.1.Data We use the JRA-55 reanalysis as the main atmospheric dataset, which has 6-hr temporal and 1.25° horizontal resolutions [11].To depict tropical convections, we use daily NOAA outgoing longwave radiation (OLR) [12] and GPCP precipitation [13].The OLR and precipitation have 2.5° and 1° horizontal resolutions, respectively.The analysis period is December-February from 1997/98 to 2014/15.

Isentropic cold air mass and cold air outbreak index
A recent study proposed an isentropic method to diagnose polar cold air mass (CAM) below a threshold potential temperature [14].This approach has quickly inspired other studies to develop CAO indices [3,9,10] because the method allows us to estimate CAM fluxes that regulate CAO events.Before defining the CAO index, let us start with the formulation of CAM below a threshold potential temperature (  ) [14].The amount of CAM at each grid point () is defined as where   and (  ) denote pressure at the ground surface and pressure at   , respectively.We can obtain the flux of cold air mass () by integrating horizontal wind ( ) from   to the surface, so that The Tendency of  is controlled by the convergence of  and genesis/loss rate due to diabatic heating (), and  are calculated from 6-hr reanalysis, and  is estimated from residual in equation (3).Consistent with previous studies, we use 280 K as   .This is justified by the fact that the zonal-mean equatorward mass flux below 280 K represents the lower part of extratropical direct circulation.We can say that the mass flux below this threshold clearly denotes "hemispheric" cold air outflow [14,15].The hemispheric outflow is greatly attributed to a cold air stream in East Asia, which elongates from inland Asia to the central North Pacific (see [3,14] for climatologies of ,  and ).
The East Asian cold stream is a manifestation of many CAO events occurring in there [3].The stream exhibits two dominant modes of variability (western and eastern modes), which simply indicate two distinct CAO pathways and suggest two types of CAOs in East Asia: western CAO and eastern CAO [9].While both CAOs cause robust impacts over the mid-latitudes, only western CAO seems to be capable of triggering remote influence over the tropics because the outflow of eastern CAO does not reach low latitudes [10].Thus, here we only discuss the impact caused by western CAO.The magnitude of western CAO (hereafter referred to as CAO or East Asian CAO) can be measured by defining a CAO index at 45°N [3,9,10]: where −  is the equatorward component of CAM flux and  is the longitude.We use daily CAOI.The climate mean of CAOI is removed and then, CAOI magnitude is normalized by standard deviation.In section 3, to depict mean temporal evolution associated with CAO events, we perform day-lagged regressions with CAO index for all days in the 18 winters.The statistical significances are assessed by using the two-sided student's t-test, where the numbers of degree of freedom are reduced following a decorrelation method [16].In section 4, we perform case studies on several unique CAO events aiming to highlight variability among the events and its possible causes.The identification method of CAO events is shown in that section.

Mean synoptic evolution of CAO event and its impact
Figure 1 shows the synoptic evolution of CAO event in East Asia constructed from lagged regressions with CAOI.The CAO is led by an increase of CAM amount north of Tibetan Plateau and by the strengthening of Siberian High from day -4.The anomalous cold air and high pressure become larger and propagate southeastward.The equatorward CAM flux then intensifies, reaching its maximum at day 0. The CAO is also associated with anomalous low pressure over the northwestern Pacific, which provides eastward pressure gradient force and thus facilitates southward geostrophic flow.The cold air spreads up to southern China, Japan, and the western Pacific.The spread over the ocean and low latitudes causes intense surface heat fluxes and leads to massive disappearance of CAM.These synoptic conditions are discussed thoroughly in [3,4,7].
Although the CAM amount and its equatorward flux rapidly disappear following the peak of CAO, the low-level northerly wind anomalies persist and even propagate to low latitudes, potentially affecting tropical weather (Figure 1b).The northerly momentum is guided by the southward expansion of Siberian High.The mean impacts of the enhanced northerlies are shown in Figure 2. Despite the cold-and dryfeatures of CAO, the outflow gradually gains heat and moistures when they enter low latitudes and sea surface.Therefore, we observe anomalous southward moisture fluxes approaching the Maritime Continent (Fig. 2a), strengthening the northeasterly monsoon flow.The enhanced northerlies trigger tropical convections when they interact with islands and mountains over the Maritime Continent.The locations of tropical convections agree well with the locations of moisture flux convergence and precipitation.The statistically significant precipitation signals appear over the Philippines, South China Sea, and north of Borneo at day +2 and day +4.This result is consistent with a previous study showing that regressions using an East Asian pressure index at midlatitude yield significant signals only over the north of equator [4].Prior to day 0, we observe some convective and precipitation anomalies leading the peak of CAO; these indicate signatures of tropical CAO precursor associated with Madden-Julian Oscillation (MJO), whose regression signals would be more significant in intraseasonal temporal bands [10].MJO can control the occurrence of CAO by deepening a low-pressure trough over the northwestern Pacific through poleward upper-level wave train [10].
Figure 3 exhibits southward propagation characteristics of several parameters associated with CAO along East Asia and Southeast Asia.It is captured by tracing anomaly peaks from 90° to 135°E of mean sea level pressure (MSLP), meridional wind at 925 hPa, moisture flux convergence, and precipitation.The high pressure and northerly wind propagate at the approximately same speed.It suggests that their southward movements are guided by pressure-wind imbalance, where the high pressure drives the northerlies and cold advection, and the cold northerlies give feedback to high pressure [3,4,8].Following the path of northerlies, the anomalies of moisture flux convergence develop with a lag of 1-2 days, favoring instability and initiation of precipitation over the tropics.The peak of precipitation appears at day +3 over 10°N, but the convective signals vary regionally over the tropics possibly due to complex topography over the Maritime Continent.

Case studies of three selected CAO events
The precipitation anomalies obtained from the regressions seem limited in area and rather weak.Several kinds of literatures have shown that cold surges over the South China Sea can exert wide and strong impacts on precipitation over the Maritime Continent, including regions south of equator [8,18,19].Our regression uses a base location of CAOI at midlatitude, thus the downstream impact of CAO may be subject to varying independent synoptic conditions over the subtropics and the tropics, which weaken average propagation of northerlies at low latitudes.To investigate possible variations among CAO impacts, we perform case studies of a few independent CAO events.Any CAO events can be collected by identifying local maxima of daily CAOI that satisfy a given threshold [9,10].The daily CAOI is filtered by 6-day low-pass filter before CAO identification to ensure synoptic signals.We select three CAO events by considering different southward propagation characteristics.They occurred on 4 January 2006 (hereafter Case 1), 3 February 2010 (Case 2), and 30 December 2007 (Case 3).Although the events exhibit strong outbreaks, as indicated by CAOI greater than 1.5, their equatorward intrusions significantly differ of each other (Figure 4).Case 1 shows a clear propagation of northerly along East Asia and the South China Sea.Case 2, however, shows a lack of equatorward propagation despite having large CAO amplitude (CAOI=2.26).Case 3 has relatively weaker CAO amplitude but it exhibits very strong equatorward intrusion of northerlies-even crossing the equatorial line.This nonlinearity leads to weak regressions shown in Figure 2. We examine probable causes for these differing cases in the following sub-sections.

Case 1: Clear southward propagation
The magnitude of CAO occurred on 4 January 2006 is typically strong (CAOI>2).Two days before the outbreak, there is large CAM accumulation from 45°-60°N in the north of the Tibetan Plateau (Figure 5a).During the outbreak, the equatorward flux of CAM prevails in East Asia, transporting abundant cold air that spans from inland China to Japan.A portion of cold air travels further southward and evolves into warmer air; and the rest moves eastward toward the North Pacific following the propagation of midlatitude waves.The southward movement of high pressure is also evident (Figure 5b).It propagates along the eastern boundary of Tibetan Plateau and controls the circulation in the lower troposphere.In the northwestern Pacific, there is a strong cyclonic center, which strengthens the equatorward outflow at midlatitude.In the tropics, the changes in wind direction and magnitude are robust (Figure 5c).Strong northerlies over the South China Sea and vicinity appear at day +2 and day +4 and enhance the northeasterly monsoon flow.The influence of northerly flow significantly affects the western Maritime Continent (Peninsular Malaysia, Singapore, and Sumatra Island).It confirms that the midlatitude CAO can cause substantial impact over a larger area of the Maritime Continent, which is not apparent in the regression analysis.The precipitation signals over there are quite strong and seem to be associated with a meso-tosynoptic scale disturbance.It shows a common pattern of South China Sea surges where the precipitation is induced by the interactions of incoming flow and terrains, which have been reported in previous studies [17,18].In the Philippine Sea, the precipitation band is well-aligned with a frontal line of northerly cold flow, suggesting front-like precipitation due to confluence of cold air with easterly Pacific warm air.
The associated northerly flow is well produced but it does not significantly cross the equatorial line.In this case, the anomalous high pressure only reaches subtropical regions and spreads to the east, consequently favor to development of easterly rather than northerly around the equator.We argue that the additional pressure gradient over the tropics is necessary to maintain further intrusion of the northerlies (see a discussion of Case 3).

Case 2: Nonexistent subtropical surge
A CAO case occurred on 3 February 2010 shows a strong magnitude of CAOI, which is larger than Case 1, but this case has a unique synoptic pattern in the extratropics.Prior to the peak of CAO, the CAM accumulation is centered over the eastern Siberia (Figure 6a).The outbreaks occur mainly over the East Asian coast, transporting CAM to Korea, Japan, and North Pacific.The CAM anomalies over China and lower latitudes are very small and temporary.Comparing with the previous case, this case clearly shows a weak southward propagation of cold air in East Asia (see also Figure 4).The outflow rapidly turns eastward toward the North Pacific and develops into westerly surge.
Figure 6b indicates weak southward extension of Siberian High.The MSLP pattern suggests that the driving force of this CAO comes from interactions between quasi-stationary high-pressure anomalies in high latitudes and cyclonic circulation north of Japan rather than traveling waves trapped over the Tibetan Plateau.One more important pattern is located over low latitudes.A large anticyclone appears in the north of the Philippines.This anticyclone seems to be associated with low-level vorticity response [19] due to suppression of tropical convection over the eastern Maritime Continent (supported by high pressure in Fig. 6b and positive OLR anomalies in Fig. 8b).The lack of Siberian High extension and the existence of subtropical anticyclone greatly hinder southward migration of cold air outflow along East Asian coast and the South China Sea; and therefore, the CAO does not excite precipitation over the Maritime Continent.Nevertheless, at day +4, we observe stronger northerlies propagating over the western Pacific around 140°E, enhancing easterly surge over the Philippines Sea.The flow resembles a Philippines Sea surge-type as reported by [4].But, the existence of high pressure over the western Pacific is not favorable for convection development around the Philippines.Case 2 is rather unusual but it highlights the fact that not all CAO events cause disturbances over the tropics.Examination on other CAO events reveals that there are few other cases resembling Case 2.

Case 3: Strong subtropical surge and cross equatorial surge
Contrary to Case 2, Case 3 shows strong southward propagation of northerlies and even affects crossequatorial flow (Figures 4c and 7).The patterns of CAM and high pressure evolve in a similar manner to those in Case 1, which shows the southward intrusion of high and northerlies along east of Tibetan Plateau.A unique feature occurred in this case is large scale negative pressure anomalies situated over the Maritime Continent, which elongates meridionally up to midlatitude East Asian coast (Figure 7b).The strong pressure contrasts along East Asia and Southeast Asia are believed to guide the southward momentum up to Southern Hemisphere.In the tropics, we observe the increase of northerlies from the South China Sea to the Java Sea, which leads to convections in both hemispheres (Figure 7c).
The accompanying low-pressure anomalies over the Maritime Continent appear to be induced by an active phase of MJO, which is indicated by eastward propagating intraseasonal OLR anomalies over the tropics that are coincided with the CAO (Figure 8c).The existences of precipitation signals prior to day 0 and strong westerlies in the south of equator support the MJO existence (Figure 7c).Furthermore, the MJO may affect the propagation speed of the northerlies.As shown in Figure 4, the propagation speed of northerlies in Case 3 is larger than that in Case 1.
Case 3 indicates that the midlatitude CAO does have an impact over the southern Maritime Continent, but with an influence of another forcing situated over the tropics.The role of active MJO is quite clear.It provides strong southward and eastward pressure gradients and maintains the northerly momentum to propagate further southward, which eventually enhances the precipitation.This finding is consistent with a previous study that shows the MJO influence on modulating precipitation associated with crossequatorial northerly surge (CENS) [18].Note that, tropical OLR anomalies in Case 1 does not show active MJO signal during its CAO event; and in Case 2, the tropical OLR anomalies are significant but they show inactive convection over the western Pacific (Figures 8a and 8b).

Figure 1 .
Figure 1.Synoptic evolution of an East Asian cold air outbreak (CAO) constructed by using lagged regressions with the isentropic CAO index (day -4 to day +2).(a), anomalies of 280-K cold air mass (CAM) amount (; shaded), CAM flux (; vector), and diabatic CAM loss rate (; red contour).For loss rate, the contours enclose areas of significant CAM losses regardless the amplitude.(b), anomalies of mean sea level pressure (MSLP) (shaded) and 925-hPa wind (vector).Only anomalies exceeding statistical significance (99.99%) are shown.Brown contours denote topography with 1000 m interval.

Figure 4 .
Figure 4. Propagation characteristics of three selected CAO cases.Top and bottom panels show CAM below 280 K and meridional wind at 925 hPa, respectively (tracing anomaly peaks from 100° to 135°E).The magnitude of CAOI is shown in the top.

Figure 5 .
Figure 5. Synoptic conditions during a CAO case on 4 January 2006 (Case 1).Columns denote time sequence from 2 days prior to and 4 days after CAO.Top to bottom: (a) CAM and its flux below 280 K; (b) MSLP and 850-hPa streamlines (black areas: surface pressure < 850 hPa); and (c) 925-hPa wind and rain rate.Note that, parameters in (a) and MSLP in (b) are anomalies, while those in (c) and streamlines in (b) are obtained from actual daily values.

Figure 6 .
Figure 6.As in Figure 5, but during a CAO case on 3 February 2010 (Case 2).

Figure 7 .
Figure 7.As in Figure 5, but during a CAO case on 30 December 2007 (Case 3).

Figure 8 .
Figure 8. Hovmöller diagrams showing evolution of tropical OLR (averaged at 15°S-10°N) associated with the three CAO cases (a-c) from 30 day prior to and 30 day after CAOs.Shadings are daily OLR anomalies constructed by removing the first three annual harmonic.Contours denote 30-80 day bandpass filtered anomalies.The day of CAO event is shown by thick solid line.