1 Introduction

Due to its latitudinal coverage, located between the mid and high latitudes of Northeast Asia and the western North Pacific, Northeast China (NEC) is characterized by frequent cold air invasions in winter and spring, which cause a diverse range of weather phenomena, such as extremely low temperatures, blizzard conditions, and freezing rain (Zhou et al., 2009). The cold air's variability over the region during wintertime is usually affected by a sudden establishment, a rapid southward movement, and strong winds; this climate system is called the East Asian winter monsoon (EAWM; Ding et al., 2014). The EAWM is an important climate system in the Northern Hemisphere during boreal winter (Zhang et al., 2014) and plays a major role in tropics-extratropics interactions. For example, the climate variability over the Eurasian continent and the anomalous tropical heating over the Maritime Continent are closely associated (Chang et al., 1979; Lau and Chang, 1987). As such, the EAWM significantly affects the variability of temperature and precipitation in boreal winter over the Asian and Pacific regions (Lau and Chang, 1987). To investigate the synoptic disturbances of cold air variability in East Asian winter, the best approach is to focus on the EAWM, which is characterized by a succession of cold air outbursts from the Siberian high, always leading to a sharp temperature drop and usually associated disastrous consequences (Chan and Li, 2004; Chang et al., 2005).

The EAWM usually causes strong northerly winds and anomalously low temperature over northeastern East Asia and South China, and shows a close relationship with ENSO (Chang et al., 1979; Boyle and Chen, 1987; Ding, 1994; Zhang et al., 1997; Kim et al., 2014). Previous studies have shown that the EAWM is driven by differential heating between the continent and ocean (Li, 1989; Huang et al., 2003; Wang et al., 2003; He et al., 2008; Zhang et al., 2008; Kim et al., 2014). In addition, Tao and Zhang (1998) found that during ENSO winters and the ensuing spring, the climate in southeastern China is warmer and wetter than normal. Many researchers have reported that in boreal winter, during the mature phase of ENSO, a weaker monsoon often affects the East Asian coast (Tomita and Yasunari, 1996; Zhang et al., 1996; Ji et al, 1997). In particular, significant effects of ENSO on the northern extratropical regions have been identified in observational data. Bueh and Ji (1999) suggested that a weaker (stronger) EAWM relates closely to strong El Niño (La Niña) events. Chen (2002) showed that anomalous southerly winds in the lower troposphere might be triggered by El Niño, while La Niña triggers anomalous northerly winds in its western periphery.

Over the last decade, the effect of global warming on the EAWM has attracted the attention of numerous researchers. Wang et al. (2009) showed that against the background of global warming, the EAWM system experienced a significant weakening around the late 1980s. Lee et al. (2013) revealed that the EAWM significantly weakened after the mid-1980s. Xu et al. (2006) also reported that the surface wind is declining—a finding confirmed not only by station data across China, but also by model experiments (Hori and Ueda, 2006). Although the variability of the EAWM covers multiple timescales (Jin and Sun, 1996; Chan and Li, 2004; Ding et al., 2013), the above phenomena focus increasing attention on the interdecadal variability of the EAWM. It is also important to note that most previous studies have focused on the influence of the EAWM on temperature over NEC, with little attention having been paid to understanding cold air in spring.

Focusing on NEC is necessary, as it is one of the major crop production regions of the country (spring maize, soybean, single-crop rice, and other cash crops), with a total farmland area of 1.82 × 10⁵ km² (Chen et al., 2012). The yield for this region is crucial for policymakers in making strategic decisions to guarantee food security and to stabilize the provisions market. Therefore, it is necessary to investigate crop yield responses to climate change. In particular, changes in the frequency and intensity of climatic extremes are usually associated with the advent of global warming, such as increases in drought, flooding, wind storms, low temperatures, freezing conditions, and hail damage. NEC is no exception in this regard, and thus large fluctuations in agricultural yields are common (Cheng and Zhang, 2005; Zhang et al., 2008; Zhao et al., 2009; Zhao and Sun, 2013). Parry et al. (2004) suggested that against the background of global warming, the effect of climate change on grain yield is negative in some agricultural regions and positive in others.

Abnormal low-temperature events in spring can cause severe damage to spring and autumn rice production in the major production regions, such as the mid-lower reaches of the Yangtze River and NEC. Qian et al. (2015) used a new method to improve the performance of medium-range forecasts of lowtemperature events in the Yangtze River valley. The outbreak of such extreme cold-weather events may be attributable to the transition from a positive North Atlantic Oscillation (NAO) event to a long-lasting blocking event over the eastern Atlantic and western Europe. Zhou and Wang (2015) indicated that spring NAO index is significantly correlated to the year-toyear increment of maize and rice production in NEC. Shen and Masahide (2007) documented that spring Eurasian surface air temperature correlates with the NAO of the previous winter, but has no detectable relationship with the synchronous NAO.

Thus far, however, few studies have focused on the influence of spring cold air variability, especially in NEC. Wang et al. (2007) suggested that cold days and nights decreased in NEC during the springs of 1957-2000, whereas the opposite was true for warm days and warm nights. Wang et al. (2013) found that during the spring of 2013, a long-lasting negative phase of the Arctic Oscillation (AO), active cold vortexes, and more extensive than normal snow cover over NEC, caused the sustained low temperature in this region. Qian et al. (2011) indicated that the onset of spring has been advancing over NEC, most significantly in the eastern part of the region. Zhang and Xu (2011) investigated the frequency of spring extreme low-temperature events in NEC during 1958-2007, and reported that the frequency decreased during the study period, and the key sea surface temperature (SST) area affecting the spring extreme low-temperature events was located in the northern-central Atlantic Ocean, with negative (positive) SST anomalies for years with more (fewer) spring extreme low-temperature events. However, little attention has been paid to daily cold events and associated cold air resources. In this paper, we attempt to address these issues, with special attention paid on the interdecadal characteristics and possible underlying mechanisms.

The remainder of the paper is organized as follows. Section 2 outlines the data and methods used in the study. Section 3 investigates the interdecadal variation, the temporal and spatial features of strong cold events (SCEs) and weak cold events (WCEs) during 1961-2012, the wave flux evolution and the possible reasons for the interdecadal variations of SCWs and WCEs. Section 4 presents the conclusions.

2 Data and methods

The study area spans 38°-55°N, 117°-137°E (boxed area in Figs. 3-14). This domain is selected to cover the NEC region, in which there is a succession of cold air outbursts from the Siberian high, contributing to a major portion of sharp temperature drops and related disasters in the region.

The daily zonal and meridional wind data are from the NCEP-NCAR reanalysis (Kalnay et al., 1996). This dataset has a horizontal resolution of 2.5° latitude × 2.5° longitude, and is available for the period 1961-2012. In addition, the gridded CN05.2 dataset (a 0.5° × 0.5° resolution daily temperature dataset over China) (Xu et al., 2009; Wu and Gao, 2013) is applied in this analysis. The winter [December-February (DJF)] of 1961 (for example) refers to December 1960 to February 1961. The monthly value of daily mean temperature anomaly [Niño-3.4 region: 5°N-5°S, 120°-170°W] index for the years 1950-2012 is obtained from the National Weather Service Climate Prediction Center (http://www.cpc.ncep.noaa.gov/data /indices/). The AO index is from the NOAA (http://www.esrl.noaa.gov/psd/data/correlation/ao.data).

Because the AO index is highly correlated with the NAO index, some studies argue that the AO and NAO are synonymous. That is, they are different names for the same variability, not different patterns of variability (Wallace, 2000). Following Bai et al. (2015), we define a winter as a positive AO (AO+) [negative AO (AO-)] winter when the DJF mean index is above 0.5 (below -0.5); otherwise, it is defined as AO-neutral. To better exhibit the horizontal structure of the CE patterns, the Takaya-Nakamura (T-N) flux (Takaya and Nakamura, 1997, 2001) and geopotential height anomalies are calculated. The level of statistical significance is determined based on the twotailed Student's t-test.

To distinguish the cold air variability in the NEC domain, CEs are defined according to the following two steps (Peng and Bueh, 2011): (1) the value of daily mean temperature anomaly is lower than its 10th percentile threshold, and the total grid number satisfying this condition is more than 50% of the total area of NEC; and (2) a cold event lasting for at least three days is defined as an SCE; otherwise, it is defined as a WCE. As a background and reference for the SCEs and WCEs during 1961-2012, we list their dates, durations, and intensities in Tables 1 and 2.

3 Interdecadal variability of SCEs and WCEs 3.1 Trends of cold air activity and its precursors in the previous winter

When focusing on spring cold air, it is intuitive to expect ENSO and AO-winters to be possible causes. Generally, ENSO events affect not only the oceanic and atmospheric conditions across the entire tropics, but also those across the global extratropics in both hemispheres (Rao and Ren, 2016). Chen et al. (2013) found that when an El Niño event is coupled with an AO-month, the winter climate anomalies in northern China are mainly influenced by the AO. For 48 SCEs (Table 1) and 94 WCEs (Table 2), most coincide with ENSO or AO-cases in their previous winters. Specifically, all 48 SCEs (94 WCEs) occur in 27 (37) yr, together with 21 (24) ENSO and 17 (17) AO-winters. Among the ENSO and AO-winters, the main objective is to capture their co-occurrence. It indicates that a classification of years according to the phases of ENSO and the AO is needed to elucidate the potential factors that can be used as predictors.

Table 3 indicates that, while SCEs occur, the majority of events occur against a background of an AO-winter in each decade, except for the 1990s. When the ENSO SST forcing is confined to AO-winters, the results (i.e., El Niño/AO-, La Niña/AO-, El Niño, and La Niña) are less than expected. The factors for WCEs listed in Table 4 are also below expectation. These results suggest that an AO-winter can be regarded as a predictor for both SCEs and WCEs, as advocated by Huang et al. (2007) and Wang et al. (2007).

According to the above analyses, the predictors are chosen based on a method of cause and effect; another option would be to view the results from the predictors themselves. We choose El Niño, La Niña, AO-, and El Niño (La Niña)/AO-winters as predictors. The classification of years according to the phases of ENSO and the AO for the period 1960-2012 is shown in Table 5. Over this period, 17 El Niño, 18 La Niña, and 26 AO-events occur. Hereinafter, for convenience, El Niño (La Niña) combined with an AO-winter is referred to as an El Niño/AO-(a La Niña/AO-) winter.

Table 5 shows that 10 (12) out of 17 El Niño events are associated with SCEs (WCEs) for the whole period, which is lower than expected. In the pre-1990s epoch, however, 9 SCEs (8 WCEs) occur during 10 El Niño winters, supporting the idea that ENSO is a valuable predictor for NEC climate (Liu and Ding, 1995). Meanwhile, for La Niña and AO-winters, 10 SCEs (11 WCEs) occur during 18 La Niña winters, and 18 SCEs (19 WCEs) occur during 26 AO-winters, in the whole period. In contrast, during the pre-1990s epoch, 7 SCEs (7 WCEs) occur during 10 La Niña winters, and 15 SCEs (11 WCEs) occur during 17 AO-winters. As such, the SCEs and WCEs are closely associated with ENSO events and AO-winters in the pre-1990s epoch; whereas, after this period, the relationship is less robust. When both ENSO and AO-occur in winters, there are 9 La Niña/AO-and 11 El Niño/AO-winters during the whole period; and against this background, 5 SCEs (7 WCEs) occur during 9 La Niña/AO-winters, and 9 SCEs (10 WCEs) occur during 11 El Niño/AO-winters. As far as the pre-1990s epoch is concerned, 4 SCEs (4 WCEs) occur during 5 La Niña/AO-winters, and 8 SCEs (8 WCEs) occur during 8 El Niño/AO-winters, which is consistent with the above results.

To understand the frequency of cold air intrusions over NEC, the days of each type of cold event during each decade are investigated. There is a notable decreasing trend for CE intensity (Fig. 1), which is consistent with previous studies (Wang et al., 2007; Zhang and Xu, 2011). Moreover, during periods Ⅰ (1961-1980), Ⅱ (1981-2000), and Ⅲ (2001-2010), the SCEs have a positive (negative/positive) trend over period Ⅰ (Ⅱ/Ⅲ). Compared to the SCEs, the WCEs demonstrate a near inverse tendency; namely, a negative trend during 1961-1980, a positive trend during 1981-1990, and a negative trend during 1991-2010. The highest frequency of CEs (SCEs/WCEs) appears in the 1970s (1970s/1980s), whereas the lowest lies in the 1990s (1990s/1970s).

Figure 2 indicates the lowest regional temperature anomalies during each cold air burst process for both SCEs and WCEs. We choose this indicator to describe the intensity of cold air. There is an obvious decreasing trend for SCEs, but no trend is detectable for WCEs. In comparison to WCEs, the duration is longer, and there is a lower temperature decrease for SCEs in the whole period, especially 1961-1980. This suggests a more significant influence of global warming on SCEs than WCEs.

3.2 Interdecadal change of 850-hPa wind in spring

To investigate how cold air affects surface temperature anomalies in NEC and the interdecadal variation of its pathway, the wind fields of SCEs and WCEs at 850 hPa are composited for each decade. In the five decades, three common features are apparent (Figs. 3-7). First, a typical East Asian trough stands out in both SCEs and WCEs, whose northwesterly wind component invades the NEC region and causes freezing days[supported by Wang et al. (2005)]; second, during SCEs there is a clear cold vortex centered over Okhotsk; and third, during WCEs there is a clear cold vortex centered in Karafuto.

Additionally, for SCEs, the source of cold air originates from two major currents: the northwestern route flow (NWRF), originating from the westerlies; and the northeastern route flow (NERF), from the Okhotsk region. The location (Figs. 3-7) and frequency exhibit remarkable interdecadal characteristics. More specifically, in the 1960s (Fig. 3a), both NWRF and NERF are located at approximately 60°-65°N, 125°E, and then invade the NEC together. However, in the 1970s (Fig. 4a), the main current of the NWRF shifts southward to approximately 55°-60°N, and the NERF moves to the same latitude. Due to this shift, cold air in the high latitudes is able to intrude into the NEC region more easily and more frequently, and the count of SCEs is therefore greater than in the 1960s, as shown in Fig. 1. In the 1980s (Fig. 5a), the NWRF moves back to 65°-70°N. Additionally, the principal part of the NERF nearly vanishes, and this declining phenomenon recurs in the 1990s (Fig. 6a) for the NWRF.

The number of SCEs in both the 1980s and 1990s is less than that in the 1970s, which supports the results in Fig. 1. Furthermore, the amount of SCEs in the 1980s is greater than in the 1990s (Fig. 1), indicating that the cold air from the NWRF favors cooling in the NEC relative to the NERF during SCEs. Figure 7a shows that the NWRF in the 2000s reappears and shifts northward back to 65°-70°N, and the two flows are confluent at approximately 65°N. As a consequence, the number of SCE days increases. For the NERF, we can divide the whole period into two stages. The first (earlier) stage comprises the first three decades (the 1960s, 1970s, and 1980s), which show a common feature; while the remaining decades constitute the second (later) stage. During the earlier stage, there is a distinct declining trend for the NERF, but in the later stage it becomes gradually reinforced.

For WCEs, the cold vortex over NEC is the factor that contributes the most to cold events. In the 1960s (Fig. 3b), a typical cold vortex appears over NEC. This cold vortex persists due to two cold air currents, i.e., the NWRF and NERF, the former being weaker during SCEs. For the 1970s (Fig. 4b), the winds during WCEs demonstrate similar patterns as their counterparts during SCEs, but are weaker in both flows. For the 1980s (Fig. 5b), the situation is almost the same as that of SCEs, but to a relatively weaker extent. The cold air flows come not only from westerlies but also from the Yenisei River, with the latter of the two flows being stronger during SCEs. A typical cold vortex over NEC exists in the 1990s (Fig. 6b) and 2000s (Fig. 7b). Comparison shows the cold vortex to be stronger in the former period, leading to more cold days, which partially explains the interdecadal variation of WCEs shown in Fig. 1 during the 1990s and 2000s. Taken together, the major factors forming SCEs over NEC are the NWRF and NERF (Fig. 8a), whereas cold vortexes contribute the most to WCEs (Fig. 8b).

To investigate the interdecadal variation of the wind difference between SCEs and WCEs, comparisons of the composite patterns between them are provided in Figs. 3c-8c for each decade. Overall, the NWRF is weaker, and the NERF is stronger during SCEs than WCEs in all decades (Fig. 8c). In the 1960s (Fig. 3c), an eastern flow dominates a belt at approximately 45°-50°N; whereas in the following four decades, this belt shifts to approximately 50°-60°N. In general, there are significant changes in the location of the northerly wind component. Specifically, in the 1960s, the main body of the northerly wind component is approximately 130°E, but in the 1970s and 1980s, it moves eastward to 135°E. Furthermore, in the last two decades, it is located near 140°E.

3.3 Rossby wave energy propagation and dispersion

To further investigate the circulation anomalies maintained by the Rossby wave energy dispersion when SCEs and WCEs occur over and around NEC, a T-N flux (Takaya and Nakamura, 2001) diagnosis is applied to the composited anomaly fields at 250 hPa. The horizontal wave-activity flux is predominantly eastward across a cold vortex type anomaly center, located at approximately 45°N. Significant wave-like anomalies appear across East Asia, which are associated with the propagation of the T-N flux, as indicated by the arrows in Figs. 9-14. To better reflect this persistence, we analyze the T-N flux in each decade. Here, the climatology refers to the long-term average in the spring from 1981 to 2010.

Clear decadal variations of Rossby wave energy dispersion for SCEs are found. In the 1960s, for SCEs, a wave pattern starts from NEC and bifurcates into two branches: one extending northeastward to Okhotsk, and the other turning eastward to Korea and Japan (Fig. 9a). In the divergence field, both NEC and its downstream regions feature positive values, not like a wave train (i.e., positive and negative values appear in turn). Additionally, the divergence center over downstream regions appears stronger than over NEC and, as a result, cold days tend to last longer. Otherwise, the wave energy decays quickly via a downward migration of zonal-mean zonal wind and temperature anomalies. The main difference between the 1970s and 1960s lies in a strong wave packet emanating from the Mongolian Plateau to NEC and its downstream regions during SCEs (Fig. 10a). The wave source in the 1970s is stronger and similar in terms of its propagation and dissipation; therefore, the frequency of cold days is higher. However, in the following three decades (Figs. 11a-13a) of SCEs, the mechanism of cold-day persistence is similar to that in the 1970s, but the wave source for NEC weakens over time.

Compared to SCEs, during WCEs, in all decades (Figs. 9b-13b), the wave trains are clear within 40°-70°N, indicating that wave energy can propagate and dissipate more easily. This is displayed clearly in the difference between SCEs and WCEs (Figs. 9c-13c). Therefore, cold days during WCEs cannot last for a long time. More specifically, in the 1960s, 1970s and 1980s, a divergence center located over NEC splits into two branches of wave trains, and the flux divergence traces positive and negative centers along each branch. In particular, in the 1960s, the branches are not obvious; whereas, in the 1970s, both branches enhance, which favor wave energy propagation and hence cold days in NEC. In contrast, in the 1980s, the strength of the southerly wave train current declines, and the number of cold days increases. In the 1990s, the strength of the flux divergence over NEC decays significantly. Compared to the 1980s, the southerly wave train branch enhances, and the number of cold days declines. In the 2000s, divergence over NEC declines, and its center shifts to the northerly branch. Moreover, simultaneously, this northerly wave train greatly enhances. Divergence, tracing negative and positive centers along this branch, could propagate and dissipate more easily. Therefore, a steady decline in cold days is apparent in the more recent decades, as shown in Fig. 1.

If we focus on the whole period for both SCEs and WCEs, the composite patterns are very similar, except for the location of the divergence center (Fig. 14). Clearly, a divergence center covers NEC, and a zonal belt of positive flux divergence appears in both cases, except that the center values are different (i.e., during SCEs, the strong divergence center is located downstream of NEC, while during WCEs, the counterpart is located over the NEC region).

4 Summary

In this paper, the characteristics of the interdecadal variation in the intensity and zonal position of the cold air events over the NEC region are investigated for the period 1961-2002. To explore the physical mechanisms behind the interdecadal variation of SCEs and WCEs, the associated anomalies in wind and the T-N flux fields are analyzed. The conclusions are as follows:

(1) During the pre-1990s epoch, El Niño, La Niña, AO-, La Niña/AO-and El Niño/AO-events are closely associated with SCEs and WCEs, whereas after this period the relationship is less robust.

(2) In terms of the multidecadal variability of SCEs and WCEs, SCEs show positive, negative, and positive trends over 1961-1980, 1981-2000, and 2001-2010, respectively; while WCEs exhibit a near inverse tendency, i.e., a negative trend during 1961-1980 and positive trends during both 1981-1990 and 1991-2010. Meanwhile, the intensity trend for SCEs is negative, but non-significant for WCEs.

(3) When SCEs or WCEs occur, the most dominant feature of the wind field at 850 hPa is a typical East Asian trough for each decade, whose northwesterly wind component invades the NEC region and causes freezing days. Moreover, a cold vortex centered over Okhotsk during SCEs is a second feature of note. Additionally, a third characteristic is a cold vortex located in Karafuto, whose northwesterly airflow intrudes into NEC during WCEs. Comparatively, the northwesterly airflow is weaker and the northeastern counterpart is stronger during SCEs than WCEs, in all decades.

(4) In the T-N flux and divergence fields, during SCEs, a divergence center exists over NEC; while over its downstream regions, a stronger divergence center appears, but not like a wave train. For WCEs, the divergence center pattern is opposite to that for SCEs. Moreover, a wave train during WCEs is clear, which means that the wave energy can propagate and dissipate more easily. Therefore, cold days during WCEs cannot persist for a long time.

Acknowledgments: The authors would like to thank the reviewers for their comments and suggestions. We also appreciate the editors for their help to improve our paper.