Trend turning of North China summer extreme precipitations around early 2000s and its possible reason

This study focuses on regional extreme precipitation (REP) in North China. We found a trend turning in summer (July to August) REP frequencies and intensities from a decrease trend in 1961–2002 to an increase trend in 2003–2020, accompanied by a southward shift of the extreme rain belt, and an enhanced connection with the Ural blocking (UB) and the Western Pacific Subtropical High Pressure (WPSH) in 2003–2020. Rains in North China are accompanied by a west–east low–high dipole at upper troposphere. During the REP, the high of the low–high dipole at Northeast China (the NEH) is strongly amplified from a pre-existing stationary ridge over northeast Asia under the influence of eastward propagating Rossby wave energies along the subpolar/subtropical wave guide over Eurasia. For the REP years, an enhanced stationary ridge over the Ural Mountains in the period 2003–2020 replaces the stationary Ural trough in the period 1961–2002, favouring the development of the UB and leading to a change of the Rossby wave propagation path from along the subtropical waveguide in 1961–2002 to along the polar wave guide in 2003–2020. Therefore, a connection between the NEH and the UB forms, which may lead to a higher probability of extreme precipitation in North China since blocking is a major source of strong circulation anomalies. The mean summer WPSH expands more westward-northward in 2003–2020 than in 1961–2002, which provides a background conditions for a further westward-northward expanding of the daily WPSH under the influence of the NEH leading to a strong moisture transport from north Pacific. As a result, the intensity and probability of extreme precipitation over southern North China increase.


Introduction
North China is at the north edge of the East Asian summer monsoon (EASM).Summer rains in North China occur at the third stage of seasonal march of the East Asian summer monsoon.For rains in north China, the EASM brings major water vapor.Annual precipitations, especially extreme precipitations in North China, are mainly concentrated in the late summer from July to August (Ding and Chan 2005;Wang et al. 2015;Zhou et al. 2020).As in the northern rain belt of EASM, precipitations in North China is also influenced by the mid-latitude circulations and the western North Pacific subtropical high (WPSH) (Zhang 1999;Li and Zeng 2003;Ding and Chan 2005;Huang et al. 2006;Liu and Ding 2011).Atmospheric blocking events significantly affect the distribution of rainfall over all Eurasian regions in summer (Antokhina et al. 2016) and have an impact on the intraseasonal variability of summer monsoon rainfall in China (Ding and Chan 2005).Precipitations over North China was found to be also connected to two dominant wave trains over summer Eurasian continent, the wave train along the subtropical Asian jet and the wave train pattern along the polar front (Enomoto 2004;Ding andWang 2005, 2007;Kosaka et al. 2009;Zhu et al. 2011;Schubert et al. 2014).They also have an impact on the intraseasonal variability of precipitations over north China (Gao et al. 2022).Moreover, the subtropical wave train is a bridge between tropical intraseasonal oscillation and North China rainfall (Gao et al. 2022), and the subpolar one is a bridge between Arctic sea ice variability and East China rainfall (He et al. 2018).
Though summer precipitation in North China is much less than that in south China and east China, extreme precipitation can lead to destructive impacts on society (Zhu and Xue 2016).Under a background of southern-flooding-northerndrought pattern over China in the last decades (Wang 2001;Yu et al. 2004), less attentions are paid to extreme precipitations in North China.The weakening of the EASM circulation since the end of the 1970s (Yu et al. 2004;Yu and Zhou 2007;Zhou et al. 2009) leads to a drying tendency in North China (Zhou et al. 2009;Xu et al. 2015;He et al. 2021).However, Zhou et al. (2013) pointed out that the 2012 flood in North China occurred in the context of a multidecadal drying tendency.Li et al. (2013) found that the summer extreme precipitation in North China showed an increasing trend in the early twenty-first century.Zhao et al. (2019) found that frequency of extreme precipitation increased from 1997 to 2016 over the North China Plain.In recent years, more attentions have been paid to the topic.Extreme precipitation in North China was found to be related to the westward extension and northward lift of the WPSH (Zhao et al. 2019;Han et al. 2021a).Due to its intraseasonal variation, the two Eurasian wavetrains also play an important role in causing extreme precipitation in North China (Orsolini et al. 2015;Han et al. 2021b).Yang et al. (2021) further pointed out that heavy precipitations over North China are connected to anomalous anticyclone over the Korean Peninsula conveying more moisture from the North Pacific.
On the other hand, the intensity of the WPSH increases significantly in the twenty-first century (Gong and He 2002;Zhou et al. 2009;Matsumura et al. 2015;Choi and Kim 2019;Fu and Guo 2020).The strengthening and westward expansion of the WPSH leads to significantly positive correlation between summer rainfall in North China and the WPSH after 1990 (Gao et al. 2014).Due to Arctic warming, sea ice has decreased rapidly (Kwok et al. 2009;Maslanik et al. 2011) since the beginning of twenty-first century, which was found to have an important impact on rainfall in east Asia through its impacts on mid-latitude circulations (Nakamura and Sato 2022;He et al. 2018).Liu et al. (2022) pointed out that positive summer shortwave cloud radiative anomalies over northern Russia favor the generation of the Ural blocking, and trigger the positive Eurasian pattern, resulting in an increase of precipitation in northern China.The previous studies have discussed the factors that influence summer extreme precipitation in North China, and pointed out the possible impacts from WPSH and middle-to-high latitude circulations under global warming.However, we still have many problems need to be answered.How do North China extreme precipitations change in the twenty-first century, how the changes of extreme precipitations are connected with the changes of polar and subtropical circulation backgrounds?Why extreme precipitations occur on the weakening tendency of summer monsoon?Whether the WPSH and the middle-to-high latitude circulations have a combined influence on extreme precipitation?The above questions need to be further explored.
This paper focus on the regional extreme precipitation in North China, and discuss the trend turning of extreme precipitations in July-August in North China.We explore how the changes of summer circulation over Eurasian continent and the WPSH influence extreme precipitation in North China in the early twenty-first century.Our study may shed a new light on the cause of extreme precipitation change in North China in the early twenty-first century.The data and analysis methods are presented in Sect. 2. We analyze changes in extreme precipitation after 2003 in the Sect.3. Section 4 discusses possible reasons for change of the REP.Section 5 gives the conclusion and discussion.

Data
The data used in this study are the daily and monthly mean data in July to August from 1961 to 2020, they include: (1) the station-based observations of daily precipitation over the 107 stations in North China region (the 34.5°-44° N, 110°-123° E) from China's surface climate data set (V3.0) provided by the National Meteorological Information Center of China Meteorological Administration, (2) the daily and monthly sea level pressure, geopotential height, wind speed (u, v) and specific humidity at vertical 17 pressure levels from the National Centers for Environmental Prediction-National Center for Atmospheric Research (NCEP-NCAR) reanalysis data (Kalnay et al. 1996), (3) the monthly West Pacific Subtropical High Ridge Index and West Ridge Point Index from the National Meteorological Information Center of China Meteorological Administration.The monthly and daily climatologies are the averaging over the period 1961-2020.Trend turning of North China summer extreme precipitations around early 2000s and its possible… 1 3

Definition of regional extreme precipitation (REP) and regional normal precipitation (RNP) events
Percentile-based extreme precipitation index (Alexander et al. 2006) is defined at each station.For any station in North China, the wet day (precipitation > 0) precipitations from July to August from 1961 to 2020 are sorted from small to large, and the 95th percentile value is selected as the threshold.When the daily precipitation of a station exceeds the threshold, the day is defined as the extreme precipitation day for the station.The empirical orthogonal function (EOF) was performed on the normalized extreme precipitation days from July to August at all stations, and the variance contribution of the first mode (EOF1) is 13.78% (Fig. 1a).The EOF1 shows that the changes in extreme precipitation days in North China are regionally consistent, and the high-loading areas are distributed in the Beijing-Tianjin-Hebei and middle and lower reaches of the Yellow River.We selected a total of 68 stations at which the EOF1 is greater than 0.3 (big circles in Fig. 1a) as the representative stations of extreme precipitation in North China.A day when (1) at least 30% of the 68 representative stations have daily precipitation not less than 1 mm, and (2) at least 10% of the representative stations with daily precipitation exceeding the extreme precipitation threshold, is defined as a regional extreme precipitation (REP) day.Precipitation day that satisfies condition (1) but has no extreme precipitation is defined as a regional normal precipitation (RNP) day.A consecutive precipitation day period with at least 1 REP day or a single REP day is defined as a REP event.A consecutive precipitation day period with only RNP days or a single RNP day is defined as a RNP event.The interval between two REP/RNP events must be 7 days or more, otherwise they are considered to be the same process.The first day of a REP/RNP event is defined as the Lag 0 day of a REP/RNP event.

Definition of the WPSHI index
The westward extending and northward lifting index of the WPSH (WPSHI) is defined based on the ridge line index and the western boundary index constructed by Liu et al. (2012), as follows.

Other methods used
We use running slope difference (RSD) t test (Zuo et al. 2019) with an 11-year window to detect trend turnings of the interannual variabilities of the REP frequency in the period 1961-2020.Rossby wave envelope propagation is analyzed using the wave activity flux defined by Takaya and Nakamura (2001).Background condition for the Rossby wave propagation is diagnosed by meridional gradient of potential vorticity ( PV y = − U yy + F r U z − F r U zz ) (Luo et al. 2019a;Luo and Zhang 2020).The baroclinicity is measured by the Eady Growth Rate (EGR) (Lindzen and Farrell 1980), which is calculated with the NCL function, eady_ growth_rate (https:// www.ncl.ucar.edu/ Docum ent/ Funct ions/ Contr ibuted/ eady_ growth_ rate.shtml).Characterization of elements such as circulation associated with REP events using composite methods.
According to the definition of Tibaldi and Molteni (1990), the daily blocking high index is calculated using the following scheme: where According to the definition in Luo et al. (2016) and Zhang et al. (2018), we defined in this study an Ural High (the UH, hereafter) day when there is at least one longitude satisfying (1) and (2) within the range of 40° E-80° E.Moreover, a consecutive process with UH day no less than 5 days is defined as an Ural blocking (UB, hereafter).The numbers of the UH days and the UB days are then counted respectively in the July-August in each year from 1961 to 2020.

Trend turnings
According to the definition in Sect.2, the number of the REP days (REP frequency here after) and the intensity of REP (the mean daily precipitation averaging over all the REP days in July-August) in July-August is counted in each year from 1961 to 2020.The interannual variations of the normalized REP frequency and intensity time series are shown in Fig. 1b, c. Figure 1b shows that the interannual variabilities of the normalized REP frequency and the principle component corresponding to the EOF1 (PC1) of station-based extreme precipitation frequencies are nearly overlapped.This indicates that the REP event reflects a regional consistent extreme precipitation in North China represented by the pattern of the EOF1.The interannual correlation coefficient between the normalized REP frequencies and intensities is 0.97 (Fig. 1c), which indicates that the year with high REP frequency regularly corresponds to the strong REP intensity.We use the REP frequency time series to show the interannual variability of the REP in the following analyses, which can be an indicator of the inter-annual variability of both the frequency and the intensity of the REP.
The RSD t test (Zuo et al. 2019) detected three trend turnings in the REP frequency time series of the period 1961-2020. They are 1986, 1995, and 2003 respectively (Fig. 1b).Therefore, the trends extracted by the RSD t test are short-term trends in the entire time series.Though the extracted trends appear to swing up and down, they are the linear part of the original oscillations of the time series.The REP frequencies in 1961-1984 show a decreasing trend, and the REP frequencies in 1985-1996 and in 1997-2002 show an increasing and then a decreasing trend afterward, so the overall trend of REP frequency in 1961-2002 is decreasing.In contrast, the REP frequency shows a significant increasing trend from 2003 to 2020.In this study, we aim to investigate the increase trend of the REP in 2003-2020, thus the period 1961-2002 of overall decrease trend is treated as a whole.So we divide the period 1961-2020 into two parts, one is 1961-2002 and the other is 2003-2020.We show the difference between the circulations corresponding to the REP in the two periods by comparing the composite anomalies based on the REP events in each of the two periods.The differences between the background field patterns of the REP events are reflected by the regressions to the detrended time series of the yearly REP frequencies in the two periods.The significance of the difference of two means is tested by a t-test.Though, lengths of the two periods are unequal, according to the rules of the t-test the unequal sample size is allowed (Livingston 2004).Moreover, the running slope difference (RSD) t test can be used for both equal and unequal test samples (Zuo et al. 2019).
In 1961-2002, the REP frequency has a weak negative correlation with the RNP frequency (Fig. 1d, black curve).However, the negative interannual correlation between the REP frequency and RNP frequency increases significantly from − 0.

Typical circulation conditions corresponding to the REP and their changes after the trend turning
According to the REP events defined above, we obtained 78 and 31 REP events in 1961-2002 and 2003-2020, respectively.In order to reveal the circulation conditions at high and low levels of atmosphere and the corresponding water vapor transports, we composite 850 hPa water vapor transport vectors (qu, qv) and their divergence/convergence and the 200 hPa geopotential height anomalies at the Lag 0 day based on the REP events.The results are shown in Fig. 2a-f.As a comparison, we also composite the circulation and water vapor transport (Fig. 3a-f) based on RNP cases.
For both RNP and REP, the main moisture sources are at the Bay of Bengal and the South China Sea (Figs. 2a,d,3a,d).Compared with RNP, the REP moisture transport and convergence are significantly enhanced.In 1961-2002, when the REP occurs, the location of the WPSH shows no significant anomaly and is basically consistent with the climatic mean.The airflows on the west side of the WPSH contribute little to the water vapor transport of the REP.This means that REP is not closely related to the day-to-day variation of WPSH.However, in 2003-2020, the WPSH is significantly strengthened at Lag 0 day, and the 5880 gpm contour extends westward to the lower reaches of the Yangtze River (Fig. 2a,  d), which leads to a stronger moisture transport from south China sea and an extra moisture transport from the northern west Pacific to North China (Fig. 2a, d).Although the composite circulations corresponding to the RNP also show that the subtropical high in 2003-2020 expands more northward and westward than that in 1961-2002, its western boundary is located along the east China coast.Thus the WPSH circulation during the RNP has very small impact on the moisture transport during the RNP (Figs. 2d,3a).So, we have a conclude that in 2003-2020, the expanding of the WPSH to the lower reaches of the Yangtze River makes it to be a significant impact on REP.
According to Figs. 2 and 3, the patterns of circulation and moisture transport corresponding to the REP are obviously different from those corresponding to RNP for both the period 1961-2002and 2003-2020. Comparing Figs. 2b, e and 3b, e the composite anomalies of the 850 hPa moisture transport vector show that the difference between REP and RNP lies in the strong cyclonic circulation over North China on the Lag 0 day and the strengthened anti-cyclonic circulation over the region from northeast China to the Japan Sea.The southwesterly flow between the cyclone and the anticyclone is significantly enhanced, thereby enhancing the transport and convergence of water vapor from the Bay of Bengal and the South China Sea to North China.However, the 850 hPa moisture transport anomaly field on the RNP only shows that there is a weak cyclonic circulation over North China.At 200 hPa, both REP and RNP are accompanied by a structure of low pressure on the west and high pressure on the east of North China.The 200 hPa low pressure is located mainly at Mongolia (ML, hereafter), and the high pressure is located over northeast China and the Korean Peninsula (NEH, hereafter).Therefore, the low pressure and high pressure (the low-high dipole) are located on the northwest side of the 850 hPa cyclone and anticyclone, respectively.This constitutes a typical baroclinic structure.So, air rising exists at the east side of the 200 hPa low pressure center (Figs.2c, e and 3c, e).However, the NEH corresponding on the Lag 0 day of the REP events is increased significantly, and the divergent west of the NEH is strong.Thus, the vertical velocity reached more than 5 times that corresponding to the RNP (Figs. 2c, f, and 3c, f).Therefore, the strong upward movement and abundant water vapor caused the extreme precipitation in North China.
Moreover, compared with that in 1961-2002, there are significant changes in the circulations corresponding to the REP in 2003-2020.The composite results show that the low-high dipole associated with REP are located in the subtropical jet near 40° N, the subtropical wave train is significant, whereas, the geopotential height anomalies over the  2c).However, the 200 hPa composite circulations corresponding to the REP in 2003-2020 show that the geopotential height anomalies over the subpolar region are strengthened significantly.Specifically, a UH exists over the Ural Mountains region, the ML is located more northward, at a location away from the subtropical jet area, and the NEH is located more southwestward (Fig. 2f).Thus, in 2003-2020, the 200 hPa low-high dipole and the corresponding 850 hPa cycloneanticyclone pair show a clockwise tilting of the axis of the low-high dipole when comparing to that in 1961-2002 (Fig. 2b, c, e, f).With the cooperation of the westward expanding WPSH, the 850 hPa anticyclone circulation expands southwestward (Fig. 2a, b, d, e).As a result, the low-level moisture convergence belt in 2003-2020 is located more southward and extends more westward than it is in 1961-2002 (Fig. 2a, b, d, e).In other words, in 1961-2002, the moisture convergence belt is mainly located in the northern part of North China, whereas in 2003-2020 the moisture convergence belt lies in the southern part of North China, that is, the lower reaches of the Yellow River and the Beijing-Tianjing-Hebei region.
For the RNP, the corresponding composite 200 hPa geopotential height anomaly field at Lag 0 is dominated by a more typical Silk Road Pattern (Kosaka et al. 2013) in the subtropical jet region in 1961-2002 (Fig. 3e).Whereas, the Fig. 3 As in Fig. 2, but for the RNP events composite 200 hPa circulations at Lag 0 of the RNP events in 2003-2020 showed a significant enhancement of the subpolar wave train and a weakening of the subtropical wave train, a change in the two Eurasian zonal wave trains that is similar to the case related to the REP events in 2003-2020 (Fig. 3f).Although the RNP events in 1961-2002 are accompanied by a similar low-high dipole as that related to the REP events, the eastern high pressure is much weaker, and the location of the low-high dipole is more eastward.In 2003-2020, the composite 200 hPa circulation at Lag 0 of the RNP events is dominated by a large-scale Mongolian low, and a much weaker high pressure located east of the Korean Peninsula.The anomalies of the RNP 850 hPa moisture transport vector field only show a strengthened northerly flow on the west side of the cyclone.Therefore, the southerly flow and moisture transport during RNP is much weaker than that of REP.Moreover, the vertical velocity related to the RNP is much weaker than that in the case of the REP (Fig. 3c, f).In summary, the significant enhancement of the upper level NEH is a key mid-latitude local circulation feature that distinguishes REP from RNP.

Strengthening connection with Ural blocking
after the trend turning

Low-high dipole amplified by upstream Rossby wave activity
The above composite analysis indicates that one of the most significant features of the circulation corresponding to the REP is the amplification of the NEH in the upper troposphere.In order to explore the cause of the NEH development, the composite daily evolution of the circulations one week before the Lag 0 day of the REP events is shown in Fig. 4. For both in 1961For both in -2002For both in and in 2003For both in -2020, the composite daily evolution of the circulation corresponding to the REP shows the following common features.One week before the occurrence of REP, a mid-latitude high pressure anomaly exists near 120° E as a precursor of the NEH.From Lag − 7 to Lag 0, the Rossby wave activity fluxes distributed zonally along a path across the Eurasian continent (Figs. 4 and 5) indicates an eastward Rossby wave energy propagation and leads to a continuous growth of the ML and the NEH.At Lag 0 the ML and the NEH reach the strongest (Figs. 4 and 5) and compose an amplified upper level low-high dipole.In addition, during the development of the low-high dipole from Lag − 7 day to Lag 0 day, the positions of the centers of the zonal wave train are geographically stable, showing a quasi-stationary wave pattern.As a comparison, during the daily evolution of the circulations corresponding to the RNP, we find a similar eastward propagating Rossby wave energy and a formation of low-high dipole structure at Lag 0. However, the action centers of the wave train move along the westerly jet, which leads to a much weaker and travelling wave pattern, thus results in a weaker low-high dipole (Figs. 6 and 7).The above analysis shows that the occurrence of REP is accompanied by the preexisting high pressure anomaly near 120° E and a quasistationary wave train over Eurasian continent.The eastward propagation of Rossby wave energy along the wave train leads to a persistent growing of the stable low-high dipole and finally forms a strongly amplified one.
Comparing the two periods before and after 2003, we further found that in 1961-2002 the Rossby wave activity flux that accompanies the development of the low-high dipole of the REP mainly distributed along the subtropical wave train.The amplified low-high dipole at Lag 0 shows a zonal distribution along the subtropical jet.Though the evolution of the NEH is accompanied by the UH persists at northern Ural, there is no significant wave energy connection between this subpolar region high and the NEH (Fig. 4).However, in 2003-2020, the UH exists at central and southern Ural before the occurrence of the REP.From Lag − 7 to Lag 0, a large part of Rossby wave activity flux propagates northeastward and leads to a continuous strengthening of the UH.With the gradual strengthening of the UH (from Lag − 6 day), the wave energy propagating along the subtropical weakens significantly.Due to the persistent growth of the UH, there is a stronger Rossby wave energy propagating southeastward from the UH, and the ML strengthens significantly.As a result, the low-high dipole related to the REP becomes more meridional in 2003-2020 than that in 1961-2002.In summary, the enhancement of low-high dipole before the occurrence of the REP is a result of the eastward propagation of upstream Rossby wave energy along the subtropical/polar waveguide in 1961-2002/2003-2020.But the low-high dipole is more zonal/meridional in 1961-2002/2003-2020.The composite sub-seasonal evolution of the wave trains corresponding to the RNP also shows a significant enhancement of the subpolar wave train in 2003-2020.However, the wave activity flux that leads to the amplification of the low-high dipole corresponding to the RNP is along the subtropical wave guide both in 1961-2002 and in 2003-2020.

Forming of the connection between the UB and the NEH after the trend turning
The composite daily evolutions of the REP events show that the local low-high dipole are connected with the daily evolutions of upper level wave trains over Eurasia Continent.
In order to show the background circulation for the daily evolution of the wave trains, we regress the July-August mean 200 hPa geopotential height anomaly field to the    (Lindzen and Farrell 1980).The PV y is a reflection of the meridional gradient of the potential vorticity both from the meridional shear of background westerly wind (the dynamical part) and the meridional background temperature gradient (the thermodynamic part).Luo and Zhang (2020) indicated that large background westerlies and strong PV y lead to strong eastward propagation of Rossby wave energy and weak background westerlies and PV y favor the occurrence of blockings.These findings in Luo and Zhang (2020) helps to explain the how the background conditions (Fig. 9) influence the intr-seasonal evolutions of the wave trains over the Eurasian continent (Figs. 4 and 5).For the two periods 1961-2002 and 2003-2020, the regressed distribution of background zonal winds and the dynamic conditions are consistent with the distribution of the background stationary waves (Figs. 8 and 9).The regression results (Fig. 9) indicate that the westerly wind speed, the PV y and the EGR are weak along the polar front, but are strong over subtropical jet for the REP years in 1961-2002.The weak westerlies and PV y at polar front favors the occur- rence of high latitude blockings over coastal region of the Kara Sea, it is persistent and shows slow westward retrogression, but very weak downstream propagation of Rossby wave energy (Fig. 4).So, there is very weak connection between the blocking and the NEH.Oppositely, in 2003-2020, the distinguish change in the stationary wave pattern is the significant enhancement of the stationary Ural ridge accompanying the REP years (Fig. 8).As a result, in the REP years the background westerlies and the PV y over the 500 Ural Mountains are weakened, which benefits the occurrence of a lower latitude UB (Luo et al. 2014;Luo and Zhang 2020); while the zonal wind, the EGR and the PV y are enhanced north and downstream of the Ural ridge (Fig. 9), which favors the Rossby wave energy propagations north of the UB and in the path from the UB to the NEH.Moreover, the meridional gradient of the Coriolis frequency (ƒ) is lager at lower latitudes than that at the higher latitudes (Luo et al. 2019b), the downstream wave energy propagation will be Fig.7 As in Fig. 4, but for the RNP events in the period 2003-2020 larger from the UB at lower latitudes in 2003-2020 than from the UB at the higher latitudes in 1961-2002 (Figs. 4 and5).Moreover, in 2003-2020 the REP years are not significantly accompanied by an increase of background westerlies, the PV y and the EGR along the subtropical jet (Fig. 9d-f), indicating a weaker Rossby wave activities wave energy propagation along the subtropical waveguide.Thus, the eastward propagating Rossby wave energy divided into two branches near 45 °E, one along the strengthened subpolar waveguide leading to an amplification of the precursor Ural ridge and the wave energy along the subtropical wave guide shows weakened (Fig. 5).As a result, a connection between the UB and the NEH forms.
The above results indicate that the enhancement of the polar stationary wave and the existence of the stationary southern Ural ridge provide a background condition leading to the change of wave energy propagation path from along the subtropical westerly jet in 1961-2002 to the polar front in 2003-2020 (Fig. 5) and constructer a linkage between the UH and the NEH.To verify the above speculation, we calculated the interannual variation of yearly July-August UH days and the UB days (Fig. 10).The result shows that the UH days almost equal to the UB days, which indicates that the existence of an UH day means an occurrence of a blocking event.The correlation between the UB days and the REP frequency has no statistical significance in 1961-2002 but it is significantly enhanced in 2003-2020 (Fig. 10).This result confirms the above analysis that the linkage between the REP and the UBs appears in 2003-2020 due to the enhancement of the southern Ural stationary ridge and the change of Rossby wave propagation path.

Strengthening connection to the WPSH after the trend turning
To further analyze the connection between the WPSH and the REP, we composite the Lag − 7 day to Lag 0 day evolution of the 5880 gpm geopotential height contour which  Whereas, in 2003-2020, the composite 5880 gpm contour corresponding to the REP continue to move westward and northward especially in the period from Lag − 4 to Lag 0 day.This is different from the REP cases in 1961-2002 and the RNP cases in the both periods.On Lag 0 day, the west edge of the 5880 contour is located at the middle and lower reaches of the Yangtze River (Figs. 2d,11b), enhancing the water vapor transport from the western Pacific and the South China Sea to North China (Fig. 2d, e).Therefore, the 2003-2020 REP events are significantly associated with the daily variations of the WPSH due to its contribution to moisture transport.On the other hand, earlier studies found that the intraseasonal west-east shift of the subtropical high is connected to the propagation of the stationary Rossby waves along the Asian Jet in the upper troposphere (Tao and Wei 2006;Enomoto et al. 2003;Dong and He 2020).Comparing the daily evolutions of the composite 200 hPa circulations (Fig. 5) and the composite 5880 gpm contour corresponding to the 2003-2020 REP events (Fig. 11b) from Lag − 4 day to Lag 0 day, we can find that the daily variation of the WPSH is closely related to the development of the NEH which is a result of the Rossby wave activity propagation along the polar wave train.From Lag − 4 day to Lag 0 day of REP, the 200 hPa NEH extends to the Japan Sea and intensifies gradually (Fig. 11b).However, before the Lag − 4 day, the NEH is located more westward thus it has almost no impact on the WPSH.For the period 2003-2020, the westward expanding July-August mean WPSH corresponding to the REP year locates near the Japan Sea (Fig. 11e) which is much more westward than that in REP years in 1961-2002 and in the RNP years.The WPSH can be influenced by the upper-troposphere NEH, because the subsidence on the southeast side of the NEH (Fig. 2c, d) contributes to the strengthening of the low-level high pressure over west Northern Pacific.As a result, the westward and northward expanded background WPSH in 2003-2020 provides a precondition for a further continuously westward and northward extending of the daily WPSH as the 200 hPa NEH continues to expand eastward and strengthen before the REP.The connection between the NEH and the WPSH is thus formed.
The inter-annual variability of the WPSHI (the westwardnorthward expanding index of the WPSH defined in Sect.2) shows that there is no significant trend in WPSHI in the period 1961-2002. However, in 2003-2020 it shows an increase trend (Fig. 12), indicating that the WPSH extends westward and northward after 2003.This is consistent with the composite results shown in Fig. 11.Though the WPSHI and REP frequency are generally positively correlated in both periods of 1961-2002 and 2003-2020, the correlation between the WPSHI and REP frequency is not significant in 1961-2002.This confirm the result that the WPSH has little impact on the REP in the period 1961-2002 because it is located more eastward far from the east coast of Asian continent.However, in 2003However, in -2020, the correlation between the WPSHI and the REP frequency increases significantly, especially after 2007, the interannual variation of WPSHI is almost consistent with the interannual variation of REP frequency.It confirms the enhancing impact of the WPSH on the REP in 2003-2020 when the WPSH expanding more westward and northward.As a conclusion, the connection between the WPSH and the REP in 2003-2020 arises from the below two conditions.First, the strong westward and northward expanding of the WPSH results in an extra moisture transport from the west Northern Pacific and South China Sea adding to the moisture transport from the Bay of Bengal which strengthens the total moisture supply to the REP in North China.Second, the forming of connection between the daily evolutions of the upper troposphere NEH and the WPSH when the mean WPSH expands to west of 130° E.Moreover, the enhanced connection between the WPSH and the REP and the increase trend of the WPSHI may explain the increase of the strength and frequency of the REP in 2003-2020.

Conclusion
Based on the July-August REP events and RNP events distinguished in the period 1961-2020, we explored the trend turning and the corresponding changes of the REP in North China during July-August and its relationship with UBs and the WPSH.The frequency and intensity of the REP in North China turned from a weak decrease trend in 1961-2002 to a significant increase trend in 2003-2020.The REP frequency and intensity increased significantly in Beijing-Tianjin-Hebei region, the middle and lower reaches of the Yellow River and the western Liaoning province, which means a southward shift of REP after 2003.
By comparing the differences in circulation and water vapor transport between REP and RNP events, and between REP events before and after 2003, we reveal the key circulation conditions that lead to the REP in northern China and explain how they changed and lead to a trend turning of the REP around 2003, which is shown schematically in Fig. 13.The REP in North China is characterized by a strongly amplification of upper troposphere low-high dipole with their center near Mongolia (ML) and Northeast China (NEH) respectively.The amplified upper level low-high dipole leads to a strong enhancement of lower level southerly flow between the low level cyclonic-anti-cyclonic pair, which further transports the moisture brought by summer monsoon circulation in the periods before and after 2003 and the WPSH from Bay of Bengal and western north Pacific to North China after 2003, therefore leads to a strong amplification of moisture convergence and precipitation there.For the REP both in 1961REP both in -2002REP both in and in 2003REP both in -2020, the daily variation and enhancement of NEH stems from the following two conditions.First, in the REP year, there is a background stationary northeast ridge.Second, the eastward propagating of Rossby wave energy along the polar wave guide or the subtropical wave guide reinforces the preexisting northeast ridge, so that it develops into a NEH and leads to REP in North China.
Obvious differences exist between the circulations corresponding to the REP in the periods 1961-2002 and 2003-2020   , the west edge of the WPSH corresponding to the REP years is located east of 130° E, so its contribution to the moisture transport to North China is weak.The summer monsoon circulation act as the major moisture source for the REP.In 2003-2020, the July-August mean WPSH corresponding to the REP year expands westward and northward to west of 130° E and lifts northward to near 35° N, which is located very near to the upper level stationary northeast China ridge.Correspondingly, the daily WPSH continues to extend northward and westward to the middle and lower reaches of the Yangtze River under the influence of the continuous enhancement of the 200 hPa NEH during the developing stage of the REP events (Lag − 7 day to Lag 0 day).Therefore, the southeasterly flow on the western edge of the WPSH brings additional moisture besides those transported by the monsoon southwesterly flow, thereby enhancing moisture transport during the REP.Under the combined influence of the NEH and the WPSH, the moisture convergence belt (rain belt) shifts southward and extends westward.Correspondingly, REP rain belt has a southward shift, and extreme precipitation intensifies in the middle and lower reaches of the Yellow River, Beijing-Tianjin-Hebei and western Liaoning province in 2003-2020.And the interannual correlation between the WPSH and REP is also enhanced in 2003-2020 (Fig. 12).Thus, due to the significant correlation between the REP and the WPSHI in 2003-2020, the increasing trend of WPSHI (Fig. 12, continuously westward and northward expanding of the WPSH) implies stronger extra moisture supply from the north western Pacific, which can partly explain the increase and enhancement of REP in 2003-2020.Moreover, the enhanced linkage of the REP to the UB is another possible reason for the increase of the extreme precipitation in North China, since blockings are a major source of persistent circulation anomaly.

Discussions
The results in this study found that both in 1961-2002 and in 2003-2020, moisture transported by summer monsoon circulation is the major moisture source for extreme precipitation in North China.This is consistent with the previous studies (Ding and Chan 2005;Huang et al. 2006;Liu and Ding 2011).Therefore, the northward advance and southward withdrawal of summer monsoon are consistent with the movements of extreme precipitation (He et al. 2021;Cui et al. 2019).However, this study further indicated that extreme precipitations are also impacted by the mid-latitude NEH or the low-high dipole located over west and east of North China, which is a direct local circulation condition that reinforce moisture transport and air rising at North China, thus leads to extreme rains.Because, the NEH is amplified majorly by the two zonal wave trains over the Eurasian continent (Figs. 4,5).To some extent, extreme precipitations can occur when the summer monsoon is weak.So this may explain the result pointed out by Zhou et al. (2013) that the 2012 flood in North China occurred in the context of a multidecadal drying tendency.As for the connection with the WPSH, this study found that the westward and northward expanding of the WPSH in 2003-2020 brings extra moisture from the western north Pacific thus lead to stronger extreme rains than those in 1961-2002 when the WPSH was located more eastward.This is consistent with the recent findings that the intensifying of the WPSH since 2000s leads to a northward stretching of the front rain belt and a reinforcing of precipitations in North China (Han et al. 2021b;Chen and Zhang 2020).However, this study shows some new findings.The intensification and expanding of the July-August mean WPSH in 2003-2020 provides a precondition for a further westward and northward expanding of the WPSH under the influence of the intra-seasonal developing of the upper level NEH (Figs. 2c,f,5,and 11b).The inland expanding WPSH together with the southward tilted upper level low-high dipole modulates the path of the low level westerly jet and results in a southward shift and westward extending of the extreme rain belt.Moreover, this study further shows that the extreme precipitations in North China changes from a weak connection with the WPSH in 1961-2002 to a significant connection in 2003-2020.This may explain the increase of extreme precipitations in the recent two decades, since the WPSH shows an intensifying trend (Choi and Kim 2019;Fu and Guo 2020).The research of Huang et al. (2020) showed a westward and northward shift of the WPSH after 1999, and they attributed the change of the WPSH to the more frequent occurrence of central Pacific El Niño events after 1999.
This study not only shows what the circulation corresponding to the REP is, it also explains the possible reasons that cause the amplification of the local circulation corresponding to the REP.The some features in circulation patterns of the REP shown in this study are consistent with Zhao et al. (2019).They also showed that the extreme precipitation is accompanied by a strengthened southwesterly or southeasterly low level jet and low pressure anomalies at west of North China.In this study, we further showed that the intra-seasonal amplification of the NEH at east of the ML is a key factor that leads to the REP.And the low-high dipole that leads to the REP is a result of both the preexisting stationary northeast ridge at Asia and the downstream propagation of Rossby wave energy along the polar front or subtropical jet.Thus, our study confirm the results pointed by Orsolini et al. (2015) and Gao et al. (2022) that North China extreme precipitations are connected with midlatitude wave trains.And this study also confirms the result of Yang et al. (2021) that extreme precipitation is accompanied by July-August mean anticyclone over the Korean Peninsula after the 1970s.Beyond the results that consistent with the earlier research indicated above, this study further found that accompanying the trend turning of the REP days from a decrease in 1961-2002 to an increase in 2003-2020, the connection between the UB and the REP in North China becomes significant in 2003-2020.As for the generation of the UB, Liu et al. (2022) attribute it to positive summer shortwave cloud radiative anomalies over northern Russia.Nakamura and Sato (2022) pointed out that East Asian rainfall is being enhanced by high-latitude atmospheric circulations due to the Arctic warming.Wang and Luo (2020) found that the high-latitude UB, which resembles to the condition of 1961-2002, results from the propagation of Rossby wave trains due to the decay of Greenland blocking, mainly related to strong ice melting over Greenland; while the mid-latitude UB, which resembles to the condition of 2003-2020, originates from the decay of North Atlantic blocking mainly related to positive extratropical North Atlantic sea surface temperature (SST) anomalies associated with a positive Atlantic Multidecadal Oscillation.Thus the change of the distribution of the UBs may result from the influence from the variabilities of North Atlantic SST and the Arctic sea ice, which indicates a connection between Arctic change and extreme precipitation in North China.This needs to further study in future.

Fig. 1 a
Fig. 1 a The first EOF (EOF1) of July-August extreme precipitation days at each station in North China from 1961 to 2020, b the corresponding PC of the EOF1 (black curve), the yearly REP days (red curve), and the trends of the yearly REP days (red dashed lines, with detection parameter T = 11 years, confidence level at 90%) and their turnings (break point of the dashed red lines), c yearly REP days (red curve) and yearly precipitation averaging over all the REP day in July-August (black curve), d yearly REP days (red curve) and RNP days(black curve), e, f spatial distributions the average (color shad- 3 in 1961-2002 to − 0.73 in 2003-2020.This indicates that the sharp increase in REP frequency is accompanied by a decrease in RNP frequency in North China after 2003, whereas the REP and RNP are relatively independent before 2003.The trend turning around 2003 not only indicates the change of the linear trend in the two periods, it also indicates changes in other features of the REP.The spatial distributions of the extreme precipitation frequency before and after 2003 are shown in Fig. 1d, e.The larger July-August mean REP frequency in 1961-2002 is located nearly north of 37° N with an overall trend of decrease.The significant decrease trend is along a southwest-northeast direction belt from north of Shanxi province to west Liaoning province, and also over west Shandong province.But the larger July-August mean REP frequency shifts to a more southward region (south of 42° N) with a trend of increase in 2003-2020.The most significant increase exists over the lower reach of the Yellow River, the Beijing-Tianjin-Hebei region and west Liaoning province.

Fig. 2
Fig. 2 Composites of the original values (a, d) and the anomalies (b, e) of moisture transport vector (black arrow) and moisture transport divergence (color shading) at 850 hPa, and 200 hPa geopotential height anomalies (contour) and vertical velocity anomalies (Pa/s, color shading, only values exceeding 90% confidence level are shown) c, f at the Lag 0 day, based on all the REP events in 1961-2002 (a-c) and 2003-2020 (d-f) respectively.The red and green yearly REP frequency time series.The common features of the stationary waves for high REP frequency years (the REP years, hereafter) in both1961-2002 and 2003-2020  are the existence of a strong stationary high geopotential height anomaly over northeast Asia (the northeast ridge, hereafter) (Fig.8a, b).However, the stationary wave patterns upstream of the northeast ridge are much different between those in1961-2002  and in 2003 -2020 .In 1961-2002, the high geopotential height anomaly with its center east of the Caspian Sea and the northeast ridge construct a typical Circumglobal Teleconnection (CGT)(Ding and Wang 2005).And the subtropical Asian westerly jet is located between the low geopotential height anomaly with its center located at the south of the Ural Mountains and the high geopotential height anomaly east of the Caspian Sea, a background pattern that leads to a stronger westerly jet.However, in 2003-2020, the subtropical stationary wave pattern is different from the CGT pattern, which shows a low geopotential height anomaly east of the Caspian Sea accompanying the northeast ridge.The south of the Ural Mountains is dominated by a strong stationary high geopotential height anomaly (the Ural ridge, hereafter) which is opposite to the circulation in 1961-2002.Thus the subtropical jet is weakened as it is located at south of the Ural ridge.Moreover, in 2003-2020 the northeast ridge is stronger and located more southwestward than that in 1961-2002.And the amplitudes of the stationary waves

Fig. 4
Fig. 4 Composites of geopotential height anomalies and wave activity flux vectors at 200 hPa from Lag − 7 day to Lag 0 day, based on all the REP events in period 1961-2002, color shading denote composite geopotential height anomalies that exceed 90% confidence level

Fig. 5
Fig. 5 As in Fig. 4, but for the REP events in the period 2003-2020

Fig. 6
Fig. 6 As in Fig. 4, but for the RNP events in the period 1961-2002

Fig. 8
Fig. 8 Regression of July-August mean 200 hPa geopotential anomalies (contours) to the inter-annual variabilities of the REP days (a, b) and RNP days (c, d) respectively, in the period 1961-2002 (a, c) and the period 2003-2020 (b, d), the color shading denotes the regressed value with statistical significance larger than 90%, the gray shading denotes the July-August mean zonal wind velocity, both the 200 hPa geopotential anomalies and the inter-annual variation of the REP days are detrended

Fig. 9
Fig. 9 Correlations between the REP days and the July-August mean 200 hPa zonal wind (U) velocity (a, d), 200 hPa effective beta (b, e), and 300 hPa Eady growth rate (c, f) respectively in the period 1961- . The first difference lies in the middle-to-high latitudes zonal wave trains.In 1961-2002, the NEH is amplified by the upstream wave energy propagating along the subtropical wave train, since the polar wave guide is very weak.Whereas, in 2003-2020, the existence of the stationary Ural ridge provides background conditions for the formation of lower latitude UBs and the existence of a northwest-southeastward wave train linking the UB and the NEH.Therefore, the NEH is amplified by the wave energy release from the UB propagating along the polar wave train.The correlation between the REP frequency and the UB is then significantly enhanced in 2003-2020.The shift of wave propagation path from along the subtropical wave guide in 1961-2002 to along

Fig. 11
Fig. 11 Composite daily variabilities of the 5880 gpm contours from Lag − 7 to Lag 0 based on the a REP and b RNP events in 1961-2002, and c REP and d RNP events in 2003-2020, and composite July-August mean 5880 gpm contours based on the e REP and f RNP

Fig. 12
Fig. 12 Inter-annual variabilities of normalized REP days (black curve) and WPSHI (red curve) respectively, the red dashed lines denote trends of the WPSHI in 1961-2002 and 2003-2020 respectively, the green number denotes the correlation coefficient exceeds 95% confidence level where a i and b i are the ridge line index and the