Progress and Prospects of Research on Subseasonal to Seasonal Variability and Prediction of the East Asian Monsoon

Congwen ZHU, Boqi LIU, Lun LI,2, Shuangmei MA, Ning JIANG, and Yuhan YAN

1 State Key Laboratory of Severe Weather, Chinese Academy of Meteorological Sciences, China Meteorological Administration, Beijing 100081

2 Collaborative Innovation Center on Forecast and Evaluation of Meteorological Disasters,Nanjing University of Information Science & Technology, Nanjing 210044

ABSTRACT Subseasonal to seasonal (S2S) variability represents the atmospheric disturbance on the 10–90-day timescale,which is an important bridge linking weather and climate. In 2015, China Meteorological Administration (CMA) listed the S2S prediction project that was initiated by WMO programs three years ago as one of its key tasks. After five years of research, significant progress has been made on the mechanisms of the East Asian monsoon (EAM) S2S variability, related impact of climate change, as well as the predictability on the S2S timescale of numerical models.The S2S variability of the EAM is closely linked to extreme persistent climate events in China and is an important target for seasonal climate prediction. However, under the influence of global warming and the interactions among climate systems, the S2S variability of the EAM is so complex that its prediction remains a great challenge. This paper reviews the past achievement and summarizes the recent progress in research of the EAM S2S variability and prediction, including characteristics of the main S2S modes of the EAM, their impact on the extreme events in China, effects of external and internal forcing on the S2S variability, as well as uncertainties of climate models in predicting the S2S variability, with a focus on the progress achieved by the S2S research team of the Chinese Academy of Meteorological Sciences. The present bottlenecks, future directions, and critical research recommendations are also analyzed and presented.

Key words: East Asian monsoon (EAM), subseasonal to seasonal (S2S) timescale, change mechanism, predictability of climate models

1. Introduction

The annual cycle is the most prominent feature of the East Asian monsoon (EAM). The subseasonal timescale(10–90 days) changes of the EAM, which can be treated as fluctuations superimposed on its annual cycle, reflect the variations in the atmospheric intraseasonal stability.It is regarded as the dominant carrier of extreme weather events in China, such as persistent heavy rainfall, heat waves, and cold surges. The complexity of multi-timescale changes and interactions of the EAM leads to large uncertainties in subseasonal-to-seasonal (S2S) climate predictions in China. The subseasonal prediction is a gap between the weather forecast and climate prediction,which has become the focus of the current operational climate prediction. The accurate S2S forecasting exhibits scientific and technological frontiers and extensive economic value in social economic society like agriculture and transportation. In 2012, the World Meteorological Organization (WMO) jointly proposed the S2S prediction project in conjunction with the World Weather Research Programme (WWRP) and World Climate Research Programme (WCRP). The goal of the project is to bridge medium-term weather forecast (two weeks) and climate prediction (three months) in order to enhance a seamless weather–climate service (Vitart et al., 2017).Phase 1 of S2S prediction project has successfully finished in 2018. In the same year, the WMO launched Phase 2 of the S2S prediction project. The most significant achievements of Phase 1 of the S2S prediction project were summarized as follows (WWRP and WCRP,2018b): (1) establishment of a database of subseasonal(0–60 days) model forecast products and three data nodes[ECMWF, IRI (International Research Institute for Climate and Society), CMA (China Meteorological Administration)]; (2) assessment of the potential predictability of climate events at subseasonal scales, such as the conditions favoring the subseasonal forecasting skills for high-impact weather and climate events (i.e., windows of opportunity; Mariotti et al., 2020); (3) estimation of the forecast skill of S2S dynamic models on the Madden–Julian oscillation (MJO), North Atlantic Oscillation(NAO), and summer monsoon in the Northern Hemisphere (Jie et al., 2017; Vitart, 2017; Feng et al., 2021);and (4) facilitating the development of the Earth system models and weather-climate-unified models (Lang et al.,2020; Meehl et al., 2021).

In 2015, CMA listed the “S2S climate prediction and climate system model” as one of three key tasks. In the same year, the Chinese Academy of Meteorological Sciences (CAMS) has listed “theory and technology of S2S prediction over EAM” as one key research directions and organized a research team. The goal of this team is to investigate the S2S variability and mechanism of the EAM,and to find a new way to improve the leading time of S2S prediction. In the past five years, the team members have revealed the S2S dominant modes of the EAM,demonstrated their critical physical processes, and revealed the influence factors from the perspective of internal interaction of climate system. We also elucidated the bottleneck of the state-of-the-art numerical models in the S2S prediction of the EAM to provide a scientific basis for the new theory and method of the S2S prediction of the EAM (Liu et al., 2020; Yan et al., 2021,2022). This paper aims to review the primary progress of the CAMS S2S research team in the past five years and to point out the currently unsolved problems, future development directions, and future critical research issues from a scientific point of view.

2. Climatological features of the dominant S2S modes of the EAM

The EAM exhibits a significant annual cycle (namely,seasonal cycle) due to the seasonality of solar radiation.The alternation of winter and summer monsoon is the most dominant feature in the annual cycle of the EAM.Because of geographical latitude and seasonal phase-lag of land–sea thermal contrast over the Asian–Pacific region, the EAM shows significant temporal phase differences between the tropics and mid–high latitudes as a response to the annual cycle of solar radiation. Jiang et al.(2020) decomposed the annual cycle of the EAM into vernal equinox and summer solstice modes, which reflect the circulation and rainfall pattern of South China and the Chinese mainland in the different seasons, respectively. In this way, the subseasonal timescale (10–90 days) variability of the EAM acts as an intraseasonal fluctuation superimposed on the annual cycle. Therefore,compared with the MJO, the S2S dominant modes of the EAM and its variation exhibit more pronounced seasonal change (Song et al., 2016; Guo et al., 2019).

The subseasonal modes of the EAM indicate the circulation interaction of EAM components at the subseasonal timescale. The East Asian summer monsoon (EASM)shows a dominant 40–80-day oscillation superimposed on the annual cycle of solar radiation (Song et al., 2016).The circulation regime of the modes features a “gearlike” coupling of the western Pacific subtropical high(WPSH), the South Asian high (SAH), and the Mongolian cyclone (MC) (Song et al., 2016). These modes depend on the non-uniform spatial distribution of atmospheric diabatic heating in East Asia and reflect the uniqueness of the subseasonal variability of the EASM, which cannot be fully explained by the MJO. Driven by the phase difference of atmospheric diabatic heating (short-wave,long-wave, sensible, and latent heat) and the time-varying inhomogeneity of the zonal sea–land thermal gradient, the modes become remarkable in May–August, with the oscillation center in the north of the western North Pacific. The diversity of the “gear-like” coupling of the WPSH, SAH, and MC corresponds to the distinct precipitation distribution in China. When the changes in the three circulations are consistent, the rainfall shows a“sandwich” pattern with the anomalous center in the mid–lower reaches of the Yangtze River. However, when the MC and the WPSH move eastward and northwest, respectively, along with the weakening of the SAH, the rainfall exhibits a dipole pattern centered in northern China (Fig. 1). Observation and numerical simulation verify that the condensation heating released by the convection over western North Pacific and local sea surface temperature (SST) maintains this intraseasonal oscillation of the EASM. The seasonal phase locking of the“gear-like” coupling of the EASM circulation with the subseasonal variation of the summer rain belt in China provides potential predictability of subseasonal variability of the summer rainfall in China.

Fig. 1. Spatial patterns of the (left) first and (right) second multivariable empirical orthogonal function (MV-EOF) modes of the intraseasonal component of daily climatological winds (vectors; m s-1) and precipitation (shading; mm day-1). The western Pacific anticyclone is abbreviated as WPA [replotted based on Song et al. (2016)].

The subseasonal modes of the East Asian winter monsoon (EAWM) can be defined as the subseasonal fluctuations superimposed on the annual cycle. In North China,they show dry–cold (warm–wet) fluctuations but cold–wet (warm–dry) variations in South China with a quasi-30-day oscillation in timescale (Yu et al., 2019). In the negative phase of the Arctic Oscillation (AO), the subseasonal fluctuation of the geopotential height at 500 hPa propagates southward. It leads to the southward intrusion of the cold front and cooling in situ. The local cold cyclone further enhances the southward shift of the Siberian high, which increases the atmospheric instability and causes the cold and wet climate in South China.Therefore, the cold air activity is a precursor for the subseasonal prediction of winter rainfall in South China (Yu et al., 2019). In particular, the winter temperature exhibits short-term cold and warm fluctuations, known as the“temperature whiplash” events in North China (Ma and Zhu, 2021). Our results show that the subseasonal winter temperature fluctuations in North China are closely associated with the southeastward propagation of a dipolepattern temperature anomaly over Eurasia, which comes from the Rossby wave train originating from the Kara Sea–Barents Sea (Ma and Zhu, 2021). Temperature budget diagnosis analyses suggested that the anomalous meridional temperature advection determines the drastic warm-to-cold transition. In contrast, the cold-to-warm process depends partly on the adiabatic heating due to the abnormal descending motion and partly on the anomalous meridional temperature advection.

The annual cycle of the EAM has shown a significant interannual variability, presenting the changes in phase and amplitude of the vernal equinox and summer solstice modes. These changes regulate the seasonal cycle of the East Asian summer and winter monsoon and modulate the seasonal precipitation variation in China. The phase changes are positively correlated between the vernal equinox and summer solstice modes on the interannual timescale. In the previous winter, El Niño–Southern Oscillation (ENSO) and SST in the western North Pacific could induce the interannual anomaly of the annual cycle in the EAM, causing the out-of-phase changes in rainfall between South and North China in the different seasons.Therefore, the annual cycle anomaly of EAM serves as a good reference for the crossing-seasonal climate prediction of rainfall in China (Jiang et al., 2020).

3. Impact of the EAM S2S variability on extreme climate events in China

Under the impact of global warming, the frequency,intensity, and duration of extreme climate events in China have changed significantly in recent decades. S2S variability of the EAM links climate change and persistent extreme weather and climate events. For instance,East China experienced the most prolonged and intense heat wave event from July to August 2013 since 1951.The persisting high-pressure anomaly directly induced this event over East China (Ma et al., 2017). In the midsummer of 2018, the WPSH shifted extraordinarily northward to cause an extreme heat wave in Northeast Asia because of the North Atlantic SST anomaly (Liu B. Q. et al., 2019). From April to June 2019, the negative AO and ENSO combined to make the WPSH extend westward continuously and result in a historic temperature soaring in Yunnan Province to produce a severe meteorological drought (Ma et al., 2021). In autumn 2019, under the influence of strong warming in the central equatorial Pacific, the persistent westward extension of the WPSH caused severe autumn droughts in the mid–lower reaches of the Yangtze River (Ma et al., 2020).

The frequency of extreme rainfall events is gradually increasing with global warming, which profoundly affects the country’s economy, society, production, and life. Under the impacts of tropical and subtropical circulation, the monsoonal rainfall over East Asia has shown a significant intraseasonal oscillation. Therefore, the S2S variability is vital for extreme precipitation in China(e.g., Hsu et al., 2016; Chen and Zhai, 2017; Ren et al.,2018; Li et al., 2021). For example, the 1998 catastrophic flood in the Yangtze River caused about 3000 deaths and a substantial economic loss of 260 billion Yuan. The intraseasonal oscillations determine extreme precipitation’s intensity and location (Zhu et al., 2003; Sun et al., 2016).In the summer of 2020, China experienced an extremely long-lasting Meiyu season. The accumulated rainfall in the Yangtze River broke its historical record in 1961.Our research shows the Meiyu front and its related circulation exhibited a significant subseasonal variability,showing the warm front in late June and the cold front in mid–late July. This transition was closely associated with the change in the NAO phase from positive to negative(Liu et al., 2020).

The Tibetan Plateau (TP) is the highest terrain in the Asian continent. The remarkable subseasonal variation of the atmosphere around the TP is one of the low-frequency oscillation centers outside the tropics (Zhang et al., 1991). The TP vortex (referred to as TPV) is a critical precipitation system over the TP (Ye and Gao, 1979).Studies have shown that after the TPV moves out of the TP, it could trigger severe weather processes, such as heavy rain and thunderstorms in the vast areas of eastern China. For example, the TPV directly induced the extreme heavy rainfall in the Sichuan Basin in July 2018(Li et al., 2020). Furthermore, the S2S variability of the EAM could modulate the TPV activity, which sets a basis for the S2S forecast of the TPV-related precipitation. Past studies have found that the activity of the TPV has a period of 10–20 days. Most TPV occurred in the positive phase of the 10–20-day quasi-biweekly atmospheric oscillation (QBWO) over the eastern TP (Li et al., 2018a). The QBWO can further affect the timing and direction of the moving-off TPV. Most TPV moved in the energy propagating direction of the QBWO. The QBWO also alters the diurnal variation of the TPV frequency (Li et al., 2018b). In general, the frequency of the TPV generation is minimum in 0600–1200 local time(LT) but peaks in 1800–0000 LT. Although the diurnal variation of environmental fields is similar between positive and negative phases of the QBWO, the dynamic and thermodynamic conditions favor the TPV generation at 0000 LT but inhibit the TPV formation at 1200 LT. The favorite conditions for the TPV generation get enhanced in the positive phase of the QBWO to increase the TPV.By contrast, the TPV decreases and weakens in the negative phase of the QBWO, along with the larger diurnal variation of the TPV frequency (Li et al., 2018b).

4. Effects of internal and external forcing on the S2S variability

The S2S timescale of the EAM is far beyond the memory limit of the initial atmospheric disturbance of 7–10 days. The external slowing forcing (e.g., ocean,land surface, and sea ice) thus should regulate the S2S variation of the EAM. Meanwhile, their effects on the S2S variability of the EAM have changed significantly under a warming climate. Since the annual cycle of the EAM links to climate change, external forcing, and subseasonal variability, it will be a hammer to knock on the nuts of the subseasonal predictable time limit of the EAM.

4.1 Impact of ENSO

As the strongest signal of interannual variability in the tropical Pacific Ocean–atmosphere coupling system, impacts of ENSO on the global climate have already become a critical predicting source of the S2S variability of the EAM. The influences on the EAM from ENSO in different stages are distinct. In boreal summer with El Niño development, precipitation becomes below-normal in North China but above-normal in the Yangtze and Huai Rivers. However, more precipitation exists south of the Yangtze River in summer after a decaying El Niño event with more significant climatic effects (Huang and Wu,1989; Chen, 2002). In boreal winter with mature ENSO events, the EAWM tends to be weaker in El Niño years but enhanced in La Niña years. Central Pacific (CP) ENSO events, characterized by SST anomalies in the central Pacific, occurred more frequently in recent decades.Compared with the traditional ENSO, the CP ENSO events with significant interdecadal changes have caused different impacts on the EAM (Xu et al., 2013; Chen et al., 2018). Due to the different responses of the EAM to the two types of ENSO, the anomalous location and intensity of the WPSH change in different seasons can alter the anomalous rainfall distribution in China (Xu et al.,2013). Our recent study has suggested that changes in ENSO diversity are closely related to global warming.Under global warming, the significant warming in the Indian and western Pacific Oceans has been found since the 1950s, accompanied by an expansion of the warm pool and enhanced convection in the western Pacific in the mid-1980s. It then intensified and shifted the Walker circulation westward, leading to the thermocline uplift in the central and eastern Pacific and the sudden change in the cold-tongue mode (Jiang and Zhu, 2020). In the meantime, the air–sea coupling center migrated westward over the Pacific. It shifted the SST warming center from the eastern to central Pacific, resulting in an asymmetric response to the diversity of ENSO events (Jiang and Zhu, 2018). The different impact of ENSO diversity is important for global warming effects on the EAM.

South China Sea summer monsoon (SCSSM) is an important component of the EASM. The SCSSM onset indicates the winter-to-summer transition of the EAM circulation pattern. Previous studies have attributed the interannual variability of the SCSSM onset to the ENSO events. The SCSSM onset generally delays following the warm phase but advances in the cold phase of ENSO events (Tao and Zhang, 1998; Zhou and Chan, 2007).However, significant decadal differences appear in the relationship between SCSSM onset time and ENSO events. From 1979 to 2020, two decadal shifts of SCSSM onset occurred in 1993/1994 and 2009/2010, respectively. The SCSSM onset has advanced since 1993/1994, after which the relationship between SCSSM onset and ENSO strengthened. Some studies have ascribed this change to the western Pacific warming (Kajikawa and Wang, 2012) or the interdecadal differences in ENSO (Ding et al., 2016; Liu et al., 2016). After 2010, the SCSSM onset has undergone a decadal shift again. Although the western Pacific continues to warm, the SCSSM onset tends to be late. ENSO events cannot indicate the SCSSM onset in this period. Under global warming, the frequent cold-tongue La Niña events in the past decade seem to break the traditional relationship between ENSO and SCSSM onset at the interannual timescale.Moreover, the warm pool warming triggers more subseasonal disturbances, which greatly increases the prediction uncertainty of the SCSSM onset (Jiang and Zhu,2021). For example, Typhoon Fani advanced the SCSSM onset in 2019 via an upscaling process (Liu and Zhu, 2020).

In addition to the variation of ENSO diversity, the responses of the EAM alter evidently in the extreme El Niño events. In general, the extreme El Niño events could increase the winter precipitation in South China in 1982/1983, 1997/1998, and 2015/2016. Compared with the case in 1982/1983 and 2015/2016, the 10–20-day oscillation contributed to the above-normal winter rainfall in South China without the low-frequency wave trains with a period above 30 days (Guo et al., 2018). The SST anomalies in the eastern Pacific (EP) El Niño stimulated the cyclone anomalies in the upper troposphere of East Asia to strengthen the midlatitude westerly waveguide.Afterward, the low-frequency wave trains propagated eastward onto South China and maintained the ascending flow in situ. However, the CP El Niño and the La Niña cannot induce the upper-level cyclone anomalies over East Asia, corresponding to the weaker subseasonal winter rainfall in South China (Fig. 2). These conclusions reveal the physical processes of how the ENSO affects subseasonal precipitation through the seasonal cycle of the EAM and provide the theoretical basis for improving the timeliness of subseasonal forecasts of winter rainfall in South China (Guo et al., 2021).

In addition, ENSO can influence the EAM in conjunction with other ocean basins. Studies have shown that the interaction in the Indo-Pacific–Atlantic SST anomaly(SSTA) determines the interannual variability of the seasonal cycle of the EASM via the air–sea interactions (Liu et al., 2021b). Simultaneously, the effects of the SSTA interaction on the EASM are also diverse and complex.For example, the relationship between ENSO and the WPSH is difficult to maintain when the capacitor effect in the Indian Ocean is damped. This process could explain the anomalous weakening of the WPSH in midsummer 2016 after the super El Niño event in 2015/2016(Liu et al., 2018). Other studies have also pointed out that before 1985, the North Atlantic and tropical Pacific SSTs had an offsetting effect on the winter precipitation in South China. However, the SST anomalies in the two basins had a superimposed effect, which enhanced the SST effects on the winter climate in South China after 1995 (Yu et al., 2021).

4.2 Role of land surface

The memory of soil moisture ensures the land surface effects on the subseasonal variation of the EAM. Spring rainfall anomaly in eastern China regulates the largescale EAM circulation via the anomalous soil moisture between the Yangtze and Yellow Rivers lasting more than two months (Zuo et al., 2015; Zuo and Zhang,2016). The spring anomalous soil moisture from the Yangtze to the Yellow River can further change the lowlevel temperature and humidity to modulate the summer precipitation in Northeast China. This result confirms the role of soil moisture in the S2S prediction and provides a reference for the processes related to soil moisture and large-scale circulation.

The atmospheric heating source over the TP is another way of land surface processes to affect the subseasonal process of the EASM. Interaction between the land surface process and atmospheric circulation over the TP maintains a quasi-biweekly oscillation of the TP atmospheric heat source in spring (Liu et al., 2021a). Thus,the circulation anomalies in the mid–high latitude amplify after crossing the TP via the air–land feedback (i.e., the“TP amplifier” effect). They further affect the critical processes of the EASM, such as spring rainfall in the south of the Yangtze River and the SCSSM onset (Jin et al., 2019; Liu and Zhu, 2021).

4.3 Influences of Arctic amplification

The Arctic is the cold air source for East Asia and China. The Arctic polar vortex modulates the S2S variability of the EAM (Yu et al., 2019; Dai et al., 2022), while the Arctic sea ice decline/variability is another external signal for climate prediction in China (Gong et al., 2004;Huang et al., 2012). As a cold polar on the earth, the Arctic plays an essential role in the global climate, weather, and environmental systems. With global warming,the Arctic region is not only a sensitive area for global climate change but also an important source of climate change affecting the Northern Hemisphere. The warming rate in the Arctic is much higher than that of the middle and low latitudes, about twice the global average rate, which is called the Arctic amplification (AA). With the rapid warming of the Arctic, the AA effect on climate has become increasingly prominent, and the Arctic has become a hot spot for global change research (Cohen et al., 2014; Coumou et al., 2018; Screen et al.,2018). Except for the local climate effect, the Arctic rapid warming directly or indirectly affects the weather and climate in the midlatitudes, leading to more frequent extreme weather and climate events in the Northern Hemisphere. The AA reduces the temperature gradient between the Arctic and midlatitudes to increase the atmospheric barotropic facilitating extreme weather and climate events in the midlatitudes (Cohen et al., 2014;Shepherd, 2016). The S2S variability of the EAM bridges the AA and extreme climate events in China.

Although a super El Niño with the warmest recorded global average surface air temperature occurred in the winter of 2015/2016, large-scale extreme cold surges attacked East Asia. The extreme temperature events presented the polarization characteristics. The cold surge in late January 2016 broke the records of the lowest temperature in many places and was called the “boss-level” cold wave. The S2S variability of the EAM enhanced evidently during the era of the recent AA to support the polarization of winter temperature in East China (Fig. 3).This phenomenon results from a combination of the thermodynamic effects of global warming and the dynamic effects of AA (Ma et al., 2018). Global warming increases surface temperature and extreme warm events,while the AA decreases the temperature gradient in the middle and high latitudes and the westerly wind over Eurasia. As a result, the meridional fluctuation of the circulation enhanced to maintain the Urals blocking and strengthen the Siberian high, which remarkably intensifies the cold advection in East Asia with more extreme cold events. Finally, extreme warm and cold events coexist over East Asia in a warming climate. The “boss-level”cold wave over East Asia is induced by the extremely positive anomaly of the Urals blocking and the recordbreaking Siberian high due to the internal atmospheric variability (Ma and Zhu, 2019). However, the AA increases the probability of extreme cold waves by the stronger dynamic process in the global warming background caused by human activities. In addition, because of the interdecadal variation of the North Pacific SST anomaly, the regional response of cold winter temperature extremes to the AA is opposite between eastern Eurasia and North America (Ma et al., 2020).

Fig. 3. Histogram of the cold-season daily temperatures averaged over eastern China (2792 stations) during the period of 1988/1989–1998/1999 (blue bars) with the Arctic surface air temperature (SAT)anomalies primarily negative and the other period of 2004/2005–2015/2016 (red bars) with the Arctic SAT anomalies primarily positive. White bars show the overlap distribution [adapted from Ma et al.(2018)].

Fig. 4. Box chart of (a) anomaly correlation coefficient (ACC) and (b) root-mean-square error (RMSE) between the ECMWF S2S real-time predicted and observed inland accumulated rainfall anomalies in different lead times for the Meiyu event in the warm- and cold-front period, respectively [from Liu et al. (2020)]. Scatterplots of (c) temporal correlation coefficients (TCCs) of the NAO index and (d) downstream circulation similarities [lag +7 days, similarity index (SI)] in the cold-front period versus the rainfall ACCs among the WMO S2S models. The upper-left numbers represent the intermodel correlation coefficients (r) and confidence level [adapted from Yan et al. (2022)].

Fig. 5. (a, b) Simulation ability and (c, d) phase deviation of different CMIP6 models for the (a, c) summer monsoon mode and (b, d) spring equinox mode of the EAM annual cycle [from Yan et al. (2020)].

5. The EAM S2S variability predicted by climate models

Operational S2S forecasting systems rely on the development of climate models. WMO holds an S2S database under the S2S prediction project, including the S2S forecast outputs of 11 operational centers worldwide (i.e., the S2S model). S2S models must consider the influence of the initial atmospheric variability and the effect of external forcing to predict the meteorological variables in the next 1–60 days (Vitart et al., 2017). At present, S2S models have a high prediction skill on the tropical intraseasonal oscillations (for example, the prediction skill of MJO has reached 32 days). However, the prediction skill of the subseasonal variations in the mid–high latitude (such as NAO and midlatitude atmospheric teleconnections) remains less than two weeks, not exceeding the lead time of the weather forecasting. Therefore, the limited predictability of the extratropical circulation is the bottleneck restricting the S2S prediction ability of the EAM by the numerical models.

5.1 Forecast skill of S2S variability

The sources of the predictability for S2S variability include the MJO, soil moisture, snow and sea ice, stratospheric–tropospheric interactions, oceanic states, and tropical–extratropical interaction (Vitart et al., 2017). Since the EAM is influenced by the complex interaction of tropical and mid-to-high latitudinal circulation, its predictability is relatively low at the S2S timescale. In particular, the eastward propagating MJO and the northward propagating boreal summer intraseasonal oscillation (BSISO) are the dominant modes of intraseasonal variability of the tropical atmosphere (Madden and Julian,1971, 1972; Wang and Xie, 1997). They are typical S2S predictability sources of EAM in the S2S models (Jie et al., 2017; Liang and Lin, 2018; Liu Y. Y. et al., 2019).The skillful forecast lead time for the MJO has reached 3–4 weeks by now. Some models, such as the ECMWF model, have a lead time of up to 5 weeks (31 days),which is closely related to the state at the initial time (Xiang et al., 2015; Kim et al., 2018; Wang S. G. et al.,2019). The prediction skill is higher when the forecasts are initialized during the strong MJO (Kim et al., 2014;Lim et al., 2018). The forecast lead time of BSISO in the S2S models is about 2–3 weeks, slightly shorter than that of MJO (Lee et al., 2015, 2017; Jie et al., 2017). The forecast lead time of intense BSISO events could reach 20–30 days. The MJO and BSISO influence the subseasonal prediction of the EAM by altering the extratropical atmospheric circulation through Rossby wave response and stratosphere–troposphere interaction (Stan et al., 2017). S2S models have certain predictability on these atmospheric teleconnections. However, it depends on the spatial location of the tropical convection, that is,the phase of MJO or BSISO at the initial time (Jie et al.,2017; Lim et al., 2019).

Although the model has certain predictability for MJO and BSISO, the subseasonal prediction skill of the EAM is much lower due to the atmospheric variability in the mid–high latitudes. The Meiyu rainfall is typical in the S2S variation of the EASM, and its dynamic prediction skill directly affects the subseasonal forecast skills of the EASM. Taking the extreme Meiyu event in 2020 as an example, although most seasonal prediction systems can predict the above-normal summer precipitation in the Yangtze River under the role of the warmer Indian Ocean,the predicted rainfall amount is much lower than the observation (Takaya et al., 2020; Zhou et al., 2021). The ECMWF S2S model exhibits higher prediction skills of Meiyu precipitation in the warm-front stage dominated by tropical forcings. By contrast, the skill decreases dramatically in the cold-front stage dominated by midlatitude systems (Figs. 4a, b) (Liu et al., 2020). Other S2S models show a similar poor skill of Meiyu rainfall in the cold-front stage. The relatively high skills in early June and mid–late July are primarily contributed by the fidelity of the tropical atmosphere circulations. In contrast,the mid–high latitude circulations reduce the skill in mid–late July due to the large bias in the predicted extratropical forcings (Yan et al., 2022). Moreover, the upscaling effect of heavy weather events (e.g., typhoons)also restricts the prediction skill of several subseasonal variations of the EASM (such as the SCSSM onset) (Liu and Zhu, 2021). Further research shows that the NAO-related variation and its impact on the downstream circulation in East Asia are less predictable in the S2S model,which contributes to the failure of the subseasonal forecast on the Meiyu rainfall in 2020. In contrast, the models with better prediction of the NAO showed a higher S2S skill on the Meiyu rainfall compared with the observation (Figs. 4c, d) (Liu et al., 2020; Yan et al., 2022).Therefore, the forecasts of midlatitude atmospheric variability are the key to improving the subseasonal predictability of the EAM.

5.2 Uncertainties in CMIP models

The prediction skill of the monsoonal climate depends on the ability of climate models to simulate the mean state and interannual variability (Lee et al., 2010). The developments of the model and observing system improve the overall simulation performance. For instance,following the development of the Coupled Model Intercomparison Project (CMIP), the simulation bias of the EAM has been decreasing in CMIP3, CMIP5, and CMIP6 (Chen and Frauenfeld, 2014; Song and Zhou,2014b; Kusunoki and Arakawa, 2015; Xin et al., 2020).However, systematic biases still exist widely in climate mean states (Huang et al., 2013; Feng et al., 2014; Xin et al., 2021), and interannual (Song and Zhou, 2014a; Fu et al., 2021; Yang et al., 2021) and decadal variability(Song and Zhou, 2014a; Park et al., 2020). In particular,complex and high orography causes a large bias in temperature and precipitation and underestimates the EASM rainfall and its southeast–northwest gradient (Salunke et al., 2019; Chen et al., 2020). In addition, the main factors affecting the intraseasonal variation of the EAM, such as ENSO and MJO, also have notable errors in the latest generation of CMIP6 multimodels (Chen and Jin, 2021;Le et al., 2021). These factors have profound implications on simulations and predictions of monsoon rainfall(Kim et al., 2017; Feng et al., 2019; Wang P. et al., 2019).

Except for the seasonal-mean monsoon intensity, the simulated monsoon processes or seasonal cycles also suffer from modeling errors (Sperber and Annamalai, 2014;Yang et al., 2015; Park et al., 2020). The simulation bias in the seasonal cycle will cause errors in seasonal mean and subseasonal variation, which are often ignored in the current climate simulation research. In the CMIP6 historical experiments, most models can capture the two dominant modes of the seasonal cycle of the EAM (Jiang et al., 2020). However, apparent differences still exist in the spatial distribution and temporal evolution of the simulated seasonal cycle modes (Yan et al., 2020). On the one hand, the model has better simulation performance for the summer monsoon mode (EOF1) associated with large-scale land–sea thermal contrast. On the other hand,the simulation bias is prominent in the spring equinox mode (EOF2) during the transition between spring and summer. The simulated phase of EOF2 lags the observations in most models. Some models even lag the observation by more than one month (Fig. 5) (Yan et al., 2020).This substantial error in the phase will affect the simulation of the EAM subseasonal process. To improve the simulation and prediction of the EAM, we should consider the bias of the seasonal cycle in the models.

6. Challenges and future directions

6.1 Challenges

Recent studies have shown that the annual cycle of the EAM accompanies strong subseasonal fluctuations, characterized by the coupling of the monsoon circulation.The annual cycle also exhibits a large interannual variability due to external forcing of SST anomalies and global warming. In contrast to the theoretical understanding of synoptic and annual cycle, the fundamental theory on the formation of subseasonal variability is still poor. For instance, it has been generally accepted that the MJO is the primary source of subseasonal variability and the target of model prediction. The prediction skill of the S2S model for MJO has reached 30 days or beyond; however, the model prediction skill of the subseasonal variability of EAM is still around the upper limit of weather forecasting. Therefore, it is a great challenge to improve the subseasonal forecasting skill of EAM based on the MJO prediction. In fact, the EASM is a typical subtropical monsoon in nature, and its subseasonal variability is largely affected by mid–high latitude and tropical circulations under the influences of multiple external forcing. In addition, accumulated evidence suggests that global warming exerts a significant impact on the persistent extreme climate events in China by enhancing the subseasonal variability of the EAM.

Observational studies have shown that the AA effect caused by rapid Arctic warming increases the East Asian temperature fluctuations at subseasonal timescale. The AA effect increases the probability of extreme cold waves, winter floods, and other extreme climate events.The TP is a sensitive area for climate change and the origin place of the severe weather systems in China. Under global warming, the subseasonal variability of atmosphere around the TP has enhanced significantly to induce more persistent extreme climate events in China.The Indo-Pacific warm pool is the active source of MJO.The observation suggests that the subseasonal variability of the EAM has enhanced under the influence of tropical convection due to the SST warming, which increases the difficulty of S2S prediction of the EAM. So far, we have little knowledge about how these external forcing affects the subseasonal variability of the EASM through the interaction of the tropical and mid–high latitude circulations. In addition, the current climate monitoring cannot accurately describe the S2S variability of the EAM, due to the overlook of the interannual variability of annual cycle. Because of the less physical understanding of the target and originality of annual cycle anomaly, the present benchmark of the seasonal evaluation criteria for the dynamical model needs to improve.

6.2 Future priorities for S2S research

The Phase 2 of WMO’s S2S research project aims to strengthen the construction of the S2S database, propose a series of research programs, and enhance the link between scientific society operation centers and user applications (WWRP and WCRP, 2018a). The research tasks include the following five fields: (a) MJO prediction and its teleconnection effects, including MJO effects on tropical/extratropical high-impact weather, potential forecasting skills for high-impact weather in two weeks to two months, MJO-related tropical–extratropical teleconnection, and its potential prediction techniques for extratropical systems; (b) impacts of land surface process and data assimilation, including soil moisture, snow cover, and vegetation changes, the impact of observing systems on land surface initialization and S2S prediction, the ability of S2S models to simulate land–atmosphere coupled processes, and the contribution of anomalous land surface processes to extreme weather/climate events; (c) influence of ocean and sea ice and their data assimilation, including direct effect of air–sea coupled processes on subseasonal variability and their performance in the S2S models, impact of ocean model resolution, impact of initial ocean state on S2S predictability, maintenance mechanisms and predictability of ocean subseasonal variability (e.g., ocean heatwaves), ability of seasonal prediction and the impact of sea ice initialization on S2S prediction,and the key process of subseasonal variability of sea ice and its performance in S2S models; (d) generation method of model ensemble members, including initial perturbation strategy for subseasonal prediction systems (starting time, number of ensemble members, ensemble member generator, etc.), establishing the error growth scoring index of the subseasonal forecasting system, exploring the influence of coupled initial disturbance on subseasonal forecasting, develop coupled initial disturbance technology, and studying stochastic parameterization impact on subseasonal prediction; (e) research on the influence of atmospheric composition, including the influence of aerosol precursor signals on S2S prediction through radiative processes, optimal configuration of model complexity, and S2S predictability of aerosols and its application value; and (f) research on the impact of stratospheric processes, including further improve the stratosphere–troposphere coupling process in the S2S model, develop an encrypted vertical level diagnostic system for the stratosphere, and strengthen the connection with other international stratospheric research programs.

Of which, the MJO is still the priority of WMO S2S programs for the future research. The MJO and its remote teleconnection enhance the model forecast of subseasonal variability of global atmosphere. However, based on our understanding of S2S variability of the EAM, we suggest to establish new theories and technology to reduce the uncertainty of the S2S model in the mid–high latitude over East Asia to improve the leading time of subseasonal forecast. In particular, the physical linkage between the annual cycle and subseasonal variability of the EAM may be effective in building a novel subseasonal prediction theory for the extreme climate events in China. In terms of this theory, we should explore the impact of the annual cycle of EAM on subseasonal variability of EAM and identify new predictable sources of subseasonal variability and extreme climate events in China.In this regard, we need to examine the fundamental physical processes and mechanisms of the timescale interaction of the S2S variability of the EAM. In addition, we need to recognize the physical process and predictability of the S2S variability of the EAM under the regulation of global warming and external atmospheric forcings. In the future, we need to build an objective forecast scheme for the subseasonal process, expand the application of subseasonal prediction, and provide climate forecast service in the ecological environment, hydrological processes,and clean energy.