Midwinter Breakdown of ENSO Climate Impacts in East Asia

The El Niño-Southern Oscillation (ENSO) in�uence on the East Asian winter monsoon (EAWM) exhibits remarkable non-stationarity on subseasonal timescales, severely limiting climate predictability. Here, based on observational and reanalysis datasets, we identify a robust subseasonal variability in the EAWM response to ENSO, with a notable synchronous break in mid-January lasting about 10 days. We suggest that this breakdown is largely caused by interference from the abrupt phase reversal of the ENSO-driven North Atlantic Oscillation (NAO), which occurs about a week earlier in early January. During El Niño years, the NAO phase transition from positive to negative triggers a rapid change in the mid-latitude atmospheric circulation via the quasi-stationary Rossby wave adjustment. This results in the strengthening of the Siberian high, which produces strong northerly wind anomalies over East Asia, while the anomalous western North Paci�c anticyclone weakens and shifts to the southeast, eventually leading to the collapse of the teleconnection.


Introduction
As the most prominent interannual climate variability on the planet, the El Niño-Southern Oscillation (ENSO) affects the weather and climate over most of the globe through a chain of oceanic or atmospheric teleconnections [1][2][3] .For the East Asian region, the seasonal-to-interannual climate response to ENSO is robust and has been extensively studied.It has been demonstrated that ENSO exerts its in uence in East Asia mainly by modifying the atmospheric circulation over the western North Paci c (WNP) with an anomalous low-level Philippine Sea anticyclone/cyclone [4][5][6][7][8] and a Kuroshio anticyclone/cyclone [9][10][11] .During El Niño winters, the southerly wind anomalies embedded in the western part of these anticyclonic circulation anomalies bring abundant warm and moist air to East Asia, leading to abnormal wet and warm climate states and a weakened East Asian winter monsoon (EAWM) 5,7,8,[11][12][13] .
When La Niña events occur, a roughly opposite picture is generally observed, with non-negligible asymmetric characteristics 14,15 .
However, this ENSO in uence in winter has been shown to be unstable on subseasonal timescales 16,17 .Compared to the late winter (January-February) EAWM, the early winter (November-December) EAWM has been shown to be more responsive to ENSO and thus more predictable in operational climate models 9,18,19 .This subseasonal change is suggested to be related to the sudden disappearance of the anomalous Kuroshio anticyclone/cyclone in January, which signi cantly weakens the ENSO in uences thereafter 9,20 .There are competing effects on the North Paci c atmospheric circulation between the ENSO-induced rainfall anomalies in the tropical Central Paci c (CP) and the WNP.While the former excites strong cyclonic circulation anomalies, the suppressed convection in the WNP is conducive to an anomalous anticyclone over the North Paci c.From early to late winter, the precipitation response to ENSO decreases in the WNP but increases in the tropical CP.These changes in the magnitude of convection anomalies result in the overwhelming effect from the tropical CP precipitation and the subseasonal disappearance of the anomalous Kuroshio anticyclone [9][10][11]20 . Othe studies argue that the disparate ENSO effects between the early and late winter are associated with the different impacts of ENSO on the tropical Walker circulation during the two periods 18,19,21 .During early winter, ENSO induces a double-cell anomalous Walker circulation over the tropics.This leads to pronounced precipitation responses in the tropical eastern Indian Ocean-western Paci c, which excites a systematic extratropical Rossby wave train from the tropical Indian Ocean towards East Asia, favoring an enhanced in uence on the EAWM 18,21 .In contrast, during late winter, ENSO is accompanied by a single-cell anomalous Walker circulation over the tropics.The precipitation response in the tropical eastern Indian Ocean-western Paci c is thus weakened, which can hardly produce an obvious wave train to exert much in uence on the EAWM 19,21 .
Although the above studies have documented the late winter weakening of the ENSO footprint in East Asia, the evolutionary feature of the EAWM responses to ENSO during the entire boreal winter season is still unclear.When, how, and why the ENSO in uence begins to weaken requires further systematic investigation.In this paper, based on the latest high temporal resolution observations and reanalysis datasets, we examine the daily evolution of the ENSO in uence on the EAWM and nd that the teleconnection is generally stable throughout the whole winter season, but exhibits a unique break for about 10 days in the midwinter.The possible mechanisms responsible for this collapse of the teleconnection are then analyzed.

Results
Midwinter breakdown of the ENSO in uence on the EAWM In Fig. 1a, we rst display the ENSO-regressed spatial pattern of the boreal winter seasonal-mean (NDJFM) 850-hPa wind anomalies to show the in uence of ENSO on the East Asian atmospheric circulation.As demonstrated in many previous studies [4][5][6][7][8][9][10][11] , when El Niño events occur, there are signi cant anticyclonic circulation anomalies over the WNP, which leads to pronounced southerly wind anomalies over East Asia and thus a weakened EAWM.To explore the subseasonal variations of this in uence, the monthly evolution of the EAWM response to ENSO is examined in Fig. 1b.It can be seen that, unlike the responses in November, December, February, and March, which show a signi cantly weakened EAWM, there is an apparent and unique breakdown of the ENSO in uence on the EAWM in January.We further examine the daily evolution of the EAWM response to ENSO in Fig. 1c.The weakened EAWM can be detected stably throughout almost all wintertime days.An exception occurs during midwinter, January 14-23, when the EAWM responses are uniquely strengthened, suggesting a breakdown and reversal of the ENSO-EAWM teleconnection.Correspondingly, the East Asian precipitation and SAT responses exhibit similar characteristics with the EAWM, i.e., El Niño events favor an increase in precipitation and SAT in East Asia, but this relationship cannot be established during the midwinter.Additionally, composite analysis is conducted to examine the daily evolution of the East Asian winter climate anomalies during El Niño and La Niña winters in Supplementary Fig. 1.It is also characterized by a unique midwinter collapse both in El Niño and La Niña winters, consistent with the results shown by regressions.We note that for East Asia, midwinter is typically a period with the coldest annual SAT (Fig. 1c).This unique breakdown during this period improves our understanding of why the ENSO-related predictability in East Asia is relatively low in January 22 .
To further illustrate this unique subseasonal change of the ENSO in uence, we divide the entire wintertime into three periods, namely P1 (November 1 to January 10) and P3 (January 27 to March 31) with weakened EAWM responses, and P2 (January 14 to January 23) with a strengthened EAWM anomaly (Fig. 1c).The spatial patterns of the ENSO-regressed atmospheric circulation (Fig. 2a-c), precipitation (Fig. 2), and SAT (Supplementary Fig. 2) anomalies during these three periods are then presented.It can be seen that during P1, El Niño events are accompanied by large-scale anticyclonic circulation anomalies over the WNP, with one center over the Philippines (i.e., Philippines Sea anticyclone) and another over the Kuroshio extension region (i.e., Kuroshio anticyclone).The anomalous southerly wind in the western part of these anomalous anticyclones strongly dominates over most of East Asia, resulting in a weakened EAWM, increased precipitation, and warm SAT responses (Fig. 2a, d).During P3, while the Kuroshio anticyclone disappears, the Philippine Sea anticyclone is still evident and has a large spatial scale.Thus, we can also observe distinct southwesterly wind anomalies over East Asia and similar climate anomalies to those in P1 (Fig. 2c, f).However, the circulation pattern is different for P2.
Northern East Asia is controlled by an anomalous cyclone rather than an anticyclone, which appears to be associated with the changes in the mid-to-high latitudes.The Philippine Sea anticyclone still exists but is mostly con ned to the tropics.As a result, we can detect evident northwesterly wind anomalies over the East Asian region and a strengthened EAWM (Fig. 2b).Correspondingly, the precipitation anomaly in eastern China and South Korea is reduced (Fig. 2b, e), and there also exist colder SAT responses (Supplementary Fig. 2b, e).Based on these results, we suggest that during midwinter, the ENSO in uence in East Asia undergoes a remarkable breakdown, which is manifested in both atmospheric circulation and surface responses.

Possible mechanisms responsible for the midwinter breakdown
We now turn to explore the possible mechanisms responsible for this midwinter breakdown of the EAWM response to ENSO forcing.It is unlikely that this abrupt subseasonal EAWM change can be directly interpreted by ENSO-related SST anomalies as they are highly persistent during the boreal winter season 23 .Although the ENSO-related SSTs show some subseasonal variations from November to March, for example, the wintertime SST anomaly gradually decreases in the tropical CP but increases in the tropical Indian Ocean (Supplementary Fig. 3a, c), they can hardly account for the abrupt collapse of the ENSO in uence.The tropical precipitation anomalies associated with ENSO are then analyzed.They are characterized by positive precipitation anomalies in the tropical CP and western Indian Ocean (WIO), and negative precipitation anomalies in the western Paci c (WP) (Supplementary Fig. 3b).Previous studies have suggested that the subseasonal variation in the response of the East Asian atmospheric circulation to ENSO is associated with the convective forcings in the tropical CP and western Paci c 9,10 , and the precipitation anomaly in the Indian Ocean can also play a role 18,19,24,25 .To test these hypotheses, we de ne three precipitation indices to measure the precipitation variability in the tropical CP (Pr_CP), WP (Pr_WP), and WIO (Pr_WIO), respectively (see Methods).Their time evolution during the boreal winter season is displayed in Supplementary Fig. 3d.Compared to those in P1 and P3, the precipitation indices show little unique or abrupt change in intensity during P2, suggesting that the subseasonal changes in the magnitude of the ENSO-induced tropical precipitation anomalies cannot directly explain the unique ENSO-EAWM teleconnection breakdown in P2.We next display the partial-regressed atmospheric circulation patterns with respect to the three tropical precipitation indices to examine the contributions of the changes in their effects (Fig. 3).It is found that while the CP convection exerts a cyclonic circulation anomaly over the North Paci c (Fig. 3a-c), the negative precipitation anomaly in the western Paci c is associated with an anomalous anticyclone there (Fig. 3d-f), which is consistent with previous studies 9,10 .
The effects of the WIO precipitation are relatively weak (Fig. 3g-i).Regarding the in uence change among the three periods, while the patterns related to the western Paci c precipitation anomaly remain almost unchanged, the CP precipitation forcing exerts some different effects over East Asia.In particular, the northerly wind anomalies over East Asia during P2, which are crucial for the EAWM response collapse, are mostly contributed by the ENSO-related local CP convective forcing.Therefore, it seems that the ENSO local forcing is more important to produce this midwinter breakdown.But how this ENSO local forcing leads to the anomalous northerly wind over East Asia is the next scienti c question to be resolved.
We then directly examine the regressed pattern of the northern hemispheric atmospheric circulation with respect to the Niño-3.4index during P1, P2, and P3 to explore the possible clues (Fig. 4, note that the regression with respect to the Pr_CP index (not shown) shows almost the same pattern).It can be seen that in P1 of the El Niño winter (Fig. 4a), there are pronounced large-scale low-level anticyclonic circulation anomalies over the Philippine Sea and the Kuroshio region.As a result, strong southerly wind anomalies prevail over East Asia, weakening the EAWM and producing enhanced precipitation.From a large-scale perspective, we can observe prominent Rossby wave trains extending from the Indo-western Paci c and the central Paci c to the North Atlantic, which projects onto a signi cant positive NAO response.However, the pattern in P2 shows a dramatically different picture (Fig. 4b).First, the Kuroshio anticyclonic circulation is replaced by an anomalous cyclone, and the Philippine Sea anticyclone is also observed but shrinks to the southeast of East Asia.Second, the Rossby wave train associated with the Indo-western Paci c forcing disappears.Only the wave train from the central Paci c reaches the North Atlantic, which is associated with the negative phase of the NAO pattern.
Since the wave source associated with the convection anomalies in the tropical Indo-western Paci c is still there, this disappearance of the wave train may be a result of interference from other forcing.Indeed, we observe a distinct atmospheric wave train in the mid-latitudes of Eurasia with a strengthened Siberian high and East Asian trough (i.e., weakened Kuroshio anticyclone), which brings strong northerly wind anomalies to East Asia and strengthens the EAWM.In P3 (Fig. 4c), although no signi cant change over the Kuroshio region, the Philippine Sea anticyclone is slightly restored to its original position to exert a signi cant in uence on the EAWM.Meanwhile, the atmospheric wave train activity in the mid-latitudes is largely weakened, which may be associated with the receded NAO response.
To sum up, the main reason for the collapse of the EAWM response in P2 appear to be the strengthened Siberian high which is embedded mid-latitude wave train adjustment associated with the NAO phase transition.We further show the daily evolution of the ENSO-regressed North Atlantic SLP anomalies in Fig. 4d.There is a clear abrupt NAO phase reversal from positive to negative in early January, which occurs about a week before the breakdown of the EAWM response.As demonstrated in the previous study 26 , this abrupt NAO phase reversal is primarily caused by the climatological change in the meridional shear of the Atlantic jet shifting the ENSO-induced low-frequency Rossby wave propagation direction, and is further facilitated or ampli ed by the North Atlantic's intrinsic positive feedback between the eddy and the low-frequency ow.It is well known that the NAO has important downstream in uences on the East Asian climate, mainly via quasi-stationary wave propagation of upper-tropospheric anomalies along the Asian jet, with a negative NAO generally associated with a strengthened Siberian high and EAWM [27][28][29][30][31] .Therefore, it appears that the Siberian high and East Asian trough strengthening, which are key reasons for the collapse of the EAWM response, are actually the results of the NAO phase transition.Considering that it usually takes about a week for the NAO signal to maximumly affect the East Asian atmospheric circulation on the sub-seasonal timescale 32 , the breakdown of the ENSO-EAWM teleconnection in the mid-January seems to be a consequence of the abrupt ENSO-induced NAO phase reversal that occurs about one week earlier.
To further probe the hypothesis, we focus on the spatio-temporal characteristics of the atmospheric circulation and wave activity related to the NAO. Figure 5a-c displays the horizontal wave activity ux corresponding to the 250-hPa geopotential anomalies associated with the sign-reversed NAO index at different lag days, together with the anomalous 250-hPa streamfunction and 850-hPa meridional wind.When the NAO is in its negative phase (lag = 0), there is a strong wave ux emanating from the NAO anomaly center over the North Atlantic and directed south-eastward across Europe and the Middle East to reach India (Fig. 5a).In addition to the south-eastward branch, a high-latitude pathway also emanating from the North Atlantic region emerges at lag = 3 days: this wave ux propagates eastward across Siberia and turns progressively south-eastward to affect the East Asian atmospheric circulation (Fig. 5b).6 days after the -NAO peak, we can still detect this high-latitude Rossby wave train (Fig. 5c).Correspondingly, along with the quasi-stationary wave propagation, we also see a strengthened Siberian high and East Asian trough at lag = 3 days and lag = 6 days.These suddenly appearing atmospheric circulation systems weaken and shift the anomalous WNP anticyclone to the southeast, and East Asia is taken over by the strong northerly wind anomalies, which strengthens the EAWM and thus leads to the breakdown or even reversal of the conventional ENSO-EAWM teleconnection mediated by the anomalous WNP anticyclones.In P3, the downstream effects of the NAO weaken as the negative NAO response to ENSO recedes (Fig. 4d).As a result, the Siberian high and the East Asian trough return to their normal states, and the anomalous Philippine Sea anticyclone regains its position to exert a weakening effect on the EAWM.
These results are consistent with a previous study that focused on the subseasonal in uence of the NAO on the East Asian winter climate 32 .
We further present the relationship between the changes in the NAO and EAWM indices before and after the reversal in Fig. 5d.They show a signi cant in-phase relationship, which means that when El Niño forces the NAO index to decrease from P1 to P2, the EAWM index will be also decreased subsequently.In other words, the winter monsoonal circulation will be intensi ed, further consolidating our previous hypotheses that the ENSO-driven abrupt NAO phase reversal around January 8 is responsible for the midwinter breakdown of the ENSO in uence on the EAWM.However, it should be noted that, in Fig. 5d, not all ENSO events are accompanied by signi cant NAO phase transition in the early January, suggesting that this ENSO-driven NAO phase reversal bears a considerable extent of uncertainty.Although beyond the scope of the current study, we assume that this uncertainty is likely related to the spatiotemporal complexity of ENSO events 33 and the nonlinear and nonstationary feature of the North Atlantic atmospheric response to ENSO [34][35][36] .Further details on this are still the subject of future investigation.

Conclusions and Discussions
The ENSO footprints in East Asian winter climate have been widely investigated.While the bulk of previous studies has focused on the in uence at seasonal-to-interannual timescale, more and more recent studies suggest that ENSO effects in East Asia may change at shorter timescales.For example, the early winter climate in East Asia has been shown to be more responsive to ENSO and more predictable in climate models, compared to the counterpart in late winter 9,10,18,19,21 .This motivates us to systematically explore the subseasonal evolution of the ENSO in uence on the East Asian winter climate.
The daily observations and reanalysis datasets during the past several decades are utilized in this paper, which allows us to detailly identify the possible subseasonal changes of the ENSO in uence.It is found that the ENSO teleconnection to the EAWM is generally stable throughout the whole boreal wintertime days, that is, El Niño events are typically accompanied by a weakened EAWM with enhanced precipitation and SAT in East Asia.However, it is interesting to nd that there is a remarkable and unique breakdown of the ENSO in uence on the EAWM during the midwinter.This collapse lasts about 10 days and mostly occurs in mid-January, when East Asia's climate is at its coldest point of the year, posing a potential threat to local socio-economic activities 37 .
The possible mechanisms responsible for this midwinter breakdown are then analyzed.It is suggested that the break is largely caused by an interference from the ENSO-driven NAO abrupt phase reversal in the early January.This ENSO-induced abrupt NAO phase transition has been investigated in our previous work 26 , which suggests that it is primarily caused by the climatological change in the Atlantic jet meridional shear, which shifts the propagation direction of the ENSO-induced low-frequency Rossby waves, and is further ampli ed by the intrinsic positive feedback between the eddy and the low-frequency ow in the North Atlantic.It is well known that the NAO has important downstream in uences on the East Asian climate, with a negative NAO generally associated with a strengthened Siberian high and EAWM 27- 31 .On the subseasonal timescale, the NAO exerts the strongest in uence on the East Asian atmospheric circulation about a week after its peak phase 32 .In our case, the abrupt NAO phase transition from positive to negative during early January of the El Niño winter leads to a rapid adjustment of the midlatitude atmospheric circulation via wave eastward propagation along the Asian jet.This leads to a strengthened Siberian high, producing strong northerly wind anomalies over East Asia, and the anomalous Philippine Sea anticyclone is shifted southeastward and con ned to the tropics, which eventually gives rise to the collapse of the teleconnection.During the La Niña winters, a generally opposite picture is presented.
Our ndings provide new insights into the non-stationarity of the ENSO effects on the EAWM, and carry important implications for the subseasonal predictability of the ENSO in uence on the East Asian winter climate.However, we note that this ENSO-NAO-EAWM linkage bears a considerable extent of uncertainty, possibly be related to the spatiotemporal complexity of the ENSO characteristics 33 , as well as the nonlinear and nonstationary NAO response to ENSO [34][35][36] .Anyway, the underlying reasons for this uncertainty are an interesting topic worthy of future study.In addition, as our current conclusions are based on observations and reanalysis datasets with a limited time record, it would be worthwhile to analyze the model representation of this midwinter breakdown of the ENSO in uence based on model simulations, such as the simulations in the Coupled Model Intercomparison Project phase 6 (CMIP6) 37 .

Re-analysis and observational products
The daily mean observational datasets used in this study are derived from the 699 weather stations in China (China Meteorological Administration, CMA) and 60 weather stations in South Korea (Korea Meteorological Administration, KMA).Two variables, precipitation and surface air temperature (SAT) are analyzed.For better reliability, we adopt the data during the recent period from Jan 1979 to Dec 2021.We also utilize the National Oceanic and Atmospheric Administration (NOAA) Extended Reconstructed global monthly SST V5 dataset 38 for the same period with a horizontal resolution of 2° longitude × 2° latitude.In addition, the daily atmospheric circulations, along with the daily 2-m temperature (i.e., SAT) and precipitation, are also derived from the fth generation of the European Centre for Medium-Range Weather Forecasts (ECMWF) reanalysis (ERA-5) 39 for the same period with the horizontal resolution of 1°l ongitude × 1° latitude.
All datasets are analyzed for the boreal winter season (November-March, NDJFMAM) in a broad sense, and the winter of 1979 refers to the winter of 1979/1980.Anomalies were derived relative to the daily mean climatology over the entire period of record (1979-2020).Linear trends have been removed to avoid possible in uences associated with global warming or long-term trends.

Climatic index
We employ several climatic indices to facilitate our analyses.The intensity of the EAWM is measured by the area-averaged boreal winter 850-hPa meridional wind anomalies (e.g., a positive EAWM index is associated with anomalous southerly winds and thus a weakened EAWM) within the domain of 20°-40°N and 100°-140°E 40 .According to the previous study 41 , the performance of the low-level wind indices is relatively better than the others in capturing the interannual ENSO-EAWM relationship.The Niño-3.4 index, de ned as the area-averaged SST anomalies in the Niño-3.4region (5°S-5°N, 120°-170°W), is used to describe the amplitude of ENSO.Following conventions, we utilize a threshold of ± 0.5 standard deviations of the Niño-3.4index during the December-February season to de ne ENSO events.With this method, 14 El Niño and 15 La Niña events are identi ed (Table 1).The North Atlantic Oscillation (NAO) index is de ned by the difference in normalized SLP anomalies between 35°N and 65°N over the North Atlantic sector 28 .All statistical signi cance tests are performed using the two-tailed Student's t-test.
Table 1 El Niño and La Niña events during the 1979-2020 period.

WAF
To analyze the source and direction of wave energy propagation, the wave activity ux (WAF) developed by Takaya and Nakamura 42 is applied.The WAF is parallel to the local group velocity that corresponds to the stationary Rossby waves and is independent of the wave phase 42 .It has been considered a useful tool for supplying information about wave propagation and is de ned as: where p is the pressure normalized by 1,000 hPa, a is Earth's radius, φ is the latitude and λ is the longitude.The geostrophic stream function ψ is de ned as z/f, where z is the geopotential, and f (= 2Ωsinφ) is the Coriolis parameter with the Earth's rotation rate (Ω).Also, |U|, U, and V represent the basic states of wind speed and zonal and meridional wind, whereas ψ' denotes the perturbed stream function.

Statistical signi cance test
All our results are tested based on the two-sided Student's t-test.

Declarations
Figures     and P1, and the ΔNAO index is de ned as the difference of the NAO index between the periods of January 9-19 and November 1-January 8.The correlation coe cient (R) between these two variables is also displayed.

Supplementary Files
This is a list of supplementary les associated with this preprint.Click to download.