Remote linkage of record-breaking U.S. Tornado outbreaks to the tropical cyclone in western North Pacific in December 2021

The frequency of tornadoes usually peaks during spring to summer rather than winter in climatology. However, the United States (U.S.) experienced more than 200 tornadoes in December 2021, which broke the historical record and caused 87 fatalities. Historically, the frequency of tornadoes in December tends to increase under El Niño conditions. Our results show that the monthly large-scale weather regime conducive to these record-breaking tornado outbreaks under a La Niña condition is closely associated with Typhoon Nyatoh in the western North Pacific. As the tropical cyclone (TC) recurved into the mid-latitudes, its interaction with the extratropical flows has caused distortions in the Asian jet stream and the dramatic development of anomalous anticyclone west of the dateline, which in turn strongly regulated the response of the monthly atmospheric teleconnection to La Niña forcing. Accurate forecasts of the monthly mean circulation for December 2021 first appeared in the European Center for Medium-Range Weather Forecasts sub-seasonal to seasonal forecasts from 29 November, with a forecast skill closely related to that of Typhoon Nyatoh. Given most studies on the warm seasons with frequent tornadoes, the present results advance our understanding of the TC effect on the monthly atmospheric response to El Niño-Southern Oscillation forcing and its linkage to the tornado occurrence during boreal winter.


Introduction
Prediction of tornadoes at extended lead times of more than a few days remains an arduous task (Gensini et al 2020, Miller et al 2020. To extend the lead time of tornado forecast, a substantial body of studies has been devoted to investigating the favorable environments of tornadoes at sub-seasonal to seasonal (S2S) timescales, which has advanced our understanding of the climate signals conducive to the U.S. tornado outbreaks (Tippett et al 2015).
Environmental conditions strongly influence the physical processes of tornado formation. Several environmental parameters, including convective available potential energy (CAPE), vertical wind shear (VWSH), and storm-relative helicity (SRH), are often used to identify the conducive environment for tornado activity (Tippett et al 2012). Specific large-scale weather patterns can modulate tornado activity via changes in these environmental parameters, which provides a basis for extended-range forecasts of tornado activity. For example, Miller et al (2020) have indicated that persisting favorable weather regimes contribute to greater than 70% of tornado outbreak days in the U.S., and weekly tornado activity can be predicted out to three weeks in advance using modelpredicted weather regimes.
A large-scale weather regime modulated by lowfrequency climate signals may last for weeks, resulting in tornado occurrence (Tippett et al 2015). These low-frequency climate signals may serve as sources of predictability for tornado occurrence on the S2S time scale. Many previous studies have shown that the tornado frequency in the U.S. was associated with the El Niño-Southern Oscillation (ENSO) (Cook and Schaefer 2008, Lee et al 2013, Allen et al 2015, Madden-Julian Oscillation (MJO) (Barrett and Gensini 2013, Thompson and Roundy 2013, Barrett and Henley 2015, Baggett et al 2018, Tippett 2018, Gensini et al 2019, Kim et al 2020, Moore and McGuire 2020, Miller et al 2022 and Arctic ice (Trapp and Hoogewind 2018).
Recent studies also explored the impact of the interactive relationship between MJO and global wind oscillation, which is defined by the global atmospheric angular momentum (AAM), on the U.S. tornado activity Marinaro 2016, Moore andMcGuire 2020). The onset of the significant MJO event is often immediately followed by above-normal AAM, the subsequent Rossby wave retrogression, and a favorable atmospheric regime for tornadoes (Miller et al 2022). Based on the above mechanism, Gensini et al (2019) successfully forecasted favorable tornado conditions in late May 2019, nearly four weeks in advance.
However, the connections between these climate signals and tornado occurrence are controversial (Tippett 2018, Kim et al 2020. Due to the limitation of the sample size, most studies focused on tornadoes during the warm seasons (e.g. Barrett and Gensini 2013, Lee et al 2016, Molina et al 2016. In a rapidly varying seasonal environment, the connections between tornado occurrences and the climate signals remain elusive, showing even monthto-month variations (Tippett 2018, Chu et al 2019. Therefore, whether the previous results are still valid in the cold season is a question.
In December 2021, more than 200 tornadoes occurred in the eastern U.S., breaking the historical record. The magnitude of the tornadoes is comparable to that in the warm season. It allows us to examine the validity of previous theories in the cold season. The local mesoscale processes on a short time scale that resulted in tornadoes are quite complicated. In the present study, we focus on the large-scale circulation features associated with the record-breaking U.S. tornado occurrence on the S2S time scale. Some findings may provide insights into S2S forecasts for tornado activity.

Data description
The U.S. tornado dataset is taken from the Storm Prediction Center (SPC) Severe Weather Database. The typhoon data was extracted from the International Best Tracks Archive for Climate Stewardship (IBTrACS version 04) (Knapp et al 2010, Knapp et al 2018. The NOAA interpolated daily outgoing longwave radiation (OLR) (Liebmann and Smith 1996) and the monthly sea surface temperature (SST) from the Hadley Centre Sea Ice and SST data set version 1 (HadISST1) with a 1 • × 1 • grid (Rayner et al 2003) are used. The global atmospheric general circulation records are from the Japan Meteorological Agency JRA-55 reanalysis dataset (Kobayashi et al 2015, Harada et al 2016. The NCEP-NCAR Reanalysis-2 dataset (Kanamitsu et al 2002) and North American Regional Reanalysis (NARR) dataset (Mesinger et al 2006) are also used to support the results in the supplemental material.
The Pentad MJO indices (Baxter et al 2014) and Oceanic Niño Index (ONI) are available online from the Climate Prediction Center. The Pentad MJO indices represent the negative projection of 200 hPa velocity-potential (χ200) anomalies onto the ten time-lagged patterns of the first extended empirical orthogonal function (EEOF) of pentad χ200 anomalies. The Pentad MJO indices can be viewed graphically to examine the evolution of 200 hPa divergence (30 • S-30 • N) ( figure 3(b)). In addition, the OLR MJO Index (

AAM and WAF
AAM was calculated using daily averages of zonal wind via the following formula: where a is Earth's radius, g is the gravitational constant, ϕ is latitude, λ is longitude, u is zonal wind speed, and p is pressure. Anomalously high (low) AAM is typically associated with anomalously strong (weak) zonal mean winds. The 200 hPa Rossby wave source (RWS) is calculated as (Sardeshmukh and Hoskins 1988): RWS is computed using the magnitudes of the divergence D, the absolute vorticity ζ a , and the irrotational component of the wind v χ . Following Takaya and Nakamura (2001), the horizontal 200 hPa wave activity flux (WAF) is also used to explore the stationary RWS and wave propagation, which is calculated as: where the overbars and primes denote the climatology and its deviation, the subscripts x and y represent the zonal and meridional gradients, respectively, ψ is the stream function, and U is the magnitude of the climatological wind. The climatological seasonal cycle from 1991 to 2020 is removed to derive the anomaly of each variable.

Event summary
The tornado frequency usually peaks in the warm season during spring to summer in climatology (figure 1(a)). The likelihood of an early June tornado occurring in the U.S. reaches 90%, in contrast to only 12%-15% likelihood of a tornado days in December, as estimated by NOAA. According to SPC, 202 preliminary tornados occurred in December 2021, causing 87 fatalities. This number is more than sevenfold the past years' (1979-2020) average of 26.1 in December, breaking the record (figure 1(b)). Notably, on December 10-11 and 15, there were 102 and 58 tornadoes, respectively (figure 1(c)). The former was the deadliest December tornado outbreak (93 direct and indirect fatalities) recorded in the U.S., surpassing the Vicksburg, Mississippi tornado of December 5, 1953 (38 fatalities), while the latter produced the first December tornado on record in Minnesota since 1950 (NOAA National Centers for Environmental Information 2022). These December tornado outbreaks were across the southeastern U.S. and associated with a favorable weather regime (figures 1(d) and S1). The weather regime is characterized by a zonally elongated low and a high center over the eastern U.S., which has been recognized as one of the most effective weather regimes generating tornadoes since it often juxtapositions abundant and deep lower-tropospheric moisture from the Gulf of Mexico, strong tropospheric wind shear, and steep mid-tropospheric lapse rates (Miller et al 2020). This unexpected tornado occurrence in December 2021 is associated with the record monthly mean CAPE and the third-largest SRH since 1979 (figures S2 and S3). Strong VWSH is north of the tornado region (figure S3(c)). Though such a conducive weather regime for tornadoes is usually reported in spring and summer (Gensini et al 2019, Miller et al 2020, it is still valid in December (figure 4 and text S1).

Influence of ENSO
Though the variability of tornado activity explained by ENSO is rather limited (Lepore et al 2017), ENSO has a detectable impact on the U.S. tornado activity in December (Allen et al 2018). The tornado frequency in December seems to be enhanced during El Niño ( figure 1(b)). Except for 2021, the top four active tornado months in December (1982, 2002, 2015, and 2018) were all during El Niño episodes. A composite of months with active tornadoes during El Niño phases was performed in figure 2, where ONI >0.5 and tornado numbers >26.1 (marked by red ovals in figure 1(b)). This shows the large-scale circulation conditions in most cases when tornadoes are active in December. Under the impacts of El Niño, anomalous divergence is produced from deep tropical convection, with flow peaks at the edges of the heating region, resulting in anomalous convergent regions in the subtropics ( figure 2(a)). The position of the jet anchors RWS and often determines the major Rossby wave response of the North Pacific (figure 2(a)). For most active tornado months in December during El Niño, a pressure pattern similar to Pacific-North America (PNA) pattern emerges as a stationary Rossby wave, which is characterized by a deepened Aleutian low (L) and an increased high (H) downstream (figure 2(b)). Such a large-scale weather pattern over the U.S. (figure 2(b)) is roughly similar to that of December 2021 (figure 1(d)), facilitating the occurrence of tornadoes.
To first order, the composites of the La Niña months (figures 2(c) and (d)) and the active tornado months during El Niño (figures 2(a) and (b)) are roughly symmetric, but the asymmetrical components lead to rather different climate impacts. The most notable asymmetry occurs over the North Pacific, where the response of the Aleutian low (AL) during La Niña is much weaker (figure 2(d)). Moreover, the downstream center (L) in figure 2(d) is located at higher latitudes, where its influence on tornadoes is not apparent. The distribution of the tornado frequency during La Niña is similar to that during the ENSO-natural years ( figure 1(b)).
During La Niña, the large-scale circulation pattern in December 2021 (figure 2(b)) exhibits a negative PNA-like pattern, which is featured by three centers over the AL region (H), western Canada (L), and southeastern U.S. (H), respectively. Notably, the anomalous GPH over the AL region plays an important role in modulating North American weather (Gibson et al 2020). Compared to the La Niña composites, the anomalous center over the AL region is further west. The averaged GPH over the AL region (30 • N-60 • N, 165 • E-145 • W) is the strongest since 1958 (figure S6), corresponding to the recordbreaking U.S. tornado outbreaks in December 2021.
The monthly mean atmospheric circulation is influenced by lower boundary forcing (e.g. ENSO) and internal variability. There is a pronounced month-to-month variation in the extratropical response to ENSO. It is reported that during El Niño, the westerly jet extends eastward, allowing forced signals to account for a greater fraction of the total variability and leading to increased potential predictability. In contrast, the Northern Hemispheric atmospheric response to La Niña forcing in December is most susceptible to internal variability  (Chapman et al 2021). Compared to the La Niña composite (figure 2(c)), the jet stream in December 2021 contracts more westward (figure 2(e)). Therefore, it is speculated that, besides the La Niña forcing, the large-scale circulation pattern in figure 2(f) is likely to be related to the sub-month or sub-seasonal atmospheric activities.

Sub-seasonal evolution
Following previous studies (Gensini et al 2019, Miller et al 2020, it is possible to anticipate highimpact forecasts of opportunity by identifying slowvarying processes involving the tropical forcing, upper-level divergence, and subsequent damping of AAM (figure 3). The pentad upper-level divergence since mid-November damped first over the Maritime Continent (MC) and began to strengthen suddenly in the western Pacific (WP) from 24 November to 4 December ( figure 3(b)), corresponding to the enhanced convection ( figure 3(a)). Afterward, the upper-level divergence weakened quickly. At the same time, the AAM anomalies around 30 • N strongly decreased to negative, and the AAM anomalies around 50 • N increased to positive (figure 3(c)). The period of high tornado occurrence (December 5-15) was immediately following the strongest upper-level divergence (about 4 December) dominated by persisting low mid-latitude AAM. The low AAM in the mid-latitude, in conjunction with the high AAM at 50 • N, is speculated to represent the breakdown of the Pacific jet with the development of large-scale anticyclones in the mid-latitude. The decreasing AAM in the mid-latitude facilitated a transition from a more zonal to more meridional orientation, leading to high-impact weather events (figure 3(c)). It seems that the changes in AAM are associated with the suddenly strengthened convection ( figure 3(a)) that induced the upper-level divergence ( figure 3(b)) from 29 November to 4 December 2021.
The pentad MJO indices in figure 3(b) easily reminds us that the processes may be related to the MJO activity. However, the propagation characteristics of tropical convection are not obvious ( figure 3(a)). It is well-known that MJO indices which incorporate large-scale circulation features are susceptible to influence from convectively coupled equatorial waves, and different indices can vary widely as to their depiction of MJO characteristics (phase/amplitude) or its existence at all (Kiladis et al 2014). The pentad MJO index only incorporates 200 hPa velocity potential equatorward of 30 • , which thus incorporates influences from the subtropics ( figure 3(b)). With further examination, it is found that the suddenly strengthened upper-level divergence from 29 November to 4 December 2021 is caused by Typhoon Nyatoh in the western North Pacific rather than MJO activity. The initial disturbance, which eventually became Typhoon Nyatoh, formed on 26 November 2021 and rapidly strengthened and recurved into the mid-latitude ( figure S8(a)). The commonly used RMM index ( figure S9(b)) is shown to become active just as Typhoon Nyatoh becomes a named storm ( figure  S8(a)). The OMI, which more directly tracks the convective component of the MJO, briefly becomes marginally active, but for a very short period ( figure  S9(a)). In particular, there are no large-scale circulation features similar to those of MJO (figure 4), and the shape of the enhanced convection is highly coincident with the typhoon track (figure 4(c)).
Previous studies (Archambault et al 2013(Archambault et al , 2015 have revealed that a tropical cyclone (TC) recurving into the mid-latitudes can regulate the extratropical flow pattern, suggesting a possible connection between the U.S. tornado outbreaks and Typhoon Nyatoh. As mentioned above, the most striking feature in December 2021 (figure 2(f)) is the unexpected anomalous GPH over the AL region, including its record intensity and westward shift. Therefore, the following main focus will be on the typhoon's role in the development of the anomalous GPH over the AL region.
Before the end of November, the AAM was stable ( figure 3(c)). Significant changes for the AAM were concentrated over the period of the early December ( figure 3(c)). During the end of November, a prominent wave train originated over the North Atlantic and propagated downstream over Eurasia, which resembled the Eurasian (EU) teleconnection, and no apparent poleward-propagating Rossby wave train   was triggered by the weak convection over the MC ( figure 4(b)). Afterward, Typhoon Nyatoh developed and moved northward over the WP (figure 4(c)). The RWS to the north of the enhanced convection in the WP arose from southerly TC-induced divergent wind crossing the Asian jet, resulting in the distortions in the jet stream and dramatic development of anomalous anticyclone over the jet exit ( figure 4(b)), forming an obvious wave train ( figure 4(d)). Subsequently, a wave train characterized as the negative PNA pattern was formed ( figure 4(f)). It resembles a preferred transition route from EU teleconnection to PNA (Mori and Watanabe 2008).
Though a PNA-like response can be forced by La Niña (figure 2(d)), the dramatic development of the anomalous GPH over the AL region was tied to the interaction between Typhoon Nyatoh and the Asian jet ( figure 4(c)). It means that besides La Niña, the monthly mean circulation for December 2021 is also strongly influenced by Typhoon Nyatoh. In contrast to the prediction for ENSO, the current numerical forecasting skills for typhoons are limited to only a few days. Therefore, it is presumed that the subseasonal forecasting skill of the monthly mean circulation for December 2021 is closely related to the prediction for Typhoon Nyatoh.
The predicted circulations for December 2021 are available from the ECMWF S2S predictions initiated on the 15th, 18th, 22nd, 25th, and 29th of November 2021. They all well predicted the La Niña state in December 2021. While only the predictions initialed from 22 November 2021 onwards can predict typhoon activities, and a relatively accurate forecast for Typhoon Nyatoh is first found in the prediction initialed on the 29th of November 2021 (figure 5(a)). A comparison between the predictions initialed on the 15th and 29th of November 2021 is made to confirm the importance of TC Nyatoh on the monthly mean circulation pattern.
Without the TC disturbance, the predicted Asian jet during 1-4 December ( figure 5(b)) is somewhat similar to that during 26-30 November in the observation ( figure 4(a)). The jet stream is flat, and its exit position is located east of the dateline ( figure 5(b)). The corresponding monthly circulation anomalies (figure 5(c)) resemble those for the La Niña composite ( figure 2(d)). Accordingly, the large-scale circulation as a PNA-like pattern (figure 5(c)) is speculated as a response to La Niña. On the other hand, the interaction between the typhoon and the Asian jet is well predicted in figure 5(d). The corresponding monthly circulation pattern (figure 5(e)) as the westward shift of that in figure 5(c) is nearly consistent with the observation. In contrast to the three GPH centers of comparable intensity in figure 5(c), the intensity of the GPH center over the AL region (figure 5(e)) is extreme, with a westward shift, which is consistent with the observation.

Summary and discussion
Under a La Niña condition, the U.S. experienced record-breaking tornadoes in December 2021. The present results show that the monthly large-scale circulation conducive to tornadoes is closely linked to Typhoon Nyatoh. The sub-seasonal predictions for large-scale circulation in December 2021 strongly depend on the accurate forecast for Typhoon Nyatoh.
Although our current research focuses on the role of Typhoon Nyatoh, the importance of La Niña as a slowly varying boundary forcing should not be ignored. Without the TC disturbance, the atmospheric circulation under La Niña is also similar to that of December 2021 as a negative PNA-like pattern, except for its position and intensity. This means that the large-scale circulation in December 2021 appears to result from a monthly circulation response to La Niña forcing regulated by Typhoon Nyatoh. Consistently, Chapman et al (2021) have revealed that along with the contraction of the Asian jet stream, the atmospheric teleconnection response to La Niña forcing in December is most susceptible to internal variability. Moreover, the high-frequency convection activities, including the TCs over the western Pacific, tend to be more active under a La Niña state. The combined effect of these can reduce the predictability of the monthly mean circulation.
However, the underlying mechanisms may provide us with an extended prediction opportunity. A preferred transition route from EU-like teleconnection to PNA implies the possible indicative role of the EU-like wave train. In addition, the sub-seasonal forecasting skill for the monthly mean circulation may be relevant to the prediction of the interactions between TCs and jet.
In present study, we mainly discussed the conducive large-scale circulation and its linkage to the TC. However, the factors contributing to tornadoes are complex, involving many local mesoscale processes. The large-scale weather conditions are not sufficient conditions for tornado formation. Especially, the connections between tornadoes and climate signals show month-to-month variations (Tippett 2018, Chu et al 2019). Extreme tornadoes in cold seasons, like the one in December 2021, have never happened before ( figure 1(b)), so it is difficult to find some statistical patterns. In addition, the present study suggests a potential framework in which to evaluate numerical model forecast error and uncertainty associated with the TC-Asian jet interaction to provide more information for the probabilistic forecasts, as discussed in some previous studies (Archambault et al 2013(Archambault et al , 2015. In the future, we will focus on how much the interaction of typhoons and the Asian jet affects the monthly large-scale circulation under ENSO forcing.