On the representation of major stratospheric warmings in reanalyses

1 On the representation of major stratospheric warmings in reanalyses Blanca Ayarzagüena1,2, Froila M. Palmeiro1a, David Barriopedro2, Natalia Calvo1, Ulrike Langematz3 and Kiyotaka Shibata4 1Dpto. Física de la Tierra y Astrofísica, Universidad Complutense de Madrid, Madrid, 28040, Spain 2Instituto de Geociencias (IGEO), Consejo Superior de Investigaciones Científicas - Universidad Complutense de Madrid, 5 Madrid, 28040, Spain. 3 Institut für Meteorologie, Freie Universität Berlin, Berlin, 12165, Germany. 4 School of Environmental Science and Engineering, Kochi University of Technology, Kami, 7828502, Japan. a now at: Barcelona Supercomputing Center (BSC-CNS) 10 Correspondence to: Blanca Ayarzagüena ([email protected]) Abstract. Major sudden stratospheric warmings (SSWs) represent one of the most abrupt phenomena of the boreal wintertime stratospheric variability, and constitute the clearest example of coupling between the stratosphere and the troposphere. A good representation of SSWs in climate models is required to reduce their biases and uncertainties in future projections of 15 stratospheric variability. The ability of models to reproduce these phenomena is usually assessed with just one reanalysis. However, the number of reanalyses has increased in the last decade and their own biases may affect the model evaluation. Here we compare the representation of the main aspects of SSWs across reanalyses. The examination of their main characteristics in the preand post-satellite periods reveals that reanalyses behave very similarly in both periods. However, discrepancies are larger in the pre-satellite period than afterwards, particularly for the NCEP/NCAR reanalysis. All datasets 20 reproduce similarly the specific features of wavenumber-1 and wavenumber-2 SSWs. A good agreement among reanalyses is also found for triggering mechanisms, tropospheric precursors and surface response. In particular, differences in blocking precursor activity of SSWs across reanalyses are much smaller than between blocking definitions. 1 Introduction Major sudden stratospheric warmings (SSWs) constitute the most important phenomena of the Northern Hemisphere polar 25 stratospheric variability in wintertime. They are abrupt warmings of the polar stratosphere that lead to a deceleration of the polar vortex and a reversal of the typical westerly circulation (Andrews et al., 1987). SSWs can be classified into two different types according to the structure of the polar vortex during the event. Accordingly, the polar vortex is either displaced from the polar cap (vortex displacement, D SSWs) or split into two parts of similar size (vortex split, S SSWs) (Charlton and Polvani, 2007). 30 SSWs represent a clear example of stratosphere-troposphere coupling in both directions. First, they are usually preceded by an enhancement of wave activity (e.g. Matsuno, 1971). Although this enhancement can take place in the lower troposphere, recent 2 studies have shown that it often happens within the stratosphere or tropopause region and depends on the stratospheric mean flow conditions (Sjoberg and Birner, 2014; Birner and Albers, 2017; de la Cámara et al., 2017; White et al., 2019). The sources of upward-propagating wave activity are mainly located in the mid-to-upper troposphere and correspond to anomalous 35 circulation events such as a deepened Aleutian low (e.g.: Garfinkel et al., 2010) or blocking highs, among others (e.g. Martius et al., 2009; Nishii et al., 2011; Ayarzagüena et al., 2011; Barriopedro and Calvo, 2014). Based on the wave activity preceding SSWs, they are commonly classified into wavenumber 1 (WN1) or wavenumber 2 (WN2) events (e.g. Bancalà et al., 2012; Barriopedro and Calvo, 2014). This classification produces subsets of events similar to the D/S catalogue. However, there are differences since the former is based on the precursory wave activity while the D/S classification accounts for the shape of the 40 polar vortex during the post-warming phase (Bancalà et al., 2012). Depending on the type of SSWs, the tropospheric precursors are different and/or located in different geographical locations (Martius et al., 2009; Cohen and Jones, 2011; Bancalà et al., 2012). In particular, differences in blocking precursors are larger when SSWs are classified into WN1/WN2 than D/S (Barriopedro and Calvo 2014). In terms of downward coupling, the SSWs signal propagates downward and reaches the troposphere as revealed from 45 composite analyses (Baldwin and Dunkerton, 2001), although there is still uncertainty about this tropospheric response when analyzing individual events (e.g.: Gerber et al. 2009). One of the suggested factors that may contribute to the spread of the surface signature of SSWs is the type of event. However, while some studies have shown that only S SSWs have large effects on surface climate (Mitchell et al., 2012, Seviour et al., 2013), others have not found consistent differences between S and D SSWs in its significant surface impact (Charlton and Polvani, 2007; Cohen and Jones, 2011). Thus, there is not yet a consensus 50 in this regard, probably due to the differences in the algorithms used to identify S and D SSWs (Maycock and Hitchcock, 2015). As for WN1 and WN2 SSW, their surface signature has not yet been explored. SSWs are a key element when analyzing stratospheric variability. The frequency and seasonality of SSWs are common metrics to assess the effects of tropospheric and oceanic phenomena on the polar night jet (PNJ). These metrics are also used to evaluate the stratospheric response to climate change (e.g.: Taguchi and Hartmann, 2006; Charlton-Perez et al., 2008; Ayarzagüena et 55 al., 2018). Indeed, in modeling studies most of them use simulations that are previously validated by comparing their results with reanalysis datasets (e.g.: Charlton et al., 2007; McLandress and Shepherd, 2009; Kim et al., 2017). However, the number of reanalyses has increased in the last decade, and although the observational data used in the assimilation process is the same, the reanalysis models are different, and so may the final products be (Fujiwara et al., 2017). As it happens with other atmospheric models, reanalyses also have biases and this can affect the model evaluation (Fujiwara et al., 2017). 60 Due to quality improvements associated with the assimilation of satellite data, modern reanalyses, such as ERA-Interim, NASA-MERRA, and NCEP-CFSR, only cover the post-satellite period since 1979. This means that the number of available reanalyses to assess the model performance in the pre-satellite era is smaller than in the post-satellite period. In addition, the amount of data to assimilate is also limited in the former period. All this might produce artificial differences in results before and after the inclusion of satellite data. Gómez-Escolar et al. (2012) documented a change of some SSW features from the pre65 satellite to the post-satellite era in NCEP-NCAR and ERA-40 reanalyses. For instance, the intra-seasonal distribution and the 3 amplitude of the SSW-associated warming showed differences between both periods, potentially due to a change in the type of the assimilated data. With the availability of the new JRA-55 reanalysis, which is the only one that applies an advanced data assimilation scheme to upper-air data during the pre-satellite era, revisiting this topic seems appropriate. In this study, we aim to assess the performance of the most widely used reanalyses in representing SSWs. To do so, first, the 70 main characteristics of SSWs are examined for all datasets to quantify the degree of agreement across reanalyses. Both preand post-satellite periods are compared to investigate whether discrepancies among reanalyses in the representation of the main SSW characteristics depend on the examined period. Secondly, we address the dynamical forcing of SSWs in all datasets, including precursors such as blockings. Finally, the surface impact of SSWs retrieved from the different reanalyses is analyzed. Special emphasis is given to the assessment and robustness of the potential differences in the forcing and surface impact of 75 WN1 and WN2 SSWs, as well as S and D events. Our work is a contribution to the Chapter 6 of the Stratosphere-troposphere Processes And their Role in Climate (SPARC) Reanalysis Intercomparison Project (S-RIP) initiative, which aims to assess stratosphere-troposphere coupling in reanalyses. In the framework of this initiative, a few recent studies have addressed some aspects of the representation of polar stratospheric variability in reanalyses. In particular, Martineau et al. (2018) and Hitchcock (2019) also investigate SSWs-related aspects. 80 The former analyzes the momentum budget during SSWs restricted to the post-satellite period, while Hitchcock (2019) compares the representation of stratosphere-troposphere coupling in both pre and post-satellite period, with the emphasis on the impact of including pre-1979 data. Different from these studies, our work provides a comprehensive inter-reanalyses comparison of the most important and typical aspects and processes associated with SSWs in both preand post-satellite eras. Additionally, we explore further the characteristics of WN1 and WN2 SSWs that have not yet been investigated. 85 The structure of the paper is as follows. The data used and methodology applied are described in Section 2. Section 3 compares the performance of the main characteristics of SSWs across reanalyses. Section 4 focuses on the dynamical forcing of the events and Section 5 addresses the performance of reanalyses in representing the surface impact of SSWs. The main conclusions are summarized in Section 6. 2 Data and methodology 90 2.1 Data We have used daily data from the following reanalyses: ERA-40 (Uppala et al., 2005), ERA-Interim (Dee et al., 2011), JRA25 (Onogi et al., 2007), JRA-55 (Kobayashi et al., 2015), NASA-MERRA (Rienecker et al., 2011), NCEP-CFSR (Saha et al., 2010), NCEP-DOE (Kanamitsu et al., 2002), and NCEP-NCAR reanalysis (Kalnay et al., 1996). More details about the different reanalyses can be found in Fujiwara et al. (2017). For the comparison across different reanalyses, all data was used 95 at the common regular grid of 2.5° lon x 2.5° lat. When not directly available from the reanalysis centers, a first order conservative remapping was applied. 4 The methodology for the intercomparison follows the S-RIP specifications. As such, the analysis has been carried out for two different periods: historical (1958-1978) and comparison (1979-2012). Given the periods covered by each reanalysis, only ERA-40, NCEP-NCAR, and JRA-55 are employed in the historical period. In contrast, all the above listed reanalyses are used 100 in the comparison period with the exception of ERA-40, because it ends in 2002. The performance of each reanalysis is evaluated against a multi-reanalysis mean (MRM), herein considered as an “unbiased” reference. In the historical period, the MRM refers to the average of the three reanalyses that cover that period, while in the comparison period, the MRM is defined as the average of the most recent reanalyses of each center (ERA-Interim, NCEP-CFSR, JRA-55 and NASA-MERRA). Hereafter, anomalies for each reanalysis are defined as the departure of the field from the daily climatology of each reanalysis. 105 In the historical period, the climatology covers the whole period (i.e. 1958-1978), whereas the comparison period uses the 1981-2010 baseline. Unless otherwise stated, statistical significance of the results is computed with a Monte-Carlo test of 1000 permutations, each one containing the same number of cases and dates as the SSWs of each composite but with random years of occurrence. 2.2 Criteria for the identification of SSWs 110 We have used the list of SSWs and common dates identified in Butler et al. (2017) and provided for the S-RIP initiative (Chapter 6), unless otherwise indicated. First, for each reanalysis, SSWs are identified based on the reversal of the zonal mean zonal wind at 60ºN and 10hPa between November and March, with at least 20 days of separation between events. Stratospheric final warmings are excluded by requiring at least 10 consecutive days of westerly winds before the end of April (Charlton and Polvani, 2007). The first day of reversal of winds determines the date of occurrence of the SSW (the so-called central date). 115 Common SSWs are those identified by at least two of the three reanalyses in the historical period and by at least four out of seven reanalyses in the comparison period around the same date (usually within one or two days). The central date of these common events is computed as the median of the central dates from the SSWs detected for each reanalysis. Thus, with this approach, the same events and central dates apply for all reanalyses even if the reversal of the winds does not occur in all of them. This is useful to ensure that the differences between datasets are not due to the selection of different events or dates. The 120 common SSWs are listed in Table 1 for the comparison period. Nevertheless, in the very first part of our study, we have addressed the opposite question and quantified the possible discrepancies in the frequency of SSWs among reanalyses when the same criterion is applied to all datasets. In that case, we have imposed the WMO definition for the identification of SSWs in each reanalysis. The definition is based on the reversal, within 5 days) of zonal-mean zonal wind at 10hPa and 60ºN and zonal-mean temperature difference between 90ºN and 60ºN 125 at the same level (Labitzke, 1981). 2.3 Types of SSWs SSWs are classified following two definitions: D vs S SSWs, and WN1 vs WN2 events. In this study, D and S SSWs were identified according to the algorithm by K. Shibata, which is similar to that of Charlton and Polvani (2007). It is based on the 5 identification of cyclonic vortices and their relative sizes by means of the non-zonal absolute vorticity at 10hPa from 5 days 130 before to 10 days after (i.e. [-5,10]-day) with respect to the occurrence of an SSW, according to the definition of Section 2.2. More specifically, S SSWs are identified when two local maxima of the absolute vorticity are located diametrically opposed and the size ratio of the sectors around those maxima is larger than 0.5 during at least one of the 16-day period surrounding the SSW. Otherwise the SSW is defined as D. The events were classified individually in each reanalysis. The classification into S/D events of common SSWs in the comparison period (used in Sections 4 and 5) was based on the predominant type of 135 each single event across the different reanalyses, following a similar procedure to that employed for the identification of the common dates (Table 1). WN1 and WN2 SSWs were selected by applying a zonal Fourier decomposition of the daily 50hPa geopotential height data at 60ºN into WN1 (Z1) and WN2 (Z2) amplitudes for the [-10,0]-day period before each SSW (Barriopedro and Calvo, 2014). An SSW was defined as a WN2 event if [Z2] ≥ [Z1] (brackets denote the averaged amplitude for the [-10,0]-day period before the 140 SSW) or if Z2 - Z1 ≥ 200 m at least for one day within the [-10,0]-day period before the SSW. Otherwise, the SSW was defined as a WN1 event. See the list of events of each type in Table 1 and Barriopedro and Calvo (2014) for more details on the algorithm. 2.4 Dynamical benchmarks We have applied the following diagnostics proposed by Charlton and Polvani (2007) to evaluate the dynamical signatures 145 associated with the occurrence and development of SSWs: - Amplitude of the SSW in the middle stratosphere (hereafter amp010) computed as the area-weighted mean 10hPa temperature anomaly over the polar cap (50°N-90°N) and averaged for the [-5,5]-day period with respect to the central date of the event. - Amplitude of the SSW in the lower stratosphere (hereafter amp100), defined as amp010 but at 100hPa. It provides a measure of the coupling between the middle and lower stratosphere around the occurrence of SSWs. 150 - Deceleration of the PNJ (hereafter decelu), corresponding to the difference of the 10hPa zonal-mean zonal wind at 60°N between the [-15, -5]-day period prior to the central date and the [0, 5]-day period after the central date. - Wave activity prior to SSW (hereafter actwav), computed as the area-weighted mean 100hPa meridional eddy heat flux (HF) anomaly averaged over 45°N-75°N for the [-20,0]-day period before the occurrence of the event. 2.5 Upward-propagating wave activity 155 The anomalous meridional eddy HF averaged over 45°N-75°N at different pressure levels was used as a metric to measure the upward propagation of wave activity. This latitudinal band corresponds to the climatological area with the strongest vertical wave propagation from the troposphere to the stratosphere (Hu and Tung, 2003). As a second step, the methodology by Nishii et al. (2009) was applied to analyze the role of different forcing processes in the occurrence of SSWs. This methodology is based on the decomposition of daily anomalous eddy HF into two components, 160 6 which correspond to the interaction between climatological waves and anomalous waves (second and third right hand terms of Eq. 1) and the inherent contribution of anomalous waves (first right hand term of Eq. 1): [𝑣∗𝑇∗]𝑎=[𝑣𝑎 ∗𝑇𝑎 ∗]𝑎+[𝑣𝑐 ∗𝑇𝑎 ∗]+[𝑣𝑎 ∗𝑇𝑐 ∗] (1) where brackets and asterisks indicate zonal mean and deviation from it, respectively, v is meridional wind, T is temperature 165 and the a and c subscripts denote daily anomalies and climatological values, respectively. Eq. 1 has been applied to each pressure level. 2.6 Blocking definitions The precursor role of blocking in SSWs has been discussed across studies (see e.g., Castanheira and Barriopedro (2010) for an overview), although there is not a clear consensus on this topic. The divergent results of previous studies may partially be 170 attributed to different methodologies of blocking detection (e.g., Woollings et al., 2008). In this study, three different blocking definitions have been used to address this question. The three methodologies use daily geopotential height at 500 hPa (Z500) and span almost all approaches to blocking definition. The first method is based on the occurrence of regional and persistent meridional Z500 gradient reversals (the absolute method, ABS; e.g., Scherrer et al., 2006). The second metric involves the detection of persistent and quasi-stationary Z500 anomalies, computed with respect to the local climatological field (the 175 anomaly method, ANO; e.g., Sausen et al., 1995). Finally, a combined method of absolute and anomaly Z500 fields (the mixed method, MIX) is used, providing a two-fold perspective of blocking (Barriopedro et al., 2010). Several criteria are imposed to ensure that the detected episodes represent large-scale, quasi-stationary, and persistent high-pressure systems. See Woollings et al. (2018) for more details about blocking definitions. 3 Main SSW characteristics 180 In this section, the main signatures of SSWs (frequency, type of events and process-based diagnostics) are analyzed for each period and compared among the different datasets. 3.1 Frequency, seasonality and type of events First, we have analyzed the results for the frequency and type of events when the same criterion is applied to each dataset. Table 2 shows the mean frequency of events and the ratio of D to S SSWs for each period and reanalysis. The main differences 185 are found in the historical period when the reanalyses show a large spread in both frequency and type of events. In particular, the NCEP/NCAR reanalysis displays the results that deviate the most from the other two datasets, although the differences are not statistically significant at the 95% confidence level (Student’s t-test). The short period of analysis and hence the reduced sample might explain part of these discrepancies. More importantly the unavailability of satellite data in the pre-satellite era 7 leads to a strong dependency of the reanalysis data in the stratosphere on the characteristics of each reanalysis model. Note 190 that NCEP/NCAR reanalysis is the only reanalysis with a low-top model and a lid in the stratosphere (3hPa), whereas JRA-55 and ERA-40 have the top in the mesosphere (0.1hPa). The low top typically dampens variability close to the top and so, reduces the probability of the occurrence of an SSW (Charlton-Pérez et al., 2013). In fact, the standard deviation of daily polar temperature and zonal wind at 10 hPa in December and January of the historical period is much lower in NCEP/NCAR than in the other two reanalyses, although the differences are not statistically significant at the 95% confidence level (F-Fisher test) 195 (Figure 1a, c). In contrast, at lower levels, we do not find such discrepancies (see 100 hPa temperature in Fig. 1b, d), supporting that the occurrence of SSWs during this period is strongly influenced by the model performance and hence should be considered reanalysis-dependent. Conversely, in the comparison period, there is a good agreement in both the frequency and ratio of D/S SSWs. Small differences are found, particularly, in the D/S ratio, but this might be due to the specific thresholds or other methodological 200 issues of the applied criterion, since such deviation does not appear when classifying SSWs into WN1 and WN2 events (Barriopedro and Calvo, 2014). More details about these classifications of SSWs can be found in the Chapter 6 of S-RIP. Regarding SSWs seasonality, Figure 2 shows the smoothed seasonal distribution of SSW per decade. This distribution has been computed by counting the number of SSWs within the ±10-day periods centered on each winter days. Additionally, the distribution has been smoothed with a 10-day running mean. Similar to the winter mean frequency of SSWs, historical 205 reanalyses show the largest spread in the seasonal distribution. A substantial part of this spread is due to the NCEP/NCAR reanalysis whose distribution is statistically significantly different from that of the other two reanalyses at a 99% confidence level (two-samples Kolmogorov-Smirnov test). In contrast, ERA40 and JRA-55 distributions display similar (statistically undistinguishable) distributions. In particular, they show an increasing SSW occurrence from early winter that maximizes in January and decreases by late winter (Fig. 2a), in agreement with the temporal evolution of the standard deviation of the zonal210 mean zonal wind at 60ºN and 10 hPa in the historical period (Fig. 1c). In contrast, SSWs for NCEP/NCAR are more uniformly distributed with three sharp maxima in early, mid and late winter. The early winter peak of SSWs in NCEP/NCAR agrees well with the climatological polar stratospheric state, which shows a weaker PNJ and a warmer polar stratosphere than the other two reanalyses (Fig. 1a and c). These NCEP/NCAR differences in the PNJ are only statistically significant for the polar stratospheric temperature and ERA-40 though, likely due to the short sample and the general large interannual variability of 215 the winter polar stratosphere. However, they agree with an artificial positive temperature trend of 8ºC at 10 hPa for 1948-1998 in the NCEP/NCAR reanalysis, as documented by Badin and Domeisen (2014). On the other hand, the lower wind variability in January in NCEP/NCAR would agree with the reduced frequency of SSWs in that month and reanalysis, as compared to the other datasets. In the comparison period the results are similar across reanalyses, which show statistically indistinguishable distributions (Kolmogorov-Smirnov test, Fig. 2b). In this period, the maximum occurrence shifts to late winter in all datasets 220 compared to the distributions of ERA-40 and JRA-55 in the historical period. Similar differences in the intra-seasonal distribution of events were already documented by Gómez-Escolar et al. (2012) between the preand post1979 periods. Despite the large uncertainty of the earlier period, their distributions are statistically significantly different at the 99% 8 confidence level and this result supports the hypothesis of multi-decadal variations in the intra-seasonal occurrence of SSWs, which adds to the reported variability in the total winter frequency of SSWs (Schimanke et al. 2011; Reichler et al. 2012; 225 Domeisen 2019). 3.2 Process-based diagnostics The processes involved in the occurrence of SSWs have been compared across reanalyses by using the diagnostics defined in Section 2d. In this case, and in the rest of the paper, we have used the common dates of SSWs to make sure the differences found across reanalyses are not due to the inclusion of different events. 230 Figure 3 shows the statistics (mean, median and interquartile range) of the dynamical benchmarks for all reanalyses in the two periods. A quick comparison of the MRM of these benchmarks for both periods reveals that SSWs are preceded by a similar anomalous strengthening of wave activity at 100hPa, are associated with a comparable deceleration of the PNJ and have a similar amplitude in the middle and lower stratosphere in both periods. Only slight differences are found in the median of decelu and amp100 (compare Fig. 3b,c with Fig.3f,g). However, given that the median and mean of these magnitudes for one 235 period are included within the interquartile range of the other, we can conclude that SSWs characteristics are similar in both periods of study. The comparison period shows good agreement among all reanalyses as all datasets are characterized by similar median, mean and spread values (Fig. 3e-h). Nevertheless, slight deviations can be found for NCEP/NCAR in the distribution of decelu, which is shifted towards lower values and shows a reduced spread among events, as compared to the rest of the datasets (Fig. 240 3g). These deficiencies are even clearer in the historical period, when a similar discrepancy is detected in amp010 (Fig. 3a), consistent with the reduced strength and variability of the PNJ in NCEP/NCAR reanalysis (Figs. 1c). As the deviation of decelu in the NCEP-NCAR reanalysis is common for both periods, this might point to a bias of the model, whose effects are amplified in the first period by the lower amount of assimilated data. As mentioned before, this bias is very likely linked to the low top of the model and the low vertical resolution in the stratosphere, provided that the SSW characteristics at lower levels (i.e. 245 amp100, actwav) do not differ much from those of other reanalyses. Note that these differences are still noticeable in NCEPDOE, in agreement with Long et al. (2017) that identified similar biases in the climatology and interannual variability of temperature and zonal winds for both NCEP reanalyses. The model of NCEP-DOE is basically the same as that of NCEP/NCAR reanalysis although with an updated version (1995 vs 1998) (Fujiwara et al. 2017). This implies that both reanalyses use a model with a low resolution in the stratosphere and with assimilated temperature data instead of direct 250 radiances that reduce their ability to represent the stratosphere (Fujiwara et al. 2017). Despite their similarities, the NCEPDOE performs better with respect to the MRM, particularly for decelu, arguably due to improvements introduced in the updated version of the reanalysis model. Primarily, NCEP-DOE was run with a new ozone climatology (Kanamitsu et al., 2002). Other differences in the concentration of CO2 or the radiation scheme between both reanalyses might also explain the differences between both NCEP reanalyses (Fujiwara et al., 2017). 255 9 A similar analysis has been carried out separately for WN1 and WN2 SSWs in the comparison period (Fig. S1). All datasets reproduce a similar behavior for both types of events and all diagnostics, with the exception of the associated deceleration of the PNJ in the middle stratosphere: WN2 SSWs are related to larger decelerations of the PNJ, probably because they are usually preceded by a stronger polar vortex than WN1 SSWs (Albers and Birner, 2014; Díaz-Durán et al., 2017). These results also confirm the overall good agreement across reanalyses except for the deficiency of NCEP/NCAR concerning decelu. 260 Unfortunately, these findings cannot be confirmed in the historical reanalyses due to the very low frequency of WN2 events in that period (not shown). 4 Dynamical forcing 4.1 Upward-propagating wave activity Figures 4 and 5 show the composited anomalous eddy HF, area-averaged between 45° N and 75°N, at different levels around 265 the SSWs onset date for the historical and comparison period, respectively. Only results from 300 to 10 hPa are presented, as the [300-100] hPa layer corresponds to the communication region for the stratosphere-troposphere coupling (de la Cámara et al. 2017), and the levels above this layer typically show the strongest HF anomalies. The MRM shows a strong anomalous peak of HF around the central date of SSWs in both periods. This strong peak is preceded by a weak pulse around [-20, -15] days in the middle stratosphere in the comparison period but not in the historical one. The largest differences across reanalyses 270 are detected in the middle stratosphere in agreement with Martineau et al. (2018), and they are more pronounced for the historical than for the comparison period. By applying the methodology by Nishii et al. (2009) we have analyzed the contributing role of the different HF terms to the occurrence of SSWs. The MRM decomposition of the HF in the comparison period shows that the strongest peak ([-5,0]-day interval) is mainly due to the action of anomalous waves (first right hand term of Eq. 1), albeit with a relevant contribution of 275 the constructive interaction between climatological and anomalous waves (second and third right hand terms of Eq. 1, Figs. 4c, e and 5c, e). Conversely, the preceding weaker pulses of the comparison period seem to be more dominated by the interaction term. The agreement among reanalyses concerning the relative roles of these terms is higher for the comparison period, mainly in the middle stratosphere, than for the historical period (compare Fig. 4d, f vs Fig. 5d, f). Given the documented differences in the dynamical forcing of different types of SSWs (e.g. Smith and Kushner, 2012; 280 Barriopedro and Calvo, 2014), we have repeated the analysis separately for WN1 and WN2 SSWs (Fig. 6). It has only been done for the comparison period, due to the low sample size of WN2 events for the historical one. Although there is not a univocal relationship between D and S SSWs and WN1 and WN2 events (Waugh, 1997), our results for WN1 and WN2 events agree well with those of Smith and Kushner (2012) for D and S SSWs. WN1 events are mainly triggered by persistent but moderately intense anomalies of HF during different periods ([-20, -15] and [-10, 0] days), which are associated with the 285 constructive interference of climatological and anomalous waves (Figs. 6e and i). In contrast, WN2 events are related to intense but short pulses of eddy HF in the five days prior to the central date. These pulses are predominantly due to the anomalous 16 Cohen, J. and Jones, J.: Tropospheric precursors and stratospheric warmings, J. Climate, 24, 6562-6572, 2011. Davini, P., Cagnazzo, C., Gualdi, S., and Navarra, A.: Bidimensional diagnostics, variability, and trends of Northern Hemisphere blocking, J. Climate, 25, 6496-6509, 2012. Dee, D. P. and coauthors: The ERA-Interim reanalysis: configuration and performance of the data assimilation system, Q. J. R. Meteorol. Soc., 137, 553-597, 2011. 480 De la Cámara, A., Albers, J. R., Birner, T., García, R. R., Hitchcock, P., Kinnison, D. E. and Smith, A. K.: Sensitivity of sudden stratospheric warmings to previous stratospheric conditions. J. Atmos. Sci., 74, 2857-2877, 2017. Díaz-Durán, A., Serrano, E., Ayarzagüena, B., Ábalos, M., and de la Cámara, A.: Intra-seasonal variability of extreme boreal stratospheric polar vortex events and their precursors, Clim. Dyn., 49, 3473–349, doi: 10.1007/s00382-017-3524-1, 2017. Domeisen, D. I. V.: Estimating the frequency of sudden stratospheric warming events from surface observations of the North 485 Atlantic Oscillation, J. Geophy. Res., 124, 3180-3194. Fujiwara, M. and coauthors: Introduction to the SPARC Reanalysis Intercomparison Project (S-RIP) and overview of the reanalysis systems, Atmos. Chem. Phys., 17, 1417–1452, 2017. Garfinkel, C. I. and Hartmann, D. L.: Different ENSO teleconnections and their effects on the stratospheric polar vortex, J. Geophys. Res., 113, D18114, doi: 10.1029/2008JD009920, 2008. 490 Garfinkel, C. I., Hartmann, D. L., and Sassi, F.: Tropospheric precursors of anomalous Northern Hemisphere stratospheric polar vortices, J. Climate, 23, 3282-3299, 2010. Garfinkel, C. I., Butler, A. H., Waugh, D. W., Hurwitz, M. M., and Polvani, L. M.: Why might stratospheric sudden warmings occur with similar frequency in El Niño and La Niña winters? J. Geophys. Res., 117, F19106, doi:10.1029/2012JD017777, 2012. 495 Gerber, E. P., Orbe, C., and Polvani, L. M.: Stratospheric influence on the tropospheric circulation revealed by idealized ensemble forecasts. Geophys. Res. Lett., 36, L24801, https://doi.org/10.1029/2009GL04091, 2009. Gómez-Escolar, M., Fueglistaler, S., Calvo, N. and Barriopedro, D.: Changes in polar stratospheric temperature climatology in relation to stratospheric sudden warming occurrence, Geophys. Res. Lett., 39, doi:10.1029/2012GL053632, 2012. Hitchcock, P.: On the value of reanalyses prior to 1979 for dynamical studies. Atmos. Chem. Phys. 19, 2749-2764, 500 https://doi.org/10.5194/acp-2018-879, 2019. Hu, Y. and Tung, K. K.: Possible ozone-induced long-term changes in planetary wave activity in late winter, J. Climate, 16, 3027-3038, 2003. Kalnay, E. and coauthors: The NCEP/NCAR40-year reanalysis project, Bull. Amer. Meteor. Soc., 77, 437-471, 1996. Kanamitsu, M. and coauthors: NCEP-DOE AMIP II reanalysis (R-2), Bull. Amer. Meteor. Soc., 83, 1631-1643, 2002. 505 Kim, J., Son, S.-W., Gerber, E., and Park, H. –S.: Defining sudden stratospheric warming in climate models: Accounting for biases in model climatologies, J. Climate, 30, 5529-5546, 2017. Kobayashi, S. and coauthors: The JRA-55 reanalysis: General specifications and basic characteristics, J. Meteor. Soc. Japan, 93, 5-48, 2015. 17 Lehtonen, I. and A. Y. Karpechko: Observed and modeled tropospheric cold anomalies associated with sudden stratospheric 510 warmings, J. Geophys. Res., 121, 1591-1610, doi:10.1002/2015JD023860, 2013. Limpasuvan, V., Thompson, D. W. J., and Hartmann, D. L.: The life cycle of the Northern Hemisphere sudden stratospheric warmings, J. Climate, 17, 2584-2596, 2004. Long, C. S., Fujiwara, M., Davis, S., Mitchell D. M., and Wright, C. J.: Climatology and interannual variability of dynamic variables in multiple reanalyses evaluated by the SPARC Reanalysis Intercomparison Project (S-RIP), Atmos. Chem. 515 Phys., 17, 14593–14629, 2017. Martineau, P., Son, S-W, Taguchi, M., and Butler, A. H.: A comparison of the momentum budget in reanalysis datasets during sudden stratospheric warming events, Atmos. Chem. Phys., 18, 7169-7187, 2018. Martius, O., Polvani, L. M., and Davies, H. C.: Blocking precursors to stratospheric sudden warming events. Geophys. Res. Lett., 36, L14806, doi:10.1029/2009GL038776, 2009. 520 Matsuno, T.: A dynamical model of stratospheric sudden warming, J. Atmos. Sci., 28, 1479–1494, 1971. Maycock, A. C. and Hitchcock, P.: Do split and displacement sudden stratospheric warmings have different annular mode signatures? Geophys. Res. Lett., 42, 10943-10951, 2015. McLandress, C., and Shepherd, T. G.: Impact of climate change on stratospheric sudden warmings as simulated by the Canadian Middle Atmosphere Model, J. Climate, 22, 5449-5463, 2009. 525 Mitchell, D. M., Gray, L. J., Anstey, J., Baldwin, M. P., and Charlton-Perez, A. J.: The influence of stratospheric vortex displacements and splits on surface climate, J. Climate, 26, 2668–2682, 2013. Nakagawa, K. I. and Yamazaki, K.: What kind of stratospheric sudden warming propagates to the troposphere? Geophys. Res. Lett, 33, L04801, doi: 10.1029/2005GL024784, 2006. Nishii, K., Nakamura, H., and Miyasaka, T.: Modulations in the planetary wave field induced by upward-propagating Rossby 530 wave packets prior to stratospheric sudden warming events: A case-study, Q. J. R. Meteorol. Soc., 135, 39-52, 2009. Nishii, K., Nakamura, H., and Orsolini, Y. J.: Geographical dependence observed in blocking high influence on the stratospheric variability through enhancement and suppression of upward planetary-wave propagation, J. Climate, 24, 6408-6423, 2011. Onogi, K. and coauthors: The JRA-25 reanalysis, J. Meteor. Soc. Japan, 85, 369-432, 2007. 535 Palmeiro, F. M., Barriopedro, D., García-Herrera R., and Calvo, N.: Comparing sudden stratospheric warming definitions in reanalysis data, J. Climate, 28, 6823-6840, 2015. Plumb, R. A.: Instability of the distorted polar night vortex: A theory of stratospheric warmings, J. Atmos. Sci., 38, 2514– 2531, 1981. Reichler, T., Kim, J., Manzini, E., and Kröger, J.: A stratospheric connection to Atlantic climate variability, Nat. Geosci., 5, 540 783-787, 2012. Rex, D. F.: Blocking action in the middle troposphere and its effect upon regional climate. Tellus, 2, 196–211, doi:10.1111/ j.2153-3490.1950.tb00331.x, 1950. 18 Rienecker, M. M. and coauthors: MERRA: NASA’s Modern-Era Retrospective Analysis for Research and Applications, J. Climate, 24, 3624-3648, 2011. 545 Saha, S. and coauthors: The NCEP climate forecast system reanalysis, Bull. Amer. Meteor. Soc., 91, 1015-1057, 2010. Sausen, R., König, W., and Sielmann, F.: Analysis of blocking events from observations and ECHAM model simulations, Tellus, 47A, 421–438, 1995. Scherrer, S. C., Croci-Maspoli, M., Schwierz, C., and Appenzeller, C.: Two-dimensional indices of atmospheric blocking and their statistical relationship with winter climate patterns in the Euro-Atlantic region, Int. J. Climatol., 26, 233-249, 2006. 550 Schimanke, S., Körper, J., Spangehl, T., and Cubasch, U.: Multi‐decadal variability of sudden stratospheric warmings in an AOGCM, Geophys. Res. Lett., 38, L01801, oi:10.1029/2010GL045756, 2011. Seviour, W. J. M., Mitchell, D. M., and Gray, L. J.: A practical method to identify displaced and split stratospheric polar vortex events, Geophys. Res. Lett., 40, 5268–5273, doi:10.1002/grl.50927, 2013. Sjoberg, J. P., and Birner, T.: Stratospheric wave-mean flow feedbacks and sudden stratospheric warmings in a simple model 555 forced by upward wave activity flux. J. Atmos. Sci., 71, 4055-4071, 2014. Smith, K. L. and Kushner, P. J.: Linear interference and the initiation of extratropical stratosphere–troposphere interactions. J. Geophys. Res., 117, D13107, doi: 10.1029/2012JD017587, 2012. Taguchi, M. and Hartmann, D. L.: Increased occurrence of stratospheric sudden warmings during El Nio as simulated by WACCM, J. Climate, 19, 324-332, 2006. 560 Taguchi, M.: Is there a statistical connection between stratospheric sudden warming and tropospheric blocking events? J. Atmos. Sci., 65, 1442-1454, 2008. Treidl, R. A., Birch, E.C., and Sajecki, P.: Blocking action in the Northern Hemisphere: A Climatological study, AtmosphereOcean, 19, 1-23, 1981. Tripathi, O. P. and coauthors: Examining the predictability of the stratospheric sudden warming of January 2013 using multiple 565 NWP systems, Mon. Wea. Rev., 144, 1935–1960, 2016. Uppala S. M. and coauthors: The ERA-40 re-analysis, Q. J. R. Meteorol. Soc., 131, 2961-3012, 2005. Waugh, D. W.: Elliptical diagnostics of stratospheric polar vortices, Q. J. R. Meteorol. Soc., 123, 1725-1748, 1997. White, I., Garfinkel, C. I., Gerber, E. P., Jucker, M., Aquila, V., Oman, L. D.: The downward influence of sudden stratospheric warmings: Association with tropospheric precursors, J. Climate, 32, 85-108, 2019. 570 Wiedenmann, J. M., Lupo, A. R., Mokhov, I. I., and Tikhonova, E. A.: The climatology of blocking anticyclones for the Northern and Southern Hemispheres: Block intensity as a diagnostic, J. Climate, 15, 3459-3473, 2002. Woollings, T.J., Hoskins, B. J., Blackburn, M., and Berrisford, P.: A new Rossby wave-breaking interpretation of the North Atlantic Oscillation. J. Atmos. Sci., 65, 609–626, 2008. Woollings, T., Charlton-Perez, A., Ineson, S., Marshall, A. G., and Masato, G.: Associations between stratospheric variability 575 and tropospheric blocking, J. Geophys. Res., 115, D06108, doi:10.1029/2009JD012742, 2010. 19 Woollings, T., Barriopedro, D., Methven, J., Son, S-W., Martius, O., Harvey, B., Sillmann, J., Lupo, A. R., and Seneviratne, S.: Blocking and its response to climate change. Curr. Clim. Change Rep., 4, 287-300, 2018. 580 20 Table 1: Classification of the common SSWs into WN1 and WN2 events in the comparison period. (In brackets the S/D classification). WN1 SSWs WN2 SSWs 29 02 1980 (D) 11 02 2001 (D) 04 03 1981 (D) 31 12 2001 (D) 04 12 1981 (D) 18 01 2003 (S) 24 02 1984 (D) 05 01 2004 (D) 23 01 1987 (D) 21 01 2006 (D) 08 12 1987 (S) 24 02 2007 (D) 14 03 1988 (S) 09 02 2010 (S) 15 12 1998 (S) 24 03 2010 (D) 26 02 1999 (S) 22 02 1979 (S) 01 01 1985 (S) 21 02 1989 (S) 20 03 2000 (D) 22 02 2008 (D) 24 01 2009 (S) Table 2: Frequency of SSWs per decade and ratio of vortex displacement (D) vs vortex split (S) SSWs for each reanalysis and period of study. 585 Historical period (1958-1978) Comparison period (1979-2012) Reanalyses Frequency (SSWs/dec) Ratio D/S Frequency (SSWs/dec) Ratio D/S ERA-40 6.2 1.6 NCEP-NCAR 4.8 0.7 6.2 1.6 JRA-55 5.7 1.0 6.8 1.2 ERA-Interim 6.2 1.6 JRA-25 6.5 1.8 NCEP-CFSR 6.5 1.4 NCEP-DOE 6.5 1.4 NASA-MERRA 6.5 1.2 21 590 Figure 1: 21-day running mean of the daily climatology (solid line) and standard deviation (dashed line) in the historical period (1958-1978) of: (a) polar-cap (50ºN -90ºN) averaged temperature at 10 hPa, (b) polar-cap (50ºN -90ºN) averaged temperature at 100 hPa, (c) zonal mean zonal wind at 60°N and 10 hPa and (d) heat flux at 100 hPa averaged over 45ºN -75N. The left (right) y-axis 595 refers to the mean (standard deviation) in each plot. Thick lines indicate values of ERA-40 or JRA-55 that are significantly different from those of NCEP-NCAR reanalysis at the 95% confidence level. Magenta crosses correspond to JRA-55 values that are significantly different from ERA-40 ones at the 95% confidence level (Student’s t-test). 600 Nov Dec Jan Feb Mar Apr 210 215 220 225 230 T (K) Polar cap T @ 10hPa 0 2 4 6 8 T (K) Nov Dec Jan Feb Mar Apr 210 215 220 225 230 T (K) Polar cap T @ 100hPa 0 0.75 1.5 2.25 3 T (K) Nov Dec Jan Feb Mar Apr -10 0 10 20 30 40 u (m s-1) Zonal mean u @ 60N & 10hPa 0 5 10 15 20 25 u (m s-1) Nov Dec Jan Feb Mar Apr 0 5 10 15 20 Heat flux (K m s-1) Heat flux @ 100hPa (45-75N) ERA-40 JRA-55 NCEP-NCAR std. ERA-40 std. JRA-55 std. NCEP-NCAR 0 2.5 5 7.5 10 Heat flux (K m s-1) a c b d 22 Figure 2. SSW total frequency distribution within ±10 day periods from the date displayed in the x-axis for: (a) the historical period (1958-1978) and (b) the comparison period (1979-2012). 23 605 Figure 3. Box plots showing the distribution of the dynamical benchmarks of SSWs (amp010, amp100, decelu and actwav) in the historical (1958-1978) and comparison (1979-2012) periods. The interquartile range is represented by the size of the box and the red line (black cross) corresponds to the median (mean). Whiskers indicate the maximum and minimum points in the distribution that are not outliers. Outliers (red crosses) are defined as points with values greater than 3/2 times the interquartile range from the ends of the box. See text for the definition of dynamical benchmarks. 610 24 Figure 4. (a) Composited time evolution of the total anomalous heat flux averaged over 45°N-75°N (K m s-1) at different pressure levels from 29 days before to 30 days after the occurrence of SSWs in the historical (1958-1978) period. Contour interval is 20 K m s-1. (b) Same as (a) but for the standard deviation of the reanalyses with respect to the MRM divided by the square root of the number 615 of reanalyses. Contour interval is 1 K m s-1. (c) and (d) Same as (a) and (b) but for the interaction between climatological and anomalous waves. Contour intervals are 10 K m s-1 and 2 K m s-1, respectively. (e) and (f) Same as (a) and (b) but for the contribution of the anomalous waves to the total anomalous heat flux. Contour intervals are 10 K m s-1 and 2 K m s-1, respectively. Shading in (a), (c) and (e) denotes statistically significant anomalies at the 95% confidence level of the same sign in at least 66.7% of all reanalyses (Monte-Carlo test). 620 25 Figure 5. Same as Fig. 4 but for the comparison (1979-2012) period.