The role of cyclones and PV cutoffs for the occurrence of unusually long wet spells in Europe

. The synoptic dynamics leading to the longest wet spells in Europe are so far poorly investigated, despite these events’ potentially large societal impacts. Here we examine the role of cyclones and potential vorticity (PV) cutoffs for unusually long wet spells in Europe, defined as the 20 longest uninterrupted periods with at least 5 mm daily accumulated 10 precipitation at each ERA-Interim grid point in Europe (this set of spells is hereafter referred to as 𝓢 !" ). The 𝓢 !" occur predominantly in summer over the eastern continent, in winter over the North Atlantic, in winter or fall over the Atlantic coast, and in fall over the Mediterranean and European inland seas. Four case studies reveal distinct archetypal synoptic storylines for long wet spells: (a) A seven-day wet spell near Moscow, Russia, is associated with a single slow-moving cutoff-cyclone couple; (b) a 15-day wet spell in Norway features a total of nine rapidly passing extratropical cyclones and illustrates serial 15 cyclone clustering as a second storyline; (c) a 12-day wet spell in Tuscany, Italy, is associated with a single but very large cutoff-complex, which is replenished multiple times by a sequence of recurrent anticyclonic wave breaking events over the North Atlantic and western Europe; and (d) a 17-day wet spell in the Balkans features intermittent periods of diurnal convection in an environment of weak synoptic forcing and recurrent passages of cutoffs and thus also highlights the role of diurnal convection for long wet spells over land. A systematic analysis of cyclone and cutoff occurrences during the 𝓢 !" across Europe 20 reveals considerable spatial variability in their respective role for the 𝓢 !" . For instance, cyclones are present anywhere between 10% and 90%, and cutoffs between 20% and 70% of the 𝓢 !"


Introduction
The nature of precipitation is episodic.In Europe, precipitation episodes vary strongly regarding peak and mean precipitation rates as well as duration.Unusually large peak precipitation rates, often termed precipitation extremes, have received much attention in the atmospheric dynamics community, and the synoptic-scale dynamical processes leading to precipitation extremes have been studied in detail for precipitation extremes on a wide range of time-scales, from minutes (e.g.Lenderink and Van Meijgaard, 2010) to seasons (e.g., Flaounas et al., 2020).Research on extreme precipitation is often motivated by these events' relevance for a wide range of natural hazards such as landslides and flooding (e.g., Moore et al., 2012;Rössler et al., 2014;de Vries, 2021).However, studies on individual high-impact events often report an extended duration of the event, for instance four consecutive days of heavy rain leading to the record floods in Germany in early summer 2013 (Grams et al., 2014), a sequence of multi-day precipitation episodes causing the devastating Pakistan flood in summer 2010 (Martius et al., 2013), or a temporal clustering of wet periods leading to an emergency spill-over of California's Oroville Dam in February 2017 and mass evacuation of the population living downstream (White et al., 2019;Moore et al., 2020).Clearly, in addition to peak intensities, the temporal characteristics of precipitation episodes are thus highly relevant too.
Previous studies have assessed these characteristics often by focusing on continuous or quasi-continuous episodes of precipitation, so called wet spells (e.g., Berger and Goossens, 1983;Schmidli and Frei, 2005;Tolika and Maheras, 2005;Zolina et al., 2010Zolina et al., , 2013)).For example, Zolina et al. (2013) found that the mean duration of wet spells in Europe over the last 60 years is between 1.5 days over the Ukraine and southern Russia and 5 days over the Scandinavian Atlantic coast.They reported that the longest wet spells last between 3 (eastern Europe) and 12 days (Scandinavia).It should be noted that these durations depend on the specific definition of wet spells.Furthermore, they found an increasing duration of wet spells during the last 60 years, especially over northern and central Europe, and a slightly decreasing duration over southern Europe during the cold seasons.In Scandinavia and eastern Europe, the duration of wet spells decreased significantly during the warmer seasons.Schmidli and Frei (2005) focused on observational data from Switzerland and identified seasonally and regionally varying trends in the duration of the longest wet spells per year.However, despite a considerable body of literature on statistical characteristics of wet spells, the synoptic-scale dynamical mechanisms leading to precipitation episodes with an unusually long duration -typically one to two weeks -have so far not been investigated in detail.
For midlatitude extreme precipitation events on timescales from a few hours to a few days it is well known that a range of weather systems, which often occur jointly, can serve as dynamical precursors.These are extratropical cyclones (Ulbrich et al., 2003;Field and Wood, 2007;Pfahl and Wernli, 2012), fronts (Catto and Pfahl, 2013;Rüdisühli et al., 2020), so-called "warm conveyor belts" (WCBs; Pfahl et al., 2014), long-range horizontal moisture transport (Winschall et al., 2014), often in the form of so-called "atmospheric rivers" (Zhu and Newell, 1998;Ralph et al., 2004;Lavers and Villarini, 2013), upper-level short-wave troughs and cutoffs (Massacand et al., 1998;Martius et al., 2006), and even atmospheric blocks (Sousa et al., 2017; Feldfunktion geändert hat gelöscht: large-scale hat gelöscht: which often occur jointly weather systems that leads to an extreme precipitation event ( de Vries, 2021).For instance, a narrow upper-level trough is often linked to an elongated surface cold front and intense poleward moisture transport, along which frontal-wave cyclogenesis and the subsequent formation of a WCB can occur.All of these weather systems act in concert to produce the heavy precipitation.
The mechanisms and weather systems contributing to the occurrence of unusually long-lasting wet spells are far less clear.
Few studies investigated the causes of long-lasting precipitation episodes and mostly focused on multi-day heavy precipitation events rather than (potentially even much longer) wet spells.Moore et al. (2021) investigated multi-day (i.e., longer than 3 day) episodes of heavy precipitation along the North American west coast and argued that, on a general level, multi-day heavy precipitation events either occur when individual rain producing weather systems stall or when multiple such weather systems occur in a serially clustered manner.
In particular in the North Atlantic region, serial clustering of extratropical cyclones is well documented (Mailier et al., 2006;Pinto et al., 2014;Priestley et al., 2017a, b;Dacre and Pinto, 2020).This phenomenon occurs predominantly at the downstream end of the North Atlantic storm track (Mailier et al., 2006;Dacre and Pinto, 2020) and arises over western Europe preferentially when the North Atlantic jet is extended towards Europe, remains at a similar latitude for a prolonged period (typically more than a week) and thereby steers entire cyclone families (i.e., primary cyclones as well as frontal wave cyclones forming on trailing cold fronts of the primary cyclones) into the same region (Pinto et al., 2014;Priestley et al., 2017a;Dacre and Pinto, 2020).During such clustering periods, the jet is kept in place by momentum fluxes arising from cyclonic and anticyclonic Rossby wave breaking on the poleward and equatorward side of the jet, respectively.The process of serial cyclone clustering is clearly relevant for long precipitation episodes, as both the extremely wet winter 2013/14 in the United Kingdom and a large set of multi-day heavy precipitation in California have been related to serial cyclone clustering (Priestley et al., 2017b;Moore et al., 2021).
The stalling of individual cyclones as well as upper-level flow features as the cause of long-lasting heavy precipitation events has also been documented in multiple cases, e.g., for an event in Spain (Doswell et al., 1998) when a single slow-moving cyclone associated with an upper-level cutoff caused a seven-day heavy precipitation event.Such cutoffs typically form from synoptic-scale Rossby wave breaking, during which upper-level troughs meridionally amplify to an extent where their evolution is no longer governed by linear wave dynamics and isentropic PV contours start to overturn (McIntyre and Palmer, 1983).Thereby, narrow and meridionally elongated filaments of stratospheric air (i.e., air with PV > 2 PVU), so-called PV streamers (Appenzeller and Davies, 1992) form.As the wave breaking process continues, these PV streamers tend to break up into one or several PV cutoffs, which consist of stratospheric air that is cut-off from the main stratospheric air mass on lower isentropes, but remains connected to the stratosphere on higher isentropes (Portmann et al., 2021).Upper-level PV streamers from narrow filaments of stratospheric potential vorticity (PV), socalled PV streamers (Appenzeller and Davies, 1992).
and cutoffs are accompanied by a cyclonic wind field with a far field effect down to the lower troposphere (Hoskins et al., 1985).Below such PV features the static stability is reduced, which can contribute to the occurrence of deep convection and heavy precipitation (Massacand et al., 1998;Romero et al., 1999;Martius et al., 2006;Portmann et al., 2018;Moore et al., 2019).Moreover, cutoffs in a baroclinic zone are associated with quasi-geostrophic forcing for ascent on their downstream side, cloud formation and precipitation, even if they remain stationary.Since some cutoffs are relatively long-lived and stationary, they can play an essential role in the formation of multi-day precipitation extremes (e.g.Grams et al., 2014).Thus, individual unusually stationary PV streamers and cutoffs also need to be considered as potential dynamical precursors of unusually long-lasting wet spells.
Furthermore, some recent studies have highlighted the role of recurrent upper-level dynamics for long-lasting wet periods.For example Lenggenhager et al. (2019) documented a case where recurrent PV streamer formation in association with an atmospheric block induced a prolonged wet period and flooding on the Alpine South side.Along a similar line of arguments Ali et al. (2021) showed that recurrent synoptic-scale Rossby wave packets [i.e., a succession of wave packets that are each in phase such that multiple troughs form repeatedly in the same region, see Röthlisberger et al. (2019) for details] significantly increase the duration of summer wet spells in parts of Central Europe and Iberia.
In summary, previous research focused on the statistics of wet spells or on the dynamics of short-term to multi-day heavy precipitation events, but rarely on the dynamical mechanisms responsible for unusually long-lasting wet spells.However, considering the potentially high societal impact of unusually long wet spells and the reported trend of increasing wet spell duration, it is crucial to improve our understanding of the dynamical processes that lead to these events.Previous studies clearly identified serial clustering of extratropical cyclones and stalling individual or recurrent upper-level cyclonic flow features (e.g., PV cutoffs) as key for multi-day heavy precipitation events.Here we hypothesize that extratropical cyclones and PV cutoffs also play an important role for the quasi-continuous precipitation during unusually long wet spells.Therefore, we pragmatically choose these two weather systems and examine their role for the formation of unusually long wet spells in Europe.Specifically, the purpose of this study is to identify unusually long wet spells in Europe using ERA-Interim reanalysis data (Dee et al., 2011), and to quantify the occurrence of cyclones and cutoffs during these spells.Hereby, we focus on the 20 longest wet spells (as per our definition of wet spells, see below) at each ERA-Interim grid point in Europe and address the following research questions: 1) How do the duration, accumulated precipitation, average precipitation rate and seasonality of the longest wet spells vary across Europe?
2) What synoptic storylines accompany these unusually long wet spells and how are cyclones and cutoffs involved in the generation of these wet spells?
3) How do the roles of cyclones and cutoffs in these synoptic storylines vary across Europe?
hat gelöscht: or 4) Where and how do the characteristics of cyclones and cutoffs during the longest wet spells differ significantly from climatology?
The structure of the paper is as follows.In Section 2 we introduce the data used in this study and elaborate on the statistical analyses performed here.The results of this study are presented in Section 3. We first discuss climatological characteristics of the longest wet spells in Europe (Section 3.1) and then present four case studies of unusually long wet spells at different locations, which each feature a distinct archetypal synoptic storyline (Section 3.2) and finally address research questions 3) and 4), by systematically analysing the occurrence of cyclones and PV cutoffs during the 20 longest spells at each grid point (Section 3.3).The paper ends with a discussion of these results (Section 4) as well as a summary and the conclusions of this study (Section 5).

ERA-Interim
We use data from the European Centre for Medium-Range Weather Forecasts (ECMWF) ERA-Interim re-analysis from 1 January 1979 to 31 December 2018 for identifying wet spells and for the additional weather system investigations.ERA-Interim data has originally been produced with a T255 resolution and is interpolated here to a regular 1° latitude by 1° longitude grid.In ERA-Interim, precipitation is not an assimilated variable but rather stems from short-range model forecasts, with lead times of 6-18 hours, and is thus subject to model limitations and forecast errors.Pfahl and Wernli (2012) showed that indeed the intensity of sub-daily ERA-Interim precipitation extremes is often underestimated compared to satellite observation-based estimates, in particular in the tropics.However, the timing and location of intense precipitation is well represented in ERA-Interim in comparison with satellite observations (Pfahl and Wernli, 2012).

Definition of the longest wet spells
Here we focus on episodes with uninterrupted and significant, but not necessarily extreme precipitation.We therefore define a wet spell  as a sequence of consecutive days, each with at least 5 mm accumulated precipitation (which is the sum of largescale and convective precipitation).At each grid point with coordinates (, ) the 20 longest wet spells are considered, and this set of spells is referred to as  !", (, ), where the -th longest spell (i.e., an individual event, with a duration, starting date, accumulated precipitation value etc.) is denoted as  $ (, ).The coordinate specification (, ) is omitted wherever possible without loss of clarity.Note that alternative definitions of wet spells, for example based on a different daily precipitation threshold or by allowing short gaps between precipitation episodes, would yield distinct sets of events for the top 20 longest wet spells per grid points.Thus, the identification of "the longest wet spells" is to some degree subjective, but the  !" as defined here certainly classify as unusually long wet spells, which, due to the relatively high threshold for daily precipitation, also have the potential to lead to societal impacts.Furthermore, we tested the sensitivity of our results to different sample sizes (top 5 and 10 longest wet spells) and found no qualitative differences, but for these smaller samples the results were statistically less robust.

Identification of cyclones and PV cutoffs
To identify cyclones the identification scheme of Wernli and Schwierz (2006)  (2017) for details of the tracking] and only cyclones with a lifetime of at least 24 hours are considered in this study.
PV cutoffs are identified and tracked according to the method of Portmann et al. (2021).This method first identifies PV cutoffs on individual isentropic surfaces between 275-360 K with 5 K intervals as isolated regions with PV values above 2 PVU following Wernli and Sprenger (2007).Then, PV cutoffs are defined as 3-dimensional objects that are vertically connected to the stratosphere.This largely removes high PV features of tropospheric origin that arise due to e.g.surface friction or diabatic heating.PV cutoffs are tracked in time based on Lagrangian air parcel trajectories and only cutoffs with a lifetime of at least 24 hours are retained [for details of the identification and tracking see Portmann et al. (2021)].A key advantage of this approach compared to earlier PV cutoff identification and tracking routines is that it is independent of the selection of single vertical levels.For subsequent analyses we use the projection of the 3-dimensional cutoff objects onto the Earth's surface, which, as for cyclones, yields binary cutoff fields with individual cutoff objects.

Additional synoptic variables
In the synoptic discussion of example cases we furthermore consider the integrated vapor transport vector (IVT) and its magnitude (IVT, in units of kg m -1 s -1 ) which are defined as (e.g., Newell et al., 1992;de Vries, 2021) and IVT = ||, respectively.Here,  is the gravitational acceleration,  is specific humidity and  is the horizontal wind vector.Moreover, to quantify the effect of upper-level flow features (e.g., cutoffs) on vertical motion we examine the quasigeostrophic w forcing (in hPa hour -1 ), computed as in Graf et al. (2017) and Besson et al. (2021).Specifically, we use exactly the same data as Besson et al., (2021) and examine the quasi-geostrophic w at 500 hPa, forced from the atmospheric layers above 550 hPa and hereafter refer to this quantity simply as QGw (see Besson et al., (2021) and references therein for further details on the QGw data and its computation).To quantify the characteristics of cyclones/cutoffs occurring during the  !" (, ) we consider for each grid point (, ) cyclones and cutoffs whose masks overlapped with a circle of radius  around (, ) during at least one 6-hourly time step of the  !" (, ).For such cyclones and cutoffs we use the terminology "occurring at (, )" hereafter for simplicity.
We compute four quantities for these cyclones and cutoffs: 1) The "cyclone/cutoff fraction", which corresponds to the fraction of all  !" time steps with cyclones/cutoffs occurring at (, ).This quantity is hereafter referred to as  .(,  ) , whereby  corresponds to the respective feature (cyclones or cutoffs).
We tested values of  of 200 km, 400 km and 600 km and found little qualitative differences for any of these  values.For brevity, only the results for  = 400 km are discussed in this manuscript, while those for  = 200 km and  = 600 km are shown in Figs.S6-9.Note that in the computation of these metrices we consider cyclones and cutoffs also if they overlap with the respective circle only during a short period, e.g., a single time step.Thus, in particular the residence time should be interpreted with such situations in mind.Furthermore, metrices  and  allow for distinguishing between recurrent features and stalling features in the following way: Recurrent cyclones or cutoffs manifest themselves with a short cyclone/cutoff period (), while long cyclone/cutoff residence times () are expected for stalling cyclones and cutoffs.
To identify locally anomalous behaviour of cyclones and cutoffs during the respective  !" we perform a Monte Carlo simulation in which we test the null-hypothesis that the four quantities and the occurrence of the  !" are unrelated, and thus the observed  ,  ,  and  correspond to respective climatological values plus some random noise.The Monte Carlo approach is analogous to that in Röthlisberger et al. (2016) and detailed here exemplarily for the cyclone fraction,  /0/12$3 (, ), but is applied in exactly the same fashion for all four metrices and both weather systems.First, the occurrence of the spells is randomized by shuffling the years of the  !"( ,  ) , while retaining the actual calendar days of these spells.
Second, using these randomized spell dates, a randomized cyclone fraction  /0/12$3 4 (, ) is computed exactly as the true cyclone fraction  /0/12$3 (, ).Third, this process is repeated 1000 times, and the resulting distribution of  /0/12$3 4 values is the distribution of  /0/12$3 (, ) under the null-hypothesis that  /0/12$3 (, ) and the occurrence  !"( ,  ) are which at least one of the  !" (, ) occurred.Then, we identify all cyclones/cutoffs that occurred at (, ) during these days of the year in any year from 1979 to 2018.From this set of cyclones/cutoffs we then compute "climatological values" for the four metrices, in exactly the same fashion as described above.This procedure implies in 265 particular that the days of the year which are considered for constructing the climatological values differs for each grid point, which is necessary because the seasonality of the  !" (i.e., the days of the year on which the  !" occur) also differs from grid point to grid point.Based on these climatological values we compute

Climatological characteristics of the longest wet spells
We begin by discussing basic characteristics of the  !" and their geographical variations.In many regions of Europe, the 285 median duration of the  !" is on the order of four to seven days (Fig. 1a).The  !" are longest predominantly along the Atlantic coastal regions as well as in areas of elevated topography (see also Fig. S1f), and shortest over the Barents, Black and Baltic Seas.Median durations range from less than 4 days over parts of northern Scandinavia, Crimea, Germany, and Poland to more than two weeks in western Norway (Fig 1a).The overall longest European ERA-Interim wet spell occurred at 6°E/62°N, about 300 km north of Bergen, Norway (not shown).It started on 9 September 2018, lasted for 28 days, and finally ended on 6 290 October 2018.The accumulated precipitation during the  !" co-varies with their median duration (Fig. 1a,b), and largest accumulations are found in Norway, north-western Iberia and Scotland (>200 mm in the median).The largest precipitation accumulation during a single wet spell again occurred in western Norway, at 6°E/61°N in January 1989 (not shown).The spell lasted for 24 days, and the accumulated ERA-Interim precipitation amounted to 471 mm.Note that at each grid point the variability in the duration and accumulated precipitation of the  !" is large, and both the duration and accumulated 295 precipitation can differ by more than a factor of two between the longest ( & ) and twentieth longest spell ( !" ) at each grid point (Fig. S1a-d).This is an expected result as the  !" constitute the top few percent of all wet spells with regard to duration at each grid point (Fig. S1e).
The average daily precipitation rate during the  !" exceeds 5 mm day -1 by definition (Fig. 1c), and its spatial variability differs 300 somewhat from those of the median  !" duration and accumulated precipitation.Largest daily precipitation rates during the  !" are again found along the Iberian Atlantic coast and western Norway, with values in excess of 15 mm day -1 , but also in hat gelöscht: The statistical significance of these anomalies is assessed based on a Monte Carlo approach, in close analogy to that of Röthlisberger et al. (2016).The procedure is detailed here 305 exemplarily for the cyclone fraction,  #$#%&'( (, ), but is applied in exactly analogous fashion for all four metrices and both weather features.First, the occurrence of the spells is randomized by shuffling the years of the  !"( ,  ) , while retaining the actual months and days of these spells.Then, using these randomized spell dates, a 310 randomized cyclone fraction is computed exactly as the true cyclone fraction  #$#%&'( (, ).This process is repeated 1000 times and twosided p-values are estimated through comparing the true cyclone fraction with the distribution of randomized cyclone fractions.A pvalue of zero is assigned if the observed cyclone fraction lies outside 315 the range of the 1000 randomized cyclone fractions.Note that the randomization through the shuffling of the years of the  !" (, ) conveniently circumnavigates issues arising from the varying seasonality of the  !" (, ).In this study, anomalies are deemed significant at a grid point wise significance level of 0.01.¶ Mediterranean coastal regions.Average daily precipitation rates during the  !" are locally anomalous, although not overly extreme.In most regions of Europe, they exceed the 95 th percentile of the accumulated daily precipitation during all wet days (defined here as all days of the year with > 0.1 mm accumulated precipitation) but are below the 97.5 th percentile (Fig. 1c), implying that between 1 out of 20 and 1 out of 40 wet days feature comparable precipitation accumulations as the ones observed on average during the  !" .This result underlines that the  !" constitute a set of potentially high-impact precipitation episodes that is different from heavy precipitation events commonly identified based on very high (³99 th ) percentiles of (sub)daily precipitation in previous studies (e.g., Pfahl and Wernli, 2012;Lenggenhager and Martius, 2019;Moore et al., 2021;de Vries, 2021).
The seasonality of the  !" exhibits a clear spatial pattern.Over the North Atlantic and the eastern Mediterranean, most of the  !" occur in winter (December-February, DJF), and likewise in southern Iberia and southwestern France.Over several coastal seas (the north-western Mediterranean, the North Sea, the Gulf of Bothnia and the Atlantic coastal seas near Iberia), the  !" occur predominantly in fall (September-November, SON), while over continental eastern Europe (including Sweden, Finland, and western Russia), the vast majority of the  !" happen during summer (June-August, JJA).At most grid points in central Europe, though, no season dominates the occurrence of the  !" .In a spatially aggregated sense, spring (March-May, MAM) has by far the lowest number of spells as only 12.8% of the  !" spells over land and ocean occurred in MAM, followed by 22.2% in JJA, 29.1% in SON and 35.9% in DJF.When only considering land grid points these contributions change markedly, with 17.6% of land  !" spells occurring in MAM, 46.4% in JJA, 19.0% in SON and 17.0% in DJF.Next, we illustrate the palette of synoptic storylines of unusually long wet spells with four case studies of selected wet spells in different regions and seasons.These cases have been selected from a large number of cases we analysed, due to their archetypal and illustrative nature.

Long wet spell in western Russia -Quasi-stationary cutoff-cyclone couple
At 37°E/55°N, close to Moscow, Russia, the longest wet spell [ & (37°, 55°)] occurred in July 2013 and lasted for 7 days (Fig. 2a).An examination of the large-scale flow pattern reveals a very persistent upper-level PV cutoff with a surface cyclone underneath during the entire spell (Fig. 3).At 12:00 UTC on 20 July, the first day of the spell, the cutoff C1 was located west of Moscow, reaching from 75°N southward to the Black Sea, with a weak surface cyclone L1 to its east (Fig. 3a).Precipitation occurred at the southeastern flank of the cutoff.Closer analysis shows that precipitation was primarily convective at 37°E/55°N (visible, e.g., from the diurnal pattern in precipitation at 37°E/55°N, Fig. 2a), supported by QGw ascent (Fig. S2a considerably between 23 and 25 July, when its local sea level pressure minimum was no longer identified as a cyclone by the cyclone identification scheme (Fig. 3f).However, it re-appeared as an identified cyclone at 12:00 UTC on 26 July (labelled L1* in Fig. 3g,h) and propagated eastward between 26 and 27 July, in tandem with the persisting cutoff C1 (Fig. 3f-h), which ultimately terminated the wet spell at 37°E/55°N on 27 July (Fig. 3h).Throughout the entire period, precipitation during daytime (06:00-18:00 UTC) strongly exceeded precipitation during nighttime (Fig. 2a), suggesting primarily diurnal convective precipitation throughout the entire period.However, upper-level forcing too contributed to precipitation at 37°E/55°N during the first days of the spell (Fig. S2).This case illustrates a first synoptic storyline of unusually long-lasting wet spells, in which a single quasi-stationary upper-level cutoff-surface cyclone couple repeatedly produces substantial precipitation in the same regions, by providing quasi-geostrophic forcing for ascent, destabilization of the troposphere underneath and thereby fosters diurnal convection (Hoskins et al., 1985;Portmann et al., 2018).

Long wet spell in central Norway -sequence of cyclones
A contrasting synoptic storyline is evident for the longest wet spell at 14°E/66°N [ & (14°, 66°)], in Norway (Figs. 2b, 4 and S3).This wet spell occurred between 24 January and 7 February 1989 and thus featured 15 continuous days with more than 5 mm accumulated ERA-Interim precipitation (Fig. 2b).At 12:00 UTC on 24 January 1989 (Fig. 4a), the large-scale flow over the North Atlantic was dominated by a broad trough, with a large surface cyclone L1 with two centers (local SLP minima) north of Iceland and east of the southern tip of Greenland.Over Europe, an amplified ridge R1 was present, but at its northern fringe a first pulse of onshore moisture transport (Fig. S3a) led to precipitation around 14°E/66°N (Fig. 4a), and started the wet spell there.Within the next two days (25-26 January), cyclone L1 moved eastwards, and three new SLP minima developed over the North Atlantic (contained in L2 and L3, Fig. 4b).The grid point at 14°E/66°N was located continuously in westerly flow associated with onshore moisture transport, and precipitation fell in association with two shortwave troughs propagating across Scandinavia in rapid succession (the latter of the two is visible in Fig. 4b as an upper-level PV filament between the British Isles and Scandinavia), which in addition provided upper-level forcing for ascent (Fig. 4b, S3b).Within the next five hat gelöscht: considerable variability in the cyclone and cutoff characteristics across these events, but also some similarities.
hat gelöscht: high PV anomaly hat formatiert: Schriftart: Fett, Kursiv hat gelöscht: ), reminiscent of a cyclonic wave breaking event (Thorncroft et al., 1993)  produced significant precipitation at 14°E/66°N.From 4 February onwards, the upper-level flow was remarkably zonal for three days, leading to continuous onshore moisture transport in particular south of 14°E/66°N, at the southern fringe of cyclones L6-L8, which passed to the north of the considered grid point (Fig. 4f,g).Finally, on 7 February, cyclone L9 developed south of Greenland and followed a more meridional track (Fig. 4g,h).L9 rapidly deepened until 9 February in association with upperlevel cyclonic wave breaking (Fig. 4h).Rapid upper-level ridge-formation (R3) occurred downstream, presumably aided by diabatic processes occurring in L9's strong warm conveyor belt (not shown).The formation of R3 interrupted the predominantly zonal flow and moisture transport over the eastern North Atlantic (Fig. 4h, S3h) and thereby terminated the wet spell.The few PV cutoffs were mostly small and filamentous and their influence on the wet spell thus seems less obvious.In summary, the example of  & (14°, 66°) illustrates a second archetypal synoptic storyline for unusually long wet spells, in which a sequence of cyclones cross the same region in rapid succession.Hereby the moist North Atlantic airmasses impinging on the western Norwegian mountains conceivably generated orographic precipitation and thereby ensured the uninterrupted formation of precipitation, in particular during transition periods between individual cyclones.Moreover, the synoptic configuration of this spell is reminiscent of the North Pacific "zonal jet configuration" of Moore et al., (2021), within which numerous long-lasting heavy precipitation events in Northern California occurred.These authors emphasized the pivotal role of landfalling atmospheric rivers for long-lasting heavy precipitation events occurring in such a flow configuration.In this study we focus on cyclones and PV cutoffs and thus leave exploring the role of atmospheric rivers for the longest European wet spells to future work.Furthermore, it is noteworthy that a total of nine distinct cyclones identified by the Wernli and Schwierz (2006) algorithm were involved in the initiation, continuation and termination of  & (14°, 66°), which is in stark contrast to the synoptic evolution of the previously discussed case near Moscow [ & (37°, 55°)], in which only two objectively identified cyclones and a single cutoff appeared to be relevant.
We next examine the composite structure of the entire set of  !" (14°, 66°) (Fig. 7b).During these events, the composite large-scale upper-level flow was predominantly zonal over the North Atlantic (Fig. 7b), with a negative SLP anomaly north of 66°N, but a positive IVT anomaly (with peak values of roughly 120 kg m -1 s -1 ) extending across the western North Atlantic and directed towards 14°E/66°N.Thirteen of the  !" (14°, 66°) featured five or more distinct cyclones, while the remaining spells featured between two and four distinct cyclones.For the  !" (14°, 66°) we find an  /0/12$3( 14°, 66° ) of 5.15, a  /0/12$3 ( 14°, 66° ) value of 0.51 and  /0/12$3( 14°, 66° ) equals 1.04 days, which underline the contrasting characteristics of cyclones affecting the  !" at 14°E/66°N (numerous, recurrent, fast moving) and at 37°E/55°N (few and stationary), despite comparable cyclone fractions at these two grid points (0.51 vs. 0.64).Based on synoptic analyses of several of the  !" (14°, 66°), PV cutoffs seemed to be less relevant to the  !" (14°, 66°) than cyclones and are not discussed here.2c, 5, and S4].This 12-day wet spell occurred in association with a large cutoff-complex over the Mediterranean that first formed after an anticyclonic wave breaking event over the North Atlantic/Europe (S1 in Fig. 5a) and was then replenished multiple times by a sequence of further wave breaking events occurring in a similar location (Fig. 5a-e).At 12:00 UTC on 1 May 2018 (Fig. 5a), the PV streamer S1 was located over western Europe and substantial precipitation fell to its east, where QG ascent and enhanced northeast moisture transport took place.At its southern fringe, cyclone L1 started tracking northeastward.At the same time, the incipient streamer S2 was already apparent west of the UK (Fig 5a).Within the next two days streamer S1 formed cutoff C1, while S2 developed into a next elongated PV-filament that, on 320 K, reached all the way to Morocco (Fig. 5b).The cutoff C1 and cyclone L1 aligned vertically in an equivalent barotropic manner, and precipitation fell underneath this cutoff-cyclone couple (Fig 5b).Within the next two days, parts of the high-PV air of streamer S2 were absorbed into cutoff C1, which was thereby substantially enlarged (Fig. 5c).At the same time, a strong anticyclone formed over northern Europe and a next wave breaking event (S3) occurred at its downstream flank on 5 and 6 May (Fig. 5c,d).The resulting PV streamer S3 also produced two small cutoffs (the more southerly one is labelled C2 in Fig. 5d), which tracked westward (Fig. 5d,e) and ultimately merged with C1 on 9 May (Fig. 5f).Between 5 and 9 May, a Rex-type blocking pattern (Rex, 1950) was present over Europe and the Mediterranean.The cutoff-complex C1 hereby acted as the positive PV anomaly on the equatorward side of the blocking pattern, covered large parts of the Mediterranean, destabilized the air underneath and led to (primarily daytime, e.g., Fig. 2c) precipitation from Iberia all the way to Turkey (Figs. 5c-f).Thereby, both IVT and QGw were remarkably small around 12°E/43°N between 9 and 12 May (Fig. S4e-g), as the cutoff-complex C1 weakened.
Nevertheless, still sufficient (convective) precipitation fell at 12°E/43°N to prolong the wet spell there for another three days, until it finally ended on 12 May.By 12:00 UTC 13 May, C1 had decayed entirely, although a next cutoff (C3) already approached from the west.Interestingly, only one surface cyclone (L1) was involved in the 12-day spell  & (12°, 43°), and was only present during roughly the first half of the spell.Thus, the destabilizing effect of the cutoff-complex C1, its formation, quasi-stationarity and replenishment due to recurrent wave breaking events, are key to this spell.
The 12-day spell  & (12°, 43°) is perhaps surprising in that it featured only one cyclone, despite its relatively long duration.
However, a small number of cyclones is a common characteristic of the  !" (12°, 43°), as 11 of them featured two or less cyclones.The synoptic analysis of this case suggests that the cutoff-complex C1 can be regarded as an individual, long-lived system, but two cutoffs that merged with the cutoff complex (visible e.g., in Fig. 5d) increased the cutoff count of this spell to and a negative SLP anomaly across the western Mediterranean.However, the  !" (12°, 43°) also contain at least one spell with a strongly differing synoptic storyline: Contrary to  & (12°, 43°), the spell  + (12°, 43°) occurred in winter and featured five distinct cyclones and five distinct cutoffs that were steered towards 12°E/43°N by a southward displaced jet over the eastern North Atlantic (not shown).

Long wet spell in eastern Europe -Intermittent periods of diurnal convection and recurrent cutoff formation
A fourth synoptic storyline is illustrated based on the longest wet spell at 25°E/48°N, in the Carpathian Mountains at the border between Romania and Ukraine (Figs. 2d,6,and S5).This wet spell lasted for an impressive 17 days, from 19 May to 4 June 1988 and featured several days with substantial precipitation without the presence of any cutoff or cyclone near 25°E/48°N.
During intermitted periods of this spell a total of three cutoffs appeared in the vicinity of 25°E/48°N, which, however, did not induce substantial QGw, but conceivably contributed to the persistence of the spell by reducing the static stability underneath.
At 12:00 UTC on 20 May 1988 a breaking wave (S1 in Fig. 6) was present over western Europe, while a surface anticyclone was located over western Russia.In-between, an area of weak SLP gradients extended across much of the Balkans and eastern Europe, and precipitation occurred over widespread areas in this region (Fig. 6a), without any considerable QGw or IVT (Fig. S5a).At 25°E/48°N the ERA-Interim precipitation on the first two days of the spell (19 and 20 May) fell exclusively between 06 and 18 UTC (Fig. 2d), consistent with primarily convective precipitation.Over the course of the next four days, a weak positive PV feature (S1) propagated slowly eastward (not shown) and eventually broke up into several remnants by 12:00 UTC on 24 May, including a cutoff C1 which covered 25°E/48°N at that time step (Fig. 6b).Between 21 May and 24 May, precipitation at 25°E/48°N exhibited less of a diurnal cycle (Fig. 2d), likely due to the influence of the upper-level PV feature which continuously destabilized the troposphere and provided quasi-geostrophic forcing for ascent (not shown).From 24 to 26 May the cutoff C1 hoovered in the vicinity of 25°E/48°N, but gradually weakened until its dissipation on 27 May (Fig. 6bd).At 12:00 UTC on 27 May a flow situation very much reminiscent of that on 12:00 UTC 20 May had established, with a PV streamer S2 approaching 25°E/48°N, and widespread precipitation around 25°E/48°N, but again without the apparent influence of a cyclone or cutoff (Fig. 6d, S5d).The streamer S2 again formed a cutoff (C2), which approached 25°E/48°N by 12:00 UTC on 29 May (Fig. 6e) and dissipated thereafter until 12:00 UTC on 30 May (not shown).It is noteworthy that during the first 13 days of the spell (19-31 May 1988) no surface cyclone is apparent in the immediate vicinity of 25°E/48°.The large-scale flow situation changed considerably between 29 May and 1 June (Fig. 6e,f): a next PV streamer (S3) approached 25°E/48°N, induced upper-level forcing for ascent as well as a pulse of poleward moisture transport on its downstream side (Fig. S5f), and a surface cyclone L1 developed to its east (Fig. 6f).The cyclone L1 deepened from 12:00 UTC on 1 July to 18:00 UTC on 2 June (Fig. 6g), and passed over 25°E/48°N, which led to the largest daily precipitation accumulations during the entire spell (Fig. 2d).By 12:00 UTC on 4 June, the streamer S3 had formed yet another cutoff (C3 in Fig. 6h), which slowly propagated eastward together with L1, and the spell finally ended on 5 June.The synoptic storyline of this spell is interesting in two regards: Firstly, during its first part precipitation fell due to diurnal convection, and the influence of surface cyclones or PV cutoffs appeared to be modest at best.Secondly, during the remainder of the spell it involved multiple cutoffs and PV streamers.Based on this case study, the intermittent occurrence of days with daytime convection in absence of direct upper-level forcing, alternated by days with recurrent wave breaking and cutoff formation thus emerges as a further archetypal storyline for unusually long wet spells.It should be noted that a similar synoptic storyline has recently been reported for a multi-week period of recurrent convective events in central Europe (Mohr et al., 2020).

A systematic analysis of cyclone and PV cutoff characteristics during the 𝓢 𝟐𝟎
We next examine the occurrences of cyclones and cutoffs during the  !" across Europe more systematically to elucidate geographical differences in their role for generating unusually long-lasting wet spells (the left columns in Figs. 8 and 9).
Furthermore, we assess whether or not the behaviour of cyclones and cutoffs during the  !" is locally anomalous (right columns in Figs. 8 and 9).We begin by discussing the cyclone fractions,  /0/12$3 , and their anomalies (Fig. 8a,b).Cyclones occur during more than 60% of the  !" time steps over vast parts of the North Atlantic, Scandinavia as well as the northern and eastern Mediterranean (Fig. 8a).Moreover,  /0/12$3 is generally larger over the ocean than over land, with exceptions over large parts of Scandinavia and the UK.In the Balkans, the Caucasus and parts of the Alps, however, cyclones occur during only roughly 10-30% of the  !" time steps, suggesting that they do not play a major role in the longest wet spells there.The  /0/12$3 anomalies (Fig. 8b) are positive and statistically significant almost everywhere, with largest anomalies in the Mediterranean and the subtropical North Atlantic.Insignificant  /0/12$3 anomalies are found in the aforementioned areas of particularly low  /0/12$3 (which is consistent with the Balkan case study, Section 3.2.4),and, interestingly, also along the west coast of Norway.
The Norway case study (Section 3.2.2) revealed a large number of cyclones contributing to the longest spell at 14°E/66°N, suggesting that serial clustering of extratropical cyclones (Pinto et al., 2014;Priestley et al., 2017a;Dacre and Pinto, 2020) hat gelöscht: the might be crucial for the occurrence of unusually long wet spells in this region.Figure 8e now reveals that in this region the cyclone period,  /0/12$3 , is indeed shorter than anywhere else in Europe, with values between 1.5 and two days.Also, anomalously large  /0/12$3 and anomalously short  /0/12$3 at some grid points around 14°E/66°N somewhat support the hypothesis of anomalous serial cyclone clustering as cause of long wet spells there.However, for numerous grid points along the west coast of Norway, none of the four quantities' anomalies are statistically significant, a result that will be further discussed in Section 4. A much clearer indication for anomalous serial cyclone clustering as a cause of the longest wet spells is found e.g., across central Europe, in southwestern Scandinavia and parts of the United Kingdom (UK), where  /0/12$3 is increased due to significantly more distinct cyclones (positive  /0/12$3 anomalies) occurring at higher rate (reduced  /0/12$3 ), with  /0/12$3 not significantly different from climatology.
A further region with particularly noteworthy cyclone characteristics (and their anomalies) during the  !" are found in the seas south of Italy.This region features some of the largest cyclone residence times ( /0/12$3 in excess of two days, Fig. 8g) and the largest significant anomalies in  /0/12$3 (up to 1 day, Fig. 8h) anywhere in the study region.Notably in the seas south of Italy,  /0/12$3 and  /0/12$3 do not differ significantly from climatology (Fig. 8d,f) despite  /0/12$3 anomalies of up to 40%, which corresponds to roughly a doubling of  /0/12$3 during the  !" compared to climatology (compare Figs. 8a and b).These large  /0/12$3 anomalies thus come about primarily due to increased  /0/12$3 , i.e., anomalously persistent, i.e., slower moving and/or longer-lived cyclones compared to climatology.
Next, we discuss the four quantities and their anomalies for cutoffs (Fig. 9) and contrast them with results for cyclones (Fig. 8).The cutoff fraction,  /562.. , as well as the number of distinct cutoffs per  !" spell,  /562.. , are considerably smaller than  /0/12$3 and  /0/12$3 (compare Fig. 9a,c with Fig. 8a,c).Much fewer of the  /562.. and  /562.. anomalies are statistically significant compared to  /0/12$3 and  /0/12$3 anomalies, but wherever  /562.. and  /562.. anomalies are significant they are also positive, showing that enhanced PV cutoff fractions and numbers are associated with the wet spells (Fig. 9a,c).Largest  (Figs. 9g), suggesting that in these regions, the cutoffs involved in the  !" are more persistent than, e.g., those involved in the  !" along the Norwegian coast or in western Europe (note, however, that the  /562.. anomalies are insignificant at most grid points in these regions).Nevertheless, throughout Europe's land area  /562.. is below 50% almost everywhere (Fig. 9a), indicating that during at least half of the total  !" time no cutoff is present within a 400 km radius.
In summary, this section reveals geographically varying and, in some regions, locally anomalous behaviour of cyclones and cutoffs during the  !" . /0/12$3 is anomalously large during the  !" almost everywhere, however, the causes of these positive cyclone frequency anomalies differ in space.Increased numbers of cyclones during the  !" explain the positive  /0/12$3 values along the north-western Atlantic coast as well as in central Europe, while over the Mediterranean anomalously large residence times of cyclones are the primary reason for positive  /0/12$3 anomalies.The Mediterranean is also the region where  /562.. deviates most from its climatological value, which is caused by anomalously persistent cutoffs during the  !" .
Elsewhere, the characteristics of cutoffs during the  !" vary in space, but rarely differ significantly from climatological cutoff characteristics.

Discussion
The synoptic storylines for unusually long wet spells presented in Section 3.2 feature individual stalling cyclones and cutoffs as well as multiple recurrent such weather systems [as anticipated by Moore et al., (2021)], although some storylines are more complex and also involve daytime convection over complex topography without apparent upper-level forcing.The four case studies were selected due to their archetypal nature, however, manual analyses of a large number of further long wet spells revealed numerous storylines that combined various features of the four archetypal storylines (e.g., multiple stationary cutoffcyclone couples or multiple recurrent cyclones preceding a particularly stationary cutoff, etc.).Furthermore, this manual analysis revealed that distinct wet spells at a single location often do not follow the same synoptic storyline.This diversity raises the question whether or not these synoptic storylines at all stratify according to geographical regions, which motivated our climatological analysis of cutoff and cyclone characteristics during the  !" presented in Figs. 8 and 9.
The two columns in Figs. 8 and 9 address distinct research questions.The left column assesses how cyclone and cutoff characteristics during the  !" vary in space and thus addresses research question 3) in the Introduction.This information is valuable, since the general relevance of cyclones and cutoffs for precipitation is well established (e.g., Hawcroft et al., 2012;Portmann et al., 2020) and thus, e.g., spatial variations in  /0/12$3 point to a spatially varying relevance of cyclones for the  !" .Nevertheless, the spatial variations in these cyclone and cutoff characteristics during the  !" are governed in part by climatological characteristics of cyclones and cutoffs.Therefore, the right column in Figs. 8 and  during the  !" than in climatology.However, not all grid points feature significant anomalies, in particular not for cutoffs.The lack of statistically significant anomalies in any of the four quantities may results from three causes: (a) at some grid points, there is simply no preferred synoptic storyline of the  !" with a clear signature in the four cyclone and cutoff characteristics, , ,  and , which may be the case for example in regions where the  !" occur in different seasons.(b) The sample size (20) is relatively small for a statistical hypothesis test, consequently our Monte Carlo test has only limited power to detect significant departures from climatology (e.g., Wilks, 2011).(c) In certain regions, e.g., the west coast of Norway, the climatological precipitation variability is itself characterized by long wet spells (Zolina et al., 2013).In such regions, unusually long wet spells do not need to be associated with anomalous cyclone and cutoff characteristics, but rather with close to climatological characteristics over a prolonged period.
Nevertheless, for cyclones the anomalies of ,  and  are significant in vast areas of the study region.Comparing the right columns in Figs. 8 and 9 we find overall weaker and much fewer significant anomalies in , ,  and  for cutoffs than for cyclones, which suggests a generally weaker link between anomalous cutoff characteristics and the occurrence of the  !" than for cyclones.In part this could be a consequence of the cutoff definition of Portmann et al., (2021), who defined cutoffs as three dimensional objects, which can persist as vertically shallow objects on relatively high isentropes, where their influence on lower and mid-tropospheric static stability as well as ascent is limited (see with at least 5 mm daily accumulated ERA-Interim precipitation strongly affects which events ultimately end up in the  !" . Nevertheless, choosing a higher precipitation threshold than previous studies (e.g., Ali et al., 2021;Zolina et al., 2010Zolina et al., , 2013) ) ensures that our  !" are unusually long-lasting periods of sustained relatively intense precipitation, and thus potentially highimpact events.Thirdly, our analyses focus primarily on only two types of synoptic systems, cutoffs and cyclones.The choice of these two systems is motivated by previous studies, who documented the particular relevance of these two types for longlasting heavy precipitation episodes (Doswell et al., 1998;Raveh-Rubin and Wernli, 2015;Moore et al., 2021), and our case studies further underline this pivotal role of cutoffs and cyclones for long wet spells.Subsequent research should nevertheless explore to what extent other weather systems such as stagnating atmospheric rivers (e.g., Moore et al., 2021) or blocking (e.g., Mohr et al., 2020) foster the occurrence of unusually long wet spells in their vicinity.

Summary and conclusions
This study investigates the role of cyclones and PV cutoffs for the formation of unusually long wet spells in Europe, which are identified at each ERA-Interim grid point as the 20 longest uninterrupted periods with at least 5 mm daily accumulated ERA-Interim precipitation ( !" ).The  !" are longest along the Norwegian coast and northern Scotland, where the median duration of the  !" reaches up to two weeks.The  !" are shortest e.g., in Poland and north-eastern Scandinavia, where their median duration is only 3-5 days.There is a clear seasonality associated with the occurrence of the  !" : over eastern continental Europe they occur predominantly in summer, while over the North Atlantic most of the  !" occur in winter, and the majority of the  !" along European coastal seas and the Mediterranean occur in winter or fall.In central and western Europe, no season clearly dominates.
Four case studies reveal distinct synoptic storylines of selected long wet spells, that each involve cyclones and/or PV cutoffs in distinct ways.The longest wet spell in Moscow occurred in association with just one cutoff-cyclone couple that formed from a single wave breaking event, subsequently stalled over western Russia and thus produced >5 mm/day precipitation at this grid point for seven consecutive days.An even longer spell (12 days) associated with a single cutoff-complex occurred in Tuscany, Italy.Initially, this cutoff-complex also formed from a wave breaking event but, in contrast to the Moscow case, was then replenished multiple times by multiple further wave breaking events over the North Atlantic and Europe.The Tuscany case thus illustrates recurrent wave breaking and subsequent cutoff replenishment as a further, hitherto not documented synoptic storyline for unusually long wet spells.In contrast, a substantial body of literature documents the tendency for North Atlantic extratropical cyclones to serially cluster in the North Atlantic region (Dacre and Pinto, 2020 and references therein).
Here we document such behaviour of extratropical cyclones during a 15-day wet spell in Norway, which also involved orographic precipitation in westerly onshore flow.Finally, a 17-day wet spell in the Balkans reveals the importance of diurnal convection for this long summer-time wet spell over continental Europe, and also featured recurrent cutoffs that occur intermittently with periods of diurnal convection in absence of evident upper-level forcing.
During wave breaking upper-level troughs or ridges meridionally amplify to an extent where they are no longer governed by linear wave dynamics and the ¶ 110 ¶ ¶ compare the four metrices to respective "climatological" values, which are constructed in the following way: At each grid point (, ) we identify the days of the year during 260

270
anomalies for all four metrices.¶ unrelated.Fourth, a two-sided p-value is then estimated at each grid point by comparing  /0/12$3 (, ) with the distribution of the  /0/12$3 4 (, ) values.Hereby, a p-value of zero is assigned if  /0/12$3 (, ) lies outside the range of the  /0/12$3 4 (, )values.Fifth we reject the above null hypothesis for  /0/12$3 (, ) values at a grid point wise significance level of 0.01.Note that the randomization through the shuffling of the years of the  !" (, ) conveniently circumvents issues arising from the 275 spatially varying seasonality of the  !" (, ).Furthermore, in Section 3.3 we present anomalies of the four quantities for both weather systems.These are computed relative to climatological , ,  and  derived as the mean over the respective 1000 randomized values (e.g., the climatological  /0/12$3 (, ) is the mean of the 1000  /0/12$3 4 (, )).Note that due to the spatially varying seasonality and duration of the 280  !" the exact calendar days as well as the number of days contributing to the climatological values of , ,  and  vary in space.

Figure 1 .
Figure 1.Average characteristics of the 20 longest wet spells at 340 ) and the decreased tropospheric static stability induced by the upper-level cutoff.Within the next four days, little changed in this synoptic configuration over western Russia, although from 21 July onwards, a second cutoff-cyclone pair (L2, C2) formed over the British Isles, resulting in a large-scale omega-type blocking pattern (Fig. 3b-e).The surface cyclone L1 weakened hat nach unten verschoben [1]: Figure 2. Time evolution of the wet spells illustrated in Figs.3-6.The panels show the precipitation evolution during the longest wet spell at (a) 37°E/57°N, near Moscow, Russia in 2013, (b) at 380 14°E/66°N, in Norway in 1989, (c) at 12°E/43°N in Tuscany, Italy in 2018 and (d) at 25°E/48°N, in Romania in 1988.Bars depict 6-hourly precipitation (multiplied by four) while the bold blue line depicts daily precipitation accumulations.Daily precipitation is depicted for each day at 00:00 UTC and refers to the precipitation accumulation 385 over the preceding 24 hours.Red bars mark the onset and termination of each spell, while purple lines indicate the times for which fields are shown in Figs.3-6.Light vertical lines locate 00:00 UTC of each day.¶ hat gelöscht: analyzed 390 hat gelöscht: most likely hat gelöscht: quasi-geostrophic forcing for ascent and with similarities to the upper-level flow 430 pattern during the spell  + (37°, 55°) (Fig. 3) low-level shortwave trough (which can be regarded as the warm frontal extension of the cyclone north of Iceland) induced onshore flow hat gelöscht: and hat gelöscht: , days (Fig. 4b-e) cyclones L3-L5 formed over the western North Atlantic, rapidly propagated into the Norwegian Sea, and Figs. 2c, 5, and S4].This 12-day wet spell occurred in association with a large cutoff-complex over the Mediterranean that three (according to the Portmann et al (2021) cutoff data).The composite fields of 320 K PV, SLP, IVT and QGw during the  !" (12°, 43°) are consistent with the synoptic evolution of  & (12°, 43°) (Fig. 7c), with an amplified trough upstream of 12°E/43°N that is associated with anomalous IVT on its southern and eastern fringes, negative QGw centred on 12°E/43°N hat gelöscht: and hat gelöscht: -…treamer S1 was located over western Europe and substantial precipitation fell to its east where quasi-geostrophic 555 forcing for ascent is expected to be large ... [10] hat gelöscht: Between … and 12 May (Fig. S4e-g),,…as ... [11] hat gelöscht: ,…but ... [12] hat gelöscht: upper-level dynamics of the hat nach unten verschoben [4]: hat gelöscht: discussed above was tracked…can be regarded as an individual, long-lived system by the Portmann et al., (2021) On average, 2.3 distinct cyclones and 1.9 distinct cutoffs occurred at 12°E/43°N during the  !" .hat gelöscht: K 570 hat formatiert: Schriftart: Fett, Kursiv hat gelöscht: field hat gelöscht: is hat formatiert: Schriftart: Fett, Kursiv Hochgestellt hat gelöscht: with a wide ridge over Europe hat gelöscht: indicating hat gelöscht: unusually long-lasting wet spells…at 25°E/48°N is much more subtle ... [14] hat gelöscht: Cutoffs occurred at 25°E/48°N during larger 705 fractions of the  !" ( #,-&.. = 0.31) than cyclones and in slightly larger number ( #,-&.. = 2.10 vs.  #$#%&'( = 1.90).However, a hat gelöscht: ¶ ¶ values of  /562.. are found in the western Mediterranean, where they significantly exceed the respective climatological values by roughly a factor of two (Fig.9a,b).The cutoff period  /562.. reveals considerable geographical variations (Fig 9e), with shortest cutoff periods around 2 days in the Norwegian Sea, but as for  /562.. , the  /562.. anomalies are significant only in few and separate regions.The most striking result for cutoffs, though, is the significantly increased residence time  /562.. in the Mediterranean.There,  /562.. values in excess of two days are observed, which is more than 1 day more than the climatological value.Thus, the anomalously large  /562.. during the  !" in this region predominantly results from increased residence times of cutoffs, i.e., persistent cutoffs that are either slower-moving, longer-lived or both.A similar behaviour of cutoffs during the  !" is found in the southwestern corner of our study domain, over the subtropical North Atlantic.hat gelöscht: and  #$#%&'( hat gelöscht: at 14°E/66°N… a result that will be further discussed in Section 4. A much Clearer ... [16] hat gelöscht: relatively close to hat gelöscht: Cyclone 810 hat gelöscht: that clearly contrast those in central Europe and the UK …re found in the seas around …outh of Italy.These seas ... [17] hat gelöscht: more than…p to 1 0.5…days… Fig. 8h) anywhere in the study region.Notably in the seas around ... [18] hat gelöscht: A further region with similar and also significant cyclone characteristic anomalies during the  !" are the Baltic Seas.¶ .whereverthey are large, but overall they are considerably smaller than those of  #$#%&'( and  #$#%&'( , and hat gelöscht: only the  #,-&.. anomalies are significant over large 835 coherent areas hat gelöscht: not …ignificant only in few and separate regions, which suggests that the rate at which cutoffs occur during the  !" does not differ strongly from its climatological value (Fig. 9d,f) ... [20] hat gelöscht: western 840 hat gelöscht: 0.5… days ... [21] hat gelöscht: over the northern Black Sea as well as…n the smost s… ... [22] hat gelöscht: ¶ Over land, comparatively long absolute  /562.. values (up to 1.75 days) are found in the Balkans and in north-eastern Europe

Figure 1 .
Figure 1.Characteristics of the 20 longest wet spells at every ERA-Interim grid point (  ).(a) median duration, (b) median accumulated precipitation, and (c) average daily precipitation rate; (d) the season in which the largest fraction of the   start.Hatching and stippling in (c) depict regions where the average precipitation rate during the   exceeds the 95 th and 97.5 th percentile, respectively, of all ERA-Interim wet days, defined as days with > 0.1 mm precipitation.Hatching in (d) shows regions where at least 12 of the   start in the same season.Crosses identify grid points for which the   are further examined in Figs.2-7.

Figure 2 .Figure 3 .
Figure 2. Time evolution of the wet spells illustrated in Figs.3-6.The panels show the precipitation evolution during the longest wet spell at (a) 37°E/57°N, near Moscow, Russia in 2013, (b) at 14°E/66°N, in Norway in 1989, (c) at 12°E/43°N in Tuscany, Italy in 2018 and (d) at 25°E/48°N, in Romania in 1988.Bars depict 6-hourly precipitation (multiplied by four) while the bold blue line depicts daily 1200

Quantifying the characteristics of cyclones and PV cutoffs during the 𝓢 𝟐𝟎
/cutoff characteristics during the  !" to climatological values, in order to identify anomalous weather system characteristics during the  !" (and thus address research question 4).The left columns in Figs. 8 and 9 reveal that the relevance of cyclones and cutoffs for the  !" indeed varies greatly across space.For example, the range of  /0/12$3 values with lowest values below 0.2 and largest values above 0.9 implies that in some regions, cyclones are present during almost the entire period of the  !" , e.g., in Scandinavia, the UK, and Mediterranean, while cyclones appear to be largely irrelevant to the  !" in other regions such as the Balkans.Furthermore,  /562.. varies from around 0.2 (e.g., over Finland) to 0.7 (over the western Mediterranean), indicating that a major fraction of the total  !" duration occurs without a cutoff within a 400 km radius, even in the regions with the largest  /562.. .The right columns in Figs.8 and 9show single-signed or insignificant anomalies of , ,  and  almost throughout the study region for both weather systems, with positive anomalies in ,  and , and negative anomalies in .Thus, wherever the , ,  and  anomalies are significant, the respective weather systems are more prevalent (positive  anomalies), occur in larger cyclonenumber (positive  anomalies), at a higher rate (negative  anomalies) and tend to be more persistent (positive  anomalies) Portmann et al., 2018for an example of such a case).
hat gelöscht: features hat gelöscht:  hat gelöscht: both hat gelöscht: features hat gelöscht: and 925 hat gelöscht: The anomalies of  take either sign for both features, h… hat gelöscht: only relatively few (and mostly positive) hat gelöscht: are significant for both types of weather systems.