The unusually long cold spell and the snowstorm Filomena in Spain in January 2021

. In early January 2021, Spain was affected by two extreme events – an unusually long cold spell and a heavy snowfall event associated with extratropical cyclone Filomena. For example, up to 50 cm of snow fell in Madrid and the surrounding areas in 4 days. Already during 9 days prior to the snowfall event, anomalously cold temperatures at 850 hPa and night frosts prevailed over large parts of Spain. During this period, anomalously cold and dry air was transported towards Spain from central Europe and even from the Barents Sea. The storm Filomena, which was responsible for major parts of the snowfall 5 event, developed from a precursor low-pressure system over the central North Atlantic. Filomena intensiﬁed due to interaction with an upper-level potential vorticity (PV) trough, which was the result of anticyclonic wave breaking over Europe. In turn, this wave breaking was related to an intense surface anticyclone and upper-level ridge, whose formation was strongly inﬂuenced by a warm conveyor belt outﬂow of a cyclone off the coast of Newfoundland. The most intense snowfall occurred on 09 January and was associated with a sharp air mass boundary with an equivalent potential temperature difference at 850 hPa 10 across Spain exceeding 20 K. Overall, the combination of pre-existing cold surface temperatures, the optimal position of the air mass boundary, and the dynamical forcing for ascent induced by Filomena and its associated upper-level trough were all essential – and in parts physically independent – ingredients for this extreme snowfall event to occur. 19 January 2021, and a heavy snowfall 235 event from 07 to 10 January associated with the passage of extratropical cyclone Filomena. The snowstorm was exceptional due to the widespread snow cover and large accumulations of snow even in areas like Madrid that are rarely troubled by snow. Prior to the snowfall, a cold spell established and served as an important preconditioning for the snowfall event. We now summarise the key results by addressing the research questions raised in the Introduction.

initialised only from grid points in the Iberian Peninsula that were affected by the snowstorm, i.e. grid points where the accumulated snowfall between 00 UTC 07 January and 00 UTC 11 January 2021 exceeded 5 mm (Fig. 1a). From these grid points, trajectories are started every 3 hours during the cold-spell period from 00 UTC 30 December 2020 to 00 UTC 09 January 90 2021 at 10, 30, 50, 70, and 90 hPa above ground level, similar to, e.g., Zschenderlein et al. (2019) and Papritz (2020). Physical parameters traced along the trajectories include temperature, potential temperature, and specific humidity. Note that the cold spell lasted beyond 9 January, but since we are primarily interested in the origin of the cold air prior to the peak of the heavy snowfall, we have only calculated trajectories until this date. This trajectory analysis for the event in January 2021 is complemented by a climatological analysis, in which trajectories are initialised at the same grid points and vertical levels, and on the 95 same December and January days in all previous winters from 1980 to 2020.
Trajectories are also used to identify WCBs over the North Atlantic and the Iberian Peninsula during the study period in early January 2021 and for other heavy snowfall events. For that, we computed 2-day forward trajectories initialised between 20 • -50 • N and 90 • W-15 • E in the lower troposphere from 1000 to 700 hPa in 25 hPa increments. Trajectories ascending more than 600 hPa in 2 days were selected as WCBs, and in addition, we identify the ascent phase of a WCB, which corresponds to 100 WCB trajectories in the pressure layer between 800 and 400 hPa, following the definition of Madonna et al. (2014).

Identification of further heavy snowfall events and cold spells
In order to compare the snowstorm Filomena with other snowstorms in the Iberian Peninsula, we identify events in all winters from 1979 to 2019. For that, we compute the accumulated snowfall in the Iberian Peninsula at grid points over land between 36 • N-44 • N and 10 • W-4 • E in a 4-day window and determine the area affected by snowfall larger than 5 mm. We use a time 105 window of 4 days because the snowfall event in early January 2021 lasted for 4 days. In addition, we identify cold spells in all winters from 1979 to 2019, but only for the area where the accumulated snowfall between 00 UTC 07 January and 00 UTC 11 January 2021 exceeded 5 mm (Fig. 1a). In this area, we average the daily 2-m minimum temperature and define a cold spell as a continuous period with daily 2-m minimum temperatures below 0 • C.
3 Precipitation and temperature evolution in Spain in early January 2021 110 Many areas in Spain and Morocco experienced large amounts of precipitation between 07 and 10 January 2021. Most of it fell as rain in southern Spain and Morocco, and as snow in the rest of Spain and in the Atlas Mountains (Fig. 1a). The highest snow accumulation occurred near Madrid, as well as in eastern and southeastern Spain. Precipitation was confined to a corridor extending from Morocco, Andalusia, Madrid to provinces along the east coast of Spain, while Portugal, France and other parts of North Africa were not affected by the event. Overall, precipitation was most intense in Andalusia and off the Moroccan coast 115 (more than 75 mm in 4 days), where it however fell exclusively as rain. Compared to all winters from 1979-2019 in the Iberian Peninsula, the 4-day snow period associated with Filomena has, according to ERA5, the highest recorded snow accumulation and is one of the events with the largest extent (Fig. 1b). The area enclosed by the 5 mm snowfall contour in Fig. 1a (thereafter called snowfall area) experienced night frosts since 30 December 2020 and negative 2-m and 850 hPa temperature anomalies (Fig. 2a). In addition, the period was mostly dry 120 and only two short precipitation events on 28 December 2020 and 01 January 2021 affected the region prior to the snowstorm (Fig. 2b). The precipitation between 07 and 10 January was associated with cyclone Filomena and, in contrast to the two earlier precipitation events, most of the precipitation fell as snow. The diurnal cycle of temperatures on 09 January, i.e. at the peak of the snowfall event, was less pronounced than during the other days. After the snow event, night-time temperatures dropped, but at 850 hPa, a new air mass arrived and temperatures at this level were increasing again (Fig. 2a). Due to the existing snow cover, 125 2-m temperatures during night were still anomalously low and an inversion formed. Interestingly, the 2-m temperature and its anomaly curve have an opposite diurnal cycle, which means that the low day temperatures were particularly anomalous. As an example on 09 January, daily minimum temperature was about 3 K colder than climatology, while daily maximum temperature was more than 7 K colder than climatology, which is presumably related to the lack of incoming solar radiation. The cold spell, i.e. the period with daily minimum temperatures below the freezing point, lasted until 19 January 2021 (not shown).

130
Due to the cold spell in early January, the surface progressively cooled down and this prevented the snow from melting when reaching the surface on 07-10 January. Therefore, this cold spell preconditioning appears to be an important ingredient of the extreme snowfall accumulation. In the next section, we analyse the origin of this cold spell and put its duration in a climatological context. Before discussing the origin of the cold air near the surface over Spain in early January 2021, we first quantify how anomalous the duration of the cold spell was. We therefore compare the duration of this event with previous cold spells in the winters 1979 to 2019 that occurred in the area affected by snowfall on 07-10 January 2021 (Fig. 1a). Figure 3 shows the number of continuous days with minimum temperatures below 0 • C at 2 m and 850 hPa, respectively. The cold spell in early January 2021 lasted for 21 days at 2 m, and for 14 days at 850 hPa. The difference is presumably associated with the inversion that 140 formed after the snow event due to warm air advection at 850 hPa (Fig. 2a). The longevity of the cold spell at the surface can be classified as extreme because it exceeds the 99th percentile of the climatology of continuous night frost days at 2 m, while the duration of the cold spell at 850 hPa is slightly below the 99th percentile (Fig. 3). The 99th percentile of the climatological cold spell duration at 2 m is 16 days, and with 21 days the January 2021 event is the 3rd longest in the period 1979-2019. The longest cold spell with 29 days occurred from 29 December 1999 to 26 January 2000 and the second longest from 10 January 145 1992 to 04 February 1992. Both events were not followed or accompanied by heavy snowfalls, hence, the combination of long-lasting cold temperatures and high accumulations of snow is a unique feature of the event in early January 2021, at least since 1979. Figure 4 illustrates the origin of the cold air in the snowfall area during the cold spell period from 30 December 2020 to 09 January 2021, based on the trajectory method described in Section 2.2. The locations of air parcels are shown as frequencies 150 per km 6 , i.e. the trajectory count at each grid point is normalised by the number of all trajectories, such that the spatial integral yields 100 %. The normalisation was applied separately to the trajectories of the cold spell and the climatology. The purple lines in Fig. 4 show the position of air parcels three days prior to their arrival over Spain in early January 2021. Most of them are located over central Europe, France, the British Isles, and some even over the Barents Sea (Fig. 4). In comparison with climatology, i.e. with the origin of air on all days between 30 December and 09 January in the years 1980 to 2020, the origin 155 of the cold air in January 2021 is unusually far north. Transport of cold air from the north has been reported as an important feature of cold spells in general (Bieli et al., 2015).
The air mass origin in 2021 is not only unusual when considering the position three days prior to arrival in Spain, but also for the whole 10-day period of the backward trajectories ( substantially lower isentropes compared to climatology (Fig. 5e). The lower isentrope (median value of 285 K compared to 295 K in the climatology) is related to the northern origin. This anomalously cold air is also drier than normal (Fig. 5f). It is interesting that the temperature of the air parcels at their origin is only slightly colder than normal (Fig. 5d), which can be explained by compensating effects of being further north (Fig. 5a) and at a lower altitude (Fig. 5c). During the last five days of the transport, they were less warmed compared to climatology and therefore stayed below 0 • C (Fig. 5d). Physically, air 165 parcels change their temperatures due to adiabatic processes (vertical motion), or diabatic processes, e.g. radiative cooling or heating by surface sensible heat fluxes. Potential temperature changes along the backward trajectories associated with the cold spell are weak and temperature changes are therefore primarily due to adiabatic warming during descent. Indeed, air parcels leading to the cold spell descend less than climatologically expected (Fig. 5c). Therefore, the cold air is eventually cold due to (i) slightly colder origin temperatures, and (ii) weaker subsidence and therefore less adiabatic warming. Hence, over a period 170 of approximately ten days, cold and dry air was steadily advected towards the snowfall area and helped to progressively cool down the surface. This trajectory analysis explains why 2-m temperatures in Spain were exceptionally cold in early January 2021. We now continue with investigating the lifecycle of cyclone Filomena, which was responsible for the snowfall event at the end of this exceptional cold spell.

175
Cyclone Filomena evolved from a decaying low-pressure system over the central to eastern North Atlantic (marked with L1 in Fig. 6c). It is therefore interesting to first discuss the evolution of this precursor cyclone L1. The cyclogenesis of L1 occurred at 12 UTC 02 January near Newfoundland (Fig. 6a). The weak upper-level PV trough intensified into a narrow PV streamer until 00 UTC 04 January 2021, and L1 underwent rapid intensification reaching a core pressure of nearly 980 hPa in the western North Atlantic (Fig. 6b). Within the next hours, the elongated PV streamer broke up into a PV cut-off. Until 00 UTC 06 180 January, L1 weakened and propagated further east, but remained at fairly low latitudes (36 • N, Fig. 6c). The large-scale flow pattern over the North Atlantic then quickly changed in the next 36 hours. A second low pressure system, L2, evolved near Newfoundland at almost the same location as L1 three days earlier (Fig. 6a,c). This low was associated with strong warm conveyor belt activity at its leading edge (Fig. 6d). This warm conveyor belt helped to intensify the downstream ridge and the https://doi.org/10.5194/nhess-2021-396 Preprint. Discussion started: 4 January 2022 c Author(s) 2022. CC BY 4.0 License. associated surface high-pressure system over the central North Atlantic, a situation that is often observed in this region (Grams et al., 2011). The strong ridge contributes to anticyclonic wave breaking over western Europe, which is important for the reintensification of the decaying low L1 and the cyclogenesis of Filomena near 32 • N at 12 UTC 07 January 2021 (marked with F in Fig. 6d). The anticyclonic wave breaking pushed the upper-level PV trough towards the Iberian Peninsula, and therefore, Filomena slightly intensified while propagating north-eastwards towards the Iberian Peninsula (Fig. 6e). At 00 UTC 09 January 2021, Filomena was located over the Iberian Peninsula where it led to intense rain and snow in this area (Fig. 6f, cf. Fig. 1a and   190 2b). In addition to the anticyclonic wave breaking, the transverse vertical circulations associated with the subtropical and the polar jet influenced the development of Filomena and the associated warm conveyor belt. For example at 00 UTC 09 January 2021, Filomena is located at the intersection of the left exit region of the subtropical jet and the right entrance of the polar jet stream (Fig. S1). A similar situation was observed for snowstorms in the eastern US (Uccellini and Kocin, 1987) and for cyclogenesis in the eastern Mediterranean (Prezerakos et al., 2006).

195
The storm Filomena led to strong contrasts of equivalent potential temperature (θ e ) at 850 hPa exceeding 20 K over Spain at the peak of the snowfall event at 00 UTC 09 January 2021 (Fig. 7c). Four days earlier, on 05 January 2021, the air was cold and dry over the Iberian Peninsula with θ e values below 300 K (Fig. 7a). With the approach of Filomena from 07-09 January 2021, θ e values in northern Spain decreased even more and over Northern Africa, θ e values increased due to the approaching warm sector of Filomena with values exceeding 310 K (Fig. 7b,c). Most of the precipitation occurred in areas with very large 200 θ e gradients, i.e. fronts. The warm front on 09 January led to the strongest snowfall in central Spain associated with a warm conveyor belt, which ascended exactly at the location of strong snowfalls (Fig. 6d-f). The extreme accumulation of snow is also related to the slow propagation of Filomena and its warm front ahead of the quasi-stationary upper-level trough. Figure 8 illustrates the stationarity of the front with the aid of meridionally oriented vertical cross sections of θ e and cloud variables. At all three timesteps, the upper part of the clouds are completely glaciated, at medium levels liquid and ice clouds coexist, while 205 the lower part consists of cloud water with rain on the warm side of the front and (intense) snow on the cold side. On the warm side of the front clouds were quite shallow, while on the cold side they were comparably deep with cloud tops up to 350 hPa.
South of the front, the air in the warm conveyor belt inflow was very moist. It is nicely visible that the ascent of the warm conveyor belt at the warm front is co-located with the heaviest snowfall at all three time steps. Altogether, the combination of a slow-moving front, sufficient lower-tropospheric moisture on its warm side, and pre-existing cold surface temperatures prior 210 to and during the snowstorm were essential ingredients for the large amounts of surface snow accumulation.

Comparison with other heavy snowfall events
In the final part of this study, snowstorm Filomena is compared to the second and third strongest snowstorms in the period 1979-2019, based on the method described in Section 2.3. These events are labelled as E2 and E3 in Fig. 1b, respectively, and we briefly compare the snow cover, the synoptic situation, and the air mass origin with the Filomena event in 2021. For this The second strongest snowstorm, the E2 event, occurred from 02 to 05 January 1997 and mainly affected mountainous regions in Northern Spain, in particular the Pyrenees (Fig. S2a). Hence, the snowstorm was associated with significant orographic effects, in contrast to snowstorm Filomena. Another difference to Filomena was the higher surface temperature, which led to stronger snowmelt in the lowlands. Similar to Filomena, temperatures decreased before the onset of the snowfall, however, 220 with still higher temperatures than during Filomena (not shown). The third strongest snowstorm, the E3 event, lasted from 28 to 31 January 1986 and also affected mountainous regions (Fig. S2b). Cold air from Greenland was transported and moistened over the North Atlantic, and when reaching northern Spain, it was orographically lifted and contributed to the cloud and precipitation formation (not shown). East of the Pyrenees, an intense cyclone with a core pressure of about 970 hPa enabled the strong snowfall (Fig. S3). In both events, the horizontal θ e contrasts were weaker than during the passage of Filomena 225 (Figs. S4a, S4b, and 7c). It is interesting that the very deep cyclone during the E3 event is accompanied by weaker θ e gradients than the weak cyclone Filomena with core pressures of about 1000 hPa (Fig. 6f). High θ e gradients can influence the intensity of a cyclone, as for example the winter storms hitting Europe in December 1999, which were very intense (Ulbrich et al., 2001). However, it appears that the strong θ e gradient during the passage of Filomena did not have a large impact on the intensification of the cyclone, but rather on the precipitation intensity. In comparison with events E2 and E3, it is remarkable 230 that the accumulation of snow during Filomena is so large without substantial orographic influences from the Pyrenees, which underlines the importance of the cold spell preconditioning in early January 2021.

Conclusions
In this study, we analysed the synoptic-scale dynamic evolution of two extreme events in early January 2021 affecting large parts of the Iberian Peninsula -a long-lasting cold spell from 30 December 2020 to 19 January 2021, and a heavy snowfall 235 event from 07 to 10 January associated with the passage of extratropical cyclone Filomena. The snowstorm was exceptional due to the widespread snow cover and large accumulations of snow even in areas like Madrid that are rarely troubled by snow. Prior to the snowfall, a cold spell established and served as an important preconditioning for the snowfall event. We now summarise the key results by addressing the research questions raised in the Introduction.
1. How unusual was the cold spell and which processes led to the anomalously low temperatures?

240
The cold spell lasted 21 days and exceeded the 99th percentile of the number of continuous days with 2-m minimum temperature below 0 • C. Due to the steady advection of dry and cold air from the north, surface temperatures could progressively cool down, which later prevented the snow from melting. A comparison with the climatological origin of near-surface air over Spain revealed that during the cold spell, exceptionally cold and dry air was transported from central and western Europe, but also from the Barents Sea. This air mass origin is shown to be unusual. Filomena evolved from a decaying low in the central North Atlantic at fairly low latitudes and intensified along a quasistationary upper-level PV trough, which formed due to anticyclonic Rossby wave breaking over Europe. This wave breaking in turn was related to an intense anticyclone over the central North Atlantic, which was strongly influenced by the WCB outflow of an intense extratropical cyclone off the coast of Newfoundland. Filomena was located at the 250 intersection of the left exit of the subtropical jet and the right entrance of the polar jet -both regions are associated with large-scale forcing for ascent, and therefore, the configuration of the two jets further influenced the intensification of Filomena. However, Filomena was not very intense with a minimum pressure of about 1000 hPa. This sequence of events, including the upstream influence from a cyclone and its warm conveyor belt in the western North Atlantic illustrates the complex dynamics responsible for the formation of Filomena. Large parts of Spain, especially the area in and around Madrid, were affected by intense snowfall that occurred along a sharp air mass boundary identified by a horizontal contrast of θ e of more than 20 K across Spain. Along the warm frontal part of this boundary, dynamical forcing for ascent, imposed by the surface cyclone and upper-level trough, led to warm conveyor belt ascent and intense snowfall. Hence, the combination of pre-existing cold near-surface temperatures, the 260 "optimal" position of the air mass boundary, and the ascent of the moist air along steep moist isentropes associated with the warm front of Filomena were essential ingredients for this extreme snowfall event to occur. This case study reveals how different processes on various time scales co-produced this hazardous event. In addition, this event is also an example of the relevance of diabatic processes for large-scale weather patterns in the North Atlantic (e.g. Grams et al., 2011;Magnusson, 2017). A comparison with two heavy snowfall events and two long-lasting cold spells has revealed the 265 uniqueness of the events in early January 2021. First, snowfall events with comparable accumulated 4-day snowfalls as during Filomena in the Iberian Peninsula are usually strongly influenced by orographic effects, i.e. the Pyrenees typically record the highest snow accumulation. In such situations, northwesterly winds transport cold and humid air towards the Pyrenees and produce precipitation by orographic lifting. Second, the co-occurrence of such a long-lasting cold spell and heavy snowfall is unusual -the two longest-lasting cold spells in that region were not associated with significant amounts of snow. Overall,   (a) 2 0 2 0 -1 2 -2 7 2 0 2 0 -1 2 -2 9 2 0 2 0 -1 2 -3 1 2 0 2 1 -0 1 -0 1 2 0 2 1 -0 1 -0 3 2 0 2 1 -0 1 -0 5 2 0 2 1 -0 1 -0 7 2 0 2 1 -0 1 -0 9 2 0 2 1 -0 1 -1 1 2 0 2 1 -0 1 -1 3