These authors contributed equally to this work.
Cold and snowy spells are compound extreme events with the potential to cause high socioeconomic impacts. Gaining insight into their dynamics in climate change scenarios could help anticipating the need for adaptation efforts. We focus on winter cold and snowy spells over Italy, reconstructing 32 major events in the past 60 years from documentary sources. Despite warmer winter temperatures, very recent cold spells have been associated with abundant and sometimes exceptional snowfall.
Our goal is to analyse the dynamical weather patterns associated with these events and understand whether those patterns would be more or less recurrent in different emission scenarios using an intermediate-complexity model (the Planet Simulator, PlaSim). Our results, obtained by considering RCP2.6, RCP4.5 and RCP8.5 end-of-century equivalent
Cold and snowy spells are driven by the midlatitude atmospheric circulation through the amplification of planetary waves
In this study, we focus on the dynamics of compound extreme cold and snowy events, for which the response to mean global change might be different from that of the individual variables (temperature and snowfall). Indeed, taking this complementary compound extreme events point of view
In this paper, we focus specifically on Italy: recent cold and snowy spells in this country have caused casualties in the population, strongly affected ground and air transport, and caused disruptions in services. Our strategy to tackle these questions is to analyse simulations produced in a global circulation model (GCM) under different emission scenarios. We first validate the cold and snowy spells produced by a simplified GCM of intermediate complexity with historical forcing, i.e. the Planet Simulator (PlaSim)
Our study is based on the detection of synoptic meteorological configurations leading to cold spells over Italy in PlaSim, considering a control run based on the recent historical climate and a set of three increased emission scenarios at steady state. In order to do so, we will proceed with the following steps:
identify large-scale, high-impact winter cold spells over Italy; describe the dynamic and thermodynamic conditions associated with such cold spells; detect cold-spell analogues in a historical climate dataset; detect cold-spell analogues in PlaSim runs and evaluate whether climate change can significantly modify their frequency and in which direction; characterize the PlaSim cold-spell analogues by analogy with point 2 to assess the potential of the considered dynamic configurations in producing relevant winter phenomena in a sensibly warmer climate.
In order to identify relevant cold spells over Italy, we consider documented events that have produced at least a record low temperature and/or a record snowfall amount (or snow at locations where snowfall has never been previously reported) at one or more locations in Italy. We combine official sources and both professional and avocational websites dedicated to weather and climate, where collections of weather event reports are available, and we countercheck their validity with station data and trusted documentary sources
Cold spells from documentary sources. Data recorded in
The in-depth description of the effects of each cold spell at the country level is presented in Appendix
Given the heterogeneous and, in some cases, unofficial origin of the considered data, we only aim to draw a qualitative picture. Overall, our analysis indicates that extreme snowfalls have occurred in recent years, despite warming temperatures (Fig.
Dynamics for clusters of cold spells over Italy: 500 hPa geopotential height [m]
Besides the qualitative analysis involving the cities of Bologna and Campobasso briefly presented in Sect.
Although our analysis is focused on cold spells affecting an area containing Italian borders, the dynamic determinants of such cold spells span much larger scales. For this reason, we consider a larger area, including Europe, European Russia, and the North Atlantic, over a 2.5
For
In order to choose the optimal number of clusters, we first performed a scree plot (not shown), obtained by plotting the within-groups sum of squared differences from the cluster centroids. This analysis did not give clear indications about the ideal number of clusters. Therefore, we compared clustering results at different values of
In Fig.
Cluster 1 presents a pattern resembling a Scandinavian blocking but with positive SLP anomalies displaced to the south, with an anticyclone stretching in a SW–NE direction rather than elevated along the meridians, and low-pressure values centred over the central Mediterranean, mainly confined below 40
Physics for clusters of cold spells over Italy. Temperature at 850 hPa [
Cluster 2 is characterized by an Omega wavy structure associated with an Atlantic high-pressure ridge
Figure
Cluster uncertainty. Standard deviation of 500 hPa geopotential height [m]
We measure the uncertainty associated with the cluster composites discussed above by computing the standard deviation of the standardized anomalies used for the clustering, shown in Figs.
Cluster uncertainty. Standard deviation of temperature at 850 hPa [
In order to understand how the frequency of cold-spell events may change in a warmer climate, we simulate different emission scenarios using PlaSim
In this study we consider four simulations performed with
The first simulation is the control run (hereinafter CTRL) with radiative forcing levels representative of the recent past climate: equivalent
This way, we explore three scenarios where excess heat is stored in the atmosphere in different amounts, and we investigate which differences in the dynamics associated with cold and snowy spells in the present climate appear, if any. We will analyse the
Climate models, even those with higher complexity than PlaSim, are characterized by a finite resolution, thus leaving smaller scales unresolved, and contain several physical and mathematical simplifications that make climate simulations computationally feasible, while also introducing a certain level of approximation. This results in statistical biases that can be easily observed when comparing control runs to observations or reanalysis datasets. In order to mitigate the effects of these biases, a bias correction step can be performed. Bias correction usually consists of adjusting specific statistical properties of the simulated climate variables to a validated reference dataset in the historical period. The target statistics can be very simple, such as a central tendency index like the mean
Given the lower complexity and the relatively coarse grid of PlaSim compared to other regional or global circulation models, we rely on simple methodologies. We apply BC only to the
For the variables characterized by approximately symmetric distributions (
For PRP, we must rely on a different method, given the strong asymmetry characterizing the distribution of precipitation. We choose quantile mapping based on regularly spaced quantiles, with a wet-day correction to obtain an equal fraction of days with precipitation in the reference and corrected data: the empirical probability of nonzero precipitation is found, and the corresponding modelled value is selected as a threshold. All modelled values below this threshold are set to zero. This technique is described by
We base our analysis of cold spells in PlaSim on the search for dynamic analogues
Let
Now let
The procedure is carried out, for each cluster, according to the following steps:
define compute the metric determine the critical value now take PlaSim compute the metrics estimate probabilities regard all events satisfying
The figure shows the 500 hPa geopotential heights in PlaSim simulations: average of the 500 hPa geopotential height [m] for cluster 1
Steps 1–3 select the (
Our first result follows from the estimation of the probabilities described in step 6. in the previous paragraph. These are the probabilities
The results of this analysis are summarized in Table
Change in frequency of cold-spell analogues for each of the considered events, divided by cluster. The values corresponding to the RCPs are adjusted by subtracting the values relative to the CTRL run, which quantify PlaSim's bias in analogues frequency; unadjusted values are shown in parentheses.
The results for the CTRL run inform us about the capability of PlaSim to reproduce the frequency of the two dynamical fingerprints of cold spells associated with the two clusters of events. Both configurations are much more frequent in PlaSim than in NCEPv2, with a
Sea-level pressure in PlaSim simulations: average of the sea-level pressure [hPa] for cluster 1
Both configurations become increasingly more frequent with growing radiative forcing. It is worth mentioning that this analysis does not discriminate between an increased number of Atlantic ridge and Scandinavian blocking episodes and their longer persistence, both of which may lead to more analogues. Nevertheless, these result clearly suggest a higher number of days characterized by configurations leading to a flow of Arctic air towards the Mediterranean area under climate change.
The figure shows 850 hPa temperature in PlaSim simulations: average of the 850 hPa temperature [
Figure
Figure
The precipitation patterns associated with these analogues are shown in Fig.
In summary, in a warmer climate, the frequency of dynamic configurations leading to
The figure shows 2 m temperatures in PlaSim simulations: average of the 2 m temperature [
Figures
Daily precipitation rates in PlaSim simulations: average of the daily precipitation rates [mm d
We have characterized high-impact cold spells that affected Italy in the course of the past 68 years by assessing their common dynamical large-scale signature. Despite the differences in duration, snowfall and temperature recorded during each event, the corresponding
Then, after assessing the capability of PlaSim to reproduce dynamic analogues of these events in the CTRL run, we studied the influence of climate change on the frequency of such analogues using three steady-state increased emission scenarios. The PlaSim control run showed a tendency to overestimate the frequency of both configurations. All three RCP runs are associated with more frequent configurations potentially leading to cold spells, with frequency increasing with equivalent
Since temperatures are projected to be contextually higher, cold spells and snow are naturally expected to decrease overall, especially under RCP4.5 and RCP8.5; however, we argue that the formation of cold air over the Arctic winter would not be completely suppressed, hence making cold-spell events still possible, and they remain relatively likely under the mitigated RCP2.6 scenario. This observation is particularly important, as RCP2.6 is representative of the current target to comply with the requirements of the Paris Agreement
Moreover, the temperature fields shown in Figs.
This study comes with some caveats and limitations: although we have validated the behaviour of PlaSim against the NCEP reanalysis, results for frequency changes for cold spells crucially depend on the position and the destabilization of the jet stream. It is known that different climate models have a different response of jet stream dynamics to climate change (Arctic amplification
The use of an intermediate-complexity model like PlaSim allowed us to evaluate climate change in atmospheric dynamics associated with cold spells in a steady, much warmer climate, showing how the frequency and intensity of cold spells may decrease less than expected, due to a higher likelihood of synoptic configuration favourable for cold air to flow towards the Mediterranean.
In this section, we describe each extreme cold event selected as a cold spell in this study. The mains characteristics of the events are the occurrence of snowfalls in regions where snow cover has usually been rare or absent for a long time (e.g. lowlands and coasts), large socioeconomic impacts (e.g. in 2017), extreme minimum temperatures and extreme amount of snowfalls. The date reported at the beginning of each event is the one selected as the most representative day of each cold-spell event, and it is the one used for the analogue search. The information about the duration of the events is reported in the text for each description.
4 January 1954: a cold spell rapidly built up in the Mediterranean in January 1954 (an exceptional month in Spain). Heavy snowfalls affected all of northern Italy, including lowland areas in the Po Valley. In 24 h, up to 60 cm of snow fell over Turin, Brescia, Milan, Piacenza, Cremona, Reggio Emilia, Bologna and Vicenza according to information found in the press 4 February 1956: one of the coldest and snowiest events of the 20th century in Europe. On 2 February, the 17 December 1961: December was a very cold month for most of Italy with historical snowfall in southern Italy coastal areas, with up to 30 cm accumulation in Bari 31 January 1962: Sicily reported several historical records of daily low temperature as in Lentini ( 22 January 1963: the winter of 1963 was one of the coldest in western European records. Sea frost trapped Norway's islanders, while a record low temperature of 12 January 1968: between 9 and 15 January 1968, Tuscany and nearby areas were affected by one of the strongest cold spells on record for the region. Extreme daily low temperatures were recorded: Città di Castello (Umbria, 295 m), 28 February 1971: on 24 February, the presence of an omega blocking with an anticyclone meridionally elevated towards the British Isles and a trough with a pressure minimum over the central Mediterranean, triggered a flow of Arctic air towards the Mediterranean. After affecting northern Europe, the cold spell reached Italy, causing a severe temperature drop between 28 February and 1 March. On the morning of 1 March, almost all of Italy recorded minimum temperatures below zero even in lowland and coastal areas: 1 December 1973: at the beginning of December, a cold air mass associated with a low-pressure area reached Italy from Scandinavia, with the 15 January 1979: a large pool of Arctic air stretching up to the North African coasts brought a cold spell that affected most of Europe, causing several fatalities. The cold air caused wind storms in the Tyrrhenian Sea, followed by a severe temperature drop. Snowfall occurred in Tuscany, Sardinia, and most of central and southern Italy, with snowstorms in the Marche, Abruzzo, Molise and Basilicata regions. The most abundant snowfalls were observed on 19 January with the advection of more temperate and humid air from the south-west. Traffic problems due to frost on the roads and to iced pipes were reported 8 January 1981: a very cold air mass penetrated deeply into the central Mediterranean Sea, accompanied by an intense storm over the south of Italy. On 8 January, western central Sicily was disrupted by unprecedented amounts of snow for the area, with 30 cm of snowfall even on the coasts. Extremely unusual snowfall was observed even on Pantelleria, a small island located south of Sicily, with only 5 m elevation above sea level. Some cities in the provinces of Palermo, Trapani, Messina and Enna remained isolated for days. The temperature reached a historical minimum of 7 January 1985: from 1 to 17 January 1985, Italy and most of western Europe were affected by a disruptive and persistent cold spell. A cyclogenesis over central Italy, between Tuscany and Lazio, triggered strong bora winds and historical snowfalls that affected Florence with 40 cm of accumulation (up to 80 cm in Val di Cecina) and Rome with 30 cm. The pressure minimum moved towards the south-east between 6 and 9 January, and the snow also reached Campania and the rest of the south with accumulations of up to 25 cm in the hilly zones of Naples, which had not happened since 1956 24 December 1986: Christmas Day 1986 was characterized by strong winds and 850 hPa isotherms of 3 March 1987: cold air and stormy weather reached the extreme south-east of Italy, with a peak on 8 March 1987 when the 31 January 1991: cold air entered the Mediterranean as strong bora winds, causing temperature to drop to 1 January 1993: a zonally tilted anticyclone with pressure maxima between the UK and Scandinavia drew a large Arctic air patch from Russia towards Italy. Due to the peculiar configuration, cold air flowed from Russia through Ukraine, Romania and the Balkans and then mostly affected southern and central Italian regions, especially the Adriatic side, where the snow fell also in coastal areas. The absolute minimum temperature record was broken in Bari ( 27 December 1996: this cold spell also affected the UK and France ( 31 January 1999: Arctic air reached Italy, particularly affecting the central and northern regions, on 5 February. The snow affected the entire Po Valley, from Venice to Turin, with accumulations of up to 30 cm on the plain. Snow also fell abundantly in the coastal cities of Rimini, Ancona, Grosseto and Genoa and in the Tuscan cities of Florence and Lucca. A snowstorm struck Viterbo and snowflakes were also observed in Rome with a remarkable temperature of 8 December 2001: before affecting northern Italy, cold air reached parts of central eastern Europe from Russia. On the evening of 13 December, the air mass entered the Po Valley in the form of a strong bora wind, causing convection accompanied by a blizzard-like snowfall that caused transport, electricity and phone line disruptions, and isolated several small towns in northern Italy. In Trieste the bora wind blew at 116 km h 20 January 2004: during this event, icy currents flowed from the north-east towards northern Italy, with weak snowfalls over Emilia, up to medium–low altitudes. In the following days, however, it snowed again in the north and in the central regions at lower altitudes: snow reached Tuscany and Lazio, with snowflakes even in Rome. The temperatures in these days of January were particularly low in the north-eastern Alps and in central and southern Italy, where markedly negative values were recorded over the usually mild Tyrrhenian plains ( 22 January 2005: western and central Europe experienced below average temperatures throughout the winter, with the cold peaking during the month of January. In northern Italy snow fell abundantly: in Lombardy snow height reached 30 cm, with peaks of up to 40–45 cm, and even the coastal city of Genoa suffered snowfalls 2 March 2005: a cold and snowy spell hit central and southern Italy. Snow fell in the hills of Naples, at medium–low altitude in Calabria, and with abundant accumulations on the central Adriatic coast (from southern Marche to Molise). A few days later, snowfall spread to the north and Tuscany 13 December 2007: the peculiarity of this cold spell was the exceptional occurrence of abundant snowfalls, blizzard conditions and an extreme low in temperatures over most of the Sardinian territory, at altitudes on average above 400 m. Also noteworthy are the 2 m of snow accumulated over an altitude of 1000 m on the slopes of Mount Limbara. Towns were largely unprepared to manage the event. An electricity blackout affected Cagliari for several hours, and schools remained closed for 2 d. Disruptions were reported in road connections: the main road of the Sardinian network of state highways suffered numerous blocks due to some trucks blocking the roads. In Nuoro, the snowfall exceeded 50 cm, breaking the record 17 December 2009: most of central and northern Europe was struck by this cold spell. On 19 December, snow fell over most of northern Italy, and it was especially copious in Tuscany 12 February 2010: snowfalls affected several regions, from Emilia-Romagna to Calabria, Marche and Sardinia 11 December 2010: an Arctic air mass reached the eastern side of Italy, giving rise to intense snowfalls on the Adriatic and Tyrrhenian coasts (especially the coasts surrounding the city of Livorno). Snow also turned Tuscany (25 cm fell in Florence), Umbria and part of Lazio white, with snowflakes observed in Rome. An exceptional snowstorm hit Ancona and the surrounding areas between 14 and 15 December. There, the Adriatic-effect snow contributed to reaching snow heights of up to 30 cm in the Chieti area and 40 cm in Lanciano. The temperatures were extremely cold over most of Italy. Milan Malpensa airport measured a minimum temperature of 2 February 2012: the February 2012 cold spell affected a large part of Europe and spread down to North Africa in the period between 27 January and 20 February 2012, causing over 650 deaths in the areas concerned. The event was characterized by extremely low temperatures, especially over eastern Europe, with an absolute minimum of 7 February 2013: this cold spell consisted of a polar trough spreading towards the Mediterranean region: the 28 December 2014: this cold spell affected the south of Italy with locally exceptional snowfalls, especially in Sicily and Apulia, the latter recording important accumulations on the plains and coasts. Snow also appeared in Naples and on the Amalfi coast. Snowfalls affected Sicilian coasts including the city of Messina, the hills and the hinterland of Palermo, with large accumulations, and Syracuse. The event was modest in Catania, with accumulations only in the hills of the city. The extreme south-eastern tip of Sicily experienced snowfall on New Year's Eve, an extremely rare event, as these southernmost areas of the country had not received snow since January 1905. Historic snowfall were also recorded in Pachino, a city famous for the production of a special type of cherry tomatoes. Snowfall also affected Sicilian towns on the Ionian side, like Avola and Noto. This cold spell was extraordinary also in the south of Sardinia, where Cagliari and surrounding areas were covered by snow 5 February 2015: Italy was affected by stormy and snowy conditions. It snowed extensively in Piedmont as well as in Liguria, a region also affected by strong winds. Snow fell at low altitudes and in the lowlands in the northern Italy and in part of central Italy. During this event, due to the strong winds, Sicily was isolated and the connections with the smaller islands were interrupted. In the surroundings of Etna, snow, wind gusts and ice caused blizzard conditions. In Enna the temperature dropped to 16 January 2016: on 17 January cold air flowed from Russia, passed over the barrier of the Alps and reached the Apennine chain. Due to the strong winds and rough sea, the Aeolian Islands were isolated and the highest peaks of the islands (Stromboli and Salina) were covered in white. Storm surges struck the Sicilian north coast 5 January 2017: from 5 to 21 January 2017, a cold spell affected most of eastern and central Europe and part of southern Europe, causing the death of at least 60 people. The cold and snowfalls mainly affected central and southern Italy. The regions most affected by this cold spell were the Adriatic ones, namely Marche, Abruzzo, Molise, Apulia and Basilicata. Snow reached almost all coastal areas of these regions, with snow totals of up to 40 cm. On 8 January, the beach of Porto Cesareo in Apulia was covered at some point with accumulations of 22–23 cm, resulting as the third most snowy Italian beach since 2000. The situation was worse in inland areas, where snow often exceeded 2 m height. A strong snowstorm affected the entire Marsicano sector (Abruzzo), with temperatures ranging between 18 February 2018: the cold spell affected Europe between the end of February 2018 and the beginning of March. The major anomalies concerned the central and northern sector of Europe with temperatures between 5 and 9
The figure shows 500 hPa geopotential height uncertainty in PlaSim analogues: root mean squared difference in standardized 500 hPa geopotential height (standard deviations) between analogues and cluster centroids for cluster 1
Sea-level pressure uncertainty in PlaSim analogues: root mean squared difference in standardized sea-level pressure (standard deviations) between analogues and cluster centroids for cluster 1
The figure shows 850 hPa temperature uncertainty in PlaSim analogues: root mean squared difference in standardized 850 hPa temperature (standard deviations) between analogues and cluster centroids for cluster 1
The figure shows 2 m temperature uncertainty in PlaSim analogues: root mean squared difference in standardized 2 m temperature (standard deviations) between analogues and cluster centroids for cluster 1
Daily precipitation rate uncertainty in PlaSim analogues: root mean squared difference in standardized daily precipitation rates (standard deviations) between analogues and cluster centroids for cluster 1
“The Planet Simulator (PlaSim): a climate model of intermediate complexity for Earth, Mars and other planets” is an open-source model available for download from
The data used in this study are available, upon request, for free by emailing the contact author of this study.
DF conceived the idea of the study and designed the methodology with CN. MD and ST performed the simulations in consultation with FL, and FP executed the statistical analysis of the results. MD, FP and DF performed the study. PY, CN, FL and DF discussed the results and implications and commented on and edited the paper.
The contact author has declared that neither they nor their co-authors have any competing interests.
Publisher's note: Copernicus Publications remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
This article is part of the special issue “Understanding compound weather and climate events and related impacts (BG/ESD/HESS/NHESS inter-journal SI)”. It is not associated with a conference.
This work was supported by ANR-TERC grant BOREAS. We thank Fabio D'Andrea and Aglaé Jézéquel (Laboratoire de Météorologie Dynamique, Paris, France) for useful discussion on the paper. Frank Lunkeit acknowledges support by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) through the University of Hamburg's Cluster of Excellence Integrated Climate System Analysis and Prediction (CliSAP) and under Germany's Excellence Strategy – EXC 2037 “Climate, Climatic Change, and Society” (CliCCS) – Project Number: 390683824, as a contribution to the Center for Earth System Research and Sustainability (CEN) of the University of Hamburg. The authors acknowledge the support of the INSU-CNRS-LEFE-MANU grant (project DINCLIC).
This research has been supported by the ANR-TERC Boreas LEFE-MANU-INSU “Dinclic” (grant no. LEFE-MANU-INSU “Dinclic”).
This paper was edited by Jakob Zscheischler and reviewed by Hylke de Vries and one anonymous referee.