Reconstruction of a Storm Map and New Approach in the Definition Of

water

Article Reconstruction of a Storm Map and New Approach in the Deﬁnition of Categories of the Extreme Rainfall, Northeastern Sicily

Francesco Fiorillo 1,*, Nazzareno Diodato 2 and Massimiliano Meo 3

1 Dipartimento di Scienze e Tecnologie, Università degli Studi del Sannio, via dei Mulini 59/A, Benevento 82100, Italy 2 Met European Research Observatory, HyMex Mediterranean Experiment Network—Via Monte Pino, Benevento 82100, Italy; [email protected] 3 CMG Testing S.r.l., Via Piano Alvanella, Monteforte Irpino, Avellino 83024, Italy; [email protected] * Correspondence: francesco.ﬁ[email protected]; Tel.: +39-0824-305-195

Academic Editor: Brigitte Helmreich Received: 16 June 2016; Accepted: 21 July 2016; Published: 5 August 2016

Abstract: After more than 350 mm of rainfall fell in a few hours on 22 November 2011, thousands of landslides and floods were induced in two main zones of Northeastern Sicily. The total rainfall has been reconstructed integrating available rain gauge data with Tropical Rainfall Measuring Mission (TRMM) satellite data from NASA (National Aeronautics and Space Administration); the landslide distribution in the field has confirmed the pattern of rainfall accumulated on 22 November 2011. Precipitation maxima of 1, 3, 6, 12, and 24 h was recognized as the hazardous events, which marks the evidence of a changing climate, with a shift toward more intense rainfalls in recent times. To investigate the sequence of the annual maxima, the historical time series have been transformed in the Standard normal distribution, from the cumulative probability of the GEV (Generalized Extreme Value) distribution. Following a similar definition of the Standard Precipitation Index (SPI), the transformation of the historical data in the standardized values allows the definition of categories of hourly maxima in term of extreme, severe, moderate, or mild. This transformation allows to eliminate the asymmetry of the time series, so that trends and fluctuations have been highlighted by the progressive accumulation of data (Rescaled Adjust Partial Sum). This statistical approach allows the improvement of the interpretability of the hydrological extreme events, and could also be used in other cases.

Keywords: rainfall; extreme event; return time; Standard normal distribution; Sicily

1. Introduction Despite hydroclimate millennia reconstruction not supporting a general unprecedented intensiﬁcation of the Northern Hemisphere hydrological cycle in the twentieth century associated with both more extreme wet and dry conditions [1], rainstorms and their injuries are becoming more dangerous as population and infrastructure continue to increase and to occupy areas exposed to ﬂood risk [2,3]. However, few literature sources are available worldwide regarding extreme precipitation and, especially, about rainstorm effects on regional and subregional terrestrial ecosystems and water resources [4–7]. This also poses another question related to the development of dynamic hydrological models, hampered by incomplete understanding of spatially-varying processes and the lack of adequate datasets to spatially characterize varying rain inputs. According to [8], secular and more records of historical precipitation data are required to deal with long-term studies and cross-site comparisons. This is especially so in the Mediterranean region, where human pressure and erratic

Water 2016, 8, 330; doi:10.3390/w8080330 www.mdpi.com/journal/water Water 2016, 8, 330 2 of 15 Water 2016, 8, 330 2 of 15 rainstorm patterns with marked inter-annualinter‐annual variability expose landforms to exacerbated, damaging hydrological processes [[9–12].9–12]. This is also so to spur thethe emergence of new hazards, suchsuch as coastalcoastal and urban floodingflooding [[13,14].13,14]. Mediterranean rainstorms and cyclones tendtend toto be characterized by short lifelife cycles,cycles, with with average average radii radii ranging ranging from from 300 to 300 500 kmto 500 (after km [15 (after]), many [15]), of which many are of a which combination are a ofcombination both frontal of and both convective frontal storms.and convective Heavy flooding storms. and Heavy storms flooding occurring and at storms Mediterranean occurring sites at wereMediterranean found to be sites characterized were found by to abe complex characterized property, by a known complex as property, multifractality, known which as multifractality, is the spatial distributionwhich is the organized spatial distribution into clusters organized of high-rainfall into clusters localized of high cells‐rainfall embedded localized within cells a larger embedded cloud systemwithin a or larger clusters cloud of lower system intensity or clusters [16]. of lower intensity [16]. Sicily has been hit recentlyrecently byby intenseintense rainstorms,rainstorms, which caused floodsfloods and landslides in many places.places. Particularly affected by recent and intense storms has been the northwestern sector, where a high andand continuous continuous ridge, ridge, the the Peloritani Peloritani Mountains, Mountains, extends extends between between the the Ionian Ionian and and Tyrrhenian Tyrrhenian seas. Duringseas. During recent recent years, years, the most the most catastrophic catastrophic event event occurred occurred on 1 Octoberon 1 October 2009 along2009 along the Ionian the Ionian side, withside, morewith thanmore 200 than mm 200 recorded mm recorded in a few in hours, a few and hours, hit the and village hit the of Giampilieri,village of Giampilieri, where landslides where causedlandslides 36 deaths caused [17 36,18 deaths]. After [17,18]. 13 November After 1855,13 November the event of1855, 2009 the was event the most of 2009 catastrophic was the to most have occurredcatastrophic in the to Messinahave occurred province, in the and Messina follows province, other catastrophic and follows events other that catastrophic have occurred events recently. that Inhave particular, occurred on recently. 25 October In 2007 particular, an intense on rainstorm25 October induced 2007 an many intense landslides rainstorm and floods induced along many the Ionianlandslides coast and [19 floods,20]. along the Ionian coast [19,20]. The stormstorm of of 22 22 November November 2011 2011 hit a hit wide a areawide the area Tyrrhenian the Tyrrhenian sector of thesector Peloritani of the mountains, Peloritani andmountains, caused and three caused deaths three in Saponara deaths in village Saponara (Figure village1). It (Figure induces 1). thousands It induces of thousands shallow landslides of shallow inlandslides two distinct, in two wide distinct, areas wide (Figure areas2), (Figure one located 2), one in located the middle in the catchment middle catchment of the Saponara of the Saponara torrent, andtorrent, the otherand the across other both across the catchments both the catchments of Mela and of Longano Mela and torrents. Longano Huge torrents. alluvial Huge phenomena alluvial occurredphenomena in theoccurred urban in area the ofurban Barcellona area of PozzoBarcellona di Gotto Pozzo and di Milazzo.Gotto and Recent Milazzo. landslides Recent landslides and flash floodsand flash hit floods the same hit area the same on 11 area December on 11 December 2008, 2 November 2008, 2 November 2010, and 102010, October and 10 2015. October 2015.

Figure 1.1. (A) The mouth of Longano River on thethe morning of 2323 NovemberNovember 2011,2011, andand thethe bridgebridge collapsed after the the flood; flood; ( (BB)) tree tree trunks trunks obstruct obstruct the the river river flow flow in in town town center center of of Barcellona Barcellona Pozzo Pozzo di diGotto; Gotto; (C ()C debris) debris flows flows hit hit Varela Varela village, village, Mela Mela river river catchment; catchment; and and ( (DD)) a a debris avalanche struck houses in Scarcelli village,village, SaponaraSaponara municipality,municipality, and and caused caused three three deaths. deaths.

Water 2016, 8, 330 3 of 15 Water 2016, 8, 330 3 of 15

Figure 2. The northeastern sector of Sicily, characterizedcharacterized by the PeloritaniPeloritani ridge, extended between the Tyrrhenian andand IonianIonian sea; (1) rain gauge network working and rainfall recorded on 22 November 2011; (2) catchment boundary of the Peloritani ridge, Tyrrhenian side; (3) catchment of Longano (A), (A), Idria (B), Mela (C) and Saponara (D) torrent affected by ﬂoodsfloods on 22 November 2011; and (4) the area hit by landslides onon 2222 November November 2011. 2011. Damage Damage occurred occurred mainly mainly in in the the villages villages of of Barcellona Barcellona Pozzo Pozzo di Gottodi Gotto and and Saponara, Saponara, which which are are highlighted highlighted by by the the white white circles. circles.

In this study we estimate the rainfall that fell on 22 November 2011 by the rain gauge data available, integrated with satellite data; the time of the landslide occurrences has helped in an attempt to deﬁnedefine the main features of thethe storm.storm. The historical time series of the annual maxima of rainfall, for different time duration, have been analyzed to estimate the frequency of extreme rainfall events events and and to to highlight highlight trend trend and and fluctuations. ﬂuctuations.

2. Materials Materials and and Methods Methods

2.1. Main Physical and Geomorphological Characteristics of the Area Hit by the Storm Due toto its its climate, climate, affected affected by long, by drylong, periods dry followedperiods byfollowed heavy rainstorms, by heavy the rainstorms, Mediterranean the regionMediterranean is particularly region is prone particularly to multiple prone damaging to multiple hydrological damaging hydrological events (MDHE; events [21 (MDHE;]) since [21]) it is characterizedsince it is characterized by high land by high vulnerability land vulnerability and poor and vegetation poor vegetation coverage. coverage. From thisFrom point this point of view, of Sicilyview, isSicily one theis one recognized the recognized areas particularly areas particularly sensitive sensitive to changes to in changes extreme in precipitation extreme precipitation events and currentlyevents and threatened currently bythreatened flash-floods. by flash The‐floods. climate The of Sicily,climate described of Sicily, in described detail by in [22 detail], is the by typical[22], is Mediterraneanthe typical Mediterranean climate (Köppen’s climate (Köppen’s Csa type), Csa with type), rainy with winters rainy and winters long, and dry long, periods dry in periods summer. in Insummer. Sicily, remarkableIn Sicily, remarkable differences differences can be recognized can be betweenrecognized the southwesternbetween the southwestern portion, characterized portion, bycharacterized dry sub-humid by dry to semi-aridsub‐humid conditions, to semi‐arid and conditions, the northeastern and the portion, northeastern where portion, sub-humid/humid where sub‐ conditionshumid/humid prevail. conditions The average prevail. annual The average temperatures annual rangetemperatures between range 11 ˝C between and 20 ˝ 11C, with°C and summer 20 °C, maximawith summer higher maxima than 40 higher˝C, while than 40 rainfalls, °C, while mainly rainfalls, concentrated mainly concentrated in autumn andin autumn winter and (Figure winter3), show(Figure mean 3), show annual mean values annual ranging values from ranging 385 mm from up 385 to 1200 mm mmup to and 1200 more. mm and more. The 22 November 2011 storm hit a wide area along the northern sector of the Peloritani Mountains, betweenbetween Barcellona Barcellona Pozzo Pozzo di di Gotto Gotto and and Saponara Saponara villages. villages. The The main main floods floods occurred occurred in the in Longano,the Longano, Idria, Idria, Mela, Mela, and Saponara and Saponara catchments catchments (Figure2 ),(Figure associated 2), toassociated diffuse landslide to diffuse phenomena; landslide phenomena; however, other minor catchments were also affected by flash‐floods and landslides, and the entire area of the Milazzo plain was affected by generalized flood phenomena.

Water 2016, 8, 330 4 of 15 however, other minor catchments were also affected by flash-floods and landslides, and the entire area ofWater the 2016 Milazzo, 8, 330 plain was affected by generalized flood phenomena. 4 of 15

250

50°percentile 200 95°percentile mm

150 rainfall,

100

monthly 50

0 Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec

Figure 3. Seasonal regime of different percentile values for the Messina municipality (mean data, 1965–1994Figure 3. Seasonal period). regime of different percentile values for the Messina municipality (mean data, 1965–1994 period).

The area hit by the storm extends between the Tyrrhenian sea and water divide of the Peloritani The area hit by the storm extends between the Tyrrhenian sea and water divide of the Peloritani Mountains.Mountains. The coastal sector is characterized by a wide flat flat area, constituted by recent (Quaternary) alluvialalluvial and marine marine deposits, deposits, with with ground ground elevation elevation up upto a to few a fewtens tensof meters of meters above above sea level sea (a.s.l.), level (a.s.l.),and includes and includes the Milazzo the Milazzo plain. Inland, plain. ground Inland, elevation ground elevation rapidly increases rapidly increases and substratum and substratum outcrops; ˝ outcrops;here the slope here theangle slope can anglebe higher can bethan higher 45°, thanand characterizes 45 , and characterizes the steep slopes the steep of the slopes Peloritani of the Peloritanimountains. mountains. FollowingFollowing thethe recent recent geological geological map map of of Italy Italy [23 [23],], scale scale 1:50,000 1:50,000 (Map (Map n. n. 600 600 “Barcellona “Barcellona Pozzo Pozzo di Gotto”di Gotto” and and Map Map n. n. 601 601 “Messina “Messina and and Reggio Reggio Calabria”), Calabria”), the the substratum substratum belongs belongs to to a a metamorphic metamorphic complexcomplex ofof PaleozoicPaleozoic age,age, and toto marinemarine sedimentarysedimentary complexes of Tertiary ageage (Flysch(Flysch didi CapoCapo d’Orlandod’Orlando and Argille scagliose), both involved in Tertiary tectonic activity of the Apennine chain. MarineMarine sinorogeneticsinorogenetic depositsdeposits ofof MioceneMiocene (Formation(Formation ofof SanSan PierPier Niceto)Niceto) andand Pliocene-PleistocenePliocene‐Pleistocene (Formation(Formation ofof Rometta)Rometta) lielie onon thethe olderolder terrains,terrains, andand outcropoutcrop betweenbetween thethe coastalcoastal andand inlandinland sectors.sectors. AA complexcomplex ofof normalnormal faultsfaults havehave causedcaused thethe generalgeneral upliftuplift ofof thethe areaarea duringduring thethe PliocenePliocene andand Pleistocene;Pleistocene; marine deposits of thethe lowerlower PleistocenePleistocene actuallyactually lielie atat 560560 mm a.s.l.a.s.l. (Rometta village), testifyingtestifying thethe highhigh upliftuplift raterate duringduring thethe Quaternary. AsAs aa consequence,consequence, thethe hydrographichydrographic networknetwork isis cachedcached inin thethe substratum,substratum, andand determinesdetermines steepsteep andand unstableunstable slopes.slopes. AA diffusediffuse soilsoil mantlemantle coverscovers thethe slopes,slopes, constitutedconstituted byby colluvialcolluvial andand eluvialeluvial depositsdeposits originatedoriginated fromfrom thethe weatheringweathering andand erosionerosion ofof thethe substratum;substratum; thisthis soilsoil mantlemantle constitutesconstitutes thethe mainmain materialmaterial involvedinvolved inin thethe landslidelandslide processes.processes. 2.2. Main Features of the 22 November 2011 Storm 2.2. Main Features of the 22 November 2011 Storm The meteorological conditions leading to 22 November 2011 storm were connected to The meteorological conditions leading to 22 November 2011 storm were connected to an an anomalous barometric gradient occurring between the northern side (Tyrrenian basin) and anomalous barometric gradient occurring between the northern side (Tyrrenian basin) and southwestern side (Ionian sea) of the Peloritani Mountains, inducing high-speed sirocco winds. southwestern side (Ionian sea) of the Peloritani Mountains, inducing high‐speed sirocco winds. These These winds generally cause intense rainfall on the Ionian side of the Peloritani Mountains, where the winds generally cause intense rainfall on the Ionian side of the Peloritani Mountains, where the highest accumulated rainfalls were expected the day before the storm. However, due to the very high highest accumulated rainfalls were expected the day before the storm. However, due to the very high intensity of winds during 22 November 2011, the rainfall accumulated mainly on the other side of the intensity of winds during 22 November 2011, the rainfall accumulated mainly on the other side of the Peloritani water divide, causing unexpected landslides and flash floods along the Tyrrhenian side of Peloritani water divide, causing unexpected landslides and flash floods along the Tyrrhenian side of the Peloritani Mountains. This storm behavior is known locally as the “Alcantara effect”, because the the Peloritani Mountains. This storm behavior is known locally as the “Alcantara effect”, because the wind from the south rises along the valley between Etna volcano and Peloritani Mountains and crosses wind from the south rises along the valley between Etna volcano and Peloritani Mountains and the watershed of Peloritani thanks also to strong winds in altitude [24]. crosses the watershed of Peloritani thanks also to strong winds in altitude [24]. The storm few hours later reached the Southern Calabria, inducing further landsides and flash floods; a train derailed near Lamezia Terme, fortunately without further deaths.

Water 2016, 8, 330 5 of 15

Water The2016, storm8, 330 few hours later reached the Southern Calabria, inducing further landsides and5 ﬂash of 15 ﬂoods; a train derailed near Lamezia Terme, fortunately without further deaths. Figure4 4 shows shows the the developmentdevelopment of of thethe atmosphericatmospheric disturbancedisturbance fromfrom satellite,satellite, where,where, fromfrom 6:00,6:00, a plume beginsbegins toto develop,develop, with with its its vertex vertex on on Northeastern Northeastern Sicily Sicily and and it it spreads spreads to to the the southern southern part part of theof the Italian Italian peninsula. peninsula. The The storm storm has has a typical a typical enhanced-V enhanced‐ shapeV shape from from the the satellite satellite images. images.

Figure 4. SatelliteSatellite imagines imagines (EUMETSAT, (EUMETSAT, IR IR 10.8), 10.8), every every 2 h 2of h the of 22 the November 22 November 2011, 2011, between between 03:00 03:00and 19:00. and The 19:00. white The zones white are zones the colder are the clouds, colder and clouds, the enhanced and the‐V enhanced-V shape develops shape on Northeastern develops on NortheasternSicily. Sicily.

Figure 2 shows the rain gauge network available during the storm and Table 1 the rainfall Figure2 shows the rain gauge network available during the storm and Table1 the rainfall accumulated. The highest rainfall was recorded at Castroreale where 351 mm were recorded; with accumulated. The highest rainfall was recorded at Castroreale where 351 mm were recorded; the exception of Torregrotta, with 188.8 mm, the other rain gauges recorded lower rainfall, especially with the exception of Torregrotta, with 188.8 mm, the other rain gauges recorded lower rainfall, for those located along the Ionian side of the Peloritani Mountains. In Messina town, only a few especially for those located along the Ionian side of the Peloritani Mountains. In Messina town, millimeters of rainfall were recorded. However, the distribution of accumulated rainfall of Figure 5 only a few millimeters of rainfall were recorded. However, the distribution of accumulated rainfall of cannot explain the diffuse landslide phenomena and flash floods occurred in the Saponara catchment, Figure5 cannot explain the diffuse landslide phenomena and flash floods occurred in the Saponara but only that of Longano, Idria, and Mela catchments (cf. Figure 2). This is due to low rain gauge catchment, but only that of Longano, Idria, and Mela catchments (cf. Figure2). This is due to density distribution, which is unable to record the areal peaks distribution of the rainfall, connected low rain gauge density distribution, which is unable to record the areal peaks distribution of the to storm development (Figure 4). Additionally, in the area of Saponara, landslides developed during rainfall, connected to storm development (Figure4). Additionally, in the area of Saponara, landslides the afternoon of 22 November 2011, several hours later than those occurring in the Mela and Longano developed during the afternoon of 22 November 2011, several hours later than those occurring in catchments. These characteristics indicate that storm migrated toward the northeast, according to the the Mela and Longano catchments. These characteristics indicate that storm migrated toward the satellite images. northeast, according to the satellite images. Figure 6 shows the accumulated hourly rainfall recorded at rain gauges during the 22 November Figure6 shows the accumulated hourly rainfall recorded at rain gauges during the 22 November 2011 storm. The Castroreale rain gauge recorded the most intense rainfall; it began to rain since the 2011 storm. The Castroreale rain gauge recorded the most intense rainfall; it began to rain since early hours of 22 November, but between 6:00 and 15:00 the storm caused the major part of the the early hours of 22 November, but between 6:00 and 15:00 the storm caused the major part of the precipitation. In particular, 72.5 mm fell between 7:00 and 8:00 and caused the first flood phenomena precipitation. In particular, 72.5 mm fell between 7:00 and 8:00 and caused the first flood phenomena in the minor gullies. Towards the northeast, (Torregrotta, San Pier Niceto, and Ziriò) the most intense in the minor gullies. Towards the northeast, (Torregrotta, San Pier Niceto, and Ziriò) the most intense rainfall occurred one or few hours later. Many eyewitnesses described that, in the area of Saponara, the major precipitation occurred in the afternoon of 22 November 2011, highlighting that mainly in this zone the rain gauge network was insufficient to catch the spatial distribution of the rainfall.

Water 2016, 8, 330 6 of 15

Table 1. Rainfall recorded on 22 November 2011 by the rain gauges network.

Elevation Accumulated Rainfall Network Rain Gauge (m a.s.l.) between 0:00 and 24:00; (mm) Milazzo 2 90.9 Ziriò—(Saponara) 860 70.9 Castroreale 383 351 OSSERVATORIO Montalbano E. 929 13.6 DELLE ACQUE S.Stefano Briga 139 20 Messina (Ist.Geof.) 50 3 S.Piero Patti 440 3.2 Torregrotta 60 188.8 Montalbano E. 1250 8.4 San Pier Niceto 460 57.8 Water 2016, 8, 330 6 of 15 Antillo 796 116.2 SIAS Novara di Sicilia 750 54.8 rainfall occurred one or few hours later.Patti Many eyewitnesses70 described that, in5.4 the area of Saponara, the major precipitation occurredFiumedinisi in the afternoon of 22440 November 2011, highlighting57.6 that mainly in this zone the rain gauge network wasMessina insufﬁcient to catch230 the spatial distribution11.8 of the rainfall.

Figure 5. 2222 November November 2011 2011 rainfall rainfall reconstructed reconstructed by by kriging kriging interpolation interpolation on on available available rain rain gauge gauge of ofTable Table 1. 1.

Table 1. Rainfall recorded on 22 November 2011 by the rain gauges network.

Accumulated Rainfall Network Rain Gauge Elevation (m a.s.l.) between 0:00 and 24:00; (mm) Milazzo 2 90.9 Ziriò—(Saponara) 860 70.9 Castroreale 383 351 OSSERVATORIO Montalbano E. 929 13.6 DELLE ACQUE S.Stefano Briga 139 20 Messina (Ist.Geof.) 50 3 S.Piero Patti 440 3.2

Torregrotta 60 188.8 Montalbano E. 1250 8.4 San Pier Niceto 460 57.8 Antillo 796 116.2 SIAS Novara di Sicilia 750 54.8 Patti 70 5.4 Fiumedinisi 440 57.6 Messina 230 11.8 Water 2016, 8, 330 7 of 15 Water 2016, 8, 330 7 of 15

FigureFigure 6.6. HourlyHourly rainfallrainfall ofof 2222 NovemberNovember 20112011 forfor severalseveral rainrain gauges.gauges.

2.3. Methods Used 2.3. Methods Used To reconstruct a more likely spatial distribution of rainfall occurring on 22 November 2011, rain To reconstruct a more likely spatial distribution of rainfall occurring on 22 November 2011, gauge density has been integrated with data recorded from the Tropical Rainfall Measuring Mission rain gauge density has been integrated with data recorded from the Tropical Rainfall Measuring satellite (TRMM). TRMM is a satellite launched by NASA for research purposes and provides rainfall Mission satellite (TRMM). TRMM is a satellite launched by NASA for research purposes and data as a function of the heat release which manifests contextually to precipitation. Data download provides rainfall data as a function of the heat release which manifests contextually to precipitation. is through a portal that can be accessed through the NASA website, with an acquisition interval of Data download is through a portal that can be accessed through the NASA website, with an acquisition 95.2 min. Acquisition points of TRMM provide coverage to the entire Earth’s surface, they form a interval of 95.2 min. Acquisition points of TRMM provide coverage to the entire Earth’s surface, regular grid of dots arranged at intervals of 0.25° of latitude and longitude. they form a regular grid of dots arranged at intervals of 0.25˝ of latitude and longitude. Available rain gauge data series on the annual maxima rainfall (1, 3, 6, 12, and 24 h) have been Available rain gauge data series on the annual maxima rainfall (1, 3, 6, 12, and 24 h) have been analyzed to find some statistical characteristics, as the return time and trend in the time series. analyzed to find some statistical characteristics, as the return time and trend in the time series. The return time has been estimated by the Generalized Extreme Value (GEV) [25] frequency The return time has been estimated by the Generalized Extreme Value (GEV) [25] frequency distribution, where the cumulative probability, P, for values below a specific threshold (X ≤ x), is: distribution, where the cumulative probability, P, for values below a specific threshold (X ď x), is: 1    x ´ 1 PX()exp1 x x ´ µ ξ (1) PpX ď xq “ exp ´1` ξ  (1)  σ  # „ ˆ ˙ + wherewhereξ ξ isis the the shape shape parameter parameter (dimensionless), (dimensionless),µ μ andandσ σ areare the the position position and and scale scale parameters. parameters. TheThe actualactual valuesvalues werewere plotted plotted by by their their order order in in the the time time series: series: i (2) P piq “ (2) N1` 1 where N is the number of the series data, and i is the position in the ordered set of data. where N is the number of the series data, and i is the position in the ordered set of data. The cumulative probability found by the GEV frequency distribution (Equation (1)) has been The cumulative probability found by the GEV frequency distribution (Equation (1)) has been transformed in the Z value of the Standard normal distribution (Figure 7). This transformation has transformed in the Z value of the Standard normal distribution (Figure7). This transformation has been been carried out for the annual maxima series of rainfall (1, 3, 6, 12, and 24 h), and obtains the carried out for the annual maxima series of rainfall (1, 3, 6, 12, and 24 h), and obtains the symmetric symmetric series, with zero mean and standard deviation equal to 1. After the transformation in the series, with zero mean and standard deviation equal to 1. After the transformation in the normal normal standardized distribution, the progressive summation of the values highlights trends and standardized distribution, the progressive summation of the values highlights trends and fluctuations fluctuations of the time series; this summation coincides with the Rescaled Adjusted Partial Sum of the time series; this summation coincides with the Rescaled Adjusted Partial Sum (RAPS; [26]). (RAPS; [26]). The RAPS of the time series are defined as follows: The RAPS of the time series are defined as follows: _ kk YY RAPS X Ytt ´ Y (3) RAPSkkk“ Xk “  (3) SS t“t1 YY ÿ1 where is sample mean, SY is the standard deviation, k is counter limit of the current summation. The plot of the RAPS versus time is the visualization of the trends and fluctuations of Yt [26]. Decreasing patterns in the RAPS are the result of periodicity of mostly below‐average Yt values,

Water 2016, 8, 330 8 of 15

where Y is sample mean, SY is the standard deviation, k is counter limit of the current summation. TheWater plot 2016 of, 8, the 330 RAPS versus time is the visualization of the trends and ﬂuctuations of Yt [26]. Decreasing8 of 15 patterns in the RAPS are the result of periodicity of mostly below-average Yt values, whereas increasing patternswhereas areincreasing the result patterns of periods are of the mostly result above-average of periods of Ymostlyt values. above The‐average RAPS highlights Yt values. the The tendency RAPS ofhighlights values to the cluster tendency together; of values that to is, cluster the consecutive together; that values is, the below consecutive or above values the mean below provides or above a decreasingthe mean provides or increasing a decreasing trend in or the increasing RAPS, respectively. trend in the RAPS, respectively.

1 1

0.9 GEV 0.9 ST. NORM. DISTR.

0.8 0.8

0.7 0.7

0.6 0.6

0.5 0.5 probability probability

0.4 0.4

0.3 0.3 cumulative cumulative

0.2 0.2

0.1 0.1

0 0 0 50 100 150 200 250 30 ‐3 ‐2 ‐10123 mm Z value

Figure 7. TransformationTransformation from GEV distribution to Standard normal distribution.

3. Results and Discussion 3. Results and Discussion

3.1. Reconstruction of the Storm Map 3.1. Reconstruction of the Storm Map To improve the rainfall spatial distribution on 22 November 2011, rain gauge data have been integratedTo improve with data the from rainfall the spatial TRMM distribution satellite platform. on 22 November The original 2011, values rain provided gauge data by TRMM have been has integratedbeen multiplied with databe a fromconstant, the TRMM obtained satellite from platform.the ratio between The original rainfall values recorded provided at several by TRMM rain hasgauges been and multiplied their nearest be a TRMM constant, measure obtained points; from the the constant ratio between found has rainfall a value recorded of 6. On at the several network rain gaugesformed andby rain their gauges nearest and TRMM TRMM measure points, points; a new the spatial constant rainfall found distribution has a value has of 6. been On themapped network by formedkriging interpolation. by rain gauges Figure and 8 TRMM depicts points, probably a new a more spatial realistic rainfall spatial distribution distribution has of been rains mapped from 22 byNovember kriging 2011 interpolation. with respect Figure to Figure8 depicts 5; it seems probably to agree a more with realistic what was spatial derived distribution from EUMETSAT of rains fromsatellite 22 images November of Figure 2011 with4, even respect if it is to also Figure based5; iton seems rainfall to data agree derived with what indirectly was derived from satellite from EUMETSATsensors. The satellite map shows images that of Figurethe storm4, even event if it of is 22 also November based on rainfall2011 is data characterized derived indirectly by two fromrainy satellitecenters, sensors.elongated The in mapthe direction shows that of theSouthwest–Northeast, storm event of 22 November one located 2011 transversely is characterized to Longano, by two rainyIdria centers,and Mela elongated catchments, in the directionand the ofother Southwest–Northeast, in the area of Saponara one located catchment. transversely The to Longano,result of Idriainterpolation and Mela of catchments, the data in andFigure the 8, other with in the the two area centers of Saponara of rain, catchment. is in fact the The result result of of the interpolation stationing of the dataplume in vertex Figure 8(mentioned, with the two before) centers for ofmost rain, of is the in facttime the on resultthese ofareas. the stationingIt can be observed of the plume that, vertexin the area (mentioned of Milazzo before) plain for at most least of 200 the mm time of on rain, these would areas. have It can fallen, be observed according that, to in flash the‐ areafloods of Milazzooccurring plain in this at area. least However, 200 mm of the rain, distribution would have of rainfall fallen, accordingin the Saponara to ﬂash-ﬂoods area appears occurring still be incoarse this area.(locally, However, it is also the missing distribution a TRMM of rainfall point, inFigure the Saponara 8), as suggested area appears by the still landslide be coarse distribution (locally, it in is alsothis missingzone (Figure a TRMM 2). point, Figure8), as suggested by the landslide distribution in this zone (Figure2).

Water 2016, 8, 330 9 of 15 Water 2016, 8, 330 9 of 15

Figure 8. RainfallRainfall of of 22 November 2011 reconstructed by kriging interpolation on available rain gauges (1) and TRMM satellite data (2).

3.2. Frequency Analyses SpecificSpecific to the area hit by the 22 November 2011 storm, the rain gauge at Castroreale is the only one where a long historical time series of maxima exists (since 1930), even if some missing data are present. This This rain rain gauge gauge is is located located in in one one of of two two zones zones characterized characterized by bythe the maximum maximum amount amount of the of therainfall rainfall that that fell fellduring during the the storm, storm, as ascan can be be deduced deduced from from Figure Figure 88 and the landslidelandslide distributiondistribution occurring inin the the field field (Figure (Figure2). Thus, 2). Thus, the maximum the maximum intensity intensity of the storm of the can storm be analyzed can be statistically, analyzed andstatistically, provides and details provides on the details recurrence on the of recurrence the intense of storms the intense in this storms specific in point. this specific point. Even if a a considerable considerable accumulated accumulated rainfall rainfall was was also also recorded recorded at at the the Torregrotta Torregrotta rain rain gauge, gauge, it ithas has only only operated operated since since 2002. 2002. Table Table 22 shows shows that that the the rainfall rainfall recorded at thethe CastrorealeCastroreale rainrain gaugegauge during the 22 November 2011 was the highest for each hourly hourly duration duration (from (from 1 1 to to 24 24 hourly hourly rainfall). rainfall). In particular, the rainfall of 6 6 and and 12 12 h h has has been been well well higher higher than than previous previous historical historical maxima maxima (Table (Table 2).2).

TableTable 2. Historical maxima recorded at the Castroreale rain gauge, for different timetime intervals.intervals. Time Series 1 h 3 h 6 h 12 h 24 h Time Series 1 h 3 h 6 h 12 h 24 h period 1930–2010, 65 years (mm) 62.8 124.8 185 222 318 period 1930–2010, 65 years (mm) 62.8 124.8 185 222 318 22 November 2011 storm (mm) 72.5 162.5 254.8 348.2 351 22 November 2011 storm (mm) 72.5 162.5 254.8 348.2 351

Figure 9 shows the statistical analyses carried out on annual maxima for 1, 3, 6, 12, and 24 cumulativeFigure 9rainfalls, shows plotting the statistical results as analyses described carried in [27]. out If annual on annual maxima maxima of 2011 are for excluded 1, 3, 6, from 12, andthe dataset 24 cumulative (Figure 9A),rainfalls, each statistical plotting results curve (GEV) as described fits the historical in [27]. data If annual well; the maxima return of time 2011 of arethe excluded22 November from 2011 the datasetwould reach (Figure values9A), eachof 450–600 statistical years curve between (GEV) three ﬁts and the historical12 h of accumulated data well; therainfall. return time of the 22 November 2011 would reach values of 450–600 years between three and 12 h of accumulatedIf the annual rainfall. maxima of 2011 are included in the dataset (Figure 9B), the statistical curves (GEV) are notIf the able annual to fit the maxima extreme of 2011value are of includedthe series in for the 3, dataset6, and 12 (Figure h cumulative9B), the statisticalrainfall (Figure curves 9B); (GEV) the arereturn not time able of to the ﬁt the22 November extreme value 2011 of storm the series fall to for 160–175 3, 6, and years 12 hbetween cumulative 3 and rainfall 12 h accumulated (Figure9B); therainfall. return Thus, time the of thepresence 22 November of annual 2011 maxima storm of fall 2011 to 160–175in data set years conditions between heavily 3 and 12 the h estimation accumulated of the return time.

Water 2016, 8, 330 10 of 15

Water 2016The, 8exceptional, 330 rainfall of the 22 November 2011 storm caused a considerable shift from10 each of 15 frequency distribution adopted to fit the historical data, and raises the well‐known complication in rainfall.this type Thus, of statistical the presence analysis of annual [28,29]. maxima This aspect of 2011 highlights in data set the conditions outlier characteristic heavily the estimation of the 2011 of theevent, return and time. will be considered later.

Figure 9. 9. ReturnReturn time time for for different different rainfall rainfall durations durations (Castroreale (Castroreale rain gauge). rain gauge). Historical Historical data for data 1, 3, for6, 12, 1, 3,and 6, 12, 24 and h (white 24 h (white and blueand bluecircles)circles) are fitted are ﬁtted by bythe the GEV GEV (cumulative) (cumulative) frequency frequency distribution. distribution. Triangles showshow thethe returnreturn timestimes ofof 2222 NovenberNovenber 20112011 onon thethe GEVGEV curves. (A) Analysis with n.65 data for the 1930–2010 period (without 2011 data); and (B) analysis with n.66 data of the 1930–2011 period (with 2011 data).

3.3. Transformation in the Standard Normal Distribution The exceptional rainfall of the 22 November 2011 storm caused a considerable shift from each frequencyTo investigate distribution the adopted sequence to ﬁtof the the historical annual data,maxima, and raisesthe historical the well-known time series complication have been in thistransformed type of statistical in the Standard analysis [ 28normal,29]. This distribution, aspect highlights from the the outliercumulative characteristic probability of the of 2011 the event,GEV anddistribution will be considered (Figure 10). later. The transformed series of Figure 10 are symmetric with respect to the zero value of the mean and have a standard deviation equal 1.

Water 2016, 8, 330 11 of 15

3.3. Transformation in the Standard Normal Distribution To investigate the sequence of the annual maxima, the historical time series have been transformed in the Standard normal distribution, from the cumulative probability of the GEV distribution (Figure 10). The transformed series of Figure 10 are symmetric with respect to the zero value of theWater mean 2016, 8 and, 330 have a standard deviation equal 1. 11 of 15

4 16 4 16

1-hour 12 3-hour 12

2 8 2 8

4 4

0 0 0 0 RAPS RAPS -4 -4

-2 -8 -2 -8

standardized rainfall rainfall (Z value) standardized -12 rainfall (Z value) standardized -12

-4 -16 -4 -16 1930 1935 1940 1945 1950 1955 1960 1965 1970 1975 1980 1985 1990 1995 2000 2005 2010 1930 1935 1940 1945 1950 1955 1960 1965 1970 1975 1980 1985 1990 1995 2000 2005 2010

4 16 4 16

6-hour 12 12-hour 12

2 8 2 8

4 4

0 0 0 0 RAPS RAPS -4 -4

-2 -8 -2 -8

standardized rainfall rainfall (Z value) standardized -12 rainfall (Z value) standardized -12

-4 -16 -4 -16 1930 1935 1940 1945 1950 1955 1960 1965 1970 1975 1980 1985 1990 1995 2000 2005 2010 1930 1935 1940 1945 1950 1955 1960 1965 1970 1975 1980 1985 1990 1995 2000 2005 2010

4 16

24-hour 12

2 8

0 0 RAPS -4

-2 -8

standardized rainfall (Z rainfall value) standardized -12

-4 -16 1930 1935 1940 1945 1950 1955 1960 1965 1970 1975 1980 1985 1990 1995 2000 2005 2010

FigureTable 10. 10. AnnualAnnual maximum maximum hourly hourly rainfall, rainfall, 1, 1, 3, 3, 6, 6, 12, 12, and and 24 24 h time series ( blue circles), expressed as Z value ofof thethe StandardStandard normalnormal distribution;distribution; progressive accumulation (RAPS) is also plotted (black line).line).

With the exceptionexception of 1978,1978, which appears randomly distributed in the time series, the most intense storms are concentrated in recent years (Figure 10),10), especially considering durations of of 1, 1, 3, and 6 h. h. In In particular, particular, large large storms storms occurred occurred in in2008 2008 and and 2010, 2010, and, and, locally, locally, induced induced landslides landslides and andfloods, floods, but the but rainfall the rainfall accumulated accumulated for different for different durations durations was waslower lower than thanthe 2011 the 2011storm. storm. For Fordurations durations of 12 of and 12 and 24 h, 24 the h, storms the storms of 1943 of 1943 and and 1949 1949 appear appear intense intense in the in thehistorical historical series. series. In the transformed series of Figure 1010,, the values that exceed 2 have a probability of occurrence lower than 0.023; as commonly used in hydrology, for example in the definition definition of drought categories by the SPI (Standard Precipitation Index; [[30]),30]), the values that exceed 2 can be considered as extreme events (in drought analyses the negative values of Z areare considered).considered). Following a similar definitiondefinition of the SPI, but considering positive values of Z in thisthis case, the categories of hourly maxima are defineddefined asas extreme, extreme, severe, severe, moderate, moderate, or mild. or Themild. transformation The transformation in the Standard in the normal Standard distribution normal providesdistribution dimensionless provides dimensionless values, and allowsvalues, a and direct allows comparison a direct betweencomparison the between time series the of time Figure series 10 of Figure 10 (1, 3, 6, 12 and 24 h). Besides, it also allows the comparison of different hydrological parameters, such as spring discharge and rainfall [31]. The transformed historical data of Figure 10 have been plotted in Figure 11, fitted by the Standard normal distribution. Values are defined as extremes for Z ≥ 2, severe for 1.5 ≤ Z < 2, moderate for 1 ≤ Z < 1.5, and mild for 0 ≤ Z < 1. The Z = 2.43 limit has been found following the Chauvenet’s method, and would allow to identify the outliers in the series; values higher than this limit appear shifted from the distribution fit (as the event of 2011 in Figure 9B). It has to be outlined that the transformation in the Standard normal distribution includes the 2011 data and, thus, it covers all of the available dataset.

Water 2016, 8, 330 12 of 15

(1, 3, 6, 12 and 24 h). Besides, it also allows the comparison of different hydrological parameters, such as spring discharge and rainfall [31]. The transformed historical data of Figure 10 have been plotted in Figure 11, fitted by the Standard normal distribution. Values are defined as extremes for Z ě 2, severe for 1.5 ď Z < 2, moderate for 1 ď Z < 1.5, and mild for 0 ď Z < 1. The Z = 2.43 limit has been found following the Chauvenet’s method, and would allow to identify the outliers in the series; values higher than this limit appear shifted from the distribution fit (as the event of 2011 in Figure9B). It has to be outlined that the transformation in the Standard normal distribution includes the 2011 data and, Waterthus, 2016 it covers, 8, 330 all of the available dataset. 12 of 15

1 1 hour 3 hours 0.9 6 hours 12 hours 0.8 24 hours Stand. Norm. Distr.

0.7 outlier, Z ≥ 2.43 0.6 extreme, Z ≥2 0.5 severe, 1.5 ≤ Z < 2 0.4 moderate, 1 ≤ Z < 1.5 0.3

cumulative probability 0.2

0.1

0 -4 -3.5 -3 -2.5 -2 -1.5 -1 -0.5 0 0.5 1 1.5 2 2.5 3 3.5 4

Z value

Figure 11. 11. AnnualAnnual maxima maxima hourly hourly rainfall, rainfall, transformed transformed from from GEV GEV distribution, distribution, fitted fitted by by the the Standard Standard normal distribution. Higher Higher values values of of ZZ areare classified classified following following the the SPI; SPI; the the outlier outlier zone zone has been defineddefined following Chauvenet’s method.

Table 3 shows all the storms of historical series characterized by hourly maxima falling in the Table3 shows all the storms of historical series characterized by hourly maxima falling in the severe and extreme categories. The event of 22 November 2011 is an extreme event for all rainfall severe and extreme categories. The event of 22 November 2011 is an extreme event for all rainfall durations, and it is an outlier for 3, 6, and 12 h. For a duration of 6 h, the storm of 2008 was also of durations, and it is an outlier for 3, 6, and 12 h. For a duration of 6 h, the storm of 2008 was also of the the extreme type. In particular, the recent storms of 2008 and 2010 are extreme/severe events for extreme type. In particular, the recent storms of 2008 and 2010 are extreme/severe events for several several hourly durations, and were induced by similar meteorological conditions as the 2011 storm. hourly durations, and were induced by similar meteorological conditions as the 2011 storm.

Table 3. Z values reached for the most intense storm recorded. Table 3. Z values reached for the most intense storm recorded. Categories of Maxima Storm Year 1 h 3 h 6 h 12 h 24 h Categories of Maxima Storm Year 11978 h 3– h 2.04 6– h – 12– h 24 h Extreme 1978Z ≥ 2 –2008 2.04– – 2.05 – – –– – Extreme Z ě 2 2008 – – 2.05 – – 2011 2.212011 2.21 2.47 2.47 2.52 2.53 2.532.13 2.13 1943 –1943 – – – 1.56 1.561.66 1.66 1949 –1949 – – – 1.83 1.831.97 1.97 1978 1.911978 1.91 – – 1.98 – – – – Severe 1.5 ď Z < 2 1995 1.74 – – – – Severe 1.52004 ≤ Z < 2 1.821995 1.74 – – – – –– – 2008 –2004 1.82 1.97 – – – 1.80– – 2010 1.61 1.82 1.86 – 1.75 2008 – 1.97 – 1.80 – 2010 1.61 1.82 1.86 – 1.75

Water 2016, 8, 330 13 of 15

The 1943 and 1949 storms were severe only for 12 and 24 h, indicating that the hourly distribution of these storms were characterized by the absence of particular peaks. In contrast, some recent storms (1995 and 2004) were severe only for a one hour duration, and they lost their power for a longer duration. As the annual maximum hourly rainfall time series have been standardized, their progressive accumulation provides the RAPS of the series (Figure 10). Looking at the RAPS of the annual series of Figure 10, the clustering appears in two main periods, the first occurred between 1965 and 1974 with values below the mean; the second period has occurred since 2007 and it is characterized by values well above the mean. The latter clustering highlights the general increasing of the occurrence of the intense storms, which seems common in all northern parts of Sicily. This is in agreement with the results of [32], by which sub-grid scale convection and intensification phenomena indicate, for Sicily, an increasing trend towards shorter rain durations (about one hour) during the period 1927–2004. Additionally, the increase in the intense precipitation has been recently highlighted in Western Sicily by [33]. The recent increasing of the intense rainfall, marked by the inconstant mean and variance in time series, highlights the difficulties in the estimation of the frequency of recent storms.

4. Conclusions Placed at the crossroads of Mediterranean flows, Sicily could be considered representative of a geographic scenario of regional climate change associated with warming, reflected in the Mediterranean cyclogenesis, and that would have contributed to the exceptional rainfall rates observed in recent times [2,15]. The assessment of current and future management systems should focus on the hazard of extreme precipitation events, which likely cause accelerated urban and rural land degradation. This must be considered together with the findings that, although rainfall amounts are not always increasing, erratic spatial and temporal storm patterns in some seasons or months drive the power of rainfall to increase its hazard. According to [34], investigations are needed especially in terms of statistical analysis of long pluviometric series, rain intensity and erosivity occurring at decadal and interdecadal scales, which mark substantial climatic fluctuations and changes during last centuries. However, the increase of extremes of precipitation cannot be assumed as a rule in Southern Italy, and in the Calabria region this increase was not observed [35]. Specific to the investigated area, on the basis of the landslide distribution in the field, and on the numerous eyewitnesses describing the storm of 22 November 2001, it clearly appears that the gauge network is insufficient to define the map of intense storms. Thus, some historical storms could be unrecorded or underestimated by the actual rain gauge network, as well as floods occurring in the ungauged catchment of Mela and Saponara. Historical data available for the Castroreale rain gauge, the only one useful for a statistic, have shown the intensification of the powerful storms in this area, as occurred in 2008, 2010, 2011, and 2015. This intensification could be connected to new and different atmospheric circulation, and the consequent develop of the “Alcantara effect” in this area; before 2008, on the basis of the historical records, this phenomenon seems to have occurred more rarely in the past. To estimate the probability of occurrence of intense storms, the transformation of the historical data in the standardized values, passing through their best distribution fit (GEV in our case), has allowed to define several categories of the intensity of storms, similarly to SPI. Even if the effect of storms in the field (floods and landslides) strongly depends on the previous wet conditions of the soil that is on the antecedent rainfall [36], the power of storms appear useful to be expressed in term of the Z value. This approach can help in the evaluation of extreme data, as the use of the Standard normal distribution allows a more simple statistical management of data. Trends and fluctuations in time series need specific considerations, because the effect of the recent increase in frequency of intense storms could cause an overestimation of the return time values. Water 2016, 8, 330 14 of 15

Acknowledgments: Authors are grateful to anonymous reviewers for helpful advises. Rainfall data were provided by Osservatorio delle Acque and Servizio Informativo Agrometeorologico Siciliano (SIAS), Regione Sicilia. Author Contributions: Francesco Fiorillo has organized the research and carried out the statistical analyses; Francesco Fiorillo and Nazzareno Diodato have written the manuscript; Massimiliano Meo has developed the GIS maps. Conﬂicts of Interest: The authors declare no conﬂict of interest.

References

1. Ljungqvist, F.C.; Krusic, P.J.; Sundqvist, H.S.; Zorita, E.; Brattström, G.; Frank, D. Northern Hemisphere hydroclimate variability over the past twelve centuries. Nature 2016, 532, 94–98. [CrossRef][PubMed] 2. Diodato, N.; Bellocchi, G. Storminess and Environmental Changes in the Mediterranean Central Area. Earth Interact. 2010, 14, 1–16. [CrossRef] 3. Cevasco, A.; Diodato, N.; Revellino, P.; Fiorillo, F.; Grelle, G.; Guadagno, F.M. Storminess and geo-hydrological events affecting small coastal basins in a terraced Mediterranean environment. Sci. Total Environ. 2015, 532, 208–219. [CrossRef][PubMed] 4. Diodato, N. Spatial uncertainty modeling of climate processes for extreme hydrogeomorphological events hazard monitoring. J. Environ. Eng. ASCE 2006, 132, 1530–1538. [CrossRef] 5. Clarke, M.; Rendell, H. Climate, extreme events and land degradation. In Climate and Land Degradation; Sivakumar, M.V.K., Ndiang’ui, N., Eds.; Springer: Berlin, Germany, 2007; pp. 137–152. 6. Curtis, S.; Salahuddin, A.; Adler, R.F.; Huffman, G.J.; Gu, G.; Hong, Y. Precipitation extremes estimated by GPCP and TRMM: ENSO relationships. J. Hydrometeorol. 2007, 8, 678–689. [CrossRef] 7. Zolina, O.; Simmer, C.; Belyaev, K.; Kapala, A.; Gulev, S. Improving estimates of heavy and extreme precipitation using daily records from European rain gauges. J. Hydrometeorol. 2009, 10, 701–716. [CrossRef] 8. Wei, W.; Chen, L.; Fu, B. Effects of rainfall change on water erosion processes in terrestrial ecosystems: A review. Prog. Phys. Geogr. 2009, 33, 1–12. [CrossRef] 9. Cannarozzo, M.; Noto, L.V.; Viola, F. Spatial distribution of rainfall trends in Sicily (1921–2000). Phys. Chem. Earth 2006, 31, 1201–1211. [CrossRef] 10. Sanchez, J.; Gallardo, E.C.; Gaertner, M.; Arribas, A.; Castro, M. Future climate extreme events in the Mediterranean simulated by a regional climate model: A first approach. Glob. Planet. Chang. 2004, 44, 163–180. [CrossRef] 11. Van Rompaey, A.; Bazzoffi, P.; Jones, R.J.A.; Montanarella, L. Modeling sediment yields in Italian catchments. Geomorphology 2005, 65, 157–169. [CrossRef] 12. Diodato, N.; Bellocchi, G.; Romano, N.; Guadagno, F.M. Modelling the Rainfall Erosivity of the Rhone Region Associated with Climate Variability and Storminess. Adv. Meteorol. 2016, 21, 7626505. 13. Papathoma, M.; Dominey-Howes, D. Tsunami vulnerability assessment and its implications for coastal hazard analysis and disaster management planning, Gulf of Corinth, Greece. Nat. Hazards Earth Syst. Sci. 2003, 3, 733–747. [CrossRef] 14. Barroca, B.; Bernardara, P.; Mouchel, J.M.; Hubert, G. Indicators for identification of urban flooding vulnerability. Nat. Hazards Earth Syst. Sci. 2016, 6, 553–561. [CrossRef] 15. Lionello, P., Bhend, J., Buzzi, A., Della-Marta, P.M., Krichak, S.O., Jansà, A., Maheras, P., Sanna, A., Trigo, I.F., Trigo, R., Eds.; Cyclones in the Mediterranean region: Climatology and effects on the environment. In Mediterranean Climate Variability; Elsevier: Amsterdam, The Netherlands, 2006; pp. 325–372. 16. Mazzarella, A. Multifractal dynamic rainfall processes in Italy. Theor. Appl. Climatol. 1999, 63, 73–78. [CrossRef] 17. Aronica, G.T.; Brigandí, G.; Morey, N. Flash floods and debris flow in the city area of Messina, north-east part of Sicily, Italy in October 2009: The case of the Giampilieri catchment. Nat. Hazards Earth Syst. Sci. 2012, 12, 1295–1309. [CrossRef] 18. Schilirò, L.; Esposito, C.; Scarascia Mugnozza, G. Evaluation of shallow landslide-triggering scenarios through a physically based approach: An example of application in the southern Messina area (northeastern Sicily, Italy). Nat. Hazards Earth Syst. Sci. 2015, 15, 2091–2109. [CrossRef] Water 2016, 8, 330 15 of 15

19. Aronica, G.T.; Brigandì, G.; Marletta, C.; Manfre, B. Hydrological and hydraulic analysis of the flash flood event on 25 October 2007 in North-Eastern part of Sicily, Italy. In Proceedings of the Floodrisk 2008, Oxford, UK, 30 September–2 October 2008. 20. Cama, M.; Lombardo, L.; Conoscenti, C.; Agnesi, V.; Rotigliano, E. Predicting storm-triggered debris flow events: Application to the 2009 Ionian Peloritan disaster (Sicily, Italy). Nat. Hazards Earth Syst. Sci. 2015, 15, 1785–1806. [CrossRef] 21. Petrucci, O.; Polemio, M. The use of historical data for the characterisation of multiple damaging hydrogeological events. Nat. Hazards Earth Syst. Sci. 2003, 3, 17–30. [CrossRef] 22. Drago, A.; Cartabellotta, D.; Lo Bianco, B.; Lombardo, M. Atlante Climatologico della Regione Siciliana; Assessorato Regionale Agricoltura e Foreste, UO di Agrometeorologia: Palermo, Iatly, 2000. 23. Istituto Superiore per la Protezione e la Ricerca Ambientale (ISPRA). Geological Map of Italy, 1:50,000 Scale. Istituto Superiore per la Protezione e la Ricerca Ambientale (ISPRA): Rome, 2009. Available online: http://www.isprambiente.gov.it/Media/carg/ (accessed on 16 June 2016). 24. Ingemi, D. Alluvione nel Messinese Tirrenico: Analisi di un Fenomeno Eccezionale Provocato Dall’effetto “Alcantara-Agrò”. 2011. Available online: http://www.meteoweb.eu/2011/11/alluvione-nel-messinese- tirrenico-analisi-di-un-fenomeno-eccezionale-provocato-dalleffetto-alcantara-agro/99210/ (accessed on 17 July 2016). 25. Jenkinson, A.F. The frequency distribution of the annual maximum (or minimum) values of meteorological elements. Q. J. R. Meteorol. Soc. 1955, 81, 158–171. [CrossRef] 26. Garbrecht, J.; Fernandez, G.P. Visualization of trends and fluctuations in climatic records. Water Resour. Bull. 1994, 30, 297–306. [CrossRef] 27. Fiorillo, F.; Guadagno, F.M.; Aquino, S.; De Blasio, A. The December 1999 Cervinara landslides: Further debris flow in the pyroclastic deposits of Campania (Southern ltaly). Bull. Eng. Geol. Environ. 2001, 60, 171–184. [CrossRef] 28. McCuen, R.H. Modelling Hydrologic Change; Lewis Publishers, CRC Press, Taylor & Francis Group: Boca Raton, FL, USA, 2003; p. 434. 29. Battaglia, F.; Orfei, L. Outlier detection and estimation in nonlinear time-series. J. Time Ser. Anal. 2005, 26, 107–121. [CrossRef] 30. McKee, T.B.; Doesken, N.J.; Kleist, J. The relationship of drought frequency and duration to time scales. In Proceedings of the 8th Conference on Applied Climatology, Anaheim, CA, USA, 17–22 January 1993; American Meteorological Society: Boston, MA, USA, 1993; pp. 179–184. 31. Fiorillo, F.; Guadagno, F.M. Karst Spring Discharges Analysis in Relation to Drought Periods, Using the SPI. Water Resour. Manag. 2010, 24, 1867–1884. [CrossRef] 32. Bonaccorso, B.; Cancelliere, A.; Rossi, G. Detecting trends of extreme rainfall series in Sicily. Adv. Geosci. 2005, 2, 7–11. [CrossRef] 33. Notaro, V.; Liuzzo, L.; Freni, G.; La Loggia, G. Uncertainty Analysis in the Evaluation of Extreme Rainfall Trends and Its Implications on Urban Drainage System Design. Water 2015, 7, 6931–6945. [CrossRef] 34. D’Asaro, F.; D’Agostino, L.; Bagarello, V. Assessing changes in rainfall erosivity in Sicily during the twentieth century. Hydrol. Process. 2007, 21, 2862–2871. [CrossRef] 35. Polemio, M.; Petrucci, O. The occurrence of floods and the role of climate variations from 1880 in Calabria (Southern Italy). Nat. Hazards Earth Syst. Sci. 2012, 12, 129–142. [CrossRef] 36. Fiorillo, F.; Wilson, R.C. Rainfall induced debris flows in pyroclastic deposits, Campania (southern Italy). Eng. Geol. 2004, 75, 263–289. [CrossRef]

© 2016 by the authors; licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC-BY) license (http://creativecommons.org/licenses/by/4.0/).