A database on flash flood events in Campania , southern Italy , with an evaluation of their spatial and temporal distribution

This study presents an historical database of flash flood events in the Campania region of southern Italy. The study focuses on small catchments characterized by intermittent flow, generally occurring during and after heavy rainstorms, which can be hydrologically defined as small Mediterranean catchments. As the outlet zones of these catchments (consisting mainly of alluvial fans or fan deltas) are highly urbanized in Campania, the population living in the delivery areas is exposed to high risk. Detailed scrutiny and critical analysis of the existing literature, and of the data inventory available, allowed us to build a robust database consisting of about 500 events from 1540 to 2015, which is continuously updated. Since this study is the first step of a longer project to perform a hazard analysis, information about time and site of occurrence is known for all events. As for the hazard analysis envisaged, collecting information about past events could provide information on future events, in terms of damage and also spatial and temporal occurrence. After introducing the issue of flash floods in Italy we then describe the geological and geomorphological settings of the study area. The database is then presented, illustrating the methodology used in collecting information and its general structure. The collected data are then discussed and the statistical data analysis presented.


Introduction
Italy as a country is highly exposed to a variety of geological hazards, whose occurrence reaps heavy casualties every year.In such a context, geohydrological disasters, including all types of slope movements and floods, are undoubtedly among the most frequent, and probably those causing most impact on the built environment.Assessment of the risk related to landslides and floods in Italy, covering the time span from 1850 to 2008, shows that many regions are seriously affected by such geohydrological disasters, with the largest landslide risk in Trentino Alto Adige and Campania and the highest flood risk in Piedmont and Sicily (Salvati et al., 2010).However, discriminating among these different processes, especially in regard to past events, is a difficult task due to the lack of specific terminology and/or to very generic descriptions of the phenomena and their effects (Guzzetti and Tonelli, 2004).This may, as a direct consequence, lead to great uncertainty in the reconstruction of the flood history of a region.
A flash flood is a flood caused by heavy or excessive rainfall in a short period of time.Such floods are localized hydrological phenomena, occurring in small catchments of a few to a few hundred square kilometers, with response times typically being a few hours or less (Borga et al., 2007).Flash floods are typical of small Mediterranean catchments (SMCs) that have three main features: limited water resources, dry summers, and high-intensity rainfall events (Merheb et al., 2016).The Mediterranean climate is associated with intense rainstorms; convective thunderstorms are frequently less than 10-14 km in diameter and result in highly concentrated local rainfall events varying considerably both spatially and temporally (Perrin et al., 2009).In such contexts, after heavy rains, the main stream may trans-C.Vennari et al.: A database on flash flood events in Campania, southern Italy port down-valley mixtures of water and sediment in varying proportions, which play an important role in the behavior and hazard of the resulting flows.Three types of flows can be differentiated on the basis of the different amount of sediment and of the flow behavior: water flow, hyperconcentrated flow, and debris flow (Pierson and Costa, 1987;Costa, 1988;Komar, 1988;Iverson and Vallance, 2001;Pierson, 2005).During a flash flood the sediments captured by the stream and transported down valley generally derive from the bed and banks of the torrent and may eventually contribute to build alluvial fans or fan deltas.Another important aspect is that in small catchments a substantial difference of magnitude exists between low water periods, with a seasonal lack of surface water within ephemeral streams, and flood events with the power to make substantial changes to riverbed topography (Kirkham et al., 2000;Coe et al., 2003).
Flash floods are a frequent natural hazard in many parts of Europe (Montz and Gruntfest, 2002;Gaume et al., 2009Gaume et al., , 2014;;Marchi et al., 2010), including Italy.Due to the particular orography and climate of the country, they occur in many different settings, from mountain valleys (Crosta and Frattini, 2004;Tropeano and Turconi, 2004;Gaume et al., 2009) to coastal and inland plains (Esposito et al., 2011;Santangelo et al., 2011;Porfido et al., 2013), in volcanic areas (Alessio et al., 2013), and in semiarid and/or karst environments (Parise, 2003;Cossu et al., 2007;Delle Rose and Parise, 2010).Flash floods generally occur in ungauged watersheds where the lack of information on precipitation and discharge is significant due to the lack of spatially well-distributed rain or flow data.Hence they often remain poorly documented phenomena (Gaume et al., 2009;Ruiz-Villanueva et al., 2010) and, despite their widespread occurrence, they are often described together with landslides and floods in alluvial plains, making it no easy task to distinguish the various processes occurring from documents and reports.This holds especially when examining events occurring at various times in the past in which recorded damage is generally attributed to landslides.
The study area of this paper is situated in Campania, in the Southern Apennines of Italy, a region which in recent decades has been affected by severe flash floods with serious damage and fatalities (Calcaterra et al., 2000(Calcaterra et al., , 2003;;Santo et al., 2002Santo et al., , 2012Santo et al., , 2015;;Del Prete and Mele, 2006;Santangelo et al., 2006Santangelo et al., , 2011Santangelo et al., , 2012;;Chirico et al., 2012;Alessio et al., 2013).The focus of the study consisted of small catch- ments with intermittent flow, generally occurring during and after heavy rainstorms, which can be hydrologically defined as SMCs.All these basins are smaller than 10 km 2 , are characterized by low concentration times (from 30 min to several hours; Santo et al., 2002;Santangelo et al., 2012;Scorpio et al., 2016), and are highly prone to flash flood events.Aiming at collecting all the available information on floods occurring in such catchments, we consulted and scrutinized a variety of sources and scientific papers, with a view to building a reliable catalogue of these events.
As the first step in the process of hazard evaluation, we compiled the database with the following main aims: (i) to identify over the whole region the areas most susceptible to flash floods and (ii) to discriminate whenever possible, in the available literature, flash floods in small catchments from floods in alluvial plains and from gravity processes, such as rapid earth or debris flow (Del Prete et al., 1998;Crosta and Dal Negro, 2003;Guadagno et al., 2003Guadagno et al., , 2005;;Revellino et al., 2004;Zanchetta et al., 2004b;Di Crescenzo and Santo, 2005;Cascini et al., 2008b).Our database is not aimed at a hydrological characterization of the study areas due to lack of hydrological data, especially as concerns the historical events.Nevertheless, given the lack of similar catalogues for the Southern Apennines of Italy, it might be useful in different ways for land use and civil protection planning.In detail, it may help in the selection of sites where monitoring procedures and/or prevention and mitigation works need to be adopted.In this perspective the main users of such data will be local administrators and the civil protection agency.

Study area
The region of Campania in southern Italy extends from the Tyrrhenian Sea to the Southern Apennine Chain, covering about 13 500 km 2 (Fig. 1).The orographic setting is characterized by the presence of a central mountain ridge made up mainly of Mesozoic carbonates, elongated for more than 200 km in a NW-SE direction, with maximum peaks reaching 2000 m a.s.l.(above sea level).On the western side the chain is bounded by a deep (up to 2 km) coastal graben originated by Plio-Quaternary extensional tectonics, which was filled by marine/transitional sedimentary successions and is now occupied by wide flat coastal plains (Ascione et al., 2008).During the late Quaternary strong volcanic activity was registered in the coastal plain with the growth of the Somma-Vesuvius and the Campi Flegrei volcanoes (Romano et al., 1994).The landscape of the western portion of Campania is thus characterized by a wide flat area with isolated volcanic reliefs and islands.On the eastern side of the region the carbonate ridges transition to hilly landscapes of lower elevation, made up mainly of Miocene and Pliocene flysch successions.In this general orographic setting, SMCs are a widespread geomorphic unit along the flanks of the main carbonate ridges, as well as the slopes of the volcanoes, where they have higher longitudinal gradients.In the remaining hilly part of the region, stream catchments present lower mean longitudinal profiles, and wide alluvial plains linked to perennial river systems prevail.
The general climate is humid temperate with mean annual precipitation ranging from 1000 to 2000 mm.In short, the main situations responsible for abundant rains over the region are generally northwesterly or westerly winds bringing eastward-moving cyclonic depressions.Due to the rugged topography of the region, heavy convective precipitation often results in flash floods, with concomitant widespread landslides.

Data collection
The main problem in searching for information about flash floods in Italy is that they are usually grouped with landslides and floods.In most cases the databases refer to landslides or floods or put different types of events together (landslides, flash floods in small catchments, floods in alluvial plains).
As before stated, in Campania a regional-scale database on flash floods is not available.For this reason, first of all we consulted the most complete archive of landslides and floods produced in Italy, the AVI project (an Italian acronym for Aree Vulnerate in Italia, areas affected by landslides or floods in Italy).It was commissioned by the Minister for Civil Protection to the National Group for the Prevention of Hydrogeologic Hazards of the Italian National Research Council (CNR) with a view to compile an inventory of information on areas historically affected by landslides and floods in the country (Guzzetti et al., 1994).We scrutinized the AVI archive to extract information about flash floods in small basins in the region.Since the archive is divided into landslides or floods, careful research was essential.Following this scrutiny, we also consulted and critically analyzed the existing literature related to flood events in Campania.In order to discriminate flash floods from other hydrogeological events, we first adopted a "geographical" approach.Based on the evidence that all the recently occurring and well-documented flash flood events (i.e., with accurate event description and availability of rainfall data; Alessio et al., 2013;Chirico et al., 2012;Santangelo et al., 2011Santangelo et al., , 2012;;Santo et al., 2002Santo et al., , 2012Santo et al., , 2015;;Scorpio et al., 2016) affected SMCs, we catalogued all the events recorded in the outlet zone of this kind of hydrographic basin as "flash floods".
According to the above approach we distinguished the following four event types: floods in alluvial plains flash floods in SMCs landslides mixed and/or doubtful cases.
All the hydrogeological events that may be easily discriminated, based on general description and geographic location, were included in the database.For mixed and doubtful cases a second analysis was carried out, searching and consulting additional and new sources of information, in the attempt to classify them into one of the four types or to exclude them from the database.Once all the events were grouped into different types, we ruled out classes 1 and 3 for further study and focused on class 2, which represented 56 % of the collected data (Fig. 2).We thus obtained a database of about 500 flash floods occurring in small catchments in Campania (Table 1).Unfortunately, information about damage and rainfall was not available in all cases.That said, the available chronicles and historical descriptions testify to sudden events, lasting just a few hours.Whenever possible, we tried to link the documented events to hourly rainfall data obtained from the Campania Weather Forecasting Center (CFDC) (Fig. 3a) and rainfall data from the literature (Fig. 3b), obtaining a total record of 34 events.In all cases hourly rainfall values were very close to or greater than 30 mm h −1 , and the daily rainfall obtained from the Italian Hydrologic and Oceanographic Service (SIMN) (Fig. 3c) was generally close to or greater than 100 mm.These values are generally associated with high-intensity storms (Santo et al., 2012(Santo et al., , 2015) ) and in the context of small catchments like those under study, characterized by very low concentration times (see Sect. 3.2), may be responsible for flash flood occurrence.
The repetitiveness of flash floods in the same area was also investigated.As shown in Fig. 7, among the 86 municipalities hit, 16 recorded more than 10 events.Nine out of these 16 municipalities were located at the outlet zone of coastal carbonate catchments (Sorrento Peninsula-Lattari ridge), whilst  the others are in the piedmont areas of inland carbonate massifs (Picentini and Matese mountains) or in the volcanic area of Mt.Somma-Vesuvius.In all these cases the mean time period of occurrence of the events is very low, ranging from 39 to 3 years, with a mean value of 15 (Fig. 8).If only the most damaging events are taken into account, the mean time period of occurrence increases to 50 years.

Geomorphological features of the catchments involved
The SMCs in Campania most severely affected by flash floods in recent decades are mainly located in carbonate and volcanic settings (Palma et al., 2009;Santangelo et al., 2012;Santo et al., 2012Santo et al., , 2015;;Alessio et al., 2013).Regarding their geographical location, they may have both an inland outlet (generally represented by a fan or by a well-defined foothill area; Santangelo et al., 2012) or a coastal outlet, represented by a fan delta (Esposito et al., 2011;Santo et al., 2012).They also show similar morphometric conditions (Santangelo et  V = number of victims ( * = total number of victims per event on the same date); R = references (see Table 2).
al., 2012; Alessio et al., 2013), which can be summed up as follows: limited catchment area (from a few km 2 to 10 km 2 ) high relief energy (from hundreds of meters up to 1000 m) high slope gradient (generally greater than 35 • ) high mean gradient of feeder channel (greater than 15 • ) low concentration time (from 30 min to a few hours).
As the sediments captured by the stream and transported down valley generally come from the bed and banks of the torrent, we tried to discriminate among different catchment types, based on the following parameters: bedrock of the catchment (carbonate/volcanic); presence/absence of detrital cover and, if present, its origin (weathered bedrock, soil, pyroclastic cover, etc.); type of outlet zone (alluvial fan or fan delta in coastal area).
Thus the collected events were eventually grouped into the following five classes: carbonate catchment with pyroclastic cover carbonate catchment without pyroclastic cover carbonate catchment with pyroclastic cover and outlet to the sea volcanic catchment volcanic catchment with outlet to the sea.

Results
The database contains about 500 events (Table 1) and is being continually updated.Each event is identified by an ID and by its geographical coordinates (events are located with Google Earth © ).In order to include an event in our database, knowledge of its location and its temporal occurrence was essential.The events were located as points, usually at the fan apex or where the main damage was recorded.The catchments were catalogued according to their size (column CE in Table 1) and, as shown in Fig. 4, 67 % of the events in question affected catchments smaller than 3 km 2 , while 91 % of the catchments were smaller than 10 km 2 .Particular attention was given in the database to damage type (column D  in Table 1) and to the presence of victims, if reported (column V in Table 1).The documented events caused serious social damage, primarily involving buildings, infrastructures, and lifelines (roads, pipelines, etc.).Damage to the population includes the number of victims, injured, homeless, and missing people.In several cases the information source does not allow casualties to be quantified with precision, since the information reported is typically generic, such as "some victims" or "several victims".Thus, in order not to loose any useful data, information about damage is also reported as text, as in the sources.Further, additional useful information can be included in the field notes.
The information thereby gathered may be useful to depict a regional scenario on the spatial and temporal distribution of flash floods in Campania.The oldest event occurred in 1540 in the town of Amalfi.The temporal distribution of the events in the region (Fig. 5) reveals that most of the municipalities have been affected by flash floods more than once during the time period covered by the database.Although there are several catchments with similar geological and geomorphological characteristics, many villages have no event recorded, probably related to the absence of inhabited areas.More than 60 % of the events have occurred during the last 50 years, a finding which is probably related to the greater availability of information sources in recent times, the numerous scientific studies carried out in the last few decades and the growing attention to geological hazards.Five degrees of temporal accuracy were used to classify the different levels of knowledge for time occurrence of the events.Events for which only the year of occurrence is known have a low accuracy, while a high accuracy is assigned when hour, day, month, and year of occurrence are known.The histogram in Fig. 6 indicates that most of the events in question have a medium-to-high accuracy, meaning that day, month, and year of occurrence are known.The accuracy degree declines, as expected, for the oldest events.
Regarding the geomorphological features of the catchments involved, Fig. 9 shows the regional distribution of events, classified according to the five classes mentioned above.Most of the events took place in carbonate catchments with pyroclastic cover, both with and without an outlet to the sea (Figs. 9 and 10a).The widespread presence of carbonate catchments, as well as the high urbanization, affected the information source: as mentioned above, intense urbanization typically means greater availability of data, due to higher attention towards the damage generated by natural hazards.Distinguishing between different catchment types is important in order to discern the different bed and bank materials available which the flow could entrap and transport down valley.In a typical carbonate catchment the surge could generally take gravels without or with a low content of matrix (Fig. 11a); in carbonate catchments with pyroclastic cover, however, medium and coarse gravels with a high percentage of matrix are expected to be available (Fig. 11b).Further, in pure volcanic catchments, due to greater erodibility of the material, mainly silts and clays can be found (Fig. 11c).The mean time period of occurrence of the events in each catchment type ranges from 3 to 35 years.Carbonate catchments with pyroclastic cover, both with and without an outlet to the sea, show a mean time period of occurrence of 6 years, which becomes 10 years for carbonate catchments without pyroclastic cover.Regarding the volcanic catchments, which in general present a lower number of events, the mean time period of occurrence is 35 years at Ischia and 3 years for the area surrounding Mt.Somma-Vesuvius.
The monthly distribution of flash flood events is quite variable in the different types of catchments.Figure 10b shows that all the events peak in October, with autumn being the season with the highest frequency.After the dry period, heavy rainfall can generate sudden high discharge, and runoff can carry downstream sediments accumulating as a result of the erosion processes.Locally, this could also be related to occurrence of wildfires during the dry season (Calcaterra et al., 2007a, b).
On the basis of the collected data, in the carbonate catchments with pyroclastic cover the lowest number of events is recorded in April, or generally during the spring, if there is an outlet to the sea.In the carbonate catchments without cover during spring (March-May) there are no events, and the lowest number of alluvial events is in January.In volcanic catchments, February and March are the months with the smallest number of events.For catchments with an outlet to the sea, during the spring (March-May) no event was recorded, and November-December are the months with the fewest events.
Regarding damage to people, in most cases the reported information is very generic, or in other cases the source provides the total number of victims per event but not for the individual municipalities.A precise estimate of the number of victims is thus not easy.In some cases divergences exist in the numbers of casualties reported by different sources for the same event.The differences could be due to several reasons, including the fact that the exact number is typically available only at the end of the search and rescue operations, from a few days to weeks after the event.During this period, newspapers and even official reports may provide different and changing data (Salvati et al., 2010).Typically, information sources mainly document the most severe events in terms of the number of deaths and damage caused.
About 18 % of the events in the database caused at least one victim.Figure 12 depicts their distribution in the region.The most dangerous events hit the province of Salerno, affecting the carbonate catchments with pyroclastic cover, both with and without an outlet to the sea.This means that carbonate catchments with pyroclastic cover (class 1) are the most hazardous.All the events with more than 100 victims took place in October; further, for the most damaging events the total number of deaths was also caused by landslides.Hence it is not easy to evaluate the victims caused only by flash flood events in small catchments.Further, attribution of the number of casualties to each municipality was very difficult, since the source typically provides the total number of victims per event, not distinguishing the different villages.This applies, for instance, to the events in 1581 (700-1000 victims), in 1899 (approximately 100 victims), in 1924 (approximately 100 victims), and in 1910 (over 200 victims) which were the most devastating in recorded history.
The mean time period of occurrence for the events with more or less than 100 victims is very different: 62 years for the events with more than 100 victims and only 3 years for all the other events.
Taking into account the total number of events, the database contains few cases with injured, homeless and missing people.With regard to the homeless, for instance, the number is too small when compared with the numbers of events that caused complete destruction of the buildings.This suggests that the documented data in some ways underestimate reality.
The flash floods documented caused severe social damage, primarily involving buildings, lifelines and infrastructures.In particular, roads and private buildings were the most severely affected categories.In this regard, the most dangerous catchment types are carbonate catchments with and without pyroclastic cover.The damaged elements by each event are reported in classes in the database.In Fig. 13 some examples of damage produced by flash floods are shown.

Final remarks
Through an analysis of the existing literature, we focused on flash flood events in Campania, a region that has repeatedly been affected by severe flooding causing serious damage and fatalities.We collected temporal and spatial information on about 500 flash floods, thus building the first specific database concerning this type of geological hazard in the region.For this purpose, a critical scrutiny of the existing literature was performed to provide a degree of reliability for the collected information.We also defined the accuracy related to the temporal information available, and for most of the events accuracy proved to be medium to high, meaning that day, month and year of occurrence of the single event are known.
In order to reconstruct flood scenarios, we classified different catchment types on the basis of the main geological (bedrock lithology and presence/absence of detrital cover) and geomorphological parameters (type of outlet zone).This differentiation may be useful to understand the type of transported bed load (coarse vs. fine-grained) and to characterize the deposition area.
We also collected information about damage to people and society, which occurred as a consequence of flash floods.Most of the events took place in carbonate catchments with pyroclastic cover, both with and without an outlet to the sea.Among the 86 municipalities that were damaged by torrential flooding, 16 recorded more than 10 events.In Campania the mean occurrence interval of floods is very low, ranging from a few years for damage to buildings and infrastructures up to some decades in the case of events with victims.The C. Vennari et al.: A database on flash flood events in Campania, southern Italy widespread presence of carbonate catchments in the region and the intense urbanization of the area affected the information sources, and these represent the main weaknesses of the historical documents.Although the volcanic areas in Campania are densely populated, the alluvial events in carbonate catchments with pyroclastic cover caused more damage to people and society than those occurring in volcanic catchments.The study also showed that a significant number of catchments were affected by floods with high mean time period of occurrence of the events.The loss of historical memories of these events certainly lies at the origin of an increase in the risk conditions.
Such a database may be useful in different ways for land use and civil protection planning.First of all, it may help to locate the areas more susceptible to flash floods all over the region.Secondly, it may facilitate local authorities in charge of land management to select sites where monitoring procedures and/or prevention and mitigation works need to be adopted.In this perspective the main users of the database will be local administrators and the civil protection agency.

Data availability
The database was built for the author's PhD project; we have chosen to make it public with this publication.Any additional information about data sets can be provided by the authors.

Figure 2 .
Figure 2. Event types collected while constructing the database.

Figure 3 .
Figure 3. Rainfall data associated to flash floods: (a) maximum hourly rainfall (official data from CFDC); (b) maximum hourly rainfall (data from the literature); (c) maximum daily cumulative rainfall (data from SIMN).

Figure 6 .
Figure 6.Temporal accuracy distribution of the events collected in Campania.The histogram shows the distribution in five classes of temporal accuracy: low (L), medium-low (ML), medium (M), medium-high (MH), and high (H).The histogram on the left shows the distribution of the accuracy classes in five temporal classes.

Figure 7 .
Figure 7. Recurrence of alluvial events in the municipalities.

Figure 8 .
Figure8.Mean time period of occurrence of the events in the municipalities that have recorded more than 10 events.

Figure 9 .
Figure 9. Geographical representation of the events, according to the catchment class.

Figure 10 .
Figure 10.Number (a) and monthly distribution (b) of flash floods events for the different classes of catchments.Catchment class key: (1) carbonate catchment with pyroclastic cover; (2) carbonate catchment without pyroclastic cover; (3) carbonate catchment with pyroclastic cover and outlet to the sea; (4) volcanic catchment; (5) volcanic catchment with outlet to the sea.

Figure 12 .
Figure 12.Distribution of victims recorded in the territory.

Table 2 .
References used in Table 1 (R column).