The Effects of Flood Damage on Urban Road Networks in Italy: The Critical Function of Underpasses

Abstract

The urban areas of Mediterranean Europe, and particularly Italy, have experienced considerable expansion since the late 19th century in terms of settlements, structures, and infrastructure, especially in large population centers. In such areas, the geohydrological risk is high not only for inhabited areas but also along roadways exposed to flooding. This scenario is worrying, especially in road underpass sections, where drivers are unlikely to perceive a real risk due to the high degree of confidence that comes from the habit of driving. Underpasses have been widely used to obviate the need to find shorter alternative routes and manage vehicular traffic in urban settings impeded by previous anthropogenic and natural constraints. To assess the numerical consistency, frequency, and areal distribution of flood risk around road underpasses, several hundred pieces of data were selected (mostly from international, national and local newspapers, CNR IRPI archive and local archives) and cataloged in a thematic database, referring mainly to the Italian territory. The behavioral aspects in the face of risk were also examined in order to provide a better understanding and raise awareness for preventive purposes. The results of this specific CNR research, which lasted about two years, confirm the exposure of underpasses to extreme risk events, affecting road users. In Italy alone, between 1942 and 2023, 698 underpasses were identified as having experienced a flooding event at least once. The database shows that 680 vehicles were involved in Italy, with a total of at least 812 individuals, of whom 19 died. Despite incomplete and uneven information, the findings of the analysis regarding the increment in underpasses flooding and the drivers action in front of a flooded underpass may be useful for undertaking the appropriate mitigation strategies.

1. Introduction

Since the early decades of the 20th century, rapid building expansion has characterized the entire territory along the European Mediterranean basin; this has resulted in an imbalance of the urban environment to the disadvantage of the natural environment. Extensive residential complexes sprung up close to major cities, merging with the suburbs and using all available space, including riverine areas [1,2,3,4,5,6]. The massive increase in land consumption [7] led to immense sealed areas, which led to radical transformations of natural areas, watercourse patterns, and coastal areas [8,9,10].

Many cities on the coasts have become tourist attractions or places of intense trade (e.g., ports), and cities have undergone periods marked by strong commercial and residential appeal, with massive increases in population due to the centralization of employment and study opportunities, alternating with periods of estrangement motivated by better livability in the suburbs (in terms of cost and quality of life). However, this fluctuation resulted in a profound redistribution of the areas used for services, facilities, and infrastructure, including roads. The latter has always been a functional and discriminatory element for multiple purposes (economic, social, logistical, cultural, etc.).

All this resulted in urbanization dramatically reducing natural areas in both lowland and coastal areas [10]. A similar territorial remodeling moved principally by touristic needs also occurred in some mountain areas, increasing the geohydrological criticalities. The development of urbanized areas resulted in the construction of road networks on different levels to make every point of interest accessible both in touristic areas and in bigger cities.

As a result of continuous and increasingly severe geohydrological effects, the scientific community has begun to talk about climate change scenarios in studies related to rain regimen, which can manifest as drastic changes occurring in the annual hydrological supply (e.g., during 2022 in Italy) or in abundant rainfall being concentrated in a few days or sometimes in few hours. The study of climate and its manifestations is a subject of scientific and economic interest around the world because of the potential medium-term and long-term consequences for many ecosystems, including human activity [11,12,13,14,15,16,17,18,19,20]. Just in 2023, Italy recorded an increase in extreme climate events of 22% in comparison with 2022, with an increase in floods by 107% [21]. This increase in extreme climate events and floods is a result of a long-term trend that is still evolving [22,23], whose effects on hydraulic works and urbanized areas will become more and more severe [24]. Therefore, it is essential to better understand such processes: the causes that lead to underpasses flooding, the actions purchased by institutions before and during flooding events, and people’s actions in front of a flooded underpass and the motivation behind their actions. However, the connection between heavy rainfalls and flood effects on urbanized areas, especially on the effects on road network and underpasses, is a new topic for the scientific community and is still not well studied [25,26].

Changes in rainfall regimes are associated with socioeconomic impacts, the severity of which is amplified in densely populated settings deprived of ecosystem services, such as large urban areas. In fact, according to recent research [27], changes in hydrological cycles also result in an increased frequency of extreme hydrometeorological events, with important consequences for people and property. The risk of population-damaging geohydrological events increases as the population density of urban centers increases and the efficiency/effectiveness of runoff systems from urban areas decreases. According to projections provided by the United Nations, by 2050, 68% of the global population will be concentrated within urban centers [19,28,29,30]. About 44% of all natural disasters that occurred between 2000 and 2019 were floods, so the need to develop and adopt new measures to prevent these processes becomes urgent [31,32]. To meet the many transportation and communication needs, the urban fabric has also expanded vertically with services and utilities, through road and rail overpasses, subways, and road underpasses.

Therefore, the main aim of this study is a better understanding of the severity of the underpasses flooding and their recurrence: the intention was to examine the risk associated with geohydrological processes in the urban environment in Italy in relation to underpasses, which is also significant for the European territory and, above all, the Mediterranean area, where the problems were found to be very similar in different states (Figure 1) [10,33,34,35,36,37,38,39]. The very first challenge in such a study is the identification of the underpasses in the road network: at present time, an Italian Cadastre of underpasses still does not exist; therefore, in many cases for this study, manual research for underpasses and their eventual flooding was carried out [40,41,42,43,44,45].

The intention of the study is to upgrade the work with more complex investigation strategies once the dataset is shared on a national scale, probably after a process of comparison with the numerous territorial authorities.

A further aim of the study is to develop these aspects of urban planning with applications in terms of risk mitigation, encouraging the use of specific warning protocols for exposed sectors and also identifying targeted actions in terms of information and dissemination activities for road users and operational guidelines for land management (e.g., real-time monitoring).

2. General Settings

2.1. Environmental Legislative Framework

In order to regulate flood risk management, the support of appropriate legislation is needed to structure and coordinate effective preventive actions and to plan the transformation of urban settlements in favor of reducing exposure and vulnerability to geohydrological events. Italian legislation on soil defense and environmental constraints is substantial; regional regulations often overlap with national or European regulations, creating more complex situations, although they are not always homogeneous in application. In many cases, environmental regulations are introduced in Italy after major natural events. Other instruments contribute to environmental regulations, including those for civil protection, with different scales of application (e.g., regional, municipal), and those for urban planning, which regulate the municipal territory in relation to different parameters, such as exposure to natural hazards. These include reference standards for areas with possible flooding by watercourses and secondary hydrographic networks (natural and otherwise) and by anthropogenic or natural conditions [46].

Italian regulations at the national scale include the Piano Assetto Idrogeologico (PAI) [47] and European Directives that offer more homogeneous actions, with operational lines and guidelines for Member States.

Water Directive 2000/60/EC [48] issued by the European Community (EC), modified in Italy as Legislative Decree 152/2006, provides guidelines for the identification of river basins in the territories of Member States, their assignment to individual River Basin Districts, and the identification of the competent authority to apply the rules for each district.

The subsequent Floods Directive 2007/60/EC [49] was issued with the aim of providing indications for the management of flood risk, with the ultimate goal of reducing the negative consequences to human health, the environment, cultural heritage, and economic activities. This directive was implemented in Italy through Legislative Decree 49/2010, which established specifications for the creation of flood hazard and risk maps. The territory was divided into three classes of flood occurrence probability: low (L), with return time (Tr) up to 500 years; medium (M), with Tr = 100 ÷ 200 years; and high (H), with frequent floods Tr = 20 ÷ 50 years.

The EC delegates the Member States to define River Basin Districts (RBDs), which are responsible for preparing management plans. These plans must incorporate urgent emergency plans and take into account the aspects of forecasting, monitoring, surveillance, alerting, hydraulic territorial surveillance, regulation of runoff, and support for the activation of emergency plans.

For this purpose, the Italian territory was organized into seven RBDs (Figure 2). Based on the flood hazard and flood risk maps, EC Member States are required to prepare Flood Risk Management Plans (PGRA in Italian) according to the provisions of the legislation [48,49].

The main characteristics of Italian RBDs are shown in

Table 1. Main features of Italian RBDs.

RBD Name	RBD Area (1 × 103 km2)	Population Density (within RBD)
Po	82.8	238
Oriental Alps	34.7	189
North Appennine	24.3	221
Central Appennine	42.3	198
South Appennine	67.5	203
Sardinia	24.1	186
Sicily	25.7	74

.

2.2. Urban Sprawl and Geohydrological Risk

The industrial push that began in Italy in the late 19th century and early 20th century, the post-war recovery, and then the economic boom that occurred since the 1960s, which prompted the relocation of the population to urban centers, led to the rapid expansion of construction (also referred to as the “cement industry”). As a result, in order to adapt to the increasing number of people and service needs, cities underwent rapid transformation, in several cases resulting in urban sprawl. This exponential growth of built-up areas, which in some areas began in the first decades of the 20th century, but boomed from the late 1950s, often neglecting the presence of natural features, especially waterways and riverine areas [2,4,10,50].

During the period 2000–2018, Italy was among the EC Member States that had the highest percentage of artificially covered land (with over 5.5% of land sealed as early as 2000 and a further 6% increase subsequent to that) [51]. The increase seemed to be especially evident in nations such as Spain (with more than 20% of sealed area compared to 2018), where building continued despite the obvious critical issues highlighted for European coastal countries bordering the western Mediterranean Basin [10]. The current landscape in Italy is thus the result of complex processes of urban transformation, originally linked to water availability for various purposes, from sanitation to industrial activities. Urban planning over the decades has shown little respect for the land and has paid little attention to the consequences of altering the balance of natural systems. Urbanized areas, and the environment, have been affected by urban planning, leading to a chaotic, sometimes random structure, compromising the necessary strategies for land self-protection.

Over the past 150 years, the list of flooding events causing damage to urban areas is long, with ever-increasing economic repercussions [52]. The factors contributing to the occurrence of flooding in the urban environment are many and include natural and anthropic causes [51,53,54]. In new urban areas, road and railroad infrastructure (which, due to structural needs, has remained almost unchanged in elevation and layout since the beginning of its construction) has occupied riverine areas of the floodplain, interfering with natural geomorphological processes and consequently increasing geohydrological risk (Figure 3) [55,56,57,58]. This is also evidenced by changes in the patterns and width of coastal watercourses, which have shrunk by 10% to 90% over about 150 years, with the covering over or alteration of more than 60% of watercourses [10]. One result of urban expansion in Italy has been increased occupation of already densely populated areas with high geohydrological risk. The reduction in natural spaces has been accompanied by a loss of historical memory of flood risk by the population and sometimes by land managers. In some cases, the transformation of watercourses has led to their being covered over, and in many cases, stretches with culverts are not shown on the most recent maps and are only rarely identifiable by local toponyms [59,60,61,62,63,64,65].

These aspects are most evident from the increased urban sprawl highlighted by a comparison of CORINE Land Cover (CLC) data [51] for each RBD in relation to the hydrographic network between 1990 and 2018 [66]. Built-up areas increased from 1990 to 2018 (CLC comparison) on the various districts by an average of 22.7%, ranging from 6.5% to 31.9% (

Table 2. Increase in sealed areas based on a comparison of CORINE Land Cover (CLC) data for Italian RBDs [51,66].

RBD Name	RBD Hydrographic Network (km)	Increase in Urbanized Area Per CLC, 1990–2018 (%)
Po	16,058	22.9
Oriental Alps	5774	24.0
North Appennine	5626	24.5
Central Appennine	7040	19.9
South Appennine	10,786	29.0
Sardinia	735	31.9
Sicily	4149	6.5

).

The comparison of CLC over a span of about 30 years (1990–2018) shows the areas most affected by the increase in land use for human activities, with significant sealing at the expense of natural drainage areas and rainwater infiltration capacity (Figure 4) [51].

2.3. Rainfall Characteristics

Alterations in precipitation regimes have been recorded worldwide in recent decades [27,67,68]. However, the Italian national territory is repeatedly and historically subjected to localized or extensive flood events [10,69,70,71,72], as shown in

Table 3. Main flood events that occurred in 2010–2023. Floods that resulted in fatalities are shown in bold [10,49,69,70,71,72].

RBD Name	Main Flood Events (dd/mm/yyyy)
Po	26–27/08/2023; 13/08/2023; 21–24/07/2023; 16–17/05/2023; 2–3/05/2023; 23/01/2023; 26–28/07/2021; 6/12/2020; 2–3/10/2020; 29/08/2020; 23–24/11/2019; 21–22/10/2019; 2/02/2019; 3/07/2018; 12/12/2017; 24/11/2016; 14/09/2015; 9/08/2015; 15/11/2014; 12/11/2014; 13/10/2014; 19/01/2014; 11/06/2011; 5/06/2011; 2/03/2011; 1–2/11/2010; 30–31/10/2010; 11–15/08/2010; 14–17/06/2010; 2–5/05/2010
Oriental Alps	21/12/2019; 11/2019 and 12/2019; 3/07/2019; 26–30/10/2018; 19/11/2016; 2/8/2014; 16/05/2013; 21–26/12/2010; 1–2/11/2010; 31/10/2010
North Appennine	2–3/11/2023; 20/10/2023; 4/10/2021; 2–3/10/2020; 23–24/11/2019; 21–22/10/2019; 27/07/2019; 23/04/2019; 14/08/2018; 9–10/09/2017; 22–24/11/2016; 23–24/04/2016; 24/08/2015; 15/11/2014; 10/11/2014; 5/11/2014; 15/10/2014; 9–10/10/2014; 31/01/2014; 19/01/2014; 24/10/2013; 21/10/2013; 5/10/2013; 4/12/2012; 28–29/11/2012; 10–12/11/2012; 7/11/2011; 4/11/2011; 24–25/10/2011; 4/10/2010; 7/09/2010
Central Appennine	15–16/09/2022; 27/07/2019; 25/11/2018; 20/08/2018; 26/05/2015; 26/03/2015; 15/10/2014; 3/5/2014; 02/12/2013; 31/10/2012; 20/10/2011; 3/06/2011; 2/03/2011; 30/11/2010; 28/11/2010
South Appennine	3/04/2023; 26/11/2022; 9/08/2022; 9/11/2021; 6/12/2020; 7/10/2020; 4/10/2018; 12/11/2019; 9/09/2016; 31/10/2015; 15/10/2015; 6/09/2014; 1/12/2013; 8/10/2013; 21/08/2013; 8/12/2012; 20/11/2012; 6/11/2011; 21/10/2011; 7/10/2011; 2/12/2010; 2/11/2010; 9–10/09/2010
Sardinia	25/01/2023; 14/11/2021; 28/11/2020; 8–10/10/2018; 22–23/07/2015; 18/11/2013; 24–25/10/2011; 4/10/2010; 25–26/01/2010; 13/01/2010
Sicily	24–29/10/2021; 15/07/2020; 25/10/2019; 3/11/2018; 22/01/2017; 25/11/2016; 2/02/2014; 21/09/2013; 31/01/2012; 5/03/2011; 19/02/2011; 1/02/2011; 1/10/2009

. Numerous historical cases of flooding in many Italian areas in the 19th and 20th centuries are also documented [10,49].

The observed changes in rainfall characteristics are not limited to an increase or decrease in the annual or seasonal accumulation in some regions but are also manifested as an increase in the frequency of heavy rainfall and the intensity of rainfall events [67,68,69,70]. These changes affect millions of people around the world, who now have to adapt to new conditions and more frequent extreme events such as heavy rains, droughts, and floods [67]. A link between the increase in extreme precipitation events and the magnitude of flooding has been demonstrated; however, the observations indicate a greater influence of local factors that can affect precipitation, such as basin characteristics and land use [67]. In addition, in order to study precipitation, forecast models include more complex evaluations, such as increasing global temperature. This will lead to an expansion of flood-prone areas [27] and an increase in the population affected by flooding by up to 400% [67].

Precipitation was also examined for the analysis of factors predisposing the flooding of the surveyed underpasses in Italy. Many floods should be analyzed as individual cases based on critical precipitation characteristics. However, annual rainfall was considered at the hydrographic district scale for the period 2010–2023 [73,74]. This information, although it has less statistical value, is useful for a qualitative analysis of the data, which must necessarily take into account the total number of underpasses surveyed, the territorial extent of the districts, the frequency of flooding events in the entire dataset by district, and the frequency of flooding events that occurred during 2010–2023 by district (

Table 4. Annual cumulative precipitation (mm) by RBD area recorded in 2010–2023. Values were obtained by averaging annual cumulative precipitation in different cities [74,75,76].

RBD	2010	2011	2012	2013	2014	2015	2016	2017	2018	2019	2020	2021	2022	2023	Variation % (2010–2023)
Po	1232	747	847	1052	1346	762	979	675	985	997	832	699	555	908	−70
Oriental Alps	1412	822	984	1215	1590	728	1080	929	958	1251	990	866	584	877	−48
North Appennine	1403	705	893	1235	1473	763	971	754	1049	1203	962	948	742	767	−45
Central Appennine	950	649	848	1017	952	809	795	756	868	810	729	712	603	826	−13
South Appennine	1036	702	757	894	822	788	768	590	913	809	679	780	1074	1046	+1
Sardinia	723	682	538	595	481	561	496	431	843	723	555	715	339	280	−61
Sicily	746	786	643	676	602	925	579	515	728	568	456	679	530	341	−54

).

A first examination of the problem in terms of its temporal variability and the possible influencing factors led to assuming a significant variation in the rainfall pattern in relation to the possible increase in the number of flooding events. The analysis of annual cumulated rainfall alone was insufficient to give a reliable explanation, except in terms of an overall average reduction of about 41% in the annual rainfall amount, considering all districts, with a more important value for the Po RBD (70%) (Table 4).

After establishing the annual reduction in the amount of precipitation, the increase in heavy rainfall was examined in relation to the number of underpasses floodings (Figure 5). The graph illustrates that the number of flooded underpasses recorded in the database is visibly increasing, especially since 2018, despite the effects of the COVID-19 pandemic lockdowns; similarly, there is a significant increase in the heavy rain events on a national scale for the same period. The dataset does not allow significant statistical evaluations, also in light of the integration of the dataset that could be performed with integrative remote sensing surveys, described in the following paragraphs.

3. Materials and Methods

3.1. Geographic Database

In order to analyze the problem of urban flooding events at road underpasses, it was necessary to carry out extensive research on the subject, in order to provide a reference tool.

The creation of databases is one of the processes on which some research dedicated to geohydrological risk prevention is based. Many historical records provide evidence for the evolution of the relationship between geohydrological risk and the development of human settlements [76]. The sources, which are often uneven and discontinuous, are methodically brought together in a homogeneous and systematic cataloging system comprising databases. This way, the data will be easy to use, allowing the evaluation of mitigation, management, and land development solutions by comparing past events.

The data collection, which lasted about 24 months, was mainly based on newspapers and online news media, social media, and published and unpublished research by the Italian National Research Council’s Research Institute for Geo-Hydrological Protection (CNR-IRPI). For data retrieval and location in a dedicated GIS project (qGIS version 3.36.2), Google research was carried out in about 10 languages, with specific keywords, to confirm the internationality of the issue. The spatial database was implemented using important open source resources [51,78,79,80,81,82,83,84,85,86,87].

A comparison of historical maps with current ones revealed major transformations of natural areas, particularly to the detriment of riverine environments (Figure 6). The elaboration for the evolution of river paths are based on historical maps and aerial photography available in the CNR-IRPI archives and on satellite and drone imagery. An example of river path evolution is presented in Figure 6, which exemplifies the evolutions occurred from 1878 to 2000 for Stura di Lanzo River (Po RBD). This kind of study aims to evaluate the degree of bank width reduction [10] and the proximity of new urbanization to historical river paths, which, in case of floods, can be prone to flooding.

Measuring river sections on historical (19th century) and current maps, there is a clear reduction in useful discharge sections in active riverbeds by over 55% [10]. Floodplain changes are seen in the form of morphological pattern changes and culverting rivers. Among the watercourses, 63% became streams with paths that were changed or covered (Figure 7) [10].

3.2. GIS Project

The dataset was transposed into qGIS (reference system WGS 84, UTM Zone 32N, EPSG 32632) using point geometry shapefiles. Each flooded underpass was located and given an alphanumeric identifier and a site certainty scale (1 to 3, with increasing uncertainty) validated from different sources and map details at 1:1000 scale.

For underpasses with maximum certainty, when flooding repeated over time, the degree of recurrence was calculated.

The analysis of cases in which one or more people were involved in flooded underpasses in motor vehicles allowed us to select some information, when available, about the age range of the victims, either drivers or passengers (<17 years, 18–29 years, 30–59 years and >60 years). For those involved, their occupation, the driver’s behavior (distraction, emulation, unawareness of the danger, etc.), and the reaction to the accident (requesting or waiting for help, exiting from the vehicle by themselves or waiting on the roof of the vehicle, expressing emotion or panic, etc.) were also indicated where possible. If described in the sources, during the stages of underpass flooding, the presence and characteristics of the garrison (instrumental or human), either temporary (barriers) or permanent (mobile bars), and visual warning systems (flashing lights, traffic lights, signs, etc.) were highlighted. The time when vehicles were stuck in the flooded underpass (morning, afternoon, evening, night) and the season or time of year were also indicated (for most, the actual date of occurrence was available).

The presence of a system aimed at preventing underpass flooding (e.g., water pumps) was specified if these were visible on Google Street View or reported by the source. Based on detailed street maps, any alternative roads within about 500 m that could have been taken were also highlighted.

To check whether underpasses were distributed over areas with flood risk, the PGRA subdivision by degree of risk was used. No other reference cartography at the national scale was chosen because it was not uniformly provided throughout the territory. The different risk classes defined by PGRA were evaluated only for underpasses with maximum certainty, highlighting the degree of risk by location.

An aspect of interest for the purposes of this research was an analysis of the spatial distribution of flooded underpasses and the main arterial roads or directions of the railway network. This is related to the fact that the extent of railway lines Italy is the fourth largest in Europe, after Germany, France, and Poland [88]. Italy is also among the top four European countries with the largest extent of highways, along with Spain, Germany, and France [89]. The distribution of roads and railways was analyzed in terms of the density and number of flooded underpasses for each district.

Another important aspect is the population density in each hydrographic district, with the assumption that higher density corresponds to a higher number of passing vehicles and, thus, a greater exposure to risk.

3.3. Geomorphological Predisposition to Flood Events

The Italian territory has suffered from overflowing watercourses in recent flooding events, but also from effects related to pluvial flood or coastal flood processes, with the latter connected to the thermal reaction of air masses coming from the Mediterranean Sea [49]. In this complex scenario, urban drainage systems, the size of rainwater storage and drainage, the absence of suitable lamination systems, and unnatural modifications of flow outlines in urban stretches all play a significant role. Culverted or channeled river stretches are frequently responsible for damaging effects during floods [49]. In case of flood events, urban areas have an elevated density of weak elements, including underpasses. Among all underpasses analyzed in this study, 75% of them are located in areas identified as “urban fabric”, “industrial or commercial units”, or “artificial, non-agricultural vegetated areas” by 2018 CLC (Figure 7).

Historical data and maps have proven to be fundamental for understanding ancient routes and the susceptibility to flooding of vast areas within districts. Often, in fact, flood areas are defined by reactivations of historical river paths. A number of historical documents preserved in the archives contain highly detailed geomorphological information, including on flood paths; these made it possible to remap the most recently flooded areas with a high degree of overlap. This experience demonstrates the importance of knowing the territory through historical information [59,60,61,62,63,64,65].

The main data collection for this study was made manually, but an attempt in a semi-automatic compilation was carried out. In order to evaluate the implementation of the dataset, satellite images of flooded areas with evidence of harmful effects were used, from which maps could be derived. The intention was to check whether, in extensive flooding events, a number of underpasses might have been excluded from the census, since the severity of the flooding led to an echoing of newspaper information based on the priority of the reports, which may have overshadowed the information of underpass flooding by importance.

Even though this automatism could not be used in the whole area analyzed, due to the lack of a National Cadastre of underpasses. When it is available, it would be possible to directly extract the flooded ones from alluvial maps, which are not always disposable, especially for historical floods. Alluvial maps for recent flood take some time to be elaborated, but this issue could be bypassed by using Sentinel 2 [90] imagery of the Normalized Difference Water Index or Copernicus emergency data [91].

The check for the incrementation of flooded underpasses using satellite imagery was carried out using official maps of flooded areas obtained by Piedmont Geoportal, and the shapefiles of road underpasses obtained by selecting categories from the regional database of Piemonte (Po RBD) [92] (Figure 8).

An approach to support the investigation of underpasses consisted of verifying, through satellite and radar images, the meteo-pluviometric conditions responsible for damaging events, identified through a search for underpass flooding data, the planning instruments related to the areas invaded by water, and the geomorphological transformations of the river patterns. Having this in mind, we used data made available by Copernicus [7,90,91] and some regional radar monitoring services.

Since 2010, there has been a dramatic increase in the number of heavy rainfall events causing flooding, landslides, and whirlwinds throughout Italy, resulting in considerable damage to urban areas. Often, in the events responsible for the greatest damage, the normally recorded annual rainfall amounts, calculated for the same locations over previous decades, accumulated in a few hours [93,94]. A total of 684 flooding events in Italy caused by heavy rain have been recorded, among which 57.7% have occurred since 2020, and 166 river floods have been recorded [93]. For example, in the Province of Ravenna and Forlì (Central Appennine RBD), two exceptional rainfall events occurred in 2023, between 1 and 3 May, with a total of 229.4 mm, and from 10 to 21 May, with a significant peak on 16–17 May of 323.2 mm. The rainfall sum of the two May events was 564.4 mm, which represents roughly the total annual rainfall of the same localities in 2022 [95].

The floods reported in the dataset are all attributable to flash floods [96], intended as a rapid inundation, usually by less than six hours after the heavy rainfall event.

Table 5. Main types of flooding designed by Institute of Catastrophic Loss Reduction [96] and significative for Italian cases.

Types of Flood	Description
Flash flood	Rapid inundation, usually caused by heavy rainfall, that occurs in less than six hours from the beginning of the trigger cause.
Fluvial flooding	Flooding caused by the water level increase, in rivers or other watercourses, and the water overflow into sourrounding areas.
Pluvial flooding	Flooding caused by a large amount of precipitation in a very short period of time; the ground condition (e.g., impermeabilization, dryness, over-saturation, etc.) do not consent the drainage of such a large amount of water in such short time, causing an inundation.
Urban flooding	Flooding caused by an incapability of the urban drainage system to dispose of an over-amount of water (e.g., heavy rainfall, sewer breakage, fluvial flooding, etc.)
Coastal flooding	Flooding of lake, sea or ocean shorelines that normally are above the water level. This type of flood may be caused by high tides, storm surge or tsunami, even in combination with other type of events.

reports further flood typologies referable to the dataset.

To obtain a better comprehension of the procedures used and suggested in this study, a synoptic frame is presented as Supplementary Material S1.

4. Results

4.1. Underpass Database Consistency

A total of 1521 underpass flooding events were surveyed, occurring between 1942 and 2023, of which 1439 (95%) occurred in Italy and 82 (5%) occurred in Europe and South America. The events that occurred outside Italy were reported in the dataset but were not considered exhaustive, since no foreign local newspapers had been consulted, and the only information available had been echoed through Italian news reports and newspapers. For some cases (about 1% of the total recorded), there is no temporal reference due to the lack of original information; therefore, they were not considered in the statistical analysis (

Table 6. Annual underpass flooding reported in the Italian database.

Date of Floods (Year)	No. Flooded Underpasses	No. People Involved in Floods	No. Vehicles Involved in Floods
1942	1	0	0
1953	1	0	0
1957	1	0	0
1981	1	2	2
1983	2	4	3
1987	1	9	9
1988	1	0	0
1989	2	0	0
1991	5	6	6
1992	5	5	5
1993	11	10	10
1994	4	0	0
1995	2	1	1
1996	10	25	25
1997	5	18	17
1998	5	2	2
1999	7	4	4
2000	9	5	5
2001	4	1	1
2002	10	3	3
2003	1	0	0
2004	1	4	2
2005	4	2	2
2006	2	0	0
2007	3	2	1
2008	4	7	5
2009	5	6	4
2010	5	6	2
2011	27	21	20
2012	19	8	3
2013	42	21	20
2014	116	29	26
2015	59	32	21
2016	60	27	19
2017	82	35	31
2018	163	67	61
2019	135	110	85
2020	97	84	67
2021	186	78	64
2022	106	72	60
2023	220	106	94
No date	15	0	0
Total	1439	812	680

and Figure 1).

The database shows that 680 vehicles were involved in Italy, with a total of at least 812 people, of whom 19 died (Figure 9).

The unevenness and incompleteness of the time series can be explained by the fact that it is easier to find more current information. The greatest availability of data is for the period after 2010 (Figure 10). Figure 10 illustrates that population trends in Italy show a main period of growth from the late 1950s through the 1980s. A lower growth is observed between 2000 and 2015, with a slight inflection in the last decade. In contrast, the number of cars, always growing since the 1960s, increased steeply until the 1990s; thereafter the growth curve is weakly, but steadily, increasing. The same graph shows the trend in underpass flooding events, which shows a marked increase from about 2010 onward.

4.2. Spatiotemporal Distribution of Underpass Dataset

The spatial distribution of surveyed flooded underpasses by RBD shows extreme variability over the period 1942–2023. The last 10 years (2014–2023) shows an overall increase in events (Figure 11).

The monthly distribution of underpass flooding events is in line with the average annual rainfall pattern in Italy, with 8.2% in winter, 12.8% in spring, 36.7% in summer, and 42.3% in autumn (Figure 12). Analyzing the same information based on RBD better shows the seasonal variability of the entire Italian peninsula in relation to thermo-pluviometric characteristics.

4.3. Flooded Underpass Typology

The flooded underpasses in the GIS project with maximum certainty (No 698) can be divided into three categories: road (59%), rail (29%), or mixed (12%) (

Table 7. Distribution of surveyed flooded underpasses, divided by type, in relation to road and rail network development in RBD areas.

RBD	Area (km2)	No. of Road Underpasses	No. of Railway Underpasses	No. of Mixed Underpasses	Railway Length (km)	Road Length (km)	Density of System (Raylway and Roads/RBD Area)
Po	82,788	143	93	33	5958	10,437	0.20
Oriental Alps	34,724	39	38	8	3781	4249	0.23
North Appennine	24,284	86	16	9	2238	4126	0.26
Central Appennine	42,275	80	27	14	2787	6076	0.21
South Appennine	67,490	53	18	19	5483	10,736	0.24
Sardinia	24,118	2	0	0	1218	3512	0.20
Sicily	25,718	9	10	1	1618	3904	0.22

). The historical involvement of railways underpasses in floods can be explained by the structural rigidity of the railway layout, which imposes an elevation constraint on the roads because of the need to maintain a constant elevation, or with reduced variations, for the railway’s functionality.

In addition, Italy’s natural constraints, such as the long coastline and reliefs, together with the presence of large urban centers, may have led to a thickening of the routes in the most frequently used or built-up areas.

The proposed research could be improved by having a homogeneous and complete National Cadastre of underpasses, which could be related to the areas actually flooded during flooding events. The current database of surveyed underpasses, while representing a step forward in terms of knowledge, is not complete for some RBDs. This reduces the possibility of identifying the real criticalities of the road network and underestimates the potential harmful consequences for drivers. The flooded areas could derive from traditionally surveyed maps or from automatic (or simplified) processing of highly detailed images acquired with satellite techniques. An example proposed is the one in Figure 13, in which the analysis process is simplified: having the satellite data (a, d) proceeding with the restitution on thematic maps of the flooded areas (b, e), and finally identifying the underpasses affected by the flooding events (c, f). The areas affected by flood events can vary in extension depending on the severity of the flood, but most of the times are in the same location, as shown in Figure 13g, where three different flood events occurred in Alessandria municipality (Piemonte, Po RBD), impacting similar areas. In fact, out of the thirty-three underpasses present in Alessandria of the Piemonte Cadastre, three were flooded thrice.

The result of overlapping the aerial and satellite images (or maps of the trends in heavy weather forecasts) and the Piemonte Cadastre of underpasses allowed us to identify the flooded areas in 2000 and 2016 (in the Po RBD) and the actual number of underpasses that were likely involved in flood events. From the dataset of flooded underpasses surveyed for this research, only two cases emerged in the 2000 Piemonte flood and three cases in the 2016 flood, while the use of satellite maps would have increased the data by at least 50% (Figure 8).

The densification of the population in tourist seasons increases the geohydrological risk: about 17% of the surveyed underpasses are located in a 1 km buffer of Italy’s coastal areas, which is 1% of the national territory, mainly following along the railroad track.

4.4. Main Factors Involved in Recorded Flooding of Underpasses

At least 84% of the underpass flooding events for which information on causes was available can be associated with precipitation events. However, the wide geographic extent of the study area and the lack of original information did not allow a detailed analysis of the characteristics of the triggering rainfall. The remaining 16% of events were mostly drainage system failures (obstructions or disruptions) caused as a consequence of heavy rain, as the drainage systems did not work as planned and caused water to stagnate in underpasses, and, in a few cases, caused overflowing streams (fluvial floods) and coastal floods (Figure 14a). Similar percentages are observed condensing the database in the restricted period 2010–2023; with the exception of 2010, the cases of pluvial flooding turned out to be the main cause of underpass flooding, followed by urban flooding. The cases of coastal flooding are always under 10% of the total annual events (Figure 14b).

Among the inland underpasses surveyed, 58% are not in any PGRA risk class; this percentage is reduced to 37% in coastal areas, where the natural expanses of watercourse mouths are located, but also where urban settlements occupy depressed and flat topologies. These areas of natural river expansion are often affected by backflow processes in channelized stretches. The distribution according to the PGRA flood risk classes shows a high number of flooded underpasses in uncategorized areas; in fact, a mean of at least 30% of the underpasses analyzed over the whole Italian territory are located in a flood-prone classified area, while the remaining are located in areas that should not be such prone to flooding (

Table 8. Distribution of underpasses among PGRA flood risk classes out of total high certainty underpasses in each RBD (%).

RBD	H Risk	M Risk	L Risk	Not Classified
Po	5	22	11	62
Oriental Alps	13	3	20	64
North Appennine	21	28	39	12
Central Appennine	3	17	16	64
South Appennine	13	1	2	84
Sardinia	50	0	0	50
Sicily	25	0	0	75

).

4.5. Rainfall Observations in Underpass Flooding Areas

The trend in annual cumulative rainfall per RBD (Table 4) does not always match the number of cases of underpass flooding (Figure 15). In fact, more significant data would have been the annual maximum for 24 h rain or the annual number of rainy days, but these data are not available for each RBD; therefore, out an analysis only on the cumulative rainfall was carried. Some critical aspects of an analysis on cumulative rainfall have been highlighted, mainly attributable to the following:

• The small amount of information available (e.g., as already mentioned, number of rainy days, intensity of precipitation).
• The small number of intense precipitation events or precipitation events that may have simultaneously affected large areas of the RBD.
• The annual cumulative precipitation calculated over the entire RBD do not represent the characteristic of precipitation responsible of flood damage. It is possible that intense rainfall may have affected the territory in very localized events in different or distant areas (e.g., tributary basins) than areas where the underpasses are located. For numerous cases occurring in the last decade in Italy, registrations for a single precipitation event occurred in just few hours reached cumulative values with percentages of over 50% of the annual cumulate. The last event occurred on 29 June 2024, and, in some localities, maximum cumulative values on 3 and 6 h were recorded (Noasca town, north Piemonte), which correspond at return period of over 200 years [97].
• Possible preventive actions put in place through regulation works.
• Inefficient drainage systems due to obstruction or malfunction rather than particular amounts of rain being the main cause of underpass flooding.

To examine the recurrence of underpass flooding, their frequency in each RBD was analyzed for the same period. By dividing the number of flooded underpasses by the number of total events in the respective districts, we obtained an index of the underpass flooding frequency (

Table 9. Flooded underpasses in the full dataset and during 2010–2023 in each RBD, and average frequency of occurrence calculated as the total number of floods over the number of flooded underpasses.

RBD	Area (km2)	No. Flooded Underpasses	No. Floods	No. Floods (2010–2023)	Flood Frequency Per Year (2010–2023)	Underpass Flooding Frequency
Po	82,788	269	539	460	35	2.00
Oriental Alps	34,724	85	123	120	9	1.44
North Appennine	24,284	111	281	257	20	2.53
Central Appennine	42,275	121	228	221	17	1.88
South Appennine	67,490	90	180	172	13	2.00
Sardinia	24,118	2	7	7	1	3.50
Sicily	25,718	20	81	80	6	4.05

).

A high repetitiveness index was found for some RBDs (e.g., Sicily District), whereas for others, significant weather/rain events occurred, justifying the number of underpass flooding events (e.g., North Appennine District, 3–5 October 2021). The high value obtained for Sardinia District has a limited statistical significance due to the limited amount of data available for the number of flooding events and the number of underpasses, and the small distribution and extension of roads and railways.

4.6. Analysis of Prevention Methods Used near Underpasses for Flood Risk

Dividing the territory into different risk classes should lead to different reactions by emergency management agencies, with the implementation of proportionate preventive measures. However, from the dataset (considering only records with maximum certainty regarding the location), it appears that the percentage of underpass flooding events where the presence of a preventive measure was reported does not consistently increase between underpasses identified external to the PGRA classified areas and those in the high-risk class (Figure 16). In fact, the mean of flooded underpasses with preventive measures, included in one of the PGRA classes, is 70% while the mean of flooded underpasses in a PGRA classified area that did not have any preventive measures is 30%. Similar percentages are reported for flooded underpasses external to PGRA delimitations: 66% of the underpasses presented some sort of preventive measure, while the remaining did not.

Possible actions by emergency management agencies against flooding events in the past were examined. Underpass flooding events were also assigned a degree of recurrence, which is useful for determining the succession of flooding events occurring in the same underpass (

Table 10. Flooded underpasses in full dataset, and their average frequency of recurrence.

PGRA Class	No. of Underpasses	Recurrence of Flooding 1–3	Recurrence of Flooding 4–6	Recurrence of Flooding >7
H	69	104	6	3
M	114	156	9	4
L	113	164	12	1
External	402	590	68	31

). One underpass in a PGRA risk class was flooded more than seven times, about 32% of all underpasses were flooded at least twice during the considered period, and 5.7% flooded at least four times. A similar analysis was carried out based on the information reported about the presence of a preventive measure at the time of the event; the number was reduced to 1148 cases of flooding with maximum certainty to 422 with prevention. A comparison of the distribution of underpass flooding events divided by PGRA class, degree of recurrence, and the distribution of manned events shows a number of flooded underpasses equipped with measures at the first flooding event, and the number decreases as the degree of recurrence increases; thus, they offered poor mitigation activity. The underpasses surveyed to date that are equipped with a warning system for approaching drivers represent a mean of 55% of the total. Observing the spatial distribution of signaling systems reveals further details. The equipping of underpasses with warning signs proves to be homogeneous throughout Italy, unlike the presence of traffic light systems, the frequency of which increases near major urban centers. The presence of barrages, although numerically insignificant in the dataset, also does not appear to be spatially clustered (

Table 11. Flooded underpasses equipped with flood alert systems, by RBD.

RBD	No. of Flooded Underpasses	Underpasses Equipped with Flood Alert (%)	Type of Flood Alert
			Warning Signs	Traffic Lights/Beacons	Traffic Barriers
Po	269	68.7	52%	45%	3%
Oriental Alps	85	76.5	45%	52%	3%
North Appennine	111	68.5	30%	69%	1%
Central Appennine	121	68.6	52%	47%	1%
South Appennine	90	58.9	43%	53%	4%
Sardinia	2	50	100%	0%	0%
Sicily	20	60	42%	58%	0%

).

In 67% of the cases, flooding occurred between 6:00 a.m. and 6:00 p.m., when visibility conditions are generally better (Figure 17). Even without detailed information on the reasons why drivers had little or no perception of risk, it is still possible to suppose a lack of caution when approaching an underpass in adverse weather conditions, even when minors were present.

5. Discussion

The results of the analyses indicate the usefulness of knowledge in order to identify critical issues and plan an integrated design for structural and nonstructural interventions. Regarding the first aspect, action can be taken by readjusting drainage systems based on the current context and by monitoring the systems and considering the actual extent of sealed areas and civil users. Related to this are the expected scenarios of increasingly intense and concentrated rainfall, derived from climate change studies.

Thanks to the development of innovative technologies and a growing awareness of solutions such as natural-based solutions [98], which are inspired by the principles of ecosystem functioning, a diverse range of solutions are now available that can address different local needs. Considering the number of events for which the cause has been traced mainly to rainfall, investing in enhanced drainage systems would greatly reduce the possibility of criticality in the future, leading, for an initial investment, to a reduction in costs resulting from future interventions. On the nonstructural side, several risk mitigation interventions are possible, which can act on the source of the hazard or the object of protection. Investing in maintenance plans for drainage systems, given the high number of floods without their implementation, would contribute to reduced emergency management costs in the long term.

Determining areas prone to flooding would also be important to prevent future harmful effects on the population if the floods extended to the whole territory based on the data of past flood events.

In addition to reducing the vulnerability of underpasses to flooding events, it is possible to alter the behavior of drivers by creating disincentives for attempted crossings, based on better knowledge of the flood hazard and accurate preventive warning.

An action aimed at developing risk awareness among drivers presupposes that they can identify elements of danger in good time while driving. Implementing visual warnings of the presence and level of water at underpasses that do not have them makes it easy to identify a potential hazard. However, in the driving context, the lack of knowledge of other routes and overconfidence in one’s ability to assess the water level could lead drivers to cross. This is why it is advisable to equip underpasses with all types of signaling devices, including barriers, which at least imposes a physical stop for the ones that are the hardest to see.

In addition, for the 117 subways (16%) that can be bypassed by an alternative passage within a distance of 0.5 km, it would be possible to equip signposts with panels indicating that drivers should bypass the obstacle. This would reduce the psychological pressure linked to the lack of knowledge of alternative routes that forces people to cross subways.

Although efforts are being made to implement strategies to mitigate the harmful effects of floods, risk awareness actions for travelers are still underdeveloped. The proof of this is the events that occurred in March and May 2024 in northern Italy. In particular, a video of the event that occurred in Po District illustrates a series of driver behaviors, highlighting their extreme subjectivity (Figure 18 and Figure 19) [99].

This paper describe a situation that helps policymakers and land managers understand how different environment choices and land use trends could influence flood effects, adaptation efforts, and societal resilience in the face of climate change impacts in the urban areas.

The resident population is often unaware of the danger posed by environmental and man-made conditions in which they spend their daily lives. On numerous occasions of post-event surveys, the CNR staff collected testimonies from the people directly or indirectly involved who did not even know of the existence of watercourses or the presence of artificial hydropower dams at short distance from their homes or places of study and work.

The territory is not known in its present form, let alone its past linked to damaging events (loss of historical memory).

Urban areas are the most compromised, because it is often not possible to intervene adequately in restoring the conditions that are more favorable to the natural course of the processes. The possibilities for reducing the damaging effects focus on a better education of people, exposing the potential risks and identifying appropriate behavioral procedures. Land managers, in their various competences, should have a greater attention to the criticalities that have already been identified in order to reduce the aggravation of risk exposure. A particularly useful measure, in the specific case of underpasses, would consist of instrumental (visual and analytical) monitoring of the underpasses with regard to the conditions of practicability during rain events or sudden water flows. Underpasses structures should allow an immediate perception of flooding risk by means of active solutions (e.g., visual sensors). Last but not least, users should be properly trained on the risks of the road, including flooding events (e.g., driving license courses).

6. Conclusions

The rapid consumption of undeveloped areas has made building space an increasingly valuable resource, including urban underground space. This has set the stage for the development of critical situations of underpass flooding, which are increasingly being reported through news and social media, often caused by heavy rains whose occurrence is increasing over time. The study of this problem is crucial to understand its dynamics and the vulnerability that underpasses may have. Over the underpasses analyzed in this paper, a mean of at least 30% are in a flood-prone area delimited by PGRA. In such cases, it is crucial for local authorities to have a higher level of attention for the dangers caused by flooded underpasses and that signals of unusable underpasses or physical barriers are quickly collocated. Over 70% of underpasses are outside the flood-prone areas, and over 40% were flooded more than one time. This high recurrence highlights a lack of attention to the security of underpasses, especially during heavy rains events. Drivers in the face of a flooded underpass will still try to use it, putting themselves in danger because, in many cases, an alternative way was not indicated and no physical barriers were collocated. The aim of this study is to identify critical points related to underpasses along roads to develop functional prevention systems and effective intervention plans.

Floods can pose a threat to the safety of people, infrastructure, and property, with repercussions in terms of productive activities and people’s lives, other than causing economic damages. Numerous structural and nonstructural solutions can be relied upon to protect urban centers from such events. Particular situations of urban development and spatial characteristics limit the application of these new technologies. For this reason, there should be multiple integrated mitigation strategies, favoring a functional and diversified system of prevention and response.

Anticipating heavy rains or situations that could lead underpasses flooding is critical to enable the timely activation of defense actions by both administrations and citizens. In this sense, early warning systems can alert the population of imminent danger, so as to encourage the adoption of safety measures and more cautious behaviors. To act for the territory and promote safeguarding activities, it is fundamental to assess and understand the area’s risk, the local forecasting systems connected to the warning services, the communication and dissemination of information on flood risk, and the community’s ability to respond. These systems are designed to reduce flood risk through early warning, providing forecasts of the timing and scale of events and the possible damage that may result, as well as providing adequate knowledge of the risk and appropriate responses. When well developed, these systems can improve flood management, reduce the number of casualties, and reduce the costs of recovery after the event. Their effectiveness depends strictly on developing a communication network and maximizing the number of people that it can reach in time.

Since the implementation of this kind of foresight is still being developed, it is crucial to have an adequate response from institutions to underpass flooding events. One of the very first actions that institutions can take in cases of flooded underpasses, especially before the flooding of underpasses when a heavy rain event is forecast, is to communicate to drivers alternative roads that can be taken.

Another action more focused on the prevention and on the divulgation of cautious behaviors is connected to driving schools and road security courses. If drivers are better educated on the effects of high-water situations on vehicles, such as impaired maneuverability, engine failure, or difficulties in opening doors, it is more likely that flooded underpasses would be perceived as a real danger.