Spatial Analysis of Flood Exposure and Vulnerability for Planning More Equal Mitigation Actions

Abstract

The issue of spatial equity of Nature-Based Solutions in cities generally concerns the spatial distribution of their benefits to local residents and other city users. In the context of flood risk management, planners are challenged to identify effective mitigation and adaptation measures that can generate benefits to the higher number of people and, more specifically, to people with highest levels of exposure and vulnerability. To address these issues, an essential step is to identify the geography of needs for mitigation, intended as prior areas in which to locate measures for flood risk mitigation. This study combines geospatial layers of multiple dimensions of exposure and vulnerability to flooding and identifies prior areas suitable for design scenarios for mitigation of flooding risk, for a regional case study located in Sicily. The results show patterns of exposure and vulnerability that vary according to locally relevant physical and social urban dimensions. Based on these results, proposals for mitigation actions are advanced with the overall objective of generating equal benefits to the most vulnerable exposed social subjects. Moreover, this study argues about the particular implications of implementing stormwater green infrastructure planning for equal beneficial distribution of the potentially achievable risk reduction.

1. Introduction

1.1. Flood Risk in Cities: Exposure and Vulnerability of Urban Contexts

Urban and infrastructure development generate dramatic impacts in the natural process of water drainage and regulations, replacing draining soil and vegetation with impervious covers, decreasing the amount of water that can be infiltrated in the ground and increasing the amount and velocity of water run-off [1,2]. This therefore increases the risk of flooding in cities. Urban development also strong consequences on regulating services supplied by urban ecosystems [3], increasing the economic costs associated with damages occurring to buildings, infrastructure and other urban facilities [4], agriculture [5], water quality [6] and eventually human health [7].

Urban systems present high levels of exposure and vulnerability to different climate change-related issues, especially urban flooding, which is a growing concern due to its direct impacts on residents, infrastructure, public services and commercial activities [8]. Urban flooding risk can include different dimensions of exposure and vulnerability, related to social and physical domains where the damages by urban flooding can be generated [9,10], and recent research on flood disasters has stressed the link between urban flooding and vulnerability [11].

A social dimension includes the direct possible harm that the flow of water can cause to people moving in cities during flooding events. Potential harm can vary according to the social group considered, with children, the elderly and people with disabilities being the most vulnerable subjects, due to their limited movements or slower reaction to harmful events [12].

A public dimension refers to the potential damage that can occur to public facilities and commercial activities. Public facilities include buildings and areas that host a wide range of functions and services, which are crucial in cities and are normally widely used by residents. Such facilities include schools/universities, hospitals, administrative buildings, cultural hotspots, squares, parks and gardens. A number of commercial activities are usually located in infrastructure, and especially in city centers, rather than in more peri-urban areas. Urban flooding can therefore have a direct and long-lasting impact on the buildings where these activities are located, as well as on the goods stored.

Infrastructure is another element of urban environment with high exposure and vulnerability to urban flooding: in urban roads, water run-off is the main cause of casualties, especially after flash flooding events. Direct impacts of these events are the traffic disruption [13], physical damage to infrastructure [14] and direct damage to people who may be stuck in vehicles, with fatal consequences in some instances [15]. This last impact is highly dependent on the nature of roads and the specific traffic conditions.

1.2. Urban Flooding Risk and Equity Issues

Solutions to tackle urban flooding can lead to unjust performances of the planned measures, as they can decrease the potential damages in areas where levels of social vulnerability is lower, or they can generate unequal distribution of the benefits [16]. This represents a problem of environmental justice related to urban flooding risk [17,18], occurring when water can flow to different areas after heavy rainfall events and produce damages to different social groups, depending on the configuration of physical and social characteristics of the urban system [19].

Floods generate unavoidable inequalities in terms of subjects who can be damaged by their impacts, because the spatial distribution of a hazard naturally depends on the hydrologic pattern and hydraulic network of a catchment. Researchers that have addressed the link between environmental justice and flooding have found disparities in the exposure to flood and pollution hazard among different racial and socio-economical communities [20]. Most of these researchers have found that minorities, residents with lower socio-economic status and other marginalized subjects live in areas with higher levels of pollution or flooding risk than other social groups [21]. However, spatial planning and urban policies can generate avoidable inequities, as planning decisions (i.e., zoning) can expose people and buildings to flooding hazards in different ways. Additionally, flooding mitigation solutions can generate inequities in terms of generated benefits for urban communities, when these benefits are not spatially distributed [16,22].

From this perspective, it is important to identify scenarios of solutions to urban flooding that can be able to distribute the potential benefits where high levels of vulnerability are present, in a spatial dimension of equity [16,23].

Distributional equity involves designing planning policies so that flood mitigation services and infrastructure are directed to the neighborhoods and households most in need [24]. Indeed, understanding the quantity/quality of subjects who are mostly at risk of floods, as well as their locations, can be used to tailor mitigation strategies to target those most in need, and leverage the results to inform the process of identification of priority locations where interventions can mitigate both physical and social aspects of flood vulnerability. This requires understanding where scenarios of urban flooding mitigation are more urgent for the high levels of hazard, exposure and social/physical vulnerability.

This article tries to address this issue by proposing a method of identifying areas with a high need for solutions to mitigate urban flooding risk.

Previous authors have stressed the importance of developing methodologies to identify priority neighborhoods for adaptation to flooding risk based on factors such as social vulnerabilities and flood risk [25]; and to highlight potential opportunities for urban planners to increase environmental equity in the siting of GI, as it relates to pluvial flood risk [26]. However, researchers very often analyze places’ vulnerability and exposure separately, ignoring risk drivers such as the traffic conditions on the road networks that represent an additional risk factor for which it is very difficult, if not useless, to decouple the social characteristics (e.g., age, gender, ethnicity, income, disability) and exposure (e.g., number) of car users. This study develops a method in which exposure and vulnerability can be analyzed in combination. It also includes the traffic conditions as a relevant element of risk, which is underrepresented, if not absent, in similar research and with regard to the scale and resolution of the presented case study.

The article is structured in the following way. Section 2 introduces the method and the geographical context of the case study investigated. Section 3 presents the results, identifying hotspots based on exposure/vulnerability (EXVUL) analyses and the related scenarios for prioritizing localization of planning mitigation interventions. Section 4 discusses the results obtained in light of recent related literature and research, advancing possible proposals for mitigation solutions according to the results obtained. Section 5 summarizes the main findings and identifies further streams of research.

2. Materials and Methods

2.1. Analysis of Place Exposure and Vulnerability

Spatial approaches and GISs are widely used for risk mapping and to examine, in particular, the geography of risk components, either as a whole or in isolation [27]. The research on social and built environment vulnerability to flood risk, in particular, has identified the risk posed by multiple hazards at the local level, suggesting both mitigation and adaptation inputs, contributing to more equitable risk reduction [28,29].

According to the ‘Directive 2007/60/EC’ [30] on the assessment and management of flood risks, ‘flood risk’ is defined as the combination of the probability of a flood event and the potential adverse consequences for human health, the environment, cultural heritage and economic activity associated with a flood event. The potential adverse consequences are expressed in terms of the number of inhabitants potentially affected; the type of economic activity of the area potentially affected; the installations which might cause accidental pollution in case of flooding and potentially affected protected areas and the other significant sources of pollution.

The potential of a hazard is filtered through the social fabric and the geographic context which interact, in their own condition of exposition, to create the overall vulnerability of a location [31]. Many definitions of vulnerability have been provided based on the disciplines of their origin. Among the long-standing interpretations, one by (Watts & Bohle, 1993) [32] defines vulnerability in terms of exposure, capacity and potentiality. Exposure and vulnerability are mutually related in generating potential damage, because the vulnerability of an element at risk tends to increase with its exposition and susceptibility to forces and impacts of the negative event [33].

In theory, it would be correct to affirm that elements can be exposed without being simultaneously vulnerable but, in reality and cautiously, flood-proofing of urban environments cannot reduce the vulnerability to zero.

Accordingly, this study explores the geography of exposure and vulnerability (EXVUL) in flood-prone areas. In particular, the place EXVUL is assessed based on an integrated analysis to combine geospatial layers of multiple dimensions (physical and social) of exposure and vulnerability to flooding, which are considered to be interrelated. Each dimension is expressed by an indicator whose corresponding value is attributed to the reference spatial unit where it is calculated (the census tract, namely the smallest territorial entities for which the Census of Population and Housing survey was carried out by the Italian National Institute of Statistics (ISTAT)). Indicators are derived using available data on fluvial and pluvial flood risk extent (official flood risk maps), the condition of human beings (number and age of exposed population) and the potential level of people fruition of places (area of public complexes and number of shops; road traffic speeds in exposed parcels of the census tract).

Thereafter, the values of EXVUL indicators were used to highlight hotspots of exposure and vulnerabilities (where scores of indicators show the highest values). Suggestions on potential planning strategies for mitigation actions are also provided for census tracts where high values of considered EXVUL factors occur contextually.

Hierarchy and relations of researched topics are graphed in Figure S1.

2.2. Data and Case Study

The applied methodology is based on a spatial analytic approach that returns the geography of flood risk hotspots based on the EXVUL assessment.

The method is designed to be applied at the regional scale so that the results can guide strategic planning of solutions for the mitigation of urban flooding. Despite the regional scale, the method is designed to provide results at the maximum resolution available, which is the one of census tracts. The method is applied in the Sicily region (Italy), which has experienced severe urban flooding events in the last 40 years [34].

Coastal watersheds, in particular, have experienced intense processes of urban sprawling and resulting soil sealing: they represent the highest-density urban landscapes and host the majority of the population.

There, largely, interferences between the natural drainage system network and the human settlements have caused heavy criticalities in terms of hydraulic risk. Sicily, similar to other Italian regions, has suffered from catastrophic landslide and flooding events which have created 683 victims and damages of more than EUR 15,000 million. Hydrologic and hydraulic failures are widely diffused, especially with regard to the minor drainage network and within urban areas, with serious risky conditions along the coasts [35].

These factors could also be exacerbated by climate change processes. Rainfall statistics changes in Sicily have been studied and provided results of extreme events analysis, as well as precipitation events of long durations. An increasing trend for rainfall of short durations and, conversely, a decreasing trend of precipitation events of long durations have been highlighted. In particular, an increase in short-duration precipitation has been observed, especially in stations located along the coastline, and heavy–torrential precipitation events tend to be more frequent at regional scale [36].

In the current study, the following geo dataset was used. Census tracts from the 2011 national census were available as a polygonal shapefile, with associated socio-demographic attributes. Census Tracts were considered as the reference spatial units for all further analysis. The primary source for delineating regional flood hazard and risk were the maps by the Regional River Basin District authority [37]. Maps depict the spatial extent of the areas at flood risk, based on an assessment method where the damage is a function of the combined exposure and vulnerability. Only areas at high and extremely high risk of flooding were investigated. In these areas, valued societal elements such as people and buildings might potentially be affected by high and extremely high adverse consequences caused by flood hazard associated with the 50-year, 100-year flood and 300-year flood scenarios (Figure 1 and Table S1).

We also used Open Street Map to identify the locations of shops, because it is one of the most-used sources of open access geographic information [38] and because this kind of information is not available from the official cartographic sources. To extract this information, the QGIS plugin QuickOSM was used to download the data from Overpass server.

Finally, Traffic Data were extracted from the World Traffic Service of ArcGis Online Resources, which includes different type of traffic data, such as historical, live and predictive traffic [39]. The color-coded traffic map layer can be used to represent relative traffic speeds. The service works globally and can be used to visualize traffic speeds in many countries; however, the data coverage varies worldwide and for the Italian region, it is based on the “predictive traffic” speed source.

2.3. Indicators for Exposure and Vulnerability

2.3.1. The Social Dimension

Because social vulnerability is not a directly measurable phenomenon, different indicators have been widely used as proxies [40]. Among these indicators, the spatial distribution of the population represents an aggregated measure of the extent of people exposure, which can also be combined with information on demography. In this study, the exposed population is disaggregated based on age characteristics: age strongly influences mortality for flood events [41], and the elderly and children are considered particularly vulnerable to the direct and the indirect health consequences of floods [42].

To estimate the number of residents vulnerable to flooding, we first selected the census tracts intersecting areas of very high or high risk (risk level 4 and 3, respectively). However, the portions of census tracts that intersect areas under risk can vary greatly and the number of residents in census tracts cannot be a precise indicator of people actually at risk. To overcome this limitation, we adjusted the number of residents in each census tract with a weight that accounts for the area of residential buildings that are located in risk areas (Figure 2).

The residents exposed to floods were then computed as a portion of the total residents in the i-th census tract with the following formulas:

$$I_{p_i}=\frac{A_{R_i}}{A_{to{t}_i}}·Po{p}_{to{t}_i}$$

(1)

$$I_{e_i}=\frac{A_{R_i}}{A_{to{t}_i}}·El{d}_{to{t}_i}$$

(2)

$$I_{c_i}=\frac{A_{R_i}}{A_{to{t}_i}}·Chil{d}_{to{t}_i}$$

(3)

where:

$I_{p_i}$,$I_{e_i}$ and $I_{c_i}$ are the adjusted values of residents living in the residential buildings in census tracts located in risk areas, respectively, for all of the population, for elderly people (over 74 years old) and for children (under 9 years old);

$A_{to{t}_i}$, is the total area of buildings in the i-th census tract;

$A_{R_i}$, is the area of the buildings located in risk areas of the i-th census tract;

$P_{to{t}_i}$ is the total value of population in the i-th census tract;

$El{d}_{to{t}_i}$ is the number of people aged over 74 years;

$Chil{d}_{to{t}_i}$ is the number of people under 9 years old.

Residential buildings have been extracted via regional vector cartography.

2.3.2. The Road Traffic Dimension

Flooding poses an important threat to roads and can lead to massive obstruction of traffic and damage to road structures, with possible long-term effects [43]. Parameters used to assess risk from natural hazards on roads often include so-called static traffic values such as the length of the hazardous section, the average number of vehicles per time unit and the speed of vehicles [44].

To understand the potential negative impact of flood events, we referred to a typical condition of workday traffic. A rush hour of a recent workday (1 pm, 17 October 2022) was selected. To this end, we used the ArcGis Traffic, an online service providing traffic conditions on the network of main streets. The service returns values of traffic ranging from 4 (free traffic condition) to 1 (stop and go condition).

The images of traffic conditions generated by the Traffic Service were exported as raster files and then converted to a polygonal vector shapefile, representing the road network with the associated traffic value.

Roads intersecting areas of very high and high risk were thus selected. In order to attribute an overall indicator of traffic condition to each census tract, the following indicator was calculated:

$$I_{t_i}={{\displaystyle \sum }}_{j=1}^n{A}_j·t_j$$

(4)

where:

$I_{t_i}$ is an indicator of traffic condition in the i-th census tract;

$A_j$ is the area of the j-th road exposed to risk in the i-th census tract;

$t_j$ is the traffic value associated with the j-th road, ranking from 1 (stop and go) to 4 (free traffic condition).

An example of selected roads with traffic values and census tracts is reported in Figure 3.

2.3.3. The Dimension of Public and Shopping Complexes

The public built-up spaces, including shopping complexes, bring together many different users living in urban environments and act as centers of attraction. This implies a concurrent increase in exposure and vulnerability for people and buildings located in census tracts where high levels of risk are present.

To estimate the quantity of such public buildings, we used a simple indicator accounting for the areas covered by public buildings, schools, religious, social and sport facilities (Formula (5)) and the number of shopping complexes (Formula (6)). Both these indicators have been calculated in the census tracts that intersected the areas at very high (R = 4) and high risk (R = 3) (Figure 4).

$$I_{p{c}_i}=A_{p{c}_i}$$

(5)

$$I_{s_i}=n_{s_i}$$

(6)

where:

$I_{p{c}_i}$ is Indicator for EXVUL of public complexes;

$I_{s_i}$ is the indicator of EXVUL for shops;

$A_{p{c}_i}$ is the total area covered by public buildings included in the vector regional cartography;

$n_{s_{i }}$ is the number of shops located in the i-th census tract, extracted from Open Street Map data.

2.4. Vulnerability/Exposure Hot Spot Analysis and Scenarios for Risk Mitigation

The identification of hotspots of EXVUL intends to select areas with high levels of exposure and vulnerability where different scenarios of mitigation measures can be planned.

Given the multi-dimensional nature of flood EXVUL, the scenarios are designed according to the different degrees of EXVUL for each of the evaluated dimensions, therefore identifying spatial priorities for mitigation measures [45]. These scenarios are based on the number of census tracts where indicators used to assess EXVUL dimensions present different values.

For each dimension of EXVUL, two scenarios of prior mitigation measures are identified:

• The max_scenario is obtained by selecting the census tracts where one or more of the considered indicators of EXVUL dimension assume highest values.
• The min_scenario is obtained via selection of the census tracts, where all of the considered indicators at the same time show the highest values.

Highest values are considered here as values above the 1 standard deviation for each indicator considered, a condition often used for hot spot identification [20].

The max scenario results in the highest number of census tracts being selected, as this scenario follow additive criteria, grouping census tracts that have the highest values for each of the indicators considered; the min_scenario returns a scenario with a minimum number of census tracts, as it means that multiple conditions are respected at the same time. The min_scenario thus represents a scenario with the highest level of vulnerability and exposure, as all the indicators of EXVUL dimensions present, at the same time, the highest values. This scenario thus identifies the census tracts at which to plan mitigation measures with high priority.

To select the census tracts to be included in the scenarios, specific conditions were elaborated on in GIS with SQL queries investigating the values of the different indicators of the social, public and shopping complexes and road traffic dimensions, as indicated in

Table 1. Conditions for the values of indicators for the identification of scenarios of mitigation measures.

Exposure/ Vulnerability Dimensions	Conditions	Scenario	Priority
Social	highest total pop OR highest_children OR highest_elderly highest_total_pop AND highest_children AND highest_elderly	Max_scenario Min_scenario	High Medium
Public and shoppingcomplexes	highest_public_complexes OR highest_shops highest_public_complexes AND highest_shops	Max_scenario Min_scenario	High Medium
Road traffic	medium-high traffic highest_traffic	Max_scenario Min_scenario	High Medium

.

Since the previous scenarios focused on single EXVUL dimension, a further step of the methodology is to identify scenarios that take into account and integrate all the EXVUL dimensions at the same time.

This approach is summarized in Figure 5, reporting all different combinations of values of the three EXVUL dimensions (as reported in Table 1). For example, the scenario (max_Social, max_Public and shopping complexes and max_ Road traffic) includes those census tracts belonging to the Max scenarios previously identified for the three EXVUL dimensions (Table 1). This scenario results in the highest number of census tracts because it follows additive criteria in the selection of census tracts. In contrast, the scenario (min_ Social, min_ Public and shopping complexes and min_ Road traffic) includes census tracts belonging to the Min_scenarios for the three EXVUL dimensions. All other scenarios are similarly identified by different combinations among Max and Min scenarios.

3. Results

3.1. EXVUL Analysis

3.1.1. The Social Dimension

Results for the social dimensions return 90,106 residents living in areas at high and very high risk of flooding, corresponding to the 30% of the entire population living in those census tracts that intersect areas at risk. In total, 8869 of these are people aged over 74 years, while 8630 are children under 9 years old.

Figure 6 maps the distribution of values of the indicator $I_{p_i}$ in the census tracts intersecting the areas at very high and high risk.

Increasing values of $I_{p_i}$,$I_{e_i}$ and $I_{c_i}$ express increasing degrees of exposure and vulnerability, therefore identifying prior census tracts at which to concentrate measures for mitigation of flooding risk. In particular, values of $I_{p_i}$ correspond to degrees of exposure, while values of $I_{e_i}$ and $I_{c_i}$ suggest higher levels of vulnerability. Census tracts with higher values of $I_{p_i}$,$I_{e_i }$ and $I_{c_i}$ are small areas located in urban areas, with an overall high density of residents.

3.1.2. The Road Traffic Dimension

The road traffic network is mainly characterized by conditions of free flow (value 4) for 52% of the entire network. Moderate traffic (value 3) conditions occur for 30%, while slow (value 2) and stop and go (value 1) conditions can be found for 16% and 2%, respectively.

Figure 7 maps the distribution of values of the indicator $I_{t_i}$ in the census tracts intersecting the areas at very high and high risk.

Likewise, for the social dimension, increasing values of the indicator express increasing degrees of exposure and vulnerability, which is equal to zero only when there are no road areas at risk.

3.1.3. The Dimension of Public and Shopping Complexes

A total of 115 buildings, corresponding to 2,160,697 m² of public complexes, are included in areas at high and very high risk. Very importantly, 35 of these are schools covering about 149,118 m². Among the selected elements are hospitals areas (38,823 m²), social spaces (33,115.8 m²), religious complexes (12,646 m²), touristic complexes (2,152,154 m²) and outdoor sporting centers (88,280 m²). A total of 477 shops have been extracted from the Open Street Map, accounting for 4% of the total shops included in the database in the region. Census tracts included a very varied number of shops, ranging from 1 to 60, with the highest number in urban areas. Figure 8 shows the map of the sum of values of the indicators $I_{p{c}_i}$ and $I_{s_i}.$

3.2. Exposure/Vulnerability Hotspot Analysis and Scenarios

Table 2. Values of variables for the Population dimension and number of census tracts included in scenarios for mitigation measures.

Social Dimension	Mean	Std Dev	# Census Tracts > Std Dev
Total population (#)	104	134	204
Children (#) Elderly (#)	10 8	13 14	195 185
	# census tracts
Max_scenario	263
Min_scenario	124

,

Table 3. Values of variables for the Public and shopping complexes dimension and number of census tracts included in scenarios for mitigation measures.

Public and Shopping Complexes	Mean	Std Dev	# Census Tracts > Std Dev
Public complexes (m2)	24,553	98,831	5
shops (#)	3	6	12
	# census tracts
Max_scenario	17
Min_scenario	0

and

Table 4. Values of variables for the Road Traffic dimension and number of census tracts included in scenarios for mitigation measures.

Road Traffic	Mean	Std Dev	# Census Tracts > Std Dev
Traffic	7965	18,524
High medium-high			145 358
	# census tracts
Max_scenario	358
Min_scenario	145

report the main statistics for the values of indicators considered and the number of census tracts included in the scenarios for each of the EXVUL dimensions. Based on the tables, it is evident that the Social and Road Traffic are the dimensions with the highest number of census tracts in both scenarios considered, and therefore the dimensions with the most widespread vulnerability. This depends on the spatial density of the variables of population and traffic, present in all the region and significantly higher than the density of public buildings and shops.

The maps hot spot scenarios for each of the vulnerability dimensions are shown in Figure 9, Figure 10 and Figure 11.

For the social dimension, the scenarios identified include census tracts mainly located in urban areas, as a direct consequence of the high values of all population variables in urban census tracts. Furthermore, the total proportion of the population living in the census tracts included in the max_scenario and min_scenario account, respectively, at 64,365 and 43,213, highlighting the high number of people exhibiting the highest level of vulnerability. The min_scenario for the public and shopping complexes did not return any census tract.

The scenarios resulting from the integration of the scenarios of the single EXVUL dimensions are reported in

Table 5. Number of census tracts after the integration of the scenarios of EXVUL dimensions.

Scenario		# Census Tracts	Scenario		# Census Tracts	Scenario		# Census Tracts
Max Max Max	Social Public and shopping Road Traffic	470	Max Max Min	Social Public and shopping Road Traffic	301	Max Min Max	Social Public and shopping Road Traffic	309
Min Max Max	Social Public and shopping Road Traffic	253	Max Min Min	Social Public and shopping Road Traffic	213	Min Max Min	Social Public and shopping Road Traffic	209
Min Min Max	Social Public and shopping Road Traffic	223	Min Min Min	Social Public and shopping Road Traffic	189

. They include a number of census tracts ranging from 470 in the most inclusive scenario (max_Social, max_Public and shopping complexes, max_road traffic) to 189 in the most restricted scenario (min_ Social, min_ Public and shopping complexes, min_ road traffic). These 189 census tracts are areas where all the three EXVUL dimensions have values above the one standard deviation at the same time and thus have the highest level of priority for the implementation of mitigation measures.

Figure 12 and Figure 13 map the scenarios of (max_Social, max_Public and shopping complexes, max_road traffic) and (min_ Social, min_ Public and shopping complexes, min_ road traffic), respectively.

Figure 14 and Figure 15 show more in detail an example of census tracts included in these latter scenarios of mitigation measures, in a town located between the northern seaside and surrounding mountains. Figure 14 maps census tracts at highest priority, satisfying the conditions Min_Social, Min_Public and shopping complexes, Min_Road Traffic. Figure 15 highlights census tracts satisfying the conditions Max_Social, Max_Public and shop-ping complexes, Max_Road Traffic. In both scenarios, census tracts are specially located along the main roads with high traffic and close to river beds. Blue colors indicate areas at very high risk (as reported in Section 2.2).

4. Discussion

This work intends to deepen research on environmental justice, focusing on the specific dimension of the geography of risk [46] and beneficial distribution of benefits that are potentially achievable via mitigation measures [47] in a spatially explicit way [16].

This research links the evaluation of exposure and vulnerability to flooding to the location of the possible mitigation actions that can be implemented, according to the overall principle of generating equal benefits to the most vulnerable/exposed census tracts. While an extensive body of research is working on improving approaches to evaluation of urban flooding in different cross-cutting fields [48,49], a more limited effort has been made to identify possible strategies and actions for flooding mitigation in areas of high vulnerability and exposure, so as to generate benefit to the higher number of communities.

Results obtained highlight that, when offering the same level of risk, scenarios of EXVUL are related to the combination of the considered dimensions, showing a complex spatial behavior, which mainly depends on the location of residents, hydrological hazard and infrastructure density. Typically, census tracts with highest need for mitigation action are the ones in dense urban areas, where the concentration of residential buildings, high-traffic streets and commercial and other public activities is higher (Figure 14 and Figure 15). Although limited in number, there are examples of high levels of EXVUL in rural parts of the region: Figure 16 shows particular situations of a high level of EXVUL due to the presence of a high-traffic street built beside the bed of a stream prone to flooding.

The combination of different values of EXVUL variables can be explored further to identify more precise priorities in mitigation actions to be planned and implemented, so they can be better perceived as more socially equitable [50]. In particular, the implementation of stormwater green infrastructure (SGI) as planned actions to reduce EXVUL can be designed to be upstream of the involved census tract, as well as within its area.

The provision of benefits and impacts of SGI performance do not interest the sole place where they are deployed and nearby community. For example, SGI located and concentrated in a particular areas of the urban watershed might redistribute floodwater elsewhere in the city and generate a different level of reduction of pluvial risk by changing its spatial distribution with respect to the original scenario without SGI [51].

At the same time, planning the implementation of SGI primarily within hotspot census tracts, as identified in Section 3.2, is supported by the principle of dealing with runoff locally (source control), which is always effective for the management of drainage exceedance during extreme events.

Many types of SGI are designed to reduce run-off under normal conditions and mitigate more frequent and less intense events. Despite this, the interest in SGI as a climate change adaptation planning strategy stays high, in the face of potential changes in precipitation patterns. The latter, in particular, called into question the reliability of the “design storm” or “return period” approach in supporting flexible long-term planning under climate and land use processes uncertainties [52,53].

Whenever possible, for urban areas with high values of EXVUL, the integration of local measures in public/private areas with upstream interventions to mitigate water run-off should be chosen. For more rural areas, with a limited amount of residential buildings and public functions, mitigation measures such as natural drainage or detaining systems should be implemented to avoid water reaching the street network.

Furthermore, the differences in urban and rural areas call for different stormwater installations according to their suitability to physical characteristics, land availability, ecological and local hydrological processes [54]. For example, in more compact urban areas, the availability of public areas in which to implement SGI could be limited, so mitigation actions must also include private areas, i.e., retrofitting residential buildings with green roofs [55] or public spaces with permeable pavement or vegetated swale [56]. In peri-urban areas, the usual higher availability of open spaces can suggest the design of multifunctional public spaces able to detain/retain water run-off [57] and to provide cultural services for peri-urban communities that often have a more limited access to green spaces [58,59].

The results of the hotspot analysis may also be leveraged to consider diverse adaptation actions for urban census tracts, directly addressed to lower the exposure of people and built structures, namely relocation of buildings, urban regeneration and soil compensation for de-sealing.

4.1. High Resolution of Analysis

Previous research indicated that assessment of exposed population is highly sensitive to the resolution of geographical data about flood hazard and population data [60]. Results from this research have returned an evaluation of exposure and vulnerability to flooding at the highest resolution available, with respect to many previous studies at the national level and considering the possible data that are publicly available. A recent national assessment by Italian Institute for Environmental Protection [34] analyzed exposure and vulnerability for Sicily using national census data and flooding maps and evaluated the total number of residents exposed to high hazards in the range of 126,000 to 138,000 (2.5% to 2.8% of the entire Population). Differences from the estimation of this research (slightly higher than 90,000 residents living in existing residential buildings) are related to the diverse applied methodologies. Other, more relevant differences between the methods used by national assessment and this research are reflected in the estimation of the total number of buildings exposed to high flooding hazard. This difference is mainly due to the fact that this research considered the sole residential/social/civil buildings included in the regional cartography (see Section 2.2), which are more precise data for refining the level of exposed resident population.

4.2. Limitations and Future Improvements

Variables considered in the population and infrastructure dimensions had a greater impact than variables of urban environment dimension: derived from census tracts, variables of populations are sampled and available for the entire region, and traffic information is also available for almost all type of streets (see Section 2.3.2). Variables of the urban environment have a much more scattered geographical extent due to their nature (Section 2.2), resulting in a general lower number of census tracts with high EXVUL. This difference in relevance should be taken into account when selecting suitable mitigation actions.

Not all the combinations of scores of EXVUL values have been explored, and the scenarios of EXVUL presented in Section 3.2 are derived from combinations of scores of variables that were higher than one std. dev (see Section 2.3). This choice has therefore not considered scenarios of “average” level of risk, with the general aim of identifying hotspots of EXVUL deserving prior attention from mitigation measures.

Although census tracts represent the geographical units with the highest possible resolution, there are some situations where they do not represent high levels of EXVUL in the best way. This typically happens where census tracts in rural areas intersect areas of high risk to a limited extent, but the entire census tracts are considered under hazard. Though this situation could be inaccurate from a geographical point of view, it depicts a more prudent scenario of attention to the intersected census tracts.

Mitigation solutions can be located in hotspots of exposure–vulnerability, but interventions located upstream of those areas should also be considered, in order to reduce the water run-off that is generated upstream. As a consequence of the complex relationship between SGI and hydrological processes, the use of modeling becomes a future compulsory step in the distributive analysis of SGI positive impacts in terms of number and location of beneficiaries and to evaluate which scenarios could lead to more equal benefits among the social subjects involved.

Other variables related to different aspects of EXVUL can be integrated in the assessment to improve the relevance of final scenarios of EXVUL hotspots. Examples include more specific social indicators [20], characteristics of buildings [61] or detailed characterization of the urban environment [62]. However, all these possible improvements require specific modeling and data. This can hamper the replicability of the model in large areas such as regions, which are appropriate territorial units for planning decisions about where to concentrate mitigation solutions. The variables used in this research represent an appropriate choice, due to their geographical extent (the entire region), their periodic update, and their availability and ability to capture the most important dimension of EXVUL.

The results cannot be considered conclusive and permanent. Data used to map flood hazard and risk are continuously updated by the regional River Basin District Authority, which is also a requirement of the Flood Directive.

Very importantly, the occurrence of flash floods as a consequence of climate change has very recently been included in the update of the Regional Flood Risk Management Plan, which is finalizing the analysis of flash floods vulnerability for a limited number of river catchments in Sicily [63].

Flooding is as much a dynamic process as urbanization. The mutual and continuous changing status of flooding and urbanization phenomena must be carefully considered in planning actions, which cannot neglect the importance of scenario simulation for reliably envisioning equitable transformations, the processes used to create and implement them and the resultant distributions of goods and hazards. As cities increase investment in SGI, they must grapple with embedded, multifaceted uncertainties; in fact, patterns of inequitable distributions of flood EXVUL might be reinforced by inefficacies of SGI planning, positively influencing the distributional equity of flood hazard and improving unequal urban conditions [64].

5. Conclusions

Flood risk in cities continues to be a challenge for policies and spatial planning managed by administrations, from local to wide area scales. New interdisciplinary and integrated analytical tools, spatially explicit methods, experimental and modeling approaches are continuously proposed by researchers to investigate the risk components.

Based on these premises, this paper has spatially evaluated the combined flood exposure and vulnerability at the regional scale, based on the use of proxy indicators for seizing the multidimensional nature of EXVUL.

Accordingly, this study has also defined a comprehensive and valid methodological framework in which diverse complex aspects could be concurrently and uniformly analyzed, including factors that are rarely explored in risk assessment research, such as traffic conditions.

Based on hotspots of EXVUL and with details based on the use of census tract as the spatial reference unit, this study has thus mapped the geography of mitigation priorities, namely prior areas in which to plan and design mitigation and adaptation planning strategies for flood risk.

The previous step is fundamental to the debate on the equity of benefits generated by planning strategies and to the achievement of fairness in bearing environmental burdens. This is also because disentangling factors of flood exposure and vulnerability may determine the misrepresented geography of needs and related mitigation priorities.

Moreover, few works have discussed the customization of mitigation decisions to the spatial layout of EXVUL hotspots, especially with approaches/models used to plan flood mitigation scenarios at the regional level while maintaining high resolution (the census tracts).

Lastly, aiming at flood mitigation, this paper reviewed stormwater green infrastructure design by considering the changing characteristics of hotspots between urban and rural settings. The latter is far less investigated by literature in terms of flood EXVUL, and less understood in its relationship with equity issues.

Uncertainties remain regarding for whom and to what extent SGI delivers disservices such as changes in the distributions of value and hazards in the urban landscapes.

Further research efforts must be channeled into comparing the geography of mitigation priorities with the geography of benefits generated by stormwater green infrastructure, from the perspective of distributional equity. In fact, when planning for flood risk mitigation, environmental injustice can arise for communities, especially when the excess of water is simply re-distributed within the urban fabric from certain locations to others.