CERCA (Cascading Effects in Risk Consequences Assessment): An operational tool for geo-hydrological scenario risk assessment and cascading effects evaluation

The risks associated with severe weather events, such as floods and landslides, are major natural hazards that affect Italian territory and cause causalities, severe damage to human property and infrastructure, and economic losses. In recent years, there has been a growing awareness that systematic and comprehensive investigations into the development of damaging events are crucial to gain experience and knowledge for efficient forecasting and technical countermeasure designs. This involves an advancement towards a more holistic understanding of the sources, processes, and consequences of exposed elements in order to implement effective strategies and options for risk reduction.

Regarding floods in the USA, surveys have been undertaken to report flash flood events, including details about their magnitude and spatio-temporal extent as well as their disastrous consequences, mainly reported in two datasets: the Severe Hazards Analysis and Verification Experiment (SHAVE) and the US National Weather Service (NWS) Storm Data, which consist of trained spotter reports (Gourley et al. 2010; Calianno et al. 2013). In the European context, the standardisation of procedures for the survey and analysis of events has developed notably due to the Floods Directive (FD) of the European Parliament of 2007 (Floods Directive 2007), which specifically addresses the need for national catalogues of past flood events and the initiatives that are supported by European projects such as FLOODsite (FLOODsite Project Deliverable D9.1 2007) and Hydrometeorological Data Resources and Technologies for Effective flash flood forecasting (HYDRATE) (Gaume et al. 2009). Nevertheless, it is noteworthy that there are some inconsistencies between definitions and taxonomy of the types of hazards, as well as criteria and classification of loss indicators across different available reporting schemes (e.g., FD Reporting Schema; Guidance for Reporting under the Floods Directive – Guidance Document No. 29; the INSPIRE Natural Hazard Classification, https://inspire.ec.europa.eu/id/document/tg/nz; the JRC-DLD Guidance – Joint Research Centre, Guidance for Recording and Sharing Disaster Damage and Loss Data, https://publications.jrc.ec.europa.eu/repository/handle/JRC95505).

In Italy, the Civil Protection Department makes available to Regions and River Basin Districts Authorities a web-GIS platform, named FloodCat, which aims to function as a catalogue of flood events to address the mentioned requirements of the FD. Moreover, regarding the post-flood surveys accommodating both quantitative and qualitative analyses, several research groups have evaluated the social and economic consequences of flood events (e.g., Ballio et al. 2015; Carrera et al. 2015; Menoni et al. 2016; Amadio et al. 2019; Molinari et al. 2020) or, more generally, the damaging geo-hydrological events resulting from rain-triggered floods and landslides. The research mostly focused on a single event or a limited area (Aceto et al. 2016; Caloiero et al. 2017) or is no longer updated (Guzzetti et al. 1994).

Nevertheless, the collection of information relating to events and their consequences are not standardised and varies between subjects who collect data for different purposes (e.g., performance evaluation of the warning system, damage recognition to infrastructure, recovery plans, and insurance refunds). Building this knowledge base at the national level can be time-consuming and demanding in terms of human resources. It requires accessibility to data, which is often insufficient and uncertain, and suffers from a lack of homogeneity on a large scale, which is necessary for the creation of multi-event databases with respect to the various needs imposed by integrated risk management. The National Warning System for real-time forecasting of this type of natural hazard requires harmonic development of all its different components: forecast/monitoring of the forcing, simulation/monitoring of phenomena, warning, and estimation of consequent impacts.

Standardisation of data formats for emergency classification, improvement of the risk scenario and events knowledge, and improvement of the exchange of information between the various stakeholders involved in risk management are some of the requirements that emerged in interviews with different organisms involved in the Civil Protection and emergency management in the Italian context within the RECIPE (Reinforcing Civil Protection capabilities into multi-hazard risk assessment under climate change) project (RECIPE Deliverable 2.2 2020).

Single-hazard approaches in which hazards are treated as isolated and independent phenomena are common. However, their prevalence can distort management priorities or underestimate risks (Gill & Malamud 2016). It is evident that natural hazards intersect with one another and/or interact with other possible hazards, increasing the complexity of the overall impacts of disasters on populations and economies. This calls for a reformulation and paradigm shift from a single-hazard perspective to a multi-hazard and systemic risk perspective, a notion supported by the national commitments to the UNISDR Sendai Framework for Disaster Risk Reduction. Multi-hazard assessments may also recognise the spatio-temporal overlap of natural hazards (Kappes et al. 2010; Gill & Malamud 2014), dependence between individual hazards and the interactions with anthropogenic processes and built environment, which often amplify impacts (e.g. Arrighi et al. 2021). Many methodologies available in the literature allow for the identification and visualisation of potential natural hazard interactions (e.g., Del Monaco et al. 2006; FEMA 2011 – HAZUS multi-hazard analysis levels; Marzocchi et al. 2009; Schmidt et al. 2011; Gill & Malamud 2016; Ming 2019); tools have also been developed within the context of the EU FP7 project MATRIX, including New Multi-Hazard and Multi-Risk Assessment Methods for Europe (Aspinall et al. 2014), although they are sometimes complex and require significant amounts of data.

The importance of a holistic approach to natural hazard assessment is also needed to capture the dimensions of interconnectedness and cascading effects on social, economic, and environmental systems, also known as ripple effects or domino effects (Zuccaro et al. 2018). The growing concern over the domino effect and the forensic disaster analysis approach is not new in the incident and industrial risk assessment (highlighted in Europe by the Seveso Directives). This approach has more recently been applied to natural disasters (Mendoza 2019) to understand the fundamental causes of natural disasters and their impacts on society, building resilience, and reducing risk through learning from past experiences (Integrated Research on Disaster Risk IRDR 2011).

In this study, we examined the cascading impacts that can arise from geo-hydrological hazards and provided a methodology for an integrated scenario assessment for strategic risk management in Italy named CERCA (Cascading Effects in Risk Consequences Assessment). The criteria followed were to use and optimise existing data and instruments and to be consistent with the national Civil Protection Guidelines, in order to produce a simple and effective operational tool. The main aim was to define a working framework for the classification of events and related damage, which can be useful for building event scenarios and sufficiently easy to implement in a real-time operational context. This study addresses, from a methodological and experimental point of view, the prediction/assessment of the effects of extreme rain events in complex areas, particularly in densely inhabited and highly anthropic areas. A dense network of physical, cultural, social, and economic interrelations makes these areas particularly vulnerable.

The framework is structured to be used as a tool for typical ‘scenario analysis’ in a multi-hazard context through the qualitative assessment of cascading effects and consequences on different categories of elements at risk, particularly in terms of human losses, caused by a forcing event. All available information can be further systematised and analysed in a multi-use view of the information itself, purposed to help decision-makers, enhance critical strategic thinking, and improve citizens' preparedness.

The analysis of a significant variety of case studies of past events in the national territory was used to test and verify the proposed methodology at different scales. Information from the ad hoc collation of data or available from existing databases was organised with homogeneous criteria to build scenarios (triggering processes and impacts) that are significant for the Italian situation. The selected events ensured a satisfactory coverage of different climatic, geo-lithological, topographical, and land-use conditions of different parts of the national territory, as well as different types of forcing events responsible for the damage. The spatial unit at which events were analysed was determined from the Warning Zones (WZ) of the national warning system, which relies on a mosaic over the entire national territory (average size of approximately 2,000 km²) that represents markedly homogeneous areas by the type and severity of expected events and related effects.

In this study, the CERCA approach was assessed to determine if it can be considered an effective and consistent framework:

• to process post-disaster information at the local level; ‘ex-post’ surveys help to identify site-specific dependencies based on local hazard proneness and exposure and vulnerability conditions as well as to identify effective countermeasures;

• for the analysis and efficient active surveillance of the real-time evolution of critical situations, helping operative structures of civil protection to continuously update the picture of phenomena occurring in the WZ;

• to provide general dependency matrices to be used in the definition of ‘ex-ante’ scenarios and recurring cascading event trees through analysis of numerous past events.

In summary, the study seeks to:

• define a method for collecting and organising information relating causal relationships to the effects associated with extreme rain events;

• enhance the knowledge, documentation, and understanding of past events;

• propose examples of the use of the method according to different objectives in relation to the scale of analysis and the needs of stakeholders.

In the following sections, we first outline the concept of the proposed framework and its components. This is followed by a description of applications at various levels, local and national, to a dataset of past events with significant adverse impacts that occurred in Italy during the period 2004–2021.

The ‘Operational Guidelines of February 10, 2016’ (OG) were issued by the Italian Civil Protection Department (https://www.protezionecivile.gov.it/it/normativa/indicazioni-operative-per-l-omogeneizzazione-dei-messaggi-diallertamento-e-delle-relative-fasi-operative-per-rischio-meteo-idro) in order to promote standard languages, timelines, and operating procedures for the warning system on the national territory and the mosaic of its 20 regions, thus facilitating an effective exchange of information within the territorial levels of civil protection and communication with citizens.

We will focus on Annex 1 ‘Guidelines for the homogenization of the messages of the national warning system: levels of criticality and warning and related event scenarios’, which contains the ‘Table of weather related geo-hydrological and hydraulic warning and criticality’. The table describes the expected impacts and damage with a criticality-level alert for different risks.

Criticality levels are distinguished as ordinary, moderate, and high, in addition to a reference level characterised by the absence of relevant phenomena in relation to geo-hydrological (GH) or hydraulic (H) risk. The criticality of H risk arises from floods that affect large rivers, for which it is possible to predict the evolution of events based on monitoring of water levels. The criticality of GH risk arises from local phenomena such as certain types of landslides that can be triggered and mobilised by water, outflows in urban areas, and floods affecting small river networks. Furthermore, the OG introduce the criticality for ‘geo-hydrological risk due to thunderstorms’ (GHT), which is inherently characterised by a high forecast uncertainty in terms of location, timeline, and intensity and therefore cannot be subject to a reliable quantitative forecast. Criticality levels are uniquely associated with warning colour codes: the absence of relevant forecasting phenomena matches a green alert; ordinary, moderate, and elevated criticality correspond to yellow, orange, and red alerts, respectively.

The ‘Table of weather related geo-hydrological and hydraulic warning and criticality’ in Annex 1, for each criticality alert level, identifies two fields corresponding to expected phenomena and effects and damage. At each level, expected phenomena are characterised in terms of intensity and extent; effects and damage identify possible consequences for assets (i.e., infrastructure, residential buildings, productive activities, and people) and rely on the extent of damaged areas and severity of expected impacts. Regarding impacts on the population, yellow alerts concern ‘potential risk for the safety of people with possible loss of life due to accidental causes’, orange alerts correspond to ‘circumstances that could endanger the safety of people with possible loss of life’, and red alerts correspond to ‘high risk for the safety of people with possible loss of life’.

The CERCA methodology frames the problem as a typical scenario analysis through the assessment of possible cascading effects and consequences characterised by a cause/effect relationship produced by a triggering event (TE). In our scheme, we considered weather-related geo-hydrological and hydraulic events to be TEs. From this perspective, we considered the chain to start from the type of event generated by intense or prolonged rain and only implicitly assumed precipitation to be the main cause of the cascade of effects.

A TE can induce a single event-tree, consisting of a single chain of elements or more branches in parallel in the case of originating events inducing different phenomena or damage. For example, the cascading effects of a primary event such as a flood, characterised by different observed parallel phenomena (e.g., levee failure, flooding, bank erosion, etc.) affecting one or more classes of exposures, are described by different independent paths originating from the same TE.

Nevertheless, in a multi-hazard approach perspective, more than one TE can occur at the same time: the simplest case is to have independent non-interacting TEs and treat them as ‘parallel hazards’ (Liu et al. 2016) each having a separate event-tree; a second multi-hazard relationship between different event types consists of those that occur in chains when ‘an adverse event triggers one or more sequential events’ (e.g., landslides impacting on a river network and generating a flood); these are also called ‘domino effects’, ‘concatenated’, or ‘cascading hazards’ (Marzocchi et al. 2009).

The analysis of a relevant number of past events and related cascading effects trees can be also used to provide general dependency matrices between different elements of the chain for the assessment of ex-ante scenarios. As will be illustrated in Sections 4.2.1 and 4.2.2, matrices of cascading dependencies can be obtained representing the relative frequencies or conditional probabilities of occurrences of an element at a certain stage of the chain, given the occurrence of an element at the previous stage. This approach allows to visualise the interactions for all the possible combination of elements at different stages of the chain and to identify recurring cascading trees.

The first step of cascade scenario assessment is to identify the type of TE.

Nine categories of TE (Table 1) were identified. The main distinction was made with respect to the type of event (i.e., floods and landslides), but further distinctions were made based on the extent, evolution time, and impact on specific elements.

Fluvial or river floods that occur when the water level in a river or stream rises and overflows onto the surrounding banks, shores, and neighbouring land have been differentiated into three types based on the time of concentration of the contributing watershed (TE01, TE02, and TE03).

In addition, the ‘Pluvial flood’ events (TE04) that occur when heavy rainfall creates a flood independent of an overflowing water body were considered and described as surface water flooding that occurs when the drainage or sewer system cannot cope with the speed of rainfall and floods into roads and nearby structures. Landslides were differentiated into ‘Deep-seated slope failure’, ‘Shallow landslide’, ‘Debris flow’, and ‘Sinkhole’ (TE05, T06, TE07, and TE08). Moreover, ‘Thunderstorm’ (TE09) was included to account for heavy localised rain events, which are often accompanied by strong winds, hail, and lightning.

The characterisation of the consequences associated with a TE was accomplished through the process of identification and classification of the representative phenomena described in the OG. Phenomena support event descriptions in terms of mechanisms, characteristics, or localisation. An event scenario can exhibit more than one associated phenomenon.

Twenty-two REP were identified (Table 2). These phenomena could be related to a specific risk (GH, GHT, or H). Some REPs that pertained to minor rivers were differentiated from those that were observed in major rivers to separate REPs related to GH risk from those related to H risk.

The next step in the chain pertains to associating potential effects on human health and damage to properties and infrastructure. Damage assessments of natural hazards provide crucial information for decision support and policy development in the fields of natural hazard management and adaptation planning in response to climate change (Merz et al. 2010). Smith & Ward (1998), following the damage loss schematisation by Schuster & Fleming (1986) with reference to floods, classified damage into direct and indirect losses. Direct losses are caused by the physical contact of the floodwater with humans or property and indirectly affect networks and social activities, causing extra costs for actions taken to prevent other losses (e.g., loss of production for companies). Furthermore, it is possible to distinguish between immediate- or long-term consequences (e.g., interruption of communication networks and critical infrastructure) and tangible or intangible consequences (e.g., psychological effects of loss of life, displacement, and property damage). Similar concepts apply to landslide consequences (e.g., Petrucci & Gullà 2010) and can be valid for other natural hazards.

Methods for assessing these impacts range from quantitative to qualitative or descriptive. Here, we referred to a descriptive approach to possible direct impacts.

The major types of impacts included in the OG were clustered into eight main categories of exposed elements (Table 3). Particular attention was paid to the harmonisation with existing classification (e.g., FloodCat) and definitions in the sharing of damage and loss data among EU countries and institutions (like the Guidance for Reporting under the Floods Directive 2013).

The last additional level in the chain, when fatalities occurred, was represented by an analysis of the circumstances of death. A classification of the most common fatal accident occurrences was proposed to understand the impact of disastrous events on the population and to design recommendations for self-protective actions (see Petrucci (2022) for a comprehensive review of the classifications in literature; Papagiannaki et al. 2022). In the presented scheme, this step constituted an additional level of information that detailed exposure, E6a, referring to death or missing people.

In Table 4, various categories of circumstances that adequately explain the place and dynamics of the majority of deaths were proposed.

A catalogue of past events with significant adverse impacts that occurred in Italy during the period 2004–2021 was built. Only events for which certified information was available (accredited source) regarding the type and extent of the event and the damage that occurred were considered.

The major sources of information were:

• event reports drawn up by the Functional Centre of Civil Protection, concerning the meteorological forcing of the event, its hydro-pluviometric features, and the impact on the territory;

• declarations of state of emergency;

• the POLARIS website (https://polaris.irpi.cnr.it/) that publishes accurate and detailed information on geo-hydrological risks (Salvati et al. 2016);

• the Emergency Events Database (EM-DAT; https://www.emdat.be/) and the International Disaster Database by the Centre for Research on the Epidemiology of Disasters (CRED);

• the European Severe Weather Database (https://eswd.eu/) that contains important weather events that can endanger people or cause damage, and

• products from the Emergency Management Service (EMS) of the Copernicus Program (list of EMS Rapid Mapping activations).

From this dataset, only past events with relevant information describing cascading paths from the main TE to adverse consequences were selected to identify hazard/consequence dependencies. The methodology was thus applied to 49 case studies at the regional level, which involved 152 WZ. Figure 2 highlights the distribution of the selected events on the national territory, grouped at the regional level, after a refinement process.

The investigated area is frequently affected by high-intensity rainfall, commonly characterised by short duration and small extent (Ferrari & Terranova 2004), which historically caused flash floods and widespread triggering of landslides with loss of lives and severe effects on infrastructure and agricultural activities (the most significant events occurred in 1925, 1946, 1975, 1980, and 2000; Petrucci & Versace 2005).

The orography is peculiar; elevation ranges from sea level to 1,475 m with relatively high gradients; watersheds are small, extending from a minimum of 2.5 km² (Fellino creek) to a maximum of 66 km² (Colognati creek). The coastal zones are crossed by the Ionian Railway ‘Taranto-Reggio Calabria’, the State Road E90 (SS106 and SS106bis), and a number of provincial roads connected to the main urbanised areas and the inner mountainous zone.

The event of 12 August 2015 was one of the most severe events in the recent history of the study area.

Heavy rainfall caused flooding of various streams, rivers, and widespread shallow landslides. The rain data from the rain-gauge network of the Regional Functional Centre (Arcuri et al. 2015) showed rain maxima for durations of 3, 6, and 12 h, which can be considered as a 100-year return period event: the gauge in Corigliano recorded a cumulative rainfall of approximately 230 mm (from 00:00 am to 11:00 pm), with 155 mm occurring in only 5 h.

The area of the Citrea Stream basin (approximately 11 km²) in the former municipality of Rossano was greatly affected by rainstorm-induced ground effects (Figure 3).

In the early morning of 12 August at around 8:00 am, a levee breach occurred in correspondence with a directional change of the Citrea Stream, along a stretch of the stream confined to artificial (concrete) embankments before it flowed into the Ionian Sea.

Flat areas in the surroundings of the failure were hit by massive floods and mud; hotels and bathing establishments located in the seafront area were severely damaged (mostly at Lido Sant'Angelo at the outlet of the Citrea Stream). As a consequence of the heavy rains, a tributary of the Citrea, the San Paolo Stream, overflowed in the centre of Rossano, particularly in the urbanised areas of Rossano-Stazione. The San Paolo Stream, which mainly flowed artificially under streets and houses, damaged numerous commercial activities and houses, dragged cars, disrupted circulation, and endangered human lives.

The breaking of embankments and silting also affected Fellino, Acqua Fosso del Fico, San Mauro, and Gennarito creeks at several locations causing overflows in the surrounding floodplain areas.

According to the event report (AdB 2015), the village of Momena was severely hit as a consequence of the flooding of the Fellino (area of approximately 2.62 km²) and Acqua Fosso del Fico (area of approximately 5.27 km²) creeks due to the sudden collapse of embankments. The most severe damage to homes and commercial activities occurred in Lussemburgo and Stati Uniti Street, where the water mixed with mud and reached a height of one and a half metres. Immediately after the event, a large accumulation of sediment in the riverbed was observed, with an estimated volume of approximately 20,000 m³.

Furthermore, widespread slope erosion and shallow landslides occurred, approximately 5,300 (Rago et al. 2021), mainly soil slip and debris flows had been mapped, particularly along road SP250, which lead to the old town of Rossano.

Fortunately, no fatalities occurred, but more than 650 people had to be evacuated and housed in emergency facilities. Additionally, there was serious damage to tourist services, agricultural infrastructure, several interruptions to roads and the Jonian Railway, as well as severe problems with water and sewerage networks.

TEs identified for the case study were: TE01 ‘Flash flood and other rapid onset flood’ and TE06 ‘Shallow landslides’.

REPs and damage were associated with each TE.

Concerning TE01, REPs are:

• REP7: Flooding due to inadequate drainage or sewage systems, surface water flow;

• REP8b: Partial or total occlusion of bridges/riverbed narrowing/buried streams (minor rivers);

• REP9b: Raising water levels (minor rivers);

• REP10b: Bank erosion (minor rivers);

• REP11b: Sediment transport (minor rivers)

• REP15b: Widespread flooding (minor rivers);

• REP19b: Levee damage/break (minor rivers).

Damage reported pertaining to different assets:

• E1: Public and private buildings;

• E2: Transport infrastructure;

• E3: Defence works and hydraulic infrastructure;

• E5: Technological and service infrastructure;

• E6: Human health;

• E7: Rural land use;

• E8: Other.

The frequencies of the categorised TEs resulting from the analysis performed on 152 WZ are displayed in Table 6. In 78 of the 152 cases (over 50% of cases), more than one TE was observed; most of the time, a flood event was accompanied by a landslide type of event.

Among the total of 257 TEs, 54% were fluvial or pluvial floods, 23% were soil erosion or landslides, and 21% were thunderstorms. Only 2% were sinkholes.

TE04 ‘Pluvial floods’ was the most frequent TE, representing 30% of all occurrences, followed by TE09, ‘Thunderstorms’ (21%) and TE06, ‘Shallow landslides’ (17%). The incidence of the other TEs was less than 10%.

Table 7 shows the dependencies between the TEs and REPs resulting from the back-analysis performed on the selected damaging events in the form of the matrix of interactions. According to the findings regarding the occurrence of identified REPs (last column, TOT REP#, Table 7), REP7 ‘Flooding due to inadequate drainage or sewage system, surface water flow’ was the most observed phenomenon; REP22, REP2, REP9a, REP9b followed with 20 or more occurrences; REP12, REP13, and REP18 were never observed among the collective events.

Table 7

Percentage (%) of occurrence of REPs for each TE and total number (#) of REPs

View Large

The occurrence percentage of each REP for each TE was evaluated from the ratio between the number of REP occurrences observed for a specific TE and the total number of TE occurrences. The occurrence percentage of each REP for each TE was evaluated from the ratio between the number of REP occurrences observed for a specific TE and the total number of the TE occurrences. Specifically, the occurrence percentage of having the ith REP given the occurrence of the jth TE can be evaluated considering the relative frequency ⁠, where is the number of occurrences of ith REP and jth TE; is the total number of jth TE occurrences. For example, the value that can be read in Table 7 at the crossing of REP2 (soil slip) and TE06 (shallow landslides) means that nearly 89% of the cases where TE06 was considered as the main TE (namely 45, from Table 6), soil slips were observed. In some cases, there is a complete alignment between a REP and a TE because of an evident correspondence (e.g., REP6 and TE08 both refer to sinkholes). Overall, TEs generally implied more REPs: nine REPs were associated with TE01, six of them with an occurrence of more than 20%.

The matrix of incidence helps to evaluate the level of dependencies identified and to develop the first general qualitative evaluation of expected REPs based on the frequency of occurrence (low, medium, and high) for each TE. The assessment performed on specific territorial contexts through specialisations and customisations that consider events and specific local vulnerability conditions can provide reliable information about the chain of ground effects that are most likely to occur as a result of a given TE.

Table 8 shows the damaged assets for each REP. The percentage of occurrences of each type of exposed element related to each REP was calculated, as for Table 7, by the ratio between the number of damage occurrences for a specific REP and the number of the REP occurrences.

Table 8

Percentage (%) of damage occurrence on exposed elements for each REP

View Large

Overall, the most frequent damaged exposures were transport infrastructure (E2) and private and public buildings, particularly damage to the basement or ground floor/yards of buildings (only flooding, E1a). Impacts on transport infrastructure included either physical damage to the exposed elements or closure to traffic and operational interruption owing to imminent risk to stability (e.g., bridge pile scour) or to stakeholders (e.g., pedestrians and drivers). Only in 6% of cases REP occurred without consequent damage.

The percentage of damage to transport infrastructure was consistently above 50%, excluding REP1, 11b, 12, 13, 18, 20, and 21. Damage to the basement or ground floor/yards of buildings was mostly affected by REPs related to flooding and overflow (i.e., 15a, 15b, 16a, 16b, 19a, and 19b with percentages higher than 50%).

In addition, damage to technological and service infrastructure (E5) and evacuated population (E6c) were frequently observed (more than 20% of occurrences for eight REPs).

On 23 occasions, there were fatalities with a total of 52 deaths encompassed in category E6a. Referring to REPs, casualties were likely to occur with widespread flooding (REP15a and REP15b), debris flow phenomena (REP5), or partial or total occlusion of bridges/riverbed narrowing/buried streams (REP8b). A more detailed characterisation of the circumstances of death that further specifies the chain for category E6a is presented in Table 9. We expect that the results of this work will be helpful in designing recommendations for self-protective actions and proactive policies that can contribute to reducing the human death toll. The majority were flood fatalities (82%) occurring both indoors and outdoors and landslide-related fatalities (11%) occurring outdoors. Numerous people were affected unexpectedly along roads (35%) and travelling in vehicles (37%).

This study introduces CERCA, a framework for event scenario assessment that is coherent with the current Italian Civil Protection OG for the national warning system. The methodology was developed for weather-related geo-hydrological and hydraulic hazards and categories of elements at risk. The framework allows for flexibility in multi-purpose and multi-scale applications and proposes a common set of criteria and standards that can be valuable in the definition of ex-ante and ex-post risk mitigation strategies.

The proposed methodology was assessed using a case study concerning a local event in Rossano (Calabria, Italy) and 152 events in the WZ of the Italian territory that occurred during the period 2004–2021.

The first case details a post-disaster survey at a local level and is aimed at illustrating CERCA functionality in describing cascading effects.

The same framework can have a potentially relevant application as a tool for real-time analysis and monitoring of the evolution of phenomena for civil protection purposes. During an event or in the emergency phase, civil protection coordination centres receive and collect reports of impacts and damage from the population and involved stakeholders. CERCA, by using a base of standard classification, can support the work of real-time classification and understanding of the causal sequence of phenomena, particularly when dealing with multi-hazard events. Taking advantage of the organisation of reports and real-time direct observations about what is happening can lead to improvements in monitoring and managing the emergency from different perspectives and particularly by better coordinating all activities, prioritising interventions, taking the decision about emergency procedures (e.g., evacuation, closing of activities, reinforcement of protection works, etc.), and driving field surveillance towards emerging critical situations. An accurate verification and deepening investigation at different times in the post-event should be considered as mandatory to confirm data gathered in real time.

A second achievement concerns the possibility to identify site-specific dependencies based on hazard proneness, exposure, and vulnerability conditions and to highlight risks from cascade effects that are rarely predicted with the traditional ‘decomposition’ approach based on Hazard-Vulnerability-Exposure multiplication through simple maps overlay (e.g., consequences related to levee break as for the Rossano event). In this regard, customisation of the general theoretical framework to site-specific cases is a necessary step to produce reliable scenarios according to the vulnerability of the affected area and the magnitude of the forcing event, its duration and extension.

Highly detailed descriptions about precise location and the severity of observed phenomena as well as fatal circumstances should be considered in the development of site-specific applications aiming to effectively support decision-makers in their preparedness and disaster mitigation strategies.

At the same time, the chain here illustrated should be linked to characteristics of the forcing rainfall so as to furnish a comprehensive view of possible hazard/impact scenarios that can be used in the forecasting phase.

The national-scale back-analysis, although incomplete, offers an overview of the chains generated by TEs and the information summarised in matrices provides an overall picture of dependencies limited to the number and type of investigated events. The analysis showed that in over 50% of investigated cases, more than one TE was observed, most of the time floods accompanied by landslides, confirming the necessity for multi-risk analysis. TE04 ‘Pluvial flood’, particularly affecting urban areas, was the most frequent TE with 30%, mainly causing damage to the basement or ground floor/yards of public and private buildings (E1a) and to transport infrastructure (E2) (49 and 95% of the times, respectively). This preliminary assessment is valuable to support further analyses that could explore also the dependency from the severity of the event and, as stated above, from the forcing rainfall. In this regard, the proposed approach is part of a long-term cooperation with Civil Protection that aims at creating a comprehensive and easy-to-use framework for data storage and analysis based on common criteria and standards. Ongoing research based on the CERCA framework concerns the attribution of a severity class to REP and DEAR in order to develop a procedure for an ex-post evaluation of an event severity colour code based on ground impacts coherent with the national Civil Protection OG. We believe that this assessment is crucial to adequately define cascading effects scenarios as well as to properly assess the performance of the forecasting and warning system.

The analysis of death circumstances is instrumental in addressing risk mitigation and by helping in conveying clear and effective messages about self-protective behaviours to enhance population preparedness for specific TEs. In several cases, accident responsibility can be attributed to a victim's hazardous behaviour: indoor victims were hit by flood while trying to save belongings or seeking shelter in garages, basement or underground levels rather than on roofs or upper floors. In this perspective, well-detailed datasets offer the opportunity to implement specifically designed measures and recommendations for self-protective actions, e.g., awareness campaign, risk education, road network management, risk signage, and impact-based warnings represent possible risk mitigation policies and initiatives regarding vehicle-related fatalities (37% in the investigated dataset). Further studies about extending the dataset and better detailing the fatal circumstances (e.g., location; victim activity, age, and health conditions; vehicle type, etc.) are underway. A non-negligible limitation of the approach to large-scale application, that is common to most of the existing tools and databases, lies in the availability of detailed and reliable data regarding the event evolution, cascading effects, and circumstances surrounding fatal incidents. As a consequence, results may suffer from epistemic uncertainty that refers to uncertainty associated with incomplete or imperfect knowledge about the processes involved. Nevertheless, a systematic process of collection, accurate verification, and organisation of data about weather-related adverse events can help to create a virtuous circle, minimise biases, and improve the quality of the outcomes.

This work was partially carried out within the cooperation agreement between Italian National Civil Protection and Camilab-DIMES Unical on ‘Criteri di allertamento, valutazione delle conseguenze e preannuncio di eventi idrogeologici estremi relativi al rischio da frana e da inondazione’.

Data cannot be made publicly available; readers should contact the corresponding author for details.

The authors declare there is no conflict.