Abstract

Over the past several decades, flash floods that occurred in Attica, Greece, caused serious property and infrastructure damages, disruptions in economic and social activities, and human fatalities. This paper investigated the link between rainfall and flash flood impact during the catastrophic event that affected Attica on 22 October 2015, while also addressing human risk perception and behavior as a response to flash floods. The methodology included the analysis of the space–time correlation of rainfall with the citizens’ calls to the emergency fire services for help, and the statistical analysis of people’s responses to an online behavioral survey. The results designated critical rainfall thresholds associated with flash flood impact in the four most affected subareas of the Attica region. The impact magnitude was found to be associated with the localized accumulated rainfall. Vulnerability factors, namely, population density, geographical, and environmental features, may have contributed to the differences in the impact magnitudes between the examined subareas. The analysis of the survey’s behavioral responses provided insights into peoples’ risk perception and coping responses relative to the space–time distribution of rainfall. The findings of this study were in agreement with the hypothesis that the more severe the rainfall, the higher peoples’ severity assessment and the intensity of emotional response. Deeper feelings of fear and worry were found to be related to more adjustments to the scheduled activities and travels. Additionally, being alert to the upcoming rainfall risk was found to be related to decreased worry and fear and to fewer changes in scheduled activities.

1. Introduction

Flash floods in urban centers threaten the safety of inhabitants and their properties. What distinguishes flash floods from slow-rise stream floods is the time scale, as the first occur within some hours. The sudden nature and the fast-moving water they produce, make flash floods very dangerous. Particularly urban development can increase the severity of a flash flooding event (Schroeder et al. 2016). Extended damage to buildings and infrastructure, disruptions of economic and social activities, and human fatalities occur because of flash flood incidents. The magnitude of the impact is a result of the weather hazard, namely the space–time distribution of precipitation, and of the vulnerability of the area affected. The accumulated rainfall and rainfall intensity are among the key factors examined in recent studies (Barbería et al. 2014; Diakakis 2012; Papagiannaki et al. 2015), in conjunction with urban-related factors, such as the population density, which affect the vulnerability of an area. Vulnerability is a multidimensional concept, briefly defined as the “conditions determined by physical, social, economic and environmental factors or processes which increase the susceptibility of a community to the impact of hazards” (p. 16, UNISDR 2004). Recent studies addressed the various aspects of vulnerability to flash floods, including the sensitivity of properties and materials exposed to the flooding risk (Diakakis et al. 2011; Merz et al. 2004), the specific environmental and climatic conditions in the affected area (Brody et al. 2015; Burgess et al. 2015), and the coping capacity of individuals and implicated authorities (Kellens et al. 2013; O’Neill et al. 2015; Poussin et al. 2014; Ruin et al. 2014; Terti et al. 2015). Particularly in urban areas, flood control measures have many times proven to be inefficient, as the ongoing urban expansion may increase flood impacts (Llasat et al. 2009; Nirupama and Simonovic 2007). Behavioral patterns, demographic, and other contextual factors, such as the timing and the day of the week the storm occurs, may also have a decisive effect on the magnitude and type of damage (Miceli et al. 2008; Ruin et al. 2014; Terti et al. 2015). Depending on the type of vulnerability (human, social, economic, systemic) and flood characteristics considered, as well as the scale of interest (global versus local, short-term versus long-term impacts), the vulnerability indicators to take into account greatly differ (Jonkman 2005; Molinari et al. 2014; Ruin et al. 2008; Terti et al. 2015).

Over the last few decades, the Mediterranean area was frequently affected by flash floods due to heavy rain associated with intense cyclones (Flaounas et al. 2016; Lagouvardos et al. 2007, 1996). The complex topography of the Mediterranean, which is characterized by steep mountains close to the sea, further favors flash floods (Jansa et al. 2014). Because flash floods are extreme and sudden events that are hardly predictable, many studies have focused on lessening the impact on humans by addressing behavioral coping responses. Of particular interest was the study of the relation between hydrometeorological variables and social responses to flash floods (Creutin et al. 2013, 2009; Lutoff et al. 2016; Ruin et al. 2009, 2014). These studies have contributed to the understanding of the range of behaviors and critical hydrometeorological parameters that might affect humans (Becker et al. 2015; Diakakis and Deligiannakis 2013; Llasat et al. 2008; Papagiannaki et al. 2015; Petrucci and Pasqua 2012; Ruin et al. 2008; Špitalar et al. 2014; Terti et al. 2017). Papagiannaki et al. (2015) showed critical rainfall thresholds to be associated with an increase in the number of emergency operations and the occurrence of damaging floods. Nevertheless, even within a region restricted to the Greek capital city, Athens, and its suburbs, these thresholds varied depending on the geographic area under scrutiny, showing, if need be, the influence of vulnerability factors related to the territory and social context. The exposure of elements was the most evident vulnerability factor critical for inducing human losses, property damage, or social inconvenience. Infrastructure, properties, or people and socioeconomic activities may suffer from runoff. Nevertheless, part of the exposure may have been mitigated by appropriate and timely reactions, starting with recognizing the potential danger of the upcoming event and avoiding dangerous situations, and then protecting movable goods and people. Based on postevent interviews of flood victims, some authors found links between the hydrometeorological characteristics (slower versus faster floods) and the capability of individuals, organizations, and institutions to adapt to the pace of the physical event (Creutin et al. 2013, 2009; Lutoff et al. 2016; Ruin et al. 2009, 2014). Of course, within the context of similar hydrometeorological characteristics, variability still existed among behavioral adaptations, showing that psychological and sociodemographic characteristics were also crucial in shaping people’s coping responses (Cutter et al. 2003; Drabek 1999; Lindell and Perry 2004; Mileti 1995; Ruin et al. 2014). According to this literature, it was obvious that events with comparable hydrometeorological characteristics might affect local communities differently, because of the difference in sensitivities of those communities.

To pursue the effort of identifying mediating factors between the occurrence of a hydrometeorological event and social impacts, this paper investigated critical rainfall thresholds with respect to different urban vulnerability contexts. This was achieved by comparing the effects of the 22 October 2015 rainfall episode on four subareas of Attica, Greece, which are characterized by different topography and population densities. A dense network of meteorological stations was used to represent the space–time characteristics of the rainfall episode in each subarea. The effects of the episode were described by two datasets: (i) the emergency impact measured by the number of citizens’ calls for help to the emergency line of the fire service; and (ii) the individual coping responses, measured through the analysis of an online survey aiming at collecting perceptual and behavioral responses of the witnesses of the rainfall episode.

The rest of the paper is structured as follows: section 2 presents in detail the methodological issues related to the data collection and analysis. Section 3 details the results from the analysis of the rainfall and impact distributions across the four subareas, and the statistical analysis of the behavioral survey. A discussion of the results is included in the same section. Section 4 includes a summary, concluding remarks, and future plans.

2. Methods

In this study, two types of methodological approaches were adopted to better understand the response of four urban subareas to the rainfall episode that affected Attica on 22 October 2015. The first one consisted of analyzing the link between the spatiotemporal distribution of the rainfall and the citizens’ calls to the fire service emergency for help during the event. For this, critical rainfall thresholds for triggering flash flood damages at the four subareas of Attica were compared. These thresholds were identified based on the effect the rainfall had on the number and the temporal distribution of emergency calls.

The second approach consisted of analyzing an online survey and examining the links between the rainfall levels and the way witnesses perceived the risk and adapted their behavioral response during the emergency. Statistical analysis was applied in order to evaluate the influence of rainfall on observed social effects (emergency calls and social responses), with respect to the different urban contexts of the four subareas. Nonparametric statistics were applied for testing the relationship between variables, because the normality assumption was violated in many cases. Therefore, the statistical dependence was measured with Spearman’s rank correlation method, which does not assume normality of data and is appropriate for correlating both continuous and discrete variables (McDonald 2014; Shipley 2016). In addition, the mean square contingency coefficient (phi coefficient) was used for testing the probability of independence between binary variables. The closest the Spearman’s rank correlation coefficient, ρ, and the phi coefficient, φ, are to +1 or −1, the stronger is the dependence between the variables. Statistical significance was considered strong if p values were lower than 0.05. However, in order to assess more accurately the dependence between the examined pairs of variables, three levels of significance were defined for p values below 0.05. Linear regression analysis was also applied for modeling the relationship between selected rain and impact variables. The Stata statistical software was used for data analysis and graphics.

a. Study area

The prefecture of Attica is a flood-prone area where intense storms and urban flooding events are relatively frequent. Flash flood events in the area have been studied both in terms of the meteorological point of view (Lagouvardos et al. 1996), and in terms of their occurrence and impacts (Papagiannaki et al. 2013, 2015). As it is the most densely populated area of Greece, human exposure is very high. Attica’s territory has many vulnerabilities including extended land cover in combination with inadequate drainage networks, unauthorized and nonexistent flood proof constructions, and absence of flood protection infrastructure in streams that have been found to be particularly prone to floods (Diakakis et al. 2011; Kitsos 2004; Lasda et al. 2010). Moreover, forest fires during the recent years have seriously affected the surrounding mountains to the detriment of land absorbency and the capacity of natural streams to capture precipitation (Amiridis et al. 2012).

The study area is the urban part of the Attica region and includes the Greek capital city, Athens, and its suburbs. The area is located in the central-east part of the Greek mainland and includes the four mountains that surround it: Aigaleo in the west, Parnitha in the northwest, Penteli in the northeast, and Hymettus in the east-southeast. Two rivers, Kifissos and Ilissos, flow through Athens’s suburban plain and have both been built up to a great extent (Pistrika et al. 2014). The 22 October 2015 flash flood event severely affected the areas around the Kifissos River, which is the main water stream of Athens. The river springs from Parnitha Mountain, covers an area of 381 km2 (Evelpidou et al. 2009), and follows a NNE–SSW flow direction. The population of Athens’s suburban area is approximately 3.3 million. Population density varies among municipalities from 600 inhabitants per square kilometer to 21 000 inhabitants per square kilometer.

Based on the number of emergency calls, the most seriously affected municipalities were identified and grouped into four subareas (NW, NE, W, and E; Fig. 1). Specifically, the flash flood impact magnitude differed between the northern and southern subareas. To define the four subareas, geographical and demographic criteria were considered, as both were expected to affect the local sensitivity and exposure to flood risks. Densely populated areas were expected to be at a high risk for flash floods, due to the highly built environment that increases runoff. Areas closer to mountains were considered to be exposed to faster and more dangerous runoff and debris flow that may leave limited time to the local community to prevent flood damage (Kim and Choi 2012).

Fig. 1.

Maps of Athens’s suburbs highlighting the four subareas under examination and the locations of the 43 surface meteorological stations, named by the municipalities in which they are located. [Maps generated with QGIS using SRTM terrain data (Datum GGRS87), edited using GIMP.]

Fig. 1.

Maps of Athens’s suburbs highlighting the four subareas under examination and the locations of the 43 surface meteorological stations, named by the municipalities in which they are located. [Maps generated with QGIS using SRTM terrain data (Datum GGRS87), edited using GIMP.]

Drawing on the above, the four subareas, shown in Fig. 1, were specified based on the following: the main river stream, Kifissos River, divides the affected areas into the west and east ones. The northern subareas are half surrounded by Attica’s two highest mountains that have both recently suffered devastating forest fires, which might increase the flash flood risk. Population density is significantly lower in the northern subareas. Specifically, the NW and NE subareas are among the less densely populated areas of Athens’s suburbs, with approximately 3000 and 2500 inhabitants per square kilometer, respectively. The E and W subareas host approximately 7500 and 10 500 inhabitants per square kilometer, respectively. Finally, to enhance methodological consistency, the subareas were defined to cover approximately the same area size, which was between 42 and 44 km2.

b. Hydrometeorological dataset

For the present study, data from 43 surface meteorological stations, spread in Athens’s suburban area, were used (Fig. 1). The National Observatory of Athens (NOA) operated 32 of the 43 stations, and the Laboratory of Hydrology and Water Resources Management of the National Technical University of Athens (NTUA) operated the other 11 stations. All meteorological stations provided 10-min observations of various meteorological parameters such as temperature, pressure, humidity, wind velocity and direction, rain amount, and rain intensity (Lagouvardos et al. 2017; Mimikou and Grammatikogiannis 2006).

According to the NOA historical dataset, the October 2015 event was among the most catastrophic weather events that affected Attica or Greece in the past 15 years, as it resulted in significant damages and four human fatalities. Historic flash flood events that affected Greece since 2000, as well as detailed meteorological and descriptive impact data, are included in the NOA database of high-impact weather events. The NOA weather forecasting group (Papagiannaki et al. 2013) systematically updates the database. A simplified version of the database, as well as a geographical presentation of the recorded weather events are available online (NOA 2016). Based on this database, Papagiannaki et al. (2015) examined the flash flood occurrence in relation to the rainfall hazard in Attica and identified flood-producing rainfall intensity thresholds for 15 urban subareas of the Attica prefecture. The estimated rainfall thresholds for some of the subareas were preliminary, as they depend on the density, the location, and the period of operation of the representative rain gauges. As the network of weather stations expanded and data time series increased, reliability of the rainfall thresholds also improved. In the present paper, the rainfall thresholds for triggering flash flood damages estimated by Papagiannaki et al. (2015) were updated to include the rainfall events that occurred in Attica in 2015.

c. Emergency calls dataset

The 22 October 2015 flash flood impact on Athens’s suburbs was measured by the citizens’ calls to the fire service emergency line. Data were derived from the Integrated Emergency Coordination Centre of the Hellenic Fire Service. Emergency calls from citizens in Attica were 1300 in total. Data included the exact time and location of the reported problem, as well as the type of the problem (water extraction, fallen tree, car accident, human or trapped animal). Recent studies have shown that the number of fire service operations, or the number of requests received by the meteorological services related to insurance claims, may be efficiently used as proxy measures of the material damage and as indicators of the social impact of flash floods (Amaro et al. 2010; Barbería et al. 2014; Papagiannaki et al. 2015).

In this paper, the hypothesis that the more severe the rainfall, the higher the number of emergency calls, was tested. The spatial distribution of citizens’ calls to the fire service, the majority of which (78%) were made from the specified subareas, is highlighted in Fig. 2, in blue. Specifically, 357 and 356 calls were made from the NW and NE subareas, respectively, and 165 and 142 calls were made from the W and E subareas, respectively. Data are presented in local time (LT), which corresponds to UTC + 3 h for October.

Fig. 2.

Spatial distribution of the citizens’ emergency calls to the fire service (blue dots) and of the questionnaire respondents’ location at the time they witnessed damages/problems caused by the flash flood (red dots). The four most affected subareas of Athens’s suburbs are marked and named by their geographical location with respect to each other (NW, NE, W, and E). [Maps generated with QGIS using SRTM terrain data (Datum GGRS87), edited using GIMP.]

Fig. 2.

Spatial distribution of the citizens’ emergency calls to the fire service (blue dots) and of the questionnaire respondents’ location at the time they witnessed damages/problems caused by the flash flood (red dots). The four most affected subareas of Athens’s suburbs are marked and named by their geographical location with respect to each other (NW, NE, W, and E). [Maps generated with QGIS using SRTM terrain data (Datum GGRS87), edited using GIMP.]

d. Perceptual and behavioral response dataset

An online behavioral survey dedicated to collecting testimonies about the 22 October 2015 event in Attica was launched to study human factors of perception and behavior as a response to severe rainfall and flash flood occurrence. This questionnaire survey was initially developed in 2014, based on the experience of postflood interviews that were conducted over the past 10 years in southern France by an interdisciplinary team of social scientists and hydrometeorologists (Ruin et al. 2009, 2014). The intent of the questionnaire was to expand the qualitative results of the interviews, by collecting quantitative testimonies of people who were direct witnesses of intense runoff events. The goal was to better comprehend the link between the space–time distribution of the rainfall amount and intensity, and its impact on people’s daily activity and life. For that purpose, the questionnaire was structured around four themes: 1) the perception of the rainfall and runoff severity and damage experienced, 2) the alert status and personalization of danger through emotions, 3) the potential travel and activity adaptations conducted the day of the event, and 4) the sociodemographic factors, known to be important factors with respect to risk perception and behavioral response.

In this paper, perceptual and behavioral responses across the urban area of Attica were analyzed in order to test the influence of the rainfall severity and urban characteristics on the level of adaptation, namely the effects the event had on people’s travel patterns and daily activities (two distinct multiple-choice questions). Based on previous research findings, these multiple answers were weighted to reflect the gradation of the adjustments made in terms of travels and activities, and were treated as ordinal variables (Creutin et al. 2013; Ruin et al. 2014). Moreover, the hypothesis that the more severe is the rainfall, the higher the individual severity assessment and the intensity of emotion will be, were tested. Literature in psychology associates the emotional component of risk perception (i.e., the feeling of worry and fear), with decision-making and the adoption of behavioral adaptations by the individual (Loewenstein et al. 2001; Miceli et al. 2008). In this regard, the feelings of fear and worry were expected to be related to travel and activity adjustment, as a response to severe rainfall and flash flood occurrence.

Translated and adapted to the Greek context, the online survey was launched one month after the occurrence of the 22 October event. The invitation to fill in the questionnaire was advertised at the website www.meteo.gr, which provides weather, wave, lightning, and dust forecasts produced by the weather forecasting group at the Institute for Environmental Research, National Observatory of Athens (IERSD/NOA). The questionnaire received over 800 valid responses in 5 days, which indicated the high concern for such events and their future management. The analysis focused on 706 responses that concerned Athens’s suburban area. The spatial distribution of the respondents’ location at the time they witnessed flood-related damages and problems is depicted with red dots in Fig. 2. Out of the 706 responses, 388 (56%) where found almost equally distributed among the four examined subareas: 100 in the NW, 85 in the NE, 102 in the W, and 101 in the E. Thus, the four subareas are relatively equally addressed in the survey.

With respect to the representativeness of various age ranges in the survey, the share of respondents under the age of 25 was 10%, respondents between 25 and 40 years old accounted for 43%, and respondents between 40 and 60 years old accounted for 44% of the population sample. The youngest person was 14 years old. For the age groups between 25 and 60, the distribution of the sample population by age was representative of Attica’s population, whereas ages below 25 and over 60 were underrepresented. Only 3% were over 60 years old, while the percentage of people of that age group in Attica is 17%. This might reflect the low use of the internet by older people (Xiao et al. 2015). Over 69% of the participants were males, which were, thus, overrepresented in the sample. In what concerns the employment rates, 52% of the respondents declared themselves employees/workers and 22% were “without professional activity.” These rates were representative of Attica’s profile of employment (Hellenic Statistical Authority 2015). Specifically, according to the official labor statistics, the percentages of employed and unemployed in Attica were 54% and 22%, respectively, during the fourth quarter of 2015, when the flash flood event occurred.

3. Results and discussion

On 22 October 2015, Athens’s suburban area was affected by a severe storm system that was related to the passage of a cold front associated with a low pressure system. The Hellenic National Meteorological Service (HNMS) had issued, 2 days before the event, a warning of intense rainfall and windstorms that would affect the entire country. The last rainfall event recorded in Athens’s suburban area was on 11 October, thus the ground was dry for more than 10 days. The spatial distribution of the 24-h accumulated rain is depicted in Fig. 3, which highlights the four most affected urban subareas and the exact location of human fatalities. The rainstorm produced extended flash flood incidents, particularly in the north part of Athens’s suburban area, where the four flood-related human fatalities occurred. The victims were swept away in torrents caused by flooded streams. Three of them were driving a car, while the fourth was a pedestrian. The fire service received over 1300 emergency calls in Athens’s suburban area, 90% of which concerned flooded properties and 10% other flood-related events (e.g., trapped drivers and damages from fallen trees).

Fig. 3.

Spatial distribution of 24-h accumulated rain on 22 Oct 2015. Yellow spots show the exact location of human fatalities. The four most affected subareas of Athens’s suburbs are highlighted.

Fig. 3.

Spatial distribution of 24-h accumulated rain on 22 Oct 2015. Yellow spots show the exact location of human fatalities. The four most affected subareas of Athens’s suburbs are highlighted.

The NE and NW were the most affected subareas, each one accounting for 27% of total calls. The maximum daily rainfall observations were 101.6 and 130.6 mm in the NE and NW, respectively. The W subarea, where the maximum daily rainfall was 105 mm, accounted for 13% of the calls, and the E subarea, with maximum daily rainfall of 69.4 mm, accounted for 11% of total calls. The daily maximum rainfall observed in the rest of Athens’s suburban area was below 70 mm, while the number of calls to the fire service was rather low, compared to the four most affected subareas.

a. Rainfall versus impact magnitude

To assess the relationship between the rainfall severity and flash flood occurrence, precipitation data, of all the rainfall events that occurred in the period 2005–15 in each subarea, were investigated and linked to the occurrence of flood-related damages. Papagiannaki et al. (2015) first performed this type of analysis for the period 2005–14, while in the present work the analysis also included the most recent events, covering, thus, the period 2005–15. Figure 4 depicts the graphs of peak rainfall intensities for various durations, for each of the examined subareas, except for the W where meteorological stations have only recently been deployed. Flash flood occurrence is highlighted in red. The lines divide the graph into three parts: the lower part includes peak intensity values that did not lead to flooding throughout the analyzed period, the middle part includes peak intensities that either did or did not lead to flooding, and the upper part includes peak intensities that always led to flooding. These graphs were used in the present analysis for assessing the probability of flood-related damages occurrence, with respect to the observed peak rain intensities.

Fig. 4.

Peak rainfall intensities vs duration, for the NW, NE, and E subareas of Athens’s suburbs. Peak intensities for each subarea are derived from the rainfall data recorded by the representative rain gauges shown in Fig. 1. Blue dots correspond to the nondamaging rainfall events, while red dots highlight events with flash flood–related damages. The logarithmic x-axis scale is used. The maximum accumulated rain in 10, 30, and 60 min, and 2, 3, 12, and 24 h of each rainfall event is calculated as the maximum moving sum, through time steps of 10, 30, and 60 min, and 2, 3, 12, and 24 h, respectively.

Fig. 4.

Peak rainfall intensities vs duration, for the NW, NE, and E subareas of Athens’s suburbs. Peak intensities for each subarea are derived from the rainfall data recorded by the representative rain gauges shown in Fig. 1. Blue dots correspond to the nondamaging rainfall events, while red dots highlight events with flash flood–related damages. The logarithmic x-axis scale is used. The maximum accumulated rain in 10, 30, and 60 min, and 2, 3, 12, and 24 h of each rainfall event is calculated as the maximum moving sum, through time steps of 10, 30, and 60 min, and 2, 3, 12, and 24 h, respectively.

The 10-min rainfall records and number of calls during the day of the event are shown in Fig. 5. Figure 6 shows the respective values accumulated at hourly level. In three out of the four subareas, there was more than one meteorological station and, thus, the station that recorded the maximum daily rainfall within the respective subarea was used for the production of the respective chart and the subsequent analysis. Two phases of the storm event have been identified. The first phase started at 0200 local time (LT), lasted more than 3 h, and produced a medium-intensity rainfall and an insignificant number of calls to the fire service, only from the NW subarea. During this phase, the maximum accumulated rain observed was 40 mm in the NW, 27 mm in the W, 26 mm in the E, and 16 mm in the NE. The 1-h (10 min) maximum rain intensity observed in the NW, W, E, and NE was 19.4 (43.2), 18.2 (27.6), 18.6 (45.6), and 5.4 (16.8) mm h−1, respectively. As can be inferred from Fig. 4, the observed peak rain intensities in the examined subareas were considered to be associated with low probability of damaging flooding in the area. It should be noted that the absence of emergency calls during the night could be partly associated with the delayed discovery of flood-related damages by citizens that were probably asleep during the event. It is also likely that a number of damages were avoided due to the nighttime occurrence of the event. However, the rather low accumulated rainfalls, as well as the low rain rates observed in this phase, may have been the reasons behind the absence of flash flood damages.

Fig. 5.

The 10-min rainfall and number of emergency calls time series (local time, UTC + 3 h) that occurred in the 4 most affected subareas of Athens’s suburbs (a) NW, (b) NE, (c) W, and (d) E). Rainfall data for each subarea derived from the meteorological station that recorded the highest daily accumulated rainfall within the specific subarea.

Fig. 5.

The 10-min rainfall and number of emergency calls time series (local time, UTC + 3 h) that occurred in the 4 most affected subareas of Athens’s suburbs (a) NW, (b) NE, (c) W, and (d) E). Rainfall data for each subarea derived from the meteorological station that recorded the highest daily accumulated rainfall within the specific subarea.

Fig. 6.

Accumulated rainfall and number of emergency calls at hourly level (local time, UTC + 3 h) that occurred in the four most affected subareas of Athens’s suburbs (NW, NE, W, and E).

Fig. 6.

Accumulated rainfall and number of emergency calls at hourly level (local time, UTC + 3 h) that occurred in the four most affected subareas of Athens’s suburbs (NW, NE, W, and E).

The second phase of the 22 October event occurred in the morning until early afternoon and produced remarkable damages. Citizens’ calls for help first occurred at around 1100 LT. The respective accumulated rainfall across the four subareas, at the time, ranged from 37 mm in the NE to 61 mm in the NW. According to the findings of Papagiannaki et al. (2015), the rain events that occurred in Athens’s suburban area and were associated with maximum daily rainfall of 30–60 mm, were related to a low number of fire service operations, while the damage magnitude was found to be significant for rain amounts exceeding 60 mm. The number of calls, in each subarea, presented an impressive gradual increase when the rain exceeded these thresholds.

The maximum 1-h (10 min) rain intensity observed since the onset of rainfall was 31.8 (68.4) mm h−1 in the NW, 27.2 (81.6) mm h−1 in the W, 48.2 (111.6) mm h−1 in the NE, and 10.8 (52.8) mm h−1 in the E. Peak rainfall intensities occurred at around 1300 LT. At that moment, the accumulated rainfall was over 100 mm in the NW and W, and 70 mm in the NE and E. In the NE subarea, the one with the highest number of calls at that time, the observed 1-h and 10-min maximum rainfall intensities were found to associate with very high probability of damaging flash floods, in accordance with the thresholds shown in Fig. 4.

The comparison of the 1-h accumulated rain and related calls among the subareas, however, shows that flood-related damages in the NE occurred for lower rainfall amounts compared to the other subareas (Fig. 6). A total of 130 calls were made before the daily rainfall reached 50 mm and for low rainfall intensities. By that time, the 10-min maximum rain intensity observed was less than 34 mm h−1. Interestingly, extensive flood-control infrastructure was recently implemented in the NE area, in the frame of an 18-month project that was completed shortly before the occurrence of the 22 October rainfall event. These facts indicated that the NE subarea may be more vulnerable to flash floods, compared to the other examined subareas. Differences between the subareas’ responses to the rainfall, however, were not only a function of the rain severity. Population density may have also affected vulnerability to flash floods, as in general more densely populated areas are at a higher risk for flash flood–related damage on properties. The rainfall event, however, caused more adverse impacts on the northern subareas (NW and NE), even though they have lower population densities compared to the W and E subareas. According to the impact indices presented in Fig. 7, the northern subareas still ranked first when the impact adjusted to the population density. Geographical and environmental features of the examined areas may have, thus, escalated the differences in the impact magnitude. Both subareas in the NE and NW, which was also highly affected, are closer to the highest mountains of Attica, namely Parnitha and Penteli, which suffered destructive forest fires in the summers of years 2007 and 2009. Water flows follow a downstream direction from sources located in these mountains toward the NW and NE subareas.

Fig. 7.

Comparison of impact indices (total number of emergency calls and average number of calls per population density) between the four examined subarea of Athens’s uburbs (NW, NE, W, and E).

Fig. 7.

Comparison of impact indices (total number of emergency calls and average number of calls per population density) between the four examined subarea of Athens’s uburbs (NW, NE, W, and E).

To examine whether other vulnerability factors, such as topography, population density, or time of the rainfall, were involved in the magnitude of the induced impact, rain variables and the number of emergency calls were statistically analyzed for each subarea. Accounting only for the storm phase that produced damages, significant and strong correlations were found between the accumulated rain and 10-min calls, based on the Spearman’s rank correlation results [NE: ρ (N = 39) = 0.54, p < 0.001; NW: ρ (N = 31) = 0.65, p < 0.001; W: ρ (N = 26) = 0.51, p < 0.01; E: ρ (N = 29) = 0.47, p < 0.01]. Linear regression was applied in the same time series of accumulated rain (independent variable) and 10-min calls records (dependent variable) and showed high scores of the determination coefficient R2 for the western subareas (NW: 0.58, W: 0.54, NE: 0.24, E: 0.12). Statistical significance was very high (p < 0.001) in all cases, and lower for the E subarea. Results suggest that the accumulated rain had a significant effect on the number of the emergency calls in all the examined subareas. However, the calculated R2 coefficients suggest that the rain cannot explain a large part of the impact induced, particularly in the NE and E. Therefore, apart from the rainfall hazard, the abovementioned vulnerability factors may also have a significant effect on the magnitude of the flash flood impact on the examined subareas.

The emergency calls to the fire service continued accumulating for hours after 1400 LT when the storm ended, probably due to contextual factors related to the timing of the event. The second and more severe phase peaked during rush hour in a usual weekday, when the city activity was at its highest, schools were open, and most people were away from their homes (Duquenne and Kaklamani 2009). Thus, a number of citizens might have discovered flood-related damage on their properties later, when they returned at home. However, this is only an assumption that could not be assessed in the present analysis.

b. Behavioral responses to the storm event

The results of the statistical analysis of the survey data, in conjunction with rain and impact variables, are presented in Tables 1 and 2, which include 1) Spearman’s rank correlation coefficient, ρ, or phi coefficient, φ, for the pairs of variables studied in the context of the present analysis, 2) the respective p values in order to assess the significance of the correlation results, and 3) information regarding the rating scale for each question and the weighting factors applied to the different modalities of the multiple answers. Table 1 includes all the survey variables and a representative meteorological parameter, namely, the accumulated rain at the respondent’s location. Table 2 includes the behavioral variables and their relationship with an ordinal variable that represents the local impact magnitude.

Table 1.

Significance and strength of correlation (Spearman’s rank correlation coefficient, ρ)a between the behavioral survey variables (demographics, perception, coping response) and meteorological parameter (accumulated rain).

Significance and strength of correlation (Spearman’s rank correlation coefficient, ρ)a between the behavioral survey variables (demographics, perception, coping response) and meteorological parameter (accumulated rain).
Significance and strength of correlation (Spearman’s rank correlation coefficient, ρ)a between the behavioral survey variables (demographics, perception, coping response) and meteorological parameter (accumulated rain).
Table 2.

Significance and strength of correlation (Spearman’s rank correlation coefficient, ρ) between the behavioral survey variables (perception, coping response) and the local impact magnitude.

Significance and strength of correlation (Spearman’s rank correlation coefficient, ρ) between the behavioral survey variables (perception, coping response) and the local impact magnitude.
Significance and strength of correlation (Spearman’s rank correlation coefficient, ρ) between the behavioral survey variables (perception, coping response) and the local impact magnitude.

1) Perception of the severity of the hydrometeorological event

The participants were asked to make their own assessment of the rain and flood severity, at the time the storm peaked according to their judgment. To assess the rain severity perception at a local level, the responses (%) at each level of rain severity perception were compared between the four most affected subareas and the rest of Athens’s suburbs (Fig. 8). According to their assessment, people from the more affected subareas (NW, NE) experienced more severe rainfall. Their answers were then correlated with the corresponding accumulated rain observations at the respondents’ geographical position in order to capture how the individual perception related to actual rain measurements and what factors might explain possible deviations. People’s judgment on the event’s severity, at the time the storm peaked at a certain location, was found to be significantly correlated with the total accumulated rain (since the storm onset) at the same location (Table 1). Specifically, as shown in the box plot of Fig. 9a, rain severity perception was positively correlated with rainfall.

Fig. 8.

Responses (%) to the online behavioral survey concerning the rain severity perception in each of the four most affected subareas (E, W, NE, and NW) and the rest of Athens’s suburbs.

Fig. 8.

Responses (%) to the online behavioral survey concerning the rain severity perception in each of the four most affected subareas (E, W, NE, and NW) and the rest of Athens’s suburbs.

Fig. 9.

(a) Level of rain severity perception and (b) level of worry, at the time the storm peaked according to the respondents’ judgment, in relation to the accumulated rain at the respondents’ position.

Fig. 9.

(a) Level of rain severity perception and (b) level of worry, at the time the storm peaked according to the respondents’ judgment, in relation to the accumulated rain at the respondents’ position.

The participants were also asked three concrete questions regarding their emotional state when the rainfall intensity peaked. Specifically, they were asked (i) whether or not they felt they were in danger, (ii) to rate their feelings of worry, and (iii) to rate their feelings of fear at the threat of suffering adverse impact. All of these variables were positively correlated with the accumulated rainfall at the respondents position (Table 1). Compared to the rain severity perception levels, the respective levels of worry and fear were shown to correspond to somewhat higher values of accumulated rain (Figs. 9a,b). More specifically, according to the medians of the box plots shown in Fig. 9, the degree of the perceived severity corresponded to lower amounts of rainfall than the degree of worry, suggesting that individuals’ severity assessment may have influenced the feelings of worry. For instance, the responses at the medium level of severity perception (level 3) were centered on 55 mm of rain, while the responses at the medium level of worry (level 3) were centered on 60 mm of rain (Fig. 9b).

The level of worry was found to significantly depend on the location of the respondent during the storm peak. The more affected the subarea was, the more worried the respondents were (Table 2). Accordingly, the perception of being in danger during the event was found to increase with increasing magnitude of local impact. Females were found more concerned than males, and their feeling of worry maximized for somewhat lower rain amounts. Increase in the age of the respondents was also found to relate to deeper feelings of worry during the storm.

2) Risk perception versus coping response

The effect of emotional and cognitive components of risk perception on the adoption of protective behaviors was also examined. To capture risk awareness, people were asked to answer the dichotomous question about whether they were previously alerted, as well as to mention the source of the warning. Interestingly, only 44% of males and 33% of females were informed about the upcoming rainfall risk, 74% of which from the weather forecast news, and 21% from the social networks. Independent of the gender, 60% of the respondents were found to be unaware of the rainfall risk.

Participants were also asked two questions about whether their planned activities and travels were disturbed because of the event. The intention was to capture the level of adaptation of individuals to the particular hazardous conditions. The respondents could choose one, or more, among five answers that concerned adjustments to their scheduled activities, and among three answers concerning their travel adjustments. The answers were weighted by rating the significance of the adjustments made on the schedule, as quoted in detail in Table 1. The hypothesis tested was whether being alerted increased preparedness against flooding. In this case, preparedness was considered to be associated with fewer last-minute changes in the schedule. A negative correlation was found between the alert status and the changes in the scheduled activities during the event. Specifically, being alerted was found to be associated with a less drastic adjustment in daily activities (Fig. 10). On the other hand, increased feelings of worry, and mostly of fear, were strongly related to a higher need for adaptation, and specifically to more important changes in the planned activities and travels (Fig. 11).

Fig. 10.

Number of responses to the online behavioral survey concerning the type of adaptive activities for each status of alert.

Fig. 10.

Number of responses to the online behavioral survey concerning the type of adaptive activities for each status of alert.

Fig. 11.

Number of responses to the online behavioral survey concerning the type of adaptive travels at various levels of worry.

Fig. 11.

Number of responses to the online behavioral survey concerning the type of adaptive travels at various levels of worry.

The effect of the dichotomous alert status on the feelings of worry and fear was found to be highly significant, according to the p values given in Table 1. Being alerted was found to be related to decreased worry and fear of the flash flood threat. Subsequently, the correlation results indicated that people were more prepared and less worried when they were alerted about the rainfall risk. However, results may simply reflect the need of the respondent to show consistency between the two statements. Namely, this could be an example of cognitive dissonance, where alerted people try to present themselves emotionally and timely prepared, as they should be. Based on the statistical results given in Table 1, individuals’ coping responses were more dependent on the emotions (worry, fear) than on the cognitive dimension of risk perception (being aware of the weather severity, and being timely alerted). However, further investigation of the interrelation of all these variables is required. The links between the alert status, risk perception, and risk preparedness have been addressed in recent studies on individual behavioral response to weather-related risks. According to literature, the emotional component of risk perception was associated with decision-making and the adoption of protective measures by the individual (Loewenstein et al. 2001; Miceli et al. 2008), while the analysis of the cognitive component might be a valuable tool for the understanding of the society’s weaknesses and the effective risk communication (Jonkman and Kelman 2005; Kellens et al. 2011).

Risk awareness did not vary significantly among the examined subareas. The percentage of the respondents that declared themselves alerted ranged from 30% in the NW to 38% in the NE and E subareas. On the contrary, the amount and significance of changes in the day’s activities were found to depend on the location of the respondent. The association between the localized impact magnitude and adaptation indicated that people reacted to the rainfall risk in accordance with the local severity of the flash flood impact (Table 2). The correlation between the impact severity and adaptation, accounting for the different alert status, had a high level of significance. The more damaged the subarea was, the more adjustments people made to their schedule, whether they were alerted or not. Thus, results showed evidence of the dependence of the human coping response on the severity of the conditions they faced.

The need for adapting planned travels to the particular conditions was found higher for more serious travel reasons, which in turn were significantly related to the family status (Table 1). Specifically, cohabitation and children increased the need for traveling, as expected. The gender and professional status were both found to have a positive effect on adaptation, which was found increased for females and people with higher professional status. At the subarea scale, however, none of the above sociodemographic variables (family status, age, gender, travel reason) was significantly correlated with the impact magnitude. Family status, as expected, played an important role to the degree of worry felt by citizens. Living with other people and, most of all, having children increased the feelings of worry and fear, and were associated with considering the event as a serious threat.

4. Summary and conclusions

In the frame of this study, the relationship between severe rainfall and flash flood–related impact on property and human coping response in a highly urbanized environment was investigated. The case study was a catastrophic flash flood event that occurred in Attica (Greece) on 22 October 2015. The data analyzed were (i) rainfall observations from a dense rain gauge network operating in the area, (ii) calls to the fire service emergency line, and (iii) responses to an online behavioral survey.

Four subareas of the urban Attica region were studied in order to assess the effect of several variables on the flash flood impact magnitude. The flash flood impact was measured by the frequency of emergency calls, which started to accumulate for rain between 40 and 60 mm, depending on the subarea. The recorded maximum rain intensities were found to relate with very high probability of damage occurrence in the examined subareas. Additionally, statistical analysis of the 10-min records of rain and calls showed a significant effect of the accumulated rain on the frequency of emergency calls. However, evidence that a large part of the impact could not be explained by the rainfall parameters was found, particularly in what concerns the NE subarea. Compared to the other subareas (NW, W, and E), the NE was more affected for lower amounts of accumulated rain and for low rain rates. To explain differences in the impact magnitude between the subareas, localized vulnerabilities, namely, the level of urbanization and geographical features, were involved in the analysis. Population density was not found to significantly affect the magnitude of impact. The fact that the more affected subareas were geographically located near the highest mountains, which suffered extended forest fires in the last decade, was suggested as a possible reason of the higher local vulnerability to flash floods.

The analysis of the behavioral survey provided insight into the factors that affected the citizens’ coping responses to the severe rainfall event. The online questionnaire attracted over 800 responses and captured the cognitive and emotional components of risk perception with respect to the individual’s adaptability during flash flood crisis. Sociodemographic factors were also found to affect risk perception and behavioral response.

The emotional state was found to depend on the rainfall severity, while results suggested that emotions were probably influenced by the rain severity perception. During the emergency crisis, coping response was found related to both the emotional and cognitive component of risk perception. Deeper feelings of worry were found to be correlated with more significant adjustment on scheduled travels and activities, while risk awareness (being timely alerted) was found to prevent last minute changes. Human response, specifically the adaptation to the particular conditions by adjusting the schedule, was found more radical in the more affected subareas. This indicates that people in these subareas were more exposed to flash floods, and that they reacted in accordance to the severity of the localized impact.

Even though the statistical results were significant, the role of the cognitive risk perception in the individual preparedness still needs further investigation. The questionnaire addressed the adaptation made by citizens to the scheduled activities during the flash flood event, without however clarifying the way risk awareness affected these decisions. It is in the authors’ plans to investigate in more detail the social factor of regional and local response to the flash flood risk, by systematically launching surveys addressing people from different geographical areas of Greece. The survey applied in the present study focused on a specific flash flood event, but considered neither people’s prior experience on similar events nor other issues influencing risk perception and precautionary behavior, such as the understandability of the warning messages, the role of the local authorities and social networks, and the willingness/capacity to pay for the adaptive/mitigation measures. Thus, future questionnaires will be refined with the aim to shed light on the role of the aforementioned factors on flood risk preparedness as a function of risk perception prior to the flood occurrence.

Acknowledgments

The authors acknowledge the Integrated Emergency Coordination Centre of the Hellenic Fire Service for the provision of the emergency calls dataset for the 22 October 2015, as well as the Laboratory of Hydrology and Water Resources Management of the National Technical University of Athens for the supply of precipitation measurements from 11 stations in the study area. This work is a contribution to the Hydrological cycle in the Mediterranean Experiment (HYMEX) program. The authors would also like to thank the Centre National de la Recherche Scientifique (CNRS) researchers at the Laboratoire d’étude des Transferts en Hydrologie et Environnement (LTHE) and Politiques publiques, Action politique, Territoires (PACTE) Institutes in Grenoble for their support and assistance regarding the online behavioral survey.

REFERENCES

REFERENCES
Amaro
,
J.
,
M.
Gaya
,
M.
Aran
, and
M. C.
Llasat
,
2010
:
Preliminary results of the Social Impact Research Group of MEDEX: The request database (2000–2002) of two Meteorological Services
.
Nat. Hazards Earth Syst. Sci.
,
10
,
2643
2652
, doi:.
Amiridis
,
V.
, and Coauthors
,
2012
:
Impact of the 2009 Attica wild fires on the air quality in urban Athens
.
Atmos. Environ.
,
46
,
536
544
, doi:.
Barbería
,
L.
,
J.
Amaro
,
M.
Aran
, and
M. C.
Llasat
,
2014
:
The role of different factors related to social impact of heavy rain events: Considerations about the intensity thresholds in densely populated areas
.
Nat. Hazards Earth Syst. Sci.
,
14
,
1843
1852
, doi:.
Becker
,
J. S.
,
H. L.
Taylor
,
B. J.
Doody
,
K. C.
Wright
,
E.
Gruntfest
, and
D.
Webber
,
2015
:
A review of people’s behavior in and around floodwater
.
Wea. Climate Soc.
,
7
,
321
332
, doi:.
Brody
,
S. D.
,
W. E.
Highfield
, and
R.
Blessing
,
2015
:
An analysis of the effects of land use and land cover on flood losses along the Gulf of Mexico coast from 1999 to 2009
.
J. Amer. Water Resour. Assoc.
,
51
,
1556
1567
, doi:.
Burgess
,
C.
,
M.
Taylor
,
T.
Stephenson
,
A.
Mandal
, and
L.
Powell
,
2015
:
A macro-scale flood risk model for Jamaica with impact of climate variability
.
Nat. Hazards
,
78
,
231
256
, doi:.
Creutin
,
J. D.
,
M.
Borga
,
C.
Lutoff
,
A.
Scolobig
,
I.
Ruin
, and
L.
Créton-Cazanave
,
2009
:
Catchment dynamics and social response during flash floods: The potential of radar rainfall monitoring for warning procedures
.
Meteor. Appl.
,
16
,
115
125
, doi:.
Creutin
,
J. D.
,
M.
Borga
,
E.
Gruntfest
,
C.
Lutoff
,
D.
Zoccatelli
, and
I.
Ruin
,
2013
:
A space and time framework for analyzing human anticipation of flash floods
.
J. Hydrol.
,
482
,
14
24
, doi:.
Cutter
,
S. L.
,
B. J.
Boruff
, and
W. L.
Shirley
,
2003
:
Social vulnerability to environmental hazards
.
Soc. Sci. Quart.
,
84
,
242
261
, doi:.
Diakakis
,
M.
,
2012
:
Rainfall thresholds for flood triggering: The case of Marathonas in Greece
.
Nat. Hazards
,
60
,
789
800
, doi:.
Diakakis
,
M.
, and
G.
Deligiannakis
,
2013
:
Vehicle-related flood fatalities in Greece
.
Environ. Hazards
,
12
,
278
290
, doi:.
Diakakis
,
M.
,
M.
Foumelis
,
L.
Gouliotis
, and
E.
Lekkas
,
2011
: Preliminary flood hazard and risk assessment in Western Athens metropolitan area. Advances in the Research of Aquatic Environment, N. Lambrakis, G. Stournaras, and K. Katsanou, Eds., Springer, 147–154, doi:.
Drabek
,
T. E.
,
1999
:
Understanding disaster warning responses
.
Soc. Sci. J.
,
36
,
515
523
, doi:.
Duquenne
,
M. N.
, and
S.
Kaklamani
,
2009
: Daily mobility of the employed in Greece: A preliminary survey on the extent of the phenomenon (in Greek). Laboratory of Demographic and Social Analyses, University of Thessaly Tech. Rep., 4 pp. [Available online at http://www.demography-lab.prd.uth.gr/DEMO-NEWS-IOYLIOS-AUGOUSTOS%20No6%20a4.pdf.]
Evelpidou
,
N.
,
N.
Mamassis
,
A.
Vassilopoulos
,
C.
Makropoulos
, and
D.
Koutsogiannis
,
2009
: Flooding in Athens: The Kephisos River flood event of 21–22/10/1994. Conf. on Urban Flood Management (COST22), Paris, France, UNESCO, doi:.
Flaounas
,
E.
,
K.
Lagouvardos
,
V.
Kotroni
,
C.
Claud
,
J.
Delanoë
,
C.
Flamant
,
E.
Madonna
, and
H.
Wernli
,
2016
:
Processes leading to heavy precipitation associated with two Mediterranean cyclones observed during the HyMeX SOP1
.
Quart. J. Roy. Meteor. Soc.
,
142
,
275
286
, doi:.
Hellenic Statistical Authority
,
2015
: Labour force (quarterly), 4th quarter. Accessed 7 December 2016. [Available online at http://www.statistics.gr/en/statistics/-/publication/SJO01/-.]
Jansa
,
A.
, and Coauthors
,
2014
:
MEDEX: A general overview
.
Nat. Hazards Earth Syst. Sci.
,
14
,
1965
1984
, doi:.
Jonkman
,
S. N.
,
2005
:
Global perspectives on loss of human life caused by floods
.
Nat. Hazards
,
34
,
151
175
, doi:.
Jonkman
,
S. N.
, and
I.
Kelman
,
2005
:
An analysis of the causes and circumstances of flood disaster deaths
.
Disasters
,
29
,
75
97
, doi:.
Kellens
,
W.
,
R.
Zaalberg
,
T.
Neutens
,
W.
Vanneuville
, and
P.
De Maeyer
,
2011
:
An analysis of the public perception of flood risk on the Belgian Coast
.
Risk Anal.
,
31
,
1055
1068
, doi:.
Kellens
,
W.
,
T.
Terpstra
, and
P.
De Maeyer
,
2013
:
Perception and communication of flood risks: A systematic review of empirical research
.
Risk Anal.
,
33
,
24
49
, doi:.
Kim
,
E. S.
, and
H. I.
Choi
,
2012
:
Estimation of the relative severity of floods in small ungauged catchments for preliminary observations on flash flood preparedness: A case study in Korea
.
Int. J. Environ. Res. Public Health
,
9
,
1507
1522
, doi:.
Kitsos
,
G.
,
2004
: Implemented and under implementation flood-prevention works in Attica the last five-years and repercussions from the new transport and other works (in Greek). Department of Professional Development Rep., Office of Works Production, Athens, Greece, 9 pp. [Available online at http://library.tee.gr/digital/m2022/m2022_kitsos.pdf.]
Lagouvardos
,
K.
,
V.
Kotroni
,
S.
Dobricic
,
S.
Nickovic
, and
G.
Kallos
,
1996
:
The storm of October 21–22, 1994, over Greece: Observations and model results
.
J. Geophys. Res.
,
101
,
26 217
26 226
, doi:.
Lagouvardos
,
K.
,
V.
Kotroni
, and
E.
Defer
,
2007
:
The 21–22 January 2004 explosive cyclogenesis over the Aegean Sea: Observations and model analysis
.
Quart. J. Roy. Meteor. Soc.
,
133
,
1519
1531
, doi:.
Lagouvardos
,
K.
, and Coauthors
,
2017
:
The automatic weather stations NOANN network of the National Observatory of Athens: Operation and database
.
Geosci. Data J.
, doi:, in press.
Lasda
,
O.
,
A.
Dikou
, and
E.
Papapanagiotou
,
2010
:
Flash flooding in Attika, Greece: Climatic change or urbanization?
Ambio
,
39
,
608
611
, doi:.
Lindell
,
M. K.
, and
R. W.
Perry
,
2004
: Communicating Environmental Risk in Multiethnic Communities. Communicating Effectively in Multicultural Contexts Series, Vol. 7, Sage Publications, 272 pp.
Llasat
,
M. C.
,
L.
López
,
M.
Barnolas
, and
M.
Llasat-Botija
,
2008
:
Flash-floods in Catalonia: The social perception in a context of changing vulnerability
.
Adv. Geosci.
,
17
,
63
70
, doi:.
Llasat
,
M. C.
,
M.
Llasat-Botija
,
M.
Barnolas
,
L.
López
, and
V.
Altava-Ortiz
,
2009
:
An analysis of the evolution of hydrometeorological extremes in newspapers: The case of Catalonia, 1982–2006
.
Nat. Hazards Earth Syst. Sci.
,
9
,
1201
1212
, doi:.
Loewenstein
,
G. F.
,
E. U.
Weber
,
C. K.
Hsee
, and
N.
Welch
,
2001
:
Risk as feelings
.
Psychol. Bull.
,
127
,
267
, doi:.
Lutoff
,
C.
,
J.-D.
Creutin
,
I.
Ruin
, and
M.
Borga
,
2016
:
Anticipating flash-floods: Multi-scale aspects of the social response
.
J. Hydrol.
,
541
,
626
635
, doi:.
McDonald
,
J. H.
,
2014
: Handbook of Biological Statistics. 3rd ed. Sparky House Publishing, 299 pp. [Available online at http://www.biostathandbook.com/.]
Merz
,
B.
,
H.
Kreibich
,
A.
Thieken
, and
R.
Schmidtke
,
2004
:
Estimation uncertainty of direct monetary flood damage to buildings
.
Nat. Hazards Earth Syst. Sci.
,
4
,
153
163
, doi:.
Miceli
,
R.
,
I.
Sotgiu
, and
M.
Settanni
,
2008
:
Disaster preparedness and perception of flood risk: A study in an alpine valley in Italy
.
J. Environ. Psychol.
,
28
,
164
173
, doi:.
Mileti
,
D. S.
,
1995
: Factors related to flood warning response. U.S.– Italy Research Workshop on the Hydrometeorology, Impacts, and Management of Extreme Floods, Perugia, Italy, National Science Foundation, 4.6. [Available online at http://www.engr.colostate.edu/ce/facultystaff/salas/us-italy/papers/46mileti.pdf.]
Mimikou
,
M.
, and
A.
Grammatikogiannis
,
2006
: Real-time monitoring and management of point and areal hydrometeorological data in the Athens metropolitan area. Proc. Fifth FRIEND World Conf.: Climate Variability and Change—Hydrological Impacts, IAHS Publ. 308, Havana, Cuba, IAHS, 31–36. [Available online at http://iahs.info/uploads/dms/13632.10-31-36-48-308-Mimikou.pdf.]
Molinari
,
D.
,
F.
Ballio
,
J.
Handmer
, and
S.
Menoni
,
2014
:
On the modeling of significance for flood damage assessment
.
Int. J. Disaster Risk Reduct.
,
10
,
381
391
, doi:.
Nirupama
,
N.
, and
S.
Simonovic
,
2007
:
Increase of flood risk due to urbanisation: A Canadian example
.
Nat. Hazards
,
40
,
25
41
, doi:.
NOA
,
2016
: National Observatory of Athens. Institute for Environmental Research and Sustainable Development. Weather forecast group. [Available online at http://www.meteo.gr/weather_cases.asp and http://www.meteo.gr/meteoplus/weatherevents.cfm.]
O’Neill
,
E.
,
M.
Brennan
,
F.
Brereton
, and
H.
Shahumyan
,
2015
:
Exploring a spatial statistical approach to quantify flood risk perception using cognitive maps
.
Nat. Hazards
,
76
,
1573
1601
, doi:.
Papagiannaki
,
K.
,
K.
Lagouvardos
, and
V.
Kotroni
,
2013
:
A database of high-impact weather events in Greece: A descriptive impact analysis for the period 2001–2011
.
Nat. Hazards Earth Syst. Sci.
,
13
,
727
736
, doi:.
Papagiannaki
,
K.
,
K.
Lagouvardos
,
V.
Kotroni
, and
A.
Bezes
,
2015
:
Flash flood occurrence and relation to the rainfall hazard in a highly urbanized area
.
Nat. Hazards Earth Syst. Sci.
,
15
,
1859
1871
, doi:.
Petrucci
,
O.
, and
A. A.
Pasqua
,
2012
:
Damaging events along roads during bad weather periods: A case study in Calabria (Italy)
.
Nat. Hazards Earth Syst. Sci.
,
12
,
365
378
, doi:.
Pistrika
,
A.
,
G.
Tsakiris
, and
I.
Nalbantis
,
2014
:
Flood depth-damage functions for built environment
.
Environ. Processes
,
1
,
553
572
, doi:.
Poussin
,
J. K.
,
W. J. W.
Botzen
, and
J. C. J. H.
Aerts
,
2014
:
Factors of influence on flood damage mitigation behaviour by households
.
Environ. Sci. Policy
,
40
,
69
77
, doi:.
Ruin
,
I.
,
J.-D.
Creutin
,
S.
Anquetin
, and
C.
Lutoff
,
2008
:
Human exposure to flash floods—Relation between flood parameters and human vulnerability during a storm of September 2002 in Southern France
.
J. Hydrol.
,
361
,
199
213
, doi:.
Ruin
,
I.
,
J.-D.
Creutin
,
S.
Anquetin
,
E.
Gruntfest
, and
C.
Lutoff
,
2009
: Human vulnerability to flash floods: Addressing physical exposure and behavioural questions. Flood Risk Management: Research and Practice, P. Samuels et al., Eds., CRC Press, 174–176.
Ruin
,
I.
, and Coauthors
,
2014
:
Social and hydrological responses to extreme precipitations: An interdisciplinary strategy for postflood investigation
.
Wea. Climate Soc.
,
6
,
135
153
, doi:.
Schroeder
,
A. J.
, and Coauthors
,
2016
:
The development of a flash flood severity index
.
J. Hydrol.
,
541
,
523
532
, doi:.
Shipley
,
B.
,
2016
: Cause and Correlation in Biology: A User’s Guide to Path Analysis, Structural Equations and Causal Inference with R. 2nd ed. Cambridge University Press, 314 pp.
Špitalar
,
M.
,
J. J.
Gourley
,
C.
Lutoff
,
P.-E.
Kirstetter
,
M.
Brilly
, and
N.
Carr
,
2014
:
Analysis of flash flood parameters and human impacts in the US from 2006 to 2012
.
J. Hydrol.
,
519
,
863
870
, doi:.
Terti
,
G.
,
I.
Ruin
,
S.
Anquetin
, and
J.
Gourley
,
2015
:
Dynamic vulnerability factors for impact-based flash flood prediction
.
Nat. Hazards
,
79
,
1481
1497
, doi:.
Terti
,
G.
,
I.
Ruin
,
S.
Anquetin
, and
J. J.
Gourley
,
2017
:
A situation-based analysis of flash flood fatalities in the United States
.
Bull. Amer. Meteor. Soc.
,
98
,
333
345
, doi:.
UNISDR
,
2004
: Living with Risk: A Global Review of Disaster Reduction Initiatives. UNISDR, 429 pp. [Available online at http://www.unisdr.org/we/inform/publications/657.]
Xiao
,
Y.
,
Q.
Huang
, and
K.
Wu
,
2015
:
Understanding social media data for disaster management
.
Nat. Hazards
,
79
,
1663
1679
, doi:.

Footnotes

Publisher’s Note: This article was revised on 5 September 2017 to include the open access designation that was missing when originally published.

© 2017 American Meteorological Society. For information regarding reuse of this content and general copyright information, consult the AMS Copyright Policy (www.ametsoc.org/PUBSReuseLicenses).