Elsevier

Journal of Hydrology

Volume 541, Part A, October 2016, Pages 636-648
Journal of Hydrology

Assessment of commuters’ daily exposure to flash flooding over the roads of the Gard region, France

https://doi.org/10.1016/j.jhydrol.2016.01.064Get rights and content

Highlights

  • Compares methodology to provide realistic estimations of dynamic human exposure to flash flood hazard.

  • Illustrates the risk amplification effect of daily mobility.

  • Highlights specific geographic locations and social profiles particularly exposed to flash flood risk in the Gard region.

Summary

Flash floods are responsible for a majority of natural disaster fatalities in the USA and Europe and most of them are vehicle-related. If human exposure to flood is generally assessed through the number of inhabitants per buildings located in flood prone zone, it is clear that this number varies dramatically throughout the day as people move from place to place to follow their daily program of activities. Knowing the number of motorists exposed on flood prone road sections or the factors determining their exposure would allow providing a more realistic evaluation of the degree of exposure. In order to bridge this gap and provide emergency managers with methods to assess the risk level for motorists, this paper describes two methods, a simple rough-and-ready estimate and a traffic attribution method, and applies both of them on datasets of the Gard département, an administrative region of Southern France with about 700 000 inhabitants over 5875 km2. The first method to obtain an overall estimation of motorists flood exposure is to combine (i) the regional density of roads and rivers to derive a count of potential road cuts and (ii) the average daily kilometers driven by commuters of the study area to derive the number of people passing these potential cuts. If useful as a first approximation, this method fails to capture the spatial heterogeneities introduced by the geometry of river and road networks and the distribution of commuters’ itineraries. To address this point, this paper (i) uses a pre-established detailed identification of road cuts (Naulin et al., 2013) and (ii) applies a well-known traffic attribution method to existing and freely available census datasets.

Both methods indicate that commuters’ exposure is much larger than the number of commuters itself, illustrating the risk amplification effect of mobility. Comparing the results from both methods shows that (i) the road network geometry plays a significant role in reducing the risk of river-road dangerous intersections and (ii) not all commuters are equally exposed. Evidently commuters who have longer routes are more exposed, but residents of rural municipalities as well as professionals with highly qualified jobs are also more exposed. Finally, these exposure assessment methods applied to the Gard area allows locating road sections where commuters’ exposure to flood is high. It also sets the first step toward the implementation of a modeling platform able to combine the estimation of daily travel patterns exposure and behavioral response of motorists to road flooding, a critical input for emergency services and services in charge of the management of road networks in flash flood prone areas.

Introduction

Flash floods are responsible for a majority of natural disaster fatalities in the USA, Australia and Europe. Many studies investigating fatality circumstances have shown that more than half of the flash flood fatalities in these countries are vehicle-related (Duclos et al., 1991, Coates, 1999, Jonkman, 2005, Jonkman and Kelman, 2005, Drobot et al., 2007, Ashley and Ashley, 2008, Coates and Haynes, 2008), making mobility a primary cause of vulnerability during flash floods (Ruin, 2010). The question of people’s motivation and circumstances for driving into flooded roads have been addressed by many researchers in the last ten years (Drobot et al., 2007, Ruin et al., 2007, Ruin et al., 2008, Coles, 2008, League, 2009, Maples and Tiefenbacher, 2009, Sharif et al., 2012, Diakakis and Deligiannakis, 2013, Spitalar et al., 2014; see Becker et al., 2015 for a literature review on the subject). These studies identified several factors as age, gender and experience, suggesting that younger male drivers (under 35) often underestimate the risk of driving into flooded roads (Drobot et al., 2007, Ruin et al., 2007). Familiarity with the route and maintaining normal daily activities, especially commuting to work is also an important factor of risk taking behavior around floodwater (Coates, 1999, Coates and Haynes, 2008, League, 2009, Maples and Tiefenbacher, 2009, Ruin et al., 2009). Finally and with respect to the previously cited factors, the analysis of flash flood vehicle-related fatalities shows that middle-aged men are more prone to such fatal accidents but whether it is a question of “active1” or “passive” behavior is not clear yet (Terti et al., 2015b). In fact, not much is known about the chance that behaviors of exposed people lead to fatalities, a useful information to allow road and emergency managers to assess the risk level or the mortality and to design protection strategy accordingly. In order to address this question we need to know how much of the road traffic is concerned and which types of daily commutes and commuters are more exposed to flash flood risk. At present, mobility aspects are not much taken into account when assessing human exposure and vulnerability to natural hazards. Most of the time, resident population density data is used assuming a static distribution, which contrasts with the fast dynamics of the flash flood phenomenon. Recently, Terti et al., 2015a, Terti et al., 2015b showed that daily and sub-daily variation of population distribution may provide a more accurate and appropriate assessment of human exposure to flash flood. Several studies in transportation research focused on road network vulnerability to adverse weather conditions (Koetse and Rietveld, 2009). Different methods were developed in order to identify critical road segments where disruptions would lead to severe consequences. Berdica (2002) defined road segments vulnerability as a function of the probability of occurrence of a hazardous event and the importance of related impacts in term of serviceability of road links. Jenelius (2009) introduced the concept of link criticality quantifying the road network vulnerability by measuring the increase of global travel cost when these links are closed. Versini et al., 2010, Naulin et al., 2013 proposed a system intended to evaluate road inundation risks based on high spatial and temporal resolution rainfall estimates in the Gard region (5875 km2) in southern France. They used an extensive inventory of road submersions over the last 40 years in order to rate the susceptibility of roads flooding at points where the hydrographic network intercepts the road network. Our approach is complementary to this study as it aims at introducing the human exposure component in the risk evaluation.

While those studies focus on the sensibility of the road segments and potential impacts on the network functioning, journey-time exposure of road users have been mostly addressed by studies dealing with traffic-related air pollution (Gulliver and Briggs, 2005, Beckx et al., 2009). In this purpose, activity-based approach offers an appropriate framework to simulate individual travel-activity patterns. These activity-based models consider travel behavior as derived from the demand of activity participation and aim at predicting the sequence of activities conducted by individuals (McNally, 2000). In this paper, we consider that professional and school activities are the main purposes for daily commutes (return trip) and risk taking behavior when facing flooded roads. In fact, those regular and foreseeable activities are strong constraints of the week-day schedules of the 70%2 of the Gard inhabitants who commute. According to a questionnaire survey made in 2004 on a sample of 960 Gard residents, professionals with daily work constraints are less willing to cancel their travels when a flash flood watch or warning is announced than non-workers (Ruin, 2010). Therefore, the present study proposes to assess the number and socio-demographic characteristics of the commuters whose daily routes cross potential road-flooding points (road cuts) as defined by Versini et al. (2010) and refined by Naulin et al. (2013). We consider that commuters who cross the most potential road-flooding points are the most exposed. Identifying the characteristics of the most exposed commuters is a necessary first step toward understanding if the most exposed are also the most vulnerable or if individual’s coping capacity plays a significant role in lessening vehicle-related flash flood losses.

This paper compares two methods to evaluate daily mobility exposure to flood hazards by using existing and freely available datasets. A first rough-and-ready method derives counts of road–river intersections (road cuts) and average daily kilometers driven by commuters from standard bulk statistics from the study area. The second more elaborated method uses, on the one hand, the road flooding susceptibility established by Naulin et al. (2013) and, on the other hand, a classical traffic attribution method affecting commuters to plausible travel routes and to potential road flooding points after origin-destination disaggregated census data. The methodology proposed is generic and could be extended to other regions provided with the same type of data. To our best knowledge, comparable studies are seldom (see road exposure considerations about avalanches in Zischg et al., 2005).

The study concerns a region of southern France characterized by a typical Mediterranean climate with heavy rainfall events triggering severe flash floods during the autumn season (Delrieu et al., 2005, Gaume et al., 2009). According to Delrieu et al. (2005), Gard’s climate gathers the three necessary ingredients to trigger large amounts of precipitation during several days but also within a few hours in the case of Mesoscale Convective System (MCS): (i) proximity of the Mediterranean sea as a reservoir of energy and moisture, (ii) a southerly flow that both advects and destabilizes air masses from the Mediterranean sea toward the coast, (iii) the surrounding relief of the Alps, Pyrénées and Massif Central mountains slows down and enhances perturbations. These elements may result in very localized or much larger disastrous flash flood events.

Since 1225, the Gard region suffered 506 floods, with 66% of the 353 municipalities cumulating at least 10 referenced flood events including some that were affected more than hundreds times (CG30, 2015). Between 1316 and 1999, Antoine et al. (2001) recorded 27 fatal flood episodes and 277 deaths in the Gard. Since 1999, five fatal events added about 30 casualties to the toll. In 2013, 50% of companies and 35% of inhabitants of the Gard area were located in flood prone zone. In addition, previous studies have shown that not only the Gard region is particularly sensitive to road flooding but also that flash flood risk is often underestimated by the population, specially when related with daily mobility and motor vehicle usage (Ruin et al., 2007, Versini et al., 2010).

Section 1 describes the dataset used and introduces some terminology and mathematical expressions. At the end of the article, a section details the notations used in the different formulas and the corresponding terminology. Section 2 proposes a first rough-and-ready approximation of the exposure based on basic data and simplified calculation. Section 3 introduces the methodological basis of a refined exposure assessment using traffic attribution of census datasets. Sections 4 Hydrologic heterogeneity at road–river intersections, 5 Spatial heterogeneity of commuters exposure examine the results from the traffic attribution method in order to test the hypothesis that all commuters are not equally exposed. Section 6 gives concluding comments.

Section snippets

Data base and notations used

The database built for this study is called MobRISK and its originality is to link two broad types of information: (i) the geographical description of the region provided by IGN3 including in particular the road and river networks as well as the administrative units and the topography, and (ii) the socio-demographic description of the population provided by INSEE

Rough-and-ready estimate of motorist exposure

In order to propose a first approximation of the number of motorists exposed to road flooding in the Gard department we (i) identified the number of potential flooding points by crossing information on road and rivers and (ii) assessed the number of motorists passing each day in front of these points. This section briefly examines what result can be obtained from a rough estimate in order to have a basis to appreciate the nature of the results obtained from a more elaborated methodology

Traffic attribution

We know from the census data, the municipalities of origin and destination of all the home-work or -school commutes. Assuming that the itinerary Ti followed by the individual i obeys to some simple criterion χ of minimum time or minimum distance for instance, a traffic attribution algorithm can specify the set of traveled road segments given the chosen criterion: Ti={r,rrunbyi(D,A)|χ}.

In this work we use the classical Dijkstra’s algorithm, a single source shortest path algorithm that provides

Hydrologic heterogeneity at road–river intersections

In this section we analyze the exposure of road–river intersections with respect to the area of the drainage basins potentially causing the road flooding. The size of the upstream watershed is considered as an index of danger since it characterizes the suddenness of the flood. The smallest watershed has the fastest response time and the quickest rise of the water levels according to the space–time scale relationship of hydrological processes (Ruin et al., 2008, Creutin et al., 2009). In a first

Spatial heterogeneity of commuters exposure

The above-described results allow investigating in more detail the spatial and socio-demographic distribution of commuters’ exposure to road flooding in the Gard area.

The spatial distribution of the potential road cuts shows that road flooding is a pervasive hazard affecting the entire area (Fig. 2). We can observe that the road cut density is higher in the western half of the Gard territory. This observation actually indicates that roads and rivers are not independent geometries, which

Socio-spatial heterogeneity of commuters’ exposure

Finally, we investigated the socio-spatial heterogeneity of commuters’ exposure to road flooding. By socio-spatial we mean that we consider both (i) socio-demographic features such as: age, gender, and professional occupation9, and (ii) geographical information, i.e. the type of municipality of residency (urban

Conclusion

This paper proposes two ways to evaluate commuters’ exposure to flash flood in the Gard region. A first simplistic way that we called “Rough-and-ready estimate of exposure” assimilates the networks of roads and rivers gridding the territory to a Buffon’s Needle problem assuming that the road network is a regular tessellation of the region and that the intersecting rivers are composed of randomly oriented linear segments of equal length. The second way is using a simulation traffic attribution

Notations used

As a general rule we used (i) low case ordinary letters for integer or real values designing entities or sizes with dimension and (ii) low case calligraphic letters to design category values with no dimension. In both cases upper case letters design functions.

aArea value [km2]
A(.)Area of (an upstream watershed for instance) [km2]
c(w, r)A road–river intersection
D(.)The socio-demographic state of an individual
dCA socio-demographic category vector
F(.)Cumulative distribution function
H(.)Histogram
i

Acknowledgements

The work developed in this paper was funded by the French national research agency ANR under two complementary projects ADAPTFlood and MobiCLIMeX respectively ANR-09-RPDOC-001-01 and ANR-12-SENV-0002-01 and is part of the MISTRAL-HyMeX framework.

References (39)

  • M. Spitalar et al.

    Analysis of flash flood parameters and human impacts in the US from 2006 to 2012

    J. Hydrol.

    (2014)
  • J.-M. Antoine et al.

    Les crues meurtrières, du Roussillon aux Cévennes

    Annales de Géographie

    (2001)
  • S.T. Ashley et al.

    Flood fatalities in the United States

    J. Appl. Meteorol. Clim.

    (2008)
  • J.S. Becker et al.

    A Review of People’s Behavior in and around Floodwater

    Wea. Climate Soc.

    (2015)
  • CG30, 2015. Population en zone inondable. Technical Report Conseil Général du Gard...
  • L. Coates

    Flood fatalities in Australia 1788–1996

    Aust. Geogr.

    (1999)
  • Coates, L., Haynes, K., 2008. Flash Flood Shelter-in-Place vs. Evacuation Research: Flash Flood Fatalities within...
  • A.R. Coles

    Managing Flash Floods: Risk Perception from a Cultural Perspective

    (2008)
  • J.D. Creutin et al.

    Catchment dynamics and social response during flash floods: the potential of radar rainfall monitoring for warning procedures

    Meteorol. Appl.

    (2009)
  • Cited by (25)

    • Road criticality assessment to improve commutes during floods

      2024, Journal of Environmental Management
    • Identification and evaluation of flood-avoidance routes in Tucson, Arizona

      2020, International Journal of Disaster Risk Reduction
      Citation Excerpt :

      Exposure to hazards while traveling is path dependent. Since people may move in and out of several hazardous areas on a single journey, exposure depends on distance traveled and the distribution of hazards within transportation networks [17]. Research on commuter exposure to air pollution often acknowledges that commuters will pass through varying levels of contamination, but these studies generally emphasize differences across modes of transportation and tend to compare exposure rates along a small number of sample routes [24–26].

    • Flood exposure analysis of road infrastructure – Comparison of different methods at national level

      2020, International Journal of Disaster Risk Reduction
      Citation Excerpt :

      Despite the simplicity of Equation (1–1), the quantity of exposure included there is not defined straightforwardly or uniquely. This is because the term is defined in several ways in the literature [1,18–23] and because of interchangeability between the concepts of exposure and vulnerability [1] often arising from the differences between researchers’ research questions, aims, and disciplinary backgrounds [18,24]. Also within studies of vulnerability and risk analysis to infrastructure, the terminology of exposure, vulnerability, and risk is used diversely [22,23].

    • Assigning Travel-Activity Patterns Based On Socio-Demographics for Flood Risk Assessment

      2020, Mobilities Facing Hydrometeorological Extreme Events 2: Analysis of Adaptation Rhythms
    • Preface

      2020, Mobilities Facing Hydrometeorological Extreme Events 2: Analysis of Adaptation Rhythms
    View all citing articles on Scopus
    View full text