DANGEROUS GOOD TRANSPORTATION BY ROAD:

FROM RISK ANALYSIS TO EMERGENCY PLANNING

B. Fabiano*, F. Currò, A.P. Reverberi, R. Pastorino

DICheP – Chemical and Process Engineering Department “G.B. Bonino”,
University of Genoa, Via Opera Pia, 15 – 16145 Genoa, Italy

Abstract

Despite the relative recent move towards inherent safe materials, the relentless drive of consumerism requires increased quantities of dangerous goods to be manufactured, transported, stored and used year on year. The safety and effectiveness of road transport systems is to be considered a strategic goal in particular in those countries, like Italy, in which 80% of goods are transported by this means. In this paper, we face the risk from dangerous good transport by presenting a site-oriented framework for risk assessment and developing a theoretical approach for emergency planning and optimization. In the first step, we collected field data on a pilot highway and developed a database useful to allow a realistic evaluation of the accident frequency on a given route, by means of multivariate statistical analysis. To this end, we considered both inherent factors (such as tunnels, bend radii, height gradient, slope etc), meteorological factors, and traffic factors (traffic frequency of tank truck, dangerous good truck etc.) suitable to modify the standard national accident frequency. By applying the results to a pilot area, making reference to flammable and explosive scenarios, we performed a risk assessment sensitive to route features and population exposed. The results show that the risk associated to the transport of hazardous materials, in some highway stretches, can be at the boundary of the acceptability level of risk set down by the well known F/N curves established in the Netherlands. On this basis, in the subsequent step, we developed a theoretical approach, based on the graph theory, to plan optimal emergency actions. The effectiveness of an emergency planning can normally be evaluated in term of system quickness and reliability. As a case study, we applied the developed approach to identify optimal consistency and localisation in the pilot area of “prompt action vehicles”, properly equipped, quick to move and ready for every eventuality. Applying this method results in an unambiguous and consistent selection criterion that allows reduction of intervention time, in connection with technical and economic optimisation of emergency equipment.

Keywords Accident frequency;Hazardous Materials; Emergency; Transportation

  1. Introduction

Despite the relative recent move towards “inherent safe” materials, the relentless drive of consumerism has required increased quantities of dangerous goods to be manufactured, transported, stored and used year on year (Thomson, 1998). Of the different way of transportation, rail has higher damage potential, as larger quantities are transported by this means. However, considering the damage it may cause to life and properties, transport by road is more hazardous, as roads often pass through populated areas, especially in developing countries. The recent EEC Directive 96/82/EC implies the evaluation of risk in highly industrialized areas by means of Quantitative Area Risk Analysis techniques. It must be evidenced that certain dangerous substances are transported along particular Italian road routes in quantities that would exceed the threshold for safety notification or declaration, set down in Italy by Seveso II Directive, if stored in a fixed installation. On the other side, it must be remembered that EEC Directive 94/55/EC implies the harmonisation of the different national legislation on transport of hazardous materials by road. The safety and efficiency of road transport is to be considered a strategic goal in particular in those countries, like Italy, in which about 80% of goods is transported by this means, with a 30% increase with reference to the 2010 forecast. In particular, Italian highways are very crowded with trucks, considering that 17% of the whole good traffic by road of EU (15 Countries) is transported on these highways. Moreover, the number of cars is still steadily increasing, making a place on the road more and more a scarce commodity. Empirical evidence shows that though improvement in transport safety, in Italy a consistent number of serious accidents on motorways and highways keeps occurring, evidencing that the risk connected to dangerous goods transport is comparable with the fixed plants one.

Analysis of the risks presented by the transportation of hazardous materials presents a very different risk than a fixed facility: detailed information on shipments is not available on a national, regional, or local level in contrast with fixed facility inventories (Pine & Marx, 1997). As reported by different researchers, a specifically tailored QRA methodology can represent an effective tool to assess the risk to people associated with the transport of dangerous substance. The selection of the best route for transport has been widely investigated (List, Mirchandani, Turnquist & Zografos, 1991) and was recently formulated as a “minimum cost flow problem”, which consists of determining, for a specific hazardous substance, the cheapest flow distribution, honouring the arc capacities, from the origin to the destination vertices (Leonelli, Bonvicini & Spadoni, 2000). Poor appreciation of factors related to road conditions such as road class, designated speed limits, traffic density, as well as of the population characteristics, is likely to result in a risk assessment insensitive to route specifics and over- or under-estimating the overall level of risk (Davies, 1999).

In this paper, a site-oriented risk analysis procedure is tested in a pilot area, starting from an in-depth inventory of hazardous materials transported and from a statistical analysis of traffic and accidents observed in the area. In fact, it must be observed that to ensure that a local emergency plan is complete, it must take into account the nature and extent hazardous materials are transported by road in the area. The results are then discussed and a mathematical model for optimising emergency planning is presented. A main focus in the management of emergencies has been on resources and logistics; in other words, having what and who you need it to meet the crisis within an urgent time frame (Kowalski, 1995). The importance of the ability of the emergency response services to minimize the damage was recently highlighted by a pilot project carried out in the Netherlands, where the evaluation method of external safety risk included three new criteria, additional to individual risk and societal risk (Wiersma & Molag, 2004):

  • “self-rescue” i.e. the ability of the people in the vicinity of the accident to safe themselves;
  • “controllability” i.e. emergency response services;
  • “consequences” i.e. analysis of representative scenarios in terms of number of fatalities, injuries and material damage.

The criterion of controllability is focused on the ability of the emergency response services to minimize the magnitudo and to prevent escalation of the accident. In case of accident in hazmat transportation and subsequent release into the environment, it is very important to have at one's disposal information on each chemical hazardous product involved, trained and skilful personnel and suitable “prompt action vehicles”, properly equipped to be employed if the above mentioned hazardous release would happen. To this end, in the last phase of this paper, the optimization algorithm is developed for solving the problem of optimal location of emergency vehicles in the pilot area.

2. Theoretical structure

2.1 Transportation risk analysis

Generally speaking, the concept of risk is the relation between frequency and the number of people suffering from a specified level of harm in a given population from the realization of specified hazards (Vrijling, Van Hengel & Houben, 1995).

The model required for our purposes is focused on a proper evaluation of the expected frequency of accidents If the route is divided into road stretches, each characterized by different characteristics, the expected number of fatalities as consequence of an accident occurred on the road stretch r and evolving according to a scenario S, can be expressed as:

(1)

where:

fr=frequencyof accident in the r-th road stretch[accident·year-1]

Nr,S= number of fatalities caused by the accident evolving according to a scenario S in the r-th road stretch [fatalities accident-1]

PS=probability of evolving scenarios of type S, following the accident initialiser (i.e. collision; roll-over; failure etc.) [-]

Transportation network can be considered as a number of vertices linked one another by a number of arcs. As shown in the following paragraph, the vertices represent origin-destination points, tool-gates, storage areas on the transportation network and the arcs are the roads connecting vertices. An arc between two vertices is characterized by a different number of road stretches and the expected number of fatalities for the arc is:

(1)

where Nr,S is the total number of fatalities according to eq. (2):

(2)

being the in-road and the off-road number of fatalities calculated respectively as:

(3)

(4)

where:

=consequence in-road area associated with scenario S[m2]

=consequence off-road area associated with scenario S[km2]

=probability of fatality for accident scenario S[-]

k=average vehicle occupation factor[-]

v=vehicle density on the road area[vehicle·m-2]

dP=population density [inhabitants·km-2]

The frequency of an accident involving a scenario S, on the r-th road stretch, can be expressed as:

(5)

(6)

(7)

where:

r=expected frequency on r-th road stretch[accident·km-1·vehicle-1·year-1]

Lr=road length [km]

nr=number of vehicles[vehicle]

0,r=national accident frequency[accident·km-1·vehicle-1·year-1]

hj=local enhancing/mitigating parameters[-]

As is well known, various factors influence the accidents: mechanical, environmental, behavioural, physical, road intrinsic descriptors. A statistical multivariate analysis was performed, by comparing historical accident data related to the whole regional highways and data directly collected on the field on each stretch, in order to highlight relevant intrinsic road factors and meteorological, traffic conditions etc. (Fabiano, Currò, Palazzi & Pastorino, 2001).

Table 1 shows the parameters suitable to influence accident rates and grouped into three categories: intrinsic characteristics, meteorological conditions and traffic conditions. The values of the parameters are in the range 0.8-2.5.

2.2 Emergency planning

The effectiveness of an action aimed at facing an emergency situation can normally be evaluated in terms of systems quickness and reliability. To approach the optimisation problem we adopted the graph-theory, recently introduced by Beroggi &Wallace (1994) in computing optimal course of action for emergency response. Generally speaking, a linear graph may be defined as a set N of objects named vertices Vi (i = 1,…,n) and a set A of arcs linking couples of vertices (Vi, Vj). In details, a graph is a couple G(N,A) where N=[V1,…,n] is a set of vertices and A=[ai,j=(Vi,Vj)| Vi, Vj N] is a class of elements called arcs. Between two vertices, several oriented arcs may exist: the maximum number of the same oriented arcs between two vertices is called P and the graph is a P-graph. The set of vertices N can be run according to different ways: as tail, leading to the research in breadth (breadth-first), or as a pile, leading to research in depth (depth-first) or backtracking (Tarjan, 1972). In order to solve the minimum intervention time problem, a label d(i) is assigned to every vertex Vi, defining the path between vertices and a pointer pred(i), which shows the predecessor of Vi in the considered path. The sequence starts from a temporary value for d(i) which has to be modified, by iteration, as to reach the right value. After a comparative survey on various shortest path algorithms (Dreyfus, 1969) we considered the Dijkstra algorithm (Dijkstra, 1959) of label setting, as follows:

  1. d(s)=0; d(i) =  ; pred(i)=s;
  2. d(h)=min[d(i) / d(i) not exact]; d(h) becomes exact;
  3. if Vi A(h) and di not exact, d(i) = min[d(i), d(h) + ]; eventually pred(i)=h;
  4. if every value d(i) is exact, then stop, if not go to 2.

We developed the optimisation algorithm as schematised in Figure 1. Every vertex corresponds to a toll gate, a fire brigade station or to a storage area and the algorithm allots the exact value for d(i) at the last iteration for every vertex.

3.Case-study

The methodology previously presented was applied to a pilot area, referring to the routes starting from the Genoa port area (the most important in the Mediterranean basin) towards four direction: the industrialized North Italian and Central Europe districts, France and South of Italy. All of these highways are characterized by high truck traffic (mainly ADR) and inherent factors (ascribed to road out-of-date: the year of construction of A7 is 1935) determining to a major accident risk, with reference to both individual and social risk, defined according to the Dutch limits.

3.1 Data collection and analysis

The value of traffic and accidents for the four highways in the area are shown in Table 2. In particular, it can be noticed that A7 highway is characterized by values higher at least an order of magnitude than the accident frequency (6.0·10-8) calculated by other researchers for certain type of load threatening accidents (James, 1986), thus approaching the calculated values for urban road.

By considering the daily ADR traffic on the different highway sections, it results that the higher values of dangerous goods fluxes correspond to the intersection between the highways A10 (West riviera) and A12 (East riviera), in the stretch between the towns of Bolzaneto and Busalla and in the starting stretch, from the central port of Genoa (Genova Ovest toll-gate) to the connection between the highways A10 and A7.

The substances transported are shown in Figure 3: it is important to notice the high striking transport percentages of chlorine and ethylene oxide.

The immediate causes of accident are summed up in Figure 4.

The proportion of severe accidents on these highways during the years 1995-1999 is in the range 60-70% of the total accidents, defining a severe incident as one involving death, serious injuries, a fire or explosion, or more than EUR 25000 worth of damage.

3.2 Modelling

In order to obtain a correct evaluation of the density of the population which might be exposed to Hazmat hazards from transport it is necessary to include data on the population density along the route and on the so-called motorist density, taking into account, as well, the proportion which may be considered particularly vulnerable or protected. Otherwise, all individuals within a threshold distance from road stretches run the same risks regardless of their location.

The population density along a route segment can vary with time, such as from day to night, and from month to month: the average density on the route can be calculated starting from the collected statistical data relevant to average daily traffic, average speed and geometrical data of carriageway and lanes, in each highway stretch considered.

Also on-road population can vary during the day: in order to evaluate correctly the number of on-road population involved in the accident, the response and the variations in the motorist density as a consequence of an accident, were considered.

Two classes of motorist density are to be considered: the former refers to the carriageway, where the accident occurs, the latter considers the opposite carriageway, were the “ghoul effect” causes the slowing down of the traffic.

In order to evaluate the probability of death in the area involved, the consequence model was applied making reference to event trees for every type of accident consequence: pool fire, flash fire, jet fire, BLEVE, fire ball, UVCE, release of toxic substances.

Average individual risk, defined as “the frequency at which an individual may be expected to sustain a given level of harm from the realization of a specific hazard” (Dantzig & Kriens, 1960), has been determined averaging the estimating individual risk levels for all the individuals in the selected area, as above-described. For individual risk, we considered the upper acceptability criterion set down in the Netherlands in new situations or new developments, corresponding to 10-6 year-1.

The same technique was secondly adopted for the evaluation of societal risk. Societal risk analysis can lead, via the generation of expectation values (average number of lives lost) to the consideration of the need for, and cost benefit, of risk reduction measures, even if it involves many generalising assumptions and averaging (Purdy, 1993). In all concepts, the most stringent of the personally and the socially acceptable level of risk determines the acceptable level of risk. So both criteria have to be satisfied (Vrijling et al., 1995). The same acceptability criterion for individual and societal risk was considered by Alp and Eelensky (1996), in developing a rigorous mathematical platform on which risk assessment can be built. It must be evidenced that the societal acceptable risk criterion is not standardized in the different EU Countries. So, in the absence of a national statistical reference, we adopted again the F/N limit curves established in the Netherlands, dividing as well the so-called Alarp region into two bands: the acceptability criterion of the risk so modified is explained in Table 3, where P is the cumulative frequency per year and N the number of fatalities (Høj & Kröger, 2002).

The results of the risk evaluation for the pilot area are summarized in Figure 5.

To reduce “intervention time”, the localization of “prompt action vehicles” must found on scientific statement, as previously explained, taking into consideration the concept of “minimum pathway”. The main constraint the theoretical approach is based upon is that the emergency vehicle can not be placed in any site of the concerned provincial territory, but only in the Fire Brigades Central Department or in one of the six detachment. Central Department and the six detachments are to be considered as vertices in the “final graph”, which will solve the problem.

In the same way, it is possible to indicate the “hazardous areas”, where production, transformation and hazardous substances storage take place, as vertices of the “final graph”. To complete the construction of the “final graph”, it was necessary to take into consideration highly hazardous areas along the highway. We considered as vertices of the graph inlet and outlet toll-gates, fire brigade districts and central department, as well as production, transformation and storage sites of dangerous substances. The obtained final graph is depicted in Figure 6.

The arcs, which link vertices together, represents normal way and highway “units”, between all areas taken into consideration.

In order to evaluate the considered graph, we allotted to each arc a corresponding scalar, defined “cost of the arc”. This scalar value corresponds to a time: the Average Run Time (in minutes) needed to reach from a vertex the subsequent one. Time was calculated considering distances between vertices and assuming two average speeds: 80 km h-1 as highway speed and 30 km h-1 as urban speed.