Summary of Meeting Paper

The 1996 Annual Meeting of the Society for Risk Analysis-Europe

Assessment of Risk for the Transportation of Dangerous Goods. D.G. Leeming and S.A. Gadd, Health and Safety Laboratory, HSE, Broad Lane, Sheffield, UK; and T.N.K. Riley, CHD, HSE, St. Anne's House, Bootle, UK

INTRODUCTION

Recommendations made in the Advisory Committee on Dangerous Substances (ACDS) report of 1991, led the HSE's Major Hazards Assessment Unit (MHAU) to commission the Risk Assessment Section of the Health and Safety Laboratory (HSL) to develop Transport RISKAT, a computerised tool for the assessment of risks to people arising from the transport of dangerous substances.

Transport RISKAT (Transport Risk Assessment Tool) is designed to estimate the risks to people arising from potential releases of toxic and flammable substances during transport by road and rail. The program is based on a generalisation of the Quantified Risk Assessment (QRA) methodology developed in the ACDS report (1991). On a selected transport route, the risks to both the on and off-route populations are estimated and the total risk is expressed in terms of 'route societal risk'. Societal risk is used because of the potential for harming a significant number of people in a single incident. The risk to any one person, on the other hand, is likely to be low, since individuals along the route will only be exposed briefly to the potential hazard. Transport RISKAT enables comparisons to be made between the risks arising from different routes or modes of transport, or from different sections of the same route. In addition, it can be used to identify risk 'hot-spots' along a particular route and the lowest risk option from a number of alternatives.

In this paper, the discussion of Transport RISKAT is restricted to consideration of the risks arising from the transport of liquefied toxic gases which form dense clouds when released into the atmosphere (e.g. chlorine and ammonia). The methodology is described, and a case study outlined in which the implications of changing the mode of delivery to a site from rail to road were assessed.

METHODOLOGY OF TRANSPORT RISKAT

Route Definition

In order to define the route, the selected rail or road route is first divided into a number of sections, of variable length, but with similar characteristics. The sections are characterised by: road type (road routes only); travel direction; type and quantity of substance transported; density and distribution of surrounding resident population. The residential population is defined by three uniform bands of average density, parallel to the route. The width of these bands, and the appropriate population density within them is, at present, determined from visual inspection of 1:50,000 scale O.S. maps for each route section. A single set of representative weather conditions is chosen for the whole route.

Frequency Analysis

Once the route has been defined, the release frequencies for each potential release of hazardous material from road tankers and rail wagons, for each route section, are estimated. In the ACDS report (1991), two main causes of spillage from dangerous goods road tankers and rail tank wagons were identified; puncture or damage to the tanker/wagon following an accident; and failure or mal-operation of the tanker/wagon equipment. In the case of rail transport. collision and derailment were the two main causes of accidents.

A high degree of uncertainty is recognised in the estimation of release frequencies for toxic gases following an accident. At the time of the work for the ACDS report (1991), there were no recorded incidents in the UK where properly designed road or rail tankers for pressurised liquefied toxic gases had been punctured as the result of an accident, however, such an event remains foreseeable. Consequently, for the ACDS report (1991), an approach was adopted using an analysis by ICI Transport Engineering Section to estimate appropriate puncture probabilities. Data from incidents involving punctures of 'thin' walled tank wagons following accidents were analysed. Expert engineering judgement was then used to estimate the probability of failure had the vessel concerned been a 'thick' walled ammonia or chlorine vessel. The results of this analysis were checked using a statistical technique which gave consistent results. Equipment failure frequencies were quantified by analysing historical data. Event trees were constructed to determine release frequencies for each route section for each potential release scenario, combining accident and equipment failure frequencies with the probability of vessel puncture. Two additional sources of release frequency information which can be used are: Saccomanno et al. (1989) and Saccomanno et al. (1993).

Consequence Analysis

The consequences of a release of toxic gas depend upon the nature of the release (i.e. whether it is instantaneous or continuous), and on the mass released. In an instantaneous release, the entire contents of the tanker are released at once, whereas in a continuous release, the tanker contents are released over a period of time. Dense gas dispersion models (Fryer and Kaiser, 1979; Jagger, 1983) are used to describe how the gas cloud spreads, for each potential release scenario, for each route section. The dispersion of the toxic cloud is influenced by the prevailing atmospheric conditions.

The impact of the toxic cloud on the population is calculated using a 'probability of fatality' approach based on a probit relationship for the dose-response relationship (toxicology). The 90%, 50% and 1% fatality hazard areas are calculated for both the outdoor and indoor population. The percentage of the resident population indoors is dependent on the weather conditions, and it is assumed that a percentage of those outdoors can 'escape' indoors depending on the level of exposure.

Risk estimation

The number of people exposed to each representative release is derived from the calculated hazard areas and the population information. For each release scenario, the mean number of fatalities is combined with the estimated release frequency. Thus, an estimate of societal risk for each route section, and for the route as a whole, is calculated. Route societal risk is expressed using F-N curves, which are plots of the number of fatalities, N, against the cumulative frequency per year of N or more fatalities.

ONGOING RESEARCH

APPLICATION OF TRANSPORT RISKAT TO A CASE STUDY

Background

The case study involved the use of Transport RISKAT as part of an assessment of the overall risks to population from a major industrial facility in the UK (henceforth referred to as 'the site'). The site uses chlorine delivered from elsewhere in the UK in the manufacture of various products. The assessment was prompted by a proposal by the site owners to switch from rail delivery of chlorine by one route, to road delivery by another. Frequent "just in time" road deliveries would reduce the amount of chlorine being stored on the site and, in turn reduce the risk to population in the vicinity of the site.

The proposal generated two questions which were addressed in this study: is the risk to population along the route greater from the proposed road route than from the existing rail route, and if so, are the overall risks reduced as a result of the change in the delivery system? Transport RISKAT was used to carry out a comparative assessment of the en-route risks for transporting chlorine by rail and road to the site. A comparative assessment of the risks from on-site operations before and after the switch in transport mode was performed using HSE's tool RISKAT (Clay et al. 1994). The total risks of site and delivery system with each mode of delivery were then considered.

Methodology

The two routes were sub-divided into sections possessing similar characteristics. For the rail route, all chlorine transport took place at night when passenger traffic on the rail network was negligible, and consequently no on-route population was included in the study. For the road route, the motorist population was assumed to be constant along the route, and to be concentrated around an accident due to congestion behind the accident, and as a result of motorists on the opposite carriageway slowing down to view the accident. Motorists were also considered to be effectively outdoors. In the case study, to ensure that the results of the comparison of the supply options took account of variability in estimation, release frequency estimates from three independent sources were considered. As the same models and assumptions were applied to the risk assessment for each delivery route, the inherent uncertainty was similar in each case. This is significant when comparing the rail and road risks since then it is the relative rather than the absolute risk which is important.

Results and Outcome

Within the uncertainties of the risk estimation, no significant difference between the risks from the road and rail routes was found. The different sources of release frequencies led to conflicting conclusions regarding the route societal risk. In all cases however, the motorist population contributed on average about two thirds of the societal risk due to the road delivery system, implying that the model is very sensitive to assumptions made about the on-road population. The main uncertainties arose from the uncertainty of the release frequency information, and from the population data. The risk estimate is highly sensitive to changes in the release frequency and to the variation of population density within 1 km of the route. This effect is more pronounced for rail than road due to the larger contribution to the overall risk from the off-route population. The results indicate that the societal risk estimates are more sensitive to assumptions made about the variation of population density with distance than to variations in the frequency analysis.

The change in delivery mode significantly reduced the risks from the site itself The result, within the uncertainties of the risk estimates, did not show that a significant risk was being transferred to the population along the road route, and indicated that overall the risk was reduced. Today, chlorine is delivered to the site by road.

CONCLUSIONS

The methodology of Transport RISKAT for toxic substances has been described. The estimates of societal risk are sensitive to the assumptions which are made about the population. The density and distribution of population along possible routes may be a deciding factor in the comparison of hazardous materials transport options. The results of the application of Transport RISKAT to a case study have been presented.

REFERENCES

1) Advisory Committee on Dangerous Substances (1991). Major hazard aspects of the transport of dangerous substances. HSC, HMSO, ISBN 011 885676 6.

2) Clay, GA, Nussey, C and Corlett, TC (1994). The basic methodology of toxic RISKAT. HSL Internal Report IR/L/RAW94/1.

3) Fryer, LS and Kaiser, GD (1979). DENZ - a computer program for the calculation of the dispersion of dense toxic or explosive gases in the atmosphere. UKAEA Report SRD R152.

4) Jagger, SF (1983). Development of CRUNCH: A dispersion model for continuous releases of a denser than air vapour into the atmosphere. UKAEA Report SRD R229.

5) Leeming, DL (1993). Assessment of the Risks of Transporting Hazardous Substances by Road and Rail - Sensitivity of the QRA model (Toxics) to the Population Distribution. HSL Internal Report IR/L/HA/93/3.

6) Saccomanno, FF, Shortreed, JH and Van Aerde, M (1989). Assessing the Risks of Transporting Dangerous Goods by Truck and Rail. Final Report, WRI Award No. 1254501, IRR, Canada.

7) Saccomanno, FF, Leeming, D and Stewart, A (1993). Comparative Assessment of Risk Model Estimates for the Transport of Dangerous Goods by Road and Rail. A review of the Benchmark Corridor Exercise, part of the International Consensus Conference on the Risks of Transporting Dangerous Goods, Toronto, Canada, April 1992.