Accidental spillage of chemicals by the process industries can pose both immediate and long-term threats to the natural environment. Long-term damage from frequent, low-level releases is difficult to predict and may not be evident until the damage is done. This thesis reviews techniques for environmental risk evaluation and proposes a number of improvements. These improvements include: • Definition of an endpoint in the natural environment around which to evaluate risk; • Estimation of risk from the cumulative stresses assaulting the endpoint; • Using a specially established maximum acceptable concentration for the endpoint, • rather than using an acute toxicity based on a fractional kill; • (III Using a probabilistic analysis rather than a deterministic analysis, thus avoiding the • use of an average or a worst-case scenario; • Including a sensitivity and an uncertainty analysis. Aspects of the improved environmental risk method are demonstrated in a case study. The case study considered in this thesis evaluates the potential long-term effects to an important shellfish bed in a harbour near an oil refinery. Modelling is the analytical tool used. It is a relatively inexpensive method, compared with ongoing field-sampling and laboratory analysis, to identify specific chemicals, industrial operations, and sites which might require further testing and investigation. The case study in this thesis focuses on an estimation of the annual volume of hydrocarbons spilled at the refinery wharf and also on the estimation of BTEX exposure from these spills. A large fraction of automotive petrol is composed of BTEX (Benzene, Toluene, Ethylbenzene, and Xylene) and this was selected as the contaminant. Benzene is present in concentrations of order 5% w/w and the combined mono-aromatics, BTEX, amount to 45% w/w in New Zealand motor fuel (premium grade, 96 octane). The exposure concentration was estimated from refinery wharf off-loading spills in Whangarei Harbour, New Zealand to a nearby shellfish bed, the Mair Bank, at the mouth of the tidal estuary. The shellfish bed is not only a fishing resource, but also lends stability to the coastline where the petroleum refinery stands. BTEX is a significant contaminant to evaluate in regard to shellfish bed health because molluscs are generally more sensitive to light fuel or refined oils than crude oils. The probable annual volume of all hydrocarbons spilled, as well as the probable exposure concentration over one operational year from small-scale, oil-spills were estimated from reconstructed historical frequency data and calculations of environmental transport. The frequency of harbour spills over 24 years was used from Northland Regional Council records. Unfortunately, only those spills estimated to be over one cubic meter in size were recorded and spills under one cubic meter were of most concern in this study. Therefore, in order to estimate the total number of spills per year, including those spills under one cubic meter, additional data were required. Detailed spills data collected over five years from marine terminals were obtained from the California State Lands Commission. Using these data, two methods of data reconstruction were used. In general, the shape of the Californian spill-size distribution was maintained, although a frequency proportionality factor was used since spills were more frequent at this refinery wharf. In addition, there were large spills at the refinery wharf (4 in 24 years), so the tail of the distribution needed to be lengthened and thickened. The result was that the mean spill size at the refinery wharf was estimated at 0.166 and 2.55 cubic meters for each of the two reconstruction methods. The average number of spills per year in loading and unloading operations predicted by these methods were 112 and 98. The exposure model combined and integrated the contaminant-transport rate processes. The significant rate processes affecting the exposure of BTEX or aromatic fraction of automotive petrol to the shellfish were evaporation, advection-induced dispersion, buoyancy-induced dispersion, and spill break-up or entrainment. Other transport processes were comparatively minor for the time frame of concern. The significant contaminant-transport rate processes were then combined with the dynamics of the natural environment to result in an exposure to the shellfish bed as a function of time. These dynamics included periodic tidal flow, temperature, and wind speed which fluctuated with the time of year. The modelled exposures over time were compared with published sub-lethal, sensitive life-stage and adult lethal marine toxicity criteria. The exposure concentration range of interest in this study was the sub-lethal range which affected reproduction and growth. Not only is the exposure concentration important, but also the frequency of exposure must be considered when evaluating shellfish health. Bivalves, in general, compared to fish and crustacea, have a very low level of activity of enzymes capable of metabolising organic contaminants, such as aromatic hydrocarbons and, once exposed, the time taken to rid tissue of the contaminant is critical. Monte Carlo simulation results of spillage at the refinery wharf showed that depending on the spill-size distribution used, the annual volume may range from 15 to 283 cubic meters. The lower range of this spillage volume compared well, proportionately, with Port Tauranga's published total of annual successful prosecutions for oil spillage. Monte Carlo simulation results of Mair Bank shellfish exposure to BTEX for one year are available for several scenarios of input spill-size distribution and spill frequency. These results indicate sub-lethal exposures greater than 10 pm BTEX at average intervals of 9 to 27 days and average exposure concentrations of 5000 to 19,500 ppb, respectively. This exposure frequency may not provide enough time between exposures for the shellfish to adequately recover. While a few excursions exceed the adult lethal toxic level, it is difficult to make fatality predictions for tidal creatures which have the ability to close up for long periods of time. What is of concern is that the exposures are closer together than the ability of the creatures to eliminate the toxins from their tissues. The output of the Monte Carlo simulation was sensitive to the input probability distributions, as well as the number of transfer events in a year, and the evaporation rate. Small spills under 1 cubic meter were almost as damaging to the shellfish in terms of sublethal exposure as the large ones, 20 to 70 cubic meters. Because the evaporation rate was significant in this analysis of mono-aromatic compounds, the volatile aromatics in the smaller spills (less than about 0.060 cubic meters) would often evaporate completely, depending on the spill size and the time it took for the slick to get to the shellfish bed. Therefore, the greater the proportion of spills under 0.100 cubic meters, the smaller the exposure and the exposure frequency. Because extensive chemical analyses or biological monitoring can be so expensive and because a natural community may appear healthy even when it is sick, the proposed methodology is a comparatively inexpensive way to check the need for further analysis. The results of this analysis are the first step to determine the long-term risk to a nonhuman population. The next step is to refine population and growth models to include the effect of sub-lethal contaminants. Given the presently available distribution of spillage size and frequency, damage to the long-term health and population of this shellfish bed is possible. Further, more detailed analysis and monitoring are recommended.