Statistical analysis of oil and oil product spills at sea

Abstract

The influence of oil and its derivatives transportation processes, especially against the background of positive dynamics of their production, on the probability of environmental pollution is considered in the paper. It is noted that in such conditions it is very important to ensure timely preparedness for response and the ability to choose the most effective strategies, mainly using preventive measures to ensure emergency preparedness. The basis for the necessary set of measures is a forecast of the potential volume of a spill, allowing only the necessary number of technical or other assets to be planned for each area. It is emphasised that such a forecast is possible if the type of distribution by which it is described is understood. The type of distribution that best describes the behaviour of the oil spill volume as a random variable is defined in the paper. Statistical analysis methods have been used to examine the data presented in various databases. It is found that none of the common distributions that could theoretically describe the spill volume behaviour fulfil the accuracy condition. The observed chi-square (Pearson) criterion of agreement absolutely for all types of distributions exceeds the critical one many times, and it is obvious that the lognormal distribution law describes the behaviour of the considered random variable in the best way, which is visually confirmed by the frequency polygon and the smallest observed Pearson criterion. Using modern methods of hybridisation of distributions, it is proposed to conduct a study aimed at creating a universal distribution, presumably based on the lognormal distribution, which would accurately describe the behaviour of the random variable of oil spill volume. In addition, an approximate step function with a very high confidence coefficient can be used to predict the volume in each specific area.