Research on continuous and instantaneous heavy gas clouds

Abstract

This report describes the contribution of Brunel University to the joint CEC project 'Research on Continuous and Instantaneous Heavy Gas Clouds' under the Major Technological Hazards programme (CEC Contract EV4T.0025.UK(H)). Brunel University's main task in this project was concerned with the analysis of experimental data provided by some of the other project collaborators. Liaison with these collaborators, and with others undertaking other aspects of data analysis, was obviously also important. The experimental data were obtained both from full-scale field trials (Tuv/Risφ) and from wind tunnel experiments (TNO, University of Hamburg, Warren Spring Laboratory). Some of the data sets are very large. The main effort of data analysis has been concentrated on the data from Tuv/Ris sφ and Warren Spring Laboratory. This was mainly because of the timely arrival of substantial quantities of data from these sources, and also to avoid direct duplication of work carried out by other collaborators. Nevertheless, some analyses were made of TNO and University of Hamburg data. The Tuv/Risφ data set had one extremely valuable property, namely that the concentrations were measured by several different methods. Analysis here confirmed the view - hitherto essentially a theoretical speculation with no substantial experimental support - that the instrumentation can itself have a significant effect on the measured concentration. One consequence of the results of Brunel's analysis of the Tuv/Risφ data set is therefore that caution must be exercised in validating practical models of hazard assessment. Interest also attaches to this data set in that, in some of the experiments, obstacles were removed while the experiment was running; some analysis of "before and after" effects has been undertaken. For example, comparisons were made of such effects on levels of concentration and concentration variability, and two different algorithms have been developed to illustrate these features and, indeed, to determine, simply from the time series, when the obstacles were removed. A major and most welcome feature of the Warren Spring Laboratory data set was that it recorded many repetitions of gas releases under identical experimental conditions. Because of this, it was possible to study the variations in the concentration data from one release to another and to build up an initial simple statistical understanding of the situation. In such circumstances, statistical measures such as mean and variance may be estimated as ensemble averages, rather than by considering them as time averages within a single release; this latter approach can be questionable, particularly if the data do not exhibit statistical stationarity. The results of Brunel's analysis of this data set, though not yet complete, amply justify the "repetitions" strategy. The report illustrates this conclusion by presenting typical results that could not otherwise have been obtained, and which have important implications for real-life. The TNO wind tunnel experiments were conducted both for the purpose of comparing results with those from other wind tunnels and to provide a simulation of one of the full-scale Tuv/Risø field trials. The resulting data set is potentially very valuable, but Brunei's analysis has identified a number of points for concern. Thus there are some doubts about the behaviour of the instrumentation, while some of the experimental results are atypical of those obtained by other collaborators and occasionally seem hard to reconcile with physical intuition. Concerning the University of Hamburg data set, Brunel was aware that extensive and detailed analyses had been carried out by the Health and Safety Executive. Brunel did not wish to essentially duplicate this effort. Brunei's work here was, therefore, largely confined to replicating some of the HSE analyses for the purpose of confirming results - an aim that was always achieved. The HSE analyses are discussed formally in HSE's report under this contract, and were presented informally to meetings of the collaborators during the summer. Unavoidable resource constraints have prevented much progress in moving forward from data analysis to the development of models. However, work of this nature is still in progress after the termination of the formal contract. Such work is justified by the quantity and quality of the data, and is expected to form an important input to research under the FLADIS contract.