cloud covered a town or city, its almost inevitable ignition would cause widespread destruction.” Although elaborate safety precautions have been proposed to minimize the chance of a large LNGspill, such an accident could still take place. Spills might result from storage tank ruptures, ship collisions, or leaks on land orat sea. They could occur during transfers from ships to storage tanks or during pipeline transfer on land. Because of the large chemical potential energy associated with the fuel in liquefied form andi its ability to form a large cloud of vaporrapidly, anyspill is potentially dangerous. Thus,it is vital to ex- plore the possible consequences of such spills and how to minimize adverse effects. When we began our research, existing models disagreed by over an order of magnitude on how far the vapor cloud from a particular large spill would travel downwind. There was also very little previous work on scaling relationships that would allow us to extrapolate from the results of small experiments to full-scale effects. There was no existing facility at which we could safely carry out large experimental LNG spills under the conditions necessary to validate our numerical models. Since mid-1978 we have participated in a series of spill experiments at the Naval Weapons Center, China Lake, Califormia, primarily to evaluate instruments for the larger spill experiments we plan, but also to gather data that can be used to improve our predictive ability. The first four of these tests involved spills of 5 m° of LNG each and produced vapor clouds that were warmer, extended farther, and contained regions more enriched in the heavier (and more detonable} hydrocarbons than expected. We are now engaged in series of 40-m? spill experiments at China Lake {using the instruments and data-gathering techniques derived from the previous experiments) to further improve our models and refine our techniques for even larger experiments. We began our experimental program with 5-m° spills because an existing facility at China Lake was amenable to such tests and because spills of this size were adequate to evaluate the performance of gas sensors in the field. We realized that these tests would not be large enough to permit observation of all important phenomena associated with the dispersion of large clouds of natural gas, but there was the possibility that such experiments would teach us something new about the process. Experimental array and instrumentation Figure 1 shows the instrument array used in our 5-m° spill experiments. The array consisted of eight primary stations distributed down- wind from the spill point, each one 28 equipped with a variety of grab samplers, gas sensors, and ther- mocouples (points 1 through 8). The primary stations were supplemented at points 9, 10, and 11 by anemometers and at point 12 by a Jet Propulsion Laboratory (JPL) gas sensor. There was also a LIDAR (laser radar} systern fielded by Computer Genetics Corporation and a 12-point square array of Mine Safety Appliances (MSA) gas sensors (not shownin the figure). Cables connected each instrument station to power sources and the data-acquisition trailer. The main function of the grab samplers. gas sensors, and thermocouples was to measure the concentration of the LNG vapor in various ways at various locations. The grab samplers consisted of evacuated bottles that were opened at prescribed intervals to collect a sample of gas for later analysis by mass spectrometer to verify the performance of our gas sensors. The Shell gas sensors(stations 1 and 2), developed and loaned to us by Shell Research Ltd., of England, measured heat loss from a heated filament exposed to the gas stream. The TSI sensors (siations 3 and 6), manufactured by Thermo-Systems, Inc., of St. Paul, Minnesota, operated by forcing the gas through a sonic nozzle and measuring, with a thin-film anemometer, the flow velocity (which depends strongly on gas composition}. The infrared gas analyzer (station 4), custom-built for us by Anarad, Inc., was capable of distinguishing methane, ethane, and propane. It gave us ourfirst quantitative evidence of differential boiloff of the different hydrocarbons froin the liquid pool.