Dense Gas Plumes Research Articles

Dense gas plume rise and touchdown for Jack Rabbit II trial 8 chlorine field experiment

The series of nine Jack Rabbit II (JR II) chlorine release field experiments in 2015 and 2016 involved releases of pressurized liquefied chlorine from tanks containing 5–20 tons of liquid. This paper concerns JR II Trial 8, where the release hole (15.2 cm diameter) was oriented upwards at the top of the tank. To our knowledge, this is the only full-scale vertically-directed release of pressurized liquefied chlorine that is available for detailed analysis. Our aim is to improve understanding of the characteristics of the dense jet trajectory (e.g., maximum plume rise, downwind distance where the dense plume first touches the ground, and maximum concentration at touchdown). The chlorine emission rate was measured by instruments in the tank, and the plume rise geometry was estimated from photographs, videos, and lidar scans taken from several angles. During the first 10 s of the release, the videos show that the dense jet rises up to about 40 m, then sinks towards the ground, touching down at a distance of about 60 m. The portion of the plume that sinks to the ground then spreads out about 30–40 m in all directions due to gravity slumping. Due to the decrease with time of the momentum jet mass release rate, the jet gradually becomes less dense and less powerful and the plume remains aloft after about 30–60 s. However, the dense cloud that touched down in the early phases moves downwind as a typical dense cloud. The observed dense jet plume rise and touchdown distance are shown to approximately agree within about a factor of two with integral model predictions by 1) the analytical dense gas plume formulas suggested for vertical jets in 1973 by Hoot, Meroney, and Peterka (HMP), 2) the analytical formula for buoyant (positive or negative) plume trajectory suggested by Briggs in 1969, and 3) the DRIFT dispersion software, which is an integral model. Also, HMP and DRIFT model predictions of concentration at touchdown agree with Trial 8 observations within a factor of about two. It is concluded that the dense jet plume models are able to simulate the Trial 8 plume trajectory and ground level maximum concentration within acceptable ranges (i.e., most of time, within a factor of two).

Comparison of the predictive results from the two dispersion models PUMA and LPELLO with the JR II field data

Abstract Events triggered by unfortunate accidents or malicious antagonistic attacks bear witness to the great health risks involved in sudden releases of poisonous liquefied gases. For the society to be better prepared for these types of crises, experimental data from controlled releases is essential. To address this, a set of large scale releases of chlorine gas was conducted in the Jack Rabbit II field trials. The resulting dense gas plumes were monitored at distances up to 11 km by means of a comprising network of sensors. A group of international researchers involved in atmospheric dispersion modelling was formed to reproduce the field trials with different models as a benchmarking process to collectively improve upon the model suite and, in the end, be able to provide enhanced support to the society. The Swedish Defence Research Agency has previously developed two custom-made models for atmospheric dispersion on the local scale: PUMA and LPELLO. They use different modelling schemes but are able to handle the same types of dispersion processes. The validation process presented in this study was based on three selected field trials from Jack Rabbit II. Necessary meteorology and source data was, with some simplifications, congregated and distributed by the organizers to obtain a high degree of uniformity in the input data for the different models. Simulations were conducted by replicating these prerequisites whereby the resulting concentration plumes were analysed and compared with the measured data. There are many different metrics, properties and phenomena to scrutinize in a model validation process. After a thorough analysis, it was found that the two models showed similar plume behaviours in some aspects but also possessed quite diverse characteristics in others. In particular, the simulated concentration maxima were in general higher than the measurements indicate, which is analysed and discussed in detail. Findings and analyses presented here will have bearing on other dispersion models using similar schemes.

Large-eddy simulations of dense-gas dispersion within a high-Reynolds number turbulent boundary layer

Large-eddy simulations of flows over a backward-facing step with release of four different gases with increasing densities have been performed. The results have been analysed with particular emphasis on the part of the flow field sufficiently far downstream where the local effect of the source dynamics is no longer dominating the flow field. The dense gas plumes maintain high concentrations close to the ground and become approximately twice as shallow in comparison with a neutrally buoyant plume. The vertical mixing is significantly reduced close to the ground, and both momentum and scalar flux reversal are found in the two densest cases, indicating negative shear and buoyancy production of turbulence kinetic energy. Examination of the gradient Richardson number indicates that all dense gases are significantly affected by buoyancy. Interestingly, a narrow layer within the plumes is identified where the impact of stratification is expected to be reduced. It is found that relaminarization is most likely occurring and that fluctuations found might originate from interfacial wave modes rather than from turbulence. The large-scale structures are seen to resemble those found in the plane channel flow close to the wall and further away from it. In the middle layer, significant differences are found due to buoyancy effects. The large-scale structures become more pancake-shaped with large-scale vorticity almost exclusively about the spanwise direction in the dense gas cases.

Computational Fluid Dynamics Validation Utilizing a Tracer Gas Study Related to a Mine Mill Area Toxic Gas Release for Emergency Response Planning

Many dispersion models (e.g. DEGADIS, SLAB, INPUFF, and ALOHA) have been developed by regulatory agencies for emergency response related to dense toxic gas releases. The National Oceanic and Atmospheric Administration’s (NOAA) Area Locations of Hazardous Atmospheres (ALOHA) software is an example one such model. However, such models typically over-predict dense toxic gas plume dispersion concentrations and do not take into account complex terrain or complex building canyon geometries. This creates imprecise and inaccurate emergency response plans for industries that utilize such gases for production processes. The application of Computational Fluid Dynamics (CFD) mostly overcomes these challenges and provides a refined understanding of dense gas plume dispersion. Evaluating micro-scale atmospheric models is typically conducted in a wind tunnel tracer gas experiment; however, the one-hour tracer gas experiments utilized in this study evaluation were conducted in a real-world, full-scale, industrial setting. These same industries are often faced with worker health risks related to airborne respirable particulates and fibers with low settling velocities and where transport is dominated by air movement. Thus, it is anticipated that this evaluation may be applied to exposure characterization and risk assessment.

Fast estimation of dense gas dispersion from multiple continuous CO2 surface leakage sources for risk assessment

Surface leakage of CO2, and associated potential impacts on health, safety, and the environment (HSE) are considered hazards of geologic carbon sequestration (GCS). There are two challenges associated with impact assessment of CO2 surface dispersion. First, the fact that CO2 is a dense gas makes its dispersion in air a complex process. Rigorous numerical solutions for modeling concentration distributions are relatively time-consuming. Second, impact assessment requires consideration of uncertainty, e.g., quantification of how much uncertainty is propagated through input parameters to model outputs by carrying out large numbers of model runs. In order to assess the potential consequences of surface leakage of CO2, it is useful to have a model that executes very quickly for repeated model calculations (e.g., in Monte Carlo mode) of the atmospheric dispersion of CO2 (concentrations as a function of space and time). In addition, the model should be able to handle multiple surface leakage sources. In this study, we have extended the nomograph approach of Britter and McQuaid (1988) for estimating dense gas plume length from single leakage source to multiple leakage sources. The method is very fast and therefore amenable to general system-level GCS risk assessment including uncertainty quantification within the framework of the National Risk Assessment Partnership (NRAP) Integrated Assessment Model (IAM). The method is conservative in that it assumes the wind could be from any direction, and it handles multiple sources by a simple superposition approach. The method produces results in reasonable agreement with a sophisticated computational fluid dynamics (CFD) code, but runs in a small fraction of the time.

CFD modeling of water spray interaction with dense gas plumes

Numerical calculations are performed to reproduce the transport and dispersion of the continuous release of dense gases over flat homogeneous surfaces with and without the mitigating influence of a downwind water curtain. Frequently such plumes are released as a result of a chemical manufacturing, storage or gas transportation accident resulting in a ground-level hazard due to gas flammability or toxicity. A field situation in which cold carbon dioxide was released upwind of water curtains ( Moodie et al., 1981 ) was simulated using the open-source software FDS (Fire Dynamic Simulator) a full 3-d CFD model. Only water-spray enhancement of dispersion was considered; hence, no chemical removal or reactions were present or simulated. Wind-tunnel measurements for a 1:28.9 scale replication of the Moodie experiments are also compared with the 3-d CFD results. Concentration distributions, percent dilution and forced diffusion parameters were compared in scatter diagrams. Concentration field contours with and without active spray curtains are also presented. ► Model replicated dense gas dispersing with water-spray curtain mitigation devices. ► Cold carbon dioxide released upwind of water curtains was simulated using CFD. ► Wind-tunnel measurements were compared with the 3-d CFD results. ► Concentration, percent dilution and forced dilution were compared in diagrams.

Dense gas vertical diffusion over rough surfaces: results of wind-tunnel studies

A cooperative program of measurements of vertical diffusion of continuous, dense gas plumes over rough surfaces in neutral boundary layers has been carried out in three wind tunnels in the USA and the UK. The three environmental boundary layer tunnels were at the Chemical Hazards Research Center (CHRC) at the University of Arkansas, the Fluid Modeling Facility (FMF) of the US Environmental Protection Agency in North Carolina, and the Environmental Flow Research Centre (EnFlo) at the University of Surrey. A simple and consistent set of definitions was adopted for the plume variables like plume depth, mean plume transport speed, vertical entrainment velocity, w e, and plume Richardson number Ri ∗ , where Ri 1/2 ∗ is a ratio of buoyancy-induced flow velocities to u ∗ , the upstream-of-source ambient friction velocity. The present experiments focus on how Ri ∗ affects the ratio w e /u ∗ . In order to maintain nearly constant Ri ∗ in distance and time, continuous line sources of dense gas, primarily CO 2, were employed. Good agreement was found among the three tunnels. The results also agree with the classic Prairie grass field experiment for the “passive limit” ( Ri ∗ =0): w e /u ∗ =0.6–0.7. For Ri ∗ up to 20, the results fit the equation w e /u ∗ =0.65/(1+0.2 Ri ∗ ). For Ri ∗ >20, molecular diffusion and viscosity effects were apparently quite strong because w e was observed to collapse to values nearly commensurate with molecular diffusion alone.

A wind tunnel study of dense gas dispersion in a stable boundary layer over a rough surface

Measurements of the vertical entrainment velocity into two-dimensional dense gas plumes over fully rough surfaces were carried out as part of a co-operative research programme with wind tunnel facilities in the USA. This paper presents results obtained for stable boundary layer conditions in the EnFlo wind tunnel at the University of Surrey; a companion paper treats the neutral boundary layer case. Mean velocity and temperature, turbulent normal and shear tresses, temperature fluctuations and heat fluxes were measured and used to demonstrate that a moderately stable atmospheric boundary layer had been successfully simulated in the tunnel. Entrainment velocities, W E, were then deduced from the streamwise development of the concentration field, non-dimensionalised with respect to the friction velocity in the undisturbed flow, u *, and correlated with the plume Richardson number, Ri *. Higher non-dimensional entrainment speeds, W E/ u *, were observed for Ri *>5 in the stable boundary layer than in the neutral boundary layer, the difference growing with increasing Richardson number. Emission velocity ratios, W 0/ u *, were however larger in the stable experiments, and exceeded one at about Ri *=18. Entrainment in the stable boundary layer appeared therefore to be more sensitive to emission velocity ratio than in the neutral case. Entrainment behaviour for Ri *⩽5 followed that found in the neutral boundary layer. In this regime, use of the neutral boundary layer entrainment speed correlation is unlikely to lead to the over-prediction of plume dilution rates in moderately stable boundary layers.

Overview of Petroleum Environmental Research Forum (PERF) dense gas dispersion modeling project

This paper provides a background for and an overview of the results of a comprehensive study of transport and dispersion of dense gas plumes over rough surfaces typical of industrial sites. The Petroleum Environmental Research Forum (PERF) 93-16 project involved model development and evaluations using observations from three wind tunnels and from the Kit Fox field experiment. Detailed discussions of the results of the research are given in the other papers in this special issue. The wind tunnel experiments produced data showing that the resulting best-fit vertical entrainment formula was close to (i.e., within about 30%) the vertical entrainment formulas already in use by current models, which were derived primarily from observations over smooth surfaces. Observations from the Kit Fox field experiment demonstrated the validity of the entrainment curves derived from the wind tunnel data. The Kit Fox data were also used to evaluate algorithms for along-wind dispersion and cloud advection speeds for short-duration releases typical of an industrial site, and to evaluate the HEGADAS dense gas dispersion model.

Wind tunnel study of air entrainment into two-dimensional dense gas plumes at the Chemical Hazards Research Center

Experiments in a neutrally stable wind tunnel boundary layer were made for two-dimensional (quasi-line) sources of carbon dioxide dispersing over two types of uniformly spaced (billboard) surface roughness elements. Velocity and concentration measurements were made with each surface roughness over a wide range of source Richardson number by varying carbon dioxide release rate and wind speed. Concentration measurements were made with a FID gas analyzer using an ethane tracer in the source gas, and velocity measurements were made with independent LDV and HWA systems. For each surface roughness, this paper describes the wind tunnel boundary layer and presents alongwind and vertical concentration profiles in the gas plume. Vertical velocity and concentration profiles were measured at selected downwind distances, and the profiles were integrated to confirm the consistency of the measurements with the mass of carbon dioxide released. The data are intended for development of improved vertical turbulent entrainment correlations for use in dense gas dispersion models applied to hazardous chemical consequence analyses.

A wind tunnel study of dense gas dispersion in a neutral boundary layer over a rough surface

Measurements of the vertical entrainment velocity into two-dimensional dense gas plumes over fully rough surfaces were carried out as part of a co-operative research programme with wind tunnel facilities in the USA. This paper presents results obtained for neutral boundary layer conditions in the EnFlo wind tunnel at the University of Surrey; a companion paper treats the stable boundary layer case. Entrainment velocities, W E, were deduced from the streamwise development of the concentration field, non-dimensionalised with respect to the friction velocity in the undisturbed flow, u *, and correlated with the plume Richardson number, Ri *. Results for Richardson numbers in the range Ri *<15 were found to be well fitted by the empirical expression: W E/ u *=0.65/(1+0.2 Ri *). Flow visualisation studies showed layered plume structures with a sharp upper interface at higher Richardson numbers and in this regime turbulent motion below the interface became progressively more intermittent as Ri * increased. Measured turbulence levels collapsed within such high Richardson number plumes and flow and dispersion were significantly affected by molecular processes. Up-welling above the source was observed when the emission speed exceeded the approach flow friction velocity, though there was no clear evidence that this affected plume behaviour away from the immediate vicinity of the source.

Experimental Study of the Criteria of Flow Laminarization in Two-Dimensional Dense Gas Plumes

Laminarization of flow in a two-dimensional dense gas plume was experimentally investigated in this study. The plume was created by releasing CO2 through a ground-level line source into a simulated turbulent boundary layer over an aerodynamically rough surface in a meteorological wind tunnel. The bulk Richardson number (Ri*), based on negative plume buoyancy, plume thickness, and friction velocity, was varied over a wide range so that the effects of stable stratification on plume laminarization could be observed. A variety of ambient wind speeds as well as three different sizes of roughness arrays were used so that possible effects of roughness Reynolds number (Re*) on plume laminarization could also be identified. Both flow visualization methods and quantitative measurements of velocity and intermittency of turbulence were used to provide quantitative assessments of plume laminarization. Flow visualization provided an overall picture of how the plume was affected by the negative buoyancy. With increasing Ri*, both the plume depth and the vertical mixing were significantly suppressed, while upstream propagation of the plume from the source was enhanced. The most important feature of the flow revealed by visualization was the laminarization of flow in the lower part of the plume, which appeared to be closely related to both Ri* and Re*. Measurements within the simulated dense gas plumes revealed the influence of the stable stratification on mean velocity and turbulence intensity profiles. Both the mean velocity and turbulence intensity were significantly reduced near the surface; and these reductions systematically depended on Ri*. The roughness Reynolds number also had considerable influence on the mean flow and turbulence structure of the dense gas plumes. An intermittency analysis technique was developed and applied to the digitized instantaneous velocity signals. It not only confirmed the general flow picture within the dense plume indicated by the flow visualization, but also clearly demonstrated the changes of flow regime with variations in Ri* and Re*. Most importantly, based on this intermittency analysis, simple criteria for characterizing different flow regimes are formulated; these may be useful in predicting when plume laminarization might occur.

Lagrangian Dispersion Model for Nonneutrally Buoyant Plumes

A capability to address positive and negative buoyancy was added to the Higher-Order Turbulence Model for Atmospheric Circulation–Random Puff Transport and Diffusion (HOTMAC–RAPTAD) modeling system. The modeling system was applied to simulate dense gas plumes, and the modeled concentrations were compared with observations reported in the Modelers’ Data Archives (MDA). Sampling sites reported in MDA were located mostly 50–800 m from the source over flat terrain. To detect a peak concentration, RAPTAD sampling sites were placed on the arcs whose radii correspond to the sampling distance reported in MDA. Concentration averaging time for a peak concentration was varied from 1 to 600 s. RAPTAD simulation time varied from 4 to 30 min. The overall performance of the current model in terms of geometric mean biases, geometric variances, and residual plots was found to be at least as good as those of the better models examined previously with the same dataset.

An experimental study of the flow structure within a dense gas plume

A wind tunnel study was conducted to examine how a dense gas plume could affect the mean flow and turbulence structure of the boundary layer containing the plume. For this, a neutral atmospheric boundary layer developing over an aerodynamically rough surface was simulated in the US EPA's Meteorological Wind Tunnel. The dense gas plume was created by releasing CO 2 through a small circular source at ground level. A procedure was developed to make reasonably accurate mean velocity and turbulence measurements within the dense gas plumes by using hot-film anemometry in a range where the probe response was insensitive to the concentration of CO 2. Both the flow visualization and quantitative measurements of concentration and velocity fields indicated that, at low wind speeds, the dense gas plumes exhibited significant buoyancy effects on the flow structure. Within the dense plumes, mean velocity profiles were observed to have changed significantly in shape, with reduced speeds near the surface and increased velocities farther away from the surface. Consistent with these changes in mean velocity profiles, significant reductions in the roughness length and friction velocity were observed. Both the longitudinal and vertical turbulence intensities were also found to be greatly reduced within the dense plumes at low wind speeds. These changes in mean flow and turbulence structure were not only related to the dense-gas concentrations, but also to the local velocity gradients and the growth of the dense plumes with distance from the source. The local gradient Richardson number is found to be the most appropriate parameter for describing the changes in the mean flow and turbulence structure. Significant dense gas effects were observed when the Richardson number increased beyond its critical value (0.25) for the dynamic stability of a stratified flow. Our experimental results show that, in an existing turbulent flow, turbulence is not completely suppressed even when the gradient Richardson number exceeds one.

Negative buoyancy effects on the dispersion of continuous gas plumes downwind solid obstacles

An investigation of continuous dense gas plumes released from a surface area source and dispersing downwind solid obstacles is presented. Comparison with passive plumes released in the same experimental conditions enables us to emphasize the buoyancy effects. Some quantitative statistical results such as mean concentration, root-mean-square fluctuations, skewness and kurtosis are given. The experimental technique used, visualizations and digital image processing, allows us to characterize the geometrical evolution of the plumes.

An evaluation of validation procedures and test parameters for dense gas dispersion models

Abstract This paper reports the evaluation of dispersion models with respect to their ability to predict the dispersion of continuous dense gas plumes over flat terrain. The exercise has been carried out in order to derive guidelines for statistical model evaluation exercises. Therefore, the models selected for the evaluation are simple: the Britter and McQuaid model and three variations of the Gaussian Plume Model. Two different evaluation procedures have been used: one using short time-averaged concentrations paired in time and space and another using observed maximum concentrations at certain distances together with a measure of the plume width. The performance numbers and ranking of the models depend substantially on the data sets used and the evaluation procedure and performance number selected. The two evaluation procedures have both advantages and disadvantages. An evaluation based on maximum concentration should not be performed without taking into account the plume dimensions. The evaluation using data paired in time and space suffers from the inability of the models and the data analysis techniques to match the predicted plume path with the actual plume path during the experiment. We suggest that two newly defined performance measures be used, the Mean Relative Bias and the Mean Relative Square Error, together with the well-known ‘Factor-of-2’ statistic in future statistical evaluation exercises. One should use at least two performance measures in an evaluation exercise in order to obtain a measure of ‘bias’ and ‘variance’. Furthermore, we advise that graphical presentations to obtain an understanding of the model behaviour be used.

Fluid modeling of dense gas dispersion over a ramp

The basic nature of the transport and dispersion of a dense gas plume in the simulated neutral atmospheric boundary layer of a wind tunnel was investigated, both in flat terrain and over a ramp. Measurements were made of the concentration fields downstream of a ground-level, circular source; these measurements consisted of longitudinal ground-level, vertical, and crosswind profiles at various distances downwind. Both neutrally buoyant (air) and negatively buoyant (CO 2) source gases were used so that the specific effects of the density difference could be observed. Similarly, measurements were made in both flat terrain and over the ramp (14° slope followed by an elevated plateau) so that specific effects of the terrain could be observed. Flow visualization was done to ascertain that the dense plume was turbulent, hence, that the effects of molecular properties were insignificant. For the particular value of the buoyancy parameter used in these experiments, the plume buoyancy was significant; the resulting dense plume was significantly wider in the lateral direction and much narrower in the vertical direction, yet the longitudinal ground-level concentration profile downwind was essentially identical to that from the neutral plume. The lateral concentration profiles of the neutral plumes were essentially Gaussian in character, whereas the dense gas plumes exhibited top-hat distributions for considerable distances from the source. The vertical concentration profiles of the neutral plumes were not Gaussian, but displayed variations of the form C/ C mx = exp (− Az n ) with n ⋍ 1.5. On the other hand, the dense gas plumes displayed vertical variations of the form C/ C mx = exp [− z/ z ], where z is the centroid of the distribution. The net effect of the ramp on the dense gas plume was a small reduction in ground-level concentration (less than a factor of two, even for a source relatively close to the base of the ramp). This reduction was quite similar to that observed for the neutral plume.

Properties of industrial dense gas plumes

Hazardous gases and vapors are often discharged into the atmosphere from industrial plants during catastrophic events (e.g. Union Carbide incident in Bhopal, India). In many cases the discharged components are more dense than air and settle to the ground surface downstream from the stack exit. In the present paper, the buoyant plume model of Hoult, Fay and Forney (1969, J. Air Pollut. Control Ass. 19, 585–590.) has been altered to predict the properties of hazardous discharges. In particular, the plume impingement point, radius and concentration are predicted for typical stack exit conditions, wind speeds and temperature profiles. Asymptotic expressions for plume properties at the impingement point are also derived for a constant crosswind and neutral temperature profile. These formulae are shown to be useful for all conditions.

Integral model of dense gas plume dispersion

The injection of a dense gas stream at ground level into a flowing turbulent atmosphere produces a wide, flat plume that entrains air primarily through its upper surface. A quasi-one-dimensional flow model of an isothermal dense gas plume is developed for the purpose of comparing experiments in wind tunnels and water flumes and field tests in the atmosphere. Comparisons are made for plume width, including the width at the source, and centerline ground plane source gas concentration. All published data are used in this comparison, which cover a factor of about 100 in plume length scale and Reynolds number. Tests conducted by different experimenters were found not to be dynamically similar. Dimensionless model parameters, all of order unity, are selected to give the best agreement among all the experimental data. The dependence of entrainment rate on the plume Richardson number, a key feature of the model, is confirmed in the comparison. The entrainment rate parameter is found to be larger for the field tests than for the laboratory experiments, reflecting the much higher Reynolds number of the former.