Automated source term and wind parameter estimation for atmospheric transport and dispersion applications
Paul E. Bieringer, Luna M. Rodriguez, Francois Vandenberghe, Jonathan G. Hurst, George Bieberbach Jr., Ian Sykes, John R. Hannan, Jake Zaragoza, Richard N. Fry Jr.
Atmospheric Environment, 2015, Volume 122, pp 206–219

Accurate simulations of the atmospheric transport and dispersion (AT&D) of hazardous airborne materials rely heavily on the source term parameters necessary to characterize the initial release and meteorological conditions that drive the downwind dispersion. In many cases the source parameters are not known and consequently based on rudimentary assumptions. This is particularly true of accidental releases and the intentional releases associated with terrorist incidents. When available, meteorological observations are often not representative of the conditions at the location of the release and the use of these non-representative meteorological conditions can result in significant errors in the hazard assessments downwind of the sensors, even when the other source parameters are accurately characterized. Here, we describe a computationally efficient methodology to characterize both the release source parameters and the low-level winds (eg. winds near the surface) required to produce a refined downwind hazard. This methodology, known as the Variational Iterative Refinement Source Term Estimation (STE) Algorithm (VIRSA), consists of a combination of modeling systems. These systems include a back-trajectory based source inversion method, a forward Gaussian puff dispersion model, a variational refinement algorithm that uses both a simple forward AT&D model that is a surrogate for the more complex Gaussian puff model and a formal adjoint of this surrogate model. The back-trajectory based method is used to calculate a “first guess” source estimate based on the available observations of the airborne contaminant plume and atmospheric conditions. The variational refinement algorithm is then used to iteratively refine the first guess STE parameters and meteorological variables. The algorithm has been evaluated across a wide range of scenarios of varying complexity. It has been shown to improve the source parameters for location by several hundred percent (normalized by the distance from source to the closest sampler), and improve mass estimates by several orders of magnitude. Furthermore, it also has the ability to operate in scenarios with inconsistencies between the wind and airborne contaminant sensor observations and adjust the wind to provide a better match between the hazard prediction and the observations.

Reactive puff model SCICHEM: Model enhancements and performance studies
B. Chowdhury, P.K. Karamchandani, R.I. Sykes, D.S. Henn, E. Knipping
Atmospheric Environment, 2015, Volume 117, pp 242–258

The SCICHEM model incorporates complete gas phase, aqueous and aerosol phase chemistry within a state-of-the-science Gaussian puff model SCIPUFF (Second-order Closure Integrated Puff). The model is a valuable tool that can be used to calculate the impacts of a single source or a small number of sources on downwind ozone and PM2.5. The model has flexible data requirements: it can be run with routine surface and upper air observations or with prognostic meteorological model outputs and source emissions are specified in a simple text format. This paper describes significant advances to the dispersion and chemistry components of the model in the latest release, SCICHEM 3.0. Some of the major advancements include modeling of skewed turbulence for convective boundary layer and updated chemistry schemes (CB05 gas phase chemical mechanism; AERO5 aerosol and aqueous modules). The results from SCICHEM 3.0 are compared with observations from a tracer study as well as aircraft measurements of reactive species in power plant plumes from two field studies. The results with the tracer experiment (Copenhagen study) show that the incorporation of skewed turbulence improves the calculation of tracer dispersion and transport. The comparisons with the Cumberland and Dolet Hills power plume measurements show good correlation between the observed and predicted concentrations of reactive gaseous species at most downwind distances from the source.

Development and evaluation of a state-of-the-science reactive plume model
Karamchandani, P., L. Santos, I. Sykes, Y. Zhang, C. Tonne and C. Seigneur
Environmental Science & Technology, 2000, 34 (5), pp 870-880

We describe the development and evaluation of a new reactive plume model that combines a state-of-the-science puff model with an optimized chemistry model that accurately represents the chemistry of a power plant plume at various stages of its evolution. The puff model uses a second-order closure scheme, allowing for an accurate treatment of dispersion and the influence of turbulent concentration fluctuations on chemical rates. The model was tested using helicopter plume measurements from the 1995 Southern Oxidants Study (SOS) Nashville/Middle Tennessee Ozone Study. The model was applied for 6 days in June and July of 1995, and the model’s ability to estimate physical and chemical plume characteristics, such as plume width and reactive species concentrations, was evaluated using the helicopter measurements. The best model results are for July 7, 1995, a case corresponding to a high NOx isolated power plant plume traveling over rural regionsmodel estimates of NOx, NOy, and O3 are highly correlated with measured values, and most of the measured plume centerline concentrations are reproduced to within 30%. For scenarios involving the interaction of the tracked plume with urban plumes or with other power plant plumes, model estimates of ozone concentrations are poorly correlated with observations, emphasizing the difficulty of characterizing such plumes from both measurement and modeling perspectives.

The representation of dynamic flow effects in a Lagrangian puff dispersion model
Sykes, R. I., C. P. Cerasoli and D. S. Henn
Journal of Hazardous Materials, 1999, Volume 64, Issue 3, pp 223–247

A technique for incorporating buoyancy-driven dynamics in a Lagrangian puff model is described. The dynamic effects are non-linear and therefore proper treatment requires interaction integrals for overlapping puffs. Conservation laws for the volume integral of momentum and buoyancy over an individual puff are based on the fundamental equations of motion. A simplified representation of the field of motion associated with the buoyancy-driven dynamics is then used to move and distort the puffs. The effects associated with dense gas `slumping’ on the ground are represented by lateral divergence of the velocity field, with a magnitude based on conservation of the moment of vorticity. The model predictions are compared with a number of experimental results on buoyant plume rise and dense gas dispersion.