Predictions of plume dispersion in complex terrain: Eulerian versus Lagrangian models

Simulations of dispersion from an elevated point source in complex terrain and non-stationary flow are presented using the Lagrangian atmospheric dispersion model (LADM, Physick et al., 1992, Air Pollution Modeling and its Applications, Vol. IX, pp. 725-729, Plenum Press, New York; 1994 CSIRO Division of Atmospheric Research Technical Paper No. 24) and the Eulerian grid-based model (CALGRID, Yamartino ct al., 1989, CALGIRD: a mesoscale photochemical grid model, Vol, I; model formation document, Report, Sacarmento, California), Both models use the same predicted windfields. We find that the different algorithms used for release of pollutants into the model domains lead to initial concentrations at the release height in LADM one-third higher than in CALGRID. The CALGRID plume spreads laterally over a larger region than does the LADM plume due to the finite-difference approach of CALGRID. The pollutant mass in the extra volume occupied by the CALGRID plume is less than 10% of that released. The essentials of morning fumigation are simulated more realistically under the Lagrangian approach. In LADM the elevated plume is mixed down to the ground rapidly, causing a sharp increase in ground-level concentrations (glc), whereas, in CALGRID glc increase more gradually over a few hours. The use of hourly averaged windfields in CALGRID compared to 10 min windfields in LADM leads to a relative separation of the two modelled plumes of 5 km at a distance of roughly 6 km downwind from the sources at 1500 LST. Consequently in complex terrain and non-stationary conditions, the plumes are subjected to different three-dimensional wind regimes. For the particular terrain studied, roughly 3% of the pollutant mass emitted into CALGRID during the day is transported above 3000 m after 1700 LST whereas only 0.8% is transported above this height in LADM. During the daytime the CALGRID simulation produces maximum glc which are about 40% smaller than those predicted by LADM. Copyright (C) 1997 Elsevier Science Ltd.