Evaluation of Lagrangian Particle Dispersion Models with Measurements from Controlled Tracer Releases


Hegarty, J., R.R. Draxler, A.F. Stein, J. Brioude, M. Mountain, J. Eluszkiewicz, T. Nehrkorn, F. Ngan, A. Andrews


Journal of Applied Meteorology and Climatology, Vol. 52, No. 12, pp. 2623-2637, 2013, doi:10.1175/JAMC-D-13-0125.1


Abstract - Three widely used Lagrangian particle dispersion models (LPDMs)—the Hybrid Single-Particle Lagrangian Integrated Trajectory (HYSPLIT), Stochastic Time-Inverted Lagrangian Transport (STILT), and Flexible Particle (FLEXPART) models—are evaluated with measurements from the controlled tracerrelease experiments Cross-Appalachian Tracer Experiment (CAPTEX) and Across North America Tracer Experiment (ANATEX). The LPDMs are run forward in time driven by identical meteorological inputs from the North American Regional Reanalysis (NARR) and several configurations of the Weather Research and Forecasting (WRF) model, and the simulations of tracer concentrations are evaluated against the measurements with a ranking procedure derived from the combination of four statistical parameters. The statistical evaluation reveals that all three LPDMs have comparable skill in simulating the tracer plumes when driven by the same meteorological inputs, indicating that the differences in their formulations play a secondary role. Simulations with HYSPLIT and STILT demonstrate the benefit of using customized hourly WRF fields with 10-km horizontal grid spacing over the use of 3-hourly NARR fields with 32-km horizontal grid spacing. All three LPDMs perform better when the WRF wind fields in the planetary boundary layer are nudged to NARR, with FLEXPART benefitting the most. Case studies indicate that the nudging corrects an overestimate in plume transport speed possibly caused by a positive bias inWRFwind speeds near the surface. All three LPDMs also benefit from the use of time-averaged velocity and convective mass flux fields generated by WRF, but the impact on HYSPLIT and STILT is much greater than on FLEXPART. STILT backward runs perform as well as their forward counterparts, demonstrating this model’s reversibility and its suitability for application to inverse flux estimates.


Email me the full report