A Comparative Study between FLEXPART-WRF and HYSPLIT in an Operational Setting: Analysis of Fire Emissions across complex geography using WRF

Transport and dispersion (T&D) models are frequently used by the meteorological community to understand and predict the trajectories of anthropogenic, natural and accidental chemical releases of hazardous materials. There are several reputable T&D models that can handle a wide range of applications under the direction of global, synoptic or mesoscale forecasts. One such application is the forecast of smoke emissions from wildfires which is important to operational air quality and meteorology communities.

Fire emissions have direct impacts to property and respiratory health. The prediction of smoke plumes from wildfires using dispersion models is an important challenge for meteorologists and fire weather specialists. Meteorologists at local National Weather Service (NWS) forecast offices are responsible for providing meteorological support to emergency management agencies within their county warning area in the event of incidents involving harmful chemical releases, radiation and smoke emissions. It is important to understand the strengths and weaknesses of the available dispersion models under critical conditions. A comparative study between two dispersion models during recent wildfire events across complex geography is presented to identify the sensitivities of each dispersion model and the operational benefits of utilizing each model for smoke emission forecasts.

FLEXPART-WRF is a Lagrangian dispersion model that predicts the transport, diffusion and deposition of trace gases forward or backward from a point, line or area source. This open source model allows users the flexibility to manipulate the code according to the application purpose. Similar to FLEXPART-WRF, the HYbrid Single Particle Integrated Trajectory (HYSPLIT) model simulates the dispersive nature of the environment. HYSPLIT was built as an operational tool and is frequently used by the National Oceanic and Atmospheric Administration in collaboration with the Air Research Laboratory. Unlike FLEXPART-WRF, HYSPLIT is frequently used as a graphical interface, without easy access to the underlying code. Other model configuration differences include the prerequisite meteorological data, density correction and dispersion algorithms and removal calculations.

Mesoscale models are needed to provide the ambient environment as well as simulate the small scale flux exchanges and boundary layer eddies that can affect dispersion simulations on a local and regional scale. Also, the subgrid source releases are handled more appropriately with higher resolution data as compared to coarse input from synoptic and global models. Therefore, both dispersion models were run using meteorological data from the WRF ARW model.

Two fire events, one along the coast of the Mid-Atlantic and the other within the Appalachians, were investigated for this analysis. A control simulation is analyzed to identify the relative performance of each dispersion model given identical meteorological input. The dispersion models are evaluated for accurate dispersive simulations and also on their ability to support operational forecast needs for NWS local forecast offices. In contrast, a series of sensitivity studies are performed to identify and evaluate the impact of changes in WRF boundary layer and land use physics to the dispersion simulations. Satellite observations provided by the National Environmental Satellite, Data and Information Service along with other radar and remote sensing tools are used for evaluation of dispersion model performance.