Using NEXRAD wind retrievals as input to atmospheric dispersion models

Atmospheric dispersion models (ADMs) are used to address the consequences of potential airborne releases of harmful materials. The wind fields in ADMs are usually based on meteorological measurements and forecasts made by mesoscale models. In the U.S., routine meteorological measurements from surface stations, towers, rawinsondes, and radar wind profilers are available to drive ADMs. Large metropolitan areas usually have a network of surface stations; however, the wind speed and direction measurements from these stations are usually not representative of conditions above the surface. There are currently relatively few measurements that can characterize the winds above the surface. One potentially rich, yet untapped, source of meteorological data for routine use in ADMs is from the Next Generation Weather Radar (NEXRAD) systems with 141 nearly identical WSR-88D Doppler radar units installed throughout the United States.

This study evaluates the computational efficiency and accuracy of two significantly different variational mathematical techniques that derive the u- and v-components of the wind from radial velocities obtained from Doppler radars. We designed a series of numerical experiments in which both models employed the same horizontal domain and resolution encompassing Oklahoma City for a two-week period during the summer of 2003 so that the computed wind retrievals could be fairly compared. Both models ran faster than real-time on one typical dual-processor computer, indicating that they could be used to generate wind retrievals in near real-time. The accuracy of the wind retrievals is quantified by comparing the derived wind components with independent, meteorological measurements made by radar wind profilers. The effect of the additional wind information aloft on transport and dispersion is examined by performing a series of sensitivity simulations using widely used ADMs (CALMET/CALPUFF, and SCIPUFF). Finally, we weigh the advantages and disadvantages of both 2DVar and VDRAS based on quantitative, operational, and system capability considerations.