A sub-km-grid mini-ensemble for representing meso-gamma hazard-prediction uncertainty in complex terrain

Certain wind regimes in combination with complex terrain may excite mountain wave activity that can influence air quality and atmospheric transport and dispersion (AT&D), and model sensitivities contribute to hazard-prediction uncertainties. A sub-km-grid mini-ensemble (horizontal grid length of 444 m on innermost nest) for evaluating AT&D uncertainty of hazardous materials in the stable boundary layer (SBL) over complex terrain is described. The four to twelve member mini-ensemble is derived using diversity in initial conditions as well as surface-layer/planetary boundary layer (PBL) turbulence physics schemes within the Weather Research and Forecasting (WRF) model. The Advanced Research WRF (WRF-ARW) is centered over the Nittany Valley in central Pennsylvania. Outputs include explicit ensemble members as well as the ensemble mean and “best member”. The best member, for example, is defined here as the member with low-level winds closest to that of the ensemble mean. These WRF output data are used as input for trajectory calculations and the Second Order Integrated Puff (SCIPUFF) AT&D model for hazard prediction. Explicit mini-ensemble results (an aggregate of dispersion results from the explicit members) are compared to those derived from a single-member approach (ensemble mean or best member) that is capable of utilizing ensemble-based wind field uncertainty in SCIPUFF hazard mode. Hazard mode employs the ensemble-based wind field uncertainty, U wind variance (UUE), V wind variance (VVE), and UV wind covariance (UVE), to compute plume coverage, depth, and orientation. Case study results demonstrate that the ensemble mean appears capable of retaining some of the gravity-wave motions evident in explicit mini-ensemble members but the averaging smooths out some of the details such as rotor circulations. Additionally, in very light wind, stable conditions where the differences in ensemble-member wind directions may be very large, the ensemble mean may be non-representative due to cancellation of individual ensemble member characteristics. A best member approach, on the other hand, is dynamically consistent and can retain both gravity waves and rotor circulations important for surface dosages in rotor wind reversal regions. Also, when used in a single-SCIPUFF hazard prediction with the WRF-ensemble velocity statistics, the mean and best member capture nearly all points of non-zero surface dosage probability within the explicit ensemble and thus appear to reflect the spatial spread of the explicit mini-ensemble. In this case-study demonstration at these sub-km grid scales, a single SCIPUFF approach appears advantageous under certain conditions compared to a more time consuming and computationally costly explicit SCIPUFF ensemble approach. In regions of complex terrain, a single best-member SCIPUFF may prove to be more attractive for retaining wave-motion details over fine grid scales and capturing the spread of an explicit ensemble via the ensemble wind field uncertainty information.