Multiscale weather forecasting sensitivities to urban characteristics and atmospheric conditions during a cold front passage over the Dallas-Fort Worth metroplex

Sensitivities of microscale weather modeling to atmospheric conditions and urban layout are investigated utilizing a combination of automated surface observing systems (ASOS) data, 1-km mesoscale numerical weather prediction (NWP), and 5-m nested large-eddy simulation (LES) modeled conditions. The 1-km mesoscale predictions in analysis mode satisfactorily reproduce the observed spatiotemporal evolution of the frontal boundary in terms of wind speed, wind direction, and temperature. The 5-m nested LES simulations follow the large-scale forcing trends while improving wind speed predictions due to explicitly resolving turbulence and building interactions. Moreover, 5-min averaged nested LES results reveal improved temporal variability particularly during the stronger wind and turbulence post-frontal conditions. The skill of the 1-km mesoscale NWP model prediction is compared to coarse-grained LES fields. Probability distributions extracted from the 5-m nested LES predictions exhibit the largest sensitivity to the contrasting meteorological conditions. In contrast, cumulative distributions of TKE additionally expose a marked dependency on the unique distribution of building heights, urban density and clustering in a given area. For the first time, an ensemble forecast methodological design at building-resolving grid spacing is explored. A larger microscale ensemble spread is found for TKE than for wind speed, decreasing with height and modulated by weather conditions.