A Numerical Study of the East Coast Snowstorm of 10–12 February 1983

Abstract

A numerical study of the East Coast snowstorm of 10–12 February 1983 has been conducted with the NRL mesoscale model. The three-dimensional, hydrostatic, primitive equation model has 91 × 51 horizontal grid points with a half degree resolution in a verification domain of 100°W to 60°W and 25°N to 45°N. There are ten layers in the vertical of equal σ(=p/ps) thickness. The model uses a split-explicit method for temporal integration and a second-order accurate spatial finite differencing. Model physics include precipitation on the resolvable scale and parameterized boundary layer and cumulus convection. The NMC 2.5 degree hemispheric analyses are used as the basic dataset (called NMC analysis hereafter). Because significant details in the initial conditions contained in the original rawinsonde observations may have been smoothed out by the NMC analysis algorithm, the analyses are also altered to provide a closer fit to the rawinsonde data. Original rawinsonde data are used in this enhancement of the NMC analyses (called enhanced). Forecasts are made from both the NMC analysis and the enhanced analyses to determine whether the enhancement can improve the forecasts. The Barnes analysis scheme with parameters suitable for reducing the short wavelength noise on the model grid scale is used to enhance the NMC analyses with the original soundings. Three types of lateral boundary treatments—constant, tendency damping, and temporal relaxation boundary conditions—are tested and compared. Results from forecast experiments show that the boundary treatment has a great impact on the model performance. The constant boundary condition produces an unusable forecast after 12 h as judged by the S1 scores, while the relaxation boundary condition produces an excellent forecast. The enhancement of the initial conditions has a negligible effect on predictions when reasonable boundary updates are used for the snowstorm case. The enhanced dataset produces a slightly better but still useless forecast when constant boundary conditions are used. Numerical experiments have also been conducted to test the sensitivity of the cyclogenesis to physical processes by suppressing one or more physical processes in the model. It is found that the evaporation from the ocean modulates the location and amount of precipitation. Without the evaporation, the-intensity of the cyclone remains the same but the center stays on the coast instead of staying off shore. The track of the snowstorm is such that the sensible beating from the ocean dampens the development of the cyclone by reducing the low-level baroclinicity. Without the sensible heating, the minimum surface pressure of the cyclone is 11 mb lower. The latent heating is found to be important for this case in which the maximum beating rate is 15°–20°C/day. When latent heating is suppressed, the cyclone translates at a much reduced rate and its central pressure is 10 mb higher after two days of simulation. These results from the sensitivity tests are of course case-dependent.