Sensitivity of lake-effect snow forecasts to the choice of boundary layer parameterization

The sensitivity of forecasts to the choice of boundary layer parameterization scheme is examined for lake-effect snow. Six different schemes are tested. Results show that precipitation rates within the band can differ by as much as 4 mm hr$^{-1}$ of liquid equivalent. Consideration of equations used in the various schemes reveals there are three key differences that contribute to the different precipitation rates - these are the way in which the surface exchange coefficients for heat and moisture, the turbulent kinetic energy on the lowest model layer, and u* are calculated. The combination of equations in the quasi-normal scale elimination (QNSE) scheme yields a maximum 6-h liquid-equivalent accumulation of 35.79 mm, while the Mellor-Yamada-Nakanishi-Niino (MYNN) scheme has a maximum of 14.2 mm for the same event. Adjusting the formulae used in either algorithm substantially alters the accumulated precipitation. To gauge whether these results have generality, set of quasi-idealized experiments are conducted wherein the control parameters dictating the intensity and form of precipitation are systematically varied. These experiments show that when the low-level wind is oriented perpendicular to the long axis of the lake, all schemes produce similar amounts and patterns of precipitation. As the flow becomes more favorable for intense, single-banded forms, differences between the schemes become more apparent. At the time of this writing, it is unclear which scheme consistently produces the most accurate results.