Date of this Version
Accuracy in snowfall prediction has lagged behind other short-term weather forecasting areas. Errors in quantitative precipitation forecasts ensure that any snow ratio applied to snow may result in inaccurate snowfall amounts, and snowfall observations are not consistent or fully reliable. In this study, the Cobb Method is tested on lake-effect snowfalls to determine if the top-down ice crystal growth modeled in the algorithm can be applied to convective snowfalls. To establish the spatiotemporal and physical characteristics of lake-effect snowfalls at selected study locations near the Great Lakes, snowfall and snow ratio climatologies are produced that separate events by lake-effect and non-lake-effect snowfall type. Lake-effect snowfalls occur most frequently at all locations in December and January, and progressing from November to March there is a decreasing proportion of lake-effect to all snowfalls from around 0.6 to near 0.1. With respect to non-lake-effect snowfalls, snow ratios of lake-effect snowfalls are higher and more variable. For snowfalls calculated by the Cobb Method, lake-effect and non-lake-effect snowfalls are 60.6% and 63.7% accurate compared to observations, respectively. Adding an empirical compaction factor improves the non-lake-effect events by 4.0% and worsens lake-effect snowfalls by 8.4%, which reflects a bias towards underforecasted snowfalls for lake-effect snow of all snow amounts. Snow ratios of lake-effect snowfalls also have higher mean and variance than non-lake-effect snowfalls; however, snow ratios are lower in magnitude than for observations. The results of this study show that the Cobb Method may be applied to lake-effect snow forecasting with the knowledge that the snow ratios produced on average will be lower than what is observed, therefore snowfalls will be greater. Events that are depicted well by a numerical prediction model with the knowledge that snow ratios are too low and used by the forecaster will be associated with more accurate snowfalls. Use of a high-resolution model that resolves mesoscale processes is also an important consideration, since lake-effect snowfall is a mesoscale process.
Adviser: Mark Anderson