Abstract
Linear screening regression is used to derive relationships between parameters computed from observed upper air soundings (RAOBS) and concurrent observations of precipitation type. Precipitation type is defined as three categories: liquid (rain or drizzle), freezing (freezing rain or freezing drizzle) and frozen (snow or ice pellets). Statistical screening results indicate that of the parameters tried the following are important: the mean temperature in the surface–1000 m and 500–2500 m layers; the depth of the warm layer (temperature >0°C), if one exists; the area between the temperature profile and the 0°C isotherm in the warm layer, the depth of the surface-based cold layer, if one exists, with respect to the wet-bulb temperature profile; and the area between the wet-bulb temperature profile and the 0°C isotherm in the surface-based cold layer.
Verification of the specification equations on both developmental and independent data samples indicates that the scores are generally stable. The equations show excellent discrimination ability for liquid and frozen precipitation but have some difficulty with freezing precipitation.
Part of the problem with the freezing category is the fact that freezing drizzle, which is included with freezing rain in this category, can occur with a RAOB in which the temperature is ≤0°C at all levels (no warm layer). It is found that about 44% of the freezing drizzle RAOBs examined have no warm layer.
