It is well known that latent heat absorption or release associated with melting and freezing precipitation can modify the atmospheric temperature profile sufficiently to affect precipitation type at the surface. What is less well known is the extent to which current operational numerical weather prediction (NWP) models are capable of representing these thermodynamic processes.

When raindrops fall into a layer of sub-freezing air near the surface, they may re-freeze completely before reaching the ground as sleet (a.k.a. ice pellets). This process releases latent heat and warms the layer in which the freezing occurs. Until 27 November, the NCEP Eta model did not account for this warming mechanism; with the upgrade implemented beginning with the 12 UTC run on 27 November 2001, the Eta now accounts for this process. However during freezing rain, where the latent heat is released at the surface, the Eta model (and most other operational models) representation of this process is problematic. Here, the Eta land-surface model (LSM) must determine the phase of hydrometeors reaching the surface. The current Eta LSM determines precipitation type based solely on the lowest model air temperature. If the lowest air temperature is above (below) freezing, the LSM assumes that rain (snow) is reaching the surface. During freezing rain events (assuming that the near-surface air temperature is below freezing), the Eta LSM will determine that snow is reaching the surface, and will not account for the latent heat release. A substantial cold bias may develop in near-surface model temperature forecasts due to i.) the neglect of latent heat released by freezing rain, and ii.) the development of a spurious snow cover in the model (which in turn absorbs heat if melted, modifies the surface albedo, and may insulate the lower atmosphere from upward heat flux from the soil in some situations). A case study example, from 12 February 2001, is provided to illustrate this scenario.