Evaluation of MM5 Simulations With HTSVS With and Without Inclusion of Soil-Frost Parameterization

Abstract Permafrost and seasonally frozen ground are important surface features in highlatitudes. Because of this, a soil-frost parameterization was added to the Penn State University/NCAR mesoscale meteorological model MM5 in combination with the well validated hydro-thermodynamic soil vegetation scheme HTSVS, which takes into account among other things soil freezing and thawing. Reanalysis of temperature, wind vector, specific humidity and observations of precipitation were used to evaluate the importance of the soil-frost parameterization on the forecast for an episode in summer. R.m.s. errors for both forecasts are reasonable with the largest errors coming from the precipitation data. Improvements in r.m.s. error for the forecast with the soil-frost parameterization added are mostly in the atmospheric boundary layer. Just as for r.m.s. error, the mean error for the soil-frost forecast is closer to zero than the mean error for the forecast made without the soil-frost parameterization. The consistent negative bias of the mean error is also improved by the inclusion of soil-frost. For precipitation, the wind vector, and in the mid-troposphere for temperature, the improvement index of the forecasts decreases as time increases, meaning that the inclusion of soil-frost makes the model more accurate especially as it is projected farther into the future. However, for the atmospheric boundary layer temperatures and for specific humidity the improvement index actually increases with time, meaning that the addition of a soil-frost parameterization for these variables does not continue to improve the longer range forecast as much as it does for the other variables. Both the bias and threat scores decrease as the precipitation threshold increases and generally increase with time. In addition, according to the threat scores, inclusion of soil-frost processes improves the prediction of moderate precipitation. In general, it is shown that with the inclusion of a soil-frost parameter in the model, forecast errors are decreased.