Verification and sensitivity experiments for the WISP94 MM5 forecasts

A statistical verification of real-time forecasts from the Pennsylvania State University-National Center for Atmospheric Research Mesoscale Model version 5 (MM5) examines several model biases noted in numerical forecasts prepared for the Winter Icing and Storms Project's 1994 field experiment. Verification of MM5 forecasts against satellite and radiosonde data reveals a strong cloudy bias in the mid- to upper troposphere, significant moist biases aloft and near the surface, and a deep cold bias through much of the troposphere. The cloudy bias and upper-level moist bias are traced to an inappropriate assumption in the microphysical parameterization. Simple changes to the parameterization greatly improve the cloud forecast. A portion of the deep cold bias is attributed to the simple parameterization of atmospheric radiation used for the forecasts. The low-level cold and moist biases are in large part due to the climatological values of soil moisture availability as a function of land-use category. Experiments with a one-dimensional column model further quantify the sensitivity of low-level temperatures to the soil moisture availability values. While an immediate improvement in model results can be achieved by selection of more appropriate values of moisture availability, ultimately a detailed initialization and parameterization of soil moisture is needed.