The representation of soil moisture freezing and its impact on the stable boundary layer

The 1993 to 1996 version of the European Centre for Medium-Range Weather Fo recasts model had a pronounced near-surface cold bias in winter over contin ental areas. The problem is illustrated in detail with help of tower observ ations. It is shown that a positive feedback exists in the land surface bou ndary-layer coupling that has the potential to amplify model biases. If the surface is cooled too much the boundary layer becomes too stable, reducing the downward heat flux and making the surface even colder. This positive f eedback is believed to be stronger in the model than in the real atmosphere , resulting in diurnal temperature cycles that are too large and in excessi ve soil cooling on a seasonal time-scale in winter. An important contributor to the excessive winter cooling turns out to be th e lack of soil moisture freezing in the model. The importance of this proce ss is obvious from soil temperature observations. The seasonal soil tempera ture evolution shows a clear 'barrier' at 0 degrees C due to the thermal in ertia of freezing and thawing. A more quantitative illustration is the resu lt of a simple calculation. This shows that the amount of energy necessary to freeze/thaw 1 m(3) Of wet soil, would cool/warm this soil by about 50 K if the phase transition was not taken into account. To reduce the winter cold bias in the model, three model changes have been tested and are described: (i) the introduction of the process of soil moist ure freezing; (ii) revised stability functions to increase the turbulent di ffusion of heat in stable situations; and (iii) an increase of the skin-lay er conductivity. The effect of these changes on the seasonal evolution of s oil and 2 m temperatures is investigated with long runs that have terms tha t relax towards the operational analysis above the boundary layer. In this way the impact can be studied on the temperature forecasts for the winter o f 1995/1996, during which the operational model showed considerable soil te mperature drift over Europe. Also, short periods of data assimilation (incl uding 10-day forecasts) have been carried out to study the diurnal time-sca les and the impact on model performance. The model changes eliminate to a large extent the systematic 2 m temperatur e biases for the winter of 1995/1996 over Europe and make the soil temperat ure evolution much more realistic. The soil moisture freezing, in particula r, plays a crucial role by introducing thermal inertia near the freezing po int, thereby reducing the annual temperature cycle in the soil. The process of soil moisture freezing leads to a considerable warming of the model's n ear-surface winter climate over continental areas.