Because of the important impact of surface and soil conditions (especially soil moisture) on atmospheric model predictions, great efforts have been given at the Meteorological Service of Canada to improve the initialization and assimilation of surface characteristics. In September 2001, a new surface modelling system was implemented in the Canadian short-range regional weather forecasting system, in which a soil moisture sequential assimilation technique, based on an error feedback mechanism, has been included. The inclusion of this new surface system led to a significant improvement of the model representation of precipitation and of the low-level diurnal cycle.

We are now in the process of developping another surface assimilation system, which could eventually replace the "in-line" sequential assimilation strategy that was implemented in 2001. This new system will be based on off-line integrations of the landsurface model. Based on this idea, two parallel assimilation systems have been integrated from mid-April to the end of August 2002. In the first of these two parallel systems, the atmospheric forcing is entirely provided by outputs from the Canadian regional atmospheric model. In the second system, the precipitation reaching the surface is obtained from analyses derived from radars. Comparison between the results produced by these two systems and the "in-line" sequential assimilation system that is currently operational will be presented at the conference.