Ensemble-based data assimilation systems appear to be reaching maturity for large-scale flows. Research is also progressing on assimilation of radar data into mesoscale domains, but ensemble-based assimilation of surface observations and PBL profiles has received little attention. These data are unique because they are strongly coupled to both the land surface and the atmosphere. Model error may be large, and little is known about error-growth properties of the PBL, making research toward the effective use of those observations in ensemble data assimilation necessary. This talk describes one aspect of on-going work: the ability of in-situ PBL and surface observations to reduce uncertainty in PBL flows.

Observation-space (innovation) and grid diagnostics are used to determine the potential value of observations to an ensemble forecast over complex terrain with a grid spacing of 3.3 km. Analysis focuses on how forecast uncertainty is reduced at the observation locations and on the grid. When uncertainty is decreased, the resulting information content is tracked forward in time to understand its contribution to the ensuing free forecast. Recent results will be described, including those from profiles and screen-height observations. Observations at different heights above ground, of different variables, and at different times of day will be examined. These results will lead to a better understanding of how to optimize ensemble data assimilation systems for nowcasts and forecasts near the Earth's surface.