Previous experiments had shown that upper-level wind flows could be used to advect surface weather parameters to produce short-range (0-15 hours) forecasts. However, to achieve scores better than persistence, allowance had to be made for stationary weather patterns and also for diurnal changes in weather conditions. Two new forecast experiments were prepared and carried out, using data from 12 cases during March 1983. First, data were edited and adjusted to reduce effects of local conditions (altitude, surface roughness), and then were advected. Finally, the adjustment was removed. The forecasts using a 500 mb space-averaged flow with modified initial conditions produced improved advection forecasts, with some parameters better than persistence and MOS (Model Output Statistics) for 2-7 hours. In the second experiment, an improved objective-analysis procedure was introduced, one based on the 'Barnes' approach, which uses one-half degree (about 45 km) resolution and previous analysis as a first guess. (Prior analyses were 1 degree, single pass, 'Cressman'-type analyses.) These improved analyses resulted in a somewhat better score for 1-3 hours (using a 'change-advection' technique), but were slightly worse at longer periods. Apparently, the small-scale patterns recovered by the improved analyses were largely either short-lived or stationary. These conditions would not lead to better advection forecasts. Further examination revealed that those parameters most difficult to resolve in the objective analyses (visibility, ceiling, and wind speed) also had the lowest forecast skill scores for persistence. Keywords: Aviation forecasting; Meteorology; Mesoscale analysis and forecasting.