Most existing models of shelf currents (such as coastal-trapped wave models) usually disagree with observations, in that the cross-shelf currents are weaker than observed, and that the predicted coherence length scales of cross-shelf currents are much longer. We seek to test the hypothesis that these inconsistencies can be resolved by including, in the forcing, more realistic wind variations with alongshore length scales down to about 10 km. We use the Coastal Ocean Dynamics Experiment (CODE) Northern California, 1982, wind observatins to drive a linear, stratified stochastic ocean model which includes bottom friction. The model results are expressed as statistics of the flow field which can be checked against observations. CODE aircraft winds from 1982 low-level alongshore flight tracks were converted to stresses and then subjected to wavelengths shorter than 50--100 km than would be expected from the extrapolation of larger-scale spectral estimates based on buoy wind times series.

Thus, for the CODE region, the forcing is energetic at relatively short length scales. The model results for cross-shelf velocity support our hypothesis, in that the modeled alongshore coherence length scales are indeed much shorter than those from a large-scale-only model. For example, at a 7-day period, the present calculations predict a coherence squared of 0.3 at a 35-km separation, while the large-scale-only model predicts 0.3 at 250 km. The observations show a 35--50 km scale for 0.3 coherence squared. However, while cross-shelf current variances are increased, they are still about a factor of 5 or more smaller than the CODE shelf observations over the shelf. ¿ American Geophysical Union 1994