We present a straightforward method for estimating surface circulation on an f plane from a set of irregularly spaced vertical density profiles. The first step is to express bottom density &rgr;h as the sum of a mean &rgr;˜h for a given water depth h and an anomaly, &rgr;h'=&rgr;h-&rgr;˜h. Sea level &eegr; can then be decomposed into a dynamic height &eegr;˜ relative to a deep reference level and a correction term, &eegr;'=&eegr;-&eegr;˜. The dynamic height is estimated using a generalization of the method of Helland-Hansen <1934> for diagnosing flow through a cross-shelf section under the assumption of zero bottom geostrophic flow. The correction &eegr;' satisfies a two-dimensional elliptic partial differential equation forced by the bottom density anomaly, wind stress and the open boundary conditions. Before calculating the density-driven component of &eegr;' we first test if the &rgr;h' are statistically different from uncorrelated noise. If they are not, the correction associated with the bottom density anomaly is set to zero. Thus the method has a degree of robustness to errors in the density observations. If the &rgr;h' have well-defined spatial structure the elliptic equation is solved for &rgr;' and sea level is equated to &eegr;˜+&eegr;'. Note that even if this last step is required the only gridding of the density data is two-dimensional. This makes the proposed method simpler to use than many of the existing diagnostic models which require a three-dimensional gridding of the observed density profiles. To test the method, we use it to diagnose the flow from an idealized density field overlying an isolated topographic feature. The predicted sea level and flow fields are then compared, and shown to be in good agreement, with results from the Princeton Ocean Model. The method is then used to diagnose the winter surface circulation on the Scotian Shelf from observed density profiles. The reliability of the diagnosed flow pattern is assessed by comparing it against all available near-surface current measurements. The differences between the observed and diagnosed currents are used to estimate the remotely forced circulation on the Scotian Shelf. It is shown that the main features of the circulation can be explained by gradients in the density field. Remote forcing is important near the coast and the shelf break. The effect of local wind is relatively weak. ¿ American Geophysical Union 1996