A corrected version of an unstratified box model of potential energy anomaly , initially developed by Garvine and Whitney (2006), and a new two-layer box model that allows for stratified and well-mixed conditions are applied to Delaware Bay. The models are applied for the Garvine and Whitney (2006) 1988-1994 study period and in Spring 2003; however, only model results of potential energy anomaly from the latter period are compared to in situ observations obtained outside the bay mouth. Unstratified model results for the two study periods reveal that the river discharge (Ω1) is the largest potential energy anomaly contributor. This term is closely followed (but with opposite sign) by the coastal current efflux term (Ω2). For the two-layer model the largest contributor is the dense inflow term (Ω6). The wind term (Ω5) is the second largest, followed by the tide (Ω3), river discharge (Ω1) and coastal current terms. In both models the solar heat flux term (Ω4) makes the smallest contribution to ϕ. The available one-month comparison of model results to observations renders statistically insignificant correlation coefficients for both models. We speculate dynamical differences between conditions at the estuary mouth and the instrument location on the nearby shelf contribute to the model-observation mismatch. Other statistics, such as the root mean square error indicate that the unstratified model performs better than the two-layer model for the observation period. The latter model is, however, able to depict the importance of tides and winds in the computation of potential energy anomaly and is able to detect the response of ϕ due to strong wind events. While there is no clear model choice for the Delaware Bay, the unstratified model may be entirely inappropriate for highly stratified estuaries.