The Cabo Frio zone presents an important coastal-oceanic system of great interest for biological and economical ressources. Among others, the coastal upwelling is one of the specific feature that occurs frequently in the area. As presented in a previous feasibility study, the main parameters of the upwelling can be tracked by acoustic tomography in vertical slices, using a feature model parameterization. In this paper, acoustic propagation simulations are used to study the sensitivity of acoustic measurements to a more advanced upwelling feature model. More specifically, it is shown how the frequency diversity and source position have a strong effect upon the subjacent ambiguity function involved in the inversion process. The Bartlett processor performances are compared for different frequency and hydrophone setup, showing that an increase of the number of transmitted frequencies compensates for the lack of hydrophones, and reduces efficiently the ambiguities. Furthermore, it is shown that a frequency-coherent cost function reduces strongly correlation between feature model parameters, enabling to resolve more efficiently the environment.