Hydrodynamic models can predict states of interest to the coastal engineer, however, due to uncertainties in the model physics, model parameters, initial conditions, and model forcing data, large errors in prediction often result. To counter this, an ensemble sequential data assimilation scheme has been applied to the Model for Estuaries and Coastal Oceans (MECO), to constrain model predicted water temperature with remotely sensed sea-surface temperature observations. This paper describes a series of synthetic twin experiments mat contrast two ensemble sequential data assimilation schemes, the Ensemble Kalman Filter (EnKF) and the Ensemble Square Root Filter (EnSRF) in both one and three dimensional forms. The experiments show that the assimilation greatly improves the model prediction. The three dimensional form outperforms the one dimensional form, and that the EnSRF outperformed the EnKF significantly in the one dimensional form but only marginally in the three dimensional form.