AbstractAccurate forecasting of ocean currents in dynamic regions remains a critical challenge due to the sparsity of observations in global ocean observing networks and the limited resolution of present‐day regional ocean models. Lately, traditional observing platforms have been complemented by newly available data streams capable of sampling at higher spatial and/or temporal resolutions in dynamically significant regions in near‐real time. However, the relative merits and trade‐offs of incorporating these “nontraditional” observations into ocean state estimates have not been thoroughly investigated. Here, we perform a detailed statistical and dynamical comparison of two high‐resolution reanalysis products assimilating different combinations of traditional and nontraditional observations in the East Australian Current (EAC) system, a vigorous western boundary current. We show that sea‐surface height and temperature are well‐constrained by satellite measurements; however, below the surface, a reanalysis incorporating fully available observations better represents the ocean state. The core of the EAC jet is effectively constrained by subsurface observations from deep water moorings upstream of jet separation, while radar‐derived nearshore surface velocities in the separation zone are found to resolve the submesoscale cyclonic band inshore of the EAC. Cost function sensitivity analysis of both products reveals excessive model adjustment at depth causing the reanalyzes to overestimate alongshore transport relative to a 22‐year freely evolving simulation. Overall, the assimilation of nontraditional observations delivers marked improvement in representing dynamical features of the EAC. However, this improvement is not as pronounced in the model forecast due to the introduction of nonphysical dynamics or forcing, suggesting that other improvements such as increased model resolution are required.