Monitoring and predicting global ocean biogeochemistry and marine ecosystems is one of the biggest challenges for the coming decade. In operational systems, biogeochemical (BGC) models are forced - or coupled - with physical ocean models fields that are generally constrained by data assimilation of temperature, salinity and sea level anomalies. Yet, while physical data assimilation substantially improves simulated physical fields, BGC models forced by such analyses are commonly degraded, and more especially in equatorial regions. Here impacts of physical data assimilation on surface chlorophyll and nitrate concentrations are investigated in the tropical Pacific, based on three ocean reanalysis runs using the same physical-BGC model configuration but differing in their level of physical data assimilation. It is shown that, in the Mercator Ocean operational system, the assimilation of satellite altimetry and sea surface temperature in addition to temperature and salinity in situ profiles leads to spurious vertical velocities in the western equatorial Pacific. Our analysis suggests that these unrealistic vertical velocities are explained by the use of an inaccurate mean dynamic topography for the assimilation of altimetry that modifies the pressure-driven horizontal circulation in the upper ocean layer. Moreover, the biases found in this key region modify the subtle dynamical and BGC balances in the whole tropical Pacific and result in unrealistic trends of ocean heat content and nitrate concentration. This study demonstrates that looking into details of the physics is indispensable to improve physical data assimilation systems and to ensure that they make the best use of observations. This is also a key point to refine the strategy of the BGC models forcing and further improve ocean predictions.