This paper proposes an ocean-only dynamical framework to mitigate the influence presentday biases of Earth System Models (ESMs) on future regional ocean physical and biogeochemical projections. Initially, a control experiment is conducted using fluxes derived from an atmospheric reanalysis, excluding climate change signals. Subsequently, a climate change simulation is performed by adding historical and future fluxes perturbations from a selected ESM to these background realistic fluxes. Since part of the ESM surface heat fluxes perturbation is a direct feedback to the sea surface temperature (SST) warming, these fluxes perturbations are split into SST-dependent and independent components. The climate change simulation is forced by the independent component, while the SST-dependent component is modeled online as an SST relaxation to the control experiment, accounting for Newtonian cooling and long-wave radiative feedback. This approach demonstrates that ESMs present-day biases can heavily impact the reliability of regional physical and biogeochemical ocean projections. For instance, the strong cold-tongue bias simulated by the IPSL-CM6A-LR model causes greater warming and chlorophyll decrease in the western than in the eastern equatorial Pacific, while our bias-corrected simulation shows opposite projected patterns. Sensitivity experiments applying heat, freshwater and momentum fluxes perturbations separately further indicate that thermodynamical and dynamical processes equally contribute to this warming pattern, highlighting the strong role of the Bjerknes feedback. This cost-effective method can be applied to any ESMs oceanic component to produce more reliable regional oceanic projections and understand the mechanisms driving the projected patterns.