Terrestrial Biosphere Models (TBMs) commonly use the Farquhar biochemical model and the Ball-Berry stomatal conductance (gs) model to simulate carbon and water fluxes. Two key parameters in TBMs are m, which indicates the sensitivity of gs to the photosynthetic rate, and Vcmax25, which represents the photosynthetic capacity. However, the accuracy of m retrieval for ecosystems is uncertain due to a lack of validation. It is unclear how accurately estimated m and Vcmax25 can enhance simulations of carbon and water fluxes. This study utilized the Bayesian parameter optimization to estimate m and Vcmax25 and validated the optimized parameters with measurements from a mixed forest stand in Ontario, Canada. Three scenarios were tested for optimizing m and Vcmax25, including carbon, water, and carbon-water coupling scenarios. Optimized parameters from the carbon-water coupling scenario showed the best correlations with measured values (R2=0.70 for both m and Vcmax25). By incorporating seasonally varied optimized m and Vcmax25, the estimated gross primary productivity (GPP) and evapotranspiration (ET) were improved compared to using constant parameters, with R2 increasing from 0.78 to 0.85 for GPP, from 0.65 to 0.71 for ET and RMSE reducing from 2.579 g C m-2 d-1 to 2.038 g C m-2 d-1 for GPP, from 1.151 mm d-1 to 0.137 mm d-1 for ET. The study proposes an effective approach to retrieve m and Vcmax25 for TBMs and demonstrates the efficacy of incorporating seasonally varied parameters for improving the accuracy of GPP and ET simulations, which is siginificant for understanding global carbon and water cycles.