Temperate inland riverine wetlands contribute meaningfully to regional carbon cycling, yet they remain poorly represented in land surface models. This study introduces CLM-Wet-MS, a modified version of the Community Land Model version 5 (CLM5) that incorporates a dedicated wetland landunit with wetland-adapted soil hydrology and a microbial soil decomposition scheme. The model was applied at Bibong wetland, a freshwater riverine wetland in South Korea, dominated by <em>Phragmites australis</em>. Hydrological representation is a dominant control on simulated carbon fluxes. The simulation configuration with observation-driven prescribed soil moisture is the only one that realistically reproduce the observed aerenchyma-dominated CH₄ transport. The microbial decomposition scheme reduces a persistent winter respiration bias that parameter optimization alone could not correct. Bayesian-based hierarchical parameter optimization was then performed against eddy covariance observations of gross primary productivity (GPP), ecosystem respiration (ER), and methane flux (FCH4). Optimization reduced root-mean-square error (RMSE) by 57.8 % (GPP), 37.1 % (ER), and 31.2 % (FCH4), with optimized parameters consistent with wetland-adapted C3 grass physiology and soil biogeochemistry. Comparison with plot-scale chamber measurements show that flux estimates diverge according to their spatial coverage, highlighting the influence of observation scale on model evaluation in heterogeneous wetlands. The CLM-Wet-MS model provides an optimized baseline for wetland modeling within CLM5 that can be extended to regional-scale simulations incorporating multiple land cover types.