In offshore wind farm management, instrumentation campaigns are often deployed to measure the vibrational response of select wind turbines in the farm. Structural identification methods utilize this instrumentation data in order to track the dynamic performance of the turbine with the goal of performance assessment and damage identification. One approach for structural identification is inverse modeling or model updating which involves integration of measured observations from the structure with a virtual model that represents the true structure by matching the response features (e.g., modal parameters). Updated models or digital twins can be used to predict the dynamic response of the structure due to loading conditions, or to estimate the remaining fatigue life. The purpose of model updating is to obtain optimal modeling parameters that match the behavior of the real structure; these optimal model parameters can be treated both deterministically and probabilistically. Probabilistic treatment of model parameters enables uncertainty quantification and modeling error estimates which provides additional information about the digital twin. This work performs probabilistic digital twinning of an operational 6 MW offshore wind turbine using a hierarchical Bayesian model updating approach. Two model parameters (effective aerodynamic stiffness in horizontal directions) are treated probabilistically such that a distribution mean and covariance can be fit and estimated from the data. Multiple versions of hierarchical Bayesian model updating are performed to account for the unique behavior of the turbine under variable environmental and operational conditions. Different distributions are fit to find model parameter distributions for power production scenarios, accounting for the unique and varying dynamic behavior of the wind turbine during operation. For instance, the dominant natural frequency of the wind turbine show a distinct relationship with wind speed during power production by the wind turbine. In this case, level 2 hierarchical Bayesian modeling is employed in which the model parameter distribution is treated as a function of wind speed. The relationship of parameter distribution and wind speed is treated as both a linear function as well as a more complex function in order to assess which function best captures the wind turbine’s behavior. Results are compared to model parameters obtained through a frequentist approach.