<p>Currently, determining the mechanical parameters by inverse analysis in deep-buried tunnels is still challenging and essential. Problems such as the limitation of unknown parameter dimensions (tremendous computational cost) and insufficient inversion accuracy may occur when using the conventional inverse method to solve the mechanical parameters of rockmass in the deep-buried tunnel. By inheriting the advantages of the gradient descent method, a novel data-simulation-driven method is proposed to determine the high-dimensional mechanical parameters of rockmass in buried tunnels based on field observations. The correlation coefficients between unknown parameters can be calculated by the spatial position relationship and its corresponding field observation. Thus, the covariance matrix between unknown parameters is formed, reducing the dimension of unknown parameters. The results of numerical verification show that the different types of objective distributed elastic modulus of tunnels can be quickly obtained by the inverse method proposed. The degree of agreement between inversion and actual distributed elastic modulus can be up to 90%. The field application results show that the numerical accuracy calculated by the conventional inverse method is insufficient, consuming tremendous computational cost. In addition, the numerical accuracy can be enhanced using the proposed inverse method: the simulation error between simulated values and field observations are basically within 20%, meeting the engineering precision requirements. Finally, the distributed elastic modulus of rockmass in the target section is obtained. The “data simulation-driven method” is potentially useful for solving the high-dimensional mechanical parameters in deep-buried tunnels during excavation and optimizing the support design based on field observations.</p>