The calibration of Earth system model parameters is a high-dimensionality problem subject to data, time, and computational constraints. In this study, we propose a practical solution for finding diverse near-optimal solutions. We argue that the effective degrees of freedom in the model performance response to parameter input is relatively small. Comparably performing parameter configurations exist and showcase different trade-offs in model errors, providing insights for model development.