Abstract: Limited by common situations of closely spaced modes and large structural dimension, damage identification based on modal parameter is difficult to perform in civil structures. A damage identification method based on multi-level modal group response reconstruction in the presence of close spaced modes is proposed. Several modes with small intervals are grouped together, and response of the entire modal group is extracted as damage sensitive characteristic. Based on the collected modal response, a multi-level damage identification strategy is adopted. In the super element level damage location, the original structure is first converted into a super element model with fewer DOFs through model reduction, and then the minimization problem is solved by defining the modal group response strain energy as a damage index to achieve the location of damaged super elements; In element level damage identification, the minimization problem is expressed as the discrepancy between reconstructed and actual modal group response to achieve elemental damage localization and quantification. A numerical simulation study and the experimental verification were conducted to demonstrate the operational process and feasibility of this method. Compared with traditional methods, the results show that the proposed method improves the accuracy and efficiency of damage identification through multi-level identification strategies and model reduction, and on the other hand, improves the shortcomings of modal-analysis-based methods that cannot accurately identify damage when faced with close spaced modes. Regardless of the presence or absence of close spaced modes. Damage identification can be performed based on multiple dynamic responses such as stress, strain, displacement, and acceleration of the structure.