For evaluating the performance of air traffic management system, a model based on linguistic decision map with multi-subnets was proposed as well as the indexes. The influences among indexes were analyzed by using a multi-subnets strategy, and the initial judgment matrices and influence connective matrices were built by multiple experts using linguistic scales. The proposed approach can achieve global priorities of indexes and evaluation results of alternatives. Finally, the case study illustrates that, in comparison with the independent situation, results of the model can represent priorities of indexes more objectively and comprehensively, and can present the gap among the alternatives more obviously.