Multi-Objective Ensemble-Processing Strategies to Optimize the Simulation of the Western North Pacific Subtropical High in Boreal Summer

The western North Pacific Subtropical High (WNPSH) in boreal summer is a major atmospheric player affecting East Asian climate, but its simulation in state-of-the-art climate models is still largely biased. Here we use a multi-objective optimization strategy, the Pareto optimality, to incorporate multiple physical constraints in processing multi-model simulations provided by the Coupled Model Intercomparison Project Phase 6. We aim to improve the simulation of WNPSH by this practice. Sea surface temperatures from three tropical oceanic basins are found highly related to WNPSH, and thus used as constraints. We also present an ameliorated strategy, which takes a subset of the raw Pareto optimality by imposing conditions of smallest errors. Results show that the overestimate of WNPSH is effectively corrected. The two multi-objective optimization schemes both perform better than the traditional approach, revealing the importance of implementing physically based links in processing multi-model ensemble simulations. The western North Pacific Subtropical High (WNPSH) in boreal summer exerts important impact on East Asian climate, but its simulation in climate models is still largely biased. In order to improve its simulation, we use an optimization strategy involving Pareto-optimality endowed with the ability to take multiple objectives into consideration to constrain climate models. Sea surface temperatures from three tropical oceanic basins are found highly related to the WNPSH, and thus used as constraining co-variables in the optimization. We also present an ameliorated strategy, by imposing additional conditions to further constrain the procedure and ameliorate the results. The two multi-objective optimization schemes are finally compared with a traditional ensemble-processing scheme that uses the same geophysical co-variables but without considering any physical constraints among them. The superiority of the multi-objective optimization is unequivocally demonstrated. Sea surface temperatures from three key basins are used to constrain the simulation of the western North Pacific Subtropical HighThe performance of models' ensemble can be improved when implementing physical links in processing multi-model ensemble simulationsSpurious states of the Pareto-optimal scheme can be eliminated with additional conditions of least errors