Marine ecosystem dynamic models(MEDMs) are important tools for the simulation and prediction of marine ecosystems. This article summarizes the methods and strategies used for the improvement and assessment of MEDM ski...Marine ecosystem dynamic models(MEDMs) are important tools for the simulation and prediction of marine ecosystems. This article summarizes the methods and strategies used for the improvement and assessment of MEDM skill, and it attempts to establish a technical framework to inspire further ideas concerning MEDM skill improvement. The skill of MEDMs can be improved by parameter optimization(PO), which is an important step in model calibration. An effi cient approach to solve the problem of PO constrained by MEDMs is the global treatment of both sensitivity analysis and PO. Model validation is an essential step following PO, which validates the effi ciency of model calibration by analyzing and estimating the goodness-of-fi t of the optimized model. Additionally, by focusing on the degree of impact of various factors on model skill, model uncertainty analysis can supply model users with a quantitative assessment of model confi dence. Research on MEDMs is ongoing; however, improvement in model skill still lacks global treatments and its assessment is not integrated. Thus, the predictive performance of MEDMs is not strong and model uncertainties lack quantitative descriptions, limiting their application. Therefore, a large number of case studies concerning model skill should be performed to promote the development of a scientifi c and normative technical framework for the improvement of MEDM skill.展开更多
基金Supported by the National Natural Science Foundation of China(Nos.41206111,41206112)
文摘Marine ecosystem dynamic models(MEDMs) are important tools for the simulation and prediction of marine ecosystems. This article summarizes the methods and strategies used for the improvement and assessment of MEDM skill, and it attempts to establish a technical framework to inspire further ideas concerning MEDM skill improvement. The skill of MEDMs can be improved by parameter optimization(PO), which is an important step in model calibration. An effi cient approach to solve the problem of PO constrained by MEDMs is the global treatment of both sensitivity analysis and PO. Model validation is an essential step following PO, which validates the effi ciency of model calibration by analyzing and estimating the goodness-of-fi t of the optimized model. Additionally, by focusing on the degree of impact of various factors on model skill, model uncertainty analysis can supply model users with a quantitative assessment of model confi dence. Research on MEDMs is ongoing; however, improvement in model skill still lacks global treatments and its assessment is not integrated. Thus, the predictive performance of MEDMs is not strong and model uncertainties lack quantitative descriptions, limiting their application. Therefore, a large number of case studies concerning model skill should be performed to promote the development of a scientifi c and normative technical framework for the improvement of MEDM skill.
