기후모형에서 모사한 북태평양 혼합층: 일반 오차

Abstract

Ocean mixed layers, the vertically quasi-homogeneous upper ocean, directly interact with the overlying atmosphere so that they influence important air-sea interaction processes including variability of sea surface temperature and ocean uptake of atmospheric CO2. The thickness of ocean mixed layer or mixed layer depth (MLD) has been one of the challenging issues to simulate in climate models. In this study, we assess biases in the MLD simulated in climate models by comparing outputs from CMIP3 (coupled model intercomparison projects phase 3) models with a MLD climatology based on observational data. The most common biases in winter MLD are a deep bias inthe Kuroshio Extension region (KE) and a shallow bias in the Oyashio region. The deep bias in the KE is mainly driven by the fact that the simulated KE is too broad and too far north. The deep bias also can be attributed to the overestimated mid-latitude westerlies in CMIP3 models. The shallow bias in the Oyashio region is related tostratification bias in the upper ocean. A preliminary analysis reveals that the CMIP5 climate models also show similar MLD biases, suggesting that CMIP5 models still suffer from substantial MLD biases that may contributeto biases in ecosystem simulation.ke of atmospheric CO2. The thickness of ocean mixed layer or mixed layer depth (MLD) has been one of the challenging issues to simulate in climate models. In this study, we assess biases in the MLD simulated in climate models by comparing outputs from CMIP3 (coupled model intercomparison projects phase 3) models with a MLD climatology based on observational data. The most common biases in winter MLD are a deep bias inthe Kuroshio Extension region (KE) and a shallow bias in the Oyashio region. The deep bias in the KE is mainly driven by the fact that the simulated KE is too broad and too far north. The deep bias also can be attributed to the overestimated mid-latitude westerlies in CMIP3 models. The shallow bias in the Oyashio region is related tostratification bias in the upper ocean. A preliminary analysis reveals that the CMIP5 climate models also show similar MLD biases, suggesting that CMIP5 models still suffer from substantial MLD biases that may contributeto biases in ecosystem simulation.