Comparison of ocean deoxygenation between CMIP models and an observational dataset in the North Pacific from 1958 to 2005

This study investigated the relationship between the observed and simulated dissolved oxygen (O2) inventory changes in the North Pacific by analyzing an observational dataset and the outputs of Coupled Model Intercomparison Project Phase 5 and 6 (CMIP5/6) between 1958 and 2005. A total of 204 ensembles from 20 models were analyzed. Many of the models in the North Pacific subarctic region have higher climatological O2 concentrations than observed at deeper water depths. Therefore, the negative trend of O2 inventories tends to be larger, and in fact, several model ensemble members have a larger negative trend in O2 inventories than observed. The variability among model ensemble members is more influenced by the uncertainty due to internal variability than by the uncertainty resulting from model dependency. An inter-model empirical orthogonal function (EOF) analysis revealed that the different simulated magnitudes of the negative O2 trend is closely associated with the first EOF mode, and ensemble members with strong negative trends are characterized by large oxygen reduction in the subarctic North Pacific, especially around the boundaries between the North Pacific Ocean and the Sea of Okhotsk as well as the Bering Seas. The modeled strong O2 decrease in the subarctic North Pacific is consistent with the spatial pattern of the observed O2 trend. Further analysis of climate models indicated that the O2 decrease in the subarctic region was primarily caused by physical factors. This conclusion is supported by the significantly high correlation is present between the potential temperature and O2 inventory trend in the subarctic region, whereas an insignificant correlation coefficient is present between dissolved organic carbon and the O2 inventory trend. However, the observations have a larger ratio of O2 inventory trend to temperature trend than any of the ensembles, and thus the relationship between O2 and temperature change in the subarctic North Pacific seen in the CMIP5/6 simulations is not exact.