Abstract
We conducted a comparative study of ocean heat content (OHC) in the top 2,000 m during the Argo‐era using 12 latest and representative global ocean data sets. The differences in the global and basins‐wide OHC trends were minor among the observation‐based data sets, and remarkable among the ocean reanalyzes (RAs). Some RAs might exhibit much higher or lower basins‐wide warming rates than the observation‐based data sets. In the top 700 m, RAs suggested similar large‐scale warming and cooling patterns, in agreement with the observation‐based data sets. Below 700 m, the major warming and cooling features were significantly different between RAs and the observation‐based data sets. All data sets suffered from relatively larger uncertainties in the highly dynamic regions. Special caution is suggested when estimating the OHC using only a single data set, especially a RA. Differences of RAs' OHC from observation‐based data sets were significantly reduced when considering their ensemble mean, to be further confirmed with a larger sample of data sets.
Plain Language Summary
Since the 1970s, the global ocean has absorbed more than 90% of the excess heat that was mainly caused by the increasing greenhouse gas emissions from the anthropogenic activities. Estimating the warming or cooling trend of the global ocean is therefore vitally important. At present, there are many global ocean data sets, but to what extent these data sets agree or disagree with each other in the ocean warming or cooling estimate? This question has not been fully addressed, especially between the observation‐based data sets and the data sets derived from both the numerical ocean models and observations, known as ocean reanalyzes. In this study, we found that the observation‐based data sets generally agreed well with each other in the estimate of warming or cooling in the top 2000 m. The ocean reanalyzes, however, exhibited significant discrepancies with the observation‐based data sets, although they could largely suggest similar large‐scale warming or cooling patterns to the observation‐based data sets in the top 700 m. The differences from the observation‐based data sets could be significantly reduced when considering the average of multiple ocean reanalyzes. Both the observation‐based data sets and ocean reanalyzes suffered from large uncertainties in the highly dynamical regions.
Key Points
Global and basins‐wide ocean heat content (OHC) trends were largely similar among the observation‐based data sets
Ocean reanalyzes (RAs) well captured the large‐scale warming and cooling patterns indicated by observations in the top 700 m, but poorly below 700 m
The ensemble mean of multiple RAs has the potential to provide a more reliable estimate of OHC than a single RA