Multi-decadal Variability And The Associated Uncertainty Of Global Ocean Heat Content

It has been shown that more than 90%of the earth's energy imbalance(EEI)in the climate system is sequestered in the ocean(increasing the OHC),the rest goes into warming the land and atmosphere and melting ice.An understanding of global and regional ocean heat con-tent(OHC)change is essential to understand both past and future climate change.It has been indicated that the global ocean exists significant OHC changes on decadal and multi-decadal scale in the past several decades and ocean heat redistribution associated with global warming has arisen a research surge.However,the existing researchers have given different observational OHC changes and then proposed different mechanisms to explain the ocean heat redistribution Where has the heat gone is still quite a question.What's more,the associated uncertainty remains unclear.In this work,the ocean heat redistribution among global ocean basins in the upper 1500-m layer is studied.By analyzing the consensuses and discrepancies between multi-datasets,we partially attribute the discrepancies between independent groups to the uncertainty associated with XBT bias correction and mapping methods.OHC change is monitored by ocean in situ observation system(OOS)with irregular and incomplete data coverage.How accurate can the past and current OOS monitor OHC and its variability is a crucial question for inferring the reliability of OHC estimates.Here we use an ocean reanalysis(CMCC Global Ocean Reanalysis System version 5:C-GLORSv5)as one realization of the past ocean,considered as "truth" for our purposes,within the 1980-2015 period and evaluate the uncertainty among mapping and the capacity of OOS on monitoring the OHC changes.The main conclusions are as follows.Firstly,all products show an increase of OHC since 1970 in each ocean basin revealing a robust warming,although the warming rates are not identical.The geographical patterns,the key modes and the vertical structure of OHC changes are consistent among the three datasets,implying that the main OHC variabilities can be robustly represented.However,large discrepancies are found in the percentage of basinal ocean heating related to the global ocean,with the largest differences in the Pacific and Southern Ocean.It is apparent that all of these regions are ocean heat sinks in the current decade,but there is no clear indication of which ocean basin dominates the global OHC change.Secondly,XBT bias correction schemes contribute to quite a lot to the total uncertainty of OHC estimate and affect both inter-annual variability and long term trend.The associated uncertainty concentrates among 1966 to 2000 with a mean magnitude of 9.9 ZJ.The main uncertainty comes from the Pacific Ocean(?75%)and locates among the western boundary current and maritime continent regions.Due to the transform of XBT instrument,the upper 300-m layer provides almost all the uncertainty before 1985 and then uncertainty among 300 to 700-m can't be neglected until 2000.Thirdly,we construct a synthetic ocean observation dataset by re-sampling the ocean from the reanalysis according to the actual spatio-temporal locations of in situ ocean subsurface temperature data(extracted from the EN4 profile database).Based on these synthetic data,uncertainty associated with mapping methods are quantified through comparing the reconstructed a-nd reanalysis value.There is large difference for the performance among different mapping schemes.In general,IAP-mapping could reconstruct the OHC variability better than the SGA and WGA schemes,with the root-mean-squared error about 0.016K(0.005K)during 1980-2004(2005-2015)period.SGA scheme is likely to underestimate the temporal variability and long-term warming trend of OHC while the WGA one overestimates the temporal variability.The overestimate from WGA scheme comes from the irregular distribution of sample.Finaly,we show that the global OHC tendency(OHCT)in the upper 2000m can be reconstructed with a root mean squared error(RMSE)of 3.61W/m2,with 3.15 W/m2 in the upper 700-m and 1.16 W/m2 within 700 to 2000-m for the 1980-2015 period.The error is more prominent on quasi-annual scale(3.43W/m2)but much smaller on the inter-annual(1.07 W/m2)and decadal scales(0.22W/m2).The RMSE decreases from 4.01W/m2(1980-2004)to 2.51W/m2(2005-2015)due mainly to the deployment of the Argo network.The impact of two individual OOS components are quantified:removing XBT(Argo)data increases the RMSE in the OHCT estimate from 4.01W/m2(2.51W/m2)to 5.51 W/m2(4.89W/m2)over the 1980-2004(2005-2015)period in the upper 700m(2000m).Spatially,very large uncertainties appear in both west boundary currents and Antarctic Circumpolar Current(ACC)regions(RMSE?500W/m2).This study revisited the ocean heat redistribution on the past several decades and highlights the importance of uncertainty among different ocean analyses.Evaluation of the historical subsurface OOS provide a basis for assessing OOS in monitoring OHC change,for quantifying the uncertainty associated with OHCT estimates,and for improving the OOS in the future.