资源环境科技发展态势分析平台

Observational estimates of global ocean heat content (OHC) change are used to assess Earth's energy imbalance over the 20th Century. However, intercomparison studies show that the mapping methods used to interpolate sparse ocean temperature profile data are a key source of uncertainty. We present a new approach to assessing OHC mapping methods using 'synthetic profiles' generated from a state-of-the-art global climate model simulation. Synthetic profiles have the same sampling characteristics as the historical ocean temperature profile data but are based on model simulation data. Mapping methods ingest these data in the same way as they would real observations, but the resultant mapped fields can be compared to a model simulation 'truth'. We use this approach to assess two mapping methods that are used routinely for climate monitoring and initialisation of decadal forecasts. The introduction of the Argo network of autonomous profiling floats during the 2000s drives clear improvements in the ability of these methods to reconstruct the variability and spatial structure of OHC changes. At depths below 2000 m, both methods underestimate the magnitude of the simulated ocean warming signal. Temporal variability and trends in OHC are better captured in the better-observed northern hemisphere than in the southern hemisphere. At all depths, the sampling characteristics of the historical data introduces some spurious variability in the estimates of global OHC on sub-annual to multi-annual timescales. However, many of the large scale spatial anomalies, especially in the upper ocean, are successfully reconstructed even with sparse observations from the 1960s, demonstrating the potential to construct historical ocean analyses for assessing decadal predictions. The value of using accurate global covariances for data-poor periods is clearly seen. The results of this 'proof-of-concept' study are encouraging for gaining further insights into the capabilities and limitations of different mapping methods and for quantifying uncertainty in global OHC estimates.