Uncertainty in Atlantic Multidecadal Oscillation Derived from Different Observed Datasets and their Possible Causes
The aim of this study is to discover the uncertainties of the resulting AMO time series and its spatial patterns derived from different observed datasets and to explore the possible causes of these uncertainties. To achieve this, the AMO signal is extracted from three different observational SST datasets (ERSSTv5, HadISSTv1.1 and COBE-SST2) for the period 1900–2019 by using an objective method that does not presume in advance the existence of SST trends due to global warming—namely, Low-Frequency Component Analysis (LFCA). The LFCA can extract the spatially uneven trend. We also explore the differences between different versions of the same dataset to see if the same AMO biases are inherited from one generation of the dataset to the next.
Uncovering the role of observational SST uncertainty that resulted from mid-20th century bias adjustment in the observed AMO will help to understand the influence of ocean/atmosphere in shaping the AMO. It is shown that the ocean heat transport convergence seems to drive the tendency of AMO in a coherent AMO pattern (like COBESST2 or HadISST1). In contrast, in the incoherent AMO pattern, the atmosphere may become the driver of the AMO tendency. We suggest that to obtain a more coherent AMO pattern from ERv5 data, it is necessary to preprocess the monthly SST to an annual SST to cancel out the seasonal difference due to the bias correction. And if monthly SST data are employed, a 16-year low-pass filter is recommended in LFCA.
We found that the known coherent spatial pattern of the AMO at the basin scale over the North Atlantic appears in two out of the three datasets. Further analysis indicates that both the warming trend and the different techniques used to construct these observed gridded SSTs contribute to the AMO’s spatial coherence over the North Atlantic, especially during periods of sparse data sampling. The SST in the Extended Reconstructed SST dataset version 5 (ERSSTv5), changes from being systematically below the other datasets during the dense sampling periods on either side of the Second World War (WWII), to systematically above the other datasets during WWII, thereby introducing an artificial 10–20-year variability that affects the AMO’s spatial coherence. This coherence in the AMO’s spatial pattern is also affected by bias adjustment in ERSSTv5 at relative cool (i.e., non-summer) seasons, and by the heterogeneous North Atlantic warming pattern. The different AMO patterns can induce the different effects of wind, surface heat fluxes, and then drive ocean circulation and its heat transport convergence, especially for some seasons. For AMO indices, both the different detrending methods and different observational data result in uncertainty for the period 1935–1950. Such SST uncertainty is important to detect the relative role of the atmosphere and ocean in shaping the AMO.