Spurious Indo‐Pacific Connections to Internal Atlantic Multidecadal Variability Introduced by the Global Temperature Residual Method

Abstract

The climate phenomenon known as "Atlantic Multidecadal Variability" (AMV) is characterized by basin-scale fluctuations in sea surface temperatures (SSTs) over the North Atlantic, accompanied by changes in atmospheric circulation and associated hydrological impacts over Africa, Europe, and eastern North America [for example, Zhang et al., 2019 (Z19) and references therein]. The spatial and temporal characteristics of AMV derive from processes internal to the coupled ocean-atmosphere coupled system, and are also influenced by external radiative forcing, primarily anthropogenic aerosols [for example, Qin et al., 2020 (Q20) and references therein]. However, quantifying the relative contributions of internal and external drivers of AMV remains an ongoing challenge, with implications for prediction. Statistical approaches aimed at separating forced and internal components of observed AMV are sensitive to methodological choices, leading to conflicting results (e.g., Frankignoul et al., 2017). In particular, simple procedures such as subtracting global-mean temperature (Trenberth & Shea, 2006) or linear detrending (Enfield et al., 2001;Murphy et al., 2017) alias the pattern of forced climate change onto the structure of internal AMV (iAMV) as shown by Deser and Phillips (2021; hereafter DP21) and Baek et al. (2022). The modeling framework of "Single Model Initial-condition Large Ensembles" (LEs;Deser et al., 2020), which provides a priori knowledge of forced and internal contributions in any individual simulation, has served as an effective testbed for evaluating these and other methods (Baek et al., 2022; DP21;Frankignoul et al., 2017; Q20).Another simple and widely-used empirical technique for isolating iAMV (and internal variability in general) is the "global temperature residual method" (G-Res method), which employs linear regression analysis to remove variability at each location that is linearly congruent with variability in global-mean SST (G; DP21; Q20; Yan

Contrasting internally and externally generated Atlantic Multidecadal Variability and the role for AMOC in CMIP6 historical simulations

“…In terms of the AMV pattern, AMV_BM appears to be related more strongly to the Pacific, with a warming over the western Pacific and less warming over the eastern Pacific reminiscent of the spatial pattern of the Interdecadal Pacific Oscillation [17]. The AMV-Glob has a much weaker relation to the Pacific in contrast to the results of [92] who show that the AMV-Glob can erroneously find a relationship with the Pacific. However, [92] only uses data from 1950 onwards, which furthermore underlines the sensitivity of results to the choice of AMV index.…”

“…iAMVnm=AMVnm−fAMVm, where n is the ensemble member and m is the model). This method of separating externally forced and internal (or residual) variability has been used previously [92,96] and is often referred to as ‘de-meaning’ the ensemble. Note that, when characterizing the iAMV relationships for each model, we compute the desired metric for each member separately and then take the ensemble mean of the metric.…”

“…However, both indices do not discriminate between variability that is unique to the North Atlantic and global scale variability [ 19 , 90 ]. Other AMV indices have been proposed to address this specific issue, for example by either removing the global-mean temperature anomalies from the basin-mean AMV index [ 1 , 90 ] or by removing the globally forced signal spatially by regressing on the global mean [ 80 , 91 , 92 ]. More complex indices have also been proposed to address this problem [ 42 ].…”

Model‐Dependent Atlantic Multidecadal Variability Modulations on North Pacific Sea Surface Temperature Variability and Decadal Prediction

“…The results of this paper are important because we highlight the strength‐dependent AMV‐North Pacific modulations in climate models. Although we analyzed the modulations in observations, the method to isolate internal variability may influence the results (e.g., Deser & Phillips, 2023). Hence, whether there is a similar trans‐basin modulation in observations still remains controversial.…”

“…Correlations between the AMV and Pacific variability were found in observations (e.g., Nigam et al., 2020), albeit debates remain (Deser & Phillips, 2023; Fenske & Clement, 2022). Such trans‐basin correlations were attributed to AMV‐forced Pacific variability by using Atlantic SST forcing experiments.…”