Spurious North Tropical Atlantic pre-cursors to ENSO

: The El Niño-Southern Oscillation (ENSO), the primary driver of year-to- 12 year global climate variability, is known to influence the North Tropical Atlantic (NTA) 13 sea surface temperature (SST), especially during boreal spring season. Focusing on 14 statistical lead-lag relationships, previous studies have proposed that interannual NTA 15 SST variability can also feed back on ENSO in a predictable manner. However, these 16 studies do not properly account for ENSO’s autocorrelation and the fact that the SST in 17 the Atlantic and Pacific, as well as their atmospheric interaction are seasonally 18 modulated. This can lead to misinterpretations of causality and the spurious 19 identification of Atlantic precursors for ENSO. Revisiting this issue under 20 consideration of seasonality, time-varying ENSO frequency, and greenhouse warming, 21 we demonstrate that the cross-correlation characteristics between NTA SST and ENSO, are fully consistent with a one-way Pacific to Atlantic forcing, even though the 23 interpretation of lead-lag relationships may suggest otherwise. interannual Niña)

reverse connection, which is characterized by a negative ENSO/NTA cross-correlation 48 with the NTA SST leading by about 8 months (Fig. 1b), is highly variable and especially 49 absent before the 1990s 22 (Fig. 1c). Despite some presumptions involved 22,36 , the 50 mechanisms responsible for the puzzling connection are less appreciated and a 51 comprehensive understanding of the two-way interaction between NTA variability and 52 ENSO is required. In this study, we use both observations and climate model 53 simulations to investigate the underlying mechanisms for the time-varying relationship. 54 We demonstrate that changes in the NTA/ENSO relationship can be explained in terms 55 of changes in ENSO frequency. The proposed mechanism is fully consistent with ENSO 56 forcing NTA, rather than the opposite. period notwithstanding a slight reduction of the correlation coefficient in the recent two 67 decades (red shading in Fig. 1c; see also ref. 38 ). In turn, the spring NTA warming 68 appears to contribute to the following La Niña development in the Pacific Ocean (and 69 similarly a spring NTA cooling contributing to a following El Niño) (Fig. 1b) with a 70 relatively weak correlation at about 8-month lag. However, we must also emphasize 71 here that an NTA warming in spring following an El Niño will automatically be 72 correlated with La Niña conditions 8 months later, because El Niño conditions are 73 usually followed by La Niña in the following year, without even involving a physical 74 NTA-to-ENSO relationship. Therefore, one needs to be careful in interpreting 75 seasonally modulated teleconnections of ENSO (see for instance discussion in ref. 21 ).

76
The 8-month leading relationship of NTA over ENSO is observed after the early 1990s 77 while it is absent in the preceding period (blue shading in Fig. 1c; see also ref. 22 ). Prior 78 to the 1990s we find a much longer characteristic lead of ~20-month (blue shading in

85
Here we hypothesize that the changing ENSO-NTA SST phase-lag relationships 86 can be explained in the context of different ENSO regimes manifested by quasi-biennial 87 and quasi-quadrennial periodicities. An El Niño is typically followed by a La Niña  relationship corresponds to what we see in the observations before the 1990s (Fig. 1c). 146 Our results clearly show that the 8-month lead of NTA over ENSO can be obtained,  Table 1). Almost all coupled models are 161 capable of capturing the robust ENSO forcing on the NTA SST ( Supplementary Fig. 5).

162
However, the models exhibit a large diversity in the statistical relationship between 163 boreal spring NTA SST variability and subsequent winter ENSO at ~8-month lead-time, 164 whereas a statistically significant relationship can only be simulated in about a quarter 165 of the CMIP6 models (Fig. 2a). To determine the underlying mechanisms responsible 166 for this, we rank the models based on their correlation between spring NTA SST 167 anomaly and subsequent winter ENSO conditions, and then select the 10 models closest

171
Although both model groups show a very similar ENSO SST anomaly pattern ( Fig.   172 2b), these two groups exhibit distinct ENSO spectral characteristics (Fig. 2c). The 173 models that have a statistically significant 8-month ENSO/NTA lagged relationship 174 exhibit a relatively shorter ENSO periodicity, analogous to the observations after the 175 1990s (Fig. 2c). In contrast, the models without a significant relationship at 8-month 176 NTA-lead-time have longer ENSO periodicities resembling the observations before the 177 1990s (Fig. 2c). In addition, there is a high inter-model linear correlation (R=0.75, 178 statistically significant at the 95% confidence level) between simulated dominant 179 ENSO periodicity and the lead-time of the most pronounced negative correlation of 180 NTA SST leading ENSO (Fig. 2d). This again supports our hypothesis that the statistical lead-time of NTA SST anomalies over the subsequent ENSO conditions is tightly 182 controlled by the ENSO periodicity.

183
There exists considerable uncertainty in the projections of trans-basin 184 interactions and the pan-tropical climate patterns that will emerge in a warming world 23 . 185 Thus, we next investigate the ENSO/NTA trans-basin interaction in CMIP6 future

207
Here we demonstrated using observational data, a simple seasonally modulated