Northward ITCZ shift drives reduced ENSO activity in the Mid-Pliocene Warm Period


 The El Niño Southern Oscillation (ENSO) is the strongest pattern of year-to-year climate variability found in the equatorial Pacific Ocean with global impacts. However, it is not fully understood how ENSO responds to different warming scenarios. In the warmer climate (~2-3K) of the mid-Pliocene Warm Period (~3 Ma BP), models consistently suggest a weakening of ENSO variability, with a mean reduction of 25% (±16%). We show that a near unanimous weakening of ENSO across models cannot be fully explained simply by mean state changes in the equatorial Pacific Ocean. Instead, robust off-equatorial mean state changes in the mid-Pliocene are not favourable for ENSO activity. A northward displacement of the Pacific Inter-Tropical Convergence Zone (ITCZ) is found to be significantly linked to the ENSO weakening across models. This is accompanied by increased south-easterly trade winds in the western Pacific and an intensified South Pacific Subtropical High, which are consistent with suppressed activity of processes that initiate ENSO. Our results provide a constraint to past and future changes to ENSO associated with the climatological ITCZ position.

interannual band (Supplementary Figure S2), a timescale that is dominated by ENSO. 108 Additionally, 75% (17 out of 23) of the models indicate a tendency for a shift towards 109 lower frequencies as indicated by either an increased amplitude at low-frequency (>10 yr) 110 or a more pronounced weakening at interannual than on longer time scales (Supplementary 111 Figure S2). However, changes in decadal or longer periods must be further evaluated using 112 longer timeseries data. Here due to data availability, we only use the last 100 years of each 113 model's simulation. Firstly, we evaluate changes in the thermocline slope as a proxy for changes in 127 equatorial ocean dynamics. Strong (weak) westward equatorial currents drive increased 128 (decreased) east-west thermocline slope, as it shoals (deepens) the eastern thermocline 129 while deepening (shoaling) its western sector 9 . We find that models with a steeper mean 6 thermocline in the mPWP (i.e. a La Niña-like mean state) are typically associated with 131 larger ENSO amplitude reductions, while a flatter mean thermocline (i.e. an El Niño-like 132 state) is associated with either an slight increase or a weak decrease in ENSO variability 133 (r=-0.52; Figure 2b). This indicates that equatorial Pacific mean state with a steeper 134 thermocline, which corresponds to intensified ocean-atmosphere circulations, is less 135 favourable to strong ENSO variability. Under such La Niña-like mean state, stronger initial 136 anomalies are required to substantially weaken the climatological states in order to provide 137 favourable conditions for strong El Niño development 9 . 138 We further examine possible changes in the equatorial oceanic conditions that could 139 be unfavourable to ENSO development 18,32 . Ocean stratification has been hypothesized to 140 influence the variability of extreme ENSO events, as an increased ocean stratification 141 would tend to increase the dynamical coupling between the ocean and the atmosphere 8 . As 142 such, we evaluate ocean stratification in the central-west Pacific near the warm pool edge, 143 a region of maximum wind variability and where wind anomalies trigger ocean waves and 144 initiate ENSO development. Indeed, we find that models with decreased ocean 145 stratification are associated with major ENSO reductions, and the reduction is weaker with 146 increased ocean stratification (Figure 2c). However, an increased stratification seen in nine 147 models cannot support the fact that the ENSO variability is reduced in each of seven of 148 those models. A similar inconsistency also applies for the thermocline slope change 149 ( Figure 2b). Thus, while changes in the thermocline and stratification help to explain inter-150 model differences in ENSO amplitude changes, there must be other processes that apply 151 across models, which provide an explanation for the overall weakening of ENSO 152 variability.

Off-Equatorial Pacific changes 155
Whilst ENSO development is closely related to the zonal equatorial dynamics 28 , 156 ENSO events are also affected by a variety of other large-scale processes beyond the 157 equatorial Pacific 10,33,34 . For instance, changes to the mean meridional SST gradient or 158 processes in the extratropics can play an important role in triggering ENSO events. In 159 particular, all PlioMIP models simulate a weaker equator-to-pole temperature gradient 160 associated with polar amplified warming 35 . 161 To investigate processes outside the equatorial Pacific, we first evaluate the role of 162 meridional SST gradients through possible displacements of the ITCZ in the mPWP. 163 Southward (northward) ITCZ displacements, due to changes in off-equatorial SST 164 gradients, have been shown to affect ENSO activity through increased (reduced) 165 probability of occurrences of deep convection in the central-eastern Pacific 11 . Here we 166 show that a mean northward ITCZ shift during austral spring-summer, i.e., during 167 developing and mature ENSO phases, is significantly related to the ENSO weakening 168 across models (Figure 3a). This northward shift in the ITCZ generally acts to supress El 169 Niño development, via a reduced probability of deep convection occurrences in the eastern 170 Pacific 11 . To illustrate this, we evaluate models' performance in simulating the non-linear 171 relationship between ENSO SST anomalies and anomalous precipitation events in the 172 eastern Pacific (see Methods; Supplementary Figure S6). Five models correctly simulate 173 this characteristic and indicate that the further north the mean ITCZ migrates the less 174 probable are occurrences of anomalous rainfall events in the eastern Pacific associated with 8 ENSO SST anomalies (Figure 3b-f). The ITCZ shift can fully explain ENSO weakening 176 across these 5 models (r=0.99; Supplementary Figure S7). 177 We also evaluate possible changes to the processes that are known to initiate ENSO 178 events. Firstly, the reversal of the circulation of easterly trade winds in the western Pacific 179 is known to initiate ENSO development 36 . In the PlioMIP models, the annual mean   With respect to future warming, paleoclimate studies have been investigating 295 whether there was a past warm climate that would serve as an analogue to the current 296 warming. Our findings indicate that, although the mPWP surface warming is comparable 297 in magnitude as projected toward the end of 21 st century under a 'business as usual' 298 scenario (~3K) 20 , ENSO shows an opposite response to that projected 8,11 . It is worth noting 299 the mPWP exemplifies an equilibrium climate with similar CO2 concentration as today, 300 indicating we could end-up in a similar-to-Pliocene climate if CO2 is maintained at present 301 levels once a steady state is reached. However, the current rate of atmospheric CO2 rise is 302 unprecedented in Earth's history, which differs from how Earth has warmed in the past. 303 Thus, linking past and future warmings is not straightforward. Here the evaluation of the 304 mPWP shows that in an empirically based equilibrium warming a northward ITCZ shift 305 drives reduced ENSO activity. If this mechanism can be applied to the 21 st century 306 projections where a southward shift of the Pacific ITCZ is projected 10 , then an increase in 307 ENSO variability 8 in the coming decades appears to be a potential outcome. 308 309

Methods 310
Models and data. Models were selected according to data availability in the PlioMIP1 and 311 PlioMIP2 databases. See Supplementary Table S1 for a list of the models included in our 312 analysis. A total of 9 PlioMIP1 and 16 PlioMIP2 models were analysed. PlioMIP1 and 313 PlioMIP2 boundary conditions are specified in Supplementary Table S2         ENSO-ITCZ inter-model relationship. a) PlioMIP2 inter-model relationship between the change in the Niño3 amplitude and mean ITCZ shift from October to February. b) to f) model relationship between DJF Niño3 SST anomalies and DJF Niño3 rainfall anomalies for pre-industrial (blue) and mid-Pliocene simulations (yellow). Models were selected according to their ability to simulate non-linear ENSO characteristics (See Methods). PlioMIP1 precipitation data for the last 100 years was not available so these models could not be included in this analysis (see Methods).

Figure 4
Changes to potential ENSO triggers. a) inter-model relationship between the change in the intensity of the western Paci c trade winds (from 160ºE to 150ºW and from 10ºS to 10ºN) and the amplitude (standard deviation) of its monthly variability. To ideally examine changes in the western wind bursts we would daily output, however high frequency output was not available for the PlioMIP models. b) Change in the amplitude (standard deviation) of the South Paci c Meridional mode time series, de ned as the mean SST anomaly between 15ºS-25ºS and 250ºW-260ºW. c) Change in the amplitude (standard deviation) of the meridional wind variability over the Southern Hemisphere Booster region (from 10ºS to 30ºS and from 140ºW to 170ºW). PlioMIP2 models in panels 'b' and 'c': a -CCSM4-UofT; b -CCSM4-2deg; c -CESM2; d -COSMOS; e -EC-EARTH3.3; f -GISS-E2-1-G; g -HadCM3; h -IPSL-CM6A-LR; i -IPSL-CM5A; j -IPSL-CM5A2; k -MIROC4m; l -MRI-CGCM2.3; m -NorESM-L; n -NorESM1-F.

Figure 5
Energetics constraints for the ITCZ position. a) DJF precipitation change in the PlioMIP2 models (mPWP minus pre-industrial). Stippling indicates where the change is signi cant at the 95% level. b) multi-model mean change zonally averaged SST for the PlioMIP1 (magenta) and PLioMIP2 (red). Banding indicates standard deviation range. c) Changes in DJF atmospheric energy ux, computed as the residual between the total top-of-the-atmosphere and surface energy uxes, in the CAM4 experiments forced with PlioMIP1 and 2 climatological SST and sea-ice (see Methods). Banding indicates standard deviation range of a 5member ensemble. d) Changes in the meridional streamfunction in the CAM4 experiment forced with PlioMIP2 SST and sea-ice (see Methods). y-axis: Pressure [mb]. Contours indicate pre-industrial streamfunction (zero contour in bold). Colours indicate change (mPWP minus pre-industrial) e) Intermodel relationship between changes in the intensity of the zonal western Paci c trades and ITCZ shift during austral summer. f) Changes in global low-level (850 hPa) winds and stream function in the PlioMIP2 models. Wind changes are only plotted where there is a signi cant change at the 95% level.
Note: The designations employed and the presentation of the material on this map do not imply the expression of any opinion whatsoever on the part of Research Square concerning the legal status of any country, territory, city or area or of its authorities, or concerning the delimitation of its frontiers or boundaries. This map has been provided by the authors.

Figure 6
Schematic of the drivers of suppressed ENSO activity in the mPWP. A northward ITCZ shift reduces the probability of occurrence of deep convection in the central-eastern Paci c. Energetic constrains for the ITCZ position indicates that higher rates of warming in Northern Hemisphere drive a northward ITCZ shift and intensi ed enhanced Southern Hemisphere Hadley circulation. These changes are also associated with enhanced subtropical high and intensi ed western Paci c trades. Enhanced trade windssuppress wind variability in the western Paci c, which are important for El Niño initiation. An intensi ed subtropical high is thought to impede zonal pressure anomalies across the tropical South Paci c and, thus, suppress the activity of the South Paci c Meridional Mode (SPMM) and Southern Hemisphere Booster that are important for the development of strong El Niño events. Note: The designations employed and the presentation of the material on this map do not imply the expression of any opinion whatsoever on the part of Research Square concerning the legal status of any country, territory, city or area or of its authorities, or concerning the delimitation of its frontiers or boundaries. This map has been provided by the authors.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.