Nonlinear response of Northern Hemisphere stratospheric polar vortex to the Indo–Pacific warm pool (IPWP) Niño

Variations in tropical sea surface temperatures (SST) have pronounced impacts on the stratospheric polar vortex, with the role of El Niño being the focus of much research interest. However, the Indo–Pacific warm pool (IPWP), which is the warmest body of seawater in the world, has received less attention. The IPWP has been warming in recent years. This paper presents for the first time the remarkable nonlinearity in Northern Hemisphere (NH) stratospheric circulation and temperature response to IPWP warming (the so-called IPWP Niño) in boreal winter. The magnitude of NH stratospheric vortex weakening is strong and significant in case of moderate IPWP Niño, but is weak and insignificant in strong IPWP Niño case. This phenomenon is robust in both the historical simulations and observations. An idealized model experiments forced with linear varying SST forcing in the IPWP region isolate the nonlinearities arising from IPWP Niño strength. Westward extension of precipitation into the Maritime Continent drives attenuation and westward shift of extratropical waves during strong IPWP Niño events. Linear wave interference analysis reveals this leads to weak interference between the climatological and anomalous stationary waves and thereby a weak response of the stratospheric vortex. These findings imply a distinct stratospheric vortex response to the IPWP Niño, and provide extended implications for the surface climate in the NH.


Results
To obtain a wide range of IPWP Niño events, we used CMIP5 outputs from a coupled climate system. Considering the good performance of the Community Earth System Model (CESM) Whole Atmosphere Community Climate Model, version 4 (WACCM4) (see Simulations for more information) in representing stratospheric variability, we analyzed simulations using CESM-WACCM4 from historical experiments covering the period 1850-2005 (see the Methods for details). Figure 1 shows scatterplots of the standardized NDJF TI (IPWP) , which indicates the strength of the IPWP Niño, versus the December-January-February (DJF) polar cap temperature anomalies between 70°N and 90°N and from 100 to 50 hPa (Fig. 1a), as well as the DJF zonal mean zonal wind anomalies averaged between 50°N and 70°N and 50 to 10 hPa (Fig. 1b). A clear nonlinearity in the stratospheric response is shown. That is, a parabolic fit of the polar cap temperature anomaly, which can be approximately expressed as a × TI (IPWP) 2 , better describes the relationship between IPWP Niño strength and the stratospheric response than does a linear fit. This is measured by the adjusted R 2 (See the Method section), which is larger in case of polynomial fit (R 2 ≈ 0.20 for both temperature and zonal wind) than in case of linear fit (R 2 = 0.01 for temperature; R 2 = 0.02 for zonal wind). Importantly, an inflection point occurs near the 1σ threshold. When IPWP Niño strength is below the 1σ threshold, the intensity of the polar stratospheric temperature and zonal wind anomalies increases with IPWP Niño strength. The correlation coefficient between IPWP Niño strength and the DJF polar vortex temperatures is r = 0.57, which suggests that the stronger the IPWP Niño event, the warmer the stratospheric anomaly. Correspondingly, the correlation coefficient for the zonal wind anomalies is r = −0.59; i.e., the stronger the IPWP Niño event, the weaker the polar vortex. In addition, the linear fit crossed zero close to, but lower than, the 0.5-SD threshold, supporting the use of 0.5σ to define IPWP Niño winters. However, when IPWP Niño strength exceeds the 1σ threshold, the response reverses so that it weakens with increasing IPWP Niño strength. An inflection point occurs near the 1σ threshold for both stratospheric temperature and circulation response. Now we have seen the nonlinear stratospheric response in CESM-WACCM4 historical experiments. Note that this relationship is based on a single CMIP5 model (CESM-WACCM4), and can be different in other models, though not extended in this paper. However, two questions must be addressed here. Does the nonlinearity really exist? If so, does it come from IPWP Niño strength? Thus, we next investigate the change in stratospheric response with increasing IPWP Niño strength in the observations to validate the nonlinearity in the CMIP5 simulations, and then isolate the signal and tracing the nonlinearities from tropics to extratropical stratosphere based on idealized modeling results.
As the 1σ threshold has been shown to be the turning point of the stratospheric response ( Fig. 1), we used this value to separate all of the IPWP Niño events into two groups: moderate and strong. That is, we defined a strong IPWP Niño as occurring when the winter mean TI (IPWP) exceeds 1σ, with the moderate IPWP Niño group containing all of the remaining IPWP Niño events. The years included in each composite are listed in Table 1. The regional-mean composite SST anomaly during strong IPWP Niño events is obviously larger than that during moderate IPWP Niño events (Fig. 2a,b). We first present composite NH stratospheric temperature and circulation anomalies during moderate and strong IPWP Niño events, for the DJF average based on European Center for Medium Range Weather Forecasting (ECMWF) reanalysis dataset (ERA-Interim) (see Data for more information) for the period 1979-2017 ( Fig. 2c-f). The general structure in the high latitudes during moderate IPWP Niño events resembles those reported by Zhou et al. 41 (their Fig. 2) in terms of a significant stratospheric warming (peaking at about 1 K) together with a robust weakening of zonal mean zonal winds throughout the stratosphere. These significant stratospheric zonal mean temperature and wind anomalies indicate a weakening and warming stratospheric polar vortex. However, during strong IPWP Niño events the stratospheric temperature anomalies show a weak and insignificant cooling rather than a linearly stronger warming (Fig. 2d). Consistent with this, the polar stratospheric zonal winds are not weakened except in the vicinity of stratopause (Fig. 2f). Thus, observational evidence implies that a strong IPWP Niño has a weaker influence than its moderate counterpart.
To overcome the limited availability of observations and isolate the impact of IPWP Niño, we performed five 30-year sensitivity runs (R1-R5; see Simulations) using CESM-WACCM4 to mimic the IPWP Niño SST forcing with linear increasing strength, ranking as weak, moderate, strong, and very strong PWP Niño forcing (Fig. 3a- www.nature.com/scientificreports www.nature.com/scientificreports/ (see Simulations for more information). Following the experimental design by Jimenez-Esteve and Domeisen 32 , Cao et al. 48 , and Lin and Derome 49 , the nonlinearities arising solely through changes in IPWP Niño strength is isolated by forcing a linear change in the amplitude of the SSTs in the IPWP region and by forcing climatological SSTs elsewhere. It should be noted that the prescribed IPWP Niño SST forcing implicitly assumes that the SSTs in the IPWP region is applied entirely as a forcing, although it is likely that some fraction of the SST pattern is generated in response to atmospheric forcing, for example by a modulation of the Pacific Walker Circulation associated with ENSO 42 . We first compare the tropical response to the four linear IPWP Niño forcings. Here the magnitude of outgoing longwave radiation (OLR) response is used as a proxy for the intensity of tropical convection response. Low-level (850 hPa) convergence of zonal winds associated with Kelvin and Rossby waves towards the IPWP region coincides with upper level (250 hPa) divergence, leading to enhanced convection (negative OLR anomalies) over the IPWP region ( Fig. 3e-h). This tropical circulation response is in agreement with previous findings from observations and ideal experiments with the linear baroclinic model 45 . Stronger IPWP Niño forcing induces stronger tropical precipitation anomalies, indicating larger adiabatic heating in the troposphere, acting as the Rossby wave source. Fletcher and Kushner 50 using multiple configurations of atmospheric general circulation models also found an approximately proportional relationship between amplitude of zonally asymmetric components of tropical SST forcing and tropical precipitation response, accept in the Pacific cold tongue region due to thresholds for tropical convection. However, in cases of weak and moderate IPWP Niño precipitation anomalies centered over the equatorial from 150°E to 180°; in strong and very strong IPWP Niño cases, stronger ascent and larger precipitation penetrate deeper into the Maritime Continent.
Many studies have found a large sensitivity of the extratropical response to the location and amplitude of the convective anomalies near the equator 49,51,52 . Goss and Feldstein 53 applied a dynamical core of a climate model to run experiments with the heating field restricted to each of seven small domains located near or over the equator. They found that the heating anomalies over the equatorial Pacific from 150°E to 150°W force an anomalous low over the North Pacific and an anomalous high over the North America. However, heating anomalies over the Maritime Continent and Indian Ocean (50°E-150°E, 15°S-15°N) force opposite-signed extratropical response. Their findings suggest that the extratropical response in cases of strong and very strong IPWP Niño is very likely to have some cancellation between contribution from precipitation anomalies over the east part the IPWP region (150°E-180, 15°S-15°N; domain 1) and contribution from precipitation over the Maritime Continent (the west part of convection; 90°E-150°E, 0-15°N; domain 2). The essence of this argument involves the distinct extratropical response to convection over the two domains. In order to identify basic atmospheric processes associated with hearting in domain 1 and 2 separately, the linear baroclinic model (LBM) is used, with two heating fields restricted in domain 1 and 2 imposed separately in two runs (See Simulations for more information). Figure 4 shows the imposed heating fields and corresponding 300-hPa geopotential height response. The response to heating fields in domain 1 and 2 shows opposite-signed anomalies over the mid-latitude North Pacific and North America regions. The modeled results are in good agreement with findings in Goss and Feldstein 53 . The LBM solutions confirms the opposite-signed response between precipitation anomalies over the east and west part of the IPWP, which is likely to attenuate extratropical Rossby waves during strong IPWP Niño event.  www.nature.com/scientificreports www.nature.com/scientificreports/ Thus, we next show the extratropical geopotential height response to the linear IPWP Niño forcing in Fig. 5. The four cases exhibit similar positive PNA-like wave trains in 250-hPa wave geopotential height field (Z * 250 hPa; Z * indicates that the zonal mean has been removed), with an anomalous deepened Aleutian low and an anomalous American high (Fig. 5). However, the strong and very strong IPWP Niño cases show a weaker negative anomaly in the Aleutian low region with a ~10° westward shift, comparing with weak and moderate cases. This is corresponding to westward extension of convection response in strong and very strong IPWP Niño cases. In addition, the weak amplitude of extratropical response suggests that contribution from the east part of IPWP Niño convection is largely cancelled out by contribution from the west part.
Previous studies have shown that when anomalous extratopical waves excited by tropical SST forcing propagate and dissipate in midlatitudes, the linear interference (phasing) of extratropical planetary waves determines the stratospheric vortex response 9,34,50,54-56 . Thus, we next examine the linear interference of extratropical planetary waves, with the wavenumber-1 and wavenumber-2 component playing the controlling role (Fig. 6). The phase difference between the anomalous wave geopotential height at 60°N (Z * 60°N; Z * indicates that the zonal mean has been removed) and its climatological mean is measured by the pattern correlation r zp (See the Method for details). The entire depth of the troposphere and lower stratosphere is considered, in order to reveal a precursory planetary wave signal from the troposphere 57,58, . The anomalous wave 1 projects weakly onto climatological wave 1 in weak IPWP Niño case (r zp = 0.45), but projects strongly onto climatological wave 1 in moderate IPWP Niño case (r zp = 0.78) (Fig. 6a,b); and the anomalous wave 2 is in quadrature with the climatological wave 2 (r zp = −0.08 for weak IPWP Niño and r zp = −0.09 for moderate IPWP Niño) (Fig. 6e,f). This pattern of positive wave-1 linear interference indicates increased wave activity flux entering the polar stratosphere, and is expected www.nature.com/scientificreports www.nature.com/scientificreports/ to weaken the polar vortex. However, for strong and very strong IPWP Niño the anomalous wave 1 is confined primarily to the stratosphere (Fig. 6c,d) and the anomalous wave project weakly onto the climatological wave for wave 2 (r zp = 0.31 for strong IPWP Niño and r zp = 0.29 for very strong IPWP Niño) (Fig. 5g,h). This wave pattern implies very weak wave activity flux into the stratosphere during strong and very strong IPWP Niño cases, and would corresponds to weakly disturbed polar vortex. Figure 7 presents the simulated stratospheric temperature and circulation response to the four linear IPWP Niño forcings. Consistent with anomalous strengthened wave activities, the moderate IPWP Niño leads to a significant warmer stratosphere at mid-to-high latitudes (Fig. 7b). However, the warming is weaker and insignificant in strong and very IPWP Niño cases, comparing with that in moderate case (Fig. 7c,d). Coherently, the zonal mean winds is markedly weakened during moderate IPWP Niño events (Fig. 7f), whereas the decrease during strong and very IPWP Niño events is not statistically significant (Fig. 7g,h). This characteristic of the stratospheric temperature and winds clearly validate the nonlinearity in the stratospheric response to IPWP Niño, which has been identified in CMIP5 simulations and observations above.

conclusions and Discussion
The combination of analyses using the historical relationship, reanalysis composites, and idealized experiments allows us to draw conclusion on the nonlinearity of IPWP Niño's impacts on the NH stratospheric vortex. Anomalous warming SST associated with IPWP Niño launches precipitation over this area, which drives extratropical waves that further weaken the polar vortex. However, a nonlinear relationship between the amplitude of the IPWP Niño strength and the NH stratospheric vortex response is identified.
When the strength of the IPWP Niño is below the 1σ threshold, the intensity of the polar stratospheric temperature and zonal wind anomalies increases with increasing IPWP Niño strength. However, the response reverses and becomes weaker when IPWP Niño strength exceeds the 1σ threshold. This nonlinear relationship between IPWP SST anomalies and the NH stratospheric response is seen in CESM-WACCM4 CMIP5 historical simulations from 1850 to 2005. As 1σ is the point at which nonlinearity begins to develop in the stratospheric response, we used 1σ as the threshold to separate the IPWP Niño events into two groups (moderate and strong). By comparing the composition of the anomalous stratospheric temperature and circulations during moderate and strong IPWP Niño events, we found that the stratospheric circulation and temperature response is weaker during strong IPWP Niño events than during moderate events based on ERA-Interim for the period 1979-2017. That is, the stratospheric polar vortex is significantly warmer and weaker during moderate IPWP Niño events, whereas there is no significant signal during strong IPWP Niño events in the observations.
Idealized model experiments with scaled IPWP Niño SST forcing have been performed to isolate the signal from IPWP Niño and to validate the nonlinearities arising from IPWP Niño strength. After the westward extension of precipitation into the Maritime Continent, there is some cancellation between the extratropical response to the west part convection over the Maritime Continent and the response to convection over the east warm pool. This leading to the attenuation of extratropical wave response, with a ~10° westward shift in the Aleutian low region during strong IPWP Niño events. Thus, the wave-1 component moves from strong in phase with the climatological wave in moderate IPWP Niño case into rather weak in phase in strong IPWP Niño case and is mostly confined in upper stratosphere. This linear wave interference produces a significant vortex response in moderate IPWP Niño case but a very weak response in strong IPWP Niño case.  www.nature.com/scientificreports www.nature.com/scientificreports/ The monthly output for the period 1850-2005 from historical simulations by CESM-WACCM4 in the CMIP5 archive (http://cmip-pcmdi.llnl.gov/cmip5/availity.html) allowed us to examine the stratospheric response to a wide range of IPWP Niño events. CESM-WACCM4 uses active ocean and sea ice components, and the model is forced using observed atmospheric composition changes from both natural (e.g., solar irradiance and volcanic aerosols) and anthropogenic (e.g., greenhouse gases, sulfate aerosols, and ozone) sources. The atmospheric component used was WACCM4. A representation of the QBO was achieved by relaxing the equatorial zonal wind between 86 and 4 hPa toward that observed 58 . To avoid the possible entanglement of QBO signals in our composite results from the fully coupled model, we regressed out the zonal wind at 50 hPa.

Methods
We calculated the monthly anomalies by subtracting the long-term mean of each calendar month from each individual month. The linear trends were removed before analysis from the temperature, zonal wind, and geopotential height data using linear regression analysis. The quasi-biennial oscillation (QBO) index used was the standardized anomaly of equatorial 50-hPa zonal winds and it was used to regress out the QBO signal. The Niño 3.4 index was defined as the area mean SST anomaly over the region 5°S-5°N, 150°W-90°W (http://www.cpc. noaa.gov/data/indices/). All statistical tests were performed using the two-tailed Student's t-test. To better represent the reversing manifestation of the nonlinearities, the composite results and model results are not scaled by IPWP Niño strength. www.nature.com/scientificreports www.nature.com/scientificreports/ To diagnose the linear interference of Rossby waves, we used the pressure-weighted correlation r zp between the anomalous wave geopotential height at 60 N (Z * 60°N; Z * indicates that the zonal mean has been removed) and its climatological mean, following the framework of Fletcher and Cassou 34 . The weights are based on the relative thickness of each of the vertical layers from 700 to 10 hPa. Since the The correlation is computed separately for the zonal wavenumber 1 and 2 components of Z * 60°N, which are filtered using a Fourier transform.
The adjusted R 2 (Eq. 3.30 of Chatterjee and Hadi 62 ) is used to quantify the added value in using a polynomial best fit (e.g., T ~ a × TI (IPWP) 2 ) instead of a linear best fit (e.g., T ~ b × TI (IPWP) ). The adjusted R 2 takes into account the likehood that a polynomial predictor will reduce the residuals by unphysically over-fitting the data. The polynomial fit could be preferred if the adjusted R 2 for the polynomial fit is larger as compared to the linear R 2 .
Simulations. The time-slice simulations were performed using CESM-WACCM4. The CESM-WACCM4, developed by the National Center for Atmospheric Research (NCAR), has 66 vertical levels extending from the ground to 4.5 × 10 −6 hPa (~145 km geometric altitude), with vertical resolution of 1.1-1.4 km in both the TTL and the lower stratosphere (<30 km). It is unable to internally simulate the QBO signals but is forced using a 28-month fixed cycle (nudged QBO). The time-slice simulations presented in this paper were performed at a resolution of 1.9° × 2.5°, with interactive chemistry. www.nature.com/scientificreports www.nature.com/scientificreports/ SST anomalous pattern for moderate IPWP Niño (Fig. 2a) is multiplied by factors of 0.5, 1.0, 1.5, and 2.0 to generate weak (R2), moderate (R3), strong (R4) and very strong (R5) IPWP Niño forcing, respectively. The anomalous SST pattern is imposed in the IPWP region (15°S-15°N, 90-160°E), and SST forcing is set to zero outside of the IPWP region. To prevent discontinuities in SST forcing on the IPWP boundary, SST anomalies on the boundary are added to the five model grids centered at the IPWP boundary, with respective weights of 1.0, 0.75, 0.50, 0.25, and 0.0 from the inside to the outside model grids of the IPWP boundary. The experiments are each run for 33 years, removing the first 3 years as spin-up. The key point is that these model integrations isolate possible nonlinearities arising solely through changes in IPWP Niño strength.
The linear baroclinic model (LBM) is used as a diagnostic tool for studying the extratropical atmospheric response to idealized forcing 46,[63][64][65] . It is constructed by linearizing the primitive equations about a 3D climatological basic state, with a T42 horizontal spectral resolution and 20 vertical levels on a sigma coordinate. The LBM is fully described in Watanabe and Kimoto 66 and Watanabe and Jin 67 . In this paper, the heating fields restricted in domain 1 and 2 are imposed on boreal winter mean climatology derived from NCEP-NCAR reanalysis, with elliptical cosine-squared horizontal distribution shown in Fig. 3 and gamma vertical profile peaking at 400 hPa 64,68 . The LBM solutions provide evidence for the cancellation effect between the two domains in a linear framework.