On the robustness of the solar cycle signal in the Pacific region



[1] The potential role of the stratosphere for the 11-year solar cycle signal in the Pacific region is investigated by idealized simulations using a coupled atmosphere-ocean general circulation model. The model includes a detailed representation of the stratosphere and accounts for changes in stratospheric heating rates from prescribed time dependent variations of ozone and spectrally high resolved solar irradiance. Three transient simulations are performed spanning 21 solar cycles each. The simulations use slightly different ozone perturbations representing uncertainties of solar induced ozone variations. The model reproduces the main features of the 20th century observed solar response. A persistent mean sea level pressure response to solar forcing is found for the eastern North Pacific extending over North America. Moreover, there is evidence for a La Niña-like response assigned to solar maximum conditions with below normal SSTs in the equatorial eastern Pacific, reduced equatorial precipitation, enhanced off-equatorial precipitation and an El Niño-like response a couple of years later, thus confirming the response to solar forcing at the surface seen in earlier studies. The amplitude of the solar signal in the Pacific region depends to a great extent on the choice of the centennial period averaged.

1. Introduction

[2] It has been suggested that large scale near surface climate variability during the 20th century is related to the 11-year cycle of the sun [White and Tourree, 2003]. The quasi decadal oscillation (QDO) reveals similar spatial characteristics as the El Niño-Southern Oscillation (ENSO) and is similarly governed by a delayed action oscillator mechanism in the tropical Pacific [White and Tourree, 2003; White et al., 2003]. While ENSO associated with 3- to 7-year period variability is an internally generated mode of the coupled ocean-atmosphere system, model studies indicate that solar forcing is necessary to generate the QDO of 9- to 13-year period [White and Liu, 2008a]. Moreover, there is evidence for a phase lock between QDO, ENSO type variability and the 11-year solar cycle resulting in a distinct temporal evolution of the solar signal [White and Liu, 2008a, 2008b]. Based on observations spanning the period from the late 19th century to present, van Loon et al. [2004, 2007] find a La Niña like response with lower sea surface temperatures (SST) in the eastern equatorial Pacific mainly for solar maximum peak years. Meehl et al. [2008] confirmed a proposed mechanism on the basis of ensemble experiments with two different ocean-atmosphere general circulation models (AO-GCM). The resulting ensemble mean response patterns are similar to the observations in the Pacific region but the amplitude is only about half the magnitude of the observed response. A possible explanation for this underestimation is the neglect of stratospheric forcing and coupling mechanisms [e.g., Shindell et al., 2006].

[3] Coupled chemistry-climate models (CCM) have so far been able to simulate important features of the stratospheric solar signal [e.g., Marsh et al., 2007]. In a recent study Meehl et al. [2009] successfully reproduce the strength of the observed response in the tropical Pacific region when employing a CCM coupled to a deep ocean model. However, their simulation reveals some discrepancies with respect to the exact shape and temporal evolution of the response. As their conclusions solely rely on a single realization with only one model, important aspects that need to be addressed are the role of (i) internal variability and (ii) ozone related sensitivities for the simulated/observed signals. In the present study we assess the associated uncertainties based on an ensemble of idealized simulations performed with a stratosphere resolving AO-GCM.

2. Model and Experiment

[4] A vertically extended version of a fully coupled AO-GCM (EGMAM) including a high resolving spectral short wave radiation scheme (FUBRad) is used. EGMAM (ECHO-G with Middle Atmosphere Model) [Legutke and Voss, 1999; Manzini and McFarlane, 1998] has already been utilized to analyse variations within the climate system [e.g., Huebener et al., 2007; Körper et al., 2009; Spangehl et al., 2010; Schimanke et al., 2011]. The atmospheric model component (MA-ECHAM4) incorporates a T30 horizontal and L39 vertical resolution with the top level located at 0.01 hPa (80 km). For better representation of the direct solar signal in the stratosphere the radiation code is partly substituted by the FUBRad radiation code [Nissen et al., 2007] which explicitly treats 49 intervals in the UV/visible part of the spectrum (121.56 to 683 nm) at levels above 70 hPa. The model does not generate a QBO resulting in predominately weak easterly winds in the tropical lower stratosphere. The coupled ocean model has a horizontal resolution of 2.8° with equator refinement and 21 vertical levels. A flux correction is used for heat and freshwater exchange to prevent climate drift. The model simulates ENSO with a maximum peak periodicity of about 2 years.

[5] EGMAM-FUBRad is used to perform idealized solar cycle simulations employing a sinusoidal forcing function with a period of 11 years. Following White and Liu [2008a, 2008b], an amplitude of ±1.25 W/m2 is used for the total solar irradiance (TSI) which is twice the observed variability. Time dependent changes in the UV/visible part of the spectrum are additionally introduced via a scaling of high resolved observation based fluxes [Lean, 2000] with the 10.7 cm radio flux. The same sinusoidal forcing function is applied using a maximum and minimum of 270 and 70 sfu (solar flux unit) respectively which is close to observations. Since the model does not include interactive chemistry, the sinusoidal forcing function is used for scaling of fixed solar maximum minus minimum ozone anomalies providing a time dependent ozone forcing. Climatological ozone concentrations [Fortuin and Langematz, 1994] are used as a background.

[6] Three simulations are performed which only differ in the used ozone anomalies spanning a small ensemble (S1, S2, S3) representing ozone related uncertainties [Soukharev and Hood, 2006]. The ozone anomalies used are: for (S1) offline calculations after Haigh [1994], for (S2) observations [Soukharev and Hood, 2006] and for (S3) observations extended with offline calculations at high latitudes. Note that ozone anomalies for the polar cap are treated differently in S2 and S3 as observations are not available for high latitudes. Model based anomalies (S1) resolve the annual cycle whereas the observation based anomalies (S2 and S3) are only available as annual mean fields. Each simulation covers 21 solar cycles with identical forcing for each cycle. Anomalies are computed by taking solar maximum years minus climatology and solar maximum minus solar minimum (defined as the 5th year after maximum) years. When calculating ensemble mean composites we omit the first 5 solar cycles of each simulation to avoid spin-up effects. Results are shown for solar maximum minus minimum years where simulations give stronger signals while Meehl et al. [2009] show solar maximum minus climatology. Additionally, in Figure 3 solar maximum minus climatology is displayed. The focus is on boreal winter season (DJF).

3. Results

[7] The zonal mean distribution of the ensemble mean atmosphere response to solar forcing is shown in Figure 1. Strongest warming is found around the stratopause and in the higher mesosphere (Figure 1a). The associated differential solar heating forms a meridional temperature gradient anomaly and strengthens the stratospheric subtropical jet. Following Kodera and Kuroda [2002] this varies the deflection and breaking of upward propagating planetary waves leading to a weakening of the Brewer-Dobson circulation. Note that the associated warming in the tropical lower stratosphere is also captured by the model (warming around 0.6 K at 70 hPa, cf. Figure 1a) though slightly underestimated when compared to the observations [cf. Matthes et al., 2004]. Cooling is found in the Northern Hemisphere (NH) polar lower stratosphere indicating a strengthening of the NH polar winter vortex. It should be noted that statistically significant cooling is only found in S1. Here the model also simulates a clear poleward and downward propagation of the zonal wind anomalies during the winter season (not shown). Compared to the observations this signal reveals a one month delay during mid-winter, which is presumably related to a prolonged radiation controlled state in the model [Kodera and Kuroda, 2002]. The ensemble mean response for zonal mean zonal wind shows a clear strengthening of the stratospheric polar winter vortex. Moreover, the model simulates an increase of the trade winds in the tropics/subtopics including a strengthening/poleward extension of the subtropical high pressure systems which is also manifested in the stronger westerly winds around 40°N (Figure 1b). Together with the changes in zonal mean precipitation (Figure 1c) there is also indication for a broadening of the Hadley circulation as suggested by Meehl et al. [2009]. Moreover, the ensemble mean anomaly for precipitation is related to a La Niña like response (cf. Figure 2c). Here the absence of a clear tropical precipitation signal in S2 (which is related to a weakening of the northern branch of the Hadley circulation in this simulation) is compensated by a comparably strong signal in S1 resembling observations. While the climatological distribution is generally close to the observations the positive anomalies are located too close to the equator which is related to a common deficit of GCMs in simulation of tropical/subtrobical precipitation variability [cf. Meehl et al., 2009]. Note that the zonal mean anomalies at SH mid-latitudes additionally indicate a southward shift of storm tracks.

Figure 1.

Solar cycle anomalies for winter mean season (DJF) in simulations with EGMAM. Solar maximum minus minimum ensemble mean anomalies for (a) zonal mean temperature [K] and (b) zonal mean zonal wind [m/s]. Statistical significance above the 95th/90th percentile value after t-test is denoted by black/grey dots. (c) Solar cycle anomalies (solid) of zonal mean of global precipitation [mm/day] in simulations (different colors). Values for observations (black) and climatology (dotted, units are [mm/day/10]) are shown additionally. Thick red curve denotes ensemble mean. Contrary to the simulated anomalies the observed anomalies are calculated as solar maximum minus climatology for JF according to Meehl et al. [2009] (3 peak solar years available, GPCP Precipitation data (GPCP version 2 is used) provided by the NOAA/OAR/ESRL PSD, Boulder, Colorado, USA, from their Web site at http://www.esrl.noaa.gov/psd/). Ensemble mean anomalies consist of 48 maximum/minimum peak years. Anomalies for the individual simulations (S1, S2, S3) consist of 16 maximum/minimum peak years each.

Figure 2.

Solar cycle anomalies of SST [K] and MSLP [hPa] for winter mean season (DJF) in observations and simulations. The observed (a) SST and (b) MSLP anomalies are shown as solar maximum minus climatology according to Meehl et al. [2009] (NOAA Extended Reconstructed SST (version V3b is used) and Hadley SLP data provided by the NOAA/OAR/ESRL PSD, Boulder, Colorado, USA, from their Web site at http://www.esrl.noaa.gov/psd/). Ensemble mean anomalies for (c) SST and (d) MSLP are shown for solar maximum minus solar minimum peak years and consist of 48 maximum/minimum peak years each. Anomalies for a selected centennial period (including 10 maximum/minimum peak years each, taken from S1) for (e) SST and (f) MSLP are also shown as solar maximum minus minimum. Statistical significance above the 95th percentile value after t-test (Welch's correction used for observations and centennial composites) is denoted by black plus signs. Black box in Figure 2d indicates area used for MSLP time coefficient approach (cf. text).

[8] Consistent response patterns are also found for SST and mean sea level pressure (MSLP). Statistically significant warming appears in the North Pacific reaching about 0.4 K (Figure 2c) which is accompanied by statistically significant positive MSLP anomalies in the region of the eastern North Pacific extending over North America (Figure 2d). Statistically significant positive MSLP anomalies also emerge in the equatorial eastern Pacific where they go along with a strengthening of the Walker circulation and statistically significant negative SST anomalies (Figures 2c and 2d). Though the SST and MSLP patterns are similar to observations (Figures 2a and 2b) the amplitude of the ensemble mean response to solar forcing might be lower even when compared with other models [e.g., Meehl et al., 2008, 2009] (remember that we show maximum minus minimum).

[9] Similar response patterns but with larger amplitudes are found for selected centennial periods. In the equatorial Pacific negative SST anomalies of up to 1 K slightly exceed the solar maximum anomalies from observations (Figures 2a and 2e). Statistically significant warming above 0.6 K (Figure 2e) and a positive MSLP anomaly of up to about 4 hPa (Figure 2f) in the subtropical North Pacific resemble the observed solar maximum anomalies though the positive MSLP anomaly is slightly displaced to the east. A simple time coefficient approach is applied to assess the complete range of the simulated response to solar forcing in the Pacific basin arising from individual centennial periods. Anomaly mean composites Cano = Cpos − Cneg are constructed according to

equation image

For SST, VALdjf(Yi) is the winter mean Nino3 index for year Yi. Each Cano spans 105 winters and a total of 11*21-1-104 = 126 values can be calculated for each simulation. The time coefficient is simply written as SSTTC(i) = Cano(i), i = 1, …, 126. For MSLP, VALdjf(Yi) is the winter mean MSLP field and the time coefficient MSLPTC(i), i = 1, …126 is calculated by projection of Cano(i) onto the ensemble mean anomaly shown in Figure 2d (area used for projection: 150°W–70°W, 20°N–60°N).

[10] For solar maximum minus minimum samples SSTTC varies between −0.84 and −0.30 K in S3 (denoted by dark blue dots, cf. Figure 3e), indicating a consistent equatorial cooling. Such negative SSTTC with variable amplitude is also found for the largest part of S1 and the final part of S2 (Figures 3a and 3c). For direct comparison with the solar maximum signal found in observations [van Loon et al., 2007] the SSTpos can be taken (for this Nino3 based index it gives the anomalies w.r.t. climatology, indicated by light blue dots in Figure 3). Here a similar tendency toward equatorial cooling with variable but generally lower amplitude (minimum around −0.37 K) is found. However, it should be noted that even stronger equatorial cooling (SSTTC reaching −1 K) is associated with moderate solar activity (Figure 3a). Moreover, even larger positive SSTTC values are found with the absolute maximum reaching +1.24 K. Such large positive SSTTC anomalies are assigned to the 3rd winter after solar maximum conditions and there is also clear evidence for a general tendency toward positive SSTTC anomalies a few years after solar maximum conditions (Figures 3a, 3c, and 3e). This result is similar to observations which also show lagged positive SST anomalies in the equatorial Pacific [cf. White and Liu, 2008b; Meehl and Arblaster, 2009]. It has to be stressed that the 2-year ENSO peak simulated by the model partly contributes to the close relationship between cooling for solar maximum and the 2nd winter thereafter and warming the 1st and 3rd winter after solar maximum.

Figure 3.

Time coefficients of 11-year solar cycle winter (DJF) mean anomaly composites over 105-year periods. Anomalies are calculated as the difference of the positive and the 5-year lagged negative composite (further information see text). SST time coefficient values (SSTTC) for individual simulations (a) S1, (c) S2 and (e) S3 represent Nino3-index mean anomalies [K] according to sampling. MSLP time coefficient values (MSLPTC) for (b) S1, (d) S2 and (f) S3 are obtained by projection of the mean fields from the 105-years anomaly composites onto the ensemble mean anomaly field shown in Figure 2d (area chosen is 150°W–70°W, 20°N–60°N, series are normalized by the ensemble mean anomaly vector length, area weighting is applied). Dark blue dots show time coefficient values assigned to solar maximum years. Light blue dots in Figures 3a, 3c, and 3e additionally represent the according solar maximum SST anomalies when compared to climatology (i.e. SSTpos, cf. text). Red dots indicate time coefficients assigned to the 3rd winter after solar maximum. Purple bars in Figures 3a and 3b mark centennial period shown in Figures 2e and 2f. Grey curve shows solar forcing.

[11] Even stronger evidence for a direct solar impact is found for MSLP. All three MSLP time coefficient series show a distinct periodicity and clear correlation with the solar forcing (Figures 3b, 3d, and 3f). This signal needs some time to amplify indicating a spin-up effect. Despite the absence of a clear ENSO like response for SST, S2 reveals a very harmonic MSLP response. By contrast the simulated MSLP response in S1 and S3 is partly related to ENSO. Though the solar response to solar forcing appears to be more persistent for MSLP the amplitude of the solar signal in the Pacific sector varies with the choice of the centennial period.

4. Conclusions

[12] The present study investigates the climate response to the 11-year cycle of the sun in the Pacific region by means of idealized simulations with a stratosphere resolving AO-GCM for boreal winter season. Differences in the shape of the ozone anomalies used for the different simulations affect the stratospheric solar signal to some extent. Still the model is found to reproduce important features of the observed response [e.g., Kodera and Kuroda, 2002]. The simulated response to solar forcing includes a broadening of the Hadley circulation, an intensification of subtropical high pressure systems and a strengthening of the trade winds, which is similar to results described by Meehl et al. [2008] and van Loon et al. [2007].

[13] Our results suggest an ENSO-like response to the 11-year solar cycle that includes a La Niña-like pattern assigned to solar maximum conditions with below normal SSTs in the equatorial eastern Pacific, reduced equatorial precipitation, enhanced off-equatorial precipitation, persistently high MSLP over the eastern North Pacific extending over North America and an El-Niño like response a couple of years later. However, these signals are small, but are most statistically significant when there are many samples. Between individual 100-year periods, their amplitude varies greatly. In a few cases the temperature response in the equatorial eastern Pacific, which is a region of large internally generated variability, is even of opposite sign.

[14] Differences in ozone forcing might additionally affect the simulated response. However, the amount of cooling in the equatorial Pacific found for individual centennial periods taken from a single simulation varies by a factor of more than two, which can not be traced back to differences in ozone forcing but indicates that internally generated variability affects the simulated response patterns. Despite the variable SST response in the equatorial Pacific all three simulations show a persistent relationship between solar activity and MSLP for the eastern North Pacific which is similar to observations as shown by Roy and Haigh [2010].

[15] We conclude that further investigations are necessary to quantify the potential contribution of stratospheric dynamics on tropical SST variability. This requires long simulations in an ensemble mode with different stratosphere resolving AO-GCMs which preferably include realistic representation of the 3- to 7 year ENSO variability. Nevertheless, results from this study show the first independent confirmation of the observed response to solar forcing in an AOGCM outside the NCAR modeling framework [e.g., Meehl et al., 2008, 2009], and thus provide greater confidence in this response to the 11-year solar cycle at the Earth's surface.


[16] Parts of this work were funded by the DFG Project ProSECCO/CAWSES. S. Bal was funded by Erasmus Mundus. The simulations were performed on a NEC-SX6 supercomputer at “Deutsches Klimarechenzentrum” (DKRZ), Hamburg. The authors are grateful to the reviewers for their valuable comments.

[17] The Editor thanks the two anonymous reviewers.