SEARCH

SEARCH BY CITATION

Keywords:

  • Solar activity;
  • middle atmosphere;
  • planetary wave

Abstract

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. The Solar Modulation of the Upper Tropospheric Planetary Waves
  5. 3. Model and Experiments
  6. 4. Model Results
  7. 5. Summary and Conclusions
  8. Acknowledgments
  9. References

[1] To assess a possible solar modulation of the planetary waves, the 300 hPa geopotential height of the two major reanalysis data sets, NCEP/NCAR and ERA40, is studied for the difference between years of solar maximum and minimum activity for three solar cycles in the period 1963–1999. We find a significant positive difference over the North Pacific ocean for December through February which resembles that found in sea level pressures in earlier studies. Furthermore, there are statistically significant amplitude differences in the climatological state of the first two planetary waves extracted from the geopotential heights. We also perform a model study of the middle atmospheric response to the tropospheric solar signal. The 300 hPa geopotential height is used as the lower boundary in a middle atmosphere model which is run with a constant solar cycle minimum radiative forcing, so that only the lower boundary is forcing the middle atmosphere. The zonal wind and temperature at 300–0.01 hPa from the model experiment are studied in boreal winter for the same years that were studied in the geopotential height analysis. Differences of the means of the two data sets reproduce the structure and amplitude seen in observations and in reanalysis data. Whether this is a direct effect of the solar forcing of the troposphere or a feedback coupling of the solar forcing of the stratosphere is not presently clear, but the modulation of the tropospheric planetary waves seems to be important for the observed solar modulation of the stratosphere.

1. Introduction

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. The Solar Modulation of the Upper Tropospheric Planetary Waves
  5. 3. Model and Experiments
  6. 4. Model Results
  7. 5. Summary and Conclusions
  8. Acknowledgments
  9. References

[2] Many observational as well as model studies of the middle atmosphere have revealed significant signals attributed to the decadal solar cycle [Haigh, 1999; van Loon and Labitzke, 2000; Kodera and Kuroda, 2002; Gray et al., 2003; Labitzke, 2003; Matthes et al., 2004, 2006] and there are significant correlations between the mid-stratospheric geopotential height and the solar cycle for both boreal winter [van Loon and Labitzke, 2000] and summer [Labitzke, 2003]. The magnitude of the solar signal generally decreases with decreasing altitude but statistically significant signals have been observed also in the troposphere [Haigh, 2003; Gleisner and Thejll, 2003; Haigh et al., 2005; Crooks and Gray, 2005]. At mid to high latitudes, the solar signal shows a stratospheric pattern of poleward and downward propagation of zonal mean zonal wind and zonal mean temperature [Kodera and Kuroda, 2002], which only recently has been simulated in a General Circulation Model (GCM) [Matthes et al., 2004, 2006]. Could a solar cycle modulation of the tropospheric planetary waves be the driving mechanism for this stratospheric pattern?

[3] The total solar irradiance forcing related to the decadal solar cycle varies by about 0.1%, with 1–8% in UV depending on the wavelength [Lean, 1997]. The visible solar forcing of the Earth's surface has been suggested to lead to localized heat anomalies [Meehl et al., 2003] which in turn could modulate the planetary waves in the troposphere [Held et al., 2002], but this effect is believed to be small due to the weak irradiance changes at visible wavelengths. Non-irradiance forcings such as a feedback through a modulation of clouds has been suggested to amplify the solar signal in the troposphere [Pudovkin and Veretenenko, 1995; Svensmark and Friis-Christensen, 1997]. The UV radiative forcing is strongest at the stratopause, where the absorption of UV by ozone is the most effective. The increased UV heating at solar maximum activity changes the temperature gradient between the tropics and the winter polar region, which gives rise to secondary dynamical effects such as a relatively stronger polar night jet (PNJ) for maximum solar activity [Kodera and Kuroda, 2002]. This secondary effect is enhanced by a change in the stratospheric ozone concentration as a consequence of the UV forcing [Haigh, 1994, 1999]. A stronger PNJ deflects the upward propagating planetary waves and thereby affects the wave breaking in the stratosphere. This has consequences for the strength of the polar vortex and for the Brewer-Dobson circulation [Kodera and Kuroda, 2002]. The significant changes in the solar forcing of the middle atmosphere could lead to changes in the tropospheric circulation [Haigh, 1999; Haigh et al., 2005], e.g., through changes in the downward propagation of stratospheric anomalies to the troposphere [Kodera and Kuroda, 2002; Matthes et al., 2006]. If for example this leads to a shift in the stormtracks, the resulting change in the diabatic heating could affect the planetary waves [Held et al., 2002]. The thickness of the troposphere has been shown to be correlated to the decadal solar signal [Gleisner and Thejll, 2003] which suggests that there might also be a solar modulation of the planetary waves in the troposphere.

[4] In boreal winter the stratospheric conditions are set by a balance between radiative relaxation to a meridionally varying temperature distribution and a dynamical forcing due to upward propagation and dissipation of tropospheric planetary waves. These waves are the main source of variability in the winter polar middle atmosphere through wave mean-flow interaction [Kodera, 1995; Christiansen, 2001]. As the planetary waves propagate into the middle atmosphere they break and decelerate the westerly zonal wind. This weakens the polar night jet and thus also the polar vortex, and increases the possibility for having a stratospheric sudden warming [e.g., Andrews et al., 1987].

[5] Gray et al. [2003] used a middle-atmosphere model and found that the influence of dynamical stratospheric phenomena on the winter stratosphere depends on the strength of the planetary wave forcing. Under idealized wave number one forcings they showed that for small wave amplitudes the model stratosphere is undisturbed, while for intermediate to high amplitudes the stratosphere has frequent major warmings. They further argued that for the intermediate wave amplitudes, which is the common state for the boreal winter, there is more opportunity for other stratospheric influences such as the QBO and the solar cycle to affect the boreal winter stratosphere. Matthes et al. [2004] used an atmospheric GCM forced by solar irradiance, and with the equatorial winds relaxed to observations. The GCM simulated the observed patterns of stratospheric anomalies in the northern polar region in winter [Kodera and Kuroda, 2002] but the model response was weaker. The weaker response was argued to be due to a weaker variability in the GCM and therefore weaker magnitudes, but also a weaker SAO was suggested as a cause for the deviations. We argue that another possibility could be that the changes in the planetary wave forcing of the model simulation is too weak.

[6] In this paper we investigate the decadal solar signal in the 300 hPa geopotential heights using reanalyses data from NCEP/NCAR and ECMWF (ERA40). We study the solar signal in each of the first three planetary waves extracted from the geopotential heights, and find significant differences in the amplitude of the first two planetary waves between solar maximum and minimum solar radiative activity. To assess the importance of the changes in the planetary waves, a mechanistic middle atmosphere model is forced by the investigated reanalyses data. In this study only the lower boundary variability is forcing the model, i.e., there are no irradiance or ozone forcings in the experiment. When forced this way the model reproduces the structure and strength of the solar modulation of the winter stratosphere.

[7] In Section 2 we show the results for the study of the solar cycle signal in the troposphere of reanalyses data. The model we use is presented in Section 3 and in Section 4 we present the results of the model study.

2. The Solar Modulation of the Upper Tropospheric Planetary Waves

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. The Solar Modulation of the Upper Tropospheric Planetary Waves
  5. 3. Model and Experiments
  6. 4. Model Results
  7. 5. Summary and Conclusions
  8. Acknowledgments
  9. References

[8] In this section we study the decadal solar cycle in the 300 hPa geopotential height of the two major reanalysis data sets, i.e., the NCEP/NCAR and the ERA40 for the solar activity of the period 1965–1997. We do this by calculating the difference in the temporal mean state of the 300 hPa geopotential height for the high and low solar activity years for December through February. We select the years when the Sun is in solar radiative maximum and in minimum activity, i.e., when the Sun has its highest and lowest irradiative output. To select the maximum and minimum activity years also in the pre-satellite era, we use the 10.7 cm solar radio flux which is a well known proxy for solar irradiance (Figure 1, left panel). There are some differences between the 10.7 cm flux and irradiance but by using yearly mean values and by making sure that there is at least a one year gap between intervals of maximum and minimum solar activity in our data, so that intermediate states are avoided, these differences become negligible. The El Niño Southern Oscillation (ENSO) has a strong impact on the whole troposphere and has been shown to vary with the decadal solar cycle [Mendoza et al., 1991], so years with extreme ENSO activity are removed from the data set to avoid conflicting signals. We set the limit for a strong ENSO year when the absolute value of the global sea surface temperature anomaly exceeds 0.03 K according to the global ENSO index (Figure 1, right panel), taken from JISAO (Joint Institute for the Study of the Atmosphere and Ocean). The impact of other ENSO indices such as the CTI index (JISAO) has also been investigated and we find that the results presented in this paper are not sensitive to the ENSO index chosen, as the differences between the indices are small in this case since we use mean values for each winter. The results are not much affected by the choice of ENSO index. Large volcanic eruptions have a strong impact on the lower stratosphere and upper troposphere so we remove also the years after the large El Chichon and Pinatubo eruptions which occurred at solar cycle maximum activity [McCormick et al., 1995] and may therefore impose conflicting signals. The resulting nine solar maximum years in our study are then 1968, 1971, 1979–1982, and 1989–1991, and the nine solar minimum years are 1965, 1975, 1977, 1985–1987, 1995–1997. The selected years for the period 1978–1997 from the 10.7 cm solar flux agrees with the years that would be selected from the irradiance cycle [Frölich, 2005] during this period.

image

Figure 1. (left) The solar 10.7 cm flux for the period 1963–1999 with the selection criteria for maximum and minimum solar activity denoted by horizontal bars. (right) The ENSO index (JISAO SST ENSO index) for the same period with horizontal bars denoting the selection criteria for extreme ENSO events.

Download figure to PowerPoint

[9] Since the two data sets are of only nine years each the statistical analysis suffer from having few independent data points. We therefore use a Monte Carlo surrogate data method which uses the full data set. From the period of 37 years, we randomly select pairs of nine years each that do not include the extreme ENSO years or years with volcanic forcing, as explained above. This is performed over 10,000 times and the unique pairs are used as surrogates of solar maximum and minimum activity cycles. We are testing for a significant difference of the means of the two time series, so we compare the absolute value of the difference between the observed, or modeled, time series to the absolute difference values of the surrogate time series. The significance level is calculated as the fraction of surrogate time series with an absolute value higher than that of the observed, or modeled, signal.

[10] To investigate the planetary waves we use a Fourier filtering that retains only the mean geopotential height and the first three planetary waves, which allows us to study the planetary wave components one by one. The results for the NCEP/NCAR and the ERA40 reanalysis are practically identical so we present only the NCEP/NCAR data.

2.1. The Climatological Mean State

[11] The climatological temporal mean state of the planetary waves and the mean geopotential height for years of minimum solar activity does not vary much in boreal winter. Planetary wave number one shows a peak amplitude of about 120–140 gpm, with somewhat higher values in January compared to the other months (Figures 2d–2f). The trough is located over the western North Pacific. Planetary wave number two has a somewhat higher amplitude in January and February, 120–130 gpm, compared to December, 90–120 gpm, and shows a weak shift toward the north (Figures 2g–2i). The troughs are located over eastern Russia and eastern Canada. The same pattern of lower amplitude in December and higher in January and February is seen also for planetary wave number three (not shown). The peak amplitude of wave number three is of about 80–100 gpm and the troughs are located over the western North Pacific, eastern Europe and over eastern USA/Canada.

image

Figure 2. Climatological 300 hPa geopotential heights of the NCEP/NCAR reanalysis [gpm] in solar minimum years for December (top), January (middle) and February (bottom). From left to right, the panels show the total geopotential height for waves one to three and the mean height (a–c), planetary wave number one (d–f) and planetary wave number two (g–i). Wave amplitudes larger than ±90 gpm are shaded. The maps are in a polar stereographic projection from 20–90°N.

Download figure to PowerPoint

2.2. The Solar Modulation of the Geopotential Height

[12] Here we show the results for the analysis of the differences between the geopotential heights for solar maximum and minimum activity. The difference between the results for the unfiltered data (not shown) and the sum of the mean and the first three planetary waves (Figures 3a–3c) are very similar with similar significant regions, except for the filtered data being smoother. We therefore limit the presentation of the solar cycle changes to the filtered data because it will later be used in a model study. The largest difference between solar maximum and minimum years is in the North Pacific where there is a positive difference of 70–100 gpm, which prevails throughout the boreal winter (Figures 3a–3c). The peak of the pattern moves from the mid North Pacific in December to the western part of the ocean in February. An earlier study showed a similar pattern in the observed sea level pressure of the North Pacific for the peak years of the solar cycle compared to the climatology [van Loon et al., 2007]. The peak difference in the sea level pressure is centered in the Gulf of Alaska, while the peak of the 300 hPa geopotential height differences presented here are further to the west. A GCM study of solar UV radiative activity showed a similar positive difference in geopotential heights at 1000 hPa in the North Pacific for December and January, but the sign of the difference reverses in February [Matthes et al., 2006]. Whether the GCM study correctly reproduces this pattern is difficult to tell, but it could be that this change in the sea level pressure is important for generating the solar induced variability in the stratosphere, as will be discussed in Section 4.

image

Figure 3. The difference between 300 hPa geopotential heights of the NCEP/NCAR reanalysis [gpm] for solar maximum and minimum years. The panels are structured the same way as in Figure 2. The 90 and 95% statistically significant regions are shaded in gray and dark gray. The maps are in a polar stereographic projection from 20–90°N.

Download figure to PowerPoint

[13] In December, and to some extent in January, there is a significant negative difference over the north-western parts of Russia and its northern coast (Figures 3a–3b). This difference changes sign in February. In December there is also a small region of positive difference over northern China (Figure 3a). The small sample of years for solar maximum and minimum activity at hand leads us to investigate the consistency of the pattern for the individual solar cycles. For each solar cycle we calculate the difference in the 300 hPa geopotential height for the solar maximum and the following solar minimum period. The positive difference over the North Pacific is present in all three subsets, even though the size of the difference is weaker for the period 1979–1987 (Figure 4, only shown for January). The consistency of the result for the period before and after the Pacific climate shift in 1976 [Hartmann and Wendler, 2005] indicates that this does not influence the results much. The large significant region of positive difference over northern Africa and East Asia in December is also consistent in December (not shown). This pattern is produced from an equatorward shift in wave number one and an equatorward and westward shift in wave number two. For the rest of the Northern Hemisphere there are inconsistencies for the different solar cycles, indicating that the small significant differences found outside the North Pacific could be random coincidences.

image

Figure 4. The difference between solar maximum and the following solar minimum of the 300 hPa geopotential height of the NCEP/NCAR reanalysis in January for each of the solar cycles in the study.

Download figure to PowerPoint

2.3. The Solar Modulation of the Planetary Waves

[14] We now study the individual Fourier components, i.e., the planetary waves, to determine their respective role in producing the differences in the Northern Hemisphere geopotential heights, with a focus on the North Pacific difference. In December the positive difference in the North Pacific is dominated by an equatorward shift and a decrease in amplitude of about 40–60 gpm of the first planetary wave (Figure 3d). Wave number two increases in amplitude by about 40 gpm in January with a slight eastward phase shift (Figure 3h), and becomes the major contributor to the pattern as the decrease in wave number one weakens compared to that in December (Figure 3e). In February wave number one shows no significant change (Figure 3f), while wave number two shows a slightly smaller increase in amplitude, compared to January, and a poleward and eastward shift (Figure 3i). The mean geopotential height, i.e., the wave number “zero”, generally shows a negative difference over the pole and positive at midlatitudes but the differences are not significant (not shown). Planetary wave number three shows no significant changes (not shown). The difference in geopotential height in the North pacific seems to be formed primarily by the first two planetary waves. The reason for these changes in the waves amplitude and phase could be due to changes in the diabatic heating in the troposphere, but it could also be due to downward propagation of stratospheric anomalies which affects the upper troposphere.

[15] How the stratosphere is influenced by the tropospheric changes is difficult to foresee, so we use a middle atmosphere model to study how the tropospheric solar signal affects the middle atmosphere.

3. Model and Experiments

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. The Solar Modulation of the Upper Tropospheric Planetary Waves
  5. 3. Model and Experiments
  6. 4. Model Results
  7. 5. Summary and Conclusions
  8. Acknowledgments
  9. References

3.1. The Stratosphere Mesosphere Model

[16] The model used is the UK Met Office Stratosphere Mesosphere Model (SMM) [Austin and Butchart, 1992; Arnold and Robinson, 2000]. It has previously been used in studies of dynamical forcings of the stratosphere by planetary waves [Scaife and James, 2000; Gray et al., 2001, 2003]. The SMM is a mechanistic three-dimensional primitive equation model with 5° × 5° horizontal resolution and about 2 km vertical resolution in 32 levels from 314 hPa to 0.03 hPa. The radiation scheme is based on Strobel [1978] and Shine [1987], and accounts for carbon dioxide, water vapor and ozone in the long wave spectrum. The shortwave spectrum is divided in six spectral bands between 0.175 and 0.7 μm. The model does not generate a QBO or an SAO, and no artificial equatorial wind is used in this study. Instead the model produce a weak easterly wind in the equatorial region during boreal winter. The gravity wave drag is simulated using Rayleigh friction, which damps the magnitude of the zonal wind at high altitudes to stop the wind speed from increasing with height [e.g., McLandress, 1998]. Using a version of this model, Gray et al. [2003] showed that the winter stratosphere is responsive to planetary wave forcing of intermediate amplitude, which is argued to be what we see in nature, but that other in stratospheric conditions and preconditioning could have an impact on the stratospheric response.

3.2. The Experimental Setup

[17] We perform one experiment in this paper. The model is forced only at the lower boundary by daily 300 hPa geopotential heights from the NCEP/NCAR reanalysis data set from the period 1963 to 1999. All other forcings such as solar irradiance and ozone concentration are kept constant on inter-annual scales throughout the experiment. At the lower boundary, the NCEP/NCAR reanalysis is preferred to ERA40 in this case due to known issues with the latter in the upper troposphere [Gleisner et al., 2005]. Before the data are introduced to the model an FFT is applied in the longitudinal direction for each day and latitude. A smoothed field is then obtained by superposition of the mean and the first three components, see further Section 2. Only the first three planetary waves can propagate in the relatively strong westerly flow of the winter stratosphere and it is therefore sufficient to include these to satisfactorily model the evolution of the polar night jet and stratospheric sudden warmings.

[18] The statistical significance of the results are calculated using the same Monte Carlo procedure as explained in Section 2. In the figures the 90 and 95% levels of significance are shaded in light and dark gray respectively.

3.3. The Model Climatology

[19] This version of the SMM model has previously only been used for experiments with a repeating annual cycle in the lower boundary [Gray et al., 2001], or in perpetual January conditions [Scaife and James, 2000; Gray et al., 2003], so the climatology of the 37 yearlong experiment is investigated here. We evaluate the model by comparing the experiment data to the ERA40 reanalysis data. We use the ERA40 reanalysis rather than the NCEP/NCAR because the ERA40 extends higher into the middle atmosphere. The two reanalyses have a similar climatology. In winter, the model temperature is underestimated by about 5–20 K in the southern polar low stratosphere and by a few Kelvin in the tropical tropopause (Figure 5, top panel). For the climatological mean, the PNJ starts to develop in September and reaches its full strength in late November. It is strong throughout December and then it starts to decrease and show a slight poleward movement. It decreases in strength and disappears in late March. The peak strength of the PNJ is overestimated in the model by more than 12 ms−1 throughout the winter (Figure 5). The tilt of the PNJ also deviates from the reanalysis, a known issue in climate models which is due to the way the gravity waves are simulated by Rayleigh friction [McLandress, 1998]. The easterly stratospheric jet in the Southern Hemisphere is weaker than in the reanalysis, especially in January. Below 100 hPa, the zonal winds are underestimated due to the influence of the lower boundary of the model. To evaluate the influence of the planetary waves on the middle atmosphere, we compare the divergence of the Eliassen-Palm flux [e.g., Andrews et al., 1987] to the ERA40 data. The model compares well, in both structure and magnitude, to the reanalysis data for the northern polar region south of 70°N for December (Figure 6) and January (not shown). In February, the structure is the same as for December and January, but the magnitude is somewhat weaker than in the reanalysis (not shown). North of 70°N the model is underestimating the divergence of the Eliassen-Palm flux. From this comparison we can conclude that in general the model is capable of simulating the dynamical planetary wave forcing of the middle atmosphere, with only smaller deviations from the ERA40 reanalysis data. The use of a Rayleigh friction to simulate gravity waves leads to a too strong and upright PNJ, but the evolution of the jet over boreal winter is fairly well simulated.

image

Figure 5. Climatologies for December–February for the zonal mean temperature [K] (a–b) and climatologies for December, January and February for the zonal mean zonal wind [ms−1] (c–h) for the SMM (left) and ERA40 reanalysis (right). Zonal winds that are stronger than 36 ms−1 are shaded.

Download figure to PowerPoint

image

Figure 6. Climatology for December of the divergence of the Eliassen-Palm flux [ms−1day−1] for the SMM (left) and the ERA40 reanalysis data (right). Note that only the northern extra tropics are shown. The flux divergence is well modeled for the polar region. Divergences lower than 20 ms−1day−1 are shaded.

Download figure to PowerPoint

4. Model Results

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. The Solar Modulation of the Upper Tropospheric Planetary Waves
  5. 3. Model and Experiments
  6. 4. Model Results
  7. 5. Summary and Conclusions
  8. Acknowledgments
  9. References

[20] To see how the middle atmosphere responds to the solar signals in the high troposphere we compare the differences of the 37 year model experiment for the same maximum and minimum solar activity years used in the study on the reanalysis data in Section 2. For each of the three winter months we calculate the difference in zonal mean temperature and zonal mean zonal wind (Figure 7). In December, we see a significant positive difference in temperature above 2 hPa and a negative difference below, north of 40°N (Figure 7a). The positive difference moves downward and poleward in January as it increases in strength, while a negative difference moves in from the tropics above 1 hPa (Figure 7b). In February, the structure has reversed with a strong positive difference below 1 hPa and negative above (Figure 7c). The zonal wind shows a similar progression with a positive difference at 5 hPa in December at 40°N (Figure 7g) that moves downward and poleward in January (Figure 7h) and is replaced by a negative difference from the subtropics in February (Figure 7i). The significance test shows lower significance for December but higher significance with increased differences in January and February.

image

Figure 7. The differences in temperature [K] (column one and two) and zonal wind [ms−1] (column three and four) between the solar maximum and minimum years. The winter is divided in December (top), January (middle) and February (bottom). The model results are shown in the first and third column, and the ERA40 reanalysis data are shown in the second and fourth column. The 90 and 95% significance levels are shaded in light and dark gray respectively. The contours are [−10, −5, −3, −2, −1, −.5, 0, .5, 1, 2, 3, 5, 10] K for temperature and [−15, −10, −5, −3, −1, 0, 1, 3, 5, 10, 15] ms−1 for zonal winds.

Download figure to PowerPoint

[21] We compare the results for temperature and zonal wind from the model experiment to the corresponding ERA40 reanalysis data for the same period (Figure 7, second and fourth columns). We use the ERA40 data for the comparison, rather than the NCEP/NCAR data, because it extends further up in the atmosphere. The two reanalysis data sets compare well in the stratospheric fields so the comparison between the model, forced by NCEP/NCAR data, and the stratospheric ERA40 data is not affected much. The differences between solar maximum and minimum years in the ERA40 compare well to the observations shown by Kodera and Kuroda [2002], even though our study uses a longer time period, 1965–1997 compared to 1979–1997. The structure of the differences is very similar between our model study and the ERA40, even though the strength of the differences deviates at times. Deviations between the model and the reanalysis are expected since the natural variability in the reanalysis data is higher. The statistical significance of the differences in the ERA40 data are weaker than in the model, possibly due to the larger natural variability of the reanalysis data.

[22] Our model experiment is idealized and only includes the lower boundary forcing, thus it does not include increases in the solar irradiance and other stratospheric influences. Increasing the radiation has a direct radiative effect on the middle atmosphere, but also indirect dynamical effects [Kodera and Kuroda, 2002]. The lack of an increased UV-flux means that the meridional temperature gradient does not increase for the solar maximum conditions, as would be the case when the stratopause is heated by the more intense insolation. An increased meridional temperature gradient is expected to result in a stronger PNJ [Kodera and Kuroda, 2002], which would inhibit the upward propagation of planetary waves in the stratosphere. This would decrease the tropospheric forcing on the stratosphere in the PNJ region. By keeping the UV radiation in solar cycle minimum conditions the model stratosphere is more responsive to tropospheric forcing close to the peak in the PNJ than would otherwise be the case. For this study it means that the upward propagating planetary waves will be affected the same way by the stratospheric state for both years of maximum and minimum solar activity. The divergence of the Eliassen-Palm flux shows how the middle atmosphere responds to the tropospheric planetary wave forcing. For the difference between maximum and minimum solar activity we see a weakening of the planetary wave forcing of the stratosphere for December, and an increased wave forcing in January and February (Figure 8). The weaker planetary wave forcing in December allows the zonal winds to grow stronger, and as the wave forcing increases, the strength of the zonal wind is reduced. This effect would enhance the radiatively induced strengthening of the PNJ in December, had there been an irradiance increase in the model experiment, but would inhibit this increase later in winter. The horizontal and vertical components of the Eliassen-Palm flux show a decrease in the poleward and downward circulation for December (not shown), and an increase in January. In February the poleward component increases in the polar region and decreases at midlatitudes (not shown). The changes in the Eliassen-Palm fluxes presented above are at most weakly significant and the conclusions drawn should be considered with caution.

image

Figure 8. The difference between solar maximum and minimum years for the SMM experiment for the divergence of the Eliassen-Palm flux [ms−1day−1]. The winter is divided in December (left), January (middle) and February (right). Note that only the northern extra tropics are shown. The 90% significance level is shaded in gray.

Download figure to PowerPoint

[23] Matthes et al. [2004] succeeded in reproducing the solar modulation of the stratosphere in an atmospheric GCM forced by solar UV changes. Furthermore the tropospheric response in the geopotential heights showed a similar pattern at 1000 hPa to that presented for 300 hPa in the previous section in December and January, while the response in February was of opposite sign [Matthes et al., 2006]. The success of simulating the observed temperature and zonal wind differences in their model was speculated to be due to a relaxation of the equatorial winds to observations, i.e., including the QBO and the SAO in the model. In our study we show that the solar modulation of the winter stratosphere is possible to reproduce without a stratospheric signal of the QBO or SAO. The response in the Matthes et al. [2004] study was much weaker than in the present model, which does not include neither a QBO nor an SAO. The lack of a QBO in the model could affect the strength of the response in the early winter, when the effect of the QBO is largest [Labitzke, 1987; Labitzke and van Loon, 1988]. The effect is difficult to asses due to its biennial nature and the ambiguous connection to solar activity: the easterly phase of the QBO shows a weaker polar night jet for minimum solar activity and a stronger jet for maximum activity, while for the westerly phase, the opposite response is observed [Labitzke, 2003]. Observations show that the strength of the QBO signal in the extra tropics is strongest in early winter and diminishes by the end of January [Dunkerton and Baldwin, 1992]. Gray et al. [2001] used the same model as used here to investigate the effect of relaxing the equatorial winds to observations and found that the QBO does have an effect on the amplitude of the stratospheric response for the two winters studied. This indicates that the QBO could have some connection to the weaker signal seen in the early winter of the model study, compared to the observations, but further experiments are needed to assess the model response to such perturbations.

[24] The experimental setup where we use reanalysis data at the lower boundary to force the model does not completely separate the effects of the forcings from the troposphere from those from the stratosphere. This is because the lower stratosphere and upper troposphere are highly coupled and the boundary data that we use already have a stratospheric signal. This means that the information about the stratosphere that we already have in the lower boundary data will be transferred back to the stratosphere. Therefore we cannot from this setup deduce whether the solar signal originates in the stratosphere or in the troposphere. If a signal with a stratospheric origin is present in the upper troposphere, it could to some extent force the model stratosphere to reproduce the signal, but this is not likely to be a strong effect above the lower stratosphere. The results of this study suggest that the solar modulation of the stratosphere is dependent on a coupling and feedback process with the troposphere, through a modulation of the planetary waves.

5. Summary and Conclusions

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. The Solar Modulation of the Upper Tropospheric Planetary Waves
  5. 3. Model and Experiments
  6. 4. Model Results
  7. 5. Summary and Conclusions
  8. Acknowledgments
  9. References

[25] In a study of the NCEP/NCAR and ERA40 reanalyses 300 hPa geopotential heights of the Northern Hemisphere winter, we have found significant changes related to the decadal solar cycle. As the Sun increases in activity, there is a significant increase in geopotential height for December through February in the North Pacific, a pattern which is consistent for all three individual solar cycles studied. The planetary wave number one shows a strong decrease in amplitude and an equatorward shift in December and, to a lesser extent, in January. Wave number two shows an increased amplitude and a slight eastward phase shift in January and February, also with a poleward phase shift for the latter. We do not specifically consider the origin of the solar cycle in the troposphere in this study, but the changes in the pressure over the North Pacific seems to be a consistent phenomenon for solar variability on the troposphere [van Loon et al., 2007].

[26] We used a three-dimensional model of the middle atmosphere to study the impact of a tropospheric solar signal on the middle atmosphere, using the observed 300 hPa geopotential heights as the lower boundary forcing. The experiment lasted for 37 model years and the data series were divided into solar maximum and solar minimum years. To avoid conflicting signals from ENSO and volcanoes, the years with strong such signals were removed from the analysis. The results of the study were significant changes in temperature and in zonal winds throughout the winter. The structure and amplitude of the responses were very similar to the ERA40, and NCEP/NCAR, reanalysis data and to earlier studies of observations for both temperature and zonal winds [Kodera and Kuroda, 2002]. Over the winter, a positive temperature anomaly in the lower mesosphere in December propagates downward and poleward to the lower stratosphere in February. The zonal wind shows a similar propagation. This indicates that the solar modulation of the troposphere has a significant effect on the solar modulation of the stratosphere, and we will end the paper with a discussion on this, and relate to earlier model results.

[27] Matthes et al. [2004] reproduced the solar modulation of the stratosphere in an atmospheric GCM and suggested that a proper QBO and SAO are needed to reproduce the stratospheric pattern. The tropospheric response in that experiment [Matthes et al., 2006] shows some similarities to the pattern found in observations [van Loon et al., 2007]. In our experiment we used the observed tropospheric response to force the model, but additional stratospheric forcings. The tropospheric forcings are therefore the only similarity between the present and the Matthes et al. [2004] experiment, thus it seems like a solar modulation of the troposphere contributes significantly to the solar modulation of the winter stratosphere. The QBO and the SAO could contribute significantly to the solar forcing in forming these tropospheric anomalies, but our experiment indicates that the effect of the equatorial winds play a minor role in affecting the planetary wave forcing in situ in the stratosphere. The origin of the solar modulation of the troposphere is not clear from this experiment. A direct solar forcing, irradiative or non-irradiative, in the troposphere could affect the planetary waves, but the forcing of the troposphere could also originate in a solar modulation of the stratosphere. The stratospheric influence could arise from a transfer of the stratospheric solar signal to the troposphere through downward propagation of stratospheric anomalies [Kodera and Kuroda, 2002; Matthes et al., 2006] or through changes in the tropospheric circulation induced by solar modulations of the stratosphere [Haigh, 1999]. This stratospheric forcing of the troposphere could then feedback to the stratosphere through the planetary waves. As the experiment performed here does not separate between the forcings that produce the differences in the planetary waves, we can not say anything about the ‘direction’ of the solar modulation of the stratosphere, i.e., if the modulation originates from stratospheric changes or from the troposphere. We propose a coupling process where the stratosphere forces the upper tropospheric planetary waves, which then feeds back to the stratosphere through upward wave propagation. Additional experiments with a modeled solar irradiance cycle and equatorial wind profile would be useful to further examine the results and conclusions presented here. Together with a thorough Eliassen-Palm flux analysis, this would lead to a further understanding of the importance of the tropospheric forcing of the boreal winter stratosphere during the solar cycle.

Acknowledgments

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. The Solar Modulation of the Upper Tropospheric Planetary Waves
  5. 3. Model and Experiments
  6. 4. Model Results
  7. 5. Summary and Conclusions
  8. Acknowledgments
  9. References

[28] P. Berg acknowledges co-funding through the European Community's Human Potential Programme under contract HPRN-CT-2002-00216, Coupling of Atmospheric Layers, during 2003–2006, and the Copenhagen Global Change Initiative (COGCI). NCEP Reanalysis data were provided by the NOAA/OAR/ESRL PSD, Boulder, Colorado, USA from their website at http://www.cdc.noaa.gov/. The ERA40 reanalysis data were provided by the ECMWF organization, Reading, UK, from their website http://www.ecmwf.int/. The ENSO index data where provided by the Joint Institute for the Study of the Atmosphere and Ocean (JISAO), University of Washington, from their website http://jisao.washington.edu/data/globalsstenso/.

References

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. The Solar Modulation of the Upper Tropospheric Planetary Waves
  5. 3. Model and Experiments
  6. 4. Model Results
  7. 5. Summary and Conclusions
  8. Acknowledgments
  9. References