SEARCH

SEARCH BY CITATION

Keywords:

  • PMC;
  • SBUV;
  • coupling;
  • middle atmosphere;
  • polar mesospheric cloud;
  • teleconnection

Abstract

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Data Description and Analysis
  5. 3. Results
  6. 4. Discussion and Conclusions
  7. Acknowledgments
  8. References

[1] This paper describes an investigation using data from the Solar Backscatter Ultraviolet (SBUV) satellite instruments to explore and understand variations in the timing of the onset of Polar Mesospheric Cloud (PMC) seasons. Previous work has shown that for several recent southern hemisphere (SH) seasons, the PMC season onset was controlled by the timing of the shift from winter to summer zonal wind flow in the SH stratosphere. We extend the analysis of PMC season onset to 28 years of SBUV observations, including both hemispheres. A multiple linear regression analysis of SBUV data from 1984 to 2011 suggests that the SH PMC season onset is delayed by one day for every day that the zonal wind at 65°S and 50 hPa (∼20 km) remains in a winter-like state. In addition, we find that the solar cycle plays a role: The SH season onset is delayed by about ten days at solar maximum compared to solar minimum. In the NH, the PMC season onset is delayed by ∼7 days at solar maximum compared to solar minimum; variations in the NH stratospheric wind, however, are not correlated with the NH onset date. On the other hand, inter-hemispheric teleconnections are important in the NH; a one-day shift in the NH season onset corresponds to a shift of ∼1.4 m/s in the SH stratospheric wind at 60.0°S and 20 hPa (∼26 km). Neither the NH nor the SH season onset date is correlated with the Quasi-Biennial Oscillation, North Atlantic Oscillation, Arctic Oscillation, or El Niño Southern Oscillation.

1. Introduction

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Data Description and Analysis
  5. 3. Results
  6. 4. Discussion and Conclusions
  7. Acknowledgments
  8. References

[2] Polar Mesospheric Clouds (PMCs), Earth's highest clouds, form in the summer polar mesopause region, which is the coldest place in the atmosphere. They are observable from ground and space, and provide an indication of the thermal and dynamical conditions of the upper mesosphere. PMCs are highly dependent on mesospheric temperatures, which are modulated by the meridional circulation, planetary waves (PWs), gravity waves (GWs), tides, and the solar cycle.

[3] The upper branch of the Brewer-Dobson circulation consists of upwelling in the summer polar mesosphere and downwelling in the winter polar mesosphere. This circulation is driven by breaking GWs. GWs are vertically propagating waves that are caused by flow over topography, frontal processes, convection, and jet streams [Fritts and Alexander, 2003]. With increasing altitude GW amplitudes grow exponentially due to decreasing atmospheric density. Eventually the wave becomes convectively unstable, breaks, and deposits angular momentum in the same direction as the GW phase speed. This momentum deposition is also called GW drag.

[4] The seasonally varying stratospheric zonal wind controls GW amplitudes and breaking levels through GW filtering. GWs can only propagate vertically when their phase speed either exceeds the zonal wind speed or is of the opposite direction. During summer, the mean stratospheric and mesospheric zonal wind is directed westward, so GWs with westward phase speeds are filtered. Thus in the summer hemisphere breaking GWs cause a net eastward drag in the upper mesosphere, decelerating the mean zonal flow [Holton and Alexander, 2000]. This transfer of angular momentum can only be balanced by the Coriolis torque [Shepherd, 2007]. The Coriolis force acts on the induced eastward flow, with balance re-established once the flow is directed equatorward. Mass conservation and the principle of downward control then require that upwelling occur in the summer polar mesosphere [Shepherd, 2000]. The resulting adiabatic cooling in the summer mesosphere causes deviations from radiative equilibrium temperatures by more than 60 K [Holton and Alexander, 2000; Lübken et al., 2009]. Only this GW induced dynamical pump allows PMCs to form in the summer polar mesopause region. In summary, GWs and stratospheric/mesospheric zonal winds, because of their filtering influence, impact mesopause temperatures and therefore PMCs.

[5] Even though the mechanism driving this dynamical pump is the same in the northern hemisphere (NH) and southern hemisphere (SH), a hemispheric asymmetry in the summer polar mesospheric temperature of up to 7.5 K was observed by Hervig and Siskind [2006], Wrotny and Russell [2006], and Morris et al. [2009]. The hemispheric temperature asymmetry is the main reason for less frequent, dimmer, and less extensive PMCs in the SH [Olivero and Thomas, 1986; Bailey et al., 2005; Petelina et al., 2005; Bailey et al., 2007; Shettle et al., 2010] as well as Polar Mesospheric Summer Echoes [Balsley et al., 1993, 1995; Woodman et al., 1999]. Siskind et al. [2003] explained a hemispheric temperature asymmetry of 3–8 K as being caused by hemispheric differences in GW wind filtering and radiative effects. Such hemispheric differences might influence the timing of the PMC season onset. Gumbel and Karlsson [2011] found that the timing of PMC onset is more variable in the SH than in the NH. Based on a model study and observations of PMCs from the Aeronomy of Ice in the Mesosphere (AIM) mission, Karlsson et al. [2011]suggested that the SH PMC onset date is largely controlled by changes in the timing of the winter-to-summer stratospheric wind reversal. They argue that a more ‘winter-like’ (‘summer-like’) stratospheric mean flow would lead to a warmer (colder) polar mesosphere, using the arguments presented above. Therefore, a late wind reversal, which would correspond to a long-lasting SH polar vortex, would lead to later dynamical cooling of the mesopause region above, thereby delaying the onset of the SH PMC season. On the other hand, increased PW activity in the SH would cause an earlier breakdown of the polar vortex and lead to an earlier onset of the SH PMC season. Using the Canadian Middle Atmosphere Model, they demonstrated this mechanism in a statistically significant way.Gumbel and Karlsson [2011] confirmed this mechanism using nine years of Odin satellite data.

[6] Several studies [e.g., Becker and Schmitz, 2003; Karlsson et al., 2007] have shown that the polar summer mesopause region is also affected by the winter stratosphere through an inter-hemispheric coupling mechanism. This mechanism is triggered by PW induced changes in the zonal wind that alter GW filtering. In cases where the winter stratosphere is highly disturbed by PWs, the usually strong eastward zonal wind inside the polar vortex is disturbed and weakened, sometimes even reversed. GW filtering through the zonal wind will therefore result in weaker westward GW drag in the winter mesosphere. The combined PW and GW effects induce changes in the circulation that cause a quadrupole structure in the winter hemisphere temperature anomaly field, as described byKarlsson et al. [2009b] and Becker et al. [2004]. That is, the high-latitude stratosphere and low-latitude mesosphere warm while the high-latitude mesosphere and low-latitude stratosphere cool. The low-latitude, mesospheric warming changes the meridional temperature gradient to the summer polar mesosphere, leading to changes in zonal winds, GW filtering, meridional circulation, and upwelling that result in a warmer summer polar mesosphere. Thus, stronger PW activity in the winter hemisphere results in higher temperatures over the summer pole and fewer PMCs. Analogous arguments can be made that weaker winter hemisphere PW activity leads to a higher probability of summer hemisphere PMCs.Karlsson et al. [2007, 2009a, 2009b] presented the effect of inter-hemispheric coupling on mesopause temperatures, PMC effective radii, and PMC occurrence frequency during the PMC season, but not on the PMC onset.

[7] We also expect that the timing of the PMC season onset may be affected by the solar cycle. The solar cycle has long been known for its effect on PMCs through photodissociation of water vapor by Lyman-α radiation [Garcia, 1989; Thomas et al., 1991]. Additionally, mesospheric temperature is modulated by the changing solar UV heating rate between solar minimum and maximum conditions. These two effects lead to an ∼11-year modulation of PMC occurrence frequency, with lower frequencies during solar maximum and higher frequencies during solar minimum conditions [Thomas et al., 1991]. Thus we would expect that on average, during solar maximum conditions the PMC season may start later than during solar minimum conditions. Even though the effect of the solar cycle on season-average PMC frequency is well known, any effect of the solar cycle on the PMC season onset date has not yet been documented.

[8] In this paper we investigate variability from 1984 to 2011 in the PMC season onset date in both hemispheres using observations from the Solar Backscatter UltraViolet (SBUV) series of instruments [DeLand et al., 2003, 2006, 2007]. We extend the studies of Karlsson et al. [2011] and Gumbel and Karlsson [2011]to investigate whether their explanation for the SH PMC season onset pertains to historical data. We also look for other mechanisms that might be contributing to variability in the PMC onset data, such as the solar cycle and inter-hemispheric coupling.Section 2 contains descriptions of data and analysis methods. Results are shown in section 3 and conclusions are drawn in section 4.

2. Data Description and Analysis

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Data Description and Analysis
  5. 3. Results
  6. 4. Discussion and Conclusions
  7. Acknowledgments
  8. References

[9] For the current work we correlate variations in PMC data for years 1984 to 2011 from the SBUV series of instruments with variations in meteorological data and the solar cycle. We analyze the SBUV observations to calculate PMC occurrence frequency, from which the date of the PMC season onset is inferred. We use the meteorological analyses described below to calculate the timing of the reversal of the stratospheric zonal mean zonal wind from winter to summer conditions.

2.1. SBUV

[10] The SBUV instruments have been launched since 1978, flying on the Nimbus-7, NOAA-9, NOAA-11, NOAA-14, NOAA-16, NOAA-18, and NOAA-19 satellites [DeLand et al., 2003]. Satellites after Nimbus-7 carried the improved SBUV/2 instrument, but for simplicity in this paper all of the instruments will be referred to as SBUV. The SBUV instruments were originally designed to observe global stratospheric ozone but have also been used to measure PMCs [Thomas et al., 1991; DeLand et al., 2003, 2006, 2007]. The SBUV instruments are nadir-pointed with a horizontal resolution of 150 × 150 km at PMC altitude and measure backscattered radiation at 12 wavelengths in the UV. PMCs are detected following the original technique developed byThomas et al. [1991], which was refined by DeLand et al. [2007]. This technique is based on PMCs appearing as a spectrally enhanced signal above the backscattered signal in the UV (252.0–292.3 nm), with the maximum enhancement at the shortest wavelengths [DeLand et al., 2003]. In the following, SH seasons will be referred to using the year of their onset, e.g., the SH00 season refers to the SH season that started toward the end of 2000 and extended through the first part of 2001.

2.2. Meteorological Analyses

[11] The European Centre for Medium Range Forecast (ECMWF) Re-Analysis (ERA-40) data are available from 1957 to 2002 [Uppala et al., 2005]. Model output is provided four times daily at 0 UTC, 6 UTC, 12 UTC, and 18 UTC and has a horizontal resolution of 2.5° longitude by 2.5° latitude. The data are available on 23 pressure levels that extend from 1000 hPa up to 1 hPa. In this study, we use daily averaged zonal mean zonal wind data at 12 UTC, in both hemispheres.

[12] Global assimilation analyses from the United Kingdom Meteorological Office (MetO) Unified Model are available from 1991 to the present [Swinbank and O'Neill, 1994]. The model uses three dimensional variational data assimilation [Lorenc et al., 2000] and has a semi-Lagrangian dynamical core [Davies et al., 2005]. Model output is provided once per day at 12 UTC on a 2.5° latitude by 3.75° longitude grid. The data are on 25 pressure levels between 1000 hPa and 0.1 hPa. Zonal mean zonal winds in both hemispheres are used to extend the zonal wind data record to the present.

[13] ERA-40 and MetO overlapping wind data between 1991 and 2002 compare favorably to within half a percent. For the current work wind data from the two re-analyses were combined using ERA-40 between 1979 and 2001 and MetO between 2002 and 2011.

2.3. Definition of PMC Season Onset and Stratospheric Wind Reversal

[14] The PMC season onset is calculated using daily SBUV PMC occurrence frequency between 75° and 82° latitude with a 5-day running average applied separately for ascending and descending nodes. Season onset is defined as the date on which the frequency remains above 1.5% for at least seven consecutive days for the first time. This criterion is sufficiently conservative to filter out false cloud detections, which can cause frequencies of ∼0.5% [DeLand et al., 2007], and to reject sporadic frequency increases before the actual onset of the season. However, this criterion makes PMC season onsets calculated at lower latitudes, e.g., between 70°S and 75°S, more and more unreliable for two reasons: in some years the 1.5% cutoff frequency is a significant fraction of the maximum seasonal frequency and therefore pushes the season onset back unreasonably far; in some other years false pre-season detections at these lower latitudes result in a too early season onset. Since SBUV does not observe PMCs poleward of 82° latitude, the latitude interval of 75° to 82° was chosen. Information from Table 3 ofDeLand et al. [2007] was used to exclude SBUV data in seasons that exhibited artifacts during the season onset time period. Since analysis of data from each SBUV instrument yields two season onset dates – one for the ascending node and one for the descending node – and some seasons are observed by several instruments (up to four), the reported season onset is the mean season onset. Potential PMC onset variability due to possible instrumental/observational effects and space shuttle launches will be discussed in section 4.

[15] It should be noted that PMC season onset as determined by SBUV is delayed by up to ∼20 days compared to the date inferred from more sensitive instruments, such as the Cloud Imaging and Particle Size (CIPS) instrument and Solar Occultation For Ice Experiment on AIM; see, e.g., Benze et al. [2011]for a comparison of SBUV and CIPS detection capabilities. This can be attributed to the fact that SBUV observes only the brightest clouds, and therefore misses the dim and sparse clouds at the beginning of the season. SBUV and CIPS PMC onset correlate with a coefficient of 0.75 (0.91) in the SH (NH) (not shown), which indicates a good correlation of PMC onset dates between less and more sensitive instruments. However, since this analysis emphasizes the relative change of PMC season onset date from year to year, a consistent determination of PMC season onset date is more important for understanding the observed variability than absolute accuracy. Moreover, SBUV offers the unique advantage of more than a quarter century of satellite-based cloud detections.

[16] The daily mean stratospheric zonal mean zonal wind was calculated from the meteorological analysis data at 50 hPa (∼20 km) and 65° latitude, and a 5-day running average was applied. The timing of the wind “reversal” from winter to summer conditions was defined as the day of the final decrease of the mean zonal wind below 10 m/s. This is the definition of the polar vortex breakup date that was first proposed byWaugh et al. [1999] and used by Langematz and Kunze [2006, 2008]; it avoids false detections of early vortex breakdown dates when major warmings occur in the middle of the winter.

2.4. Solar Cycle Data

[17] Daily values of Lyman-α [Woods et al., 2000] (http://lasp.colorado.edu/lisird/lya/) are averaged from 30 to 10 days prior to solstice in the summer hemisphere to provide an indication of the state of the solar cycle at the beginning of the PMC season. Tests were run to evaluate the sensitivity of the results to the period chosen for the Lyman-α average. Conclusions were the same for all choices of averaging time period, which included −40 to −30, −30 to −20, −20 to −10, −10 to 0, 0 to 10, −40 to 10, −30 to 0, −20 to −10, and −30 to −10 days from solstice (DFS).

3. Results

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Data Description and Analysis
  5. 3. Results
  6. 4. Discussion and Conclusions
  7. Acknowledgments
  8. References

[18] Figure 1shows the timing of the NH and SH PMC season onset versus years. The dotted lines indicate years prior to 1984. Seasons between 1979 and 1983 were excluded from the analysis due to very low overall PMC frequencies in these years, the reason for which is unclear. In these years large uncertainties in the definition of the season onset are caused by the season onset threshold of 1.5% being a significant fraction of the maximum frequency. The PMC seasons 1979 to 1983 coincide with solar maximum conditions that cause higher mesospheric temperatures and lower water vapor concentrations and therefore fewer and dimmer clouds. However, the 1981 solar maximum was comparable to the following solar maximum, during which overall frequencies were twice as high. Solar maximum conditions in addition to the long-term increase in the frequency of PMCs noted byDeLand et al. [2003] and Shettle et al. [2009], could contribute to lowest frequencies being associated with the earliest dates. Despite the exclusions, 28 continuous years spanning more than two solar cycles are covered in both hemispheres. While the NH and SH PMC seasons on average start at the same time (NH: DFS −15 ± 6, SH: DFS −12 ± 12), variability in the SH onset date is twice as high as in the NH. Gumbel and Karlsson [2011] found average PMC season onset dates in units of DFS of −26 ± 3 and −24 ± 9 in the NH and SH, respectively, using PMC data from the Optical Spectrograph and InfraRed Imaging System (OSIRIS) instrument between 2002 and 2011. The earlier OSIRIS onset date is because OSIRIS is much more sensitive than SBUV, since it measures limb scattered sunlight. The overall higher SBUV variability in the PMC onset date may be due to the longer period of PMC onset observations available from SBUV. Nevertheless, measurements from both OSIRIS and SBUV show that variability in the SH season onset is significantly higher than in the NH.

image

Figure 1. Time series of SBUV PMC season onset date in the NH (red) and SH (blue). Dotted lines indicate PMC onsets before 1984.

Download figure to PowerPoint

[19] Figure 2relates the timing of the SH PMC season onset to the timing of the SH lower stratospheric wind reversal from winter to summer conditions. Error bars on the PMC onset time series indicate the range of PMC onset dates from multiple SBUV instruments. Again, the dotted lines prior to 1984 indicate the problematic years described above. Year-to-year variations in the dates of the SH PMC season onset and wind reversal over the last 28 years are highly correlated, with a correlation coefficient r = 0.85. Including the PMC onset dates observed prior to 1984 decreases the correlation coefficient to 0.68. For reasons discussed above, the following results will not include the years prior to 1984. It could be argued that a small or absent ozone hole plays a role in these early years: the loss of ozone as observed from the late 1970s onward leads to a reduction in heating via absorption of solar radiation [Thompson and Solomon, 2002; Randel et al., 2009]. A colder stratospheric polar vortex may be prolonged into austral summer, which in turn may lead to a later reversal from winter to summer conditions in the Antarctic polar upper atmosphere [Smith et al., 2010; Lossow et al., 2012]. Therefore it could be argued that the absence of the ozone hole in these early years could lead to earlier wind reversal dates in the SH stratosphere prior to 1984, and therefore cause a de-coupling of stratospheric wind and PMC onset. That is, if the wind reversal occurred long before summertime conditions were setting up in the mesosphere, one might not expect it to influence PMCs. However,Figure 2 shows that the SH stratospheric wind reversal dates were not unusually early between 1979 and 1983. Therefore we exclude this possibility.

image

Figure 2. Time series of average SH PMC season onset dates as observed by SBUV (black) and date of SH stratospheric zonal mean zonal wind (ERA40 and MetO combined) reversal from winter to summer conditions at 50 hPa and 65°S latitude (red). Dotted lines indicate PMC onsets before 1984. Error bars show the range of values from multiple SBUV instruments.

Download figure to PowerPoint

[20] For a more quantitative comparison, Figure 3shows a linear fit of the PMC season onset to the stratospheric wind reversal. The linear fit to the data (red line) is in very good agreement with the black line, which denotes perfect agreement. The slope of the fit is 0.95 ± 0.12, suggesting that the stratospheric wind reversal is indeed linearly related to the PMC season onset. That the fit is displaced above the one-to-one line indicates that the PMC onset date as observed by SBUV is delayed by about 1–2 days relative to the timing of the stratospheric wind reversal as defined here. This delay would likely be shorter or reversed for more sensitive instruments, and of course depends on the exact definition of the wind reversal.

image

Figure 3. PMC onset date versus wind reversal date (pluses) with a linear fit to the data (red) and one-to-one line for reference (black). Data before 1984 is excluded.

Download figure to PowerPoint

[21] In the NH, the stratospheric wind reversal takes place about two months prior to the SBUV PMC season onset (not shown), so the wind reversal is not expected to control the date of the PMC season onset. This is supported by Gumbel and Karlsson [2011] who found no connection between the NH PMC onset and the timing of the NH wind reversal using the more sensitive Odin/OSIRIS instrument. We investigated whether the strength of the NH stratospheric zonal mean zonal wind at the time of the NH PMC season onset has any influence, but did not find any correlation (not shown).

[22] Since PMCs are strongly influenced by solar activity [Thomas et al., 1991; DeLand et al., 2003], we also correlate Lyman-α with the PMC season onset date. Figure 4 shows SH (Figure 4a) and NH (Figure 4b) PMC season onset date expressed as DFS, and the median Lyman-α between DFS −30 and −10. In the NH there is a moderate correlation between solar signal and PMC season onset date, both with zero lag (r = 0.60) and a lag of one year (r = 0.50). Lags of more than one year decrease the NH correlation. A lag of one year is considered since several studies have found that PMC frequency correlates best with the solar cycle when a lag of 0.5 to 1.5 year is considered [Kirkwood et al., 2008; Thomas and Olivero, 2001; DeLand et al., 2003]. In the SH, small correlation coefficients of 0.12 (0.14) at lag = 0 years (1 year) indicate very little correlation between the PMC season onset date and solar signal.

image

Figure 4. Time series of SBUV PMC onset date (black) and median of Lyman-α between DFS −30 and −10 (red) in the (a) SH and (b) NH. Note the different y axis scales for the PMC season start.

Download figure to PowerPoint

[23] Although the SH correlation coefficient just described was small, visual inspection of Figure 4gives the impression that even in the SH there is a low-frequency variation of the PMC season onset date that is correlated with the solar cycle, upon which is superimposed a higher frequency variation.Figure 5compares the difference between SH PMC season onset date and the timing of the SH stratospheric wind reversal (i.e., the residual when the wind variation is subtracted from the observations) to the Lyman-α variations. This comparison reveals that the residual is correlated with the solar cycle, with a correlation coefficient of 0.61, similar to the solar cycle correlation in the NH. This indicates that variations in the SH PMC season onset date are dominated by the timing of the stratospheric wind reversal, with additional modulation from the solar cycle.

image

Figure 5. Difference between SH PMC onset date and SH wind reversal date (black) versus Lyman-α (red).

Download figure to PowerPoint

[24] In order to quantify the relative importance of these components, a multiple linear regression analysis was carried out. Since the stratospheric wind reversal and the Lyman-α signal are not correlated (not shown) and thus assumed to be independent, it can further be assumed that the PMC season onset date, Don, can be expressed as a linear combination of the timing of the stratospheric wind reversal, W, and the solar cycle as represented by the Lyman-α flux, Lyα:

  • display math

The coefficients are a = 1.03 ± 0.10 days per day delay in the wind reversal, and b = 5.00 ± 1.34 days per unit of Lyman-α equal to 1011 cm−2 s−1. That is, for every day delay in the wind reversal, the season onset date is delayed by 1.03 days. For every increase in Lyman-α by 1011 cm−2 s−1, the season onset date is delayed by 5.0 days. As shown in Figure 5, Lyman-α changes by roughly twice this much over the solar cycle. Figure 6 compares the PMC onset date Don calculated from equation (1)to the PMC onset date inferred from SBUV observations. The correlation coefficient of 0.91 shows a small improvement over the correlation coefficient of 0.85 obtained by correlating the observed PMC onset date to the wind reversal date alone. Inter-hemispheric coupling of the SH PMC onset with the NH stratospheric wind is not significant (not shown).

image

Figure 6. SBUV (black) and regressed (red) time series of SH PMC onset date using the SH stratospheric wind reversal date and solar Lyman-α as linear components.

Download figure to PowerPoint

[25] In the NH, as in the SH, a multiple linear regression was also carried out. Recall (Figure 4) that we found a clear correlation between the observed NH PMC onset date and Lyman-α, but not between the NH PMC onset date and the same hemisphere stratospheric wind. We therefore investigated a possible connection between NH PMC onsets and stratospheric winds from the opposite hemisphere. This was accomplished by first performing a linear regression with Lyman-αas the only linear component in order to find the component of the total PMC onset variability that can be accounted for by Lyman-α. The residual of the NH PMC onset and the regression result was then correlated with the SH zonal mean zonal wind at various latitudes, altitudes, and times. The maximum correlation with the wind occurred 10 days before the NH PMC onset, at 60.0°S and 20 hPa (∼26 km). Figure 7ashows the residual of the NH PMC onset and Lyman-α plotted versus this zonal wind. The correlation coefficient of 0.77 supports the idea that interhemispheric coupling with the SH stratospheric winds plays a role in controlling NH PMC onset dates. For the regression analysis the NH PMC season onset date, Don, is therefore expressed as a linear combination of the SH stratospheric wind at the above mentioned space and time, W, and the solar cycle as represented by the Lyman-α flux, Lyα:

  • display math

The resulting coefficients are c = 0.72 ± 0.13 days per m/s and d = 3.51 ± 0.79 days per unit of Lyman-α equal to 1011 cm−2 s−1. That is, for every increase in opposite hemisphere winter wind strength by 1 m/s, the NH season onset date is delayed by 0.72 days. For every increase in Lyman-α by 1011 cm−2 s−1, the season onset date is delayed by 3.51 days. Figure 7b compares the PMC onset date Don calculated from equation (2)to the PMC onset date inferred from SBUV observations. The correlation coefficient of 0.85 shows a moderate improvement over the correlation coefficient of 0.60 obtained by correlating the observed PMC onset date to Lyman-αalone. Similar to the SH, in the NH Lyman-αexplains much of the low-frequency variation of the PMC onset, whereas the opposite hemisphere winter wind strength explains much of the high-frequency variation. No correlation was found between the residual of observed and regressed NH PMC onset and quasi-biennial oscillation (QBO), North Atlantic Oscillation (NAO), Arctic Oscillation (AO), and El Niño Southern Oscillation (ENSO). Similarly, in the SH there were no statistically significant correlations between residual differences calculated by differencing the two curves inFigure 6 and the QBO, NAO, AO, or ENSO.

image

Figure 7. (a) Difference between the NH PMC onset date and solar Lyman-alpha (black) versus zonal mean zonal wind at 60.0°S, 20 hPa, 10 days before the onset date (red, right axis). (b) SBUV (black) and regressed (red) time series of NH PMC onset date using solar Lyman-α and the wind plotted in Figure 7a as linear components.

Download figure to PowerPoint

[26] Variability in the PMC season onset date may be increased beyond natural causes by several factors: space shuttle launches and time-varying instrumental effects like sampling of different scattering angles and local times. Observations at varying scattering angles lead to changing cloud detection sensitivity. This study focuses on polar latitudes, where SBUV observes the season onset at scattering angles between 100 and 120°. In this region the Mie scattering phase function is rather flat compared to forward scattering, and we find no clear dependence of PMC season onset date on scattering angle. Thus scattering angle variations may be able to influence the timing of the PMC season onset as determined from some instruments, but this is not a factor in the work presented here. Cloud frequencies are observed to vary in local time due to temperature oscillations caused by tides [von Zahn et al., 1998; Chu et al., 2003, 2006; Fiedler et al., 2005; Stevens et al., 2010]. For this study, the PMC onset for each year is calculated from at least two SBUV observations (ascending and descending node) of the PMC onset, and up to eight observations (both nodes from four instruments). Therefore, in most years the spread in local time during the PMC onset observed by SBUV is huge, usually covering five to ten hours ranging from 1 A.M. to 10 P.M. in the SH, and from 3 A.M. to 7 P.M. in the NH. We found no clear connection between the timing of the cloud onset and local time, thus we do not believe that local time is a major driver of the season onset variability observed by SBUV.

[27] Space shuttle launches inject huge amounts of water vapor into the lower mesosphere that have the potential to be transported into the polar regions and increase PMC frequency [Stevens et al., 2002, 2003, 2005]. We found that no space shuttles were launched up to five (six) days prior to any NH (SH) season observed by SBUV. Transport time of more than five or six days will cause the water vapor plume to diffuse considerably before it reaches the polar region. Another unknown is whether any SBUV instruments were observing at the right location and time to observe a potential increase in PMC frequency at all. Due to its low sensitivity and low horizontal resolution SBUV may not pick up the signal or may report it as a very faint one that does not trigger the season onset. Since the season onset detection algorithm explained in section 2.3 rejects sporadic frequency increases, only water vapor plumes injected just before the natural season onset would have the potential to skew the onset determination to an earlier onset. There were only four years in both the NH and SH when shuttle launches occurred within eight days of the SBUV observed season onset; this would not significantly affect the correlations derived here. Thus it is unlikely that space shuttle launches have affected the PMC onsets observed by SBUV in any significant way.

4. Discussion and Conclusions

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Data Description and Analysis
  5. 3. Results
  6. 4. Discussion and Conclusions
  7. Acknowledgments
  8. References

[28] Using more than a quarter century of SBUV PMC observations and re-analysis wind data, this work builds on the study carried out byKarlsson et al. [2011] and Gumbel and Karlsson [2011] who found that the timing of the SH stratospheric wind reversal from winter to summer conditions plays an important role in controlling the SH PMC onset date. This work extends these previous findings by showing a high correlation between the timing of the SH PMC season onset date and the timing of the SH stratospheric wind reversal from winter to summer conditions for observations from 1984 to 2011. The mechanism for this phenomenon was summarized by Karlsson et al. [2011]: an early reversal from winter eastward to summer westward winds causes an early onset of a net eastward GW drag that leads to an early deceleration of the mesospheric mean zonal flow and thus to equatorward meridional drift, mesospheric upwelling and adiabatic cooling. This early onset of cold temperatures then allows early formation of PMCs.

[29] Beyond showing that the mechanisms described by Karlsson et al. [2011] and Gumbel and Karlsson [2011]hold for PMC seasons back to 1984, this work also determines the role of the solar cycle in the SH and NH PMC season onset, and investigates other possible controlling factors like inter-hemispheric coupling. The solar signal is expected to have an influence on PMCs because photolysis of upper mesospheric water vapor is modulated by the changing solar Lyman-α(121.6 nm) flux, and mesospheric temperatures are modulated by the changing solar UV heating rate between solar minimum and maximum conditions. Correlation of the SH season onset date and the solar cycle did not immediately reveal any connection. However, multiple linear regression of the PMC season onset date to stratospheric wind reversal and Lyman-α showed that the SH PMC onset is controlled primarily by the SH stratospheric wind reversal and secondarily by the solar cycle. The multiple linear regression coefficients obtained from equation (1)indicate that a one-day change in the timing of the stratospheric wind reversal from a winter to summer state results in a one-day change in the SH PMC onset date. The wind reversal date from 1984 to 2011 ranges from around −37 to +3 days from solstice (seeFigure 2), for a total range of 40 days. Thus, in the absence of other effects, variations in the same hemisphere stratospheric wind reversal would cause the latest PMC season onset to be delayed by about 40 days from the earliest PMC season onset for the years analyzed here. Likewise, a change in Lyman-α of 1011 cm−2 s−1results in a change in the PMC onset date by 5.00 ± 1.34 days. Such a change in Lyman-α corresponds to roughly half the solar cycle, which suggests that if this were the only relevant mechanism, the SH PMC onset date would be about ten days later at solar maximum than at solar minimum. Since the solar cycle and wind reversal do not vary synchronously, the actual range of observed PMC onset dates is ∼42 days, or slightly smaller than the sum of the two effects.

[30] The NH stratospheric mean zonal wind reversal from winter to summer conditions takes place about two months earlier than the onset of the PMC season, and thus has no influence. This is due to higher PW activity in the NH that leads to an earlier breakup of the polar vortex and therefore earlier wind reversal than in the SH. The wind reversal causes the mesospheric temperature to drop (not shown), but these temperatures are not low enough yet for PMCs to form until one to two months later when the temperature drops below 140 to 150 K. In addition, the mean zonal stratospheric wind speed at the time of the PMC onset does not correlate with the PMC season onset either (not shown). The solar signal shows a moderate correlation with the NH PMC season onset date, with a maximum correlation coefficient of 0.60, which was obtained without a lag. Multiple linear regression coefficients obtained from equation (2)indicate that a change in Lyman-α of 1011 cm−2 s−1results in a change in the PMC onset date by 3.51 ± 0.79 days. It suggests that the NH PMC onset date should be delayed by about seven days at solar maximum compared to solar minimum. In the NH, inter-hemispheric coupling of the opposite hemisphere stratospheric winds plays a slightly bigger role in controlling the PMC onset dates than the solar cycle: A one m/s change in the zonal mean zonal wind at 60.0°S, 20 hPa (∼26 km) and ten days before the NH PMC onset results in a change in the NH PMC onset date of 0.72 ± 0.13 days. Relative to the observed variations in the wind this means that the NH PMC onset should be delayed by about fifteen days at maximum wind speed compared to minimum wind speed. In neither hemisphere is there a correlation between the QBO, NAO, AO, or ENSO with either the PMC onset date or with the difference between PMC onset dates that were observed and calculated through linear regression.

[31] Other possible causes of PMC onset variability such as instrument sampling and space shuttle effects were discussed. We found that neither varying scattering angles nor local times were major drivers of the PMC onset as observed by SBUV. While space shuttle launches may have an effect on PMC occurrence frequency as observed by SBUV, it is unlikely that they have affected the PMC onsets observed by SBUV in any significant way.

[32] In summary we conclude that PW effects on stratospheric winds and consequent GW filtering trigger changes in mesospheric circulation that significantly affect the PMC season onsets in both hemispheres. The SH PMC season onset is controlled primarily by timing of the SH stratospheric wind reversal from its winter to summer state, with a smaller but still important contribution from the solar cycle; inter-hemispheric coupling appears to have a minimal effect. The strong dependence of the SH PMC onset date on the timing of the stratospheric wind reversal could have implications for long-term trends, since any trend in the stratospheric wind reversal will also affect the PMC onset [see, e.g.,Smith et al., 2010; Lossow et al., 2012]. The NH PMC season onset is controlled primarily by changes in the opposite hemisphere stratospheric wind, but with a large contribution from the solar cycle. It is possible that the 2-day and 5-day waves have a minor effect on the PMC onset dates and could explain some of the so far unexplained variability, especially in the NH. Neither the timing of the NH stratospheric wind reversal nor the NH stratospheric wind speed at the time of the NH PMC onset have any influence on the NH season onset.

Acknowledgments

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Data Description and Analysis
  5. 3. Results
  6. 4. Discussion and Conclusions
  7. Acknowledgments
  8. References

[33] The SBUV/2 data were obtained from NOAA/NESDIS with support from the NOAA Climate and Global Change Atmospheric Chemistry Element. Funding for S.B., C.E.R., G.E.T., and V.L.H. was provided by the NASA Small Explorer program under contract NAS5–03132. S.B. was also supported by the NSF CEDAR program under grant AGS 0737705. M.T.D. was partially supported by the NASA Geospace Science program through grant NNH09CF72C. E.P.S. was partially supported by the NASA Earth Science program through grant NNH08CD48C. We thank ECMWF for producing the ERA-40 reanalysis and the National Center for Atmospheric Research for distributing the reanalysis data. We thank colleagues at the United Kingdom Meteorological Office for producing the MetO analyses and the Distributed Active Archive Center at the Goddard Space Flight Center and the British Atmospheric Data Centre for distributing the MetO data. Lyman-α data were obtained from http://lasp.colorado.edu/lisird/.

References

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Data Description and Analysis
  5. 3. Results
  6. 4. Discussion and Conclusions
  7. Acknowledgments
  8. References