SEARCH

SEARCH BY CITATION

Keywords:

  • cloud cover;
  • galactic cosmic rays;
  • solar irradiance

Abstract

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Datasets
  5. 3. Methods
  6. 4. Results and Discussion
  7. 5. No Evidence of a GCR-Based Solar-Cloud Link
  8. 6. Conclusions
  9. Acknowledgments
  10. References
  11. Supporting Information

[1] Although over centennial and greater timescales solar variability may be one of the most influential climate forcing agents, the extent to which solar activity influences climate over shorter time periods is poorly understood. If a link exists between solar activity and climate, it is likely via a mechanism connected to one (or a combination) of the following parameters: total solar irradiance (TSI), ultraviolet (UV) spectral irradiance, or the galactic cosmic ray (GCR) flux. We present an analysis based around a superposed epoch (composite) approach focusing on the largest TSI increases and decreases (the latter occurring in both the presence and absence of appreciable GCR reductions) over daily timescales. Using these composites we test for the presence of a robust link between solar activity and cloud cover over large areas of the globe using rigorous statistical techniques. We find no evidence that widespread variations in cloud cover at any tropospheric level are significantly associated with changes in the TSI, GCR or UV flux, and further conclude that TSI or UV changes occurring during reductions in the GCR flux are not masking a solar-cloud response. However, we note the detectability of any potential links is strongly constrained by cloud variability.

1. Introduction

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Datasets
  5. 3. Methods
  6. 4. Results and Discussion
  7. 5. No Evidence of a GCR-Based Solar-Cloud Link
  8. 6. Conclusions
  9. Acknowledgments
  10. References
  11. Supporting Information

[2] The Sun is one of the most important factors responsible for governing climate change over centennial and greater timescales [e.g., Versteegh, 2005; Bard and Frank, 2006; Beer et al., 2006], however the extent to which solar activity may influence Earth's climate over shorter time periods is still poorly understood. Several mechanisms have been proposed which may account for a solar-climate link, including: a connection between changes in total solar irradiance (TSI) absorbed over cloud free regions of Earth's oceans, leading to modifications of synoptic circulation patterns [Meehl et al., 2009]; a link between ultraviolet (UV) spectral irradiance changes and stratospheric temperatures resulting from alterations to stratospheric ozone production [Austin et al., 2008], which may impact large scale tropospheric variability via dynamic stratosphere-troposphere couplings [Haigh, 1996]; and, a link between the galactic cosmic ray (GCR) flux and cloud properties, via either an ion-mediated (clean-air) pathway, or a global electric circuit (GEC) related (near-cloud) pathway [Carslaw et al., 2002]. The response times of these mechanisms ranges from minutes (for mechanisms concerning the GEC) [Tinsley et al., 2001], to periods of around a week (for mechanisms concerning the growth of cloud condensation nuclei or dynamic tropospheric links) [Arnold, 2007]. It should also be noted that irradiance-based mechanisms have been suggested which operate over long (annual-decadal) timescales (e.g. [White et al., 2003]), which are beyond the scope of this work.

[3] These mechanisms all have the potential to amplify relatively low energy changes in solar activity into climatologically significant effects. Such an amplification is necessary to account for the findings of palaeoclimatic reconstruction studies which have demonstrated the existence of a pervasive association between solar activity and numerous climatological parameters [e.g., Ram and Stolz, 1999; Bond et al., 2001; Mauas et al., 2011].

[4] To investigate the possibility of daily timescale solar-climate links a number of investigations have focused on the effects of high amplitude short term reductions in the GCR flux (known as Forbush Decrease (FD) events) on cloud properties. However, these studies have shown a range of conflicting results. While some have demonstrated the presence of significant positive correlations between cloud changes and solar activity [Pudovkin and Veretenko, 1995; Todd and Kniveton, 2004; Harrison and Ambaum, 2010], others have shown significant negative correlations [Wang et al., 2006; Troshichev et al., 2008], or found no compelling evidence of significant correlations [Pallé and Butler, 2001; Kristjánsson et al., 2008; Čalogović et al., 2010; Laken et al., 2009, 2011].

[5] A number of possibilities may account for these conflicting findings: 1) no relationship exists between solar activity and climate; 2) a relationship exists, but it is constrained by the atmospheric conditions at the time (i.e. not a first order relationship) [Laken et al., 2010]; 3) even at daily timescales studies have often failed to properly isolate the effects of various solar parameters, consequently this may have interfered with their results [Laken et al., 2011]; 4) FD studies deal with inherently small sample sizes, as the number of high magnitude solar events is low and consequently sample noise is high, limiting the detectability of any solar-cloud signals. With regards to addressing the third possibility, this work presents an analysis of the largest daily timescale TSI variations using an epoch-superposition (composite) approach to test for the presence of reliable daily-timescale link to satellite-detected cloud variability. Samples have been carefully selected to isolate a range of periods undergoing substantial changes in solar activity of hypothesized significance to atmospheric variability.

2. Datasets

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Datasets
  5. 3. Methods
  6. 4. Results and Discussion
  7. 5. No Evidence of a GCR-Based Solar-Cloud Link
  8. 6. Conclusions
  9. Acknowledgments
  10. References
  11. Supporting Information

[6] This investigation uses measurements of TSI emissions, cloud cover, the GCR flux, and the 10.7 cm solar radio flux (F10.7). TSI data are taken from the Active Cavity Radiometer Irradiance Monitor (ACRIM) [Willson and Mordvinov, 2003].

[7] Variations in solar irradiance are caused by the presence of surface features such as sunspots and faculae, which differ in brightness to the average surface intensity and cumulatively contribute to fluctuations in TSI. Variability at UV wavelengths is often connected with the presence of plages: these are large bright complexes connected to magnetically active regions in the chromosphere. The rotation of features such as plages and sunspots across the solar disk alters the amount of energy emitted towards the Earth (for further details see Lean and Woods [2010]). Changes in the GCR flux are linked to the solar wind and associated disturbances (such as coronal mass ejections CMEs). As the solar wind travels at supersonic speeds, variations in the GCR flux resulting from solar disturbances can take up to 2–4 days to occur [Brueckner et al., 1998], whereas irradiance associated changes are experienced on Earth almost instantaneously following a solar event.

[8] Cloud data are taken from the International Satellite Cloud Climatology Project (ISCCP) D1 dataset infrared (IR) channels [Rossow and Schiffer, 1991]. ISCCP data are created from inter-calibrated radiance measurements recorded from polar orbiting and geostationary satellites. These data are provided globally over on an equal-area grid of 280 × 280 km2, at a 3-hour temporal resolution from 1983–2008. In this investigation daily average cloud data (retrieved at IR wavelengths) is used at 3 different altitude levels: for high (>6.5 km), middle (2–6.5 km) and low (0–3.2 km) clouds. At each of these altitude levels a further distinction is also drawn between high (>45°) and low (<45°) latitude regions as well as over ocean and land regions. This area-averaging method was selected both in order to reduce noise associated with daily cloud variability and because the theoretical effects of solar irradiance and GCR flux on cloud cover may vary across spatial domains.

[9] GCR flux data are derived from the count rate recorded by the Climax Colorado neutron monitor (39.37 N, −106.18 W, 3400 m, 2.99 GV). The F10.7 (2800 Mhz) radio flux is used as a proxy of extreme ultraviolet (EUV) solar activity [Rich et al., 2003]. All used data (except the F10.7) have been normalized to a static averaging period from −20 to −10 days prior to the key composite date (i.e. all values displayed are an anomaly calculated against a 10-day static averaging period).

3. Methods

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Datasets
  5. 3. Methods
  6. 4. Results and Discussion
  7. 5. No Evidence of a GCR-Based Solar-Cloud Link
  8. 6. Conclusions
  9. Acknowledgments
  10. References
  11. Supporting Information

[10] Three composite samples were constructed for this analysis. The first sample isolates dates of significant increases in TSI occurring over a five day period where the increase is half of the key date in magnitude and shall hereafter be referred to as IncTSI. The second sample isolates dates of significant decreases in TSI over a five day period, and will be referred to as DecTSI-A. The third sample referred to as DecTSI-B is identical to the DecTSI-A sample, however it is further restricted to events which show no significant GCR variations within ±10 days of the key composite date.

[11] The composited events were selected from a population of the largest (95 percentile) daily timescale increases/decreases in TSI from 1978 to 2010. To calculate the changes in the TSI record, a seven-day running mean was subtracted from a 31-day reference period, these original populations were then reduced by excluding consecutive dates, leaving only the date of greatest TSI deviation. Finally, events were also removed from the populations if significant TSI deviations (at least half of the key date in magnitude) were observed to occur within a ±10 day period around the key date. This treatment further reduced the samples sizes to 19 events for theIncTSI sample, 48 events for the DecTSI-A sample and 37 events for the DecTSI-B sample presented in Table S1 in the auxiliary material.

[12] The correlation coefficient (r) values were calculated for each three samples during an analysis period of ±20 days between the TSI, GCR and F10.7 fluxes and the corresponding cloud data. The analysis period of the cloud data was extended by an additional 20 days allowing us to calculate the correlations with a lag of up to 20 days. Monte Carlo (MC) cased testing was employed to establish the threshold significance values for every obtained correlation: random composite samples using the whole available cloud dataset (ISCCP, 1983–2008) were constructed with sample sizes corresponding to the sample they were testing (e.g. n = 19 random events for the IncTSI sample). These random composites of cloud data were correlated with every investigated parameter, and this process was repeated 100,000 times. The resulting rvalues were found to be normally distributed, according to the Shapiro-Wilk test of normalcy (W = 0.996,p = 4.8 × 10−10) [Shapiro and Wilk, 1965]. Consequently, the statistical significance thresholds for this work are set by the two-tailed 95 percentile MC-generatedr values. This approach shows what a stochastic range of correlations should be given random sampling. If the solar variations of our samples were to affect cloud changes then the obtained correlations would be out of this stochastic range and a significant correlation would be detected.

[13] Consequently, this implies that for a solar-cloud signal to be detected the efficiency of the mechanism must be high (where efficiency here refers to the ability of a change in a solar parameter to influence a change in cloud, e.g. if a 1% change in the GCR flux induced a 1% change in cloud cover the efficiency is 100%): for example, in the case of low level clouds over ocean area regions we find that for 100,000 randomly generated samples of 48 events, there is an average sample noise of 0.83(±0.15)% over a 41-day period (Table 1). Thus, to detect a statistically significant TSI-cloud correlation with this level of noise under theIncTSI sample (TSI increase of ∼0.08%), a mechanism would have to have an efficiency greater than ([0.83/0.08]*100=) 1,038%. Whereas, in the case of the DecTSI-A sample, with a GCR reduction of 2.8%, a GCR mechanism would have to have an efficiency greater than 30% to produce a detectable signal.

Table 1. Simulated Standard Deviations of Cloud Data for Various Regionsa
Samplen = 48n = 37n = 19
  • a

    Standard deviations (1.96σ level) of 100,000 Monte Carlo simulations of ISCCP cloud data at various pressure levels: low altitude (LC), middle altitude (MC), and high altitude (HC), over low latitude (<45°), high latitude (>45°), ocean and land regions. Error range is shown in brackets. These values provide an estimate of the sample noise, to which any solar induced cloud changes must be greater than in order to be detectable. Values are presented for every composite size (n = 48, 37, 19).

LC - <45°0.95(±0.18)1.09(±0.21)1.51(±0.29)
LC - >45°1.75(±0.35)1.99(±0.40)2.78(±0.56)
MC - <45°1.22(±0.23)1.38(±0.26)1.93(±0.38)
MC - >45°1.44(±0.32)1.63(±0.36)2.27(±0.50)
HC - <45°1.51(±0.28)1.72(±0.33)2.40(±0.46)
HC - >45°3.30(±0.62)3.75(±0.72)5.20(±1.03)
LC - Land2.57(±0.60)2.92(±0.68)4.04(±0.95)
LC - Ocean0.83(±0.15)0.95(±0.17)1.33(±0.24)
MC - Land1.59(±0.37)1.80(±0.41)2.50(±0.57)
MC - Ocean1.11(±0.20)1.26(±0.23)1.75(±0.33)
HC - Land2.50(±0.47)2.84(±0.53)3.95(±0.75)
HC - Ocean1.83(±0.34)2.09(±0.39)2.90(±0.57)

4. Results and Discussion

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Datasets
  5. 3. Methods
  6. 4. Results and Discussion
  7. 5. No Evidence of a GCR-Based Solar-Cloud Link
  8. 6. Conclusions
  9. Acknowledgments
  10. References
  11. Supporting Information

[14] Figure 1 shows the variation in TSI, GCR and F10.7 occurring over the three composite samples. The IncTSI sample (denoted by blue lines in all figures) showed an increase in TSI emissions of 0.076% centered around the key date, beginning on day −4 before decreasing to approximately original values by around day +5 (Figure 1a). This sample displayed no strong GCR fluctuations over the composite until day +8, where a reduction in the GCR flux of 1.5% is observed over a two day period, followed by a fast recovery and another additional lower magnitude reduction on day +16 (Figure 1b). Closer inspection of the individual events revealed that only a few events are responsible for the GCR flux reductions (such as 26/08/1989, 06/03/1991, and 19/05/1991). The F10.7 shows a steady increase of 20% following the key date over a seven day period; values were then seen to plateau for 3 days before declining to undisturbed levels (Figure 1c). The F10.7 index shows a delay in changes of around one week. This delay corresponds to the journey time of the active regions from the solar limb to solar disc centre. Faculae responsible for TSI variations have their greatest influence at the solar limb [Carlsson et al., 2004], whereas plages (responsible for variations in the F10.7 index) have their greatest influence near the centre of the solar disc. A close inspection of the individual events in the IncTSI composite reveals that four events (07/06/2989, 13/05/1990, 22/01/1991 and 09/08/1991) are contributing to a large portion of the F10.7 increase.

image

Figure 1. Solar activity parameter variations over composite samples. This figure displays the mean variations of TSI flux, GCR flux, and the F10.7 index for the three composite samples: the IncTSI sample, which shows the largest (95 percentile) increases in TSI (displayed on the blue line, n = 19); the DecTSI-A sample, which shows the largest decreases in TSI (displayed on the green dotted line, n = 48); and, the DecTSI-B sample, which shows the largest decreases in TSI with significant GCR reductions excluded (displayed on the red dashed line, n= 37). The mean fluctuations in the individual parameters are shown for individually for the three samples as follows: (a) TSI changes, (b) the GCR flux, and (c) the F10.7 EUV index. Shaded regions indicate the standard error of the mean (at the 95 two-tailed confidence level). All values are an anomaly calculated against a 10-day static averaging period beginning on day −20.

Download figure to PowerPoint

[15] The DecTSI-A sample (denoted in all figures by red dashed lines) showed a strong decrease in TSI emissions of ∼0.13% centered on the key date over a five day period (Figure 1a). It also showed a stronger reduction in the GCR flux of ∼2.8%, centered on day +3 (Figure 1b) which can be explained by the travel time of the solar wind disturbances to the Earth. Variations in the F10.7 were found to inversely mirror the TSI emissions, showing an increase of ∼40% on the key date (Figure 1c).

[16] The DecTSI-B sample (denoted in all figures by green solid lines) showed TSI changes that corresponded closely to the DecTSI-A sample, although the peak reductions in the TSI emissions were of a marginally lower magnitude (Figure 1a). GCR reductions on day +3 are only ∼0.3%; an order of magnitude lower than those identified in the DecTSI-A sample and are considerably smaller than daily GCR variations (Figure 1b). The F10.7 values were virtually identical to the previously described DecTSI-A sample (Figure 1c).

[17] None of the samples showed any significant correlations between cloud anomalies and the TSI, GCR or F10.7 flux for any of the investigated regions (high, low latitudes or ocean and land regions) at any altitude level (low, middle and high) within the 20-day lag period (Figure 2 and Figure S1). However, we note that TSI showed slightly higher correlation coefficients than in the case of the GCR of F10.7 flux. Slightly higher, but non-significant correlation coefficients were also noticed in the case of ocean regions of theDecTSI-B sample compared to land regions, giving some support for the mechanism suggested by Meehl et al. [2009]. Closer inspection of the spatial distributions of correlation coefficients for TSI, GCR and F10.7 across the globe (not shown) revealed inhomogeneous sporadic correlations, with no field significance.

image

Figure 2. Cloud variations at high and low latitudes. Normalized ISCCP cloud cover changes (%) at (a and b) low levels (0–3.2 km), (c and d) middle levels (3.2–6.5 km), and (e and f) high levels (>6.5 km). Data are presented at low (<45°) and high (>45°) latitudes, over a −20 to +40 day period surrounding the key composite date. IncTSI sample denoted by blue line, DecTSI-A sample denoted by green dotted line, and DecTSI-Bsample denoted by red dashed line. All values are an anomaly calculated against a 10-day static averaging period beginning on day −20.

Download figure to PowerPoint

[18] Additionally, no significant correlations were identified between any dataset (TSI/GCR/F10.7) and globally averaged cloud cover, where the correlations were somewhat lower than those achieved by regional samples. This may be explained by the observation of the existence of correlations/anti-correlations in the cloud data (particularly high cloud between low and high latitude regions), which (assuming the changes represent absolute changes in cloud amount, as opposed to shifts in cloud cover between the latitude divisions) may cancel out when considering a global area.

[19] It is theoretically plausible that a correlation between solar activity and cloud may exist over smaller regions than those considered in this work. However, reliably detecting low amplitude signals with the data used in this study is highly problematic. This is primarily due to the fact that cloud datasets show a high degree of variability; the higher the spatial resolution of the experiment, the greater this issue becomes. This is demonstrated in Table 1, which displays MC simulated 95% confidence level cloud variations at low middle and high cloud levels for low and high latitude regions across the composite samples; in each instance the variations are found to strongly increase with decreasing region and sample size.

[20] This important observation indicates that the detectability of any potential signal over a given sample will be strongly constrained by the size of the area, and the size of the sample, and may account for the wide-ranging and conflicting results obtained by workers dealing with such studies. We essentially have attempted to attain a balance between area-averaging (as a method of minimizing sample noise), with selecting physically meaningful sample areas with respect to theoretical solar-climate linkages. The aim was to select our sample areas to reflect regions that may allow us to both detect a solar signal and attribute it to a mechanism, whilst minimizing potential noise-to-signal ratio issues.

5. No Evidence of a GCR-Based Solar-Cloud Link

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Datasets
  5. 3. Methods
  6. 4. Results and Discussion
  7. 5. No Evidence of a GCR-Based Solar-Cloud Link
  8. 6. Conclusions
  9. Acknowledgments
  10. References
  11. Supporting Information

[21] Comparing the cloud anomalies of the DecTSI-A and DecTSI-B samples (TSI reductions in the presence and absence of significant GCR reductions) yields no appreciable differences (Figure 1b). The same was observed in the case of ocean and land regions (Figure S1), where due to the different conditions low clouds cover ocean regions are claimed to be more sensitive to GCR changes [Yu and Turco, 2001; Marsh and Svensmark, 2003]. From this we conclude that in the case of the selected events the variations in the GCR flux do not significantly alter widespread cloud amount at any tropospheric level.

[22] Large TSI reductions have been found to accompany short term reductions in the GCR flux, and it has been suggested that the co-variations in these parameters may complicate the unambiguous attribution of changes in atmospheric properties to a specific cause, or even interfere with the detectability of potential signals [Laken et al., 2011], possibly influencing the results of studies [e.g., Pudovkin and Veretenko, 1995; Todd and Kniveton, 2004; Svensmark et al., 2009]. However, the analysis presented in this work shows that following careful isolation of TSI and GCR variations, neither is found to be significantly associated with changes in cloud cover.

[23] Recent results from the Cosmic's Leaving Outdoor Droplets (CLOUD) experiment at CERN has shown evidence that the GCR flux may enhance the formation of aerosols by ion-mediated nucleation [Kirkby et al., 2011]. However, the enhancement is low, implying that for the majority of tropospheric conditions the enhanced nucleation rate is not large enough to ultimately affect cloud condensation nuclei (CCN) numbers significantly. Such conclusions have also been made independently based on climate model results [Pierce and Adams, 2009]. The results we present here are in agreement with these findings, as we find no significant change in cloud properties following significant GCR fluctuations. Furthermore, our selection of certain regions (e.g. low and high latitudes) and altitude levels support the findings of the CLOUD experiment, which demonstrates that temperature and altitude play a primary role in determining ion induced aerosol nucleation [Kirkby et al., 2011].

6. Conclusions

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Datasets
  5. 3. Methods
  6. 4. Results and Discussion
  7. 5. No Evidence of a GCR-Based Solar-Cloud Link
  8. 6. Conclusions
  9. Acknowledgments
  10. References
  11. Supporting Information

[24] This work has attempted to test the notion of a link between solar activity and cloud cover, using several highly isolated composite samples. These samples reflected periods of increasing and decreasing TSI, the latter being in the presence/absence of significant reductions in the GCR flux. Although we successfully isolated periods of significant solar activity changes, we found no widespread detectable changes in cloud cover at any tropospheric level within a 20 day period of the solar forcing clearly associated with solar activity changes. Thus we can also conclude that TSI or UV changes occurring during reductions in the GCR flux are not masking a solar-cloud response. It is still possible that any small amplitude or low efficiency solar-cloud signals may be hidden by high meteorological variability of the cloud data, although the sample selection and large area averages utilized provide some compensation for this effect.

Acknowledgments

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Datasets
  5. 3. Methods
  6. 4. Results and Discussion
  7. 5. No Evidence of a GCR-Based Solar-Cloud Link
  8. 6. Conclusions
  9. Acknowledgments
  10. References
  11. Supporting Information

[25] The authors thank Bojan Vršnak (Hvar Observatory), Enric Pallé (Instituto de Astrofísica de Canarias) and Dominic Kniveton (University of Sussex) for comments. ACRIM data obtained from http://www.acrim.com/Data%20Products.htm. ISCCP data are available from http://isccp.giss.nasa.gov, obtained from NASA Langley Research Centre Atmosphere Science Data Center. This research received funding from the European 115 Commission's Seventh Framework Programs (FP7/2007-2013), grant 116, 218816. We also thank Geoffrey Tyndall and two anonymous reviewers. The authors would like to acknowledge the support of the European COST Action ES1005.

[26] The Editor thanks two anonymous reviewers for their assistance in evaluating this paper.

References

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Datasets
  5. 3. Methods
  6. 4. Results and Discussion
  7. 5. No Evidence of a GCR-Based Solar-Cloud Link
  8. 6. Conclusions
  9. Acknowledgments
  10. References
  11. Supporting Information

Supporting Information

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Datasets
  5. 3. Methods
  6. 4. Results and Discussion
  7. 5. No Evidence of a GCR-Based Solar-Cloud Link
  8. 6. Conclusions
  9. Acknowledgments
  10. References
  11. Supporting Information

Auxiliary material for this article contains a table with sample events and dates and a figure showing cloud variations.

Auxiliary material files may require downloading to a local drive depending on platform, browser, configuration, and size. To open auxiliary materials in a browser, click on the label. To download, Right-click and select “Save Target As…” (PC) or CTRL-click and select “Download Link to Disk” (Mac).

Additional file information is provided in the readme.txt.

FilenameFormatSizeDescription
grl28708-sup-0001-readme.txtplain text document1Kreadme.txt
grl28708-sup-0002-ts01.txtplain text document1KTable S1. List of sample events.
grl28708-sup-0003-fs01.tifTIFF image2871KFigure S1. Cloud variations over land and ocean regions.
grl28708-sup-0004-t01.txtplain text document1KTab-delimited Table 1.

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.