Abstract

Based on the results of decadal correlation studies between the International Satellite Cloud Climatology Project–detected cloud anomalies and the galactic cosmic ray (GCR) flux, it has been suggested that a relationship exists between solar activity and cloud cover. If valid, such a relationship could have important implications for scientists’ understanding of recent climate change. In this work, an analysis of the first decade of Moderate Resolution Imaging Spectroradiometer (MODIS)–detected cloud anomalies are presented, and the data at global and local geographical resolutions to total solar irradiance (TSI), GCR variations, and the multivariate El Niño–Southern Oscillation index are compared. The study identifies no statistically significant correlations between cloud anomalies and TSI/GCR variations, and concludes that solar-related variability is not a primary driver of monthly to annual MODIS cloud variability. The authors observe a net increase in cloud detected by MODIS over the past decade of ~0.58%, arising from a combination of a reduction in high- to midlevel cloud (−0.31%) and an increase in low-level cloud (0.89%); these long-term changes may be largely attributed to ENSO-induced cloud variability.

1. Introduction

In 1997 a positive correlation was noted between variations in oceanic cloud cover over geostationary satellite regions and the solar-modulated galactic cosmic ray (GCR) flux for the period of 1983–94 (Svensmark and Friis-Christensen 1997). In this study, cloud measurements were taken from the International Satellite Cloud Climatology Project (ISCCP) (Rossow and Schiffer 1999). The ISCCP data are an intercalibration of radiance measurements, recorded by a fleet of polar-orbiting and geostationary satellites, provided at an 8-times-daily temporal resolution since 1983. Data from this project have been widely utilized for studying the possibility of a relationship between solar activity and cloud cover (e.g., Pallé and Butler 2000; Todd and Kniveton 2001; Kristjánsson et al. 2002; Laut 2003; Marsh and Svensmark 2003; Kristjánsson et al. 2004; Čalogović et al. 2010; Laken et al. 2010, 2011).

Solar activity has increased over the last century (Lean et al. 1995; Solanki et al. 2004). Consequently, the amount of low-energy GCRs impinging on earth’s atmosphere has decreased during this period. This observation led Svensmark (2007) to suggest that GCR-modulated cloud decreases may be primarily responsible for the recent anomalously high global temperatures and not anthropogenic emissions; such claims have also been made by other researchers based on similar arguments (e.g., Shaviv 2005; Rao 2011). However, it is unlikely that these claims are valid, as since 1987 solar activity levels have remained constant and thus cannot account for the observed warming climate trends (Lockwood and Fröhlich 2008), and the historical cloud datasets do not seem to show evidence of such GCR–cloud connections (Pallé and Butler 2002). Furthermore, the arguments supporting an anthropogenic (and not solar based) warming of climate over the past several decades are more consistent with the observed temperature records (Solomon et al. 2007).

Despite these facts, notions of a solar–cloud link still warrant further investigation, particularly as a range of palaeoclimatic reconstructions have found strong evidence implying the presence of pervasive links between solar activity cycles and variations in earth’s climate from a range of distinct sources over periods of up to thousands of years, such as variations in monsoonal activity recorded by stalagmite growth (Fleitmann et al. 2003); periodic ice rafting events in the North Atlantic (Bond et al. 2001); variations in dust layers present in Greenland ice core samples (Ram and Stolz 1999); and hemispheric-scale climate anomalies recorded by tree rings (Yamaguchi et al. 2010). Such connections are difficult to explain via a conventional understanding of solar–terrestrial connections and may imply the presence of an unknown mechanism(s) linking small changes in solar activity to the earth’s climate.

A variety of pathways linking changes in solar activity to atmospheric variability have been proposed, and detailed descriptions of these pathways are given by Haigh (1996), Carslaw et al. (2002), and Meehl et al. (2009). Such connections may provide additional sources of unaccounted for natural climate variability, which could have been a significant factor influencing past climate and may continue to operate as a source of contemporary (albeit likely low amplitude) climate variability (Beer et al. 2000).

In the context of such palaeoclimatic studies, the notion of a cloud-based solar–climate link is particularly intriguing, as such a connection would have the potential to amplify the relatively small solar impetus into a climatologically significant effect. This is achieved as the radiative forcing effects of clouds are large enough, that even a small change in cloud amount may exert an influential radiative forcing (Slingo 1990); for example, an increase of global cloud cover by 1% would alter earth’s radiative budget by approximately −0.13 W m−2 (Ramanathan et al. 1989). An amplifying mechanism such as this is essential to the idea of a solar–climate link, as the total energy variance of the sun over an 11-yr solar cycle possesses an amplitude of only ~0.1%; assuming a conservative climate sensitivity of 0.3–1.0°C (W m−2)−1, this value is too small to be of direct significance to earth’s climate (Kelly and Wigley 1992; Schlesinger and Ramankutty 1992; ,Lean and Rind 1998).

Subsequent reanalysis of the initial Svensmark and Friis-Christensen (1997) GCR–cloud relationship found that the observed correlation was restricted to low-altitude (>680 mb) cloud cover only (Marsh and Svensmark 2000; Pallé and Butler 2000) and even this relationship was found to break down after 1994 (Laut 2003). It was claimed by Marsh and Svensmark (2003) that this breakdown was the result of a satellite calibration issue. However, this idea is questionable, as while adjustments to the ISCCP dataset occurring following the addition of new satellite instruments to the network may conceivably result in a jump in the dataset, it would not produce a change necessary to obscure a long-term correlation. As a further issue to this notion, it seems that the long-term trends in the ISCCP dataset may largely result from satellite-viewing geometry artifacts and not from physical changes in the atmosphere (Evan et al. 2007).

The accuracy of long-term ISCCP cloud studies is also limited by errors in the data: estimations of total relative uncertainties in the radiance calibrations are around ≤5% for visible (VIS) and ≤1% for infrared (IR) measurements, where absolute calibration uncertainties are estimated to be around 10% and 2%, respectively (Brest et al. 1997). Additionally, criticisms regarding long-term solar–cloud correlation studies have been raised in relation to the detected relationships being influenced by internal oscillations, such as El Niño (Farrar 2000), and also by long-term artificial errors in the ISCCP dataset itself (Norris 2000; Pallé 2005).

2. Rationale and datasets

In this study we aim to provide an overview of global- and local-scale correlations between cloud anomalies and solar activity over the last decade. Cloud data are drawn from the Moderate Resolution Imaging Spectroradiometer (MODIS) project. This dataset provides several advantages over ISCCP measurements utilized by similar studies; most significantly, MODIS provides a greatly increased spectral resolution from which to determine cloud properties. As MODIS has been active since 2000, data are available for almost a complete solar cycle, providing an intriguing opportunity to further investigate the possibility of long-term relationships between cloud cover and solar activity in an independent and modern cloud dataset.

Because of the parallels between this work and similar long-term analysis carried out with ISCCP data, it is important to highlight some of the primary differences between the MODIS and ISCCP datasets: 1) the ISCCP samples the entire globe 8 times per day from a network of geostationary and polar-orbiting satellites, whereas MODIS samples over a 1–2-day period from two instruments on board the Earth Observing Systems Aqua and Terra polar-orbiting platforms. 2) ISCCP uses 11 channels operating between 0.4 and 14.12 μm (although the actual number of channels viewed by each individual satellite in the network varies from 2 to 11), whereas MODIS operates at much higher spectral resolutions, using 36 channels (between 0.405 and 14.385 μm). 3) To determine cloud-top pressure ISCCP uses emissivity, whereas MODIS uses a method known as CO2 slicing; this technique has proven to be more sensitive to certain cloud types than the ISCCP method (Wylie and Menzel 1999). 4) ISCCP performs both VIS-and-IR-combined retrievals during daytime and IR-only cloud retrievals during nighttime and compares these values to adjust the nighttime (IR only) retrievals, whereas MODIS provides only a simple daily mean statistic of retrieved cloud pixels during daytime. 5) Over certain regions MODIS observes larger cloud amounts than ISCCP; this difference is most pronounced at low-altitude levels, over land during daytime (Pincus et al. 2006, 2012). It is important to note that there is no clear definition of cloud fraction that is applicable across distinct observing systems; MODIS cloud fractions are comparable to ISCCP, which in turn have been tested against other measures of cloud fraction from surface-based observations (Rossow et al. 1993). MODIS cloud fractions are relatively conservative and frequently reject patchy low-level cloud because of ambiguous retrievals; this may contribute to a bias in the dataset.

This investigation also uses both total solar irradiance (TSI) and GCR flux data: TSI data are taken from the Physikalisch-Meteorologisches Observatorium Davos (PMOD) World Radiation Center composite (Fröhlich and Lean 1998; Fröhlich 2000), while measurements of the GCR flux are derived from the Moscow neutron monitor (55.47°N, 37.32°E; 200 m; 2.43 GV). In addition, an index of the El Niño–Southern Oscillation (ENSO) is also used: the multivariate ENSO index (MEI) (Wolter and Timlin 1993). The MEI is created from the combined first unrotated principal component of six observed climate parameters over the topical Pacific, specifically, sea level pressure, zonal and meridional surface winds, sea surface temperatures, surface air temperatures, and total cloud fraction.

3. Methodology

In the following analysis, the mean of the seasonal cycle is removed (deseasonalized) and the cloud anomalies are compared to TSI, GCR, and MEI variations; all data are utilized at a monthly temporal resolution. The deseasonalization is performed on the cloud data by differencing individual monthly averages from the respective (grouped) monthly average over the entire dataset.

The cloud data used in this work are drawn from the combined measurements of the MODIS instruments on board both the Terra and Aqua platforms. No artificial changes are introduced into the resulting time series as a consequence of this combination, and it is noted that the individual (overlapping) Terra and Aqua cloud fraction time series show a strong correlation (r = 0.87) (Fig. 1). The combination of Terra and Aqua data should provide an additional level of robustness to the results, as averaging across independent MODIS sensors will reduce the possibility of instrumental errors influencing the results. Monthly cloud amount at high-altitude (<440 mb), midaltitude (440–680 mb), and low-altitude (>680 mb) levels are calculated from the MOD08 (daily) level 3, collection 5.1 data (from the “cloud-top pressure mean” and “cloud fraction combined mean” parameters). The distinction between low-, mid-, and high-altitude levels is based on ISCCP definitions. It is noted that collection 5.1 Terra data includes a long-term calibration drift of currently unknown origins; however, this issue does not affect the cloud fraction or cloud-top pressure parameters.

Fig. 1.

Agreement between MODIS Terra and Aqua cloud anomalies. Deseasonalized global MODIS cloud anomalies from both Terra (solid line) and Aqua (dashed line) instruments over the period of 2000–11. Data shows an R of 0.87.

Fig. 1.

Agreement between MODIS Terra and Aqua cloud anomalies. Deseasonalized global MODIS cloud anomalies from both Terra (solid line) and Aqua (dashed line) instruments over the period of 2000–11. Data shows an R of 0.87.

Evaluations of the degree of association between the datasets are made using correlation coefficients (R). Cloud data are analyzed at both a globally averaged geographical resolution and over a local (1° × 1°) resolution at varying altitude levels. To accurately determine the significance of the R values, Monte Carlo (MC) simulation techniques are employed in the following manner: detrended monthly GCR/TSI data for the entire data period (i.e., the 10-yr trend is removed) are randomized and correlated against ±24-month periods of detrended monthly cloud data 10 000 times; the critical significance thresholds are then determined based on the resulting (normally distributed) R values. An analysis of autocorrelation within the datasets showed deseasonalized MODIS cloud anomalies have only negligible autocorrelation, while the TSI/GCR and MEI datasets showed autocorrelations of around 15 months. As this period is far shorter than the total data period and the MC methodology makes no assumptions regarding independence, autocorrelation will not influence our estimates of statistical significance to a notable degree. For an examination of the significance of cross correlations in the deseasonalized and linearly detrended data, lagged significance thresholds are calculated at globally averaged resolutions. For an analysis of local significance levels, the MC procedure is performed at the individual gridcell level (at zero-month lag).

4. Results

The deseasonalized cloud anomalies are presented in Fig. 2. The total anomaly exhibits virtually no long-term trend (0.004% month−1) and possesses an average standard deviation over the data period of 1.04% (Fig. 2a). Cloud anomalies detected at high-altitude levels exhibit a slight decreasing tendency over the data period of −0.005% month−1 and show an average standard deviation of 0.90% (Fig. 2b). The midaltitude-level cloud anomalies show a small increasing tendency (0.002% month−1) and an average standard deviation of 0.78% (Fig. 2c). The low-level anomalies show an increasing trend of 0.0069% month−1 and an average standard deviation of 0.78% (Fig. 2d).

Fig. 2.

Anomalous cloud trends. Deseasonalized monthly cloud anomaly (%) detected by MODIS (Aqua and Terra combined) between March 2000 and February 2011: (a) total anomaly, (b) high-level (<440 mb) anomaly, (c) midlevel (440–680 mb) anomaly, and (d) low-level (>680 mb) anomaly. Solid line shows mean anomaly, while dotted lines above and below display one standard deviation (σ) level of the monthly averaged cloud values. Linear fits are also displayed, along with both the regression equations (Y) and R values.

Fig. 2.

Anomalous cloud trends. Deseasonalized monthly cloud anomaly (%) detected by MODIS (Aqua and Terra combined) between March 2000 and February 2011: (a) total anomaly, (b) high-level (<440 mb) anomaly, (c) midlevel (440–680 mb) anomaly, and (d) low-level (>680 mb) anomaly. Solid line shows mean anomaly, while dotted lines above and below display one standard deviation (σ) level of the monthly averaged cloud values. Linear fits are also displayed, along with both the regression equations (Y) and R values.

Figure 3 shows both the monthly normalized GCR flux (%) and the normalized TSI (W m−2) outside the atmosphere. The GCR flux is relatively low (approximately −6.82%) from 2000 to 2003 during the maximum of solar cycle 23. It then increases by around 10% from 2003 to 2006, after which time it remains relatively stable (at around 6.30%), only beginning to decrease in the latter part of 2009 (Fig. 3a). TSI demonstrates approximately similar but anticorrelated behavior (the GCR and TSI data possess an R value of −0.76 over the data period): during the first half of the decade, TSI shows a relatively large monthly standard deviation (during solar maximum) of around 0.46 W m−2; however, after this period this value drops to only 0.08 W m−2. The approximate amplitude of the solar maximum to minimum TSI change over the period is seen to be 0.90 W m−2 (outside of the atmosphere) (Fig. 3b).

Fig. 3.

GCR and TSI trends. Monthly (a) GCR flux (%) from Moscow neutron monitor and (b) TSI (W m−2) from the PMOD reconstruction. Values are normalized against their respective averages over the entire period (March 2000–February 2011). Solid line shows mean anomaly, while dotted lines above and below display one σ level of monthly averaged values. Linear fits are also displayed, along with both Y and R values.

Fig. 3.

GCR and TSI trends. Monthly (a) GCR flux (%) from Moscow neutron monitor and (b) TSI (W m−2) from the PMOD reconstruction. Values are normalized against their respective averages over the entire period (March 2000–February 2011). Solid line shows mean anomaly, while dotted lines above and below display one σ level of monthly averaged values. Linear fits are also displayed, along with both Y and R values.

An examination of the cross correlations between cloud at total, high, mid-, and low levels and the TSI/GCR flux over a ±24-month period is presented in Fig. 4. No strongly significant associations are identified, although several weakly significant correlations can be seen at a variety of time lags and inconsistent signs. The only variables found to show a statistically significant (negative) correlation above the 0.99 critical level were the low cloud and the TSI flux. However, the TSI data showed a significant correlation at a +11-month time lag, precluding the possibility of a causal relationship.

Fig. 4.

Cloud–solar cross correlations for detrended data. Cross-correlation plots for ±24-month time lags for both TSI (solid line) and GCR (dashed line) against deseasonalized monthly cloud anomalies at varying altitude levels: (a) total, (b) high-level (<440 mb), (c) midlevel (440–680 mb), and (d) low-level (>680 mb). The linear trends (presented in the Y equations of Figs. 1, 2) are removed from the datasets prior to correlation analysis. Second-order polynomial fits of MC-simulated significance thresholds are also displayed over the ±24-month lag period at the two-tailed 0.95 (dashed lines) and 0.99 (dotted lines) levels.

Fig. 4.

Cloud–solar cross correlations for detrended data. Cross-correlation plots for ±24-month time lags for both TSI (solid line) and GCR (dashed line) against deseasonalized monthly cloud anomalies at varying altitude levels: (a) total, (b) high-level (<440 mb), (c) midlevel (440–680 mb), and (d) low-level (>680 mb). The linear trends (presented in the Y equations of Figs. 1, 2) are removed from the datasets prior to correlation analysis. Second-order polynomial fits of MC-simulated significance thresholds are also displayed over the ±24-month lag period at the two-tailed 0.95 (dashed lines) and 0.99 (dotted lines) levels.

An examination of locally significant correlations (occurring at zero lag) shows the presence of small and sporadic regions of largely positive correlation occurring between total cloud and TSI over areas of the southern midlatitude ocean regions and also a single relatively large area located at low-cloud levels over the equatorial Pacific, coinciding with the region associated with the Pacific cold tongue (PCT) region (Figs. 5a–d). Although these areas of positive correlation are found to occur over high-, mid-, and low-altitude levels, it can also be observed that negative correlations appear cospatially (at different altitudes) with many areas of overlying positive anomalies (while the reverse is also true), for example, over areas of the North Atlantic. Locally significant correlations between GCR variations and cloud show a range of geographically scattered statistically significant correlations (Figs. 5e–h). However, the correlations at low-altitude levels tend to be largely negative, contradicting a suggested positive correlation between GCR and low-cloud variations (Fig. 5h). Field significance testing was performed on these results: only the low-cloud/GCR flux sample was found to be field significant above the 0.95 significance level (having a p value of 0.043).

Fig. 5.

Locally significant solar–cloud correlations. Locally statistically significant (at the 0.95 two-tailed level) R (at zero lag) plotted between 60°N and 60°S. (a)–(d) [(e)–(h)] display correlations between cloud anomalies and TSI [GCR], at all total, high-altitude (<440 mb), midaltitude (440–680 mb), and low-altitude (>680 mb) levels. Critical significance levels are determined for each pixel by MC simulation methods (see methodology for full description). Linear trends are removed from the datasets prior to analysis.

Fig. 5.

Locally significant solar–cloud correlations. Locally statistically significant (at the 0.95 two-tailed level) R (at zero lag) plotted between 60°N and 60°S. (a)–(d) [(e)–(h)] display correlations between cloud anomalies and TSI [GCR], at all total, high-altitude (<440 mb), midaltitude (440–680 mb), and low-altitude (>680 mb) levels. Critical significance levels are determined for each pixel by MC simulation methods (see methodology for full description). Linear trends are removed from the datasets prior to analysis.

The MEI shows a decreasing linear trend over the data period (of −0.004 month−1) (Fig. 6a). In relation to the solar and cloud datasets, the MEI shows (zero lag) R values of −0.16/−0.08 to TSI/GCR, respectively, and −0.06/−0.19/−0.09/0.28 to total/high/middle/low deseasonalized cloud anomalies, (the linear trends are removed from all data prior to tests of cross correlation). At local (geographical) scales, R values are found to be both large (up to ±0.8) and geographically widespread. The most apparent features are a large area of positive (negative) correlation at high (low) levels over the middle equatorial Pacific and a region of negative (positive) correlation located largely around Indonesia at high (low)-altitude levels (Figs. 6b–e).

Fig. 6.

ENSO index and cloud correlations. (a) Monthly changes in the MEI from 2000 to 2010, with the linear trend, Y, and R value displayed. Also shown are locally statistically significant correlations to variations in cloud cover at (b) total, (c) high-altitude (<440 mb), (d) midaltitude (440–680 mb), and (e) low-altitude (>680 mb) levels. Critical significance levels are determined for each pixel by MC simulation methods (see methodology for full description). Linear trends are removed from the datasets prior to analysis.

Fig. 6.

ENSO index and cloud correlations. (a) Monthly changes in the MEI from 2000 to 2010, with the linear trend, Y, and R value displayed. Also shown are locally statistically significant correlations to variations in cloud cover at (b) total, (c) high-altitude (<440 mb), (d) midaltitude (440–680 mb), and (e) low-altitude (>680 mb) levels. Critical significance levels are determined for each pixel by MC simulation methods (see methodology for full description). Linear trends are removed from the datasets prior to analysis.

The MEI shows a statistically significant (>0.99 two-tailed level) negative correlation to the GCR flux centered on month −8 and a secondary significant negative peak (>0.95 confidence level) occurring around month +16, whereas TSI shows no notable correlation to the MEI within approximately ±20 months surrounding the zero lag (Fig. 7).

Fig. 7.

MEI cross correlations. Cross correlations over a ±24-month lag period between the MEI and TSI emissions (solid line) and GCR flux anomalies (dashed line). Linear trends are removed from all data prior to testing. Second-order polynomial fits of MC-simulated critical significance thresholds are also displayed over the ±24-month lag period at the two-tailed 0.95 (dashed lines) and 0.99 (dotted lines) levels.

Fig. 7.

MEI cross correlations. Cross correlations over a ±24-month lag period between the MEI and TSI emissions (solid line) and GCR flux anomalies (dashed line). Linear trends are removed from all data prior to testing. Second-order polynomial fits of MC-simulated critical significance thresholds are also displayed over the ±24-month lag period at the two-tailed 0.95 (dashed lines) and 0.99 (dotted lines) levels.

5. Discussion

An analysis of Fig. 1 has shown that total MODIS cloud measurements exhibit a long-term increase of 0.58% over the past decade, resulting from a decrease in high- and midlevel cloud cover of −0.31% and an increasing trend in low-cloud levels of 0.89%. If real, this change implies an increase in shortwave (negative) and a decrease in longwave (positive) radiative forcing, which would suggest that over the past decade clouds may have exerted an enhanced cooling influence on climate of around −0.075 W m−2. However, it is likely that this observation is partially artificial, as satellite radiometer measurements are noncloud penetrating. Consequently, low-level cloud is frequently masked by the occurrence of overlying cloud. Thus, a physical decline in the high-/midlevel cloud amount over regions that overlap low-level cloud will give the false impression of a roughly equal magnitude increase in low-level cloud amount; such occurrences have been noted in the ISCCP dataset (Pallé 2005). This phenomenon is evident in Fig. 6, which shows cospatial significant correlations of opposing sign between altitude levels (e.g., the Indonesian region shows both negative correlations at high-altitude levels and positive correlations at low-altitude levels). Thus, it is likely that the increase in low-level cloud (which is more than twice the magnitude of the reduction in high- and midlevel clouds) is at least partially artificial. It is physically plausible that the linear reduction detected in the high-level cloud may result from the shift toward La Niña conditions over the data period [indicated by the linear decrease of the MEI (Fig. 6a)].

It is also nontrivial to note that trends detected in the ISCCP dataset over a comparable time period are found to be in disagreement with MODIS: from 2000 to 2008 ISCCP D1 IR data shows trends of −0.018% (30 days)−1 in low-level clouds and 0.027% (30 days)−1 in midlevel and high-level clouds combined. Although we make no attempt to test the physical validity or attribution of these trends, we note that this conflict underscores the difficulty in accurately measuring global long-term cloud changes from satellites and suggests that we should limit the level of confidence that can be placed in the results of such long-term cloud studies. For a detailed discussion of relevant ISCCP and MODIS comparisons, see Pincus et al. (2006, 2012) and references therein.

Although this is the first long-term examination of MODIS-detected cloud anomalies in relation to solar activity, MODIS has been previously utilized to test for the presence of a solar – cloud link by Kristjánsson et al. (2008). This study focused on high-magnitude daily time-scale decreases in the GCR flux known as Forbush decrease (FD) events to test for the occurrence of significant cloud anomalies over regions theoretically sensitive to the effects of GCR variations; the results of this work also yielded no compelling evidence of a GCR–cloud link.

The period of study (2000–10) occurs during the final phase of solar cycle 23 and the first portion of cycle 24. The latter has been an abnormal cycle, with an unusually deep minimum (Nandy et al. 2011). This is clearly evident in Fig. 2, which shows both TSI and the GCR flux exhibiting low monthly variability post 2006. Consequently, this may mean that the magnitude (and thus detectability) of any potential solar–cloud link may also be comparably low over this period. To be fully confident that a solar signal is not clearly present in the MODIS data, this study should be repeated in the future when data spanning several solar cycles becomes available.

It is plausible that artificial correlations may arise between TSI and cloud cover, as a result of varying irradiance being received by the satellite instruments. However, studies observing changes in the TSI flux of ~0.4 W m−2 over a period of several days observed no statistically significant changes in ISCCP cloud cover (Laken et al. 2011; Laken and Čalogović 2011), implying that such a situation is likely similarly incapable of producing artificial cloud changes over annual–decadal time scales also.

Because of the linked nature of the variations in the TSI and GCR flux, it is problematic to isolate one variable and claim it as causally linked to atmospheric changes. However, based on the overall significance of the correlations achieved, TSI has a higher probability of being causally related to cloud changes than the GCR flux: similar conclusions have been drawn by other long-term correlation analysis studies based on the ISCCP dataset (Kristjánsson et al. 2002, 2004). However, it is clear from the overall poor correlations achieved (Figs. 4, 5) that solar activity is not a dominant factor driving monthly and annual cloud variability. Indeed, it may be argued that several of the locally significant correlations identified in Fig. 5 may be reflecting ENSO-related variability (to which TSI/GCR are found to demonstrate a weak correspondence); arguments that long-term solar–ISCCP cloud studies are influenced by ENSO activity have similarly been made by Farrar (2000). The results of a cross-correlation analysis between the MEI and solar anomalies/globally averaged cloud variations seem to further reinforce such findings: analysis has shown that GCR anomalies yield a significant anticorrelation to the MEI around month −8; this correlation does not diminish below statistically significant values until approximately month −3 (Fig. 7). Such a correlation clearly has no physical basis, and it implies that the significant correlation detected between GCR variations and midlevel global cloud anomalies beginning around month −11 (Fig. 4c) may be due to the chance correlation between the GCR flux and ENSO variations. Similarly, the increases in R values around month +12 between TSI and total/high-altitude/midaltitude cloud anomalies (Figs. 4a–c) are also likely connected to positive correlations between the aforementioned parameters and ENSO variations at this time (Fig. 7).

Despite these largely discouraging findings with relation to the notion of a dominant solar–cloud link, our local-scale correlation analysis identifies variations in low-altitude cloud anomalies over the PCT region as being potentially linked to changes in TSI. Although this correlation is detected at low-altitude levels, there are no detected areas of cospatial cloud correlation at upper-atmospheric levels, indicating that the correlation is likely not an artifact. We, therefore, suggest that this correlation may represent a plausible region of TSI-associated cloud variability linked to ENSO, likely operating via a link between TSI and local sea surface temperatures; a mechanism for such a link may operate in a method similar to that proposed by Meehl et al. (2009).

6. Conclusions

An analysis of the first decade of monthly time-scale MODIS cloud anomalies has shown that neither variations in TSI emissions or the GCR flux are dominantly responsible for cloud variability at global or local (geographic) scales at any altitude level. Although correlation analysis suggests that some statistically significant correlations between cloud variability and TSI/GCR variations are present, further investigation of these relationships revealed that such associations either broke down during the data period or were likely connected to internal climate variability and not to solar activity.

Acknowledgments

The authors kindly thank Jasa Čalogović (Hvar Observatory), Dr. Dominic Kniveton (University of Sussex), Dr. Juan Betancort (Instituto de Astrofísica de Canarias), Dr. Steven Platnick (NASA), Dr. Robert Pincus (NOAA), Dr. Robert Wood (University of Washington), and two anonymous referees for their constructive comments. The MODIS data were obtained from the NASA website (http://ladsweb.nascom.nasa.gov), while the cosmic ray data were obtained from the Solar Terrestrial’s physics division of IZMIRAN (http://helios.izmiran.rssi.ru). The authors acknowledge the PMOD dataset (version d41_62_1102): PMOD/WRC of Davos, Switzerland, which also comprises unpublished data from the VIRGO experiment on the ESA/NASA mission SoHO. MEI NOAA/OAR/ESRL/PSD data were obtained from online (at www.esrl.noaa.gov/psd//people/klaus.wolter/MEI/).

REFERENCES

REFERENCES
Beer
,
J.
,
W.
Mende
, and
R.
Stellmacher
,
2000
:
The role of the sun in climate forcing
.
Quat. Sci. Rev.
,
19
,
403
415
.
Bond
,
G.
, and
Coauthors
,
2001
:
Persistent solar influence on North Atlantic climate during the Holocene
.
Science
,
294
,
2130
2136
.
Brest
,
C.
,
W.
Rossow
, and
M.
Roiter
,
1997
:
Update of radiance calibrations for ISCCP
.
J. Atmos. Oceanic Technol.
,
14
,
1091
1109
.
Čalogović
,
J.
,
C.
Albert
,
F.
Arnold
,
J.
Beer
,
L.
Desorgher
, and
E.
Flueckiger
,
2010
:
Sudden cosmic ray decreases: No change of global cloud cover
.
Geophys. Res. Lett.
,
37
,
L03802
,
doi:10.1029/2009GL041327
.
Carslaw
,
K. S.
,
R. G.
Harrison
, and
J.
Kirkby
,
2002
:
Cosmic rays, clouds, and climate
.
Science
,
298
,
1732
1737
,
doi:10.1126/science.1076964
.
Evan
,
A. T.
,
A. K.
Heidinger
, and
D. J.
Vimont
,
2007
:
Arguments against a physical long-term trend in global ISCCP cloud amounts
.
Geophys. Res. Lett.
,
34
,
L04701
,
doi:10.1029/2006GL028083
.
Farrar
,
P.
,
2000
:
Are cosmic rays influencing oceanic cloud coverage – Or is it only El Niño?
Climatic Change
,
47
,
7
15
.
Fleitmann
,
D.
,
S. J.
Burns
,
M.
Mudelsee
,
U.
Neff
,
J.
Kramers
,
A.
Mangini
, and
A.
Matter
,
2003
:
Holocene forcing of the Indian monsoon recorded in a stalagmite from southern Oman
.
Science
,
300
,
1737
1739
,
doi:10.1126/science.1083130
.
Fröhlich
,
C.
,
2000
:
Observations of irradiance variations
.
Space Sci. Rev.
,
94
,
15
24
.
Fröhlich
,
C.
, and
J.
Lean
,
1998
:
The sun’s total irradiance: Cycles, trends and related climate change uncertainties since 1976
.
Geophys. Res. Lett.
,
25
,
4377
4380
.
Haigh
,
J.
,
1996
:
The impact of solar variability on climate
.
Science
,
272
,
981
984
.
Kelly
,
P.
, and
T.
Wigley
,
1992
:
Solar cycle length, greenhouse forcing and global climate
.
Nature
,
360
,
328
330
.
Kristjánsson
,
J. E.
,
A.
Staple
,
J.
Kristiansen
, and
E.
Kaas
,
2002
:
A new look at possible connections between solar activity, clouds and climate
.
Geophys. Res. Lett.
,
29
,
2107
,
doi:10.1029/2002GL015646
.
Kristjánsson
,
J. E.
,
J.
Kristiansen
, and
E.
Kaas
,
2004
:
Solar activity, cosmic rays, clouds and climate - An update
.
Adv. Space Res.
,
34
,
407
415
,
doi:10.1016/j.asr.2003.02.040
.
Kristjánsson
,
J. E.
,
C. W.
Stjern
,
F.
Stordal
,
A. M.
Færaa
,
G.
Myhre
, and
K.
Jónasson
,
2008
:
Cosmic rays, cloud condensation nuclei and clouds – A reassessment using MODIS data
.
Atmos. Chem. Phys.
,
8
,
7373
7387
,
doi:10.5194/acp-8-7373-2008
.
Laken
,
B. A.
, and
J.
Čalogović
,
2011
:
Solar irradiance, cosmic rays and cloudiness over daily timescales
.
Geophys. Res. Lett.
,
38
,
L24811
,
doi:10.1029/2011GL049764
.
Laken
,
B. A.
,
D. R.
Kniveton
, and
M. R.
Frogley
,
2010
:
Cosmic rays linked to rapid mid-latitude cloud changes
.
Atmos. Chem. Phys.
,
10
,
10 941
10 948
,
doi:10.5194/acp-10-10941-2010
.
Laken
,
B. A.
,
D. R.
Kniveton
, and
A.
Wolfendale
,
2011
:
Forbush decreases, solar irradiance variations, and anomalous cloud changes
.
J. Geophys. Res.
,
116
,
D09201
,
doi:10.1029/2010JD014900
.
Laut
,
P.
,
2003
:
Solar activity and terrestrial climate: An analysis of some purported correlations
.
J. Atmos. Sol. Terr. Phys.
,
65
,
801
812
,
doi:10.1016/S1364-6826(03)00041-5
.
Lean
,
J.
, and
D.
Rind
,
1998
:
Climate forcing by changing solar radiation
.
J. Climate
,
11
,
3069
3094
.
Lean
,
J.
,
J.
Beer
, and
R.
Bradley
,
1995
:
Reconstruction of solar irradiance since 1610: Implications for climate change
.
Geophys. Res. Lett.
,
22
,
3195
3198
.
Lockwood
,
M.
, and
C.
Fröhlich
,
2008
:
Recent oppositely directed trends in solar climate forcings and the global mean surface air temperature. II. Different reconstructions of the total solar irradiance variation and dependence on response time scale
.
Proc. Roy. Soc. London
,
464A
,
1367
1385
,
doi:10.1098/rspa.2007.0347
.
Marsh
,
N. D.
, and
H.
Svensmark
,
2000
:
Low cloud properties influenced by cosmic rays
.
Phys. Rev. Lett.
,
85
,
5004
5007
.
Marsh
,
N. D.
, and
H.
Svensmark
,
2003
:
Galactic cosmic ray and El Niño–Southern Oscillation trends in International Satellite Cloud Climatology Project D2 low-cloud properties
.
J. Geophys. Res.
,
108
,
4195
,
doi:10.1029/2001JD001264
.
Meehl
,
G. A.
,
J. J.
Arblaster
,
K.
Matthes
,
F.
Sassi
, and
H.
van Loon
,
2009
:
Amplifying the Pacific climate system response to a small 11-year solar cycle forcing
.
Science
,
325
,
1114
1118
,
doi:10.1126/science.1172872
.
Nandy
,
D.
,
A.
Muñoz-Jaramillo
, and
P. C. H.
Martens
,
2011
:
The unusual minimum of sunspot cycle 23 caused by meridional plasma flow variations
.
Nature
,
471
,
80
82
, doi:10.1038/nature09786
.
Norris
,
J.
,
2000
:
What can cloud observations tell us about climate variability?
Space Sci. Rev.
,
94
,
375
380
.
Pallé
,
E.
,
2005
:
Possible satellite perspective effects on the reported correlations between solar activity and clouds
.
Geophys. Res. Lett.
,
32
,
L03802
,
doi:10.1029/2004GL021167
.
Pallé
,
E.
, and
C. J.
Butler
,
2000
:
The influence of cosmic rays on terrestrial cloud and global warming
.
Astron. Geophys.
,
41
,
4.18
4.22
.
Pallé
,
E.
, and
C. J.
Butler
,
2002
:
The proposed connection between clouds and cosmic rays: Cloud behaviour during the past 50–120 years
.
J. Atmos. Sol. Terr. Phys.
,
64
,
327
337
,
doi:10.1016/S1364-6826(01)00105-5
.
Pincus
,
R.
,
C. P.
Batstone
, and
S. E.
Platnick
,
2006
:
Systematic comparisons of clouds as observed by MODIS and ISCCP
.
Preprints, 12th Conf. on Atmospheric Radiation, Madison, Wisconsin, Amer. Meteor. Soc., 2.3. [Available online at http://ams.confex.com/ams/Madison2006/techprogram/paper_112596.htm.]
Pincus
,
R.
,
S. E.
Platnick
,
S. A.
Ackerman
,
R. S.
Hemler
, and
R. J. P.
Hofmann
,
2012
:
Reconciling simulated and observed views of clouds: MODIS, ISCCP, and the limits of instrument simulators
.
J. Climate
,
25
,
4699
4720
.
Ram
,
M
., and
M. R.
Stolz
,
1999
:
Possible solar influences on the dust profile of the GISP2 ice core from central Greenland
.
Geophys. Res. Lett.
,
26
,
1043
1046
.
Ramanathan
,
V.
,
R.
Cess
,
E.
Harrison
,
P.
Minnis
,
B.
Barkstrom
,
E.
Amhad
, and
D.
Hartmann
,
1989
:
Cloud-radiative forcing and climate: Results from the Earth Radiation Budget Experiment
.
Science
,
243
,
57
63
.
Rao
,
U.
,
2011
:
Contribution of changing galactic cosmic ray flux to global warming
.
Curr. Sci. India.
,
100
,
223
225
.
Rossow
,
W.
, and
R.
Schiffer
,
1999
:
Advances in understanding clouds from ISCCP
.
Bull. Amer. Meteor. Soc.
,
80
,
2261
2288
.
Rossow
,
W.
,
A.
Walker
, and
L.
Garder
,
1993
:
Comparison of ISCCP and other cloud amounts
.
J. Climate
,
6
,
2394
2418
.
Schlesinger
,
M. E.
, and
N.
Ramankutty
,
1992
:
Implications for global warming of intercycle solar irradiance variations
.
Nature
,
360
,
330
333
.
Shaviv
,
N. J.
,
2005
:
On climate response to changes in the cosmic ray flux and radiative budget
.
J. Geophys. Res.
,
110
,
A08105
,
doi:10.1029/2004JA010866
.
Slingo
,
A.
,
1990
:
Sensitivity of the earth’s radiation budget to changes in low clouds
.
Nature
,
343
,
49
51
.
Solanki
,
S. K.
,
I. G.
Usoskin
,
B.
Kromer
,
M.
Schüssler
, and
J.
Beer
,
2004
:
Unusual activity of the sun during recent decades compared to the previous 11,000 years
.
Nature
,
431
,
1084
1087
,
doi:10.1038/nature02995
.
Solomon
,
S.
,
D.
Qin
,
M.
Manning
,
M.
Marquis
,
K.
Averyt
,
M. M. B.
Tignor
,
H. L.
Miller
Jr.
, and
Z.
Chen
, Eds.,
2007
:
Climate Change 2007: The Physical Science Basis. Cambridge University Press, 996 pp
.
Svensmark
,
H.
,
2007
:
Cosmoclimatology: A new theory emerges
.
Astron. Geophys.
,
48
,
1.18
1.24
,
doi:10.1111/j.1468-4004.2007.48118.x
.
Svensmark
,
H.
, and
E.
Friis-Christensen
,
1997
:
Variation of cosmic ray flux and global cloud coverage—A missing link in solar-climate relationships
.
J. Sol. Atmos. Terr. Phys.
,
59
,
1225
1232
.
Todd
,
M.
, and
D.
Kniveton
,
2001
:
Changes in cloud cover associated with Forbush decreases of galactic cosmic rays
.
J. Geophys. Res.
,
106
(
D23
),
32 031
32 041
.
Wolter
,
K.
, and
M. S.
Timlin
,
1993
:
Monitoring ENSO in COADS with a seasonally adjusted principal component index
.
Proc. 17th Climate Diagnostics Workshop, Norman, OK, NOAA/NMC/CAC, NSSL, Oklahoma Climate Survey, CIMMS, and the School of Meteorology, University of Oklahoma, 52–57
.
Wylie
,
D.
, and
W.
Menzel
,
1999
:
Eight years of high cloud statistics using HIRS
.
J. Climate
,
12
,
170
184
.
Yamaguchi
,
Y. T.
,
Y.
Yokoyama
,
H.
Miyahara
,
K.
Sho
, and
T.
Nakatsuka
,
2010
:
Synchronized Northern Hemisphere climate change and solar magnetic cycles during the Maunder Minimum
.
Proc. Natl. Acad. Sci. USA
,
107
,
20 697
20 702
,
doi:10.1073/pnas.1000113107
.