Cosmic rays, aerosols, clouds, and climate: Recent findings from the CLOUD experiment

1 Introduction

The flux of galactic cosmic rays (GCR) to the Earth's troposphere is modulated by solar activity [Usoskin et al., 2002]. About 20 years ago, a series of studies showed correlations between GCR and low-cloud cover on decadal timescales with several percent variability in cloud cover between solar maximum and minimum time periods of the 11 year solar sunspot cycle [Svensmark and Friis Christensen, 1997; Marsh and Svensmark, 2000]. As clouds modulate the Earth's radiative balance, a GCR-cloud connection could have an important impact on the Earth's radiative balance and climate. More recently, studies have also shown correlations of GCR with cloud cover and atmospheric aerosols on weekly timescales [e.g., Svensmark et al., 2009]. However, these GCR-cloud correlations are controversial as other studies have shown a lack of correlation between GCR and cloud cover at the decadal and weekly timescales with discussions as to how methodological differences lead to differences in results [e.g., Laken et al., 2012, and references therein]. Additionally, there has not been a significant trend in GCR over the past 50 years, so recent warming cannot be attributed to GCR [Usoskin et al., 2002]. Regardless of the uncertainty in the correlations between cosmic rays and clouds and the lack of possible influence on recent climate, the potential of a GCR-cloud-climate connection has led to research into physical mechanisms where GCR might affect clouds and historical climate change.

The GCR-cloud-climate proposed mechanism that has received the most attention is the ion-aerosol clear-sky hypothesis [Carslaw et al., 2002], shown in Figure 1 and described as follows. An increase in GCR to the atmosphere (which happens when moving between the solar maximum and solar minimum of the solar cycle) leads to an increase in small ions (charged molecules or charged small clusters of molecules) in the troposphere. An increase in small ions may increase the nucleation rate (the formation rate) of ~1 nm diameter aerosol particles. If these new 1 nm particles grow to diameters larger than ~50–100 nm through condensation of vapors (generally sulfuric acid and low-volatility organics), these particles may act as cloud condensation nuclei (CCN). However, the growth to CCN sizes must occur before the new particles are lost by coagulation with existing CCN particles [Pierce and Adams, 2007]. An increase in CCN would lead to an increase in cloud droplet number, which may lead to an increase in cloud albedo and cloud cover, and for low clouds this albedo/cover increase has a cooling effect on climate.

There are several factors that may dampen the fractional change in cloud properties relative to the fractional change in GCR: (1) While nucleation of 1 nm particles is aided by the presence of ions, nucleation still proceeds without ions at a slower rate. Additionally, while most ions in the troposphere are formed from GCR, ions are also formed from terrestrial radioactive material. (2) The growth of 1 nm particles to CCN is limited by the amount of condensible material, so an increase in nucleation rate leads to a reduction of the 1 nm particles that survive to CCN sizes without being lost by coagulation (i.e., with increasing nucleation, growth becomes slower and coagulation faster, so pathway 3a in Figure 1 becomes more favorable than pathway 3b). (3) Cloud droplet concentrations, cloud albedo, and cloud extent generally respond sublinearly to changes in CCN concentrations. The strength of each of these three dampening factors globally and regionally determines how responsive clouds and the Earth's radiative balance are to GCR through the ion-aerosol clear-sky hypothesis. Previous estimates of the GCR-CCN-cloud system have shown the GCR influence on CCN and clouds to be weak [Pierce and Adams, 2009; Snow-Kropla et al., 2011; Yu et al., 2012; Dunne et al., 2012; Kazil et al., 2012; Yu and Luo, 2014], yet uncertainties have remained about the nucleation and growth assumptions in these studies.

The Cosmics Leaving OUtdoor Droplets (CLOUD) study was conceived by Jasper Kirkby at CERN in the late 1990s [Kirkby, 2001] to explore the potential mechanisms between GCR and clouds. CLOUD uses the CERN pion bean to simulate GCR in a state-of-the-art 26 m³ chamber as free of chemical contaminants as currently possible [Kirkby et al., 2011]. The CLOUD study has entrained many of the world's top experimental and theoretical researchers of aerosol nucleation and growth. The CLOUD experiments began in the late 2000s, and since 2011, the CLOUD team has made significant contributions to our understanding of nucleation and the early stages of particle growth to CCN-sized particles. Until recently, the CLOUD publications have focused on the details of the nucleation/growth processes and have used these findings in global aerosol models to compare how nucleation/growth may have changed due to human influence (described in the next section). While the work of the CLOUD team has made a huge impact on aerosol nucleation, growth, and their overall connection to clouds and climate, they had not yet addressed a GCR-cloud connection. In their recent article in JGR [Gordon et al., 2017], the CLOUD team present their first estimates of the effect of GCR modulations over the solar cycle on CCN (a major step toward the GCR climate impacts through the ion-aerosol clear-sky effect). In the following two sections I highlight (1) the CLOUD team's impact on our understanding of nucleation and growth as well as related results done outside of CLOUD and (2) the CLOUD team's recent GCR results.

2 Advances in Understanding of Nucleation and Growth

A majority of the particles in the atmosphere are formed by nucleation from the clustering of low-volatility vapors, as opposed to primary emissions where particles are emitted directly to the atmosphere (from combustion sources, wind-blown dust, or sea spray) [Gordon et al., 2017]. However, until recently there has been considerable uncertainty involving which vapor species participate in aerosol nucleation. Additionally, theoretical calculations of nucleation rates were often incorrect by orders of magnitude relative to measured rates [Jung et al., 2008; Yu et al., 2010]. Many of these uncertainties stem from the challenge of performing nucleation experiments in the laboratory (small amounts of contaminant species may dramatically change results) and precisely measuring the composition of the vapors and the newly formed particles. The CLOUD collaboration has largely remedied these limitations by taking great efforts to make the CLOUD chamber as free of contaminants as possible, while leveraging novel vapor and nanoparticle measurement instrumentation. In doing this, they (along with several studies elsewhere) unlocked many secrets about aerosol nucleation and growth during the past decade.

The vapors of sulfuric acid, water, and ammonia have been well-established ingredients of new-particle formation [Korhonen et al., 1999], but theoretical predictions of nucleation rates involving these species often did not reproduce measurements [e.g., Yu et al., 2010]. CLOUD studies [Kirkby et al., 2011; Kürten et al., 2016] have measured this system across relevant atmospheric temperatures, and Dunne et al. [2016] have fit these CLOUD measurements to a parameterization for atmospheric models. Almeida et al. [2013] found that amines may stabilize sulfuric acid clusters more effectively than ammonia. Recent work done outside of CLOUD group has also demonstrated the importance of diamines [Jen et al., 2016] in addition to amines [Jen et al., 2014; Glasoe et al., 2015] in nucleation and early particle growth. The CLOUD studies found that highly oxidized organic vapors may enhance sulfuric acid nucleation [Riccobono et al., 2014] and may nucleate particles on their own [Kirkby et al., 2016], which has large implications on where nucleation may occur in historical atmospheres. In parallel, studies outside of the CLOUD group have also found the importance of organics in nucleation [Zhao et al., 2013].

The 1 nm diameter new particles must grow to diameters of at least 50–100 nm in order to act as CCN and impact climate (Figure 1). This growth primarily occurs through condensation of vapors. Fast growth leads to a higher fraction of the new particles surviving growth to climate-relevant sizes rather than coagulating with preexisting particles; thus, it is necessary to understand how different vapors interact with differently sized particles. The CLOUD team has determined how effectively organic vapors condense to particles at different particle diameters and how these details impact predicted CCN concentrations globally [Tröstl et al., 2016]. These CLOUD results support and provide details on recent observations of particle-size-dependent organic condensation made in the field [Kuang et al., 2012; Häkkinen et al., 2013].

The CLOUD team has applied their experimental results to global aerosol models to evaluate how nucleation varies throughout the atmosphere, the connections between nucleation and CCN, the importance of organics in growth to CCN, and the differences in nucleation and CCN between the preindustrial and present-day atmospheres [Dunne et al., 2016; Gordon et al., 2016, 2017]. Their findings have shown that the anthropogenic aerosol forcing between preindustrial and present-day times depends on the nucleation schemes used. For example, including their organic-only nucleation scheme [Kirkby et al., 2016] creates more CCN in the preindustrial time period because these nucleating organics may be formed from volatile organic compounds emitted from trees. Having more CCN in the preindustrial time reduces the impact of anthropogenic emissions, lowering the anthropogenic aerosol forcing [Gordon et al., 2016]. Hence, the recent advances in our understanding of aerosol nucleation and growth from the CLOUD team and other groups have allowed for refined estimates of the impact of anthropogenic aerosols on climate.

3 Cosmic Rays, Nucleation, and Cloud Condensation Nuclei

In Gordon et al. [2017], the CLOUD team estimates the effect GCR on CCN, a major step in evaluating the impact of GCR on clouds and climate via the ion-aerosol clear-sky hypothesis. This study uses a global aerosol-chemistry model to estimate aerosol processes and concentrations throughout the atmosphere for both present-day and preindustrial atmospheres. Their model is informed by the major nucleation and growth findings from the CLOUD laboratory experiments described above. The use of a global aerosol model is necessary to bridge the length and timescales of the CLOUD chamber experiments to the full atmosphere.

Figure 2 shows a simplified view of the response of nucleation and CCN to GCR roughly corresponding to the results of Gordon et al. [2017]. Gordon et al. [2017] estimate that ions (formed primarily from GCR) are responsible for around half of the nucleation in both preindustrial and present-day atmospheres. This is consistent with many of the CLOUD experiments that showed that ions increase nucleation rates under many conditions. However, while GCR affecting nucleation is a necessary condition of the ion-aerosol clear-sky hypothesis, it is not sufficient on its own: The changes in nucleation over typical cycles in GCR must lead to significant changes in CCN (Figure 1) and in turn significant changes in cloud properties.

Regarding the connection of GCR to CCN, Gordon et al. [2017] found that setting ion concentrations to 0 (i.e., no GCR or terrestrial radioactive sources of ions) reduced CCN by 10–20% at the altitude of low clouds, still significant but a much smaller change than the relative change in nucleation rates when ions were turned off. This damped response of CCN is due to more nucleated particle surviving growth to CCN sizes when ions are off as well as the influence of CCN from primary emissions. When Gordon et al. [2017] modulated GCR intensity by the magnitude of the solar maximum versus solar minimum of the solar cycle (changes in ions on the order of 10–20% over much of the troposphere), they found changes in CCN of 0.2–0.3% at the altitude of low clouds. This is much too small of a CCN change to explain the observed 2% change in cloud cover with the solar cycle [Marsh and Svensmark, 2000] as there would be further damping of changes in clouds to changes in CCN. Thus, these results from CLOUD show that the ion-aerosol clear-sky hypothesis is too weak to significantly impact clouds and climate.

Interestingly, the weak GCR-CCN response in Gordon et al. [2017] is similar to several earlier estimates that all showed CCN changes of 1% or less to GCR cycles [Pierce and Adams, 2009; Snow-Kropla et al., 2011; Yu et al., 2012; Dunne et al., 2012; Kazil et al., 2012; Yu and Luo, 2014]. These earlier studies used nucleation and growth schemes that did not have the benefit of being informed by the CLOUD experiments, and some of the nucleation schemes used were quite inferior to those developed by CLOUD [e.g., Pierce and Adams, 2009]. Regardless, the response in all studies has been remarkably similar. Nucleation is damped relative to changes in GCR, and CCN is damped relative to changes in nucleation; thus, the response of CCN to GCR is strongly damped (Figure 2). Thus, the exact details of nucleation rates and mechanisms were not needed to find the weak GCR-CCN connection in earlier studies.

Does this CLOUD finding close the door on any possible GCR-cloud-climate connection? No, other work has suggested that cosmic rays could influence the freezing of liquid water to ice in clouds [Carslaw et al., 2002] or may influence the amount of condensable material [Svensmark et al., 2013], and other possible mechanisms might exist. However, no one has shown mechanistically that the ion-aerosol clear-sky mechanism is strong enough to impact clouds, and now the CLOUD team has also found this to be the case. Nonetheless, we can thank the ion-aerosol clear-sky hypothesis for creating the CLOUD experiment that has unlocked a treasure trove of information on nucleation, growth, and aerosols in general.