Impact of Biogenic versus Anthropogenic emissions on Clouds and Climate: towards a Holistic UnderStanding

Final Report Summary - BACCHUS (Impact of Biogenic versus Anthropogenic emissions on Clouds and Climate: towards a Holistic UnderStanding)

Executive Summary:
Aerosol particles affect climate through scattering and absorption of radiation and by affecting cloud properties by acting as cloud condensation nuclei (CCNs) and ice nucleating particles (INPs). Aerosol-cloud interactions play a key role in the anthropogenic radiative forcing of the climate system but remain the most uncertain of all forcing agents and are still associated with a low scientific level of understanding. A major part of the uncertainty in how aerosol and cloud processes respond to changes in anthropogenic and natural emissions is due i) to lack of fundamental understanding about ice-containing clouds and ii) to the incomplete knowledge of the coupling between biosphere and atmosphere. These two areas were explored in greater details in the BACCHUS project.
Emphasis was placed on collecting INPs from previous and new measurement campaigns in different environments. The results obtained during BACCHUS confirm earlier results that biological particles are the best INPs, initiating ice nucleation already at temperatures larger than -10 °C. Dust particles are rather ubiquitous (together with haze and smoke) and control heterogeneous ice formation in the temperature range for about -18 to -35 °C. In their absence, INP at a remote North Atlantic coastal site were largely comprised of organic carbon and/or heat labile material from biological activity but sea spray aerosols also contributed. No evidence was found for INP from anthropogenic activity to be an important source for INP. In summary, anthropogenic activities mainly affect CCN concentrations while they only indirectly affect INP concentrations.
The BACCHUS ESMs simulate relative model diversity in the direct aerosol radiative effects, often considered to be well understood, that exceeds the diversity in the indirect radiative effects. Nonetheless, the absolute diversity is larger for the indirect radiative effects. The model diversity can likely be attributed to the strong sensitivity of direct forcing to aerosol absorption, which modulates the sign of the top-of- atmosphere forcing. Resulting available global-mean all-sky direct radiative forcings range from -0.37 Wm-2 (HadGEM-UKCA) to +0.10 Wm-2 (ECHAM-HAM-CCFM). The BACCHUS ESMs simulate total aerosol effective radiative forcing (ERF) forcing ranging from - 0.96 Wm-2 (ECHAM-HAM) to -1.59 Wm-2 (HadGEM-UKCA). It is likely that these relatively strong ERFs are driven by strong increases in cloud liquid water. Fundamental research on related cloud and aerosol processes remains a key priority for reducing the uncertainty in the total anthropogenic perturbation of the climate system.
In terms of feedbacks BACCHUS investigated how changes in climate affect natural aerosol emissions and their precursors. More biogenic volatile organic carbon will be emitted in a warmer climate causing the formation of more aerosol particles. This results in more but smaller cloud droplets and more reflection of sunlight back to space. This cooling partly offsets the greenhouse gas warming. This feedback loop cannot be regarded as isolated from anthropogenic emissions because the biosphere is changed by human activities. Therefore, a proper assessment of anthro- pogenic interference requires studying anthropogenic emissions in concert with biogenic emissions.

Project Context and Objectives:
Objectives
The core idea of BACCHUS has been to quantify key processes controlling clouds and climate and their feedbacks by i) contrasting the processes occurring in climate-relevant environments such as tropical areas and the Arctic and ii) by combining advanced measurements of cloud and aerosol properties with state-of-the-art numerical modeling. Specifically, BACCHUS has aimed to characterize the importance of biogenic versus anthropogenic emissions for cloud formation and climate in regions that are key regulators of climate (tropical rain forests) as well as in regions experiencing the most profound climatic changes, and which may be prone to irreversible transitions, e.g. the Arctic. In addition to contrasting the tropical (Amazon and Barbados) and polar regions (Arctic), we augmented the focus of the BACCHUS project by using well-established sites at Mace Head (North Atlantic), Hyytiälä (boreal forest) and Jungfraujoch (mainly in free troposphere). These latter sites enabled us to identify regional differences in CCNs and INPs.

This core idea has been developed around two central objectives:
Objective 1: To develop a robust methodology to quantify the influence of anthropogenic aerosol on cloud properties based on the estimate of the background levels of natural aerosols in various environments, the identification of their sources and their role in aerosol-cloud processes. Emphasis has been placed on aerosol-cloud interactions (rather than aerosol-radiation interactions) with a particular focus on the ice-phase as well as the involvement of biogenic and organic aerosols in modifying the properties of CCNs and INPs.
Objective 2: To characterize and understand the key interactions and feedback mechanisms in the terrestrial and marine biosphere-atmosphere-cloud-climate system by building on advanced in-situ observations, remote sensing, and numerical models operating over a wide spectrum of spatio-temporal scales and complexity.
Both objectives have been achieved within the BACCHUS project.

Summary description of project context
The BACCHUS project evolved around four scientific work packages. The goal of WP1 was to collect and harmonize long-term observations, field and laboratory studies of the physical and chemical properties of aerosols particles, CCN and INP. WP1 has developed a strategy to make use and to incorporate this data into existing European networks and synthesis projects. Combining this data with in-situ and remote measurements of warm and ice-containing clouds and aerosols, WP1 has provided the best estimate of the natural (pre-industrial) background aerosol, and discriminate anthropogenic and biogenic contributions to the CCN/INP aerosol fractions. WP1 compiled aerosol/cloud data sets for key contrasting environments and provided the basis for comprehensive modelling efforts in WP2-WP4.

WP1 successfully implemented the INP database and filled the database with recent and historical observations. Numerous campaigns (short-term, long-term, and well-coordinated intensive field observations in very different continental regions of the world, on the remote Atlantic, and the pristine Southern Ocean) focusing on CCN and INP observations have been successfully completed and analyzed. CCN datasets collected in various field campaigns have been analyzed concerning the relationship between the chemical and microphysical particle properties and the potential of the particles to serve as CCN. WP1 and WP2 collaborated and elaborated the first European long-term data set, linking CCN and chemical properties via experimental and theoretical closure studies and model approaches, focusing on a variety of background conditions, and taking particle sources and atmospheric aging into account. Field data on CCN-chemistry relationships sampled at Central-North Europe, in the Mediterranean, Amazonia, Artic, and Southern Ocean have been used for this effort.

Vertically resolved field observations of aerosols and clouds in key climate regions with networks of aerosol/cloud lidars and cloud radars, and aircraft, in the framework of field campaigns as well as networks (Cloudnet) have been successfully performed and analysed. The goal was to better understand aerosol-cloud interactions (with emphasis on mixed-phase cloud formation) for fresh and aged aerosol mixtures and to support in this way efforts of process, cloud-resolving and regional modeling as well as global climate modeling. Satellite remote sensing with very high horizontal resolution (Visible Infrared Imager Radiometer Suite, VIIRS) on board NASA’s Earth-observing satellite NPP was used to study aerosol-cloud-precipitation interaction (with emphasis on liquid-water cloud systems) on a global scale. The aerosol and cloud algorithm for the spaceborne AATSR/ATSR2 dual view radiometers has been further developed and applied to four regions in contrasting environments: biomass-burning regions in Amazonia, relatively clean Europe, Saharan dust outbreaks over the Central Atlantic, and polluted China. These data have been used in WP3 for the models/MODIS/AATSR cloud properties intercomparison.

The aim of WP2 was to understand and parameterise processes associated with aerosol-cloud interactions, which are missing or are currently not well represented in ESMs. WP2 made extensive use of results from WP1 and performed process modelling at various scales to improve/develop parameterisations suitable for pristine and polluted conditions. In particular the work in WP2 was targeted to better understand and parameterise the marine sources of organic aerosols as well as the terrestrial biogenic aerosol sources. These parameterisations were incorporated in CTMs and ESMs (see WP4). Further WP2 improved the organic aerosol representation in the models accounting for all major identified processes of OA formation and transformation in the atmosphere. As a synthesis of the BACCHUS project, WP2 initiated and led an international model intercomparison exercise with the participation of 15 global model that evaluated the uncertainty in CCN calculations in the global models and how this propagates to the cloud droplet number concentrations (CDNC) (deliverable 2.5). The differences found between models point to the size distribution of the emissions as one major source of uncertainty. The spread of models for CDNC is found smaller than the spread for CCN and for aerosol number concentration (Nα). The sensitivities of CDNC to Nα and to updraft velocity are negatively correlated and these two parameters are shown to be able to describe the variability of CDNC. However, the sensitivity of CDNC to these parameters is very different among models and from observations, implying that further process understanding and model improvements are required.

The task of WP3 has been to determine the key processes controlling cloud systems in contrasting environments and the relative role of natural vs. anthropogenic aerosol (precursor) emissions in each of them. Where available, it made use of data and process understanding gained in WP1 and WP2, conducted case and process studies and evaluated and improved ESMs used for climate projections, aerosol radiative forcing and feedback studies in WP4. A joint framework and case study protocol was devised in deliverable 3.1 and has subsequently been refined and updated to incorporate links to recent international activities. Regional case studies to investigate the relative role of aerosol emissions highlight the complexity of aerosol-cloud interactions in contrasting environments. A general finding across all case studies was that their results were considerably affected by the high degree of uncertainty in ice- and mixed-phase cloud microphysics, which affects their simulation of the cloud base state as well as the response to aerosol perturbations (White et al., 2017; Heikenfeld et al., 2018). For polar stratus clouds, an Arctic case study revealed that current, state of the art, high resolution models show significant diversity in simulating key parameters of Arctic clouds, even when CDNC/ice crystal number concentration (ICNC) are prescribed, and that the model results are further sensitive to uncertain ice crystal properties. The results further strongly support the hypothesis that the liquid water content of polar stratus clouds is CCN-limited (Stevens et al., 2018). This has significant implications for our ability to explain the role of clouds for the sensitive radiation balance of the Arctic in the light of future shipping emissions.

WP4 then used the improved ESMs based on the different findings and developments obtained in WPs 1-3 concerning aerosol emissions (especially organics from the terrestrial and marine biosphere, e.g. Hantson et al., 2017; Vergara-Temprado et al., 2017), secondary organic aerosol formation and CCN and INP formation in several studies (e.g. Vergara-Temprado et al., 2018, Huang et al., 2018). The ESMs were applied to i) provide estimates of radiative forcing and effective radiative forcing due to aerosol-cloud interactions (Fiedler et al., 2016), ii) identify and quantify the feedback processes in the biosphere-aerosol-cloud-climate system and to iii) investigate the effect of shipping in a future ice-free Arctic ocean on Arctic clouds (Gilgen et al., 2018).

Impact for society/policy makers:
BACCHUS has strongly enhanced our current understanding of aerosol-cloud interactions in the Earth-atmosphere system and as such reduced the uncertainties of current state-of-the-art ESM climate projections at different scales and in several ways. Examples are a quantification of the biogenic-atmosphere-cloud-feedback loop and the importance of marine organic carbon aerosols as INP (Huang et al., 2018) as discussed in detail in WP4. Further BACCHUS contributed prominently to an expert meeting on bounding the effective radiative forcing of anthropogenic aerosols as discussed in deliverable 5.9 being a community assessment of the importance of biogenic versus anthropogenic emissions for cloud formation and climate. It summarizes the major findings from the BACCHUS project in terms of sources of CCN, composition and abundance of INP and the anthropogenic aerosol forcing.
Results that are of interest for policy makers have been summarized in deliverable 5.8. The key findings are:
• Ice nucleating particles: BACCHUS created an INP database from worldwide measurements. Despite a high temporal and spatial variability of INPs, one conclusion from BACCHUS was that most INPs are of natural origin and are dominated by mineral dust and biological particles, such as pollen, bacteria and marine organic carbon. Land use changes due to human activities could therefore affect clouds through changes in emissions and properties of ice nucleating particles.
• The Arctic environment: The Arctic has experienced and will experience rapid climate change in the future. The large temperature increase is due to a combination of increased poleward heat transport and local feedbacks in the Arctic. An ice-free ocean also leads to more aerosol emissions, both natural (sea salt and sulphate) and through the expected increase in shipping. The BACCHUS project has shown that in a polluted Arctic, clouds will be thicker because of the increase in aerosol particles. Depending on local factors (environmental conditions, time of the year, meteorology) and also on the numerical model used, the timescale and amplitude of the aerosol response differs. A quantification of aerosol effects in Arctic clouds remains challenging. Overall, the Arctic remains a poorly understood natural system, which makes it potentially vulnerable to inadvertent pollution.
• Atmosphere-biosphere interactions: Changes in climate affect natural aerosol emissions and their precursors. More biogenic volatile organic carbon will be emitted that leads to more aerosol particles. This results in more but smaller cloud droplets and more reflection of sunlight back to space. It causes a cooling, which partly offsets the greenhouse gas warming. This feedback loop cannot be regarded as isolated from anthropogenic emissions because the biosphere is changed by human activities. Therefore, a proper assessment of anthropogenic interference requires studying anthropogenic emissions in concert with biogenic emissions.

Project Results:
WP1 Collection and harmonization of aerosol, CCN, INP and cloud properties datasets
WP1 covered the BACCHUS aerosol and cloud observations. Within BACCHUS, focus was placed on the characterization of aerosol particles regarding their potential to serve as CCNs and INPs based on past and published in-situ measurements as well as on new BACCHUS observations at continental and oceanic sites at rather contrasting aerosol conditions. Observational data were also collected in other European networks (see as an example, Figure 1.1) and during well-coordinated international BACCHUS field campaigns (Mace Head campaign in August 2015 and a series of campaigns in Cyprus performed during four-week intensive field phases in each spring season of 2015, 2016, and 2017). Strongly different environmental conditions prevail at the western and eastern boundaries of the European Union (i.e. clean marine at Mace Head, Ireland, versus highly polluted and dusty conditions at Cyprus, see Figure 1.2).

Besides the numerous ground-based in situ aerosol observations, WP1 covered a variety of activities to obtain vertically resolved information on aerosols and clouds in key climate regions by means of Unattended Arial Vehicles (UAVs), by long-term monitoring with aerosol/cloud lidars and cloud radars, organized in networks (e.g. ACTRIS, Cloudnet) or arranged in super sites within large field campaigns (e.g. BAECC, CyCARE 2016-2018), and by means of state-of-the art satellite remote sensing (VIIRS, AATSR). Because of the large number of WP1 activities and key findings as well as the limited space in this final report we highlight the key activities and results in form of a compact summary and refer to the relevant publications/deliverables for more detail.

INP observations in contrasting continental environments

One of the main goals of BACCHUS was the establishment of a global INP database (INDB) containing INP-relevant aerosol properties for different natural and anthropogenic aerosol types and mixtures around the world (see Figure 1.3). In contrast to the INBD, a CCN data base is already existing since years (GASSP). BACCHUS also contributed with numerous short-term and long-term CCN field studies (including re-analyzed data sets) to the CCN data base. Individual field studies with focus on INP concentration (INPC) and INP-relevant aerosol properties were collected in the framework of BACCHUS at Svalbard (Spitsbergen, Arctic, Norway), Gotenborg (Sweden), Mace Head (North Atlantic, western Ireland), Taunus (mountain, near Frankfurt, Germany), Jungfraujoch (3466 m, Alpes, Switzerland), San Pietro Capofiume (near Bologna, Italy), Monte Cimone (GAW, 2165 m, Apennine Mountains, Italy), Capogranitola (southern Italy, facing the Strait of Sicily, Italy), Cyprus (Troodos mountains, Nicosa), Martinique (Caribbean, tropics), Amazonia (Brazil, Amazon Tall Tower), Mount Kenya (GAW, 3680 m, Kenya), Storm Peak (Colorado, USA), and Cabo Verde. All these data were carefully analyzed, compared and contrasted, and finally integrated in the INBD, shown in Figure 1.3.

Shipborne INP observations over remote oceanic areas from the Arctic to Antarctic regions

During several ship cruises (RV Polarstern, RV Akademik Tryoshnikov, PEGASO shipborne campaign) INP and CCN observational data were collected over the Atlantic and Southern Ocean. Several Polarstern cruises from Punta Arenas, Chile or Cape Town, South Africa, to Bremerhaven, Germany, provided the unique opportunity to sample aerosol and INP data in the outflow region of Africa, i.e. in air masses with complex aerosol mixtures of marine, dust, anthropogenic (haze), and biomass burning (smoke) particles. The full spectrum from very efficient INP (dust) to very inefficient INP (sea salt) could be covered by these observations. The north-south Polarstern cruises allowed contrasting northern and southern hemispheric aerosol conditions and impacts. Figure 1.4 provides an overview of the Polarstern observations. By careful comparison of the Polarstern results and the rare Southern Ocean observations from the 1970s a clear and updated view of the INP conditions in pristine regions of the Southern Ocean is now available. The observations from the 1970s overestimated the INP potential of marine particles in the Southern Ocean significantly.

Mace Head campaign in August 2015

One of the key BACCHUS efforts was the investigation of INP properties of marine aerosol (see Figure 1.5) over the North Atlantic Ocean during the BACCHUS campaign at Mace Head in August 2015. The results can be summarized as follows: Sea spray aerosol (SSA) generated by bubble bursting at the ocean surface is an important component of aerosol-cloud interactions over remote oceans, providing the atmosphere with CCN and INPs. Studies have shown that organic INPs are emitted during phytoplankton blooms, but changes in INP number concentrations (INPC) due to ocean biological activity have not been directly demonstrated in natural SSA. INPC measurements performed allowed a thorough characterization of the INP population over mid-latitude oceanic regions, in terms of number concentration, nucleation efficiency and size distribution. The obtained results contribute significantly to fill the current gap of INP observations in the remote marine boundary layer. The complexities of predicting increases in INPC due to offshore biological activity are explored and a parameterization for predicting INPs in pristine SSA (i.e. not significantly augmented by offshore biological activity) over the North Atlantic Ocean was developed and illustrates that sea spray aerosol is associated with a factor of 1000 fewer ice nucleating sites per surface area of aerosol compared to mineral dust.

CCN-related aerosol studies during the Mace Head campaign

During the Mace Head campaign, sophisticated studies of the relationship of CCN characteristics, aerosol type, chemical composition, and microphysical properties were carried out. For the first time, the chemical composition of ultrafine particles (with diameter as low as 20 nm) could be performed using an aerosol mass spectrometer. Unexpectedly for the small mean diameter of the ultrafine particles mode and for the large organic fraction of these particles, a significant fraction of them were activated to CCN. As a consequence, CCN concentrations reached 1000 cm-3 in spite of the very clean conditions at the site. The observations could be reconciled with the Koehler theory only by taking into account a surface tension effect exerted by the organic compounds present in the small particles. It has been disputed for a long time the surface tension effect is cancelled by the simultaneous reduction in the solute, or Raoult, effect. In simulating the cloud droplet activation process using mixed organic-inorganic nuclei, we revealed that surface tension can be lowered without triggering changes in the Raoult effect through a process known as liquid-liquid phase separation (essentially an organic-rich layer on the drop’s surface keeps the surfactants separated from the internal aqueous solution occupying the core of the droplet). The model was able to explain the tenfold increase in cloud droplet number concentration observed. This phenomenon could be detected in many diverse environments throughout the world, reinforcing its role in cloud brightening and global climate cooling.

Analysis of CCN long-term measurements from regionally representative observatories

Observations derived long-term regionally representative aerosol properties are indispensable to evaluate models’ performances and improve representations of aerosols and of aerosol-cloud interactions in climate models. Until ten years ago collocated long-term observations of CCN activity, particle number size distribution and chemical composition were very sparse, but since then atmospheric observatories of the Aerosols, Clouds, and Trace gases Research InfraStructure (ACTRIS) as well as other international sites were able to provide quasi continuous comprehensive aerosol observations (including CCN) over multiple years. As part of BACCHUS, harmonized data records from 11 observatories were summarized. Studied environments include coastal background, rural background, alpine sites, remote forests and an urban surrounding. In terms of particle activation behaviour, most continental stations exhibit very similar activation ratios (relative to particles >20 nm) across the range of 0.1 to 1.0% supersaturation. Over land, both anthropisized and forest sites are characterized by size-distributions with a small geometric diameter (Fig. 1), because ultrafine particles are produced in large amounts by both anthropogenic (combustion) and biogenic (new particle formation) sources. The coastal sites show the most diverse behaviors. In the Arctic (Barrow) and in the North Atlantic (Mace Head) coastal areas, the transition from particles being CCN inactive to becoming CCN active occurs over a wider range of the supersaturation spectrum indicating that the composition of the aerosols is very variable in these environments. Most of the stations show strong seasonal cycles of CCN number concentrations and particle number size distributions. This also means that short-term measurements (e.g. from intensive field campaigns) can only be representative of the season in which they were performed. The average hygroscopicity parameter kappa calculated from the chemical composition of submicron particles was highest at the coastal site of Mace Head (0.6) and lowest at the rain forest station ATTO (0.2–0.3). We performed closure studies based on kappa-Köhler theory to predict CCN number concentrations. The ratio of predicted to measured CCN concentrations is between 0.87 and 1.4 for five different types of kappa. The temporal variability is also well captured, with Pearson correlation coefficients exceeding 0.87. Information on CCN number concentrations at many locations is important to better characterize ACI and their radiative forcing.

While monthly means provide information about seasonal cycles, changes in the aerosol properties over shorter time scales have been investigated using alternative metrics. For instance, autocorrelation of the hourly CCN data was used to describe the persistence of CCN concentrations over time scales from one day to two weeks. In addition, frequency distributions of the hourly CCN concentrations as well as 2D-distributions connecting the variability in CCN concentrations to concurrent changes in the concentrations of aerosol chemical compounds (determined by long-term AMS measurements) were also used to characterize the phenomenology of CCN at the regional atmospheric observatories and to provide new metrics suitable for comparison with global models.

Overall, the dataset is appropriate for comprehensive aerosol characterization (e.g. closure studies of CCN), model-measurement intercomparison and satellite retrieval method evaluation, among others. Data have been acquired and processed following international recommendations for quality assurance and have undergone multiple stages of quality assessment. The harmonized multi-year data of three key variables (cloud condensation nuclei, particle number size distribution and chemical composition) for aerosol-cloud interactions is a unique collection to test satellite retrieval methods and to evaluate global climate models targeting the reduction of the related uncertainty in radiative forcing.

Profiling of aerosols and clouds with UAVs at Cyprus

UAVs were used to measure (in situ) INP height profiles during the Cyprus-2016 BACCHUS campaign. A large number of dust outbreaks from Africa causing dust layers up to several kilometers over the field site in Cyprus occurred. As a highlight in situ measured INP profiles were compared with INP profiles retrieved from active remote sensing with lidar. A new lidar-based methodology was developed in the framework of BACCHUS to estimate INP profiles from polarization lidar observations.

One of the key results of the Cyprus 2016 campaign is that lNP concentrations measured at surface are usually not representative for INPC in the profile. INP concentrations obtained from surface observations were on average (for the entire Cyprus 2016 campaign) an order of magnitude lower than the profile values (collected between 250 m and 3 km height above the filed site (Figure 1.6). There is a clear need for INPC observations in the profile preferably at heights at which mixed-phase clouds develop. This corroborates the importance of continuous lidar observations (in combination with cloud radar observations) to permit profiling INPC, aerosol and cloud properties simultaneously at height levels where cloud formation takes place.

UAVs were also used to study (in situ) aerosol-cloud interaction (aerosol-cloud closure studies here with focus on the relationship between aerosol and liquid-water cloud microphysical properties and related radiation fluxes). The observations included updraft and downdraft measurements and corroborated that a good knowledge of aerosol, cloud and vertical wind conditions, or, more general, of meteorological conditions is a prerequisite for the study of the impact of aerosol particles on cloud evolution.

Profiling of aerosols and clouds by means of active remote sensing (Cloudnet, Cyprus campaigns)

A key task was the height-resolved investigation of mixed-phase clouds (e.g. in terms of ice water content vs liquid water content) in different regions (with different aerosol and meteorological conditions) by means of active remote sensing (aerosol lidar, cloud radar, microwave radiometer, Doppler lidar) in the framework of the European network Cloudnet (EU ACTRIS-2). The ACTRIS network has been increased (by BACCHUS efforts) and now include stations in the tropics (Barbados) and temporal stations in Finland (mobile ARM facility, 2014), the Eastern Mediterranean (Cyprus, TROPOS, LACROS station), and in Punta Arenas (Chile, 2019-2018, TROPOS LACROS station). This activity is long lasting and ongoing. Many efforts (Cyprus, Chile) are voluntary contributions to BACCHUS. New data analysis schemes were developed in the framework of BACCHUS covering ice-nucleation aspects.
No distinct differences in the statistics of mixed-phase clouds (e.g. ice flux or ice water content as a function of cloud top temperature from 0°C to -40°C was found for the different stations from Ireland to Cyprus, from the mid latitudes to the tropics (Figure 1.7). The analysis was based on observations of shallow altocumulus layers with well-defined cloud top height and cloud top temperature. The cloud top region is the coldest layer of the cloud system and here the probability is highest to heterogeneously nucleate ice particles.

Formation of precipitation was triggered in about 50% out of all observed events by clouds with cloud top temperature of <-38°C (homogeneous freezing), and also about 50% by clouds with cloud top temperature of >-38°C and <0°C (heterogeneous freezing), cases with warm-cloud precipitation (at cloud top temperatures >0°C) were negligible. For the tropics (Barbados) this is rather different. Precipitation events were linked to 70-80% of the cases to clouds with cloud top temperature of >0°C and to a minor part (20-30%) to clouds with cloud top temperature between 0 and -60°C (heterogeneous and homogeneous freezing). In several cases (during the CyCARE 2016-2018 long-term Cloudnet field study), ice formation was observed in clouds with cloud top temperature as high as -5°C. This clearly indicates the impact of biological particles (such as pollen during the winter and springs seasons which coincide with the rain season) or biogenic material on the nucleation of ice crystals.

Furthermore, long-term 12-year Cloudnet datasets collected at the Meteorological Observatory Lindenberg of the German Weather Service allowed us to contrast heterogeneous ice formation in different seasons. This differentiation permits the study of a potential impact of biological and biogenic particles on ice formation (in the cloud top temperature range from 0 to -10°C). For this temperature range, the lowest ice flux values (sedimentation of ice crystals out of the altocumulus cloud base) are found in winter, and a factor of 2-3 higher values in summer and autumn (and probably in late spring, May). We may conclude that stronger convective processes in the summer half year transport more boundary-layer aerosol into the free troposphere (including biogenic particles). However, also long-range transport of aerosols may favor ice formation because more dust outbreaks from Africa towards higher northern latitudes take place in the summer half year. Forest fire smoke from southern Europe and North America reach central Europe in summer. In winter, the free troposphere mainly contains aged background aerosol and the aerosol levels are reduced compared to the summer months. Thus, solid conclusions could not be drawn.

Profiling of clouds by means of passive remote sensing

A highlight of BACCHUS was the intensive use of VIIRS (Visible Infrared Imaging Radiometer Suite) to investigate the impact of aerosol particles on liquid-water cloud evolution on a global scale. The method, developed in the framework of BACCHUS, mainly relies on the data products of the (VIIRS passive sensor, onboard the Suomi National Polar-orbiting Partnership (NPP) satellite. Suomi-NPP is a sun-synchronous polar-orbiting satellite that overpasses at 13:30 solar time. The VIIRS sensor has a very high spatial thermal resolution of 375 meters at nadir that is useful for detecting small cloud elements.
The new methodology is utilized to produce automated maps of retrieval of microphysical properties of convective cloud fields over large areas. The developed Automatization Mapping of Convective Clouds (AMCC) system enables to investigate natural and man-made causes on large scales. In order to produce the AMCC, a scene of satellite image of about
2400×2304 km is undergoing image segmentation to pixels of 36×36 km. The microphysical properties are retrieved for each moving window using the satellite retrievals. Afterwards, cirrus screening mechanism is applied to filter out cirrus clouds. The retrieved CCN based on vertical profiles of observed cloud microstructure was validated against measurements in BACCHUS field projects. Based on that, we fine-tuned the retrieval algorithm of CCN.

Based on CCN mapping as shown in Figure 1.8 it is shown that in unperturbed environments such as the Green Ocean Amazon and Southeast Australia CCN spatial distribution patterns are similar. convective clouds over ocean exhibit very low concentrations of CCN varying from 20-100 cm-3. CCN concentration increases at shore, and continue to increase inland until stabilizing at values of 100-200 cm-3. Precipitation scavenging might lead to a reduction of CCN both at Ocean and inland. As clouds move inland cloud base height rises, becomes colder and with stronger updrafts. Topography conditions might also change cloud properties inland. Aerosols natural perturbations in pristine environments (e.g. occasional wildfires ignited by lightning) can increase dramatically CCN concentration, which in turn might affect clouds properties such as precipitation rate and cloud radiation interactions. Aerosols perturbations are less conspicuous with respect to the already perturbed background in areas that are influenced by human activities. For example, at the southern coast of Texas CCN concentration increases close to emission sources. Background CCN concentration inland is calculated to be about 400 cm-3, which is higher by a factor of 2 than at pristine environments inland. In heavily polluted areas such as Southern China, the impact of specific aerosol emission sources on clouds properties is almost unnoticeable, since the background CCN concentration is already very high (above 1000 cm-3). This study has also shown that when comparing satellite retrievals and the WRF 3.4/CMAQ 5.01 coupled modeling system, large pollution sources are detected by both methods. Yet, there are remaining considerable dissimilarities away from these sources. This means that there is still much room for improvement in the models, and that the satellite observations can be useful for constraining the simulations.

In summary, BACCHUS triggered many activities in the field of heterogeneous ice formation. It linked the observational and the modelling community. Many activities could be completed, several are just started, others are ongoing and will keep the research community busy during the next 10 years. BACCHUS also showed that well-coordinated field campaigns with airborne in situ instrumentation and modern active remote sensing are required to investigate aerosol-cloud interaction with focus on ice nucleation and formation of precipitation in order to deepen and improve our knowledge in one of the most complex and at the same time important field of atmospheric science.

WP2 Process studies of the role of both organic and inorganic aerosol in CCN/INP
WP2 overall aim was to understand and parameterize processes associated with aerosol-cloud interactions, which were missing or were not well represented in ESMs. WP2 made extensive use of results from WP1, performed process modeling at various scales and improved/developed parameterizations suitable for pristine and polluted conditions, which were used in WP3 and WP4. Particularly WP2 improved (i) the marine aerosol, including organics, CCN and INP, the terrestrial biogenic aerosol sources parameterizations, as well as the sources of organics from vegetation and open fires for use in CTMs and ESMs (see deliverable 2.1); (ii) improved the organic and inorganic aerosol representation in the models accounting for all major identified processes of OA formation and transformation in the atmosphere and in particular the involvement of organics in the nucleation and growth to CCN size (see D2.2 D2.3); (iii) developed parameterizations of primary sources of aerosols, formation and growth of CCN, ageing of aerosols by atmospheric chemical and dynamic processing and contribution of aerosol sources to INP (see D2.3 D2.4 D2.5).

In particular for the sea-spray source function and the primary biogenic aerosol from terrestrial sources the parameterizations are related to indicators of “biological activity” (chlorophyll, terrestrial vegetation dynamic), and are found suitable for different environments (O’Dowd et al 2015; Wilson et al., 2015; Myriokefalitakis et al., 2017). A physically based approach of Burrows et al. (2014), in which the organic fraction is computed using the sea surface concentrations of five different classes of macromolecules: proteins, polysaccharides, lipids, processed compounds, and humics has been also tested. Biogenic Volatile Organic Compounds (BVOC) emissions were parameterized in the LPJ-GUESS dynamic vegetation model by Hantson et al. (2017) using a (semi-)process-based approach, linking emissions to the photosynthetic activity of the vegetation and taking into consideration the direct and indirect process-response of BVOC emissions to changing climate and CO2 concentration. The effects of climate change and changes in atmospheric CO2 levels on fire conditions, amount of combustible litter and vegetation wildfire emissions were assessed using a simple but non-linear model of fractional burned area, SIMFIRE, coupled with the LPJ-GUESS model (Hantson, et al., 2016), and provided for ESM studies in WP4. For primary biological aerosol particles extensive measurements of fluorescent biological particles from WP1 have been used (D2.1).

Simplified parameterizations have been also developed and tested to investigate the impact of aerosol origin (natural versus anthropogenic) and chemistry (organic aerosol formation and aerosol ageing) on CCN and sensitivity modelling studies are performed to estimate the associated uncertainties. The effects of aerosol size, composition and semi-volatile organic compounds (SVOCs) on CCN activation and warm rain have been studied using data from WP1 and a detailed process model and comparing natural versus anthropogenic conditions (Crooks et al., 2016; Connolly et al., 2014).

The particle mixing state (internal vs. external mixtures) has been analyzed with respect to CCN 'closure' studies of observed CCN concentrations and size distributions (Schmale et al., 2017; 2018). CCN closure studies (Schmale et al., 2017; 2018), performed on the long-term observational data of CCN number concentrations, particle number size distributions and particle chemical composition at seven regionally representative observatories, have shown that applying a simple kappa-Köhler formulation assuming internal mixture and size-independent particle hygroscopicity to predict CCN results in relatively high prediction skills (statistically significant correlation coefficients between 0.87 and 0.98;). The measured ion composition was used to derive κ for inorganic aerosol constituents, while κorg was assumed to be 0.1 and no information on BC mass concentrations was used. Similarly good agreement between measured and predicted CCN0.5 number concentrations was obtained, using an overall κ=0.3 (D2.2 D2.5). Furthermore, a simple core-shell model combined with kappa-Köhler theory has been proposed to predict the activation behavior of black carbon and improve the representation of the contribution of black carbon to the CCN population (Motos et al., in prep).

Based on a large perturbed parameter ensemble (PPE) of the HadGEM-UKCA climate model that was created as part of the Global Aerosol Synthesis and Science Project (GASSP; UK project), BACCHUS has shown that κorg contributes locally about 30-40% of the CCN variance as an annual mean with its largest effect in July. The largest effect on CCN occurs over regions with large organic emissions, particularly over the boreal forest, West Africa and Amazonia. Changing κorg affects the water uptake of the aerosol, and therefore the ambient aerosol size and thus the AOD. It also affects CDNC (both directly through changes in activation, and also indirectly through changes in the aerosol distribution due to different deposition, coagulation rates through the change in ambient size). Based on that PPE, κorg affects by less than 25% the AOD and ERFari variance over strong biogenic emission regions; it is therefore a major cause of uncertainty in aerosol radiative forcing (D2.4)

During the Mace Head measurement campaigns, cases were found where the Kelvin effect (curvature) can prevail over the Raoul effect (chemical composition), leading to substantial increases in cloud droplet concentrations and the need to account for phase separation effects into activation thermodynamics (Ovadnevaite et al. 2017; Fig 2.1).

Based on combined field measurements, remote sensing and biological/plankton/chemistry laboratory studies as well as detailed process models, BACCHUS has developed several new parameterizations for NPF as well as CCN and INP as a function of organic and inorganic aerosol components. BACCHUS improved organic aerosol representation in the models for the aerosol nucleation scheme (Makkonen et al., 2014; Määttänen et al., 2018), accounting for organic, inorganic, neutral and ion contributions to new particle formation (NPF) and for NPF growth to CCN as well as for atmospheric ageing of aerosols (D2.3 and D2.5). These have been tested in various scale models and appropriate parameterizations have been incorporated in ESMs. The new organic-related formation and growth mechanisms described above have been studied individually in several studies (Riccobono et al., 2014; Dunne et al., 2016; Gordon et al., 2016 and Troestl et al., 2016). Organic nucleation has been estimated to contribute between 28 and 45% of all nucleated particles in the atmosphere (D2.3). Observations of new particle formation (NPF) on Jungfraujoch during 9 months (Frege et al., 2017) show that NPF are characterized by either highly oxygenated molecules clustered with nitrate and low sulfuric acid concentration, or ammonia-sulfuric acid clusters in air masses that have been in contact with the boundary layer prior to the observed nucleation in the free troposphere (D2.3). Kulmala et al. (2016) analysed primary and secondary aerosol contributions to nucleation, Aitken, and accumulation mode sized aerosols in two locations: rural boreal forest (Hyytiälä, Finland) and polluted megacity environment (Nanjing, China). In both locations, secondary particles dominated nucleation and Aitken mode concentrations, but even in accumulation mode in Nanjing, secondary aerosols formed a considerable fraction of the total population. The dominance of secondary aerosols has implications on both natural and anthropogenic contributions to aerosol number loading, as well as potential perturbations to aerosol number via emission reductions (D2.3).
For INP, parameterizations have been based on laboratory studies of mineral dust and biological particles and have been validated against field measurements of INP from the BACCHUS IN database (Wilson et al., 2015; Boose et al., 2016; Vergara Temprado et al., 2016; Huang et al., 2017 details in D2.4 and D2.5). However, measurements of ice nucleating particles observed from ships transiting the Port of Gothenburg, Sweden and direct comparisons to the clean background aerosol show that the sampled shipping emissions have enhanced ice nucleating properties (Thomson et al., 2018). Based on modeling results (Possner et al., 2017), such observations might have implications for environments with low particle concentrations and non-linear aerosol cloud feedbacks, like the Arctic which is opening to transit and human activities in the region since the sea ice pack is declining due to global warming.

Studies on organic aerosol formation and ageing in the atmosphere have been performed with regard to their contribution as CCN (Gordon et al., 2017; Fanourgakis et al., 2018) and INPs (Vergara Temprado et al., 2016; Huang et al., 2017) as well as on the importance of multiphase chemistry for aerosol ageing and for organic nutrient formation in the atmosphere and deposition to the surface (Kanakidou et al., 2018) (D2.2). The importance of biogenic volatile organic compounds (BVOC) in NPF and CCN formation was found to be greater than previously thought since all secondary organic aerosol (SOA) from BVOC is calculated to increase low-cloud-level CCN0.2% by about 26% in the present-day atmosphere and about 41% in the preindustrial. About 75% of the impact of SOA on CCN0.2% is due to the tiny fraction of the oxidation products of BVOCs that have very low volatility and are thus involved in NPF and early growth (Gordon et al., 2017; Figure 2.2). These simulations also suggested that the effect of changes in cosmic ray intensity on CCN is small and unlikely to be comparable to the effect of large variations in natural primary aerosol emissions (D2.5). Because of the simulated production of significant amounts of natural atmospheric aerosol contributing to the baseline preindustrial aerosol concentrations, these finding could lead to a reduction of 27% in estimates of anthropogenic aerosol radiative forcing (Gordon et al., 2016).

An international model intercomparison exercise was also initiated and led by BACCHUS with the participation of 15 global model that evaluated the uncertainty in CCN calculations in the global models and how this propagates to the CDNC. The models have been thus evaluated by comparisons between models and between models and observations for the PM1 composition, the aerosol number concentration for aerosols at various sizes, the CCN concentrations at various supersaturation and the derived cloud droplet concentrations. The evaluation focused both on the models’ ability to simulate the average state across diverse environments, and on the seasonal and short-term variability in the simulated time series. For this comparison, the same parameterization has been used to derive CDNC from the simulated spectra of CCN (modelled CDNC) and from the observed CCN spectra (observed CDNC) (Figure 2.3).

The differences found between models point to the size distribution of the emissions as one major source of uncertainty. Supersaturation ratios (smax) control the fraction of aerosols that will activate to CCN and show high values when aerosol number concentrations (Na) is low while the opposite happens when Na is high due to competition of aerosols for the available atmospheric water vapor. Furthermore, high updraft velocities facilitate activation of aerosols and thus lead to higher number of CDNC than under low updraft velocity (for the same amount of Na). Finally, the sensitivity of CDNC to Na is low when Na is high. In this case CDNC is more sensitive to the updraft velocity. These patterns are shown in Figure 2.3 for four stations where data are available. The spread of models for CDNC is found smaller than the spread for CCN and for Na. The sensitivities of CDNC to Na and to updraft velocity, ω, are negatively correlated and these two parameters are shown to be able to describe the variability of CDNC. However, the sensitivity of CDNC to these parameters is very different among models and from observations. In most case model sensitivities are much lower than derived from observations implying that further process understanding and model improvement is required (D2.5 and Fanourgakis et al., to be submitted to ACPD summer 2018).

WP3 Role of aerosols vs. dynamics for different cloud systems
The objective of WP3 was to determine key processes controlling cloud systems in contrasting environments and the relative role of natural vs. anthropogenic aerosol (precursor) emissions in each of them. These objectives were addressed through three main lines of research:

Task 3.1 & Task 3.2: Case studies to investigate the key processes controlling cloud systems in contrasting environments and the relative role of natural vs. anthropogenic aerosol (precursor) emissions in each of them using models of different resolution and complexity.
Polar stratus clouds – Arctic Case Study
The protocol for the Arctic case study (Stevens et al., 2018) with focus on polar stratus clouds was based on the observations available during the 2008 ASCOS campaign, when observed cloud condensation nuclei (CCN) concentrations fell to exceptionally low levels of 1 cm−3. Previous analyses suggested that at these low CCN concentrations the liquid water content (LWC) and radiative properties of the clouds are determined primarily by the CCN concentrations. The intercomparison includes results from three LES (UCLALES-SALSA, COSMO-LES, and MIMICA) and three NWP models (COSMO-NWP, WRF, and UM-CASIM). The case study tested the sensitivities of the model results to different treatments of cloud droplet activation, including prescribed cloud droplet number concentrations (CDNC) and diagnostic CCN activation based on either fixed aerosol concentrations or prognostic aerosol with in-cloud processing.

As evident in Fig. 3.1 above, there remains considerable diversity in simulated cloud properties even in experiments with prescribed CDNCs and prescribed ice crystal number concentrations (ICNC) with current state-of-the-art high-resolution models. Our results therefore suggest that properly estimating aerosol-cloud interactions requires an appropriate treatment of the cloud droplet size distribution within models, as well as in-situ observations of hydrometeor size distributions to constrain them. The results strongly support the hypothesis that the liquid water content of these clouds is CCN-limited. For the observed meteorological conditions, the cloud generally did not collapse when the CCN concentration was held constant at the relatively high CCN concentrations measured during the cloudy period, but the cloud thins or collapses as the CCN concentration is reduced. The CCN concentration at which collapse occurs varies substantially between models. Global and regional models with either prescribed CDNCs or prescribed aerosol concentrations would not reproduce these dissipation events. Additionally, future increases in Arctic aerosol concentrations would be expected to decrease the frequency of occurrence of such cloud dissipation events, with implications for the radiative balance at the surface.

Shallow Convection – Barbados Case Study
The original case study protocol was extended to seize the opportunities provided by recent field measurements during of shallow convection in the tropical North Atlantic, near Barbados as part of NARVAL-II campaign (Stevens et al., 2016). In addition, during NARVAL-II ground-based measurements from the Barbados Cloud Observatory were augmented by extensive aerosol measurements, including supersaturation resolved CCN and bio-aerosol, made by MPI-C. High-resolution modelling studies were performed using the ICON model with unprecedented resolution of their horizontal grid and expansiveness of their domain.
In addition to the identification of specific case studies, and initial simulations with ICON, analysis of the ICON data has been performed for the purposes of comparing with the air-borne (droposonde) data. From the high-resolution ICON data dropsonde profiles have been constructed to mimic the measurements performed by the aircraft and these are being compared to the analysis of the actual dropsonde data. Profiles from the sondes and the ICON simulations have been analyzed to help define the cases for LES and SCM studies. Examples of this analysis are shown in Fig. 3.2 for the period of analysis in NARVAL1 and NARVAL2. In particular, this analysis highlights the very different humidity and vertical velocity structure.
The NARVAL-II cases, also the contrast to the winter cases from NARVAL-I, are being used to test the generality of the winter vs. summer cases constructed based on measurements at the Barbados Cloud Observatory as well as earlier studies of the effects of aerosol-cloud interactions on the development of fields of cumulus clouds (e.g. Vogel et al., 2016). The ability to link these simulation studies to the airborne measurements and constrain them by CCN measurements is helping to advance our understanding of the relative role of aerosol versus dynamic influences on clouds.
Deep Convection – Amazon Case Study
Multiple modelling tools have been used to investigate the cloud formation and evolution in the Amazon areas for a case study with focus on the GOAMAZON and ACRIDICON campaigns. Detailed analysis modules for different models were developed to improve our understanding of cloud microphysics, aiming to guide the improvement of future climate models. Simulations were performed using the WRF model as well as the global climate model ECHAM-HAM-CCFM with explicit representation of the convective cloud spectrum (Kipling et al. 2017; Labbouz et al. 2017) and extended cloud microphysics (developed during BACCHUS). To establish consistency, the aerosol model HAM used in ECHAM-HAM has been implemented into WRF. The simulations were performed with a setup consisting of three nested domains centred around the ARM measurement facility at Manacpuru. ECHAM-HAM-CCFM is used in Single Column Mode (SCM) in order to compare the results with the domain-average high resolution WRF simulations. To isolate the effect of cloud microphysics, microphysical aerosol effects are represented by two different values for CDNC in the simulations with both WRF and ECHAM-HAM-CCFM. The SCM performs well in reproducing the precipitations over the domain, and WRF also gives satisfactory results until 21 UTC but does not simulate any precipitation afterwards (Fig. 3.3). Increasing CDNC leads to a delay in both the onset and the peak of precipitation simulated by WRF, however the delay is small and it is hence difficult to conclude on a microphysical effect from an only one day long simulation. Moreover, there is no significant impact on the hydrometeor profile to support a CDNC impact. The ECHAM-HAM-CCFM SCM appears to be insensitive to changes in convective cloud-base CDNC; consistent with a complex response to aerosol perturbation in global model simulations (Kipling et al., in prep.).

Deep Convection – ACPC Houston Case Study
In addition to the originally proposed case studies, BACCHUS took also leadership in an international Aerosols, Clouds, Precipitation & Climate initiative (iLeaps/IGAC/GEWEX) deep convection case study over Houston (van den Heever et al., 2017). This case study applies BACCHUS diagnostics, tools and analyses from the Amazon case study in an international context.
A summer-season case study of isolated convective cells developing under onshore flow conditions in the Houston region was identified using a combination of ground-based weather radar and satellite remote-sensing observations. Aerosol profiles representing clean versus polluted conditions in the Houston area were derived from airborne in situ measurements obtained during NASA’s Deriving Information on Surface conditions from Column and Vertically Resolved Observations Relevant to Air Quality (DISCOVER-AQ) field campaign. Six different modeling groups from Germany, England, Israel and the USA are contributing to the modelling study. Some emerging results about the aerosol-perturbation impact on the convection were identified across models (e.g. shifts in precipitation intensity).
However, substantial differences between the models’ base state and response to aerosol perturbations remain, as is evident from Figure 3.4 showing an inconsistent response for clean and polluted cases.

Task 3.3: Parcel model cloud microphysics closure studies: traditionally bottom up (in-situ aerosol/CCN/INP data via parcel model to CDNC/ICNC) and novel top down (satellite via parcel model to CCN/IN)

For the bottom-up closure studies we made use of data from WP1 to conduct parcel model closure studies of cloud microphysical parameters. We used the detailed Aerosol-Cloud-Precipitation Interactions Model (ACPIM) cloud parcel model (UMAN) and evaluated the convective cloud field model (CCFM) used in ECHAM-HAM in single column mode. For the top-down approach, we retrieved the number of activated particles at cloud base based on the vertical profile of the dependence of the effective radius on cloud top temperature from the NPP/VIIRS imager in non-precipitating convective clouds. This has been used to infer CCN concentrations from satellites in a manner similar to Rosenfeld et al. (2012). We focused here on a case of cumulus congestus over the UK during the COPE field campaign over the Cornish peninsula on the 25th July 2013. This is a clean aerosol case with an active warm rain process. A summary of results from the BAe-146 aircraft and ACPIM are shown in Figure 3.5. ACPIM models the activation of cloud drops from observed aerosol particles and their collision and coalescence to form rain. It therefore models the evolution of effective radius within the cloud.
Overall, the model simulation agrees well with the observation and the satellite retrieval (Figure 3.5). The effective radius increases with decreasing temperature at a fairly steady rate; however, at around 0 °C, we see a marked change in the rate of increase. This corresponds to the onset of warm rain in the model. Generally, the maximum observed values of CDNC compare favorably with those from the model. Only few maxima of liquid water content follow the adiabat, with most of the values being lower. Hence, in this case only few regions are observed that correspond to adiabatic ascent. For this case, looking at the cloud properties at 0 °C, the satellite retrieved effective radius is 20 μm and the liquid water content is around 3 g m-3. Hence, we retrieve a drop concentration of
~100 cm-3. However, we see in Figure 3.5c that the measured CDNC is around a factor of 2 higher than this - approximately 350 cm-3. Our results suggest ways forward to improve the retrieval of CDNC from the NPP / VIRRS satellite instrument by accounting for the broadness of the droplet size distribution.

Task 3.4: Evaluation of global ESMs and connection of regional case studies through global modelling, satellite data and global compilations of in-situ data

For this task, the BACCHUS ESMs were set up for a hindcast over 2008, nudged to ECMWF data. To establish consistency with satellite retrievals, the ESMs used the MODIS COSP observational simulator package. MODIS C6 L3 daily data were used for global-scale evaluation.
The evaluation of key cloud properties in BACCHUS ESMs, shown in Figure 3.6 reveals generally a generally reasonable agreement of simulated liquid cloud properties (b,d) – and highlights the large uncertainties in simulated ice-phase clouds (a,c).

WP4 Feedback processes in the climate system
WP4's tasks were to use the improved ESMs and applied them to i) global modelling of INP, ii) provide estimates of radiative forcing and effective radiative forcing due to aerosol-cloud interactions, iii) quantify the feedback processes in the biosphere-aerosol-cloud-climate system and to iv) investigate the effect of shipping in a future ice-free Arctic ocean on Arctic clouds as discussed below.

Task 4.2. Global modelling of INP
Over the Southern Ocean climate models simulate far too little reflection of shortwave (SW) radiation compared to satellite observations (Trenberth and Fasullo, 2010; Bodas-Salcedo et al., 2014). Excess absorption of downwelling SW radiation is thought to be an important cause of the error of about 2 K in the Southern Ocean sea surface temperature observed in several global climate models (Wang et al., 2014). This error has important consequences for the ability of climate models to simulate sea ice, the jet stream and storm track location (Ceppi et al., 2012). Previous analysis shows that most of the simulated radiative bias is associated with low and mid-level mixed-phase clouds (Bodas-Salcedo et al., 2014).

We used our global model of INP concentrations based on dust and marine organics (D4.2; Vergara-Temprado et al., 2017) together with a high-resolution regional version of the global Unified Model to quantify ice formation and the impact on the radiative properties of cyclonic systems over the Southern Ocean (Vergara-Temprado et al., 2018). The regional nest within the global model includes a double-moment microphysics scheme, which is required so that we can link the concentration of INP to the number concentration of ice crystals. The global model, in common with most climate models, simulates only a single moment of ice (mass but not number). Typical INP concentrations at -20 oC in the Vergara-Temprado model over the S Atlantic region are about 10-2 L-1 (range 10-4-1 L-1) but greater than about 10 L-1 in the Meyers et al. (1992) temperature-based parameterization. We simulated three cyclonic cloud systems, each containing extensive regions of stratocumulus and cumulus mixed-phase clouds (one example is shown in Figure 4.1).

The range of INP concentrations in the Vergara-Temprado model results in a range of simulated SW fluxes that spans the observations. The same is true for LWP (Vergara-Temprado et al., 2018). In contrast, the Meyers et al. (1992) INP parameterization, which predicts unrealistically high INP concentrations over the Southern Ocean, simulates SW fluxes that are 28–36% too low and LWPs that are a factor of seven too low. Our results show that the very low INP concentrations from natural sources over the remote Southern Ocean cause mixed-phase clouds to persist in a supercooled state for longer than similar clouds in high-INP environments.
Task 4.3. Aerosol climate forcing
Aerosol-cloud interactions and the associated radiative effects of BACCHUS ESMs were evaluated in D3.5 and D4.3. As evident in Figure 4.2 the ESMs showed a distinct increase in the liquid water path (LWP) of about 6-7% to anthropogenic aerosol perturbations. Strong increases in LWP to aerosol perturbations may be exaggerated and linked to excessive aerosol radiative forcing (Quaas et al., 2009), so this distinct LWP response will require further attention. The precipitation response to aerosol perturbations was highly variable across models.
The BACCHUS ESMs simulate relative model diversity in the direct aerosol radiative effects, often considered to be well understood, that exceeds the diversity in the indirect radiative effects. Nonetheless, the absolute diversity is larger for the indirect radiative effects. This can likely be attributed to the strong sensitivity of direct forcing to aerosol absorption, which modulates the sign of the top-of-atmosphere forcing. The BACCHUS ESMs simulate total aerosol effective radiative forcing (ERF) forcing ranging from -0.96 Wm-2 (ECHAM-HAM) to -1.59 Wm-2 (HadGEM-UKCA). It is likely that these relatively strong ERFs are driven by the strong increases in cloud liquid water. Fundamental research on cloud and aerosol processes remains a key research priority.

Task 4.4. Biosphere-atmosphere-cloud-climate interactions and feedbacks

Towards quantifying Earth System feedbacks via aerosols, UiO used the upgraded and evaluated version of NorESM (Norwegian Earth System Model) and performed simulations to investigate feedbacks associated with changes in emissions of BVOC, formation of SOA, CCN, clouds and climate. We have run 3 different experiments with NorESM where we 1) double CO2, 2) increase the sea surface temperature (SST) 3) both double CO2 and increase SST. All three experiments were run once with interactive BVOC emissions (FB-ON) and once with fixed BVOC emissions at present day levels (FB-OFF) to be able to estimate the impact of the BVOC-feedback.

The global BVOC emissions in this experiment almost double compared to present-day conditions. The modelled SOA formation is also almost twice as high in the FB-ON simulation compared FB-OFF simulation indicating that including the BVOC feedback has a strong impact on the SOA budget. The aerosol number concentration and CDNC are also higher in the FB-ON simulation as a result of the increase in SOA. This is in particular true for the Arctic and it is also here that we see that the feedback has the strongest effect on clouds and climate, see Figure 4.3. The average global net cloud forcing (NCF) is -0.3 W m-2 lower in the FB-ON simulation compared to the FB-OFF simulation.

Simulations that test the model sensitivity to the parametrization of SOA, VOC emission strength and levels of natural background have also been run with NorESM. 5 experiments were set up to test these sensitivities. In the first/second experiment the yield for all reactions including SOA formation was increased/decreased by 50%. In the third experiment all formed SOA become semi-volatile SOA (SVSOA), which only condense onto pre-existing aerosols and does not participate in new particle formation. The last two experiments consisted of turning off the isoprene emissions and monoterpene emissions, respectively. These last two tests are interesting since only monoterpene can form low volatile SOA (LVSOA) and contribute to new particle formation while isoprene contributes with more than three times as much mass than monoterpene. The results from these simulations indicate that NorESM is fairly sensitive to changes in the BVOC parametrizations and background levels.

Task 4.5 Future scenario simulations

KIT prepared an emission dataset for both fire and BVOC emissions. Emissions are given for both present-day and future scenarios (RCP4.5 & 8.5) taking into account changes in climate, human population density, land use, [CO2] and natural vegetation dynamics. Emissions are simulated using the global vegetation model LPJ-GUESS (Smith et al., 2014). For BVOCs both isoprene and monoterpene emissions are produced, with both having a very different spatial pattern under present-day conditions, and different temporal pattern under global change (see figure 4.4 for RCP4.5; Hantson et al., 2017).

The impact of ship emissions in a future ice-free Arctic ocean was investigated with the global aerosol-climate model ECHAM6-HAM2 (Gilgen et al., 2018 and deliverable 4.5). Simulations were conducted for the years 2004 and 2050 and focus was placed on the periods late summer (July/August) and early autumn (September/October). Next to changes in natural aerosol emissions, the impact of enhanced Arctic shipping activity in the future was considered. The future ship emissions are based on the study by Peters et al. (2011) and consider changes in shipping related to transport and oil/gas extraction. In the simulations, the sea ice area decreases from 6.1*10⁶ km² to 3.4*10⁶ km² and from 5.7*10⁶ km² to 2.3*10⁶ km² in late summer and early autumn, respectively; present-day sea ice area is derived from observations, whereas the future sea ice area is based on future simulations from the Earth-System-Model MPI-ESM (RCP8.5).

Sea ice acts as a barrier between the ocean and the atmosphere, therefore the decrease in the sea ice area leads to higher emissions of natural aerosol particles (sea salt) and precursors gases (DMS) in the future. As an example, the vertically integrated mass of sea salt between 75° and 90°N increases by 90% in early autumn. Both changes in aerosol particles and meteorology lead to an increase in cloud droplet number concentrations (+29% in-cloud/+35% all-sky between 75° and 90°N in early autumn). Furthermore, not only the number concentration, but also the size of the cloud droplets increases, which is due to the higher specific humidity in the future. The cooling effect of both aerosol particles and clouds is larger in the future, especially in late summer when more sunlight is available in the Arctic. However, this is not predominantly caused by changes in the aerosol particles and clouds themselves, but by changes in the surface albedo. The radiative effects of aerosols and clouds strongly depend on the surface albedo; as an example, aerosol particles can have a cooling effect if the surface is dark, but a warming effect if the surface is bright. The reduction in sea ice, which is highly reflective, thus increases the cooling component of aerosols and clouds. Averaged between 75° and 90°N, the radiative forcing of aerosols decreases from 0.53 Wm-2 to 0.36 Wm-2 and the cloud radiative effect from -36 Wm-2 to -46 Wm-2 in late summer.

To detect significant changes in aerosol particles, the ship emissions by Peters et al. (2011) had to be increased by a factor of 10 in terms of mass. Despite this tenfold ship emissions, changes in the aerosol radiative forcing were small and insignificant. The radiative forcing of BC deposited on snow shows locally significant increases in early autumn (see Figure 4.5f) but both absolute values (19*10-2 Wm-2; averaged between 75° and 90°N) and absolute changes (0.64*10-2 Wm-2; not statistically significant over 75° to 90°N) are small. However, we find that the tenfold ship emissions clearly enhance the cooling effect of clouds in late summer (see Figure 4.5d). The increase in aerosol particles leads to more cloud droplets (see Figure 4.5b) which increases the total surface area of the clouds. Furthermore, the collision-coalescence process is less efficient, which increases the liquid water content of the clouds. Averaged between 75° and 90°N, the net cloud radiative effect decreases from -48 Wm-2 to -52 Wm-2. Since the changes in clouds have a much larger impact than the changes in deposited BC, the simulations indicate that future Arctic shipping could lead to a small cooling that offsets part of the expected Arctic warming. However, this cooling is likely overestimated in these simulations since the ship emissions were scaled up by a factor of ten.
Potential Impact:
BACCHUS was very successful in terms of dissemination and outreach. We had impacts in several areas:

Scientific impacts:
BACCHUS strongly enhanced our current understanding of aerosol-cloud interactions in the Earth System and as such reduced the uncertainties of current state-of-the-art ESM climate predictions at different scales and in several ways.

Scientific breakthroughs:
The numerous well-coordinated BACCHUS field activities (field campaigns and long-term observations of particle properties (aerosol size distribution, chemical composition, CCN, INP) will have a strong impact on future efforts in the research field of aerosol-cloud interaction (observations, modeling) and will continue to trigger the improvement of respective aerosol parameterizations in cloud-resolving models.

BACCHUS had a strong impact on in situ INP observations and clearly shows the way how to proceed in future with these field activities, how to design future strategies, how to apply existing and new methods of measurements, where to measure, no longer just focusing on laboratory and ground-based studies, but shifting gears to more airborne observations with drones,

Development, harmonization and dissemination of data sets: One legacy of BACCHUS will be the comprehensive and consistent ice nucleating particle database. It will be open for use for the scientific community. The long-lasting impact of this database should be visible in the years to come.

Integration of BACCHUS data into existing networks: Data obtained during BACCHUS especially in terms of aerosol properties (chemical composition, number concentration and size, CCN) have been and will continue to be integrated into the already existing databases of ACTRIS and GASSP.

Strengthening of European observational sites: BACCHUS strengthened the network of European funded observational sites that are connected to recognized global networks. BACCHUS expanded this station network of sites to incorporating new sites (e.g. Cyprus, the Barbados Cloud observatory and the Amazon site), enriching and standardizing measurements made at these sites and linking these measurements more closely to data synthesis, the development and validation of satellite retrievals (see above), and modelling activities (see results from WP3). Some observational sites have already be equipped with prototype instrumentation for long-term observations of INP, and others will follow.

Remote sensing verification capability: New remote sensing methods (based on advanced aerosol lidar, wind lidar, and cloud radar technology) to obtain vertical profiles of CCNC, INPC, CDNC and ICNC have been developed during the BACCHUS period. This will have a strong impact on future observation-based studies of aerosol-cloud interaction in the most crucial field of mixed-phase cloud evolution (mainly occurring in the middle and upper troposphere). All these methods could also be applied to space lidar activities such as CALIPSO/CloudSAT and later on in the case of EarthCARE mission with lidar and radar aboard ONE satellite. BACCHUS triggered pioneering work in the field of active remote sensing of aerosol-cloud interaction. This will clearly benefit for the ACTRIS community.

Satellite remote sensing methods applicable to satellite observations towards higher horizontal resolution (necessary to obtain reliable estimates of the impact of aerosols on liquid-water cloud evolution, small-scale processes, with scales of updraft widths) have been developed in the framework of BACCHUS for VIIRS by Rosenfeld et al. BACCHUS opened the door to modern passive remote sensing of aerosol-cloud interaction on a global scale, here with focus on boundary layer clouds affected by anthropogenic haze.

Development of improved numerical models: BACCHUS will provide an opportunity to further develop and refine key numerical models that operate over a wide range of spatio-temporal scales (ranging from cloud resolving models to global Earth systems models) that are crucial for addressing the scientific questions relevant to the current call. Models operating at the local and cloud-scale will be evaluated in details in different contrasting environments and will possibly be further refined, and will subsequently serve as a basis for the evaluating and improving Earth systems models.

Improved cloud and precipitation parameterisations in global climate and numerical weather prediction models: The focus of BACCHUS on aerosol-cloud interactions with an emphasis on ice phase processes led to improved parameterizations of ice nucleation in the GLOMAP and ECHAM6-HAM2 GCMs. During BACCHUS, also a simple plume aerosol module MACv2-SP was developed (Stevens et al., 2017) that will now be used in several CMIP6 models. This approach could be of interest also for numerical weather prediction models such as the ECMWF model, that have started to incorporate the radiative effect of some aerosols.

Reduced uncertainties in aerosol radiative forcing:
Work initiated during BACCHUS on the anthropogenic aerosol forcing will be relevant for the upcoming sixth assessment report of Intergovernmental Panel for Climate Change (IPCC). In addition, BACCHUS contributed to a one-week expert meeting in Schloss Ringberg, Germany that took place from February 26th to March 2nd, 2018 to which 36 experts were invited. The aim of this meeting was to discuss unlikely strong and weak anthropogenic aerosol radiative forcings (aerosol forcing for short) by providing lines of evidence for why the aerosol forcing cannot be more negative or more positive than certain bounds. The bounds are meant in terms of likelihood rather than certainty. The Ringberg meeting was organized by Nicolas Bellouin, Sandrine Bony, Olivier Boucher, Jean-Louis Dufresne, Piers Forster, Jim Haywood, Ulrike Lohmann, Gunnar Myhre, Johannes Quaas, Bjorn Stevens, and Philip Stier, i.e. by 3 BACCHUS Principal Investigators. Stephanie Fiedler, David Neubauer, Stefan Kinne and Ken Carslaw from the BACCHUS consortium attended the meeting as participants. The expert assessment for the final bound for the all-sky effective aerosol forcing from the Ringberg meeting concluded it to lie in the range of -0.5 and -1.5 W m-2. To have either a more positive or more negative all-sky effective aerosol forcing would require strong adjustments for which we find no evidence. The uncertainty range in all-sky effective aerosol forcing from the Ringberg meeting is reduced by 50% as compared to the expert solicitation by Granger Morgan et al. (2006) but wider than can be found in some recent literature. The results of this expert assessment are still preliminary and may need to be adjusted. A paper entitled “Bounding aerosol radiative forcing” by Bellouin et al., (2018) is in preparation.

All of these activities lead to enhancing the European scientific excellence and competitiveness. The knowledge gain acquired in BACCHUS and manifested in international assessment reports will help the European Commission to better define policy needs such as the sustainability impact assessments of EU policies and the follow up to the Kyoto and post-Kyoto actions on climate change, as well as to consolidate EU’s strength in the upcoming IPCC reports. Finally, early access to new information provided by BACCHUS will strengthen the position of the European Commission in developing and negotiating international treaties. The work conducted in BACCHUS on Arctic clouds will be incorporated and be beneficial for the Arctic Monitoring and Assessment Programme work on the warming Arctic. The UNECE convention on Long-Range Transboundary Air Pollution (LRTAP) programme and Directorate General Environment could benefit from the results obtained in BACCHUS on natural and anthropogenic aerosols. Improvements of the Earth System Models during BACCHUS furthers our understanding of climate change and benefits the World Climate Research Programme (WCRP) and Directorate General Clima.

Directions for future research:
Despite its achievements, several open questions remain. BACCHUS results highlight the importance of continued basic research in the following areas:
• Development of low-weight (and low-cost) instrumentation for unmanned measurement platforms for in-cloud observations and other in-accessible regions.
• Human activities in the Anthropocene affect natural aerosol emissions. Those changed emissions will have an influence on climate next to anthropogenic emissions. However, their detailed climate impact is not yet known and requires further investigations through improved Earth system models.
• The Arctic and other extreme environments are susceptible to rapid climate change. This makes them potentially vulnerable to socio-economic development warranting further research.

Impact on education:
BACCHUS engaged in training and education at various levels. In particular, BACCHUS had a large training component by having hired 8 new PhD students and 10 post-doctoral fellows who worked exclusively for the BACCHUS project. The students were trained and became experts in aerosol and cloud measurements, in using earth system, global climate and process models, and learned how to use meteorological information.
In addition to experimental and modelling skills, the PhD students and post-doctoral fellows gained valuable expertise and experience towards becoming talented scientists. This included the attendance at conferences, writing and publishing of manuscripts (see dissemination below), the ability to work in the field, working in a team, i.e. skills that will be useful for them in the future. The PhD students and young scientists were encouraged to present their results at the special BACCHUS session at the 2016 European Geophysical Union conference and as part of the 2017 and 2018 larger aerosol-cloud session at the European Geophysical Union conference.
For post-doctoral fellows, the project provided ample opportunities to strengthen their scientific and experimental skills by being responsible for one of the experimental and/or modelling aspects, but also to have a leadership and coordinating role. Thus, post-doctoral fellows were extensively and comprehensively trained to become effective project managers because of their role in organizing field studies, engaging into complex modelling tasks and developing collaborations at various levels in the consortium. They also gained supervisory experience and were trained for a future career in science, but also more widely for engaging themselves into leading positions in the private sector, and/or national and international agencies involved in environmental planning.
The scientific issues that are addressed in environmental sciences in general and in climate sciences in particular are becoming increasingly complex and young scientists need to acquire knowledge and expertise that span several disciplines and to learn how to examine the climate system in a holistic manner. This particularly applies to the topic of BACCHUS dealing with the complex biosphere-aerosol-cloud-climate system. In this context, BACCHUS organized specific winter schools for PhD students and young scientists in Hyytiälä (task 5.2). These schools were led by University of Helsinki, who has a long experience in organizing intensive training courses (2-3 annually) in Hyytiälä with the main focus to teach multidisciplinary approaches to investigate biosphere-atmosphere-climate interactions with various data analysis methods and tools.

Impact on society:
Climate change mitigation and adaptation: A major objective of the global environmental and atmospheric research community is to significantly reduce the uncertainties of current state-of-the-art climate projections. BACCHUS contributed to this objective, by obtaining an improved knowledge of how aerosols and clouds respond to a changing climate, which requires accurate knowledge on how both natural and anthropogenic aerosols affect clouds and climate. By providing a more accurate estimate of background aerosol concentrations representative of pre-industrial times, BACCHUS provided an improved quantification of the aerosol radiative forcing (including aerosol-cloud interactions) since pre-industrial times (see above). Such information is necessary to assess the extent to which the aerosol radiative forcing has offset the greenhouse gas forcing until now and how it will evolve in future. Clear information on the aerosol radiative forcing will help European and international policymakers in formulating cost-benefit policies aiming at both reducing air pollutant emissions (that result in an amelioration of air quality) and mitigating climate change.
In addition, in order to avoid dangerous anthropogenic interference with the climate system, the United Nations Framework Convention on Climate Change has adopted to limit global warming below 2 ºC (2 degree target). Allowable emissions to maintain the 2 ºC target depend heavily on Earth’s climate sensitivity as uncertainties in climate sensitivity translate into uncertainties in the emissions allowed to stay within a 2 ºC warming target. One source of uncertainty is the role of the negative cloud phase feedback (see Lohmann and Neubauer, 2018).
More generally, scientific results from BACCHUS will directly impact the assessment of different mitigation, adaptation and geo-engineering strategies, both through advancements in fundamental understanding of the key processes governing the Earth system and development of new tools and data sets that will be shared with the entire scientific community to ensure that BACCHUS results are used in multi-model type assessments.

Air pollution and ecosystem services: Degradation of air quality (and in particular increasing aerosol and aerosol precursor emissions) severely impacts human health, as well as ecosystem services. BACCHUS contributed to a better understanding of the processes driving the aerosol (=particular matter) distribution by quantifying natural vs. anthropogenic emissions. A better understanding of the biosphere-aerosol-cloud-climate interactions as gained within BACCHUS, allows to better inform the policy regulations and thus to ensure a better protection of humans and ecosystems’ health. In particular, by having used process models to develop parameterisations for aerosol-cloud interactions in GCMs and ESMs, BACCHUS contributed to improving these models and their reliability in terms of future aerosol and thus air quality projections. In addition, future climate projections generated by the improved ESMs can then be used to investigate the cost-benefit of political decisions aimed at improving air quality while mitigating climate change, and can ultimately guide socio-economic decisions. The results obtained during BACCHUS will be beneficial to:
• Develop, collect and validate scientific information relating to the effects of outdoor air pollution of aerosol particles, emission inventories, air quality assessment, emission and air quality projections, cost-effectiveness studies and integrated assessment modelling, leading to the development and updating of air quality and deposition objectives and indicators and identification of the measures required to reduce emissions;
• Support the implementation and review the effectiveness of existing legislation, in particular the air quality daughter directives, the decision on exchange of information, and national emission ceilings as set out in recent legislation, to contribute to the review of international protocols, and to develop new proposals as and when necessary;
• Ensure that the sectoral measures that will be needed to achieve air quality and deposition objectives cost-effectively are taken at the relevant level through the development of effective structural links with sectoral policies;
• Determine an overall, integrated strategy at regular intervals which defines appropriate air quality objectives for the future and cost-effective measures for meeting those objectives;

Innovation-related activities: While measurement devices for cloud condensation nuclei concentrations are commercially available and relatively widespread in the atmospheric research community, only recently, also devices for ice nucleating particle concentrations are becoming available, such as SPectrimeter Ice Nuclei (SPIN, Garimella et al., 2016).
During BACCHUS ice nucleating particle counters (INPC) were deployed at several locations as shown in Figure 1.3 with first INP measurements being performed in Hyytiälä, Finland, Agia Marina and Nicosia, Cyprus, in an Atlantic north-south transect and around Antarctica. The results from this worldwide INP database is in preparation for publication.
The INPC technology used in BACCHUS is state-of-the-art for in-situ INP measurement systems and will represent the basis of future decisions on INP instruments. These activities not only attest to the innovation aspect of BACCHUS but also offer the possibility for a further commercialization of the INPC technology in a way of a spin-off company.

Dissemination:
We had an inter-journal special issue ACP-AMT-GMD (https://www.atmos-chem-phys.net/special_issue911.html) with 50 publications in ACP, and 4 each in AMT and GMD. In total, the BACCHUS project lead to 133 peer-reviewed publications during its 4.5 years (9 in 2014, 15 in 2015, 43 in 2016, 39 in 2017 and 21 in 2018). We had 22 publications in high profile journals (7 in Science and Nature, 7 in PNAS, 7 in Scientific Reports, 1 in Scientific Data). The full list of BACCHUS publications can be found here: https://www.bacchus-env.eu/data/pubanddel.html

Policy makers and other relevant stakeholders (environmental planners, practitioners, etc.): Some of the BACCHUS partners are continuously involved in the preparation of assessment reports (such as IPCC reports to which Ulrike Lohmann, Bjorn Stevens, Maria Kanakidou, Maria Cristina Facchini and Sandro Fuzzi have contributed), which enable a more direct dissemination of scientific findings to policy makers and negotiators. Two policy-relevant reports (mid-term and final report for policy makers) were prepared to the attention of the European Officer and other EU policy makers with the objective of increasing their awareness and improving their understanding of climate-related issue, and ultimately guide their decisions (task 5.5).

Media and general public: BACCHUS improved the general public awareness about the fundamentals of climate and climate change through its webpage and other outreach activities. A summary of which can be found at the BACCHUS webpage: https://www.bacchus-env.eu/ and is also provided in this final report.

References:
Garimella, S., Kristensen, T. B., Ignatius, K., Welti, A., Voigtländer, J., Kulkarni, G. R., Sagan, F., Kok, G. L., Dorsey, J., Nichman, L., Rothenberg, D. A., Rösch, M., Kirchgässner, A. C. R., Ladkin, R., Wex, H., Wilson, T. W., Ladino, L. A., Abbatt, J. P. D., Stetzer, O., Lohmann, U., Stratmann, F., and Cziczo, D. J., The SPectrometer for Ice Nuclei (SPIN): an instrument to investigate ice nucleation, Atmos. Meas. Tech., 9, 2781-2795, doi.org/10.5194/amt-9-2781-2016 2016.
List of Websites:
Project webpage: www.bacchus-env.eu

BACCHUS Coordinator: Prof. Ulrike Lohmann
BACCHUS Project Manager: Dr. Monika Burkert

Contact:
ETH Zurich
Institute for Atmospheric and Climate Science
Universitätstr. 16
8092 Zurich
Switzerland

Final Report Summary - BACCHUS (Impact of Biogenic versus Anthropogenic emissions on Clouds and Climate: towards a Holistic UnderStanding)

Share this page

Download