Atmospheric and Surface Processes, and Feedback Mechanisms Determining Arctic Amplification A Review of First Results and Prospects of the (AC) 3 Project

: Mechanisms behind the phenomenon of Arctic amplification are widely discussed. To contribute to this debate, the (AC) 3 project was established in 2016 (www.ac3-tr.de/). It comprises modeling and data analysis efforts as well as observational elements. The project has assembled a wealth of ground-based, airborne, shipborne, and satellite data of physical, chemical, and meteorological properties of the Arctic atmosphere, cryosphere, and upper ocean that are available for the Arctic climate research community. Short-term changes and indications


Annual
, the interactions of sea ice and ocean processes with the atmosphere (Rinke et al. 2019a) and clouds (Huang et al. 2019), and the understanding of processes determining airmass transformations during meridional transports of heat, moisture, and momentum by atmospheric circulation and ocean currents (Pithan et al. 2018;Nash et al. 2018;Wendisch et al. 2021).Additionally, the role of trace gases and aerosol particles in Arctic amplification is still uncertain (Schmale et al. 2021).
To resolve these issues, both long-term and campaign-based observations accompanied by detailed data analysis and dedicated numerical weather prediction and climate model simulations are required.These needs have motivated the establishment of the Transregional Collaborative Research Center (AC) 3 (Wendisch et al. 2017) (www.ac3-tr.de/).The project aims to enhance the understanding of key local and remote atmospheric and surface processes and feedbacks driving Arctic amplification.This general objective is being achieved by a synergistic combination of observations using ground-based, airborne, shipborne, and satellite-borne sensors, with comprehensive data analysis and modeling over a broad range of spatial and temporal scales.For this purpose, (AC) 3 utilizes extensively the excellent Arctic research infrastructure made available by Alfred-Wegener-Institut, Helmholtz-Zentrum für Polar-und Meeresforschung (AWI). 1 Furthermore, the data collected during the Multidisciplinary drifting Observatory for the Study of Arctic Climate (MOSAiC) expedition (Shupe et al. 2022;Nicolaus et al. 2022;Rabe et al. 2022), to which (AC) 3 provided a major German contribution, will help to reach the goals of (AC) 3 , in particular with regard to the annual cycle of local processes in the inner Arctic.Spectacular new data gathered during the recently completed HALO-(AC) 3 airborne campaign (Wendisch et al. 2021) will contribute to clarify remote feedbacks of airmass transformations during warm-air intrusions (Pithan et al. 2018) and cold-air outbreaks (Geerts et al. 2022).
The specific processes and feedback mechanisms that comprise the focus of (AC) 3 are summarized in the second section.Major results as well as prospects of the project are introduced by discussing three questions in the subsequent sections: What have we done so far (third section)?What did we learn (fourth section)?Where do we go from here (fifth section)?In the concluding sixth section selected important results achieved within (AC) 3 so far are summarized.
The (AC) 3 framework of processes determining Arctic amplification Arctic amplification involves a number of intertwined chains of effects, some of which are shown in Fig. 1, which provides the major framework for the investigations carried out within (AC) 3 .The figure illustrates important and interlinked local and remote atmospheric and surface processes and feedback mechanisms that contribute to Arctic amplification.They include largely atmospheric and marine effects related to the Arctic atmosphere, upper ocean, and sea ice.The role of the land surfaces in Arctic amplification is not a focus of (AC) 3 .In the following four subsections, the processes and feedbacks in Fig. 1 are introduced.
Surface albedo feedback: Positive (black in Fig. 1).Triggered by global warming, 2 the near-surface air temperature in the Arctic increases.As a result, sea ice and snow melt and the surface albedo of the Arctic Ocean decreases.Therefore, more solar radiation is absorbed by the increasingly darker open (sea ice-free) ocean surface, leading to a radiative heating of the ocean mixed layer.This enhances the radiative (solar and terrestrial) and turbulent (heat, moisture) atmospheric energy fluxes.
1 All acronyms are listed in the appendix.
2 Terms from Fig. 1 are highlighted in the text of "The (AC) 3 framework of processes determining Arctic amplification" section in italic style.

E213
For example, the upward terrestrial radiative energy flux densities (irradiances) emitted by the warmer ocean surface are absorbed by near-surface atmospheric greenhouse gases (e.g., water vapor) and/or clouds and then reemitted toward the surface.This in turn further increases the near-surface air temperature establishing a positive feedback loop.This amplifying feedback mechanism is well known as the direct surface albedo feedback (Arrhenius 1896;Budyko 1969;Sellers 1969;Manabe and Wetherald 1975).It is most effective in summer when sufficient solar radiation reaches the surface of the Arctic Ocean.Apart from that, the surface albedo feedback works also indirectly and delayed.The solar radiative energy absorbed in the open-ocean mixed layer during spring and summer is converted into heat that is stored in the upper ocean.The saved heat energy determines the onset of refreezing in fall.Prior to the refreezing in late fall and early winter, the stored heat is released into the atmosphere via radiative and turbulent processes, warming the near-surface air, and thus causing a delayed, indirect warming effect.After the onset of refreezing, the increasing sea ice cover reduces the energy exchange between the ocean mixed layer and near-surface air due to thermal insulation.
There is general consensus that the direct surface albedo feedback represents one of the important causes for Arctic amplification in spring (Hall 2004;Screen and Simmonds 2010;Taylor et al. 2013).The indirect surface albedo feedback contributes to Arctic amplification in fall and winter (Pithan and Mauritsen 2014;Goosse et al. 2018).
Upper-ocean effects-Positive or negative (brown in Fig. 1).The expanding openocean areas, resulting from the enhanced melting of sea ice and snow in a warmer Arctic, cause intensified cryosphere-ocean-atmosphere interactions and, as a consequence, increased emissions of trace gases and aerosol particles of marine/biogenic origin.

E214
Furthermore, ocean biogeochemistry processes in the upperocean mixed layer are amplified (Ardyna and Arrigo 2020), in particular the primary productivity. 3This may cause an enhanced radiative heating of the ocean mixed layer by increased absorption of solar radiation.This heating in turn raises the temperature of the ocean mixed layer, further amplifying cryosphere-ocean-atmosphere interactions and ocean biogeochemistry processes.Additionally, the upper ocean will become less stable in terms of the halocline (Polyakov et al. 2017).Altogether, the role of the oceanic heat transport in warming the lower atmosphere is promoted (Tsubouchi et al. 2021).
Local atmospheric processes: Positive or negative (green in Fig. 1).The enhanced emissions of trace gases discussed above influence the magnitude of the Arctic amplification (greenhouse effect) and the oxidation capacity of the atmosphere.Arctic aerosol particles, including cloud condensation nuclei (CCN) and ice nucleating particles (INPs), are released over the ice-free ocean by wind drag (wind-wave driven), ocean biogeochemistry processes, new particle formation from the release of iodine, or biological activities.
Due to the radiative heating of the ocean mixed layer, the ocean surface warms up, causing an increase of water vapor concentration due to intensified evaporation.This enhances the downward terrestrial radiative atmospheric energy fluxes, thus further increasing the nearsurface air temperature.Furthermore, the higher water vapor amounts advance the formation of low-level clouds, which also increase the emission of downward terrestrial irradiances.During polar night this leads to a general warming of the near-surface air (positive feedback), while in summer the solar cloud cooling effect may outweigh the terrestrial warming (negative feedback).However, due to the bright surfaces and low solar zenith angles, the net effect of clouds in the central Arctic is mostly a surface warming (Shupe and Intrieri 2004).Thus, in a warming Arctic with possibly more clouds, an initial warming is mostly amplified over sea ice (positive feedback).However, as sea ice retreats and darker open-ocean surfaces are exposed, the cooling effect by clouds could play a bigger role in summer.On the other hand, no cloud response to sea ice loss has been detected during summer between 2006 and 2008, a period with low summer sea ice concentrations (Kay and Gettelman 2009).Instead, such a cloud response was detected in fall.Therefore, Kay and Gettelman (2009) conclude that cloud changes resulting from sea ice loss play a minor role during summer, but may contribute to a cloud-ice feedback during early fall.
The future evolution of cloud properties will have an impact on precipitation.While total precipitation in the Arctic is predicted to increase in the future, snowfall in winter is projected to stay roughly constant and to decrease in summer and fall (McCrystall et al. 2021).
The atmospheric lapse rate feedback comprises an important atmospheric process with respect to Arctic amplification (Block et al. 2020;Lauer et al. 2020;Boeke et al. 2021;Linke and Quaas 2022).In the Arctic, vertical atmospheric turbulent transport is commonly inhibited by a very stable and shallow atmospheric boundary layer (ABL).Consequently, the warming of the near-surface air due to radiative heating of the ocean mixed layer is mostly kept to the ABL.The resulting enhanced downward terrestrial radiative atmospheric energy fluxes further increase the near-surface air temperature.However, the increased surface temperature also initiates convection weakening the warming effect of the lapse rate feedback.At the top of the atmosphere, the diminished increase in outgoing terrestrial radiative energy, compared to a vertically homogeneous temperature change, is relevant and implies the overall relative warming effect.
Arctic-midlatitude linkages: Positive or negative (yellow in Fig. 1).Atmospheric circulation and meridional transports are suspected to change due to the decreasing E215 meridional geopotential gradient caused by the enhanced warming of the Arctic compared to midlatitudes (Francis and Vavrus 2015).This effect likely causes a slower west wind drift and larger Rossby wave amplitudes moving with a slower phase speed, although other teleconnections or even internal variability may also be a reason for the increased waviness (Blackport and Screen 2020).The enhanced amplitudes and stationarity of the Rossby waves would promote the meridional transport of trace gases, aerosol particles, heat, water vapor, clouds, and precipitation into and out of the Arctic by means of more frequent warm-/moist-air intrusions and cold-/dry-air outbreaks.During the meridional transport, remote feedback mechanisms involving the vertical thermodynamic structure (atmospheric lapse rate) of the transported air masses, as well as combined effects including water vapor, clouds, and precipitation as well as aerosol particles link Arctic and midlatitude processes in a complex manner (Wendisch et al. 2021).There seems to be growing evidence that such links actually exist, for example, with regard to sea ice loss (Screen 2021;Crawford et al. 2022).However, the current generation of models struggles to represent the modification of regional and large-scale atmospheric circulation and of airmass properties along meridional transports likely caused by enhanced warming of the near-surface air temperature in the Arctic (Cohen et al. 2014;Francis and Vavrus 2015;Pithan et al. 2018;Armour et al. 2019;Cohen et al. 2020).

What have we done so far?
The focus in the "What have we done so far?" and "What did we learn?" sections is to document and summarize some of the specific contributions from (AC) 3 to the international research on Arctic amplification.Whereas the current section introduces observational activities, retrieval developments, corresponding data analysis, and modeling applications, the "What did we learn?" section will elaborate the scientific results of (AC) 3 .Of course, the work conducted in (AC) 3 builds upon a rich history of Arctic research.Therefore, we have tried to concisely put the findings of (AC) 3 into context with the existing literature.For a more detailed discussion of the progress achieved in (AC) 3 , the reader is encouraged to consult the specific (AC) 3 publications, compiled at https://publons.com/researcher/3796220/ac3-arctic-amplification/.
Ground-based, airborne, and shipborne measurements.An important pillar of (AC) 3 is the collection of comprehensive datasets as a basis for the study of Arctic amplification.The Fram Strait area northwest of Svalbard and northern Greenland, as one of the most sensitive regions for the Arctic climate system, became a hotspot of ground-based, airborne, and shipborne activities during (AC) 3 .Successful measurements in this area have provided an immense treasure of data, covering a broad range of spatial and temporal scales, different seasons, as well as meteorological and sea ice conditions.
Ground-based and vertically resolved data were obtained during continuous observations at the permanent joint German-French Alfred Wegener Institute for Polar and Marine Research and the French Polar Institute Paul Emile Victor (AWIPEV) research base in Ny-Ålesund (Spitsbergen, Fig. 2).Measurements at Ny-Ålesund have shown that clouds occurred around 80% of the time, mostly below 2 km altitude (Fig. 2d) (Nomokonova et al. 2019).Most of the time clouds contained ice; the relative occurrence of liquid water and ice clouds was mainly influenced by the prevailing synoptic conditions, e.g., wind direction, and thermodynamic coupling with the surface (Gierens et al. 2020).Multilayer clouds (45%) and single-layer mixed-phase clouds (20%) were dominant.During (AC) 3 the instrumentation of the AWIPEV research base was extended and new retrieval algorithms were developed to enable sophisticated cloud and precipitation measurements (Nomokonova et al. 2019).Furthermore, instruments to detect the seasonal variability and secular trends of several E216 long-and short-lived trace gases through the global measurement networks Network for the Detection of Atmospheric Composition Change (NDACC) and Total Carbon Column Observing Network (TCCON) with enhanced observation frequency were operated at AWIPEV (Buschmann et al. 2017).In addition, the tethered Balloon-Borne Modular Utility for Profiling the lower Atmosphere (BELUGA) (Egerer et al. 2019) and the unmanned aerial system Application of Light-weight Aircraft for Detecting in situ Aerosols (ALADINA) (Lampert et al. 2020) were launched during dedicated intensive operational phases covering small scales up to 1.5 km altitude with high-resolution meteorological, turbulence, aerosol, and radiation measurements.
Five airborne campaigns were performed using the German research aircraft Polar 5 and Polar 6 of AWI (Wesche et al. 2016) (Wendisch et al. 2021).The aircraft observations provided a rich dataset covering the northwestern segment of Svalbard and the area north of Villum Research Station (Fig. 3a) over an altitude range of up to 4 km (Fig. 3b).The two aircraft were extensively equipped with partly newly developed instruments E217 (Ehrlich et al. 2019;Mech et al. 2022), such as the new airborne Microwave Radar/radiometer for Arctic Clouds (MiRAC) (Mech et al. 2019).
The shipborne research cruise Physical feedback of Arctic ABL, Sea ice, Cloud and Aerosol (PASCAL) using the Research Vessel (R/V) Polarstern of AWI was conducted in May/June 2017 (Wendisch et al. 2019).PSACAL was closely coordinated with ACLOUD, most of the flight hours of the two polar aircraft were spent sampling the vertical column above R/V Polarstern, which comprises a unique observational approach.Furthermore, (AC) 3 has significantly contributed to the MOSAiC expedition with R/V Polarstern performed between September 2019 and October 2020 (Shupe et al. 2022;Nicolaus et al. 2022;Rabe et al. 2022).The tethered balloon system BELUGA was deployed during both PASCAL (Egerer et al. 2021) and MOSAiC (Lonardi et al. 2022).Furthermore, the helicopter-borne instrument sonde HELiPOD was flown during MOSAiC (Shupe et al. 2022).Also, the OCEANET-Atmosphere facility, extended by the motion-stabilized cloud radar Mira-35 (Griesche et al. 2020) and the passive part of MiRAC, was operated continuously during PASCAL and MOSAiC (Engelmann et al. 2021;Walbröl et al. 2022).The ship-based observations conducted during PASCAL revealed a high fraction of low-level stratus clouds in summer (Griesche et al. 2020).These clouds below an altitude of 150 m were observed 25% of the time.
Satellite data analysis.Important work on developing new satellite retrieval algorithms and combining data from different sensors to retrieve aerosol particle, cloud, and surface properties was performed within (AC) 3 .A summary of the satellite data employed during (AC) 3 is given in the online supplementary materials (Table 1; https://doi.org/10.1175/BAMS-D-21-0218.2).
Here we can just introduce some of the (AC) 3 activities in this area.Mei et al. (2018) developed a novel algorithm for the derivation of cloud optical thickness (COT) and cloud effective radius (CER) for ice clouds from satellite data with promising results over snow and sea ice.Jafariserajehlou et al. (2019) developed a new cloud masking technique specialized for the Arctic, which can be used for long-term aerosol retrievals over sea ice, and whose quality depends heavily on the quality of the cloud mask.Mei et al. (2020a) proposed a unique retrieval technique to retrieve the coarse mode fraction of aerosol particles above snow and ice.Furthermore, Mei et al. (2021b) have developed and globally validated a new retrieval

E218
of snow properties based on a thorough sensitivity study (Mei et al. 2021a).Innovative retrieval techniques to derive the integrated water vapor (IWV) have been introduced merging observations from different microwave satellite sensors (Triana-Gómez et al. 2018, 2020).
The first consistent and consolidated long-term (1996-2017) dataset of bromine monoxide (BrO) over the Arctic region has been derived from four different ultraviolet-visible satellite instruments (Bougoudis et al. 2020).In addition, a new high-resolution retrieval algorithm for column densities of BrO has been developed (Seo et al. 2019).Long-term (2002-present) time series of phytoplankton groups were derived (Losa et al. 2017;Xi et al. 2021), which for the first time also deliver phytoplankton group datasets for the Arctic Ocean.Several snow and sea ice parameters have been retrieved using microwave observations (Scarlat et al. 2017(Scarlat et al. , 2020)).The accuracy of sea ice concentration measurements has been improved (Lu et al. 2018(Lu et al. , 2022)), as well as their spatial resolution, by combining infrared and microwave data (Ludwig et al. 2020).Also, the retrieval accuracy of the thickness of thin sea ice was increased (Paţilea et al. 2019).Work conducted within (AC) 3 delivered the first dataset

E219
of snow depth on sea ice for the complete Arctic including multiyear sea ice (MYI) regions in spring and first-year sea ice (FYI) from October to May (Rostosky et al. 2018(Rostosky et al. , 2020)).By including lower microwave radiometer frequencies at 7 GHz, a more reliable retrieval and an extension to MYI could be achieved, which was not possible with previous methods (Markus et al. 2006).The uncertainty of retrievals of snow depth on sea ice was calculated based on a Monte Carlo simulations (Rostosky et al. 2020), it increases with increasing snow depth and is higher over MYI than over FYI.Roughly, the relative uncertainty is about 20%-30% of the snow depth.Figure 4 shows an example of the sea ice datasets retrieved within (AC) 3 .
Atmospheric-ocean and radiative transfer models.Newly developed and further refined numerical models covering a broad range of temporal and spatial scales were applied within (AC) 3 , a summary can be found in Table 2 of the supplementary materials.A Lagrangian large-eddy simulations (LES) setup was developed for the Arctic to follow trajectories of an air mass at high latitudes (Neggers et al. 2019).A specific LES version of the atmospheric Icosahedral Nonhydrostatic (ICON) model with realistic forcing and considering a heterogeneous surface was applied representing the complex environment of Ny-Ålesund (Schemann and Ebell 2020).By combining the model output with the instrument simulator Passive and Active Microwave radiative Transfer tool for simulating radiometer and radar measurements of the cloudy atmosphere (PAMTRA) (Mech et al. 2020), the detection of mixed-phase clouds by specific instruments was modeled with unprecedented resolution down to 75 m (Fig. 5).ICON was operated in a nested mode for an Arctic domain with a horizontal resolution of down to 2 km, to improve the representation of Arctic clouds (Kretzschmar et al. 2020) and intense moisture intrusion events (Bresson et al. 2022).For larger scales ( 9 the coupled ocean-atmosphere High-Resolution Limited Area Model (HIRHAM)-North Atlantic/Arctic Ocean-Sea Ice Model (NAOSIM) regional model system was established (Dorn et al. 2019;Rinke et al. 2019a).Output from the sixth-generation atmospheric general circulation model ECHAM6 and Coupled Model Intercomparison Project phase 5 (CMIP5) global multimodel ensemble simulations has been analyzed to improve cloud representation (Kretzschmar et al. 2019) and radiative feedback understanding (Block et al. 2020).The fast ozone module fast ozone chemistry scheme for interactive calculation of the extrapolar stratospheric ozone layer in coupled general circulation models (SWIFT) (Wohltmann et al. 2017), which efficiently calculates stratospheric ozone chemistry and enables mutual ozone-climate interactions in global circulation models (GCMs), has been implemented into ECHAM6 (Romanowsky et al. 2019).
Several radiative transfer models of different complexity have been improved within (AC) 3 to investigate Arctic-specific radiative effects.For example, the coupled oceanatmosphere radiative transfer model Radiative Transfer and Retrieval Algorithm (SCIATRAN) (Rozanov et al. 2017), and the coupled sea ice-ocean-biogeochemical model Massachusetts Institute of Technology General Circulation Model (MITgcm; http://mitgcm.org/) were extended to assess the feedback of surface ocean biogeochemistry on Arctic amplification (Soppa et al. 2019;Pefanis et al. 2020).New aerosol types specifically considering Arctic conditions were incorporated into SCIATRAN (Mei et al. 2020b).In addition, several ice crystal databases (Baum et al. 2011;Yang et al. 2013) have been included, allowing optimized radiative transfer simulations for Arctic ice and mixed-phase clouds.Furthermore, new modules for the bidirectional reflectance distribution function (BRDF) of snow, white ice, and melt ponds have been implemented into SCIATRAN (Mei et al. 2022).The fast and accurate radiative transfer model Fast and Accurate Semi-analytical Model of Atmosphere-surface Reflectance (FASMAR) was newly developed to specifically consider for large solar zenith angles (larger than 80°), common in the Arctic, while other existing models are typically either limited to solar zenith angle smaller than 70°, or are too slow to process long-term datasets (Mei et al. 2020a).Furthermore, the one-dimensional libRadtran and Rapid Radiative Transfer Model for GCMs (RRTMG) software packages for radiative transfer were used to estimate the radiative forcing of Arctic clouds (Ebell et al. 2020;Barrientos-Velasco et al. 2022;Stapf et al. 2020Stapf et al. , 2021;;Stapf 2021).In addition, the new and computationally highly efficient, three-dimensional, backward Monte Carlo radiative transfer code Light Estimator Including Polarization, Surface Inhomogeneities, and Clouds (LEIPSIC) was developed and applied to quantify the impact of multiple scattering between clouds and a heterogeneous sea ice-ocean surface on satellite observations and the radiative energy budget in Arctic-specific conditions (Sun et al. 2020).
What did we learn?Surface albedo feedback (black in Fig. 1).Here we provide examples of selected results with regard to the main components of the surface albedo feedback: sea ice and snow, surface albedo, radiative heating of the ocean mixed layer, and atmospheric energy fluxes.
Sea ice and Snow: Temporal TrendS and condiTionS during moSaic.We begin with some results of (AC) 3 with regard to sea ice and snow, which represent crucial components for the surface albedo effect.Temporal changes of these variables have been observed using partly newly developed methods.For example, combined satellite and ocean mooring data have shown that the mean sea ice thickness in the Fram Strait decreased by 15% decade −1 during 1990-2014 (Spreen et al. 2020).Primarily due to this thinning, the Arctic sea ice volume export decreased by 27% decade −1 between 1992 and 2014.Previous estimates from models and observations (Spreen et al. 2009;Zhang et al. 2017) did not find a significant decrease E221 of the ice volume export.These former estimates ended in earlier years, when the ice thinning was not as dominant.
With regard to snow depth on sea ice, it was found that in March 2015, the average snow depth on MYI was 31 cm, which is about twice the snow depth determined for MYI of 16 cm (Fig. 6a).Furthermore, a significant decrease of snow depth on sea ice of 2 cm decade −1 for FYI, and 3 cm decade −1 for MYI was identified in the March time series of 2003-20 (Fig. 6b).The snow depth trends vary regionally with strongest values in the Atlantic sector and the Kara and Laptev Seas, while other regions did not show significant trends.The interannual variability of the snow depth on MYI appeared much higher than on FYI, which showed a pronounced decline between 2009 and 2014 (Figs.6c,d).
Also, it was found that during the MOSAiC expedition, sea ice was thinner than in previous years and the drift of the MOSAiC ice flow was about 25% faster than expected from climatology (Krumpen et al. 2021).However, the sea ice concentration, snow, and lead fraction were close to the climatological average.Surface albedo: parameTerizaTion for modelS.To improve the model representation of the surface albedo feedback, HIRHAM-NAOSIM has been upgraded with an adjusted sea ice albedo parameterization derived from field measurements during (AC) 3 (Jäkel et al. 2019).The former version has shown significant differences of the parameterized sea ice albedo compared to observations.Therefore, the sea ice albedo parameterization was adjusted with

E222
respect to changes in temperature-dependent snow properties and threshold temperatures describing the transition between dry and melting snow/ice.Further, the dependence of surface albedo on cloud occurrence was implemented, since the broadband albedo of snow and sea ice increased significantly compared to cloud-free conditions (Stapf et al. 2021).As a result, the root-mean-squared deviation between parameterized and measured surface albedo could be reduced from 0.14 to 0.04 for cloud-free and broken-cloud situations.The revised surface albedo parameterization has led to a more realistic magnitude of the overall sea ice loss from May to August as simulated by HIRHAM-NAOSIM.Also, for ICON, the consideration of surface albedo measurements in nested high-resolution model runs improved the representation of the Arctic surface radiative energy budget significantly (Wendisch et al. 2019).
radiaTive heaTing of The ocean mixed layer: a Self-reinforcing poSiTive feedback loop.Biological particles embedded in the upper-ocean water, such as phytoplankton groups (PG), colored dissolved organic matter (CDOM), and total suspended matter (TSM), absorb solar radiation, which heats the upper part of the ocean mixed layer.If biological activity increases with rising ocean temperatures, the heating may reinforce itself in a positive feedback loop, thus contributing to reducing the sea ice cover even further.To investigate the importance of this effect in the Arctic Ocean, satellite ocean color observations were analyzed within (AC) 3 .Respective new satellite data products were developed for PG (Losa et al. 2017), CDOM, and TSM (Soppa et al. 2019).The resulting time series revealed changes of PG in the Fram Strait significantly affecting surface albedo (Losa et al. 2017).Correspondingly, SCIATRAN and MITgcm were applied to quantify the influence of CDOM and TSM on the radiative heating of the Laptev Sea shelf waters (Soppa et al. 2019) and the Arctic Ocean (Pefanis et al. 2020).It was shown that in Arctic summer, due to high levels of CDOM, 43% more solar radiative energy is absorbed in the near-surface ocean water compared to situations with low CDOM, leading to radiative heating of the upper-ocean water of 0.6 K day −1 .Thus, the expected future increase of CDOM and TSM discharge into the Laptev Sea, due to permafrost thawing, will likely accelerate the melting of sea ice and lead to enhanced ocean-atmosphere heat fluxes.To upscale this assessment, numerical experiments were conducted with and without incorporating the effect of PG and CDOM on solar absorption (Pefanis et al. 2020).These simulations indicated higher surface temperatures and more sea ice melt in summer.As a consequence, the sea ice season over parts of the Siberian shelf shortens by up to 1 month.aTmoSpheric energy fluxeS: Turbulence parameTerizaTion and impacT of leadS.Common parameterizations tend to overestimate the turbulent energy fluxes for the often stably stratified Arctic surface layers.This issue has been overcome within (AC) 3 by Gryanik et al. (2020), who derived new Monin-Obukhov similarity theory (MOST) stability correction functions (SCFs) for momentum Furthermore, Gryanik and Lüpkes (2018) developed a new noniterative scheme to determine transfer coefficients for momentum C d (Ri b , E m , E t ) and heat C h (Ri b , E m , E t ) (where E m = z/z 0 and E t = z/z t are roughness parameters), which can be used with the above SCFs.It turned out that relative to SHEBA data, the new transfer coefficients are superior to those of previous schemes over sea ice (Fig. 7).They can be part of a package of new noniterative parameterizations based on measurements in different regions on Earth as well (Gryanik et al. 2021;Gryanik and Lüpkes 2022).The new improved stability functions have been implemented in HIRHAM5.Roughness lengths were adjusted to mean values derived from SHEBA (z 0 = 3.3 × 10 −4 m for momentum and z t = 6.6 × 10 −5 m for heat).In simulations of wintertime conditions, the significant impact of these changes was shown through modifications to the regional circulation, wind, and near-surface turbulent fluxes (Schneider et al. 2022).
In a warmer Arctic with thinner sea ice, the number of leads is likely to increase.It was confirmed within (AC) 3 that even though leads occupy only a small areal fraction, they exert a large impact on the regional temperature, stability over sea ice, and surface fluxes (Chechin et al. 2019).Therefore, a new parameterization of convective plumes and related processes over leads of different widths has been established within (AC) 3 (Michaelis et al. 2020(Michaelis et al. , 2021)).It is applicable in plume-resolving, computationally inexpensive models with much coarser resolution than LES, which enables extensive scenario studies.It was shown that the impact of leads on the surface fluxes over a typical domain of a grid cell of a climate model depends critically on geometric lead parameters (e.g., lead width).Such subgrid-scale characteristics of leads are not considered in an individual model grid cell (Michaelis and Lüpkes 2022).Furthermore, a new conceptual model of the ABL coupled to sea ice in the presence of leads was proposed by Chechin et al. (2019) describing analytically the ABL warming due to leads as a function of wind speed.In particular, the new model highlights and explains the role of leads in the formation of decoupling between the sea ice surface and the ABL temperatures, being a step forward compared to earlier studies (Lüpkes et al. 2008).Another new result of Chechin et al. (2019) was the finding that the threshold value subdividing different stability regimes in the environment of leads is a function of the lead fraction.Furthermore, it was shown by Chechin et al. (2019) that those regimes agree well with data from Russian drifting stations and that there is a clear connection between the net terrestrial surface radiative energy fluxes and wind speed.
Upper-ocean effects (brown in Fig. 1).Closely linked to the surface albedo effect are consequences for the upper-ocean biological activity that partly feedback to components of the system.Some examples of these interactions studied within (AC) 3 are given in the "Cryosphere-ocean-atmosphere interactions: Feedbacks and coupled processes" and

E224
"Cryosphere and ocean biogeochemistry: Impact on trace gases and aerosol particles" sections.
cryoSphere-ocean-aTmoSphere inTeracTionS: feedbackS and coupled proceSSeS.A positive feedback mechanism involving wind stress, sea surface temperature, and sea ice in the Nordic seas was investigated during (AC) 3 using a fully coupled Earth system model (Kovács et al. 2020).It was shown that an anticyclonic wind anomaly causes a strong surface cooling in the Greenland Sea, which is mostly due to the drift of sea ice.The cooling reduces the net surface heat flux to the atmosphere and increases sea level pressure.The pressure gradients cause southerly winds that are comparable to the prescribed forcing anomalies, suggesting a positive feedback.In another study within (AC) 3 , Metzner et al. (2020) investigated the role of changes in the Arctic Ocean cold halocline for ocean heat transfer from the ocean to the atmosphere.The cold halocline is a stable layer that separates warm Atlantic water from the overlying cold mixed layer.Previously, observational studies had suggested that the cold halocline is retreating (Steele and Boyd 1998;Polyakov et al. 2017).Based on CMIP5 model results, we found within (AC) 3 that in future climate projections, events, in which warm Atlantic water is no longer separated from the ocean mixed layer, become more frequent, facilitating an increased surface heat flux from the Arctic Ocean to the atmosphere during winter.Furthermore, in the Fram Strait, part of the circulation of Atlantic water and its subduction was found to be dominated by eddies and, therefore, prone to changes caused by atmospheric forcing (Hofmann et al. 2021).
cryoSphere and ocean biogeochemiSTry: impacT on Trace gaSeS and aeroSol parTicleS.Temporal trace gas changes are associated with sea ice trends.As an example, a positive tropospheric BrO trend of about 1.5% yr −1 during polar spring was identified within (AC) 3 , which appeared to be correlated with an increase in FYI, at the expense of MYI (Bougoudis et al. 2020).Spatial trend patterns of BrO appeared to vary, indicating that local factors such as the amount of blowing snow and meteorological parameters play an important role.
Clear indications for local marine sources of biogenic INPs were found during airborne (Hartmann et al. 2020) and ship-based (Hartmann et al. 2021) filter sampling within (AC) 3 .Efficient biogenic INPs were also found in fog droplets (Hartmann et al. 2021).Further data have shown that Arctic INPs feature a seasonal cycle with highest concentrations in summer and lowest in winter (Wex et al. 2019).In the laboratory, the INPs nucleated ice at temperatures as high as −7.5°C (Hartmann et al. 2019(Hartmann et al. , 2020) ) or even up to −5°C (Wex et al. 2019).Prior to these studies done within (AC) 3 , it had generally been assumed that Arctic INP concentrations are low (Loewe et al. 2017).In general, Arctic INPs are grossly unexplored and, therefore, still provide a large source of uncertainty for understanding Arctic clouds (Morrison et al. 2012).The new results achieved within (AC) 3 show that ice nucleation at high freezing temperatures can also occur for Arctic clouds in summer.
A closure study concerning INP number concentrations in the sea surface microlayer (SML) and the atmosphere suggested that INPs need to be significantly enriched during transfer from the SML into the atmosphere, meaning that INPs need to be emitted preferentially, compared to sea salt, to explain observed atmospheric INP concentrations based on sea spray production (Hartmann et al. 2021).To elucidate the chemical links between INPs and marine carbohydrates, an analysis method to detect free and combined carbohydrates in saline samples was developed (Zeppenfeld et al. 2020).Measurements revealed that glucose may serve as a biological INP tracer in the central Arctic Ocean.In seawater, phytoplankton composition and its overall abundance show a linkage to glucose, which is likely formed or released together with biogenic INPs.The SML of the marginal sea ice zone (MIZ) and aged melt ponds are particularly enriched in glucose and INPs (Zeppenfeld et al. 2019).Hence, there are strong E225 indications that the MIZ and melting sea ice environments represent local sources of marine INPs.Supplementary investigations of ice core samples have shown that INP concentrations in the Arctic seem not to be affected by anthropogenic pollution (Hartmann et al. 2019).
Aerosol particles acting as CCN are important for liquid water cloud processes.Hartmann et al. ( 2021) examined CCN and INP concentrations in a case study and did not find a correlation between these two types or particles.In a further study, new particle formation events were observed in the summertime Arctic, which were shown to increase the background CCN concentrations (Kecorius et al. 2019).
Local atmospheric processes (green in Fig. 1).Local processes and feedback mechanisms take place at a fixed location; their causes and impact are mostly restricted to the same place.
In this section we focus on local phenomena caused by clouds ("Clouds: Representation by models" to "Clouds: Interaction with atmospheric energy fluxes" sections) and precipitation ("Precipitation: Radar observations and regional modeling" section), which consider key aspects within (AC) 3 in general (Wendisch et al. 2019).
cloudS: repreSenTaTion by modelS.Clouds were analyzed in the ECHAM6 GCM in combination with data from the CALIPSO cloud-aerosol lidar and a satellite simulator (Kretzschmar et al. 2019).As expected, the evaluation pointed to the Wegener-Bergeron-Findeisen (WBF) process as a key determinant of the life cycle of Arctic mixed-phase clouds.Corrections of the representation of the WBF were required to obtain an improved agreement between satellite data and model results in the Arctic.To further enhance the cloud representation in regional Arctic climate models, more realistic, observation-tied modeling of CCN activation (Kretzschmar et al. 2020;Mech et al. 2020), and the use of appropriate INP concentrations (Sedlar et al. 2020) were identified as crucial factors.
Targeted LES were designed to explore the interaction of mixed-phase clouds with the large-scale flow (Neggers et al. 2019;Egerer et al. 2021).Furthermore, Lagrangian model configurations were adopted that follow the clouds as embedded in warm/moist airmass intrusions, constrained by measurements collected during PASCAL.We find that entrainment deepening driven by liquid cloud-top cooling occurs persistently, but varies very little and cannot fully explain the effective mixed layer deepening.In contrast, large-scale subsidence acts much more as a control on mixed layer evolution.It is much more episodic, including strong subsidence events that are even capable of causing cloud collapse (Neggers et al. 2019).This behavior can well be captured by idealized bulk mixed layer model approaches.
cloudS: impacT of Surface condiTionS and airmaSS properTieS.In general, and not surprisingly, in situ and remote sensing observations conducted during (AC) 3 clearly emphasized that the cloud properties were significantly impacted by surface conditions and airmass characteristics.More specifically, we could show that over sea ice, the total water path, and the mean droplet number and mass concentrations were lower, and the droplet sizes smaller than over open ocean (Mioche et al. 2017;Mech et al. 2019;Ruiz-Donoso et al. 2020).In addition, airborne radar measurements verified that clouds over sea ice were of a lower vertical extent with more frequent but rather low amounts of precipitation, as compared to clouds over open ocean (Mech et al. 2019).During ACLOUD, mixed-phase clouds and precipitation were not observed at temperatures below −14°C.Over the MIZ in particular, they were not observed at temperatures below −10°C.Higher concentrations of small droplets were encountered during southerly air flows, which were characterized by higher CCN concentrations, compared to clouds associated with cleaner air masses originating from the north (Wendisch et al. 2019).
E226 4 The SEB is generally defined as the net (downward minus upward) energy flux density of the radiative, turbulent sensible, latent heat, and conductive heat fluxes.It is given in units of W m −2 .
Dedicated LES experiments for selected ACLOUD flights and respective sensitivity tests revealed that CCN concentrations in the air mass significantly affect the efficiency of radiatively driven entrainment in warming the boundary layer.The response in the thermal inversion strength plays a key role in this impact (Chylik et al. 2023).Measurements of cloud droplet residuals, sampled in clouds by means of a counterflow virtual impactor (Ehrlich et al. 2019), and aerosol particles above and below clouds have indicated whether the cloud forming particles were linked to the surface layer below cloud base, or to the free troposphere above the cloud-top inversion (Wendisch et al. 2019).The latter was the case for clouds over sea ice, while above the open ocean cloud-forming particles likely originated from below the cloud.Whether this pattern is directly linked to the emission of aerosol particles at the surface or indirectly caused by the different thermodynamic and turbulence profiles over both surfaces is still unclear.However, the turbulence measurements during the (AC) 3 airborne campaigns verified the important role of cloud-generated turbulence in the ABL over sea ice (Wendisch et al. 2019;Chechin et al. 2022).Additionally, in situ observations identified larger cloud particle residuals over open ocean with smaller ones over sea ice, which further indicates different pathways for aerosol particles, feeding the cloud from below and/or above the cloud.Further investigations of the influence of surface coupling showed that for cloudtop temperatures higher than −10°C, surface-coupled clouds contained ice more often than clouds that were decoupled from the surface (Griesche et al. 2021).This suggests an influence of near-surface aerosol particles in the process of heterogeneous ice formation for Arctic ABL clouds at weakly supercooled temperatures.
Humidity inversions just above cloud top have been detected by the BELUGA system by Egerer et al. ( 2021) with a higher vertical resolution than provided by earlier radiosonde studies (Naakka et al. 2018).Dedicated LES experiments were designed, based on the method explored by Neggers et al. (2019), to accompany the BELUGA measurements and to function as a virtual laboratory for investigating impacts of humidity inversions on clouds below.cloudS: inTeracTion wiTh aTmoSpheric energy fluxeS.Cloud radiative impacts typically drive variability in the overall surface energy budget (SEB), 4 where other terms of the SEB most often respond to the radiative energy fluxes.To first order, cloud radiative effects control the surface skin temperature, which in turn affects the nearsurface stratification and, thus, the sensible heat flux.The vertical sensible heat flux typically acts to mitigate variability in surface radiative energy fluxes and, thus, feeds back on the near-surface air temperature.Collectively, as the partitioning of radiative and turbulent heat fluxes constrains the surface temperature, this also determines the conduction of heat through the sea ice from the warm ocean below.MOSAiC observations have shown that the winter surface temperature was 8°-10°C higher when liquid water clouds were present versus when they were not (Shupe et al. 2022).Similarly, surface sensible and conductive heat fluxes were each significantly diminished under cloudy skies.The sensible heat flux was 10 W m −2 less, and the conductive heat flux is also up to 10 W m −2 less.The magnitude of these impacts depends on many details, including ice thickness, snow depth, solar radiative energy input, and others.
It was shown that state-of-the-art Arctic regional models are, in principle, able to represent critical observed transitions of the SEB.However, the across-model spread of the results can be large (Sedlar et al. 2020), which is mainly related to the different treatments of clouds and cloud-radiation interactions.In particular, models struggle with the proper representation of the cloud phase partitioning and vertical distribution of cloud liquid water content (LWC), compared to observations (Inoue et al. 2021;Kretzschmar et al. 2019Kretzschmar et al. , 2020)).

E227
(AC) 3 airborne measurements close to and over the MIZ identified four distinct modes of near-surface terrestrial net irradiances below clouds (Wendisch et al. 2019;Kretzschmar et al. 2020;Stapf 2021), distinguishing low versus high surface albedo, and cloud-free versus opaquely cloudy states (Fig. 8).This conceptual framework generalizes the previous classification into two typical atmospheric states in the Arctic (Shupe and Intrieri 2004;Stramler et al. 2011;Graham et al. 2017b).In addition, seasonal influences of surface and thermodynamic properties on the mode structure have been revealed.During AFLUX (spring), strong surface temperature gradients between sea ice and the open ocean (up to 25 K) were observed, while during the second half of ACLOUD (melting period during early summer) smaller temperature gradients of only up to 6 K were found.A warming surface causes an increase of upward terrestrial radiation meaning a more negative net (downward minus upward) irradiance (increased loss of terrestrial radiation at the surface, radiative cooling), because more upward terrestrial radiation is emitted by the surface.Thus, the cloud-free modes over sea ice and open ocean differ strongly in response to the large horizontal temperature gradient prevailing during AFLUX (Fig. 8a).For opaquely cloudy conditions, cloud-base temperature is crucial, since it determines the downward terrestrial radiation.Cloud-base temperature is approximately the same when over sea ice or adjacent open ocean, leading to similar downward irradiance.However, the upward irradiance is significantly higher over the warmer open ocean compared to the sea ice, resulting in more negative values of terrestrial net irradiance (increased loss of terrestrial radiation) over the ocean than over the sea ice during AFLUX.During ACLOUD, with a much smaller gradient of surface temperature, the corresponding cloud-free and opaquely cloudy modes differ only slightly over sea ice and open ocean (Fig. 8b).This last statement is true for terrestrial radiative balance, but not for solar, which could be the dominant term of the total (solar plus terrestrial) net irradiances depending on sun angle.
Simulations of terrestrial net irradiances at cloud-system-resolving kilometer-scale resolutions were conducted using the ICON model and compared with the airborne observations during ACLOUD (Kretzschmar et al. 2020).The simulations represented the observed four-mode radiative structure close to the surface comparatively well if the measured surface albedo was implemented into ICON (Wendisch et al. 2019).Several model deficiencies, however, were identified when investigating the cloud microphysical state in detail.The simulations were

E228
improved when the cloud droplet activation scheme was revised, using an observations-tied profile of CCN and a representation of the turbulence impact on cloud-scale updraft speeds (Kretzschmar et al. 2020).
Depending on cloud properties, the season of the year, and surface conditions, Arctic clouds may warm or cool the surface.To quantify the cloud effect on radiative energy fluxes the concept of the cloud radiative forcing (CRF) was applied during (AC) 3 .The few available observation-based retrievals of the CRF of Arctic clouds were significantly extended by continuous ground-based and shipborne observations, and in situ airborne measurements in different seasons (Barrientos-Velasco et al. 2020;Ebell et al. 2020;Stapf et al. 2020Stapf et al. , 2021;;Stapf 2021).From the continuous observations at the AWIPEV research base, the surface CRF was calculated over more than 2 years (Ebell et al. 2020).The results confirmed a negative CRF (cooling effect of clouds) in summer and a warming by clouds from September to April/ May with an annual positive CRF of 11 W m −2 .Liquid-containing clouds were found to be the largest contribution to the CRF.Additionally, the CRF was estimated considering shipborne and satellite remote sensing observations for the summer of 2017, using data from PASCAL (Griesche et al. 2020).The results indicated a cooling effect of about −9 W m −2 ( Barrientos-Velasco et al. 2022).Airborne observations of the CRF-covered open-ocean and sea ice surfaces, as well as the MIZ (Stapf et al. 2020(Stapf et al. , 2021;;Stapf 2021).Figure 9 illustrates the warming effect of clouds over sea ice (highly reflecting surface) and the cooling over ocean (low surface albedo) as derived from all ACLOUD observations.The solar CRF depended strongly on the interaction of clouds and the surface albedo (Jäkel et al. 2019;Stapf et al. 2020).Over snow and sea ice surfaces, the estimated average solar cooling by clouds almost doubles if this interaction is taken into account.This emphasizes the importance of realistic surface albedo parameterization in models.
precipiTaTion: radar obServaTionS and regional modeling.Precipitation, in particular snowfall, is a critical component of the Arctic climate system influencing the hydrologic cycle and surface energy fluxes through impacts on surface albedo and subsurface conduction.During (AC) 3 , innovative retrieval algorithms converting cloud radar measurements to snowfall rate were developed, based on in situ and remote sensing measurements (Schoger et al. 2021).Furthermore, the skill of five reanalyses and the regional climate model HIRHAM5 in simulating precipitation associated with atmospheric river cases during the ACLOUD/PASCAL campaign was studied (Viceto et al. 2022).The total precipitation amounts were similar; however, the discrimination of the precipitation phase showed major differences.Also, the seasonal and regional distributions of snowfall in HIRHAM5 simulations over the period of 2007-10 were evaluated in two ways: using the CloudSat retrieval (classical approach), and applying the forward operator PAMTRA (von Lerber et al. 2022).The classical retrieval approach reveals that HIRHAM5 is much closer to the mean snowfall rate than ERA-Interim, and the model reproduces the surface snowfall associated with specific weather patterns as observed by CloudSat.The only exception is the northerly direction typically occurring during marine cold-air outbreaks where HIRHAM5 seems to underestimate the associated snowfall.
Arctic-midlatitude linkages (yellow in Fig. 1).Arctic amplification is not only dependent on local physicochemical and biogeochemical processes and their nonlinear feedbacks, but is also largely affected by a hierarchy of regional and global changes of atmospheric composition and dynamics, as well as by the related meridional and vertical transport of energy and matter in the atmosphere, ocean, and cryosphere.These transports into and out of the Arctic, which change over time due to internally generated and externally forced changes in large-scale atmospheric and oceanic variability patterns, link the Arctic with the midlatitudes and vice versa.Here we focus on temporal trends and processes, which are related or possibly caused by meridional transports in the atmosphere.They include water vapor and clouds, precipitation, and aerosol particles.Furthermore, we look at stratospheric pathway, ocean heat transport, and atmospheric cyclonic circulation.
waTer vapor and cloudS: Temporal TrendS.(AC) 3 has shown that a set of factors, including increasing temperature and moisture advection, increasing sea surface temperature, and reduced sea ice extent have caused an overall increasing trend of IWV averaged over the time period of 1979-2016 and the central Arctic (≥70°N) during all seasons (Rinke et al. 2019b).Recently, a shift of the maximum IWV trend from summer to fall was observed.This phenomenon is related to an accelerated increasing trend of IWV observed over the Barents and Kara Seas associated with the sea ice retreat, and respective consequences on warming and atmospheric circulation changes in that region.Notably, the Arctic-wide moistening trend is also reflected over large parts of the central Arctic Ocean in fall and winter (Rinke et al. 2019b), as well as in local long-term observations at the AWIPEV research base in Svalbard (Barthlott et al. 2017;Kulla and Ritter 2019;Nomokonova et al. 2020).This moistening has been linked to an increased occurrence and persistence of storms (Rinke et al. 2017;Zahn et al. 2018) that transport moist and warm air into the Arctic and cause a distinct winter warming of the Svalbard and North Pole regions (Dahlke and Maturilli 2017;Graham et al. 2017a).The enhanced moisture advection via the North Atlantic pathway has further been attributed to changed atmospheric circulation patterns (Mewes andJacobi 2019, 2020), namely, an enhanced occurrence of the Scandinavia-Ural high pressure blocking in early winter, and to sea ice retreat in the Barents and Kara Seas (Dahlke and Maturilli 2017;Crasemann et al. 2017).The moistening can have far-reaching impacts.For example, higher relative humidity was identified as a major factor for the growth of polar lows (Radovan et al. 2019).However, from the analysis of numerous satellite-based IWV products, strong differences of monthly mean IWV have been found in summer, likely due to the challenge of representing the complex and changing surface characteristics within the retrieval algorithms (Crewell et al. 2021).This is in line with the identified large differences in the IWV trend across reanalyses over the central Arctic Ocean in summer (Rinke et al. 2019b).
Ground-based observations at the AWIPEV research base hint at an increase of cloud occurrence, liquid water path (LWP), and ice water path (IWP) in all seasons (Nomokonova et al. 2020).Using a new retrieval algorithm (Nakoudi et al. 2021b), the long-term analysis (2011-20) of data collected at AWIPEV revealed that in winter and spring cirrus clouds are thicker and seem to appear more frequently (Nakoudi et al. 2021a).From the preliminary analysis of satellite data from previous research outside (AC) 3 , it appears that an increase in cloud LWP has resulted in a positive trend in COT for the liquid phase of 2.8% decade −1 and a negative trend in COT for the ice phase of −6.1% decade −1 (Lelli et al. 2022).
precipiTaTion: aTmoSpheric riverS.Atmospheric rivers (ARs) represent filament structures of enhanced vertically integrated, horizontal moisture transport, which are found responsible for a majority (up to 80% in winter) of the poleward moisture transport into the Arctic (Nash et al. 2018).This high percentage is despite their rare occurrence globally and even more rarely reaching the polar regions (Guan and Waliser 2019;Wille et al. 2021).ARs are often associated with intense precipitation and they can trigger sea ice and Greenland ice sheet surface melt events (Neff 2018;Box et al. 2022).Significant uncertainties exist in the detection of the ARs particularly in the Arctic (Rutz et al. 2019).Within (AC) 3 , observations at the AWIPEV research base in Ny-Ålesund were combined with satellite-derived data, models, and state-of-the-art reanalysis products to study ARs (Bresson et al. 2022;Viceto et al. 2022).The regional climate model HIRHAM5 and the Limited-Area Mode (LAM) ICON model were used to study specific AR events during the ACLOUD and PASCAL campaigns.To identify AR events, a new polar-specific algorithm by Gorodetskaya et al. (2020) was adapted to the Arctic and compared to the global algorithm by Guan and Waliser (2019).The analyzed cases highlighted the importance of the Atlantic AR pathway in winter, but also its shift eastward (toward Siberia) in spring and summer.Compared to observations during these events, it was demonstrated that model simulations with high horizontal resolution (down to 3 km) result in an improved representation of the spatiotemporal AR structure and its signature in the temperature, humidity, and wind profiles relative to global simulations and ERA5 (Bresson et al. 2022).Results also revealed the important role of ARs originating in western Siberia for influencing Svalbard and Greenland in May-June (Viceto et al. 2022).Tropospheric humidity profile measurements derived from radiosondes showed the ARrelated increase in low-level moisture, sometimes topped by a dry layer above, which were not well captured in HIRHAM5, while ICON-LAM sufficiently represented the fine vertical structure and its evolution (Bresson et al. 2022;Viceto et al. 2022).An important feature of the analyzed AR events was a transition from rainfall over the land areas to preferentially snowfall over the sea ice during a May AR event, and a dominance of rainfall during the June events (Viceto et al. 2022).Furthermore, the significant impact of the ARs on the surface radiative energy fluxes has been quantified (Bresson et al. 2022).
aeroSol parTicleS: Temporal changeS, TranSporT, radiaTive effecTS, and global SimulaTionS.The ACLOUD and PAMARCMiP airborne campaigns, conducted in the European Arctic, confirmed higher black carbon (BC) number concentrations in spring, compared to lower values in early summer (decrease by a factor of 5, almost equally at all altitudes), which is similar to observations performed in the Canadian Arctic (Schulz et al. 2019).These seasonal differences are mainly controlled by transport patterns and emission sources (Willis et al. 2019), and also by cloud processes in the ABL (Wendisch et al. 2019).Preliminary analyses of existing satellite data in the framework of (AC) 3 have revealed a long-term negative trend in aerosol optical thickness over the Arctic Ocean between −12% and −13% decade −1 (1981-2020), which is attributed to both the reduction of the transport of tropospheric aerosol particles and their precursors from Europe and North America and the reduction of stratospheric aerosol particles.However, an increase of large Northern Hemisphere wildfires in recent years has provided large smoke layers over the central Arctic up to the stratosphere (Ohneiser et al. 2021).The year-to-year variation of biomass-burning activities also likely affected BC amounts in the Arctic troposphere in spring (Ohata et al. 2021).In general, and E231 aside from lack of satellite coverage at very high latitudes, one of the main challenges in the retrieval of atmospheric aerosol (and cloud) properties over the Arctic Ocean from spaceborne passive microwave measurements is sea ice coverage and melt ponds on sea ice.(AC) 3 made some progress in considering highly reflective surfaces in the aerosol retrievals (Mei et al. 2020b,c), although the problem is not fully solved yet.
During PAMARCMiP, aerosol particles transported over long distances were observed above the Fram Strait and Ny-Ålesund.They consisted of a mixture of industrial pollution and biomass-burning particles (Nakoudi et al. 2020).In a unique approach, the radiative forcing of these aerosol particle plumes was estimated taking into account their modifications during the transport.As a result, although different aerosol size distributions were derived over Fram Strait and over Ny-Ålesund, the solar aerosol radiative forcing was similar for both locations, ranging between 4.4 and 4.9 W m −2 .Furthermore, the dependence of the radiative effects of BC on the mixing state, vertical distribution, surface albedo, and cloud properties was systematically investigated within (AC) 3 on the basis of measured data and not purely based on simulations (Kodros et al. 2018;Zanatta et al. 2018;Donth et al. 2020).In a case study, retrievals of aerosol microphysical properties were used to estimate the radiative impact of a specific biomass-burning event (Ritter et al. 2018).In addition, mass concentrations of BC particles originating either from long-range transport or local emission and then deposited at the ground and embedded in surface snow layers were measured during PASCAL; the observed values ranged between 2 and 10 ng g −1 .Their absorption effect caused a local solar radiative warming at the surface, which was estimated by radiative transfer simulations to be up to 0.7 W m −2 (Donth et al. 2020).However, Zanatta et al. (2021) have recently discovered that the BC mass detection efficiency of the respective instrument, the single particle soot photometer, drastically drops with increasing sea salt content during snow sample analysis.Therefore, the previously determined BC mass concentrations in surface snow layers might be seriously underestimated and with this their solar radiative warming as well.The onset of melting in spring and early summer led to an accumulation of BC within the snow layer at the surface, yielding BC mass concentrations exceeding 10 ng g −1 , probably even higher due to the exposure of deeper snow layers with higher BC loading.Although, the radiative effects of BC in snow and atmosphere are well documented from model simulations (Warren and Wiscombe 1980;Flanner 2013), the quantitative influences of BC on the radiative forcing of the atmosphere-surface system, based on measured BC concentrations within both, the atmosphere (especially the high concentration in spring), and the surface snow is still unclear.
Within (AC) 3 , global aerosol-climate modeling has investigated the controls of seasonal and vertical BC variability, as well as key uncertainties limiting the accuracy of model estimates of Arctic BC and its radiative effects (Schacht et al. 2019).Available literature shows that models tend to underestimate BC in the lower Arctic troposphere, but are often biased high at altitudes above 500 hPa (Sand et al. 2017;Lund et al. 2018).Schacht et al. (2019) introduced improved emission assumptions into the modeling, which helped to reproduce observations with 25% higher BC burden, while an optimized representation of aerosol aging and wet removal led to 10% lower high-altitude loadings.Schacht et al. (2019) also found values of top of atmosphere (TOA) BC radiative forcing of +0.31 W m −2 , and a BC-in-snow albedo effect of +0.12 W m −2 , averaged over the Arctic (>60°N) for years 2007-18, which places this study in the upper range of recent climate-model estimates (Sand et al. 2017;Gliß et al. 2021).
STraToSpheric paThway, ocean heaT TranSporT, and aTmoSpheric cyclonic circulaTion.Arctic amplification and related sea ice decline has an impact on the large-scale atmospheric circulation and energy transport.A tropospheric pathway expressed by more frequent occurrence of high pressure blocking situations over Scandinavia and northern Eurasia in early winter (Crasemann et al. 2017) initiates a stratospheric pathway with enhanced E232 upward propagation of wave energy and momentum weakening the stratospheric polar vortex (Romanowsky et al. 2019).The subsequent downward propagation of these stratospheric circulation anomalies contributes to persistent negative North Atlantic Oscillation (NAO) anomalies in late winter (Jaiser et al. 2016).This dynamical pathway is considered as most robust (Cohen et al. 2020), but it is still missing in many climate models (Smith et al. 2022).An improvement of the modeling of this stratospheric pathway by including interactive stratospheric ozone chemistry into GCMs was achieved within (AC) 3 (Romanowsky et al. 2019).
Two (AC) 3 model studies found evidence for positive atmosphere-sea ice-ocean feedback processes triggered over the North Atlantic sector, which is the key region for oceanic heat transport into the Arctic (Akperov et al. 2020;Kovács et al. 2020).Combined with atmospheric changes and winter sea ice decline, an increased Atlantic northward ocean heat transport was observed via the strengthening and warming of the Atlantic water inflow.Model experiments indicate that an anomalously high Atlantic water inflow through the Barents Sea opening in winter was associated with a cyclonic circulation anomaly and sea ice reduction in the Barents Sea region, which can lead to decreased atmospheric stability and increased wind shear in the lower troposphere, providing favorable conditions for cyclogenesis (Akperov et al. 2020).These shifts may lead to further increasing Atlantic water inflow, suggesting a potential positive regional feedback.
Further model experiments have shown that also changes in the strength of the cyclonic circulation over the Nordic seas in winter can potentially trigger positive feedback mechanisms (Kovács et al. 2020).Such circulation changes may cause significant anomalies in sea ice cover and sea surface temperature, especially in the Greenland Sea, which may lead to exceptional ocean-air heat fluxes modifying the atmospheric stability and possibly amplifying the initial wind anomaly.Due to more favorable conditions for cyclogenesis, cyclones might penetrate farther to the north and reach the central Arctic with important implications for further reduction of sea ice through a shifted sea ice edge, reduced sea ice growth, ice breakups in winter, and amplified sea ice retreat in the subsequent summer.The changed baroclinic cyclones impact planetary wave patterns as well and thereby introduce some degree of unpredictability of large-scale atmospheric waves into climate models (Cohen et al. 2020).
Where do we go from here?To understand and project the interlinked effects of atmospheric and surface processes and feedback mechanisms on Arctic amplification, the many pieces of the Arctic climate puzzle obtained by (AC) 3 so far, and those to be achieved in the future of this project, must be assembled.This synthesis will be achieved by two efforts: (i) enabling models to more realistically represent the relevant processes across scales and use them for sensitivity studies to quantify the relative importance of the processes driving Arctic amplification, and (ii) combining and integrating observational and modeling activities focusing on major drivers of Arctic amplification.Effort (i) will continuously be pursued within (AC) 3 by anchoring models in reality using state-of-the-art observations, and by further improving parameterizations and process representations.Activities with regard to effort (ii) are currently being intensified within (AC) 3 by focusing on research of the four following major crosscutting processes and drivers of Arctic amplification.
Lapse rate feedback.The usually low ABL height promotes an amplified near-surface warming and a rather muted free-tropospheric heating in the Arctic, which determines the lapse rate feedback.This mechanism is linked to surface energy fluxes, vertical mixing, clouds, radiation, and large-scale advection.First results on the lapse rate feedback underline the importance of the surface fluxes in particular where sea ice retreats.They

E233
show strong increases in the lapse rate response to warming in the past 40 years, which may imply opportunities for thorough model evaluation.
Key questions that are being pursued in (AC) 3 to quantitatively investigate the processes driving the lapse rate feedback, to verify and improve its representation in climate models exploiting observations, and to quantify its contribution to Arctic amplification are as follows: (i) Which processes determine the surface energy budget?(ii) How does vertical mixing and the inversion strength change in the Arctic, both spatially and temporally?(iii) Which mechanisms govern advective heating and cloud-top radiative cooling?(iv) How does the free-tropospheric temperature change?
Coupled surface-atmosphere processes.Coupling between the atmosphere and surface (snow cover, sea ice, open ocean or land) strongly constrains surface energy fluxes.For example, the transition from large sea ice cover, through fractured sea ice (leads, polynyas) to open (sea ice-free) ocean, involves dramatic changes in radiative and turbulent energy fluxes.To consider these systems, three major questions will be addressed: (i) What is the influence of changing surface types (e.g., open-water leads, rough vs smooth, warm vs cold, biochemically active vs inactive surfaces) in the Arctic on near-surface air temperatures, turbulent and radiative energy fluxes, and airmass transformation in the different Arctic subregions?(ii) How do changing Arctic surface properties affect the emission of marine aerosol precursors as a source of atmospheric particles, and aerosol-cloud interactions in the Arctic?(iii) Which degree of surface heterogeneity must be represented in models?
Arctic mixed-phase clouds and their representation in models.To better understand the mixed-phase cloud persistence as well as their macro-/microphysical and radiative properties on various scales, data collected during (AC) 3 (ground-based, airborne, ship-based, satellite observations) from different regions (Ny-Ålesund, Eurasian, western and central Arctic) will be analyzed in a synergistic way.Combining the observations with simulations will help to assess cloud parameterizations across the Arctic.For this purpose, the following major questions will be tackled: (i) How do surface and synoptic conditions and airmass transformations influence the representation of mixed-phase clouds in coarse and highresolution models?(ii) What is the impact of changes in cloud phase on cloud radiative effects?(iii) How do aerosol-cloud-turbulence interactions influence the properties and evolution of Arctic aerosol particles and Arctic mixed-phase clouds?
Airmass transport and transformation.The interplay between warm-/moist-air intrusions into the Arctic and cold-air outbreaks is decisive for the overall energy budget of the Arctic.The airmass transformations along their path over different surfaces are challenging to model.Deficiencies in the model representation of the development of clouds and precipitation both in respect to their positioning and amount have been identified.Within the recent HALO-(AC) 3 airborne campaign, successfully completed in March/April 2022 (Wendisch et al. 2021), dedicated flight patterns have been flown to map the spatiotemporal development of pronounced airmass transports to provide reference cases for Lagrangian-like investigations.In addition, anomalously strong moisture transport by atmospheric rivers with related precipitation over sea ice, as well a distinct synoptic-scale Arctic storm event transporting pulses of heat and moisture into the Arctic have been observed in detail during this campaign.Cold-air outbreaks are another facet of Arctic-midlatitude linkages that have been probed during HALO-(AC) 3 .The following questions will be investigated in future studies: (i) How do cloud microphysical and radiative properties and vertical mixing affect airmass transformation and precipitation, and how are these processes represented in models of different resolution?(ii) What is the role of anomalous moisture transport into the Arctic for

E234
precipitation, and what are its impacts on the surface energy budget and ice-ocean conditions along its pathway?Summary Arctic amplification represents a major and alarming sign of currently ongoing climate changes.In 2016, the Transregional Collaborative Research Center (AC) 3 was established to investigate atmospheric and surface processes and feedback mechanisms contributing to the phenomenon of Arctic amplification.Since then, the project has successfully performed comprehensive observations and data analysis, in concert with extensive modeling activities on different temporal (long-and short-term campaigns, satellite observations over several decades) and spatial (from centimeter-scale turbulence to global circulation) scales.Some of the results achieved within (AC) 3 since its beginning in 2016, are summarized in the following.
Trends of Arctic amplification-relevant parameters.Indications of trends of several climate-relevant parameters in the Arctic have been revealed, which shed new light on processes determining Arctic amplification.An obvious moistening of the central Arctic for all seasons with the highest absolute trend in summer (larger than 0.3 mm decade −1 ) and lowest value in winter (less than 0.2 mm decade −1 ) was detected .Over the Atlantic sector and large parts of the central Arctic Ocean, the moistening is most obvious in fall and winter.This trend has been linked with an increased occurrence and persistence of Arctic storms transporting moist warm air into the Arctic, causing an amplified winter warming in the Svalbard and North Pole regions.The increased moisture advection via the North Atlantic pathway has been attributed to an enhanced occurrence of the Scandinavia-Ural blocking in early winter.The mean annual sea ice thickness in Fram Strait decreased by 15% decade −1 (1990-2014), as well as the Arctic sea ice volume export through the Fram Straight (27% decade −1 , 1992-2014).For the time period of 2003-20, a significant decrease of snow depth on sea ice in March of 12% decade −1 for first-year ice was observed; for multiyear ice the decrease was smaller (9% decade −1 ).In addition, a positive tropospheric BrO trend of about 15% decade −1 during polar spring was identified, correlated with the increase in firstyear ice.
Enhanced process understanding.Surface albedo-cloud interactions were identified as an influential component in calculations of cloud radiative forcing.A four-mode structure of the surface radiative energy budget was revealed based on observations, and this structure was sufficiently reproduced by the ICON model.The presence of cloud layers above the dominant boundary layer clouds was found to significantly impact the lower cloud by damping cloudtop cooling and turbulent fluxes.However, in general, clouds remain a major issue in models, on all scales.The quantitative impact of black carbon on radiative forcing in the Arctic during polar day still appears unclear.A seasonal variation of INP concentrations and indications for both, terrestrial and marine/biogenic INP sources have been found.Marine/ biological particles within the upper ocean have been shown to significantly absorb solar radiation and thus to contribute to warming of the ocean mixed layer.The vertical atmospheric stability determines the coupling of the surface with cloud processes, while humidity may feed clouds from below and/or above.Airmass transformations during meridional transport are poorly represented in models.To improve this situation, the recent HALO-(AC) 3 airborne campaign has provided promising measurements (Wendisch et al. 2021).
Improved parameterizations.The exploitation of observational data has led to improved parameterizations that have been implemented in various models.To improve the model representation of the surface albedo feedback, models have been equipped with new sea E235 ice albedo parameterizations derived from field measurements, which also consider the dependence of surface albedo on cloud properties.Revised parameterizations of convective plumes and related processes over leads of different widths have been established.Turbulent energy flux parameterizations for very stable surface layers using Monin-Obukhov similarity theory stability functions have led to an improved reproduction of transfer coefficients in the very stable surface layer.Systematic comparisons of simulations with measurements have revealed further open issues in models.For example, the cloud droplet activation scheme of ICON was revised, scaling the default CCN profile, representative for a more polluted atmosphere, to be in better agreement with actually observed CCN concentrations and, furthermore, by using a representation of the turbulence impact on cloud-scale updraft speeds.
It is clear that these broad themes are all essential components of amplified Arctic change.In the coming years (AC) 3 aims to understand how these work together to determine the degree of amplification and the future trajectory of the Arctic climate system.

Fig. SB1 .
Fig. SB1.Time series of zonally and meridionally averaged, near-surface air temperature differences (anomalies).(a) The annually averaged differences of the near-surface air temperature relative to the corresponding long-term mean over the time period of 1951-80 for the Arctic (60°-90°N), midlatitudes (30°-60°N), tropics (20°S-20°N), and the globe.(b) The difference of the warming in the Arctic shown in (a), and the global average warming for winter (DJF) and summer (JJA).The thick lines in (a) and (b) without symbols indicate 5-yr running averages.The curves for spring (MAM) and fall (SON) are similar to those for DJF and, therefore, have been omitted in (b).The data are provided by the NASA GISTEMP Team, 2020: GISS Surface Temperature Analysis (GISTEMP), version 4. NASA Goddard Institute for Space Studies.Dataset accessed at https://data.giss.nasa.gov/gistemp/ on 8 Jun 2022.

Fig. 1 .
Fig. 1.The (AC) 3 simplified schematic of important local and remote processes and feedback mechanisms driving Arctic amplification.The figure illustrates the initial trigger by global warming (red), and shows examples of processes/feedback mechanisms such as (i) surface albedo feedback (black), (ii) upper-ocean effects (brown), (iii) local atmospheric processes (green), and (iv) Arctic-midlatitude linkages (yellow).Adopted from Wendisch et al. (2017) in modified form.

Fig. 2 .
Fig. 2. Time series of selected atmospheric observations collected at the AWIPEV research base at Ny-Ålesund during (AC) 3 .(a) Monthly mean solar (red) and terrestrial (blue) net (downward minus upward) surface irradiance, (b) monthly mean 2 m temperature, (c) monthly mean vertically integrated water vapor from microwave radiometer (MWR; blue) and radiosondes (red), (d) monthly frequency of occurrence of any type of clouds (black), liquid clouds (blue), mixed-phase clouds (red), and ice clouds (orange) in the atmospheric column based on a cloud radar and ceilometer synergy, (e) monthly accumulated precipitation from Pluvio weighing gauge based on original (blue) and corrected values (gray).The error bars in (a) and (c) represent the standard deviation of the daily mean values.Hatched areas indicate times where no or insufficient data are available to calculate monthly mean values.
Fig. 3. Horizontal and vertical coverage of the aircraft observations conducted during ACLOUD (82 + 82 flight hours with Polar 5 and Polar 6), PAMARCMiP (40 flight hours, Polar 5), AFLUX (66 flight hours, Polar 5), and MOSAiC-ACA (44 flight hours, Polar 5); altogether 314 h of measurements (about 10% within clouds) were collected.(a) Horizontal projections of flight pattern during the airborne measurements; LYR refers to Longyearbyen on Svalbard; (b) vertical distribution of number of flight hours spent in different altitudes over different sea ice conditions.

Fig. 5 .
Fig. 5. Radar reflectivity Z e (in dBZ) observations of airborne (curtain from lower-right to upper-left corners, measured by the flying downward-pointing 94 GHz MiRAC radar installed on Polar 5) and ground-based (curtain from lower-left to upper-right corners, measured by the locally fixed upwardpointing 94 GHz radar operated at the AWIPEV research base at Ny-Ålesund, whereby the vertically resolved column measurements were shifted with the simulated wind, indicated by the arrows) remote sensing measurements combined with high-resolution simulations (ICON-LEM, wind arrows, rendered clouds) around the area of Svalbard (topography).

Fig. 6 .
Fig. 6.(a) Average March 2015 snow depth on sea ice from AMSR2 satellite microwave radiometer observations.The black line discriminates first-year from multiyear sea ice.(b) Trend of March snow depth for years 2003-20.Stippled areas mark statistically significant trends.Time series of yearly March snow depth on (c) first-year sea ice (FYI) and (d) multiyear sea ice (MYI).See Rostosky et al. (2018, 2020).
where z is height and L is the Obukhov length; Pr 0 = 0.98 is the neutral-limit turbulent Prandtl number and the constants are a m = 5.0, b m = 0.3, a h = 5.0, and b h = 0.4.The new SCFs resulted from an optimization to Surface Heat Budget of the Arctic Ocean (SHEBA) data(Uttal et al. 2002) considering their functional form, values of constants, and dependence of ζ from the bulk Richardson number (Ri b ).

Fig. 7 .
Fig. 7. Transfer coefficients for momentum C d and heat C h as a function of the bulk Richardson number Ri b .Solid lines show 10 m values obtained with different parameterizations using surface roughnesses as given in the text.The curve notation is as follows: Black, Louis (1979); orange, Beljaars and Holtslag (1991); green, Grachev et al. (2007); red, new development of Gryanik et al. (2020); blue, Businger et al. (1971) and Dyer (1974).Black squares show SHEBA data obtained in the surface layer at different heights.

Fig. 8 .
Fig. 8. Joint two-dimensional histogram of frequency distribution of terrestrial net irradiance (measured below clouds) as a function of all-sky surface albedo observed during (a) AFLUX and (b) ACLOUD.Two modes (cloud-free and opaquely cloudy) become obvious over both open ocean (low surface albedo) and sea ice (high surface albedo).The figure is taken from Stapf (2021).

Fig. 9 .
Fig. 9. Two-dimensional histogram of frequency distribution of cloud radiative forcing (CRF) as a function of surface albedo derived from ACLOUD observations.Two cloud modes become obvious over open ocean (low surface albedo) and sea ice (high surface albedo).The figure is adapted from Stapf (2021).
Unauthenticated | Downloaded 09/17/23 06:18 PM UTC A M E R I C A N M E T E O R O L O G I C A L S O C I E T Y J A N UA RY2 0 2 3 UA RY 2 0 2 3 Unauthenticated | Downloaded 09/17/23 06:18 PM UTC