Aerosol particles can contribute to the Arctic amplification (AA) by direct and indirect radiative effects.
Specifically, black carbon (BC) in the atmosphere, and when deposited on snow and sea ice, has a positive warming effect on the top-of-atmosphere (TOA) radiation balance during the polar day.
Current climate models, however, are still struggling to reproduce Arctic aerosol conditions.
We present an evaluation study with the global aerosol-climate model ECHAM6.3-HAM2.3 to examine emission-related uncertainties in the BC distribution and the direct radiative effect of BC.
The model results are comprehensively compared against the latest ground and airborne aerosol observations for the period 2005–2017, with a focus on BC.
Four different setups of air pollution emissions are tested.
The simulations in general match well with the observed amount and temporal variability in near-surface BC in the Arctic.
Using actual daily instead of fixed biomass burning emissions is crucial for reproducing individual pollution events but has only a small influence on the seasonal cycle of BC.
Compared with commonly used fixed anthropogenic emissions for the year 2000, an up-to-date inventory with transient air pollution emissions results in up to a 30 % higher annual BC burden locally.
This causes a higher annual mean all-sky net direct radiative effect of BC of over 0.1 W m
The near-surface temperatures in the Arctic are warming at about twice the rate of the global average (
The main sources of Arctic BC are located outside of the Arctic circle and originate mainly from fossil fuel use and biomass burning.
Local emissions exist in the form of shipping, domestic fuel burning in remote locations, gas flaring and biomass burning
The concentration of BC and other aerosol types like organic carbon, sulfate and dust in the Arctic is the highest in late winter and/or early spring and shows a minimum during the summer.
The maximum is often referred to as Arctic haze and is caused by the southward expansion of the Arctic front, which promotes the transport of pollutants from the mid-latitude emission zones
Despite a good agreement between BC obtained from models and observations close to source regions
Having pointed out the potential importance of BC for the AA and the additional uncertainties in aerosol-climate models, in this study, we thoroughly evaluate the global aerosol-climate model ECHAM-HAM for the period 2005 to 2017, with a focus on BC in the Arctic. The evaluation uses a comprehensive set of ground and airborne in situ measurements of BC all across the Arctic and throughout all seasons. In order to address emissions as one of the main sources of uncertainty, we make use of different emission setups to assess the sensitivity of our model to the emission data used. The emissions are composed of different state-of-the-art and widely used emission inventories of anthropogenic air pollution and wildfires. The sensitivity studies allow for estimating the uncertainty range of the BC burden and climate radiative effects in recent aerosol-climate model simulations that are related to emission uncertainties. Estimates of BC radiative effects presented in this study comprise the atmospheric radiative perturbation and the BC-in-snow albedo effect. The model results utilizing the different emission inventories are compared among each other in such a way that the following three points can be explored: (1) the importance of considering daily varying biomass burning emissions, (2) uncertainties in current anthropogenic emission inventories and (3) the potential improvements by regional refinements, in particular in Russian air pollution sources, including gas flaring.
The methods used in this study are discussed in Sect.
For this study the global aerosol-climate model ECHAM-HAM is used.
It was first described in
The aerosol number concentration as well as the mass concentration are prognostic variables calculated using a “pseudomodal” approach.
The log-normal modes represent the following: the nucleation mode with a dry radius (
The accumulation and coarse modes contain BC, OC and DU, for both classes, and SU (internally mixed), as well as SS, for the mixed classes. Aerosol particles within a mode are assumed to be internally mixed such that each particle can consist of multiple components. Aerosols of different modes are externally mixed, meaning that they coexist in the atmosphere as independent particles. During the mixing, aging and coagulation processes, which are parameterized in M7, aerosol can grow to a bigger mode and can be coated with sulfate to transfer from the hydrophobic to hydrophilic mode. The median radius of the modes can be calculated from the number and mass concentration.
Aerosol modes of the species in ECHAM-HAM, including organic carbon (OC), sulfate (SU), mineral dust (DU) and sea salt (SS)
The removal process in ECHAM6.3-HAM2.3 is split between sedimentation, dry deposition and wet deposition.
The sedimentation process describes the removal by gravitational settling and is applied only to accumulation- and coarse-mode particles.
In the model, dry deposition is due to turbulent mixing and affects all but the nucleation mode particles.
In the wet deposition scheme, particles are removed as activated aerosol only if the cloud is precipitating.
Additionally, below-cloud scavenging is applied.
For more details on the removal processes in ECHAM-HAM, see
The modeled spatial aerosol distribution affects the climate simulations through interactions with radiation and clouds.
A lookup table with precalculated Mie parameters is used to dynamically determine the particle optical properties, considering their size, composition and water content (
While here we focus on BC, the details on the emissions of other aerosol species can be found in
We use the emissions developed for the Atmospheric Chemistry and Climate Model Intercomparison Project (ACCMIP) as described by
The global emission data set created for Evaluating the Climate and Air Quality Impacts of Short-lived Pollutants (ECLIPSE), version 5a, by
To address the importance of local emissions, we use the anthropogenic BC emission data set for Russian BC, described in
GFAS (Global Fire Assimilation System) is a data set of biomass burning emissions.
The strength of the emissions is scaled to the fire radiative power as observed by the MODIS instruments aboard NASA's Aqua and Terra satellites
Regions indicate the area used for averaging presented in Table
Arctic BC budget averaged for the years 2005–2015 (in kt month
We run ECHAM6.3-HAM2.3 at T63 horizontal resolution (approximately 1.8
Multi-year monthly mean emissions of
For the first run we use the historical year 2000 ACCMIP emissions throughout the whole simulation period.
Hereafter, this run is referred to as ACCMIP.
ACCMIP emission data are still widely used for model experiments, in some cases using the RCPs
The second run, called ACCMIP-GFAS, combines the biomass burning emissions of GFAS with the year 2000 ACCMIP emissions from anthropogenic sources (orange line in Fig.
In the third run, we use the ECLIPSE RCP4.5 emission data combined with GFAS emission.
It is referred to as ECLIPSE hereafter (blue line in Fig.
The fourth run, which is referred to as BCRUS, uses the updated spatially highly resolved BC emissions from
BCRUS is chosen as the reference run, since it uses the most up-to-date data and is therefore assumed to be the best estimate.
In BCRUS the BC emissions north of 60
Maps of annual mean BC emissions for the years 2005–2015.
Figure
Figure
For diagnostic output, the instantaneous radiative impact of all aerosol types is calculated in ECHAM-HAM by calling the radiation routine twice: once considering the interaction between aerosol particles and radiation and once without any aerosol. The difference between these two calls is then considered to be the direct aerosol radiative effect (DRE), which is free of any rapid adjustment (semi-direct effects).
To calculate the DRE by BC, the ACCMIP-GFAS and BCRUS runs were repeated, leaving BC out in the computation of radiative fluxes.
For this, BC was skipped in the calculation of the complex radiative index and the radiatively active number of particles, while the wet radius of respective aerosol modes was not adjusted further.
The DRE of BC is then derived from the difference of these two runs to their original setup.
Note that with this method, the estimate includes the semi-direct effect of BC, which is small in the large-scale average, since positive and negative effects cancel each other out, and is not statically significant in the Arctic
The aerosol transport and radiation simulations in this study consider the reduction of snow albedo due to deposited BC.
The BC-in-snow albedo effect is parameterized in terms of a lookup table based on a single-layer version of the Snow, Ice and Aerosol Radiation (SNICAR) model from
Geographic regions of Arctic aircraft campaigns. The data of these are used for model evaluation: HIPPO in blue, ACLOUD and PAMARCMiP-2017 in green, ACCESS in red, and ARCTAS in orange. Black triangles show the location of stations with BC surface measurements.
Area-weighted totals of BC emissions from anthropogenic sources and biomass burning fires for the main source regions (as shown in Fig.
Near-surface BC concentrations are taken from different measurement sites around the Arctic, shown on the map in Fig.
At five of these sites light absorption was measured with aethalometers.
Of those five stations, Alert and Summit measured at 467, 525 and 637 nm; Zeppelin Station measured at 525 nm only; and Tiksi and Pallas measured at 637 nm only.
From the light absorption, equivalent black carbon (eBC) concentrations were calculated using different mass absorption coefficients (MACs) depending on the wavelengths.
For the stations where measurements at 525 nm were available, 9.8 m
BC concentration data measured with a continuous soot-monitoring system (COSMOS), which removes volatile aerosol compounds, are available for Ny-Ålesund, Svalbard, Norway, and Barrow, Alaska, USA, for the period from 1 April 2012 to 31 December 2015 and 12 August 2012 to 31 December 2015, respectively.
The data collection and the retrieval of BC mass concentrations using a MAC of 8.73 m
In addition, we use measurements of eBC concentration at the Villum Research Station in northern Greenland that were performed with a multi-angle absorption photometer (MAAP).
We use daily averaged data from 14 May 2011 to 23 August 2013.
Further information on data sampling and processing can be found in
For Alaska we use filter-collected BC data acquired by the Interagency Monitoring of Protected Visual Environments (IMPROVE) aerosol network.
The thermal protocol used to process the measurements is described in
The correct representation of the modeled aerosol vertical distribution is a key prerequisite for estimating the aerosol radiative impact
The HIPPO (HIAPER Pole-to-Pole Observation) campaign consists of five deployments by the National Science Foundation (NSF; data set –
The BC data from NASA's campaign ARCTAS (Arctic Research of the Composition of the Troposphere from Aircraft and Satellites) were collected in two deployments,
spring (April 2008) and summer (June–July 2008), over North America and the American Arctic.
The mission design and execution are described in
The summer campaign of ACCESS (Arctic Climate Change, Economy and Society) in July 2012 took place over Scandinavia and the European Arctic
Another set of airborne measurements was collected from the 2017 PAMARCMiP (Polar Airborne Measurements and Arctic Regional Climate Model Simulation Project) campaign
Also based in Ny-Ålesund was the ACLOUD (Arctic CLoud and Observations Using airborne measurements during polar Day) campaign, with measurements from 22 May to 28 June 2017
The range of flight tracks of the aircraft campaigns used in this study are mapped in Fig.
The comparison between a coarsely resolved model and aircraft measurements is challenging because of many factors.
Any observed feature of subscale lifetime or spatial extend will be missed or at least underestimated by a model that is designed to estimate climate-relevant effects over multiple years.
In this study, we sample from the model's 12-hourly output for each measurement point during one campaign before averaging to one vertical profile, without prior interpolation.
In order to investigate the uncertainty range in the BC burden and its direct radiative impact, which results from the uncertainty in emissions, different simulations with the aerosol-climate model ECHAM6.3-HAM2.3 using four emission configurations are performed and compared as outlined in Sect.
Contour plot showing the modeled atmospheric BC burden averaged over the simulation period (2005–2015).
The atmospheric burden of BC averaged over the simulation period (2005–2015), which results from the different emission setups, is shown in Fig.
The causes and details, as well as differences between the runs, will be discussed in the following.
To estimate the range of anthropogenic emissions in currently widely used inventories, we compare the runs BCRUS and ACCMIP-GFAS. The ACCMIP run does not take recent economic changes into account, since emissions are fixed to the year 2000. BCRUS, on the other hand, is largely based on the ECLIPSE emissions that consider the economic development until 2015 and provide projections for the years after. Since both are combined with the biomass burning emissions from GFAS (which covers natural as well as human-caused fires), the differences in BC emissions are solely in the anthropogenic emissions (excluding human-caused grass and forest fires).
The use of fixed emissions in ACCMIP-GFAS causes a remarkable difference in the atmospheric burden of BC over the source regions compared with the reference run (see Fig.
Higher, more realistic estimates of emissions for Arctic sources (e.g., gas flaring) have been discussed as a requirement for reproducing observations like locally high BC concentrations in snow
The atmospheric composition and, in particular, the BC loading are strongly influenced by wildfires, which have a strong spatio-temporal variability.
The importance of considering actual biomass burning events is demonstrated by comparing the runs ACCMIP-GFAS and ACCMIP.
While ACCMIP-GFAS accounts for real fire events derived from satellite retrievals, ACCMIP uses fixed fire emissions for the year 2000.
The ACCMIP-GFAS BC emissions are higher than the ones of ACCMIP by 64.5 kt yr
The patterns of the BC burden of both runs are similar, with a higher burden over the western industrialized countries and a lower burden over China compared to BCRUS.
The area-weighted average burden of BC estimated with ACCMIP is 186
Near-surface measurements of BC mass concentrations can help evaluate the capability of ECHAM-HAM to reproduce the distribution of BC in the Arctic atmosphere and hence reasonable estimates of the warming influence of absorbing aerosol. While the data are only representative of the lowest atmospheric layer, the long time series give robust information about this specific important climate forcer. The multi-year seasonality of near-surface BC is compared with observations in the Arctic, as is the temporal correlation, with a spatial emphasis. Each measurement point is compared with the nearest grid cell at the closest time step from the model. The medians are calculated after this sampling.
Near-surface BC mass concentrations for Atlantic Arctic stations. Solid black line shows the multi-year monthly median BC mass concentration observed in
As in Fig.
Figures
As in Fig.
Of the stations used, Zeppelin Station and Ny-Ålesund are located in Svalbard.
Alert and the Villum Research Station are both situated in the north of the Greenland ice sheet.
The annual cycle of the BC concentration is shown in Fig.
For Ny-Ålesund the highest concentrations are observed in April, with a median of 30 ng m
For Zeppelin Station and Ny-Ålesund, BC is also overestimated in November and December.
Here, the model simulates monthly median values, each at 90 ng m
Figure
Results for four Alaskan stations of the IMPROVE network are shown in Fig.
Map showing the Arctic sites where the near-surface BC mass concentration was measured. Colors show the correlation coefficient between the measured and modeled daily averages. Correlation coefficients close to zero are not colored. Top right segment indicates the correlation coefficient for the BCRUS run. Clockwise are the ACCMIP, ACCMIP-GFAS and ECLIPSE runs. The label of Zeppelin Station is shifted to the north on the map for better visibility. The label of station Trapper Creek is shifted to the southeast.
The Pearson correlation coefficient between the collocated data of measured and modeled BC mass concentrations for all available aerosol stations in the Arctic region is shown in Fig.
The top right segment of each circle shows the correlation coefficient between the BCRUS model run and the measurements.
Following clockwise are the correlations for the runs ACCMIP, ACCMIP-GFAS and ECLIPSE.
The circle for Summit is not filled, since there the correlation coefficients are negative albeit close to zero (
For the other Alaskan stations of the IMPROVE network, however, a correlation between observations and BCRUS model results is found that is robustly positive. Even for the stations where the annual cycle was not reproduced, the correct timing of short-term events leads to these positive correlation coefficients. Trapper Creek shows a correlation coefficient of 0.55, Denali NP of 0.72 and Gates of the Arctic NP of 0.94. ACCMIP clearly performs the worst of all experiments, with correlation coefficients 0.14, 0.31 and 0.20 for Trapper Creek, Denali NP and Gates of the Arctic NP, respectively, while the other runs do not differ strongly from each other. Taking the position and strength of actual biomass burning events into account is crucial for correctly reproducing the near-surface BC concentrations in Alaska.
The correlation coefficient at Oulanka is below 0.3 for all runs. This, however, is computed only on the basis of 3 months of measurements. The other European stations of Pallas, Ny-Ålesund and Zeppelin Station also show relatively low correlation coefficients of 0.45, 0.50 and 0.30 for BCRUS, respectively. The other runs behave similarly.
At the four northernmost stations, Tiksi, Utqiaġvik (Barrow), Alert and Villum Research Station, correlation coefficients of 0.55, 0.65, 0.60 and 0.60 are found for BCRUS, respectively. These four stations are located north of a big land mass and likely show a good correlation, since concentrations are drastically different when the wind either comes from the land or the Arctic Ocean. With the exception of Tiksi, the ACCMIP run does not produce considerably weaker correlations with the observations than the other runs. At Tiksi, the highest correlation coefficient is expected for BCRUS, since BCRUS comprises the most recent and detailed emissions specifically for Russia. At 0.56 compared with 0.71 (ACCMIP-GFAS) and 0.61 (ECLIPSE), the correlation is, however, the lowest.
The BC mass mixing ratio from airborne measurements is a valuable source of information about the vertical distribution of BC.
However, because of the logistical difficulties and high costs, the spatial and temporal coverage is quite sparse.
The aircraft campaigns used in this study for model evaluation are described in detail in Sect.
For the winter months (December–January–February; DJF) only data from the HIPPO campaign are available, starting with the first deployment during January 2009.
We consider only data points north of 60
The observed and modeled profiles of BC mass mixing ratios from the ARCTAS spring campaign over the American Arctic (orange box in Fig.
The averaged profile of the measured BC mass mixing ratio for the HIPPO-3 campaign over the Pacific in March–April 2010 is plotted in Fig.
The ACLOUD campaign took place around Svalbard in May and June 2017 and therefore represents late spring and early summer.
As can be seen in Fig.
Vertical profiles of BC mass mixing ratios from airborne in situ measurements during the flight campaign HIPPO-1 campaign in January 2009. The modeled BC mass mixing ratios were averaged over the vertical levels. The observations are shown in black, and the different model runs are color coded (see Sect.
Results for the comparison between the ARCTAS summer campaign over the American Arctic in June and July 2008 and the model results from ECHAM-HAM are shown in Fig.
As in Fig.
Observations from HIPPO-4 (June–July 2011) and model results are compared in Fig.
As in Fig.
The profile plot for HIPPO-5 (August–September 2011) shows low observed and modeled mass mixing ratios throughout the atmosphere (see Fig.
As in Fig.
Figure
The second mission of the HIPPO campaign measured BC layering over the Pacific during November 2009.
The fall profile is shown in Fig.
Measurements overview. For aircraft campaigns, the location of the airfield is given unless no specific base can be defined (denoted by
Any difference in the prescribed anthropogenic and biomass burning emissions affects the atmospheric burden, the vertical layering and deposition of BC aerosol, as shown before.
The corresponding uncertainties of the DRE of BC in the atmosphere and those of BC in snow are explored using the calculation method described in Sect.
Since most of the effect results from the solar spectral range, the DRE is stronger in summer and close to zero in winter.
At the surface, the DRE of atmospheric BC is negative, as shown in Fig.
Arctic (60–90
The BC-in-snow albedo effect for all-sky conditions is shown in Fig.
The difference between the model runs is used to estimate the emission-related uncertainty of the Arctic energy budget.
Therefore, difference of the total radiative effect at TOA (all-sky conditions) of ACCMIP-GFAS minus BCRUS, as shown in Fig.
Multi-year mean all-sky direct aerosol radiative effect (DRE) of BC for the period 2005–2009. Top row for top of the atmosphere (TOA) and bottom row for bottom of the atmosphere (BOA).
We therefore conclude that, according to our best estimate, BC causes a net energy gain for the Arctic on the annual mean at TOA as well as BOA. The uncertainty with respect to the emission setup is roughly 25 % for TOA and BOA but stronger in absolute values at TOA. This is solely due to the uncertainties in emission; potential uncertainties in removal shown in the evaluation with observations are not included.
In this study, the representation of Arctic black carbon (BC) aerosol particles in the global aerosol-climate model ECHAM6.3-HAM2.3 is evaluated with respect to different emission inventories. As a reference BC measurements at Arctic sites and from aircraft campaigns are used comprehensively. By comparing the effects of different state-of-the-art BC emission inventories, an uncertainty range of current model estimates of the Arctic atmospheric BC burden and the local direct aerosol radiative effect (DRE) of BC is quantified. The uncertainties are explored with a focus on three influencing factors: (1) the influence of temporally variable biomass burning emissions, (2) the importance of recent air quality policies and economic developments, and (3) the potential improvements by regional refinements in Russian BC sources. This is achieved by comparing four different emission setups.
The run BCRUS represents a recent estimate of global emissions with the special feature of a high estimate in local Arctic emissions, especially in gas flaring.
It uses anthropogenic emissions from the ECLIPSE emission data set, and in Russia the BC emissions of ECLIPSE are replaced with the higher-resolution and more recent data from
The comparison between ACCMIP and ACCMIP-GFAS is used to estimate the impact of temporally variable biomass burning emissions. ACCMIP-GFAS and BCRUS are used to quantify the impact of recent developments in air quality policies and economic developments. The difference between ECLIPSE and BCRUS shows the impact of a regional refinement.
The variable biomass burning emissions are not particularly important for the annual mean of the Arctic BC burden but are crucial for reproducing high-pollution events.
The different assumptions on anthropogenic emission based on economic development and air quality policies result in an uncertainty in the BC burden of more than 50
The near-surface BC concentrations could be reproduced to a reasonable accuracy by ECHAM-HAM in most cases. The exception from this are stations that are challenging because of their surrounding orography and the horizontal model resolution, namely Summit, Ny-Ålesund and Zeppelin Station, where ECHAM-HAM falsely produced similar peak concentrations in late winter and early spring as for all other stations. The sensitivity to the different emission setups is low in the summer. This is a result of low local emissions near the measurement sites in all runs and reduced long-range transport from the mid-latitudes as well as more precipitation in the summertime Arctic.
In the months with high modeled concentrations the model shows a high sensitivity to the changing emissions for the stations closest to the Arctic Ocean. The observed monthly median BC peak concentrations in Tiksi were underestimated by the model, but the run BCRUS that includes the most accurate gas flaring emissions produced the best results. For other stations, e.g., in Barrow in February, BCRUS showed a stronger overestimation than the other runs.
A similar pattern can be observed for Zeppelin Station, Ny-Ålesund, Villum Research Station and Alert. Higher emissions lead to higher concentrations, with no significant changes in the pattern of the annual cycle. Overall, however, it is difficult to decide which emission setup provides satisfactory agreement with the aerosol observations for all cases. This means that the annual cycle of Arctic stations reproduced by ECHAM-HAM is mainly controlled by the transport. Changing the amount and location by using a different emission setup only modulates the amount of the BC concentrations but unexpectedly does not affect the seasonality significantly.
The correlation coefficients of near-surface concentrations are generally reasonably good, at 0.45 and higher for most stations. This points toward a good agreement in the timing, especially of observed peak events. These peaks are most often caused by biomass burning. The exceptions are Summit, Simeonof, Zeppelin Station and Oulanka, with correlation coefficients below 0.3. The run ACCMIP is the only one that shows significantly smaller correlation coefficients, since the biomass burning emissions for this run are fixed and not prescribed on a daily basis from satellite observations.
The evaluation using a combination of aircraft campaigns shows that, in general, the vertical distribution is reproduced well by ECHAM-HAM.
This improvement over older model versions is at least partly achieved with the aerosol size-dependent wet removal scheme by
In one summer case of an observed wet removal affecting a biomass burning plume, described by
The ECHAM-HAM simulations show that over the Arctic Ocean the net (solar plus terrestrial) TOA DRE of atmospheric BC is positive, with an annual average of over 0.4 W m
Overall, the current model version of ECHAM6-HAM2 performs considerably better than in a previous model intercomparison study
The code for ECHAM-HAM is available to the scientific community according to the HAMMOZ Software License Agreement though the following project website:
JS performed the ECHAM-HAM simulations, collected emission data and in situ measurement data from the providers, prepared the emissions, performed the analysis, and wrote the paper. BH provided support for the ECHAM-HAM simulations, suggested in situ measurement data providers, and provided advice during the analysis and on the project design. JQ, MZ, AE, JB and RC provided support in writing and designing the paper. RC gave advice on the emission data setup. WTKH provided the code and advice on the BC in snow parameterization for ECHAM-HAM. JB, AH, YK, AM, PRS, BW and MZ provided in situ measurement data and associated discussion. IT provided advice throughout the project design, setup, analysis and writing progress.
The authors declare that they have no conflict of interest.
We gratefully acknowledge the funding by the Deutsche Forschungsgemeinschaft (DFG; German Research Foundation; project number 268020496; TRR 172) within the Transregional Collaborative Research Centre “ArctiC Amplification: Climate Relevant Atmospheric and SurfaCe Processes, and Feedback Mechanisms (AC)
This paper was edited by Kari Lehtinen and reviewed by three anonymous referees.