Der anthropogene Klimawandel ist eine der größten Herausforderungen des 21. Jahrhunderts. Eine Hauptschwierigkeit liegt dabei in der Unsicherheit bezüglich der regionalen Änderung von Niederschlag und Temperatur. Hierdurch wird die Entwicklung geeigneter Anpassungsstrategien deutlich erschwert. In der vorliegenden Arbeit werden vier Evaluationsansätze mit insgesamt 13 Metriken für aktuelle globale (zwei Generationen) und regionale Klimamodelle entwickelt und verglichen, um anschließend eine Analyse der Projektionsunsicherheit vorzunehmen. Basierend auf den erstellten Modellbewertungen werden durch Gewichtung Aussagen über den Unsicherheitsbereich des zukünftigen Klimas getroffen. Die Evaluation der Modelle wird im Mittelmeerraum sowie in acht Unterregionen durchgeführt. Dabei wird der saisonale Trend von Temperatur und Niederschlag im Evaluationszeitraum 1960–2009 ausgewertet. Zusätzlich wird für bestimmte Metriken jeweils das klimatologische Mittel oder die harmonischen Zeitreiheneigenschaften evaluiert. Abschließend werden zum Test der Übertragbarkeit der Ergebnisse neben den Hauptuntersuchungsgebieten sechs global verteilte Regionen untersucht. Außerdem wird die zeitliche Konsistenz durch Analyse eines zweiten, leicht versetzten Evaluationszeitraums behandelt, sowie die Abhängigkeit der Modellbewertungen von verschiedenen Referenzdaten mit Hilfe von insgesamt drei Referenzdatensätzen untersucht. Die Ergebnisse legen nahe, dass nahezu alle Metriken zur Modellevaluierung geeignet sind. Die Auswertung unterschiedlicher Variablen und Regionen erzeugt Modellbewertungen, die sich in den Kontext aktueller Forschungsergebnisse einfügen. So wurde die Leistung der globalen Klimamodelle der neusten Generation (2013) im Vergleich zur Vorgängergeneration (2007) im Schnitt ähnlich hoch bzw. in vielen Situationen auch stärker eingeordnet. Ein durchweg bestes Modell konnte nicht festgestellt werden. Der Großteil der entwickelten Metriken zeigt für ähnliche Situationen übereinstimmende Modellbewertungen. Bei der Gewichtung hat sich der Niederschlag als besonders geeignet herausgestellt. Grund hierfür sind die im Schnitt deutlichen Unterschiede der Modellleistungen in Zusammenhang mit einer geringeren Simulationsgüte. Umgekehrt zeigen die Metriken für die Modelle der Temperatur allgemein überwiegend hohe Evaluationsergebnisse, wodurch nur wenig Informationsgewinn durch Gewichtung erreicht werden kann. Während die Metriken gut für unterschiedliche Regionen und Skalenniveaus verwendet werden Evaluationszeiträume nicht grundsätzlich gegeben. Zusätzlich zeigen die Modellranglisten unterschiedlicher Regionen und Jahreszeiten häufig nur geringe Korrelationen. Dies gilt besonders für den Niederschlag. Bei der Temperatur sind hingegen leichte Übereinstimmungen auszumachen. Beim Vergleich der mittleren Ranglisten über alle Modellbewertungen und Situationen der Hauptregionen des Mittelmeerraums mit den Globalregionen besteht eine signifikante Korrelation von 0,39 für Temperatur, während sie für Niederschlag um null liegt. Dieses Ergebnis ist für alle drei verwendeten Referenzdatensätze im Mittelmeerraum gültig. So schwankt die Korrelation der Modellbewertungen des Niederschlags für unterschiedliche Referenzdatensätze immer um Null und die der Temperaturranglisten zwischen 0,36 und 0,44. Generell werden die Metriken als geeignete Evaluationswerkzeuge für Klimamodelle eingestuft. Daher können sie einen Beitrag zur Änderung des Unsicherheitsbereichs und damit zur Stärkung des Vertrauens in Klimaprojektionen leisten. Die Abhängigkeit der Modellbewertungen von Region und Untersuchungszeitraum muss dabei jedoch berücksichtigt werden. So besitzt die Analyse der Konsistenz von Modellbewertungen sowie der Stärken und Schwächen der Klimamodelle großes Potential für folgende Studien, um das Vertrauen in Modellprojektionen weiter zu steigern. / Climate change is one of the major tasks of the 21st century. The uncertainty of precipitation and temperature change is considered as a main challenge in this context. Thus, the development of appropriate adaptation strategies is very difficult. In this study, four climate model evaluation approaches with 13 metrics in total are developed and compared. Current global (two generations) and regional climate models are evaluated to assess projection uncertainty. Based on model performances, weighting is applied to future climate projections to estimate simulation uncertainty. The evaluations are performed in the Mediterranean and eight sub-regions. Seasonal trend of temperature and precipitation are evaluated for the period 1960–2009. For some metrics, the climatological mean and the spectra of the time series are evaluated as well. In addition, six globally distributed study areas are evaluated to test the metrics’ transferability. Further, temporal consistency is assessed by the evaluation of a second slightly shifted timeframe. Finally, three reference datasets are considered in order to analyse the dependence of the evaluation results between each other. Results indicate that most metrics are suitable to evaluate climate models. Their application to different variables and regions generates reasonable model assessments which fit in the context of current publications in this field of research. In many situations, the results of the current model generation (2013) are similar or better compared to those of the last generation (2007). One single model with superior performance for all variables and situations cannot be found. Most metrics show similar estimations of performances for the same situations. Precipitation turned out to be more suitable for model weighting. Here, the differences between model weights are larger because of overall higher spread and lower model performances. Against this, there are mostly high performances on an equal level for simulations of temperature which lead to a minor added value of weighting. While metrics can easily be transferred and applied to different regions and scales, some evaluation results depend on the evaluated timeframe. Further, the model rankings for different regions and seasons show only minor correlations for most situations. This is particularly true for precipitation. However, for temperature there are some significant positive correlations. Comparing the mean ranking over all evaluation results of the main study areas of the Mediterranean with that of the globally distributed regions, there is a significant correlation of 0.39 for temperature and a correlation around zero for precipitation. The choice of reference dataset for the Mediterranean areas is subordinated in this context. For different reference datasets, the overall rankings show correlations around zero for precipitation while those for temperature are between 0.36 and 0.44. Overall, the metrics are suitable for the evaluation of climate models. Thus, they offer promising contributions to improve the range of uncertainty and therefore to enhance the general confidence in climate projections. However, dependence of model assessments on the analysed region and evaluation timeframe has to be considered. Consequently, the analyses of consistency of model evaluations and of climate model strengths and weaknesses have great potential for future studies, to further enhance confidence in climate projections.

Understanding the Earth's climate system and particularly climate variability presents one of the most difficult and urgent challenges in science. The Antarctic plays a crucial role in the global climate system, since it is the principal region of radiative energy deficit and atmospheric cooling. An assessment of regional climate model HIRHAM is presented. The simulations are generated with the HIRHAM model, which is modified for Antarctic applications. With a horizontal resolution of 55km, the model has been run for the period 1958-1998 creating long-term simulations from initial and boundary conditions provided by the European Centre for Medium-Range Weather Forecasts (ECMWF) ERA40 re-analysis. The model output is compared with observations from observation stations, upper air data, global atmospheric analyses and satellite data. In comparison with the observations, the evaluation shows that the simulations with the HIRHAM model capture both the large and regional scale circulation features with generally small bias in the modeled variables. On the annual time scale the largest errors in the model simulations are the overestimation total cloud cover and the colder near-surface temperature over the interior of the Antarctic plateau. The low-level temperature inversion as well as low-level wind jet is well captured by the model. The decadal scale processes were studied based on trend calculations. The long-term run was divided into two 20 years parts. The 2m temperature, 500 hPa temperature, MSLP, precipitation and net mass balance trends were calculated for both periods and over 1958 - 1998. During the last two decades the strong surface cooling was observed over the Eastern Antarctica, this result is in good agreement with the result of Chapman and Walsh (2005) who calculated the temperature trend based on the observational data. The MSLP trend reveals a big disparity between the first and second parts of the 40 year run. The overall trend shows the strengthening of the circumpolar vortex and continental anticyclone. The net mass balance as well as precipitation show a positive trend over the Antarctic Peninsula region, along Wilkes Land and in Dronning Maud Land. The Antarctic ice sheet grows over the Eastern part of Antarctica with small exceptions in Dronning Maud Land and Wilkes Land and sinks in the Antarctic Peninsula; this result is in good agreement with the satellite-measured altitude presented in Davis (2005) . To better understand the horizontal structure of MSLP, temperature and net mass balance trends the influence of the Southern Annual Mode (SAM) on the Antarctic climate was investigated. The main meteorological parameters during the positive and negative Antarctic Oscillation (AAO) phases were compared to each other. A positive/negative AAO index means strengthening/weakening of the circumpolar vortex, poleward/northward storm tracks and prevailing/weakening westerly winds. For detailed investigation of global teleconnection, two positive and one negative periods of AAO phase were chosen. The differences in MSLP and 2m temperature between positive and negative AAO years during the winter months partly explain the surface cooling during the last decades. / Eine der dringendsten wissenschaftlichen Herausforderungen besteht darin, das Klimasystem der Erde und die Prozesse zu verstehen, die seine Klimavariabilität bestimmen. Die Antarktis spielt eine entscheidende Rolle im globalen Klimasystem, da sie die wesentliche Energiesenke und atmosphärische Abkühlregion darstellt. In dieser Arbeit wird das regionale Klimamodell HIRHAM zur Untersuchung des Klimas der Antarktis eingesetzt, das dafür speziell angepasst wurde. Mit einer horizontalen Auflösung von 50 km und 25 vertikalen Schichten wurden Simulationen für 40 Jahre von 1958-1998 durchgeführt, wobei die Anfangs- und Randbedingungen durch die ERA40 Daten des ECMWF (European Centre for Medium-Range Weather Forecasts) geliefert wurden. Die Modellergebnisse wurden mit Daten von Beobachtungsstationen, aerologischen Vertikalsondierungen, globalen Analysedaten und Satellitendaten verglichen. Diese Validierung zeigt, dass die HIRHAM Modellsimulationen die globalen und regionalen Zirkulationsmuster mit einem vertretbaren Modellfehler generieren. Auf der jährlichen Zeitskale zeigen sich die größten Modellfehler in einer Überbestimmung der totalen Wolkenbedeckung und der kalten bodennahen Temperaturen der Atmosphäre. Die bodennahen Inversionen und katabatischen Windsysteme werden durch das Modell gut wiedergegeben. Dekadische Prozesse wurden durch Trendberechnungen analysiert. Dazu wurden die 40 Jahre umfassenden Simulationen in zwei 20 Jahre Abschnitte von 1958-1978 und 1979-1998 unterteilt. Die Trends in den 2m Temperaturen, im mittleren Bodenluftdruck, 500 hPa Geopotential, Niederschlag und der Netto Massenbilanz wurden berechnet. In den letzten zwei Dekaden wurde eine starke atmosphärische Abkühlung an der Oberfläche in der Ostantarktis simuliert, die in guter Übereinstimmung mit den Trendanalysen aus Beobachtungen von Chapman und Walsh (2005) steht. Der Trend im mittleren Bodenluftdruck weist deutliche Unterschiede zwischen den ersten Periode 1958-1978 und der zweiten Periode 1979-1998 auf. Insgesamt verstärkt sich über die untersuchten 40 Jahre der zirkumpolare Tiefdruckwirbel und die kontinentale Antizyklone. Die Nettomassenbilanz und der Niederschlag zeigen einen positiven Trend über der Antarktischen Halbinsel, Wilkes Land und Dronning Maud Land. Das antarktische Eisschild wächst im östlichen Teil der Antarktis mit geringen Ausnahmen in Dronning Maud Land und Wilkes Land an und schächt sich über der antarktischen Halbinsel ab. Dieses Resultat befindet sich in Übereinstimmung mit den Akkumulationstrends von Davis (2005) auf der Basis von Satellitendaten. Die horizontalen Strukturen der simulierten antarktischen Trends im mittleren Bodenluftdruck, in der 2m Temperatur und der Netto Massenbilanz wurden mit Trends des globalen Telekonnektionsmusters der Südhemisphäre verglichen. Dazu wurden wesentliche atmosphärische Parameter für positive und negative Phasen der Antarktischen Oszillation (AAO) analysiert. Die positiven/negativen AAO Phasen gehen einher mit einer Verstärkung/Abschwächung des zirkumpolaren Tiefdruckwirbels, verstärkten/reduzierten Stormtracks und verstärkten/abgeschwächten Westwinden. Für eine tiefergehende Untersuchung wurden zwei positive und eine negative AAO Phase miteinander verglichen. Die Unterschiede im Bodenluftdruck und der 2m Temperatur zwischen den positiven und negativen AAO Perioden können den Abkühlungstrend während der letzten Dekaden zu großen Teilen erklären.

Für die zukünftige klimatische Entwicklung in Mitteldeutschland wurden die Ergebnisse von zwölf globalen Klimamodellen ausgewertet. Im Laufe des 21. Jahrhunderts ergeben sich aus den untersuchten Simulationen für alle Jahreszeiten Temperaturzunahmen. Die Zunahmen der Jahresmitteltemperaturen bewegen sich je nach Modell und Emissionsszenario zwischen 0,5 K und 4,8 K. Während die Zunahmen im Frühjahr vergleichsweise moderat ausfallen, unterliegen die (Spät)sommer und Wintermonate einem besonders starken Erwärmungstrend. Hinsichtlich der Niederschlagsentwicklung ergeben sich insbesondere in den Sommermonaten signifikante Abnahmetrends. Besonders betroffen sind hierbei die Monate Juli und August. Deutlich zunehmende Winterniederschläge führen zu nur vergleichsweise unauffälligen Veränderungen der Jahresniederschlagssummen (Niederschlagsumverteilung). Insbesondere die kombinierte Wirkung von erhöhten Temperaturen und deutlich abnehmenden Niederschlägen sowie der damit einhergehende erhöhte Verdunstungsanspruch während der Sommermonate birgt die Gefahr angespannter Wasserhaushaltssituationen.

Large uncertainties exist in estimations of aerosol direct radiative forcing and indirect radiative forcing, and the values derived from globalmodeling differ substantially with satellite-based calculations. Following the approach of Quaas et al. (2008; hereafter named Quaas2008),we reassess satellite-based clear- and cloudy-sky radiative forcings and their seasonal variations by employing updated satellite products from 2004 to 2011 in combination with the anthropogenic aerosol optical depth (AOD) fraction obtained frommodel simulations using the Goddard Earth Observing System-Chemistry-Advanced ParticleMicrophysics (GEOS-Chem-APM). Our derived annual mean aerosol clear-sky forcing (-0.59 W m-2) is lower, while the cloudy-sky forcing (-0.34 W m-2) is higher than the corresponding results (-0.9Wm-2 and -0.2W m-2, respectively) reported in Quaas2008. Our study indicates that the derived forcings are sensitive to the anthropogenic AOD fraction and its spatial distribution but insensitive to the temporal resolution used to obtain the regression coefficients, i.e.,monthly or seasonal based. The forcing efficiency (i.e., the magnitude per anthropogenic AOD) for the clear-sky forcing based on this study is 19.9Wm-2, which is about 5% smaller than Quaas2008’s value of 21.1Wm-2. In contrast, the efficiency for the cloudy-sky forcing of this study (11 W m-2) is more than a factor of 2 larger than Quaas2008’s value of 4.7 W m-2. Uncertainties tests indicate that anthropogenic fraction of AOD strongly affects the computed forcings while using aerosol index instead of AOD from satellite data as aerosol proxy does not appear to cause any significant differences in regression slopes and derived forcings.

The formation of cloud droplets on aerosol particles, technically known as the activation of cloud condensation nuclei (CCN), is the fundamental process driving the interactions of aerosols with clouds and precipitation. Knowledge of these interactions is foundational to our understanding of weather and climate. The Intergovernmental Panel on Climate Change (IPCC) and the Decadal Survey (NRC 2007) indicate that the uncertainty in how clouds adjust to aerosol perturbations dominates the uncertainty in the overall quantification of the radiative forcing attributable to human activities. The Clouds, Hazards, and Aerosols Survey for Earth Researchers (CHASER) satellite mission concept responds to the IPCC and Decadal Survey concerns by studying the activation of CCN and their interactions with clouds and storms. The CHASER satellite mission was developed to remotely sense quantities necessary for determining the interactions of aerosols with clouds and storms. The links between the Decadal Survey recommendations and the CHASER goals, science objectives, measurements, and instruments are described in Table 1. Measurements by current satellites allow a rough determination of profiles of cloud particle size but not of the activated CCN that seed them. CHASER will use an innovative technique (Freud et al. 2011; Freud and Rosenfeld 2012; Rosenfeld et al. 2012) and high-heritage (flown in a previous spaceflight mission) instruments to produce satellite-based remotely sensed observations of activated CCN and the properties of the clouds associated with them. CHASER will estimate updraft velocities at cloud base to calculate the number density of activated CCN as a function of the water vapor supersaturation. CHASER will determine the CCN concentration and cloud thermodynamic forcing (i.e., forcing caused by changes in the temperature and humidity of the boundary layer air) simultaneously, allowing their effects to be distinguished. Changes in the behavior of a group of weather systems in which only one of the quantities varies (a partial derivative of the intensity of the weather system with respect to the desirable quantity) will allow the determination of each effect statistically.

A process-oriented climate model evaluation is presented, applying the International Satellite Cloud Climatology Project (ISCCP) simulator to pinpoint deficiencies related to the cloud processes in the ECHAM5general circulation model.ALagrangian trajectory analysis is performed to track the transitions of anvil cirrus originating from deep convective detrainment to cirrostratus and thin cirrus, comparing ISCCP observations and the ECHAM5 model. Trajectories of cloudy air parcels originating from deep convection are computed for both, the ISCCP observations and the model, over which the ISCCP joint histograms are used for analyzing the cirrus life cycle over 5 days. The cirrostratus and cirrus clouds originate from detrainment from deep convection decay and gradually thin out after the convective event over 3–4 days. The effect of the convection–cirrus transitions in a warmer climate is analyzed in order to understand the climate feedbacks due to deep convective cloud transitions. An idealized climate change simulation is performed using a+2-K sea surface temperature (SST) perturbation. The Lagrangian trajectory analysis over perturbed climate suggests that more and thicker cirrostratus and cirrus clouds occur in the warmer climate compared to the present-day climate. Stronger convection is noticed in the perturbed climate, which leads to an increased precipitation, especially on day -2 and -3 after the individual convective events. The shortwave and the longwave cloud forcings both increase in the warmer climate, with an increase of net cloud radiative forcing (NCRF), leading to an overall positive feedback of the increased cirrostratus and cirrus clouds from a Lagrangian transition perspective.

In this study, we employ the global aerosol-climate model ECHAM-HAM to globally assess aerosol indirect effects (AIEs) resulting from shipping emissions of aerosols and aerosol precursor gases. We implement shipping emissions of sulphur dioxide (SO2), black carbon (BC) and particulate organic matter (POM) for the year 2000 into the model and quantify the model’s sensitivity towards uncertainties associated with the emission parameterisation as well as with the shipping emissions themselves. Sensitivity experiments are designed to investigate (i) the uncertainty in the size distribution of emitted particles, (ii) the uncertainty associated with the total amount of emissions, and (iii) the impact of reducing carbonaceous emissions from ships. We use the results from one sensitivity experiment for a detailed discussion of shipping-induced changes in the global aerosol system as well as the resulting impact on cloud properties. From all sensitivity experiments, we find AIEs from shipping emissions to range from −0.32±0.01Wm−2 to −0.07±0.01Wm−2 (global mean value and inter-annual variability as a standard deviation). The magnitude of the AIEs depends much more on the assumed emission size distribution and subsequent aerosol microphysical interactions than on the magnitude of the emissions themselves. It is important to note that although the strongest estimate of AIEs from shipping emissions in this study is relatively large, still much larger estimates have been reported in the literature before on the basis of modelling studies. We find that omitting just carbonaceous particle emissions from ships favours new particle formation in the boundary layer. These newly formed particles contribute just about as much to the CCN budget as the carbonaceous particles would, leaving the globally averaged AIEs nearly unaltered compared to a simulation including carbonaceous particle emissions from ships.

Realistic simulations of clouds are of uppermost importance for climate modelling using general circulation models. Satellite data are well suited to evaluate model parametrizations. In this study we use the Laboratoire de Me´te´orologie Dynamique general circulation model (LMDZ). We evaluate the current LMDZ cloud phase parametrization, in which the repartition of condensed cloud water between liquid and ice is a function of the local temperature. Three parameters are used to derive a relation between liquid cloud water content and temperature, two of which are not physically based. We use the POLDER-1 satellite data to infer more realistic parameters by establishing statistical relationships between cloud top thermodynamical phase and cloud top temperature, consistently in both satellite data and model results. We then perform a multitude of short model integrations and derive a best estimate for the lowest local temperature where liquid water can exist in a cloud (Tice = -32°C in our parametrization). The other parameter which describes the shape of the transition between ice and liquid water is also estimated. A longer simulation has then been performed with the new parameters, resulting in an improvement in the representation of the shortwave cloud radiative forcing.

Cloud drop condensation nuclei (CCN) and ice nuclei (IN) particles determine to a large extent cloud microstructure and, consequently, cloud albedo and the dynamic response of clouds to aerosol-induced changes to precipitation. This can modify the reflected solar radiation and the thermal radiation emitted to space. Measurements of tropospheric CCN and IN over large areas have not been possible and can be only roughly approximated from satellite-sensor-based estimates of optical properties of aerosols. Our lack of ability to measure both CCN and cloud updrafts precludes disentangling the effects ofmeteorology fromthose of aerosols and represents the largest component in our uncertainty in anthropogenic climate forcing.Ways to improve the retrieval accuracy include multiangle and multipolarimetric passive measurements of the optical signal and multispectral lidar polarimetric measurements. Indirect methods include proxies of trace gases, as retrieved by hyperspectral sensors. Perhaps the most promising emerging direction is retrieving the CCN properties by simultaneously retrieving convective cloud drop number concentrations and updraft speeds, which amounts to using clouds as natural CCN chambers. These satellite observations have to be constrained by in situ observations of aerosol-cloud-precipitation-climate (ACPC) interactions, which in turn constrain a hierarchy of model simulations of ACPC. Since the essence of a general circulation model is an accurate quantification of the energy and mass fluxes in all forms between the surface, atmosphere and outer space, a route to progress is proposed here in the form of a series of box flux closure experiments in the various climate regimes. A roadmap is provided for quantifying the ACPC interactions and thereby reducing the uncertainty in anthropogenic climate forcing.

This paper introduces and evaluates the second version of the global aerosol-climate model ECHAM-HAM. Major changes have been brought into the model, including new parameterizations for aerosol nucleation and water uptake, an explicit treatment of secondary organic aerosols, modified emission calculations for sea salt and mineral dust, the coupling of aerosol microphysics to a two-moment stratiform cloud microphysics scheme, and alternative wet scavenging parameterizations. These revisions extend the model’s capability to represent details of the aerosol lifecycle and its interaction with climate. Nudged simulations of the year 2000 are carried out to compare the aerosol properties and global distribution in HAM1 and HAM2, and to evaluate them against various observations. Sensitivity experiments are performed to help identify the impact of each individual update in model formulation. Results indicate that from HAM1 to HAM2 there is a marked weakening of aerosol water uptake in the lower troposphere, reducing the total aerosol water burden from 75 Tg to 51 Tg. The main reason is the newly introduced k-Köhler-theory-based water uptake scheme uses a lower value for the maximum relative humidity cutoff. Particulate organic matter loading in HAM2 is considerably higher in the upper troposphere, because the explicit treatment of secondary organic aerosols allows highly volatile oxidation products of the precursors to be vertically transported to regions of very low temperature and to form aerosols there. Sulfate, black carbon, particulate organic matter and mineral dust in HAM2 have longer lifetimes than in HAM1 because of weaker incloud scavenging, which is in turn related to lower autoconversion efficiency in the newly introduced two-moment cloud microphysics scheme. Modification in the sea salt emission scheme causes a significant increase in the ratio (from 1.6 to 7.7) between accumulation mode and coarse mode emission fluxes of aerosol number concentration. This leads to a general increase in the number concentration of smaller particles over the oceans in HAM2, as reflected by the higher Ångström parameters. Evaluation against observation reveals that in terms of model performance, main improvements in HAM2 include a marked decrease of the systematic negative bias in the absorption aerosol optical depth, as well as smaller biases over the oceans in Ångström parameter and in the accumulation mode number concentration. The simulated geographical distribution of aerosol optical depth (AOD) is better correlated with the MODIS data, while the surface aerosol mass concentrations are very similar to those in the old version. The total aerosol water content in HAM2 is considerably closer to the multi-model average from Phase I of the AeroCom intercomparison project. Model deficiencies that require further efforts in the future include (i) positive biases in AOD over the ocean, (ii) negative biases in AOD and aerosol mass concentration in high-latitude regions, and (iii) negative biases in particle number concentration, especially that of the Aitken mode, in the lower troposphere in heavily polluted regions.