Direct and semi-direct effects of aerosol climatologies on long-term climate simulations over Europe

Article

Abstract

This study compares the direct and semi-direct aerosol effects of different annual cycles of tropospheric aerosol loads for Europe from 1950 to 2009 using the regional climate model COSMO-CLM, which is laterally forced by reanalysis data and run using prescribed, climatological aerosol optical properties. These properties differ with respect to the analysis strategy and the time window, and are then used for the same multi-decadal period. Five simulations with different aerosol loads and one control simulation without any tropospheric aerosols are integrated and compared. Two common limitations of our simulation strategy, to fully assess direct and semi-direct aerosol effects, are the applied observed sea surface temperatures and sea ice conditions, and the lack of short-term variations in the aerosol load. Nevertheless, the impact of different aerosol climatologies on common regional climate model simulations can be assessed. The results of all aerosol-including simulations show a distinct reduction in solar irradiance at the surface compared with that in the control simulation. This reduction is strongest in the summer season and is balanced primarily by a weakening of turbulent heat fluxes and to a lesser extent by a decrease in longwave emissions. Consequently, the seasonal mean surface cooling is modest. The temperature profile responses are characterized by a shallow near-surface cooling and a dominant warming up to the mid-troposphere caused by aerosol absorption. The resulting stabilization of stratification leads to reduced cloud cover and less precipitation. A decrease in cloud water and ice content over Central Europe in summer possibly reinforce aerosol absorption and thus strengthen the vertical warming. The resulting radiative forcings are positive. The robustness of the results was demonstrated by performing a simulation with very strong aerosol forcing, which lead to qualitatively similar results. A distinct added value over the default aerosol setup of Tanré et al. (1984) was found in the simulations with more recent aerosol data sets for solar irradiance. The improvements are largest under low cloud conditions, while overestimated cloud cover in all setups causes a common underestimation of low and medium values of solar irradiance. In addition, the prevalent cold bias in the COSMO-CLM is reduced in winter and spring when using updated aerosol data. Our results emphasize the importance of semi-direct aerosol effects, especially over Central Europe in terms of changes in turbulent fluxes and changes in cloud properties. We also suggest to replace the default Tanré et al. (1984) aerosol climatology with more recent and realistic data sets. Thereby, a better model performance in comparison to observations can be achieved, or the masking of model shortcomings due to a too strong direct aerosol forcing thus far is prevented.

Keywords

Regional climate modelling COSMO-CLM Direct aerosol effect Semi-direct aerosol effect Added value 

1 Introduction

Aerosols strongly affect Earth’s energy budget and thus the climate system (Haywood and Boucher 2000; Kaufman et al. 2002; Lohmann and Feichter 2005; Boucher et al. 2013). Two different mechanisms of aerosol forcing can be distinguished (Haywood and Boucher 2000; Lohmann and Feichter 2005). The direct aerosol effect accounts for the ability of aerosol particles to scatter and absorb solar and terrestrial radiation, which primarily occurs by reducing the amount of shortwave sunlight reaching the surface and a resulting cooling tendency. The semi-direct effect accounts for local warming within the troposphere caused by absorbing aerosols, such as black carbon, which leads to changes in atmospheric stability and cloud cover (Hansen et al. 1997; Ackerman et al. 2000; Koch and Del Genio 2010). Also, aerosol particles are essential for cloud formation because they act as cloud condensation and ice nuclei (Lohmann and Feichter 2005). Thus, changes in aerosol concentration and composition can affect cloud microphysical and optical properties (“indirect effects”, Twomey 1977; Albrecht 1989; Zubler et al. 2011b), such as cloud albedo or cloud lifetime.

Although greenhouse gases are globally well mixed, tropospheric aerosol concentrations are characterized by large spatio-temporal variability caused by a short lifetime on the order of a few days (Stier et al. 2005). Most important tropospheric aerosol types, regarding their concentration and climate impact, are dust originating from arid regions, sea salt emitted by the ocean, carbonaceous aerosols and sulphate aerosols. The latter two are predominantly anthropogenic and are derived from burning fossil fuels. These anthropogenic aerosol loads can be an order of magnitude larger when determined at a regional scale over source regions compared with when they are determined at a global scale (Kinne et al. 2013). The recent assessment report (AR5) of the IPCC highlights the role of aerosols as the largest source of uncertainty in radiative forcing since pre-industrial times (Myhre et al. 2013). Although aerosol processes are better understood now than when the previous report was published, the degree to which the complexity of these processes must be represented within a climate model to sufficiently consider their effects remains unclear. In particular, semi-direct and indirect aerosol effects are still subject to large uncertainties (Myhre et al. 2013).

Numerous studies have focused on aerosols and their impact on climate on a global or regional scale using global climate models (GCMs). Initial studies within the GCM community used uncoupled approaches with prescribed aerosols. Because of increasing computing power, fully coupled atmosphere–ocean-chemistry models are now applied for some years. In the context of regional climate models (RCMs) till today, simulations are conducted using by default coarse and often out-dated aerosol optical properties to account for aerosol–radiation interactions (Hohenegger and Vidale 2005). Investigations of aerosol effects, especially those on longer timescales, are rare. For Europe, for example, Ekman and Rodhe (2003) investigated the indirect cloud albedo effect of interactive anthropogenic sulphate aerosols for a one-year period. Zubler et al. (2011a) used different aerosol optical properties in COSMO-CLM for the period from 1997 to 2003 and compared the results with those of a simulation with the default aerosol climatology, and their results showed a partial improvement in the surface shortwave radiation (SSR); however, the authors focused on mean values and the difference to the default aerosol setup instead of the aerosol effect. Nabat et al. (2015) used prescribed aerosol optical properties in an atmosphere-ocean coupled model system for the Mediterranean Sea and adjacent areas for the period from 2003 to 2009. Despite the limited area of focus, they only addressed the aerosol effects obtained with one aerosol climatology and neglected inter-comparisons of the effects of different aerosol climatologies. Recently, Toll et al. (2016) compared different aerosol data sets in a 15-day ALADIN-HIRLAM simulation with the focus on numerical weather prediction (NWP), and they found similar responses of the meteorological quantities to different aerosol forcings, which only produced weak impacts on the forecast quality. High-resolution two-way coupled regional climate simulations, including comprehensive gas-phase chemistry and aerosol dynamics, were for instance conducted using the COSMO-ART model (Vogel et al. 2009) for only a few months. Thus far, the extreme computational demands of such simulations have prevented their application on multi-annual timescales. Moreover, the number of degrees of freedom in such a complex model system increases rapidly; therefore, a reduction in uncertainty cannot be expected a priori.

In this study, our goal was to close the gap between short-term simulations for case studies or specific weather situations and long-term studies focusing only on one specific aerosol climatology. A long simulation time of 60 years (1950–2009) was chosen to ensure robust climatic results. The objective of this study was to investigate generally valid mechanisms of direct and semi-direct aerosol effects in Europe that occur independent of the underlying aerosol forcing. Hence, several regional climate simulations were forced with different aerosol climatologies and analysed following a qualitative and comparative approach. Additionally, we investigated whether updated aerosol data yielded an added value or whether model deficits have been masked by using out-dated aerosol information thus far. Our uncoupled approach does not account for aerosol-related changes in sea surface temperature as well as high-frequency aerosol fluctuations. In addition, aerosol–cloud interactions have been neglected. However, our approach represents a standard procedure of accounting for aerosols in regional climate modelling.

The paper is structured into four main parts. First, we provide a description of the regional climate model COSMO-CLM (Sect. 2.1) and outline our modelling strategy (Sect. 2.2) as well as the used aerosol data sets (Sect. 2.3). The second part focuses on the response of the model to direct and semi-direct aerosol effects using different aerosol climatologies (Sect. 3). In the next part, model results with different aerosol forcing are compared with observation data, where also the potential added values against the default aerosol setup is discussed (Sect. 4). The final part presents the summarizing conclusions and the outlook (Sect. 5).

2 Methods

2.1 Regional climate model COSMO-CLM

In this study, we use the non-hydrostatic regional climate model COSMO-CLM (Rockel et al. 2008) version 5. The COSMO model (Doms and Schättler 2002; Doms et al. 2011; Baldauf et al. 2011), formerly known as the Lokal Modell (LM), was originally developed and applied for operational weather predictions by the German Weather Service (DWD). Further development and maintenance is organized as a joint effort within the COnsortium for Small-scale MOdelling (COSMO, http://www.cosmo-model.org/). COSMO-CLM (http://www.clm-community.eu) refers to the climate version of the COSMO model with specific adaptations for long-term simulations (Böhm et al. 2006). Climate simulation have been successfully performed using COSMO-CLM for various ranges of applications with horizontal grid resolutions from 50 km up to 1 km (Jaeger et al. 2008; Knote et al. 2010).

The Ritter and Geleyn (1992) radiation scheme is implemented for radiative transfer calculations, and it solves the \(\delta \)-two-stream solution of the radiative transfer equation for three and five spectral bands in the solar and thermal part of the spectrum, respectively. Interactions between aerosol particles and radiative fluxes occur via scattering, absorption and emission. Aerosol optical properties are described with the common parameters aerosol optical depth (AOD), single scattering albedo (SSA) and asymmetry factor (g).

Grid-scale moist processes are treated with a one-moment extended Kessler-type (Kessler 1969) bulk formulation. Prognostic number concentrations and aerosol activation is not included, but a fixed cloud droplet number concentration is used. Thus, aerosol effects are not explicitly taken into account.

2.2 Modelling strategy

To generate a number of sensitivity experiments, six simulations with different aerosol forcings were performed, including a control simulation without tropospheric aerosols, four simulations using different prescribed aerosol climatologies and a simulation using an artificial quadrupling of anthropogenic AOD. The latter was performed to provide an additional analysis of climate sensitivity for extreme aerosol loads (Sect. 3.5). Details on the aerosol climatologies are provided in the next section (Sect. 2.3). Table 1 summarizes the most important information on the underlying aerosol climatology and optical properties for the different simulations. The abbreviations in the table for the simulations are used throughout the paper.

Table 1

Simulations performed in this study and general information on the underlying aerosol climatologies. For each climatology, the European spatio-temporal mean optical properties in the mid-visible range at 550 nm are given for the aerosol types. Note that all simulations consistently include stratospheric background aerosols, namely stratospheric sulphate (AOD 0.045; SSA 1; g 0.73) and stratospheric ash (AOD 0.007; SSA 0.94; g 0.70). The abbreviations of the simulations are used throughout the paper

Abbr.

Description

Horizontal resolution

Temporal variability

References

Aerosol type

AOD

SSA

g

CTL

Control simulation

Tan_84

Default aerosol climatology

T10

None

Tanré et al. (1984)

Urban

0.050

0.64

0.60

   

Land

0.075

0.89

0.64

   

Sea salt

0.019

0.99

0.75

   

Desert dust

0.245

0.89

0.64

   

Background

0.030

0.89

0.64

   

Teg_97

Aerosol climatology of the IFS

T21

Monthly

Tegen et al. (1997)

Black carbon

0.017

0.82

0.70

   

Organic carbon

0.059

0.93

0.71

   

Sulphate

0.049

0.93

0.71

   

Sea salt

0.004

0.99

0.80

   

Desert dust

0.023

0.85

0.74

   

Aer_06

AeroCom aerosol climatology

1° × 1°

Monthly

Kinne et al. (2006)

Black carbon

0.016

0.82

0.70

   

Organic carbon

0.050

0.93

0.71

   

Sulphate

0.115

0.93

0.71

   

Sea salt

0.021

0.99

0.80

   

Desert dust

0.034

0.85

0.74

   

MAC

MPI aerosol climatology V.2

1° × 1°

Monthlya

Kinne et al. (2013)

MAC_4×

MAC, 4× higher anthrop. AOD

   

Anthropogenic

0.118

0.94

0.65

   

4× Anthrop.

0.472

0.94

0.65

   

Preindustrial

0.023

0.94

0.65

   

Coarse

0.031

0.96

0.75

   

aActually, MAC provides aerosol data varying monthly as well as interannually. To be consistent with the other simulations, a mean annual cycle for the period from 1950 to 2009 was used instead

The model domain covers the European continent, North Africa and the East Atlantic (Fig. 1). The horizontal grid distance is 0.44° (≈ 50 km) with 40 vertical layers up to 22 km and a time step of 360 s. The initial conditions, lateral boundary conditions and sea surface temperatures (SSTs) were obtained from the NCEP/NCAR-1 reanalysis data set (Kalnay et al. 1996; Kistler et al. 2001) with a spectral resolution of T62L28 (≈ 1.8°). All simulations cover the period from 1950 to 2009 with a prior spin-up time of 5 years, which was included to ensure balanced soil moisture fields. Therefore, the model is initialized on 01 January 1948 with a soil moisture field from a 3-year simulation (1948–1950), which is similar to the method described in Geyer (2014).
Fig. 1

Model domain with orography, including the sponge zone. The blue boxes mark the European subregions defined in the PRUDENCE project (Christensen and Christensen 2007) and a ninth region (“Europe” or “EU”) enclosing all subregions. The abbreviations are as follows: BI British Isles, IP Iberian Peninsula, FR France, ME Mid-Europe, SC Scandinavia, AL Alps, MD Mediterranean, EA Eastern Europe. The sponge zone of eight grid points at the lateral boundaries, were the RCM is disturbed by the constraint boundary conditions is included here, but discarded in all figures hereafter

Analyses of the results are frequently presented as spatial means (excluding grid points over the sea) for the eight European subregions (Fig. 1) used in the PRUDENCE project (Christensen and Christensen 2007). Additionally, a ninth region was defined (hereafter referred to as “Europe” or “EU”), and it encompasses all subregions.

We investigate model results obtained under clear-sky and all-sky conditions. All-sky conditions refer to all output time steps. The current official version of COSMO-CLM does not provide clear-sky radiative fluxes in the output. To overcome this limitation, clear-sky conditions are defined when the modelled total cloud cover is less than 5%. In comparisons between two model simulations, we only consider those time steps at each grid point for which this criterion is fulfilled simultaneously in both simulations. With this approach, a coherence between the all-sky and the clear-sky samples cannot be achieved, which makes direct comparisons delicate.

2.3 Aerosol climatologies

In this section, we provide an overview of the aerosol climatologies applied in our study. These climatologies employ fixed annual cycles of tropospheric aerosol loads that differ with respect to the analysis strategy and the time window for which they were designed. More precisely, each climatology was constructed originally for a certain emission period, while our simulations were performed for an equal multi-decadal period. Hence, emission and simulation years are not coherent, which lead to a conflation of effects. During the course of the simulations, major variations in the aerosol load have taken place in Europe, which is often referred to as dimming and brightening (e.g. Wild et al. 2005). This effect is neglected in our study. Nevertheless, our approach follows a standard practice in regional climate modelling.

In all climatologies, the effective radii of aerosol particles and thus the optical properties do not vary with the relative humidity. To ensure consistent results, stratospheric aerosols stemming from the default aerosol climatology (see Sect. 2.3.1) were adopted in all simulations independent of the aerosol climatology. This method ensures that only the effects resulting from the differences in tropospheric aerosol load are considered in our sensitivity experiments.

2.3.1 Aerosol climatology of Tanré et al. (1984)

In COSMO-CLM, the aerosol climatology of Tanré et al. (1984) is used by default to simulate the effect of aerosols on radiative fluxes. Aerosol optical properties for each spectral band are directly implemented in the radiation module for this climatology and are used if no deviating aerosol option is switched on. Originally, this climatology was developed for use in the low-resolution GCM at the ECMWF, although it has been implemented in many RCMs as well.

The climatology is based on the recommendations of the World Climate Research Program (WCP 1980). The spectral resolution is T10 (≈ 11.8°) and includes five types of aerosols in the troposphere (urban, land, sea salt, desert dust and background) and two types of aerosols in the stratosphere (volcanic sulphate and volcanic ash). Because of the expected long lifetime, volcanic and background aerosols are assumed to be horizontally homogeneous. All types are fixed in time with an exponential vertical decay. Concise information on the optical properties are listed in Table 1.

The climatology is characterized by a smooth and symmetric structure (Fig. 2) with a dominant desert dust component, especially in the mid-troposphere. The European average of the total AOD at 550 nm is approximately 0.42 with a desert dust amount of 0.24 (57%).
Fig. 2

Time mean tropospheric AODs (top) and European spatio-temporal mean vertical profiles of the extinction coefficients in the lowest 5 km (bottom) of the aerosol climatologies from left to right. The values are valid in the mid-visible range at 550 nm. Note the non-equidistant color bar of AOD for the high aerosol loads

2.3.2 Aerosol climatology of Tegen et al. (1997)

The aerosol climatology of Tegen et al. (1997) is an alternative global data set in COSMO-CLM to prescribe aerosols and it is used in the IFS at the ECMWF (Morcrette et al. 2009).

This climatology is a compilation of the results from different global chemistry transport models. Column-integrated AODs are calculated initially from the aerosol mass and the effective radius using a bulk formulation. The spectral resolution is T21 (≈ 5.6°) and the climatology includes carbonaceous aerosols (black carbon + organic carbon), sulphate aerosols, sea salt and soil dust. The values for sulphate aerosols are based on a simulation of Chin et al. (1996) for the year 1985. The data set is based on monthly mean values for all species to generate one annual cycle. However, the base year varies with the aerosol type. Because only the total column AODs are provided and read in by external parameters, the exponential vertical decay was adopted from the default aerosol climatology for the respective simulation.

In Europe, sulphate and organic carbon are the dominant aerosol components with highest values over Eastern and Southeastern Europe (Fig. 2) and maximum values in the summer season. The annual and European average of the total AOD at 550 nm is approximately 0.15 and it has an organic + sulphate amount of 0.11 (73%).

2.3.3 AeroCom aerosol climatology

Another more recent aerosol climatology was compiled within the Aerosol Model Intercomparison Project (AeroCom, http://aerocom.met.no) Phase 1 (Kinne et al. 2006), which is an international science initiative on aerosols and its impact on climate. Within this initiative coordinated model-intercomparisons are carried out and are evaluated against measurements.

A multi-model mean of 16 Global Chemistry Transport Models (GCTMs) with emission data for present-day conditions (year 2000) provides the aerosol optical properties for the five main aerosol types: black carbon, particulate organic matter (POM), sulphate, sea salt and dust. The data set has a 1° × 1° spatial resolution and includes monthly mean values for the base year 2000. Kinne et al. (2006) compared modelled AODs with satellite composites and the AERONET ground-based network and found a common overestimation of AODs over Europe in summer and they also emphasized the large model spread in terms of aerosol composition and the amount of black carbon causing large uncertainty in the radiative forcing. In our setup, the vertical distribution originates from the default aerosol climatology.

The AeroCom climatology is characterized by a strong contribution of anthropogenic aerosols over Central and Eastern Europe (Fig. 2), especially in late spring and summer. The annual average of the total AOD at 550 nm for Europe is about 0.24 with an organic + sulphate amount of 0.17 (71%).

2.3.4 MAC-v2 aerosol climatology

The Max-Planck-Institute aerosol climatology (MAC, Kinne et al. 2013) is a new comprehensive global aerosol data set that can be regarded as the successor to the AeroCom aerosol climatology. MAC-v2 refers to the updated MAC-v1 data set including e.g. coarse correction factors (see below).

MAC includes all of the necessary aerosol optical properties for the radiative transfer calculations (AOD, SSA, g) on a 1° × 1° grid at a monthly resolution. The general concept is based on monthly background aerosol data from the AeroCom phase 2 model mean for the year 2005, and these data were adjusted with coarse and local correction factors obtained from AERONET measurements. Aerosols are classified into coarse and fine modes, with the latter split into a pre-industrial and anthropogenic part. The vertical distribution is based on an ECHAM5-HAM2 (Stier et al. 2005; Zhang et al. 2012) simulation that employs the height-dependent fractional AOD contribution at each level. Inter-annual changes are only considered for anthropogenic aerosols by applying scaling factors from an ECHAM5-HAM2 simulation to current day conditions. For reasons of comparability with the other aerosol simulations, the inter-annual variability is neglected, but an annual climatology 〈1950–2009〉 is used. For this study, the height-dependent aerosol optical properties of MAC were interpolated to the COSMO-CLM grid and implemented into the radiation module.

The annual climatology shows large contributions of anthropogenic aerosols with maxima over Western and Eastern Europe (Fig. 2). Vertically, a well-mixed boundary layer with an exponential decay above is visible (Fig. 2). The climatological average of total AOD at 550 nm for Europe is about 0.17 with an anthropogenic amount of 0.12 (71%).

In addition to the original setup with MAC, an additional simulation (MAC_4×) with an artificial quadrupling of the anthropogenic AOD is performed.

2.4 Observational data

The model results are evaluated based on comparisons with the gridded data set EOBS, version 10.0 (Haylock et al. 2008). This data set includes more than 2300 European station measurements interpolated to a \(0.22^\circ \times 0.22^\circ \) regular lon-lat grid. In this study, we use data for daily mean, minimum and maximum temperature as well as daily precipitation sum. We also use monthly mean data from the global CRU database, version 3.2 (Jones and Harris 2011), with an original resolution of \(0.5^\circ \times 0.5^\circ \).

SSR is evaluated against a satellite-based climatology from the Surface Solar Radiation Data Set-Heliosat (SARAH, Müller et al. 2015), which is distributed by the Satellite Application Facility on Climate Monitoring (CM SAF). Hourly data are provided on a \(0.05^\circ \times 0.05^\circ \) regular lon-lat grid for the period from 1983 to 2013, which is derived from SEVIRI/MVIRI sensor images obtained by the geostationary METEOSAT satellite. The mean absolute bias for SSR against ground-based measurements is estimated at 8 W/m2 for the monthly mean and 20 W/m2 for the daily mean values. Additionally, quality-checked monthly mean in situ measurements from the Global Energy Balance Archive (GEBA, Gilgen and Ohmura 1999) at five European locations [Geneve (CH), Warschau (PL), Kolobrzeg (PL), Vlissingen (NL), Kronoberg (S)] are used.

Observational surface data are interpolated to the model grid with a bilinear approach. The same procedure is also used for the interpolation of the model results to station data.

3 Direct and semi-direct aerosol effects

This section focuses on generally valid mechanisms of direct and semi-direct aerosol effects that occur for all aerosol climatologies presented here. Also, the differences in magnitude of effects between the simulations are discussed. Hence, our analysis is based on a qualitative and comparative approach. The aerosol effect for each simulation is derived from the differences to the control simulation (CTL) driven without tropospheric aerosols. The statistical significance of a simulated signal is assessed with an one-sample t-test (von Storch and Zwiers 1999) based on the time-series of seasonal mean differences with the null hypothesis of zero mean. A standard two-sample t-test, testing the time-series against each other is not applicable here, because the condition of independence of the two samples is violated due to identical constraint boundary conditions and SSTs.

Arising from the highest aerosol load in Tan_84, we expect the strongest effects with this simulation. In particular, the high fraction of absorbing desert dust aerosols suggests a strong heating within the troposphere.

3.1 Near-surface quantities

Figure 3 shows the 60-year seasonal mean relative differences in SSR for the different aerosol climatologies. Only statistically significant differences at \(p\le 0.05\) from a two-sided t-test were considered here. Aerosol scattering and absorption lead to a distinct reduction in SSR in all simulations. The reduction is strongest in the default setup (Tan_84) and shows a spatial gradient from south to north caused by the high dust aerosol load and the spatial pattern of AOD in the Tanré et al. (1984) climatology. In Europe, the strongest and weakest reductions in SSR are evident over the Mediterranean area (annual mean −18%) and the British Isles (annual mean −9%), respectively. The other simulations show considerably weaker reductions in SSR and maxima over Central and Eastern Europe, which are consistent with the respective spatial patterns of AOD. For example, Aer_06 presents an annual mean maximum decrease in SSR over Eastern Europe (−8%), whereas MAC shows a maximum over Mid-Europe (−6%). The relative reduction in SSR has a seasonality in all simulations, with the strongest impact occurring in winter and/or autumn and the weakest impact occurring in spring and/or summer.
Fig. 3

Relative difference in the surface shortwave radiation (SSR) for Tan_84, Teg_97, Aer_06 and MAC (from left to right) compared with CTL. Seasonal mean values (from top to bottom) for the period from 1950 to 2009 are shown. Only statistically significant differences at \(p\le 0.05\) from a two-sided t-test were considered here

Absolute differences in SSR (not shown here) in all simulations are strongly influenced by the spatial and temporal astronomical availability of sunlight and different cloud cover situations, which mask the spatial patterns of AOD. Consequently, in all simulations, a gradient is prevalent from south to north with the strongest reduction in summer, reaching values of less than −50 W/m2 for the Tan_84 simulation over Northern Africa.

The strongest aerosol impact in terms of absolute values in summer is also visible in Fig. 4, which shows the aerosol responses in the surface energy budgets, temperature, total cloud cover and precipitation for Tan_84, Teg_97, Aer_06 and MAC for the European subregions. The negative shortwave radiative forcing at the surface is statistically significant in all regions in all simulations, and it reaches −40 W/m2 for the Tan_84 simulation over the Mediterranean area. This region of strongest decrease is also evident in the other simulations, albeit with half or less magnitudes. This also applies to the other subregions. The negative shortwave radiative forcing at the surface is mainly balanced by a reduction in sensible heat fluxes. Latent heat flux reductions appear to be important in Central Europe (BI, FR, ME, AL) and Scandinavia, partly reaching reductions that are similar to those of sensible heat fluxes (e.g. MAC for ME). However, distinct decreases in thermal emissions commonly occur over the Iberian Peninsula and the Mediterranean area but with weaker importance than the sensible heat flux changes in those regions. In winter, changes in energy fluxes are considerably smaller and only present stronger impacts in negative shortwave radiative forcings.
Fig. 4

Differences in the surface energy budgets (SRB solar radiation budget, TRB thermal radiation budget, H sensible heat flux, L latent heat flux; top row), temperature (middle row), and total cloud cover (TCC) and precipitation (P) (bottom row) for Tan_84, Teg_97, Aer_06 and MAC (from left to right) compared with CTL for the European subregions. Seasonal mean/sum values for summer (solid lines) and winter (dashed lines) for the period from 1950 to 2009 are shown. Mean upwelling fluxes (TRB, H, L) have negative signs and a reduction in these fluxes have positive signs. Filled circles as symbols indicate that the differences are not statistically significant (\(p>0.05\)) with a two-sided t-test for a subregion, whereas all others are statistically significant. Note the doubled scale for all quantities in the first column for Tan_84 compared to the other simulations

Temperature responses to different aerosol forcings are characterized by cooling at the surface caused by a decrease in solar irradiance that shifts into a warming with increasing height (represented by 850 hPa temperature here) because aerosol absorption becomes dominant. Because of the high aerosol load, Tan_84 shows the largest impact and a summer mean cooling that exceeds 0.5 K at 2 m and nearly 1 K at the surface over the Mediterranean area. The strongest warming at 850 hPa is evident in this simulation over Scandinavia at up to 0.3 K. Except for the Iberian Peninsula and the Mediterranean area, the effects on the 2 m temperature are similar in winter, whereas the drop in surface temperatures shows a much stronger seasonality. This indicates that the near-surface stabilization is considerably larger in summer, which is consistent with the stronger reduction in sensible heat fluxes. The effects on the 850 hPa temperatures are also weaker in winter. The shift from warming in summer to cooling in winter in 850 hPa, which is most pronounced over the Iberian Peninsula, is related to more powerful vertical cooling in summer that exceeds the 850 hPa height, whereas warming prevails at this height in winter. In general, the effects on temperatures in the Teg_97, Aer_06 and MAC simulations are qualitatively similar but at small magnitudes in all subregions. In particular, the changes in 2 m temperatures are negligible with a somewhat stronger cooling simulated over Eastern Europe (e.g. MAC in summer: −0.2 K).

The total cloud cover and precipitation decrease in all simulations with aerosol forcing. The reductions are strongest in Tan_84 with an absolute cloud cover decrease of approximately 1% similar in all subregions in summer, whereas the precipitation changes are largest in Central and Western Europe at more than 6 mm/month (O(10%)). In winter, the response to aerosol forcing is weaker, especially for precipitation. Exceptions are the regions in Southern Europe (IP, MD), which are characterized by distinct winter rain and thus stronger effects in terms of absolute values in winter. The effects in the other simulations show similar spatio-temporal characteristics, although the magnitudes are only half compared to that in Tan_84. A common peak of decreased precipitation occurs in all four simulations for Mid-Europe in summer.

In summary, all simulations show consistent values for the direct and semi-direct aerosol effects, i.e. an initial decrease in SSR, which causes a drop in surface temperature. At higher levels, the warming tendency from aerosol absorption becomes dominant. The resulting stabilization of stratification leads to a reduction in cloud cover and precipitation.

3.2 Vertical structure of the troposphere

The more detailed analysis of the vertical distributions of different meteorological quantities also provides valuable information on the direct and semi-direct aerosol effects. They are shown in Fig. 5 for two different subregions (IP, ME) in winter and summer. We have selected these two regions because they represent two different regions with pronounced and less pronounced semi-direct aerosol effects in particular in terms of changes in cloud properties. The details are discussed below.
Fig. 5

Differences in the clear-sky (dashed lines) and all-sky (solid lines) vertical profiles up to 5 km for Tan_84, Teg_97, Aer_06 and MAC (from left to right) compared with CTL for solar (yellow) and thermal (red) radiative heating rates, temperature (black) and cloud water + ice content (blue). Two subregions were chosen: a IP and b ME. Seasonal mean values for summer and winter for the period from 1950 to 2009 are shown. Note the doubled scales for all quantities in the first column for Tan_84 compared with that of the other simulations

Changes in the vertical profiles of radiative heating rates in both regions are characterized by the solar heating from aerosol absorption, which is strongest at the surface and shows exponentially declining values aloft. However, for MAC the deviating, well-mixed aerosol load within the boundary layer (see Sect. 2.3.4) is also visible in the solar heating rates. Caused by higher AODs (except in Tan_84) and stronger solar forcing in summer, the change in solar heating is considerably larger in both regions in summer compared with that in winter. The vertically more powerful solar heating with a maximum at a height of approximately 1.5 km for MAC in Mid-Europe in summer compared to a shallower profile with the maximum heating at roughly 500 m in winter, is also a consequence of the annual cycle in the aerosol extinction profile, which is soleley present in the MAC simulation. The changes in clear-sky solar heating rates over Mid-Europe in summer are larger than the all-sky values only near the surface and smaller above in all simulations; however, a stronger heating is expected in all layers caused by stronger aerosol absorption in the absence of clouds (clear-sky). Apparently, all-sky changes in clouds and cloud optical properties reinforce the aerosol-related solar heating. Thermal heating rates are dominantly negative with vertical profiles characterized by a more or less logarithmic decay with height, which is reasonable because the cooling signal originates at the surface and is transmitted upwards. For Mid-Europe in summer, thermal heating rates are even shifted into a warming signal at heights of approximately 2 km. Overall, changes in the thermal heating rates are considerably smaller than changes in the solar heating rates. Consequently, the warming tendency from aerosol absorption is unbalanced from a radiative-transfer point of view.

Changes in cloud water and ice content are most pronounced over Mid-Europe in summer, with an increase near the surface and a decrease above. For Tan_84, the increase and decrease show similar levels in terms of magnitude and vertical extension, whereas in the other simulations the decrease in cloud water and ice content dominates. The peak of strongest cloud water and ice reductions is located at a height of approximately 2 km in all four simulations.

Temperature profiles in Tan_84 over the Iberian Peninsula show a cooling trend near the surface that transforms into a warming with increasing height. Interestingly, the reversal point in summer is more than twice as high as that in winter. In addition, a column net cooling is prevalent in summer, whereas a net warming is visible in winter. The other simulations show similar behavior although at a small magnitude. A potential mechanism underlying these effect is the more horizontally than vertically advected cooling signal at the surface in winter due to weaker turbulent vertical mixing than in the summer season. Thus, the local heating from aerosol absorption might become more important in winter. Temperature profiles in Mid-Europe are more strongly affected by a warming above the surface in all simulations. In winter, we find a moderate near-surface cooling in Tan_84 and the strongest all-sky and clear-sky warming in MAC in the lowermost 1 km. This is consistent with the solar heating rates, although the temperature differences approach zero near the surface because of the cooling tendency caused by reduced SSR. Despite similar solar heating rates in the summer, the clear-sky and all-sky warming above the surface over Mid-Europe is stronger than that over the Iberian Peninsula in all simulations. This result is surprising because the aerosol composition (and thus SSA and g) is horizontally constant in Tan_84, Teg_97 and Aer_06. Only the MAC climatology has a stronger anthropogenic aerosol contribution over Central Europe, which could lead to stronger absorption and thus intensified solar heating. In addition, the all-sky temperatures above a height of 2 km in Mid-Europe in summer show a stronger warming than the clear-sky temperatures at this heights in all simulations, which is possibly related to reinforced solar heating from aerosol absorption due to reduced cloud water and ice content.

In conclusion, qualitatively similar effects on the vertical profiles of meteorological quantities were observed in all four simulations. The changes in vertical profiles over the Mediterranean area are comparable to those over the Iberian Peninsula, whereas all other subregions show similar results as discussed for Mid-Europe.

3.3 Circulation

We also find slight impacts on the circulation patterns in the simulations with aerosols. This is shown on the basis of mean sea level pressure (PMSL), geopotential height (H500–H1000 hPa) and 10 m wind for Tan_84 and MAC in winter and summer in Fig. 6. The magnitudes of effects in Teg_97 and Aer_06 are relatively similar to that in MAC. Thus, we only show the latter here.
Fig. 6

Differences in the mean sea level pressure (PMSL; filled contours), geopotential height (H500–H1000 hPa) (black contour lines), 10 m wind (arrows) for Tan_84 (left) and MAC (right) compared with CTL. Seasonal mean values for winter (top) and summer (bottom) for the period from 1950 to 2009 are shown. Values for the differences in the geopotential height are plotted every 1 gpm as solid (positive values) and dashed (negative values) black lines. The magnitude of difference wind vectors is given by arrow colors (\(\left| \Updelta \mathbf{v _h}\right| \)). Wind difference arrows are shown only, if their magnitude exceeds 0.1 m/s

A large-scale reduction in PMSL and increase in geopotential height is evident over large parts of Europe in all simulations, at least in summer. Here, the differences in PMSL are characterized by a three-pole structure with a primary maximum drop over the Baltic Sea region and secondary maxima over Southern Italy and the Black Sea.

The effects are strongest for Tan_84 in summer with pressure drops on the order of up to 1 hPa and increases in the geopotential height of more than 8 gpm (following the mentioned three-pole structure). Effects on 10 m winds reach up to 1 m/s over the Southern Mediterranean which is roughly 25% in terms of absolute wind speed in this region. In winter, the decrease (increase) in the PMSL (geopotential height) is considerably weaker, with the strongest effects over Southern Italy and the Black Sea. This also applies to 10 m winds.

In Teg_97, Aer_06 and MAC, the changes of the PMSL and the geopotential height show similar spatial patterns in summer as Tan_84, but with strongly reduced magnitudes. Weak impacts on wind vectors (> 0.1 m/s) are mainly present over the Mediterranean. In total, the ‘strongest’ effects occur in Aer_06 (not shown). In winter, the responses are negligible in all three simulations.

We the aim of getting a deeper insight into the underlying dynamical mechanisms, we investigated pattern correlations and temporal correlations between effects on column-integrated temperature (not shown), geopotential height and PMSL. Strong correlations (\(r> +0.8\)) between a column-integrated warming and an increase in geopotential height are present in all simulations with aerosols. This is physically consistent, because a column net warming leads to an increase in layer thickness which is equivalent to an increase in geopotential height. From a thermodynamic perspective, a spatially differing increase in layer thickness causes a decline in pressure at the location of the strongest heating. Our results are consistent with this, because the spatial pattern of the column net temperature increase correlates with the decrease in the pressure field. The pattern correlations are strongest (\(r\approx -0.7\)) for Tan_84 and weaker (\(r\approx -0.5\)) for the other simulations.

Undoubtedly, constrained lateral boundaries affect large-scale thermodynamic changes, that originate within the RCM domain [see e.g. Becker et al. (2015)]. Two sensitivity experiments with enlarged model domains lead to a reinforcement of reduction (increase) in PMSL (geopotential height) compared with the original domain setting, which indicates that the aerosol effects are limited because of the size of the model domain for these types of prescribed aerosol settings. Non-interactive but prescribed SSTs do not respond to the surface-based cooling signal caused by SSR decreases. Thus, the vertical total warming and the effects on PMSL and geopotential height might be overestimated in our setups over oceanic areas and adjacent continental areas.

3.4 Radiative forcing

Radiative forcing (RF) at the top of atmosphere (TOA) is frequently used to estimate the net effect of aerosols on Earth’s energy budget. Thus, the differences in radiative forcing for clear-sky and all-sky conditions during summer and winter between the aerosol simulations and the control simulation are shown in Fig. 7. The clear-sky RF at the TOA is positive in most regions on the European continent in winter. The strongest forcing is evident in all simulations over Eastern Europe (Tan_84: 6 W/m2; Aer_06: 2 W/m2). Over oceanic areas, the clear-sky RF is negative and mostly pronounced over the Mediterranean in all simulations. In summer, an overall negative clear-sky RF is evident and it is strongest over the Mediterranean (Tan_84) or the North Sea (MAC). In contrast, the high surface albedo over Northern Africa, which leads to an absorption of reflected sunlight, and longwave RF from dust absorption are responsible for a positive clear-sky RF in this area in all simulations. This effect was also reported by Nabat et al. (2015).
Fig. 7

Differences in the clear-sky (first and second row) and all-sky (third and fourth row) radiative forcing (RF) at top of atmosphere (TOA) for the Tan_84, Teg_97, Aer_06 and MAC simulations (from left to right) compared with CTL. Seasonal mean values for winter (each top row) and summer (each bottom row) for the period from 1950 to 2009 are shown

The all-sky RF in winter is positive over land and somewhat larger than the clear-sky RF in the Tan_84 simulation for Central, Western, and Southern Europe. The other simulations mainly show a reduction in positive RFs compared to the clear-sky values. In contrast to the clear-sky, the all-sky RFs in summer are positive for all subregions except for the Iberian Peninsula. A common region of strongest positive RF in all simulations is Mid-Europe. Probably, the changes in cloud properties in this region and the related reinforced solar heating as discussed in Sect. 3.2 is on cause for this phenomenon. Over the Mediterranean and the Black Sea, a negative all-sky RF is evident in summer. In these areas, the low sea surface albedo and related weak reflection of solar radiation reduces the aerosol absorption. In addition, constraint SSTs in our setting suppress semi-direct aerosol effects, so that e.g. changes in surface fluxes over the sea are much weaker than over land areas (not shown).

3.5 Sensitivity with an extreme aerosol load

As shown in the previous sections, the aerosol effects in all simulations show similar mechanisms, although, except in Tan_84, with a magnitude that is indeed statistically significant in most cases, but below the physical significance. To demonstrate the robustness of the results, a simulation with strong anthropogenic (quadruple) aerosol forcing based on MAC was performed (MAC_4×). The resulting aerosol loads create an annual mean AOD of approximately 1 over Central Europe. This value is rather unrealistic for Europe. However, such strong anthropogenic aerosol loads actually occur, such as at urban sites in South East Asia as reported by Che et al. (2015).

The condensed results for the different seasons are shown in Fig. 8 as scatterplots for the all-sky RF at TOA, 2 m temperature, total cloud cover and precipitation. Each dot shows the aerosol signal at a specific grid point within the European subregion as a function of the local AOD. The best linear fits obtained with the least squares method and the represented variances of the aerosol signal by the linear approach are shown as well. The spatial autocorrelation was not considered because of the large number of grid points that ensure a sufficient number of degrees of freedom. The all-sky RFs at TOA are predominantly positive and become larger with increasing AOD. The RFs are strongest in summer with a represented variance by the linear fit of 46%. The pool of negative RFs in summer can mainly be assigned to the Iberian Peninsula, where the cooling from the surface dominates the change in the vertical temperature profile. In autumn, only a weak positive all-sky RF AOD relationship is indicated. Generally, the robustness of the results discussed in Sect. 3.4 can be confirmed with the artificially extreme aerosol load in the MAC_4× simulation.
Fig. 8

Scatterplot of the differences (compared with CTL) in the all-sky RF at TOA, 2 m temperature, total cloud cover and precipitation (top to bottom, y-axes) against AOD (x-axes) for the MAC and MAC_4× simulations. Seasonal mean values (from left to right) for the period from 1950 to 2009 are shown. Each dot represents one grid point within the European subregion \((N=2349)\). The respective linear fits using the least squares method and related parameters are provided

Fig. 9

Reference all-sky SARAH SSR (Müller et al. 2015), Tan_84 bias, MAC bias and MBSS (Winterfeldt et al. 2010) of MAC versus Tan_84 (from left to right). Seasonal mean values (from top to bottom) for the period from 1983 to 2009 are shown. Filled points represent best coverage in situ GEBA (Gilgen and Ohmura 1999) measurements (1 Geneve, 2 Warschau, 3 Kolobrzeg, 4 Vlissingen, 5 Kronoberg). Grey-colored grid points indicate missing values

The 2 m temperature shows a decrease with increasing aerosol loads. The responses in winter, spring and summer are similar, whereby the spread is remarkably large in summer. In autumn, the temperature decrease with AOD is larger. Temperatures in autumn decrease by approximately 1 K per unit AOD with a represented variance of more than 50% in MAC_4×. Grid points with outstanding strong cooling, which are visible in MAC and MAC_4× in summer and autumn, are located over Eastern Europe, indicating a comparatively high sensitivity of temperatures to aerosol forcing in this region. Dividing the mean temperature in the daily minimum and maximum temperatures (not shown) indicates a distinct stronger cooling for the maximum compared with the minimum temperatures, which is approximately two times larger in spring and autumn and about 70% larger in winter and summer. The reduction of the daily temperature range by strong anthropogenic aerosol forcing has also been discussed by Makowski et al. (2008).

A general decrease in total cloud cover and precipitation is observed with increasing AOD. The reduction in cloud cover is strongest in winter, with a represented variance of two-thirds with MAC_4×. In spring and summer, the sensitivity is weaker but still with a good correlation between the modelled values and the linear fit. In autumn, the changes in cloud cover are weakest and show a considerably reduced correlation. Changes in precipitation are less robust, but the strongest and weakest impacts are in summer and autumn, respectively.

Linear regressions were also determined by using the non-parametric and outlier-insensitive Theil–Sen estimator (Theil 1992; Sen 1968), which indicated similar results. In conclusion, semi-direct aerosol effects appear to be most important in summer and only play a limited role in autumn, which is likely responsible for the stronger surface cooling during this season.

4 Performance of COSMO-CLM

This second part of the paper focuses on a comparison of COSMO-CLM model results with the observed climatology of different meteorological quantities for Europe, namely, SSR, 2 m temperature, total cloud cover and precipitation. Because the climatology of Tanré et al. (1984) is the default aerosol setting in COSMO-CLM, Tan_84 is considered to be the reference simulation. In addition, MAC with aerosols from Kinne et al. (2013) represents a simulation with updated aerosol data. Because the MAC, Teg_97 and Aer_06 simulations show qualitatively similar results (see Sect. 3), we only contrast here the default aerosol setting (Tan_84; Tanré et al. 1984) with the most up-to-date aerosol climatology (MAC; Kinne et al. 2013).

Comparisons are based on the seasonal mean bias, root-mean-square error (RMSE) and the modified Brier Skill Score [MBSS; Winterfeldt et al. (2010)]. The MBSS is derived from the Brier Skill Score as described by von Storch and Zwiers (1999) and estimates the skill of a simulation (F; here: MAC) relative to a reference simulation (R; here: Tan_84) in reproducing a reference data set (P; here: observation data) and is given by
$$\begin{aligned} &\sigma ^2_{FP} = \epsilon \left( (\mathbf F -\mathbf P )^2\right) \nonumber \\ &\sigma ^2_{RP}= \epsilon \left( (\mathbf R -\mathbf P )^2\right) \nonumber \\& \text {MBSS}= {\left\{ \begin{array}{ll} 1 - \sigma ^2_{FP}\sigma ^{-2}_{RP},\quad &{} \text {if } \quad \sigma ^2_{FP} \le \sigma ^2_{RP} \\ \sigma ^2_{RP}\sigma ^{-2}_{FP} - 1,\quad &{} \text {if } \quad \sigma ^2_{FP}> \sigma ^2_{RP}\end{array}\right. } \end{aligned}$$
(1)
The MBSS is based on mean square errors taking into account temporal variations and it ranges from −1 to +1. Under these circumstances positive values indicate an added value with the MAC simulation and negative values indicate a better performance with the default Tan_84 simulation in reproducing the observations. The magnitude of the MBSS gives the extent of the improvement with Tan_84 (negative signs) or MAC (positive signs).

4.1 Surface shortwave radiation

Figure 9 shows the seasonal mean SSR from the CM SAF [SARAH; Müller et al. (2015)] satellite retrievals and the best coverage in situ measurements from the Global Energy Balance Archive [GEBA, Gilgen and Ohmura (1999)] for the period from 1983 to 2009. The biases of the Tan_84 and MAC simulation against these data sets are presented. The MBSS is our metric to assess the added value of the MAC simulation over the Tan_84 simulation.

The default aerosol setup considerably underestimates the solar irradiance at the surface in most regions in all seasons. The strongest bias is evident in summer over Western, Central, and Eastern Europe and the eastern Mediterranean, with values below −40 W/m2 (e.g. ME 20%). In winter, a slight overestimation occurs in parts of Northern and Eastern Europe. Overall, the magnitude of biases over the European mainland follows the annual cycle of available sunlight at the surface (winter, autumn, spring, summer), where the lowest bias seasonality occurs over the Iberian Peninsula. The positive biases observed over the Norwegian mountains in winter and particularly in spring are probably caused by an insufficient representation of the steep orography at the 50 km model resolution and biases related to snow cover. However, also the satellite data are not reliable in mountainous snow-covered areas (Müller et al. 2015).

The reduced aerosol loads and changes in the aerosol composition in the MAC simulation lead to an overall increase in SSR compared to the Tan_84 simulation. Consequently, positive biases are induced or enhanced over Northern and Eastern Europe in winter, where negative MBSS values indicate no added value. A reduction of negative biases leads to an added value over Western Europe and the Iberian Peninsula, the Mediterranean and Northern Africa (positive MBSS values) in winter. In spring, summer and autumn, an added value occurs in all regions according to the MBSS because of the increase in SSR with MAC. The bias reductions and improvements are strongest in summer and show a gradient from southeast to northwest. This finding is consistent with the differences in spatial distribution of the AOD in the Tan_84 and MAC aerosol climatologies (see Fig. 2). For instance, comparatively low aerosol loads in MAC over the Eastern Mediterranean lead to an improvement in the so far (strong) underestimations of SSR, which generates MBSS values close to 1. In contrast, lower AOD in Tan_84 and a stronger anthropogenic aerosol load in MAC over the British Isles result in only weak bias reductions in this region. We also investigated different time periods with more available GEBA station measurements (e.g. 1990–1999), and the results for this periods show similar spatial and temporal characteristics as discussed above.

A further comparison between the modelled and satellite-observed SSR with a focus on the different ranges of solar irradiance within the individual subregions for the same period (1983–2009) during summer is shown in Fig. 10. We restrict the discussion to the summer season because the largest differences between the observations and model simulations as well as between the two simulations occur during this season. Consistent with the results discussed above, the spatial and seasonal averaged negative biases with the default setup are reduced distinctly or even shifted into weak overestimations (IP, MD) in the MAC simulation. The European mean bias is reduced by more than 50% in summer. However, the RMSEs are considerably larger for all subregions than the mean biases with only slight improvements observed in the MAC simulation. Obviously, spatial and day-to-day variations in SSR controlled by cloud cover are not fully captured by the model. A priori such errors can be expected, but indicate that prescribed monthly aerosol optical properties cannot compensate for these types of high-frequency errors. Quantile-quantile plots show that the underestimations observed with the default aerosol setup are largest for the low and medium values of SSR in most subregions, whereas a better performance (but still underestimating) in the Tan_84 simulation is observed towards the higher levels of solar irradiance. This finding indicates that the error in SSR is strongly related to the shortcoming of the model in reproducing the observed cloud cover situation, which will be discussed in greater detail in Sect. 4.3. In regions with a low mean cloud cover during summer (IP, MD), the largest errors occur for high levels of SSR.
Fig. 10

Scatterplots of the modelled all-sky SSR against the SARAH SSR (Müller et al. 2015) for the European subregions (see Fig. 1) during the summer season. Filled contours indicate the data density of Tan_84 for each daily mean value for the period from 1983 to 2009 and for each grid point of the respective subregion. Biases and RMSEs are provided for Tan_84 (black) and MAC (red). Additionally, quantile–quantile points are shown

Overall, improvements of SSR in the MAC simulation are marginal for low and medium values, but show a more pronounced improvement for high levels of SSR. Hence, the added value is most pronounced among the highest percentiles, where the underestimations of the Tan_84 simulation are reduced or even shifted into weak overestimations with MAC. This distribution is reasonable, because the values and difference in the AOD in the Tan_84 and MAC climatologies are negligible compared with the cloud optical depth under cloudy conditions (low percentiles). In turn, the AOD is more important for SSR under cloudless conditions (high percentiles). In spring and autumn (not shown here), we found qualitatively similar results at lower levels of SSR. In winter, qualitatively similar results at lower levels of SSR are observed in the southwestern part of Europe (e.g. IP), whereas weakly overestimations of SSR for low and medium percentiles and underestimations of SSR for high percentiles are observed in the northeastern part of Europe (e.g. EA). Thus, in winter, an added value with the MAC simulation is achieved for high percentiles only.

4.2 Temperature

Figure 11 shows the climatological 〈1950–2009〉 seasonal mean temperatures from the EOBS database (Haylock et al. 2008). Also, the biases of Tan_84 and MAC against these data are presented. Again, we use the MBSS to estimate the added value of the MAC simulation over the Tan_84 simulation.
Fig. 11

Reference EOBS (Haylock et al. 2008) 2 m temperature, Tan_84 bias, MAC bias and MBSS (Winterfeldt et al. 2010) of MAC versus Tan_84 (from left to right). Seasonal mean values (from top to bottom) for the period from 1950 to 2009 are shown. Grey-colored grid points indicate missing values

The well-known cold bias in RCMs and COSMO-CLM [e.g. Kotlarski et al. (2014); Geyer (2014)] is also prevalent in our default setup and shows similar magnitudes for all regions on annual mean (EU −0.44 K). On a seasonal mean basis, the spatial differences are larger. In winter, the largest cold bias is located over Southern Europe (MD −1.14 K), whereas a strong warm bias can be found over northern Scandinavia. The latter is related to the inadequate representation of snow cover and ice-albedo feedbacks. In spring, cold biases are relatively homogeneously distributed in Europe (e.g. EA −0.9). During the summer season, a bias gradient from south to north is visible with a warm bias over Southern Europe (MD +0.94 K) and a cold bias over Northern Europe (SC −1.2 K). The weakest (mainly cold) biases occur during autumn, and they only exceed −0.5 K over the Alps. Strong warm biases over the bright surfaces of Northern Africa, especially in summer, are caused by a known underestimation of surface albedo in this region (Kothe et al. 2013; Geyer 2014).

Reduced aerosol load and changed aerosol composition lead to a warming in the MAC simulation compared with Tan_84. Consequently, cold biases are generally reduced, whereas warm biases are enhanced. Except for Scandinavia, the cold biases are reduced in all regions in winter and spring. Broadly distributed low positive MBSS values indicate a weak improvement against the default setup. In summer, the warm bias, especially over the Balkans, increases. The cold bias over Central Europe is not reduced in summer, because the AOD and cooling in MAC are the largest in this region during summer, while the default aerosol setup provides a fixed annual mean value of AOD with an even weaker cooling in summer than in winter. In autumn, a shift towards positive biases with MAC is indicated over the southern half of Europe. A slight improvement occurs over northwestern Russia and parts of Scandinavia. In the annual mean, a weak added value for temperature remains over Central Europe, whereas the results over the Balkans are worsened.

COSMO-CLM is known to underestimate the diurnal temperature range, because, in particular, daily maximum temperatures are underestimated [e.g. Geyer (2014)], and these errors are also observed in our default setup. The comparison with the MAC simulation in terms of MBSS (Fig. 12) provides some interesting characteristics. Here, an improvement in the daily maximum temperatures is evident for nearly all subregions and all seasons, and it is largest in winter and autumn and weakest during summer. However, the simulation of daily minimum temperatures does not improve. This phenomenon is most pronounced in autumn. Overall, the location of most circles remains within the top-left octa of each subplot, indicating that the added value of the daily maximum temperature prevails. The results are reasonable, because aerosols have a stronger influence on daily maximum temperatures than on daily minimum temperatures as discussed in Sect. 3.5. Thus, the changed aerosol forcing in the MAC simulation leads to increased/improved daily maximum temperatures and weakly increased/worsened daily minimum temperatures.
Fig. 12

MBSS (Winterfeldt et al., 2010) of MAC versus Tan_84 for the daily minimum (x-axes) and daily maximum (y-axes) 2 m temperature. Seasonal mean values (from left to right) for the period from 1950 to 2009 are shown. Each filled circle represents one European subregion. Reference EOBS 12.0 (Haylock et al. 2008)

4.3 Cloud cover and precipitation

The simulated total cloud cover and precipitation are evaluated against data from the global CRU database, version 3.2 (Jones and Harris 2011) in Fig. 13. Based on the European and annual mean, a distinct overestimation of 18% in the total cloud cover is observed in the Tan_84 setup. An overestimation of the total cloud cover is also dominant for each season and all subregions, although it is weakest in winter over Central Europe and strongest in summer over Scandinavia and Eastern Europe (EA 18%). The annual cycle of cloud cover is not well captured by the model (except for IP and MD). With respect to the lowest observed mean cloud cover in summer during the year, the cloud cover is overestimated by 33% in Eastern Europe. The weakly enhanced cloud cover in the MAC simulation of approximately 1% is counterproductive but of little importance compared with the absolute error that is already present in the default setup. Thus, the MBSS values are negative but rather slight for all seasons. We found a strong negative temporal correlation between the total cloud cover and SSR in both simulations (\(r\approx 0.9\)). Thus, the underestimation of SSR for low and medium values as discussed in Sect. 4.1 is related to the overestimation of cloud cover.
Fig. 13

Total cloud cover (upper panel) and precipitation (lower panel) of the CRU (Jones and Harris 2011) observation data (grey), and the Tan_84 (black) and MAC (red) simulations for the European subregions. Seasonal mean values (from left to right) for the period from 1950 to 2009 are shown. Additionally, the MBSS (Winterfeldt et al. 2010) of MAC versus Tan_84 is indicated by red bars

The total precipitation with the default setup is mainly overestimated in winter and spring. In summer, a dry bias is prevalent in all subregions, except for Scandinavia. This bias is a common feature of relatively coarse resolved RCM simulations, because convective precipitation that dominates in summer is inadequately parameterized. In autumn, the European mean precipitation is simulated nearly perfectly with a counterbalancing dry bias in Western and Southern Europe, and a wet bias over Mid-Europe and Scandinavia. Reduced aerosol load and changes in aerosol composition in MAC cause an increase in precipitation in that simulation. Thus, the results worsen in winter and spring, whereas precipitation is reproduced slightly better in summer. Overall, the magnitude of changes caused by different aerosol forcing is larger than that for cloud cover, but still low compared to the bias already present in the default setup. Different reference data sets for cloud cover [ERA-40 (Uppala et al. 2005)] and precipitation [EOBS (Haylock et al. 2008); ERA-40 (Uppala et al. 2005)] yield qualitatively similar results.

5 Discussion

Long-term (1950–2009) COSMO-CLM RCM simulations with different aerosol forcing (Tanré et al. 1984; Tegen et al. 1997; Kinne et al. 2006, 2013) were performed and compared with respect to their direct and semi-direct aerosol effects. All simulations show a distinct reduction in surface shortwave radiation compared to the control simulation without tropospheric aerosols, and the relative changes are consistent with the respective patterns of aerosol optical depth. The absolute decrease in surface shortwave radiation is strongest in summer and mainly balanced by a weakening of sensible heat fluxes. In Mid-Europe, reductions in latent heat fluxes play a role in all simulations as well. The vertical temperature profiles are characterized by a shallow cooling near the surface because of the decrease in surface shortwave radiation, and a warming above caused by aerosol absorption. The resulting stabilization of stratification leads to a reduction in cloud cover and precipitation. Changes in the vertical profiles over Southern Europe are affected by stronger cooling in summer and stronger warming in winter, which might be related to a weaker turbulent mixing of the surface-based cooling signal in winter and the larger importance of local heating from aerosol absorption. In Central Europe, reductions in cloud water and ice content possibly reinforce aerosol absorption and tropospheric heating, particularly in summer. As a result, we also found weak impacts on circulation patterns in terms of a drop in surface pressure because of the spatial differing increase in layer thickness caused by the dominant mid-tropospheric warming. This was also reported by Toll et al. (2016) for short-term period NWP simulations with different aerosol forcing. The radiative forcing at the top of atmosphere, particularly in summer, is characterized by a negative clear-sky forcing and a positive all-sky forcing. This finding implies that semi-direct aerosol effects for instance in terms of changes in cloud properties are stronger than the direct aerosol effect. This mechanism appears to be most pronounced over Mid-Europe, although it is not pronounced over the Iberian Peninsula. The robustness of the results was demonstrated by a simulation with very strong aerosol forcing, which produced qualitatively similar results. In doing so, we found the weakest positive radiative forcing and changes in cloud cover and precipitation in autumn, although the strongest and most robust effects on 2 m temperature during this season, which indicated a weaker importance of semi-direct aerosol effects during autumn.

Our finding of positive all-sky radiative forcing at the top of atmosphere is in contrast to previous global climate model studies focusing on Europe and regional climate simulations [e.g. Nabat et al. (2015)]. The latter study is based on the aerosol climatology described in Nabat et al. (2013) with less absorption and less forward scattering compared with our aerosol climatologies. This might explain the stronger vertical warming and larger semi-direct aerosol effects in our study. The effects caused by the atmosphere–ocean coupling in Nabat et al. (2015) are mainly limited to the Mediterranean and adjacent areas, thus cannot explain the discrepancies in radiative forcing over the European mainland.

In the second part, the model was evaluated with regard to the default aerosol setting (Tanré et al. 1984) and an updated setting (Kinne et al. 2013). We found that the Tan_84 aerosol setup underestimates surface shortwave radiation, particularly in summer. Distinct improvements can be achieved with the more recent aerosol data set. The added value is largest for high levels of solar irradiance. Both simulations show very similar underestimations for low and medium values of solar irradiance, which is related to a shared considerable overestimation of cloud cover that is also strongest in summer. Increased temperatures in the updated aerosol setting add value by reducing cold biases that are present in the default setup in winter and spring, and enhancing or inducing warm biases, especially over the southern half of Europe in summer and autumn. Moreover, already present overestimations of cloud cover and precipitation in the initial setting are slightly enhanced. However, the lack of summer convective precipitation can be reduced slightly with updated aerosol data in most regions.

Our study emphasizes the importance and complexity of cause-effect-chains related to semi-direct aerosol effects, especially over Central Europe in summer. Consistent with Zubler et al. (2011a), we also suggest replacing the Tanré et al. (1984) aerosol climatology with more recent and realistic data sets. This more realistic aerosol settings not necessarily result in an overall better model performance since the model have undergone explicit and implicit tuning during its development. Indeed, the masking of model shortcomings due to a too strong aerosol forcing thus far can be prevented.

Future studies should include indirect aerosol effects since their magnitude can be of the same order as direct and semi-direct aerosol effects (Myhre et al. 2013). A parametrization of aerosol activation dependent on aerosol mass and upwind velocity has recently been implemented in the COSMO model and is currently being tested (Blahak 2016). Indirect aerosol effects possibly affect the magnitude of direct and semi-direct aerosol effects (and vice versa) by altering cloud properties. The strong link between the types of aerosol effects is discussed by e.g. Ghan (2013). In addition, the causes for the systematic overestimation of cloud cover need to be investigated. Currently, it is unclear to what extend this shortcoming affects direct, semi-direct and indirect aerosol effects.

Notes

Acknowledgements

This work is a contribution to the Helmholtz Climate Initiative REKLIM, a joint research project of the Helmholtz Association of German research centres (HGF). The CCLM is the community model of the German climate research (www.clm-community.eu). The NCEP/NCAR1 reanalysis data was provided by the National Center for Atmospheric Research (NCAR). We acknowledge the E-OBS dataset from the EU-FP6 project ENSEMBLES (http://ensembles-eu.metoffice.com) and the data providers in the ECA&D Project (http://www.ecad.eu). We would like to thank the Climatic Research Unit (CRU, http://www.cru.uea.ac.uk/) for providing observation data. The EUMETSAT Satellite Application Facility on Climate Monitoring (CM SAF) provided the Surface Solar Radiation Data Set—Heliosat (SARAH, https://wui.cmsaf.eu/safira/action/viewDoiDetails?acronym=SARAH_V001) for which we are thankful. Thanks to the Global Energy Balance Archive (GEBA) located at ETH Zurich for providing in situ radiation data. The authors would like to thank S. Kinne from the MPI-M Hamburg for providing the MAC-v2 aerosol data as well as for support during the implementation in COSMO-CLM. We are grateful to H. von Storch for valuable hints during preparation of the manuscript. We thank S. Wagner for fruitful discussions and proofreading of the paper, and the American Journal Experts (AJE) for English language editing. We thank two anonymous reviewers for comments that improved the manuscript.

References

  1. Ackerman AS, Toon OB, Stevens DE, Heymsfield AJ, Ramanathan V, Welton EJ (2000) Reduction of tropical cloudiness by soot. Science 288(5468):1042–1047. doi:10.1126/science.288.5468.1042 CrossRefGoogle Scholar
  2. Albrecht B (1989) Aerosols, cloud microphysics, and fractional cloudiness. Science 245(4923):1227–1230CrossRefGoogle Scholar
  3. Baldauf M, Seifert A, Förstner J, Majewski D, Raschendorfer M, Reinhardt T (2011) Operational convective-scale numerical weather prediction with the cosmo model: description and sensitivities. Mon Weather Rev 139(12):3887–3905. doi:10.1175/MWR-D-10-05013.1 CrossRefGoogle Scholar
  4. Becker N, Ulbrich U, Klein R (2015) Systematic large-scale secondary circulations in a regional climate model. Geophys Res Lett 42(10):4142–4149. doi:10.1002/2015GL063955 CrossRefGoogle Scholar
  5. Böhm U, Köcken M, Ahrens W, Block A, Hauffe D, Keuler K, Rockel B, Will A (2006) CLM - the climate version of LM: brief description and long-term applications. Tech. rep., COSMO Newsletter, vol 6. http://www.clm-community.eu/dokumente/upload/3a8e8_COSMOnewsLetter06_clm.pdf. Accessed 28 Oct 2009
  6. Blahak U (2016) Personal communicationGoogle Scholar
  7. Boucher O, Randall D, Artaxo P, Bretherton C, Feingold G, Forster P, Kerminen VM, Kondo Y, Liao H, Lohmann U, Rasch P, Satheesh S, Sherwood S, Stevens B, Zhang X (2013) Clouds and aerosols, vol 7. Cambridge University Press, Cambridge, pp 571–658. doi:10.1017/CBO9781107415324.016 Google Scholar
  8. Che H, Zhang XY, Xia X, Goloub P, Holben B, Zhao H, Wang Y, Zhang XC, Wang H, Blarel L, Damiri B, Zhang R, Deng X, Ma Y, Wang T, Geng F, Qi B, Zhu J, Yu J, Chen Q, Shi G (2015) Ground-based aerosol climatology of China: aerosol optical depths from the China aerosol remote sensing network (CARSNET) 2002–2013. Atmos Chem Phys 15(13):7619–7652. doi:10.5194/acp-15-7619-2015 CrossRefGoogle Scholar
  9. Chin M, Jacob DJ, Gardner GM, Foreman-Fowler MS, Spiro PA, Savoie DL (1996) A global three-dimensional model of tropospheric sulfate. J Geophys Res Atmos 101(D13):18,667–18,690. doi:10.1029/96JD01221 CrossRefGoogle Scholar
  10. Christensen J, Christensen O (2007) A summary of the prudence model projections of changes in European climate by the end of this century. Clim Change 81(1):7–30. doi:10.1007/s10584-006-9210-7 CrossRefGoogle Scholar
  11. Doms G, Schättler U (2002) A description of the nonhydrostatic regional model LM. Part I: Dynamics and numerics. Tech. rep., Deutscher Wetterdienst. www.cosmomodel.org/content/model/documentation/core/default.htm. Accessed 28 Oct 2009
  12. Doms G, Förstner J, Heise E, Herzog HJ, Mrionow D, Raschendorfer M, Reinhart T, Ritter B, Schrodin R, Schulz JP, Vogel G (2011) A description of the nonhydrostatic regional cosmo model. Part II: Physical parameterization. Tech. rep., Deutscher Wetterdienst. http://www.cosmo-model.org/content/model/documentation/core/cosmoPhysParamtr.pdf. Accessed 4 Sept 2014
  13. Ekman AML, Rodhe H (2003) Regional temperature response due to indirect sulfate aerosol forcing: impact of model resolution. Clim Dyn 21(1):1–10. doi:10.1007/s00382-003-0311-y CrossRefGoogle Scholar
  14. Geyer B (2014) High-resolution atmospheric reconstruction for Europe 1948–2012: coastdat2. Earth Syst Sci Data 6(1):147–164. doi:10.5194/essd-6-147-2014. http://www.earth-syst-sci-data.net/6/147/2014/. Accessed 27 May 2014
  15. Ghan SJ (2013) Technical note: estimating aerosol effects on cloud radiative forcing. Atmos Chem Phys 13(19):9971–9974. doi:10.5194/acp-13-9971-2013. https://www.atmos-chem-phys.net/13/9971/2013/. Accessed 10 July 2017
  16. Gilgen H, Ohmura A (1999) The global energy balance archive. Bull Am Meteorol Soc 80(5):831–850. doi:10.1175/1520-0477(1999)080<0831:TGEBA>2.0.CO;2 CrossRefGoogle Scholar
  17. Hansen J, Sato M, Ruedy R (1997) Radiative forcing and climate response. J Geophys Res Atmos 102(D6):6831–6864. doi:10.1029/96JD03436 CrossRefGoogle Scholar
  18. Haylock MR, Hofstra N, Tank AMGK, Klok EJ, Jones PD, New M (2008) A European daily high-resolution gridded data set of surface temperature and precipitation for 1950–2006. J Geophys Res. doi:10.1029/2008JD010201
  19. Haywood J, Boucher O (2000) Estimates of the direct and indirect radiative forcing due to tropospheric aerosols: a review. Rev Geophys 38(4):513–543. doi:10.1029/1999RG000078 CrossRefGoogle Scholar
  20. Hohenegger C, Vidale PL (2005) Sensitivity of the European climate to aerosol forcing as simulated with a regional climate model. J Geophys Res Atmos. doi:10.1029/2004JD005335
  21. Jaeger EB, Anders I, Luthi D, Rockel B, Schär C, Seneviratne SI (2008) Analysis of ERA40-driven CLM simulations for Europe. Meteorol Z 17(4):349–367. doi:10.1127/0941-2948/2008/0301 CrossRefGoogle Scholar
  22. Jones P, Harris I (2011) CRU time series (TS) high resolution gridded data version 3.10. NCAS British Atmospheric Data CentreGoogle Scholar
  23. Kalnay E, Kanamitsu M, Kistler R, Collins W, Deaven D, Gandin L, Iredell M, Saha S, White G, Woollen J, Zhu Y, Chelliah M, Ebisuzaki W, Higgins W, Janowiak J, Mo KC, Ropelewski C, Wang J, Leetmaa A, Reynolds R, Jenne R, Joseph D (1996) The NCEP/NCAR 40-year reanalysis project. Bull Am Meteorol Soc 77(3):437–471. http://rda.ucar.edu/datasets/ds090.0/docs/bams/bams1996mar/bamspapr-bm.pdf. Accessed 29 July 2009
  24. Kaufman YJ, Tanre D, Boucher O (2002) A satellite view of aerosols in the climate system. Nature 419(6903):215–223. doi:10.1038/nature01091 CrossRefGoogle Scholar
  25. Kessler E (1969) On the distribution and continuity of water substance in atmospheric circulations. Meteor Monogr 32(10):1–84Google Scholar
  26. Kinne S, Schulz M, Textor C, Guibert S, Balkanski Y, Bauer SE, Berntsen T, Berglen TF, Boucher O, Chin M, Collins W, Dentener F, Diehl T, Easter R, Feichter J, Fillmore D, Ghan S, Ginoux P, Gong S, Grini A, Hendricks J, Herzog M, Horowitz L, Isaksen I, Iversen T, Kirkevåg A, Kloster S, Koch D, Kristjansson JE, Krol M, Lauer A, Lamarque JF, Lesins G, Liu X, Lohmann U, Montanaro V, Myhre G, Penner J, Pitari G, Reddy S, Seland O, Stier P, Takemura T, Tie X (2006) An aerocom initial assessment—optical properties in aerosol component modules of global models. Atmos Chem Phys 6(7):1815–1834. doi:10.5194/acp-6-1815-2006. http://www.atmos-chem-phys.net/6/1815/2006/. Accessed 23 Feb 2016
  27. Kinne S, O’Donnel D, Stier P, Kloster S, Zhang K, Schmidt H, Rast S, Giorgetta M, Eck TF, Stevens B (2013) Mac-v1: a new global aerosol climatology for climate studies. J Adv Model Earth Syst 5(4):704–740. doi:10.1002/jame.20035 CrossRefGoogle Scholar
  28. Kistler R, Kalnay E, Collins W, Saha S, White G, Woollen J, Chelliah M, Ebisuzaki W, Kanamitsu M, Kousky V, van den Dool H, Jenne R, Fiorino M (2001) The NCEP–NCAR 50-year reanalysis: monthly means CD-ROM and documentation. Bull Am Meteorol Soc 82:247–267CrossRefGoogle Scholar
  29. Knote C, Heinemann G, Rockel B (2010) Changes in weather extremes: assessment of return values using high resolution climate simulations at convection-resolving scale. Meteorol Z 19(1):11–23. doi:10.1127/0941-2948/2010/0424 CrossRefGoogle Scholar
  30. Koch D, Del Genio AD (2010) Black carbon semi-direct effects on cloud cover: review and synthesis. Atmos Chem Phys 10(16):7685–7696. doi:10.5194/acp-10-7685-2010 CrossRefGoogle Scholar
  31. Kothe S, Lüthi D, Ahrens B (2013) Analysis of the West African monsoon system in the regional climate model COSMO-CLM. Int J Climatol 34(2):481–493. doi:10.1002/joc.3702 CrossRefGoogle Scholar
  32. Kotlarski S, Keuler K, Christensen OB, Colette A, Déqué M, Gobiet A, Goergen K, Jacob D, Lüthi D, van Meijgaard E, Nikulin G, Schär C, Teichmann C, Vautard R, Warrach-Sagi K, Wulfmeyer V (2014) Regional climate modeling on European scales: a joint standard evaluation of the EURO-CORDEX RCM ensemble. Geosci Model Dev 7(4):1297–1333. doi:10.5194/gmd-7-1297-2014 CrossRefGoogle Scholar
  33. Lohmann U, Feichter J (2005) Global indirect aerosol effects: a review. Atmos Chem Phys 5(3):715–737. doi:10.5194/acp-5-715-2005 CrossRefGoogle Scholar
  34. Makowski K, Wild M, Ohmura A (2008) Diurnal temperature range over europe between 1950 and 2005. Atmos Chem Phys 8(21):6483–6498. doi:10.5194/acp-8-6483-2008 CrossRefGoogle Scholar
  35. Müller R, Pfeifroth U, Träger-Chatterjee, Christine, Cremer R, Trentmann J, Hollmann R (2015) Surface Solar Radiation Data Set - Heliosat (SARAH) - Edition 1. Satellite Application Facility on Climate Monitoring. Tech. rep., EUMETSAT Satellite Application Facility on Climate Monitoring (CM SAF): Offenbach, Germany. doi:10.5676/EUM_SAF_CM/SARAH/V001
  36. Morcrette JJ, Boucher O, Jones L, Salmond D, Bechtold P, Beljaars A, Benedetti A, Bonet A, Kaiser JW, Razinger M, Schulz M, Serrar S, Simmons AJ, Sofiev M, Suttie M, Tompkins AM, Untch A (2009) Aerosol analysis and forecast in the European centre for medium-range weather forecasts integrated forecast system: forward modeling. J Geophys Res Atmos. doi:10.1029/2008JD011235
  37. Myhre G, Shindell D, Breon FM, Collins W, Fuglestvedt J, Huang J, Koch D, Lamarque JF, Lee D, Mendoza B, Nakajima T, Robock A, Stephens G, Takemura T, Zhang H (2013) Anthropogenic and natural radiative forcing. Cambridge University Press, Cambridge. doi:10.1017/CBO9781107415324.018 Google Scholar
  38. Nabat P, Somot S, Mallet M, Chiapello I, Morcrette JJ, Solmon F, Szopa S, Dulac F, Collins W, Ghan S, Horowitz LW, Lamarque JF, Lee YH, Naik V, Nagashima T, Shindell D, Skeie R (2013) A 4-d climatology (1979–2009) of the monthly tropospheric aerosol optical depth distribution over the Mediterranean region from a comparative evaluation and blending of remote sensing and model products. Atmos Meas Tech 6(5):1287–1314. doi:10.5194/amt-6-1287-2013 CrossRefGoogle Scholar
  39. Nabat P, Somot S, Mallet M, Sevault F, Chiacchio M, Wild M (2015) Direct and semi-direct aerosol radiative effect on the Mediterranean climate variability using a coupled regional climate system model. Clim Dyn 44(3):1127–1155. doi:10.1007/s00382-014-2205-6 CrossRefGoogle Scholar
  40. Ritter B, Geleyn JF (1992) A comprehensive radiation scheme for numerical weather prediction models with potential applications in climate simulations. Mon Weather Rev 120(2):303–325. doi:10.1175/1520-0493(1992) 120 0303:ACRSFN 2.0.CO;2 CrossRefGoogle Scholar
  41. Rockel B, Will A, Hense A (2008) The regional climate model COSMO-CLM (CCLM). Meteorol Z 17(4):347–348. doi:10.1127/0941-2948/2008/0309 CrossRefGoogle Scholar
  42. Sen PK (1968) Estimates of the regression coefficient based on Kendall’s tau. J Am Stat Assoc 63(324):1379–1389. doi:10.1080/01621459.1968.10480934 CrossRefGoogle Scholar
  43. Stier P, Feichter J, Kinne S, Kloster S, Vignati E, Wilson J, Ganzeveld L, Tegen I, Werner M, Balkanski Y, Schulz M, Boucher O, Minikin A, Petzold A (2005) The aerosol-climate model ECHAM5-HAM. Atmos Chem Phys 5(4):1125–1156. doi:10.5194/acp-5-1125-2005 CrossRefGoogle Scholar
  44. Tanré D, Geleyn JF, Slingo J (1984) First results of the introduction of an advanced aerosol-radiation interaction in the ECMWF low resolution global mode. Aerosols and their climatic effects, pp 133–177Google Scholar
  45. Tegen I, Hollrig P, Chin M, Fung I, Jacob D, Penner J (1997) Contribution of different aerosol species to the global aerosol extinction optical thickness: estimates from model results. J Geophys Res Atmos 102(D20):23,895–23,915. doi:10.1029/97JD01864 CrossRefGoogle Scholar
  46. Theil H (1992) A rank-invariant method of linear and polynomial regression analysis. Springer, Netherlands. doi:10.1007/978-94-011-2546-8_20 CrossRefGoogle Scholar
  47. Toll V, Gleeson E, Nielsen K, M”annik A, Masek J, Rontu L, Post P (2016) Impacts of the direct radiative effect of aerosols in numerical weather prediction over Europe using the ALADIN-HIRLAM NWP system. Atmos Res 172–173:163–173. doi:10.1016/j.atmosres.2016.01.003 CrossRefGoogle Scholar
  48. Twomey S (1977) The influence of pollution on the shortwave albedo of clouds. J Atmos Sci 34(7):1149–1152. doi:10.1175/1520-0469(1977)034<1149:TIOPOT>2.0.CO;2 CrossRefGoogle Scholar
  49. Uppala S, Kallberg P, Simmons A, Andrae U, da Costa Bechtold V, Fiorino M, Gibson J, Haseler J, Hernandez A, Kelly G, Li X, Onogi K, Saarinen S, Sokka N, Allan R, Andersson E, Arpe K, Balmaseda M, Beljaars A, van de Berg L, Bidlot J, Bormann N, Caires S, Chevallier F, Dethof A, Dragosavac M, Fisher M, Fuentes M, Hagemann S, Holm E, Hoskins B, Isaksen L, Janssen P, Jenne R, McNally A, Mahfouf JF, Morcrette JJ, Rayner N, Saunders R, Simon P, Sterl A, Trenberth K, Untch A, Vasiljevic D, Viterbo P, Woollen J (2005) The ERA-40 re-analysis. Q J R Meteor Soc 131:2961–3012. http://www3.interscience.wiley.com/journal/113522684/abstract?CRETRY=1&SRETRY=0. Accessed 14 Feb 2007
  50. Vogel B, Vogel H, Bäumer D, Bangert M, Lundgren K, Rinke R, Stanelle T (2009) The comprehensive model system COSMO-ART—radiative impact of aerosol on the state of the atmosphere on the regional scale. Atmos Chem Phys 9(22):8661–8680. doi:10.5194/acp-9-8661-2009 CrossRefGoogle Scholar
  51. von Storch H, Zwiers F (1999) Statistical analysis in climate research. Cambridge University Press, CambridgeCrossRefGoogle Scholar
  52. Wild M, Gilgen H, Roesch A, Ohmura A, Long CN, Dutton EG, Forgan B, Kallis A, Russak V, Tsvetkov A (2005) From dimming to brightening: Decadal changes in solar radiation at earth’s surface. Science 308(5723):847–850. doi:10.1126/science.1103215. http://science.sciencemag.org/content/308/5723/847
  53. Winterfeldt J, Geyer B, Weisse R (2010) Using QuikSCAT in the added value assessment of dynamically downscaled wind speed. Int J Climatol 31(7):1028–1039. doi:10.1002/joc.2105 CrossRefGoogle Scholar
  54. World Climate Research Program (1980) Aerosols and climate, vol 12. World Meteorological Organisation, GenevaGoogle Scholar
  55. Zhang K, O’Donnell D, Kazil J, Stier P, Kinne S, Lohmann U, Ferrachat S, Croft B, Quaas J, Wan H, Rast S, Feichter J (2012) The global aerosol-climate model ECHAM-HAM, version 2: sensitivity to improvements in process representations. Atmos Chem Phys 12(19):8911–8949. doi:10.5194/acp-12-8911-2012 CrossRefGoogle Scholar
  56. Zubler EM, Lohmann U, Lüthi D, Schär C (2011a) Intercomparison of aerosol climatologies for use in a regional climate model over Europe. Geophys Res Lett. doi:10.1029/2011GL048081
  57. Zubler EM, Lohmann U, Lüthi D, Schär C, Muhlbauer A (2011b) Statistical analysis of aerosol effects on simulated mixed-phase clouds and precipitation in the Alps. J Atmos Sci 68(7):1474–1492. doi:10.1175/2011JAS3632.1 CrossRefGoogle Scholar

Copyright information

© Springer-Verlag GmbH Germany 2017

Authors and Affiliations

  1. 1.Institute of Coastal ResearchHelmholtz-Zentrum GeesthachtGeesthachtGermany

Personalised recommendations