Interactive comment on “ Effects of aerosols on solar radiation in the ALADIN-HIRLAM NWP system ”

The paper addresses relevant scientific questions within the scope of ACP. The authors do not clearly support to present novel concepts, ideas, tools, and data, but they are interested about an important topic, i.e. the sensitivity of shortwave model outputs on aerosol and relative humidity. The overall presentation is neither well-structured nor clear; improvements in the structure and the content of each section are needed. There are three main issues that need careful attention prior to publication: 1) the language is not always fluent and precise. In particular, several phrases remain colloquial and simplistic. The repetition of the same words within the same phrase has not been eliminated. Indicative paradigms are highlighted in yellow or commented or replaced within the attached file, 2) there description of the tools and methods used is not precise and


Introduction
High resolution numerical weather prediction (NWP) models resolve details about local weather.Accurate treatment of the interactions of radiation with clouds and surfaces, and the direct and indirect effects of aerosols, is therefore important.Shortwave (SW) radiation in particular strongly impacts on weather at the surface and the development of the atmospheric boundary layer.Therefore, improvements in the representation of SW radiation are important for the ongoing improvement of NWP forecasts.Accurate SW radiation output from short-range NWP models relies on the accuracy of (1) the assumed amount and distribution of gases, cloud (liquid and ice) and aerosol particles, (2) the parametrization of the cloud, gas and aerosol inherent optical properties (IOPs) and (3) the approximations of radiative transfer through the atmospheric layers.We assumptions and clear-sky gas transmittance in NWP models (Nielsen et al., 2014;Gleeson et al., 2015).This study focusses on the direct radiative influence of aerosols.Climatological distributions of aerosols are commonly used in present-day operational NWP models when computing the direct radiative effect of aerosols.For example, monthly aerosol climatologies following Tegen et al. (1997) are assumed in ECMWF's (the European Centre for Medium Range Weather Forecasts) global Integrated Forecast System, IFS, and in the ALADIN-HIRLAM limited area modelling system used in this study.However, including a more complete representation of the effects of aerosols in NWP models can make the numerical weather forecast more accurate and is an active area of research (e.g.Mulcahy et al., 2014).Milton et al. (2008) showed that excluding the direct radiative effect of mineral dust and biomass burning aerosols during the dry season in West Africa resulted in an inaccurate representation of the surface energy budget and a warm bias in screen level temperature in forecasts using the UK Met Office Unified Model.Carmona et al. (2008) presented significant correlations between observed aerosol optical depth (AOD) and NWP model temperature errors.Pérez et al. (2006) also demonstrated that including the radiative effects of mineral dust in meteorological simulations can improve the representation of radiative balance, atmospheric temperature and mean sea-level pressure in numerical weather forecasts.Toll et al. (2015b) showed that the accuracy of ALADIN-HIRLAM forecasts were improved during severe wildfires in summer 2010 in Eastern Europe when the direct radiative effect of the realistic aerosol distribution, as opposed to the Tegen et al. (1997) monthly averages, was included in the model hindcasts.Overall, it is important to include the direct radiative effect of aerosols in order to improve the forecasting of direct and diffuse irradiance for solar energy applications because under clear-sky conditions aerosols mainly affect the SW radiation budget (Breitkreuz et al., 2009).
The IOPs (mass extinction, single scattering albedo (SSA) and asymmetry factor (g) and the direct radiative effect of aerosols depend on the physical properties of the aerosols.Changes in the IOPs of different aerosols types, induced by hygroscopic growth, change the radiative effect of aerosols (e.g.Cheng et al., 2008;Bian Introduction Conclusions References Tables Figures

Back Close
Full  et al., 2009;Markowicz et al., 2003;Zieger et al., 2013).For example, Magi and Hobbs (2003) present measurements of enhanced backscatter by biomass burning aerosols when the relative humidity is high.Pilinis et al. (1995) estimated that the global forcing due to aerosols would double for a relative humidity increase from 40 to 80 %.The direct radiative effect of aerosols is also dependent on the albedo of the underlying surfaces.Absorbing aerosols may induce a net cooling or a net warming effect depending on the surface albedo and the clouds below the aerosol layer.Chand et al. (2009) estimated that the regional warming induced by absorbing aerosols is three times higher when the covariation for aerosols and cloud cover is included in the calculation.Absorbing aerosols considerably modify the radiative heating rates in the aerosol layer and the vertical profile of temperature.For example Ramanathan et al. (2007) estimated that the 50 year average warming resulting from the so-called atmospheric brown cloud over Asia to be more than 0.5 K over a large area at the 700 hPa level.
Another important aspect to take into account when estimating the direct radiative effect of aerosols is the vertical profile of the aerosols.There are considerable variations in the vertical distributions of aerosols over Europe.Therefore, inaccuracies result from using constant climatological vertical profiles for different aerosol species as is the case in the ALADIN-HIRLAM system.For example, Guibert et al. (2005) analysed the vertical profiles of aerosol extinction over Europe and found that aerosols over southern Europe were concentrated higher in the atmosphere due to the occurence of dust episodes.Meloni et al. (2005) showed that under clear-sky conditions the direct radiative effect of aerosols on radiation at the surface has little dependence on the aerosol vertical profile, but that the vertical profile has an impact on the top of the atmosphere forcing, especially for absorbing aerosols.
The main goal of the present study is to compare the direct SW effect of aerosols in the ALADIN-HIRLAM system using three radiation schemes of varying complexity, with particular focus on an August 2010 Russian wildfire case study.We focus on the direct SW effect of aerosols which is more pronounced than the longwave (LW) effect.MUSC Introduction

Conclusions References
Tables Figures

Back Close
Full (Model Unifie Simple Colonne, Malardel et al., 2006), the single column version of three ALADIN-HIRLAM model configurations is used in the sensitivity experiments.Singlecolumn models are extremely useful for developing and testing physical parametrizations and for running idealised experiments that focus on atmospheric physics in a simplified framework.The sensitivity of the direct radiative effect of aerosols to the aerosol vertical profile, relative humidity and the spectral dependence of AOD is studied.Such a study is necessary in order to evaluate the strengths and weakness of the model regarding the treatment of the direct radiative effect of aerosols and for preparation of the model for utilisation of real-time aerosol information.Currently the model relies on the use of external aerosols in the form of monthly climatologies of aerosol optical depth (AOD) at 550 nm and parametrized IOPs (SSA, g and AOD scaling) to estimate the direct effect of aerosols on radiation.In the future, the model could obtain aerosol information from chemical transport model simulations, possibly coupled to the NWP model or from the analysis of aerosol observations.The paper is outlined as follows: the model set-ups are described in Sect.2; the datasets and experiments carried out are detailed in Sect.3; the results and discussion are presented in Sect. 4 and conclusions and future work are summarised in Sect. 5.

ALADIN-HIRLAM configurations and their single column version MUSC
The basic configuration of the ALADIN-HIRLAM system used in this study was HARMONIE-AROME based on Seity et al. (2011).In this, the acronym HARMONIE (HIRLAM ALADIN Regional Mesoscale Operational NWP in Europe) is used to denote the specific configuration of the ALADIN-HIRLAM NWP system maintained by the international HIRLAM programme.HARMONIE-AROME uses ALADIN non-hydrostatic dynamics (Bénard et al., 2010), Meso-NH physics (Mascart and Bougeault, 2011) and the SURFEX externalised surface scheme (Masson et al., 2013).Its default set-up for Introduction

Conclusions References
Tables Figures

Back Close
Full operational NWP uses a 2.5 km horizontal grid where deep convection is treated explicitly.Surface physiographies are prescribed using the 1 km resolution ECOCLIMAP database and surface elevation is based on GTOPO30 (USGS, 1998).The ALADIN-HIRLAM system can also be used for regional climate simulations (Lindstedt et al., 2015), where the direct radiative effect of aerosols can be of greater importance than in short range NWP applications.
The single column version of the ALADIN-HIRLAM model configurations (MUSC) includes all of the atmospheric and surface parametrizations but excludes large-scale dynamics, horizontal advection, the pressure gradient force and large-scale vertical motion.However, the excluded terms can be estimated by prescribed forcings.Because of the simplifying assumptions, a single-column model is not suitable for real forecasting.Its value lies in the fact that it provides the possibility to study the sensitivity of the model to different formulations of the physical parametrizations for realistic atmospheric conditions.
The time integration in MUSC is done on a single atmospheric column with 65 hybrid model levels.The inputs to MUSC include the initial conditions in the atmosphere, the initial conditions at the surface, surface properties and possibly atmospheric forcing data (i.e.temperature, specific humidity and wind speed forcings) applied during the MUSC experiment.These can be derived from the output from a 3-D model experiment.
In our experiments the initial state for MUSC was obtained from hourly snapshots of a 3-D HARMONIE-AROME simulation.Single time-step MUSC experiments were then run using each snapshot, which is sufficient for the diagnostics of radiative transfer.

Aerosols in MUSC
Regarding aerosols, monthly climatologies of vertically integrated AOD at 550 nm (Tegen et al., 1997)  IOPs are used directly by the IFS radiation scheme but are spectrally averaged over the six IFS SW intervals before being used by HLRADIA and ACRANEB2.Climatological vertical profiles of the aerosol types are assumed to distribute the aerosols on model levels (Tanre et al., 1984) which results in an exponential decrease in the aerosol attenuation coefficients with height.The spatial distribution of AOD is used to calculate the direct radiative effect of aerosols whereas the indirect radiative effect of aerosols has not yet been extensively studied in the system.
Each scheme treats the atmosphere as a 1-D column, consisting of a set of planeparallel homogeneous layers.All grid boxes are split into a cloudy fraction and a clearsky fraction with no lateral exchanges between them.The three schemes vary in complexity.The IFS scheme is the most complex of the three, consisting of several SW and LW radiation spectral bands.Its complexitity means that it is also computationally slow and is thus called intermittently in an operational forecasting set-up (by default the frequency of radiation computations is every 15 min in HARMONIE-AROME).The HLRADIA scheme is the simplest of the three schemes.As well as being broadband in SW and LW, the radiative transfer equations are parametrized which makes the scheme very fast.The complexity of ACRANEB2 lies between that of IFS and HLRADIA.It is also a broadband scheme but posesses selective intermittency where slowly-varying gaseous transmissions are updated on a time-scale of hours but quickly-varying cloud IOPs, for example, are recomputed at each model time-step.Introduction

Conclusions References
Tables Figures

Back Close
Full Atmospheric composition (i.e.aerosols, clouds and atmospheric gases) and the radiative properties of the surface are needed as input for each of the radiation schemes.
3-D cloud liquid and cloud ice particle concentrations and fractional cloud cover used by the radiation schemes are provided by the corresponding parametrization schemes and model dynamics.However, in the sensitivity tests presented in this paper the model was run under clear-sky conditions.
Monthly climatologies of ozone and a fixed composition mixture of CO 2 , N 2 O, CH 4 and O 2 are used by IFS and ACRANEB2 but the impact of ozone, oxygen and carbon dioxide on SW irradiance in HLRADIA is assumed to be constant over time and space.
Diffuse and direct surface albedos are provided by the SURFEX surface module.
Albedos for 1 ultraviolet, 1 visible and 1 infrared SW band are available from SURFEX and remapped to the 6 bands in IFS.These albedos are spectrally averaged over the six IFS SW intervals for use in ACRANEB2 and HLRADIA.In fact, HLRADIA only uses the broadband diffuse albedo because in this scheme the direct albedo is computed by adding the direct albedo correction factor (i.e.0.2/(1 + cos(SZA)) − 0.12) to the diffuse albedo (Wyser et al., 1999).In the version of MUSC used in the sensitivity tests, the effects of orography on radiation are not taken into account.
Further details on some of the differences between the three radiation schemes are given in the following sub-sections.
There cloud ice IOP parametrization.The cloud water load is modified by so-called cloud SW and cloud LW inhomogeneity factors before being used for radiation calculations, each of which is set to 0.7 in the default set-up.For detailed tests of the IFS SW cloud physics in the model see Nielsen et al. (2014), who suggested an improved SW cloud IOP scheme, the removal of the cloud inhomogeneity factor, and that the Fu SW cloud ice IOP scheme (Fu, 1996) be used.We followed these suggestions in the tests presented in this paper.
The SW radiative transfer for the clear-sky fraction is calculated using the two-stream equations (Fouquart and Bonnel, 1980) explained by Mascart and Bougeault (2011) where the reflectance and transmittance of the atmospheric layers is calculated similar to Coakley Jr. and Chylek (1975).The scheme uses IOPs (optical depth, SSA, g) of clouds, aerosols and molecules as input where the IOPs of clouds depend on the cloud droplet effective radius (r e,water ) and cloud ice effective dimension (d e,ice ).In IFS the Martin et al. (1994) scheme is used to calculate r e,water .The Ou and Liou (1995) scheme is the default scheme for cloud ice effective dimension where r e,ice = 0.5d e,ice (i.e. the crystals are assumed to be spherical).However, the advance Sun and Rikus (2000,2001) scheme is also available; this scheme assumes hexagonal ice cyrstals (i.e.r e,ice = 0.6435d e,ice ).On the other hand, the two-stream delta-Eddington approximation (Joseph et al., 1976;Fouquart and Bonnel, 1980) is used for the cloudy sky fraction radiative transfer calculations.Thus, the radiative transfer calculations in the clear-sky fractions are different to the radiative transfer calculations in the cloudy sky fractions.The results from tests of both of these radiative transfer approximations are detailed in Sect.4.3.In the IFS scheme a maxium-random overlap algorithm is employed between cloud layers.Introduction

Conclusions References
Tables Figures

Back Close
Full d e,ice is parametrized according to Sun andRikus (2000, 2001).r e,water is based on Martin et al. (1995).The effective radius used in the radiative transfer parametrization is determined from r e,water and r eq , where an ice cloud with particles of size r e,ice has the same IOPs as a water cloud with r e,water = r eq .Thus, in the SW in HLRADIA cloud water and ice are treated together.This scheme uses a maximum overlap cloud algorithm.
In older versions of the scheme, aerosols were accounted for using constant coefficients (Savijärvi, 1990).However, the scheme has recently been augmented with parametrizations of the direct and semi-direct effects of aerosols.Direct and semidirect effects were realised by implementing new fast analytical SW aerosol transmittances, reflectances and absorptances.The 2-stream approximation equations for anisotropic non-conservative scattering described by Thomas and Stamnes (2002) are used for these calculations.HLRADIA uses the GADS/OPAC aerosol IOPs of Koepke et al. (1997) and the IFS aerosols described in the introductory part of Sect. 2 are translated into GADS aerosols following the definition of Morcrette (ECMWF, 2004).The species include soot, minerals (nucleus, accumulation, coarse and transported modes), sulphuric acid, sea salt (accumulation and coarse modes), water soluble, and water insoluble aerosols.In order to make the calculations fast, the IOPs are averaged over the entire SW spectrum using spectral weightings calculated at a height of 2 km and a solar zenith angle of 45 • for a standard mid-latitude summer atmosphere (Anderson et al., 1986) using the libRadtran/DISORT software package (Mayer and Kylling, 2005;Stamnes et al., 1988).

The ACRANEB2 radiation scheme
The ACRANEB2 scheme (Mašek et al., 2015)  ice crystal IOPs, where a rough aggregate of hexagonal crystals is used as the most representative of cirrus clouds.
The Ritter and Geleyn (1992) delta two-stream system, including partial cloudiness according to Geleyn and Hollingsworth (1979), is used for both the clear-sky and cloudy radiative transfer calculations in ACRANEB2.The coefficients are computed according to Räisänen (2002), i.e. by averaging the coefficients of all of the radiatively active species, weighted by their optical thicknesses.The scheme uses a delta-linear phase function (similar to the delta-Eddington approximation used in IFS) but with hemispherically constant intensities of diffuse radiation.
Like IFS, ACRANEB2 uses intermittency to reduce computational costs.However, it differs in that the intermittency is partial with slowly evolving gaseous transmissions updated hourly (using a simple correction to the actual sun elevation), but rapidly changing cloud IOPs updated at every model time-step, followed by solving the delta two-stream equations.Such partial intermittency proved to be more accurate than the full intermittency employed by the IFS radiation scheme, which scales broadband SW fluxes using the elevation of the sun but does not recompute the IOPs.

Experiments and datasets
Three sets of experiments are detailed in this section (short names for each experiment have been included in brackets): (1) the case study of the summer 2010 Russian wildfires whose smoke plumes affected Estonia (WFEXP), ( 2

Aerosol data in MUSC
In addition to the aerosol climatology described in Sect.2, simulations were also run using observed AOD and derived IOPs from the CIMEL sun/sky radiometer measurements recorded at the Aerosol Robotic Network (AERONET, Holben et al., 1998) station in Tõravere, Estonia.The AOD was derived from measurements of the direct SW radiation at various wavelengths (Holben et al., 1998).SSA and g were derived from sky diffuse irradiance measurements using an inversion algorithm by Dubovik and King (2000).Quality controlled (level 2) AERONET data (Smirnov et al., 2000) were used in this study.

BSRN radiative flux measurements
SW radiation measurements recorded at Tõravere were used in comparisons to simulated fluxes for the August 2010 wildfire case study.These measurements are independent of the AERONET network and are in fact part of the Baseline Surface Radiation Network (BSRN) described by Ohmura et al. (1998).Downwelling SW radiation fluxes (W m −2 ) simulated by a range of different MUSC experiments for Tõravere on 8 August 2010 were compared to measurements recorded at the Tõravere BSRN station (Kallis, 2010).

Experiments using MUSC
The first suite of aerosol experiments focused on the Russian wildfire case study (WF-EXP) of summer 2010 which affected parts of Eastern Europe including Estonia.Severe thunderstorms coincided with forest fires in the Baltic region on 8 August 2010 (Toll and Männik, 2015).These wildfires (Witte et al., 2011;Huijnen et al., 2012) resulted in one of the worst cases of atmospheric pollution over Estonia in recent decades.The radiative effects of smoke from these wildfires are described by Chubarova et al. (2012).Introduction

Conclusions References
Tables Figures

Back Close
Full However, here we have conducted a focused study of the direct radiative effect of the smoke aerosols which reached Tõravere on 8 August 2010.
3-D simulations on a 2.5 km horizontal grid for 8 August 2010, described by Toll et al. (2015a), were carried out to provide meteorological input data for the MUSC experiments for the severe wildfire case study over Tõravere (58.3 • N; 26.5 • E).Atmospheric and surface fields from the 3-D simulation at hourly intervals were used to generate input data for MUSC sensitivity experiments to ensure realistic time evolution of the forcing.These single column experiments were run assuming clear-sky conditions, obtained by removing cloud water and cloud ice from each of the hourly initial profiles.For the analysis of the results, only output from the first time-step was used each hour.
For the wildfire test case three MUSC experiments were carried out, differing in their treatment of aerosols: (1) no aerosols, (2) climatological aerosols and (3) observed aerosols.For tests using observed aerosols, the aerosols were assigned to the land or continental aerosol type with the remaining aerosol categories (there are six categories in the ALADIN-HIRLAM system) set to zero.In this way, the climatological vertical distribution of land aerosols was assumed for the observed IOPs.This assignment affects the SSA, g and wavelength dependence of AOD used by default in the model.
Two variations of the experiment using observed aerosols were carried out.In the first of these, observed AOD at 550 nm and observed aerosol IOPs (SSA, g and wavelength dependence of AOD) were used, whereas in the second, climatological or parametrized aerosol IOPs were used along with observed AOD at 550 nm.Single time-step MUSC experiments were then carried out using hourly forcing from the 3-D experiments for 8 August 2010 for each aerosol experiment separately.As well as this, the experiments were repeated using the three radiation schemes: IFS, HLRADIA and ACRANEB2.
The second suite of experiments consisted of sensitivity tests using MUSC, including tests of AOD at 550 nm (AODEXP), relative humidity (RHEXP) and the vertical distribution of aerosols (VPEXP).The Tõravere location was also used for each of these tests.Introduction

Conclusions References
Tables Figures

Back Close
Full In the tests of AOD at 550 nm the AOD was varied from 0 (no aerosols) to 5 (extremely heavily polluted) to investigate its influence on SW radiation fluxes at the surface.Both observed and parametrized aerosol IOPs were tested for each of the three radiation schemes in MUSC.
The effect of relative humidity (varying from 0 to 1) was tested using 550 nm AODs of 0.1 and 1 and parametrized aerosol IOPs.This was only done using the HLRA-DIA parametrization as the IFS and ACRANEB2 schemes assume a constant relative humidity of 80 %.
To test the effects of the vertical distribution of the aerosols in MUSC the observed 550 nm AOD at 10:00 UTC on 8 August 2010 at Tõravere was arbitrarily chosen.The effect of varying the vertical scale height (h) of land aerosols in the surface-normalized vertical distribution of AOD was investigated.This exponential function distributes the AOD across the 65 model levels in the model.
The final suite of tests presented in this paper focus on aerosol radiative transfer (RT-EXP).Accurate aerosol radiative transfer is of equal importance to accurate aerosol IOPs.We therefore tested clear-sky radiative transfer through a homogeneous layer, with optical properties resembling those of aerosols, using the IFS Fouquart and Bonnel (1980) formulation, the IFS delta-Eddington approximation, the HLRADIA formulation and the Ritter and Geleyn (1992) delta-two-stream approximation compared to the accurate DISORT model (Stamnes et al., 1988).than in the default climatology for August.Figure 2 shows the AOD at Tõravere on 8 August 2010 for 7 wavelengths (data from the AERONET archive) which highlights the strong spectral dependence of AOD; AOD is higher for shorter wavelengths.This notable wavelength dependence is characteristic of biomass burning aerosols (Eck et al., 1999).The AOD at 550 nm (Fig. 2), used in some of the model simulations, was calculated using the AERONET AOD at 500 nm and the Ångström exponent in the 440-675 nm spectral interval.
In addition to 550 nm AOD, the model also requires the following inherent IOPs of aerosols in order to calculate the direct radiative effect of aerosols: AOD spectral scaling coefficients and spectral SSA and g.The AOD spectral scaling coefficients are particularly important because of the high dependence of AOD on wavelength for biomass burning aerosols.Observed AOD scaling coefficients for the six SW bands in the model, assumed valid for the land (soot) aerosol, were derived from the spectral AERONET measurements.These coefficients were then used to derive the AOD for the six SW spectral intervals for the IFS radiation scheme.
A broadband AOD scaling coefficient of 0.959, for use in HLRADIA and ACRANEB2, was derived from the six spectral values derived for IFS using spectral weightings from the standard solar spectrum in the lower troposphere described in Sect.2.2.2.The remaining IOPs, SSA and g for 8 August 2010 at Tõravere are shown in Fig. 3 for four wavelengths.The asymmetry factor varies from 0.56 to 0.7 across the wavelength range with an average value of g = 0.634 for broadband applications (HLRADIA and ACRANEB2 schemes).The aerosol scattering, represented by SSA, remains high (close to 0.96) at each wavelength with little SW spectral dependence.A value of 0.955 was used for the broadband cases.This is similar to results by Dubovik et al. (2002) who showed that the typical SSA of smoke from biomass burning in Boreal forests is high.However, the scattering or SSA from the smoke in this Russian wildfire event was in fact higher than for typical aerosols from biomass burning in Boreal forests (Chubarova et al., 2012).Introduction

Conclusions References
Tables Figures

Back Close
Full The results of the wildfire test case experiments in terms of downwelling global SW radiation (SWD) at the Earth's surface in comparison to BSRN observations are shown in Fig. 4a and b (continuous lines) for the IFS radiation scheme.The results from HLRA-DIA and ACRANEB2 are also depicted relative to observations in Fig. 4b (dashed lines and dotted continuous lines).As expected, when the direct radiative effect of aerosols was excluded SWD was overestimated by ∼ 120 W m −2 (Fig. 4b, red curves) at midday.When the direct radiative effect of aerosols was accounted for using only the observed AOD at 550 nm, but default (i.e.climatological/parametrized) SSA, g and spectral scaling factors of land aerosols, SWD was underestimated by ∼ 100 W m −2 (green curves in Fig. 4b).
This underestimation can be explained by two factors.Firstly, the climatological spectral scaling factors for the IFS land aerosol AOD give a broadband scaling factor that is more than 30 % higher than the factor estimated from the AERONET spectral measurements.Secondly, the differences in the climatological spectral SSA and g values in the model lead to a delta-Eddington optical depth scaling factor (1 Joseph et al., 1976) that is approximately 15 % lower than observed.The combination of these factors results in a broadband SW AOD of IFS land aerosols that is approximately 15 % larger than the observed value.This can be seen from the cyan curve in Fig. 4b, which shows the results obtained when all of the aerosol IOPs are based on the AERONET observations rather than climatology.
The use of AOD and IOPs derived from AERONET observations gives excellent agreement between the modelled and observed SWD for each of the three radiation schemes.The discrepancy between simulated and observed SWD after 14:00 UTC is due to the development of convective clouds (Toll et al., 2015a) which are not accounted for in the MUSC clear-sky simulations.SWD is underestimated, particularly at high solar angles when the observed 550 nm AOD is used with climatological IOPs.
Overall, the results for the three schemes are similar (mostly to within 10-20 W m −2 for global SWD), particularly in their response to the different aerosol cases.The schemes mostly agree to within a few percent at high sun elevations as can be seen in Introduction

References
Tables Figures

Back Close
Full Fig. 4b for the hours around noon.The largest discrepancies occur in the early morning but a detailed investigation into the reasons for this have not been carried out in this study.
Figure 4c shows the unscaled broadband AOD used in HLRADIA and ACRANEB2 for the following three cases: (1) Climatological aerosols (i.e.climatological 550 nm AOD and parametrized/climatological IOPs, black line), (2) Observed 550 nm AOD and IOPs (cyan line) and (3) Observed 550 nm AOD but parametrized IOPs (green line).The broadband AOD shown in this figure can be compared to the observed AOD at Tõravere shown in Fig. 2 for different wavelengths.The broadband AODs were scaled using g and SSA in the HLRADIA and ACRANEB2 SW radiation schemes to produce the broadband SWD fluxes used to generate the HLRADIA and ACRANEB2 curves in Fig. 4b.Therefore, the differences in SWD between the various Tõravere wildfire experiments are mainly due to the differences in the unscaled broadband AOD shown in Fig. 4c.

AOD sensitivity (AODEXP)
3-D model input and AERONET observations for 10:00 UTC on 8 August 2010 for Tõravere were used for the AOD sensitivity tests, which show the influence of 550 nm AOD on SWD for each of the radiation schemes for both observed and parametrized IOPs.Global SWD is shown in Fig. 5a and the direct component of this is shown in Fig. 5b.Clearly, in the case of global SWD, the differences between the SW radiation schemes is small compared to the differences between the results when observed rather than parametrized IOPs are prescribed.While this experiment is a sensitivity test, it clearly illustrates the significant effect that the IOPs have on SW radiation.Direct (Fig. 5b) and diffuse fluxes cannot be compared to observations because the direct irradiances assumed in the models include circumsolar irradiances that cover a broader field of view than standard measurements of direct irradiances.Introduction

Conclusions References
Tables Figures

Back Close
Full

Relative humidity sensitivity (RHEXP)
Tests were done for 550 nm AOD values of 0.1 and 1.0 and parametrized rather than observed aerosol IOPs.3-D model input files for 10:00 UTC on 8 August 2010 were used as initialisation and the relative humidity was hard-coded to vary between 0.0 and 1.0.In the IFS and ACRANEB2 radiation schemes, the aerosol IOP parametrizations assumes a relative humidity of 0.8 for the climatological land aerosols.In this regard, the HLRADIA scheme is more advanced as its aerosol IOPs vary with relative humidity.This is shown in Fig. 6a where global, direct and diffuse SWD for the HLRADIA scheme is plotted as a function of relative humidity for 550 nm AODs of 0.1 and 1.0.For large AODs (1.0) the influence of relative humidity is significant and of the order of 10 s of W m −2 for global and direct SWD. Figure 6b shows the relative difference between the HLRADIA and IFS global SWD as a function of relative humidity, where the latter is independent of relative humidity.For AODs close to the climatological value (i.e.0.1 here) the relative differences between HLRADIA and IFS are very small and negligible at a relative humidity of 0.8.As the humidity deviates from 0.8, particularly for larger AODs, the differences between HLRADIA and IFS grow to ±5 %.

Vertical distribution of aerosols sensitivity (VPEXP)
In these tests of aerosols a 550 nm AOD value of 0.7 for Tõravere on 8 August 2010 (i.e. the observed value at 10:00 UTC) was used.The tests were run using the IFS radiation scheme but the ACRANEB2 scheme uses the same vertical distribution as IFS.
However, the vertical profile is not considered in HLRADIA.The surface-normalized vertical distribution of AOD, α(z), has the following exponential form where z is the height above the ground and h is a vertical scale height.h has a default value of 1000 m for land aerosols in the model.We varied the scale height h in order to change the distribution of the land aerosols, and to determine its relative effects Introduction

Conclusions References
Tables Figures

Back Close
Full on SW radiation and the SW radiative heating rate.For smaller h the aerosols are concentrated closer to the ground while for bigger h they are spread higher into the atmosphere.
The effect of aerosols, and in particular the vertical scale height for land aerosols, on net SW radiation and the SW radiation heating rate as a function of pressure is shown in Fig. 7.The results are scaled relative to the case where aerosols are not included.At each model level, the relative net SW radiation for each of the vertical scale heights, varies by less than 3 % (Fig. 7a).It is reassuring that the variation with scale height is small because the HLRADIA scheme does not consider vertical profile for overall aerosol transmittance.However, in HLRADIA the vertical profile is considered for the heating rates at atmospheric levels.Unlike SW fluxes, where the differences remains within 3 % throughout the atmosphere, the impact on SW heating rates is significant (ratio of up to ∼ 1.5 compared to the default vertical profile when the vertical scale height is approximately halved).The main reason for this is presumably that the heating rate is proportional to flux divergence.

Aerosol radiative transfer (RTEXP)
The IFS radiation scheme uses different two-stream radiative transfer approximations (Joseph et al., 1976;Fouquart and Bonnel, 1980) for aerosols under the clear-sky and cloudy fractions of a given grid box.We tested both of these radiative transfer approximations as well as the two-stream approximation (Thomas and Stamnes, 2002) used in HLRADIA and the Ritter and Geleyn (1992) two-stream approximation in ACRANEB2 and compared the results to the accurate DISORT scheme (Stamnes et al., 1988) run with 30 streams.Sample transmittance results are shown in Figs. 8 and 9 where the optical thickness is varied but the cosine of SZA (µ 0 ) is set to 0.6, g to 0.7 and SSA to 0.95.
As can be seen from these figures the IFS, HLRADIA and ACRANEB2 radiative transfer approximations give transmittances that are within a few percent of the accurate 30-stream DISORT calculations for optical thicknesses which are less than 1.

Conclusions References
Tables Figures

Back Close
Full Even for optical thicknesses up to 5, the approximations remain within ±12 % of the DISORT results.Thus, it can be concluded that the three (IFS, HLRADIA, ACRANEB2) approximations are sufficiently accurate.Additional results from tests of the IFS delta-Eddington radiative transfer scheme can be found in the supplement to the paper by Nielsen et al. (2014).

Conclusions and future work
When the 550 nm AOD and IOPs (spectral AOD scaling, SSA and g) of aerosols are accurately known, the SW direct radiative effect of aerosols is accurately simulated with the IFS, HLRADIA and ACRANEB2 radiation schemes available in the ALADIN-HIRLAM NWP system.Global SWD radiation was underestimated during the wildfire event in summer 2010, especially at high solar angles, when the observed AOD at 550 nm was used with the climatological or parametrized IOPs in each of the radiation schemes rather than the observed IOPs.This highlights the need for accurate information on both aerosol concentration and aerosol IOPs in order to improve the simulated radiation budget in the model and possibly to result in an improvement in weather forecasts in the future.The model could acquire this information, including the vertical profile of the aerosol properties, from 3-D aerosol IOP estimates from the C-IFS model or chemical transport model simulations coupled to the NWP model.
The IFS, HLRADIA and ACRANEB2 radiative transfer approximations were tested and found to be accurate to within ±12 % even for large aerosol loads.The dependency of the direct radiative effect of aerosols on the relative humidity and the dependency of the SW heating rates on the vertical distribution of aerosols were shown to be non-negligible.We therefore suggest that a relative humidity dependent parametrization of aerosol IOPs should be used (like in HLRADIA) as well as realistic vertical profiles of aerosols.Nevertheless, as a first approximation, assuming a constant relative humidity and climatological vertical profiles are acceptable.Introduction

Conclusions References
Tables Figures

Back Close
Full For cases of heavy pollution the climatological AOD in MUSC is much smaller than observed and when used with either observed or parametrized IOPs, global SWD radiation is overestimated.This is similar to the case where aerosols are completely omitted.This overestimation of global SWD radiation leads to other model errors.Thus, during heavy pollution considerable improvements in the representation of the radiation budget are possible when real rather than climatological aerosols are included.
The influence of improvements in the representation of the direct radiative effect of aerosols on meteorological forecasts needs to be further studied using 3-D simulations.In this study the direct SW radiative effect of aerosols was studied; in the future the indirect effects of aerosols on cloud microphysics and the longwave effect of aerosols also need to be investigated.
Additionally, our results highlight the issues with having IOPs for each aerosol type but a fixed 550 nm AOD.This is important for aerosol data assimilation which should include all of the aerosol IOPs and not only 550 nm AOD.Introduction

Conclusions References
Tables Figures

Back Close
Full have previously investigated cloud optical and physical properties, radiative transfer Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | for six aerosol (continental, sea, urban, desert, volcanic and background stratospheric) types are available by default in the model.The aerosol IOPs (SSA, g and AOD scaling for each of the six IFS SW radiation bands) are parametrized following Hess et al. (1998) and Koepke et al. (1997).The 550 nm AOD and spectral Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | are several SW cloud liquid and cloud ice IOP schemes available in the model.By default, the SW cloud liquid and cloud ice IOPs are calculated with the Fouquart (1987) and the Ebert and Curry (1992) parametrizations respectively.It is also possible to use the Slingo (1989) SW cloud liquid IOP parametrization and the Fu (1996) Discussion Paper | Discussion Paper | Discussion Paper | Screen / Esc Printer-friendly Version Interactive Discussion Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | was developed as part of the ALARO-1 suite of physics parametrizations.Similarly to HLRADIA, it uses a single SW radiation interval (0.245-4.642 µm).ACRANEB2 uses the Reid et al. (1999) conversion formula to diagnose r e,water and the Heymsfield and McFarquhar (1996) equation to parametrize r e,ice .Using these, the broadband cloud liquid IOPs were derived from Hu and Stamnes (1993) spectral data and Key et al. (2002) spectral data were used to derive cloud 32528 Discussion Paper | Discussion Paper | Discussion Paper | ) AOD (AODEXP), aerosol vertical profile (VPEXP) and humidity sensitivity tests (RHEXP) and (3) tests of aerosol radiative transfer (RTEXP).A summary of these experiments in terms the aerosols and radiation schemes used is given in Table1.Further details are provided in subsequent subsections.Tõravere in Estonia on 8 August 2010 was selected to be studied because the smoke had reached the area by then and measurements of aerosol IOPs and broadband radiation fluxes at Tõravere were available from AERONET and BSRN measurements respectivelyDiscussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Figure 1b shows the 550 nm AOD from the MACC reanalysis (Inness et al., 2013) over much of Europe on 8 August 2010 compared to the 550 nm climatological AOD in the model (Fig. 1a) which greatly underestimates aerosols when pollution is heavy.The values of the realistic MACC AOD (up to over 3.5) are an order of magnitude higher Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper |

Figure 3 .Figure 4 .Figure 7 .
Figure 1.(a) 550 nm AOD for August from the default climatology in ALADIN-HIRLAM (Tegen et al., 1997) (b) AOD for 8 August 2010 from MACC reanalysis (Inness et al., 2013).The location of Tõravere (58.3•N; 26.5• E) for which the MUSC experiments are run is shown as a red dot in both panels.Note the factor of 10 difference in AOD between the climatology and the MACC reanalysis.