Aerosols are tiny solid and liquid particles suspended in the air. They cause direct radiative forcing through scattering and absorbing shortwave (SW) and longwave (LW) radiation in the atmosphere. They also alter cloud formation and precipitation efficiency by increasing droplet and ice particle concentrations. In this way aerosols also cause indirect radiative forcing.
Aerosols off-set a poorly quantified fraction of the greenhouse gas warming effect on the Earth. In fact, the quantification of aerosol radiative forcing is more complex than that for greenhouse gases because aerosol mass and particle concentrations are highly variable in space and time. This is mainly due to the shorter atmospheric lifetime of aerosols. Spatial and temporal information on the physical and radiative properties of aerosols is required such as size distributions, dependence on relative humidity, refractive index and solubility of the particles.
Over recent decades substantial progress has been made in reducing uncertainties related to radiative forcing due to aerosols. This progress is due to advances in global modelling, theoretical developments and improved observations. Integrated weather-chemistry models [1
] simulate the life cycle of aerosols from their formation to their deposition and their dispersion in the atmosphere at timescales of the order of several days. In such models the physicochemical processes evolve in an environment controlled by atmospheric large-scale dynamics. Advanced data assimilation, using reliable information about emission sources and conventional and space-born observations, is used to constrain the modelled processes. The Copernicus Atmosphere Monitoring Service (CAMS) global reanalysis of atmospheric composition [7
] has resulted in an extensive historical aerosol dataset. They also produce global forecasts of aerosols and atmospheric chemical constituents in near-real time [8
In the European Centre for Medium-Range Weather Forecast (ECMWF) model the use of a new global 3D aerosol climatology based on [7
], combined with updated aerosol inherent optical properties (IOPs), led to a systematic improvement in lower troposphere temperature and wind forecasts over certain regions of the globe [9
]. However, during wildfires, desert dust intrusions, volcanic eruptions and enhanced anthropogenic emissions aerosol concentrations in the atmosphere may significantly exceed climatological values, which can influence the weather on local to global scales. In such cases reliance on aerosol climatologies is insufficient for accurate forecasting of radiation and temperatures [10
]. Recent studies, such as those by [14
], provide additional motivation to improve how aerosols are taken into account in short-range regional numerical weather prediction (NWP) models and not just in global medium-range forecasting and climate models [16
]. The direct radiative effect of aerosols on weather and climate is much better quantified than indirect effects [19
] because of the greater physical understanding of the aerosol radiative effects compared to the impact of aerosols on clouds and precipitation. This suggests that focusing on improving aerosol radiative transfer parametrizations is worthwhile in limited-area NWP models.
Regional integrated weather-chemistry models [21
] are computationally demanding, thus mainly used for research. The availability of up-to-date global aerosol datasets opens new possibilities for operational limited-area weather forecasting, because the regular weather models can now import ready-made aerosol concentrations for use in the parametrizations of radiative transfer, cloud-precipitation microphysics, and the consistent treatment of cloud-aerosol-radiation interactions. However, to benefit from new aerosol datasets improvements in regional NWP models are required. More detailed information on the optical properties of aerosols, such as the aerosol optical depth (hereafter AOD) or mass extinction coefficient (ME), single-scattering albedo (SSA) and asymmetry parameter (ASY), is needed combined with the improved aerosol concentration data. Technical changes are needed to introduce external three-dimensional (hereafter denoted as 3D) aerosol load data to these models in near-real time (n.r.t. hereafter).
The work presented in this paper expands on the studies by [10
]. Three radiation schemes available in the shared ALADIN-HIRLAM Numerical Weather Prediction system [22
] (hereafter the ALADIN-HIRLAM NWP system) were compared using a single-column model approach [10
]. All schemes produced realistic results when observation-based wavelength-dependent AOD, SSA and ASY were taken into account. In a Saharan dust intrusion case, the introduction of n.r.t. aerosol data into 3D NWP model simulations led to large changes in the SW irradiance and near-surface temperatures. The results showed good agreement with local temperature and radiation observations despite the simplified treatment of the aerosol IOPs. However, updating the input aerosol load (AOD at the 550 nm wavelength, hereafter AOD550) from the default Tegen climatology [23
] to a simplified CAMS-based AOD550 climatology resulted in almost no changes in climate simulation results [11
In this paper, we report on our upgrades to the aerosol radiative transfer parametrizations in ALADIN-HIRLAM. By default, the system uses Tegen climatologies [23
] of AOD550 of land, sea, desert, urban and sulfate aerosols and the following aerosol IOPs: AOD wavelength-scaling factor, SSA and ASY based on [24
]. Our upgrades involve the combination of updated IOPs (ME, SSA and ASY for 11 aerosol species at 30 wavelengths) from ECMWF [9
] and aerosol concentrations from the CAMS reanalysis [7
] or CAMS n.r.t. data [8
]. We introduced the updated aerosol optical properties into HLRADIA, which is the HIRLAM broadband radiation scheme [27
], the simplest radiation scheme available in HARMONIE-AROME [29
]. The single column version of ALADIN-HIRLAM, known as MUSC [30
], is used as the tool to test the impact of aerosols on the radiative fluxes and atmospheric temperature tendencies due to radiation.
Our aim is to understand the importance of realistic estimations of the distributions of different aerosol species and accurate IOPs in the calculation of aerosol radiative transfer. In this study, we focus on the direct radiative effect of aerosols using clear-sky cases so that the interactions between cloud-precipitation microphysics and radiation are excluded. Comparisons with observations have not been made. Instead, sensitivities and uncertainties under real atmospheric conditions are considered. Comparisons with observations will be carried out in the next phase using the 3D ALADIN-HIRLAM NWP system.
The paper is organized as follows: Section 2
provides a description of aerosol radiative transfer in the ALADIN-HIRLAM NWP system. Section 3
describes the single-column experiments and the results are included in Section 4
. The results are discussed in Section 5
and conclusions are drawn in Section 6
. Basic information about the ALADIN-HIRLAM NWP system, its radiation parametrizations and the CAMS aerosol data used in the experiments is provided in Appendix A
2. Aerosol Radiative Effects in the ALADIN-HIRLAM NWP System
In this section, we summarize the main features of the parametrization of aerosol radiative effects in the ALADIN-HIRLAM NWP system, in particular within HARMONIE-AROME. Radiation schemes estimate the radiative heating in the atmosphere due to the vertical divergence of net LW (LWNET, terrestrial) and SW (SWNET, solar) radiation fluxes in an NWP model. This heating is a source term in the thermodynamics equation in the model and influences atmospheric temperatures and the evolution of clouds. At the surface, radiation parametrizations provide the model with downward (LWDS, SWDS; the D refers to downwards and the S to surface) and upward LW and SW radiation fluxes. The outgoing fluxes at the top of the atmosphere (TOA) are also given (LWUT, SWUT; the U refers to upward and T refers to top). The surface fluxes are part of the surface energy balance and a lower boundary condition for the calculation of atmospheric radiation transfer. At the TOA, downwelling SW radiation defines the upper boundary condition for radiation parametrizations. Surface properties such as temperature, albedo and emissivity are also required as input. In terms of aerosols, the radiation schemes include parametrization of the direct radiative effect of aerosols due to absorption and scattering in the SW part of the spectrum and absorption/emission in the LW. The indirect radiative effects of aerosols, which are related to cloud-aerosol interactions, are currently not properly included in HARMONIE-AROME.
The ALADIN-HIRLAM NWP system uses monthly climatologies of four classes of tropospheric AOD550 (land, sea, desert and urban where the ‘land’ classification includes sulfates) based on [23
] by default. In addition, stratospheric sulfates and volcanic dust are accounted for using assumed constant background values. The six AOD550 fields are distributed vertically using the assumed exponential profiles of Tanré [31
] in the same way as described by [9
]. Aerosol IOPs, namely the relation of AOD to AOD550, the SSA and ASY, of the 6 aerosol species are prescribed for 6 SW and 6 LW intervals in the default IFSRADIA radiation scheme. Two broadband radiation schemes, HLRADIA and ACRANEB, available for experimenting within HARMONIE-AROME, have been adapted to use the default AOD550 climatologies as input. Appendix A.2
contains references and details about the three radiation schemes.
We mainly used HLRADIA for the single-column study of the sensitivity of radiation fluxes to aerosol concentration and optical properties. The original HLRADIA scheme [27
] estimates the impact of aerosols on radiation by applying constant coefficients to represent SW absorption and scattering and LW absorption and emission. Thus, no climatological or other gridded aerosol load data are used. The treatment of aerosol data and IOPs by HLRADIA was renewed in the Enviro-HIRLAM modelling environment using GADS/OPAC data [24
] and software as suggested in [4
]. The aerosol optical properties were derived from GADS/OPAC ME, SSA and ASY and then remapped to the default AOD550 input of 6 aerosol species and spectrally averaged to get the broadband SW and LW values. Relative humidity from the model is taken into account. The resulting broadband optical depths of the aerosol mixture are scaled using a delta-Eddington factor [32
] in the form 1 − SSA × (ASY)
. This AOD550-based HLRADIA was first applied in HARMONIE-AROME experiments in [10
] for the Russian wildfire case study of 2010, and only involved SW effects.
For the current study we adapted HLRADIA so that it can use CAMS aerosol MMR data combined with up-to-date aerosol IOPs from ECMWF as input rather than AOD550. In this MMR-based approach, monthly 2D climatologies based on reanalysis data [7
] or 3D n.r.t. data [8
] can be used. For our purposes, data for 11 aerosol species were extracted: 3 size bins of hydrophilic sea salt (SS) aerosols, 3 size bins of hydrophobic mineral (desert) dust (DD) aerosols, hydrophilic and hydrophobic organic matter (OM), hydrophilic and hydrophobic black carbon (BC) and hydrophilic sulfate (SU). SS approximately corresponds to the Tegen ‘sea’ aerosol category, DD to ‘desert’, OM to ‘land’ without sulfates and BC to ‘urban’ aerosols. Updated exponential functions [9
] were applied to distribute the 2D MMR data vertically on model levels. The CAMS data and ECMWF IOPs are described in more detail in Appendix A.3
To summarize, in this study we test three different versions of the HLRADIA parametrization of aerosol radiative transfer: the original version, the AOD550-based version and the MMR-based version. For the comparison some results using the AOD550-based IFSRADIA and ACRANEB schemes are also shown.
The single-column version of ALADIN-HIRLAM, known as MUSC [30
], was used to carry out the various sensitivity tests. Table 1
summarizes the series of MUSC experiments. Further information about MUSC is given in Appendix A.1
. The input data (aerosols, and atmospheric and surface state) for these experiments were taken from 3D HARMONIE-AROME experiments for two locations: Lake Ladoga in Russia for the 19th of April 2019 and Badajoz in Spain for 21 February 2017. Experiments were done using both 3D n.r.t. and 2D climatological aerosol MMR data. Experiment names starting with C used climatological aerosol input data while those starting with N used n.r.t. data, 2 denotes vertically integrated (2D) aerosol data and 3 denotes column (3D) data.
The Badajoz case represents a strong Saharan dust intrusion while the Ladoga case represents regular background aerosol conditions. Often in NWP studies only extreme situations are chosen for case studies, while in daily forecasting normal conditions are prevalent. We used the Ladoga atmospheric and surface data as the basis for the MMR and AOD series of sensitivity tests, in which the aerosol compositions and concentrations were modified (Section 4.3
). To exclude soil interactions from the MUSC experiments, a water surface was assumed at both locations. The incoming solar radiation at the top of atmosphere was 895 Wm
over Badajoz and 779 Wm
over Ladoga. The surface albedo was 0.07 at Badajoz and 0.37 over Lake Ladoga (the lake was assumed to be partly frozen). The LW emissivity of the surface was 0.97 in both cases.
The ZERO experiments do not include aerosols. Experiment HSA uses constant coefficients for aerosol radiative transfer [27
]. CAOD2, which is the reference experiment, uses monthly climatological values of AOD550 for land (OM + SU), sea (SS), desert (DD) and urban (BC) aerosols. The CMMR2 experiments use vertically integrated climatological MMRs of 11 species, given on the same 3 degree resolution global horizontal grid as the climatological AOD550 data. These data are based on CAMS interim reanalysis output for 2003–2011, and IOPs applied therein [26
]. No background aerosols are added when MMR data are used.
The NMMR3 experiment uses 3D n.r.t. CAMS aerosol data. Output of this MUSC experiment included vertically integrated MMR for the 11 species, that was used as input for experiment NMMR2, and AOD550 which was used as input for the NAOD2 experiment. Thus, the experiments NMMR3 and NMMR2 differ only with respect to the vertical distribution of aerosol MMR whereas experiments NMMR2 and NAOD2 differ with respect to the optical properties. The MMR and AOD sensitivity experiment series differ in similar ways to NMMR2 and NAOD2 but only a single aerosol species was included in each specific experiment.
When the aerosol load is at a background or normal level, the aerosol concentration and optical property differences lead only to small differences in radiation fluxes and radiative heating. When the concentration of some aerosol species is higher than its background level, regionally or during pollution episodes, reliable 3D aerosol data become more important. To benefit from such data an accurate treatment of the optical properties of the different species becomes important.
The greatest differences were found between the experiments based on climatological AOD550 or MMR and those using n.r.t. data for the desert dust intrusion case over Badajoz. For example, SW transmission was 0.98 (0.97) when 2D climatological MMR (AOD550) was used but 0.78 when n.r.t. 3D MMR or 2D AOD550 was used. For the same case, LW transmission was 0.90–0.95 when n.r.t. aerosol data was used but 0.98–1.0 when climatological data were employed. In general, LW differences were smaller than SW differences. The value of 0.78 for SW transmission in the Badajoz experiment is in fact not very low. For example, based on satellite-based estimates [33
] monthly mean values of less than 0.5 were recently reported over areas of India. Our example was that of desert dust whereas the aerosol composition over India is likely to be quite different.
Sensitivity studies using a total-column AOD550 = 0.5 for each aerosol species separately, demonstrated their respective impacts. To obtain AOD550 values of 0.5, 37 g of black carbon per square meter was required, whereas almost 1 g of sea salt results in the same AOD550. The SW transmission due to this amount of black carbon was 0.56 but 0.95 for sea salt, despite having the same AOD550 values. When AOD550 = 0.5 was used in conjunction with the old prescribed IOPs, the SW transmissivity increased to 0.79 for black carbon and decreased to 0.90 for sea salt. This demonstrates that it is important to account for the scattering and absorption properties of aerosol species, and not only to estimate the total AOD. An uncertainty in the estimated concentration of the species has different impacts: an inaccuracy of one g in the mass of black carbon may influence the radiative transfer more than a ten- or hundred-fold inaccuracy in the estimation of the mass of coarser particles.
The difference between the AOD and MMR sensitivity experiments is in how they handle optical properties. Both relied on vertically integrated AOD or MMR values that were distributed on model levels using the same species-specific exponential profiles. However, the optical properties depend not only on the vertical distribution of the aerosols but also on atmospheric humidity which varies significantly with elevation. Our new method of combining aerosol optical properties and mass distribution highlighted the sensitivity of radiative transfer to the vertical distribution of aerosol species. The introduction of new dependencies and interactions, even physically well-based, may increase the uncertainties of the calculations.
The single-column model framework allowed us to diagnose effects and interactions that cannot easily be detected using results of 3D model experiments. The main limitations of our study include the following which require further experimentation and developments:
We have applied the simple HLRADIA scheme to determine the sensitivity of radiation fluxes and temperature tendencies to aerosol load and optical properties. HLRADIA was chosen for the practical reasons of availability and simplicity. Its known limitations, relating to the simplified treatment of atmospheric layers and the use of many empirical coefficients for the calculation of the SW and LW radiative transfer, must be taken into consideration. In particular, HLRADIA is unable to fully benefit from the 3D details of the aerosol optical properties suggested here. Also, as a broadband scheme, it is unable to use the spectral details of aerosol-radiation interactions that may become important in the LW range of the spectrum.
In our MUSC experiments the surface temperature was intentionally kept constant by assuming a water/ice surface at the bottom of the atmospheric column. In reality, the local near-surface temperature changes related to aerosol impacts on radiation over land areas are assumed to arise from heating of the soil and not because of direct air temperature changes. However, MUSC is a less suitable tool for analyzing the evolution of temperature and other atmospheric variables over time because large-scale dynamical processes are ignored in the single-column framework.
We have analyzed the impact of mineral dust in a Saharan dust intrusion case study. The impact of other aerosol species—sea salt, organic matter, black carbon and sulfates were only studied using artificial data under realistic atmospheric conditions. It would be interesting to study wildfire cases again, where organic matter, and possibly black carbon, impacts can be seen. Cases involving increased volcanic and anthropogenic emissions deserve further study. Stratospheric volcanic sulfates that are assumed to be a main factor in past climate cooling episodes at annual to decadal scales (see [34
] and references therein) were not included. Stratospheric aerosols are poorly parametrized in limited-area NWP models and are not represented in the version of the CAMS dataset used in this study. However, volcanic emissions contribute to the tropospheric sulfate and dust loads in the CAMS dataset.
We did not carry out tests using 3D climatological MMR data, that are available in the CAMS reanalysis dataset at high horizontal and vertical resolution and used in the ECMWF operational model [9
]. We believe that for limited-area NWP models used for short-range weather forecasting, it is a higher priority to capture episodes of high aerosol load in real time than to address small systematic errors that may be related to the use of a coarse-resolution 2D aerosol climatology.
Cloud–radiation–aerosol interactions were excluded from this study to focus on the direct radiative effects of aerosols. It is possible to estimate the first indirect effect of aerosols (Twomey effect) by parametrizing the cloud droplet number concentration (CDNC) using external aerosol data. Cloud particle effective size can be derived from CDNC and applied in the radiative transfer calculations. It is more complicated to take the impact of (hydrophilic) aerosols on the evolution of cloud droplets to precipitating particles into account. Single-column experiments can be used as a first step in formulating and testing the parametrizations.
To overcome the limitations of single-column studies, 3D HARMONIE-AROME experiments are required. Such experiments allow the study of aerosol impacts on weather parameters, and take dynamical processes and evolving surface interactions into account. They also enable quantification of aerosol-related uncertainties in weather forecasting. It is more useful to do such experiments using advanced radiation schemes which include radiative exchanges between atmospheric layers in cloudy and clear-sky cases. The results of 3D experiments should be compared to aerosol and radiation observations.
6. Conclusions and Outlook
In this study, we suggest improvements to the aerosol radiative transfer parametrizations in the ALADIN-HIRLAM NWP system. We have updated the calculation of aerosol optical properties in the HARMONIE-AROME configuration of the ALADIN-HIRLAM system. This was done using a combination of climatological 2D or n.r.t. 3D aerosol concentrations from CAMS and new pre-calculated IOPs from ECMWF. The resulting AODs, SSAs, ASYs for the aerosol mixture, at 16 LW and 14 SW wavelengths, were used to calculate broadband values of these optical parameters by applying spectral averaging over the SW (0.20–12.19 m) and LW parts (3.08–1000 m) of the spectrum.
These broadband optical properties were used by HLRADIA, the simplest radiation scheme available in HARMONIE-AROME, employing the approach in Enviro-HIRLAM [4
]. The impact of the updated aerosol optical properties on the radiative fluxes and temperature tendencies was studied in single-column MUSC experiments. Both the aerosol concentrations, that originated from the CAMS dataset, and the atmospheric states were extracted from 3D HARMONIE-AROME experiments. For additional sensitivity studies, artificial 2D AOD550 and MMR data were prepared and used as input.
Using external 3D aerosol concentration data instead of climatological AOD550 data is beneficial for limited-area NWP models for several reasons. Until now, the treatment of aerosol inputs has usually been a part of the radiation schemes. We suggest that the 3D optical properties (AOD, SSA, ASY) of the aerosol mixture at each time-step of the model’s integration, taking the atmospheric humidity into account, be prepared outside of the radiation scheme. Using the actual optical properties as input enables radiative transfer calculations to be done applying any scheme without the need to treat the specific properties of individual aerosols. The possibility to choose between n.r.t. or climatological aerosol input data provides additional flexibility. Improved consistency between radiation and cloud parametrizations can be expected, in particular regarding the derivation and use of cloud particle effective sizes. In the future, n.r.t. data on the distribution of aerosols of different sizes and species could also be used in cloud microphysics parametrizations.
In this study, we have taken external aerosol data and IOPS as given, and focused on how to use these in a limited-area NWP model. Global integrated weather-chemistry models, with advanced data assimilation, presumably produce more reliable data than any regional integrated model. The CAMS global reanalysis data [7
] are more detailed and more reliable than the older Tegen dataset [23
]. However, the results from global weather-chemistry models contain uncertainties related to aerosol emission sources, assumptions used in the data assimilation, parametrizations of aerosol dynamics and the derivation of IOPs as discussed extensively in recent papers [7
]. Additional inaccuracies arise from spatial and temporal interpolation of the coarse-resolution global aerosol data to the high-resolution limited-area NWP model grid.
Based on our results, we suggest the following steps to improve aerosol-related parametrizations in the ALADIN-HIRLAM NWP system:
Include MMR-based 3D optical properties of the aerosol mixture for use by the IFSRADIA and ACRANEB radiation schemes to benefit from their more advanced SW and LW radiation transfer parametrizations compared to HLRADIA.
Implement the method of importing n.r.t. high-resolution 3D CAMS MMR data to the ALADIN-HIRLAM system for use in operational weather forecasting. Investigate possible simplifications that would reduce the computational resource demand.
Carry out extensive model-observation inter-comparisons for cases involving biomass burning, mineral dust intrusion, anthropogenic and volcanic emission to evaluate their impacts on local weather and radiation flux forecasts.
Find optimal ways to use n.r.t. aerosol concentration data for derivation of cloud particle effective sizes, which are assumed to be the key parameters in the consistent treatment of aerosol-cloud-radiation interactions.