A parameterization for secondary organic aerosol (SOA) production based on
the volatility basis set (VBS) approach has been coupled with microphysics
and radiative schemes in the Weather Research and Forecasting model with
Chemistry (WRF-Chem) model. The new chemistry option called
“RACM-MADE-VBS-AQCHEM” was evaluated on a cloud
resolving scale against ground-based and aircraft measurements collected
during the IMPACT-EUCAARI (Intensive Cloud Aerosol Measurement Campaign –
European Integrated project on Aerosol Cloud Climate and Air quality
interaction) campaign, and complemented with satellite data from MODIS. The
day-to-day variability and the diurnal cycle of ozone (O
It is well recognized that aerosol particles have a fundamental role in the
climate system. They directly alter the budget of the radiation that reaches
the Earth's surface by scattering and absorbing the incoming sunlight
(Haywood and Boucher, 2000), and they indirectly affect cloud properties and
precipitation patterns, because they act as cloud condensation nuclei (CCN)
(Rosenfeld et al., 2008; Lohmann and Feichter, 2005). Some aerosol species as
black and brown carbon or mineral dust heat the atmosphere absorbing the
solar radiation. The local warming may increase the atmospheric stability,
leading to a decrease in cloud cover through the so-called semi-direct effect
(Hansen et al., 1997). The global mean radiative forcing associated with
aerosols, as a result of changes in anthropogenic emissions since
pre-industrial times, is highly uncertain and is estimated to be
Experimental evidence of the influence of aerosols on cloud macrophysical and microphysical properties has been shown in several works (Clarke and Kapustin, 2010; Christensen and Stephen, 2011; Koren et al., 2012; Ten Hoeve et al., 2011; Li et al., 2011). Several modelling studies show that aerosol particles have a strong impact not only on the climatic spatial–temporal scale but also at short range on the regional scale (Baklanov et al., 2014). At regional scale, online coupled mesoscale meteorology–chemistry models are useful tools to take into account aerosol feedback effects on both meteorology and atmospheric composition (Zhang, 2008; Baklanov et al., 2014).Weather Research and Forecasting model with Chemistry model (WRF-Chem), which is the model used in this study, is one of such models (Grell et al., 2005; Fast et al., 2006; Chapman et al., 2009). In this work we present and evaluate some developments of WRF-Chem for a better simulation of direct and indirect aerosol feedback.
The introduction of the aerosol–cloud–radiation feedback leads to non-linear chains and loops of interactions between meteorological and chemical processes that are inhomogeneous in space and time (Baklanov et al., 2014). Furthermore, the prediction of meteorological variables significantly improves when the direct and indirect aerosol effects are taken into account in numerical simulation. For example, Yang et al. (2011) found that the inclusion of aerosol feedback produces significant benefits in the simulated optical and microphysical properties of marine stratocumulus, and these improvements positively affect the simulation of the boundary layer structure and energy budget. Yu et al. (2014) reported an improvement of the simulation of shortwave and long-wave cloud forcing when the aerosol feedback is added to the model.
Recent studies conducted with global models, predict an important
contribution of secondary organic aerosol (SOA) to direct and indirect
aerosol feedback. O'Donnell et al. (2011) calculated an annual mean direct
and indirect shortwave forcing of
Previous studies over USA and Europe demonstrated that the “traditional”
configuration of WRF-Chem (Grell et al., 2005), using the secondary organic aerosol model (SORGAM) (Schell et al., 2001), presents a negative bias of
simulated PM
A pre-release of version 3.4 of the WRF-Chem (Grell et al., 2005) was used in this study. WRF-Chem is a community model that has many options for gas chemistry and aerosols. One of these has been updated in order to include a new chemistry option for simulation of direct and indirect effects with an updated parameterization for SOA production. The modifications introduced by Fast et al. (2006), Chapman et al. (2009), and Ahmadov et al. (2012) are the basis of our work. The technical details of the implementation are summarized in Appendix A.
The gas-phase mechanism used is an updated version of the regional
atmospheric chemistry mechanism (RACM) (Stockwell et al., 1997). The
inorganic aerosols are treated with the Modal Aerosol Dynamics Model for
Europe (MADE) (Ackermann et al., 1998). The updated parameterization for SOA
production is based on the VBS approach (Ahmadov et al., 2012). MADE-VBS
model has three log-normal modes: Aitken, accumulation and coarse. The
species treated are the sulfate (SO
SOA parameterization implemented by Ahmadov et al. (2012) is based on a four
bin volatility basis set, with a saturation concentration of 1, 10, 100, and
1000
The implementation of aerosol–cloud–radiation interaction within RACM-MADE-VBS follows the methods described by Fast et al. (2006) and Chapman et al. (2009). We modified the WRF-Chem code by preparing the inputs for the modules devoted to calculation of the aerosol optical properties and the aerosol activation, starting from the mass of each aerosol type as predicted by the new chemistry package. In the approach of Fast et al. (2006), the three modes of the log-normal distribution are divided into eight bins, and each chemical constituent of the aerosol mass is associated with a complex refractive index. The refractive index is calculated for each bin with a volume average. Mie theory is used to find the scattering and absorption efficiencies. Aerosol optical thickness (AOT), single scattering albedo, and asymmetry parameter calculated with the optical package developed by Barnard et al. (2010) are used as input to the radiative scheme (Goddard and RRTMG). The aerosol direct effect on long-wave radiation is included following Zhao et al. (2010).
Aerosol–clouds interaction is a complex problem that involves the activation and resuspension of the aerosol particles, aqueous chemistry, and wet removal. Following Chapman et al. (2009), aerosols within cloud droplets are treated as “cloud borne”. Aerosols that do not activate as cloud droplets are treated as “interstitial”. In WRF-Chem the activation process is based on the parameterization developed by Abdul-Razzak and Ghan (2000, 2002). The number and mass concentration of the activated aerosols are calculated for each mode. The activation of aerosols is based on a maximum supersaturation determined from a Gaussian spectrum of updraft velocities and bulk hygroscopicity of each aerosol compound for all log-normal modes of particles. Bulk hygroscopicity is based on the volume-weighted average of the hygroscopicity of each aerosol component. In addition to the activated aerosols at environmental conditions, the CCN spectrum is also determined; i.e. the aerosol particles acting as CCN at some given maximum supersaturations (0.02, 0.05, 0.1, 0.2, 0.5, and 1 %) are calculated.
Within the dissipating clouds, the droplets evaporate and the cloud borne aerosols are resuspended to the interstitial state. Cloud borne aerosols and dissolved trace gases may be modified by aqueous chemistry. In this chemistry option, cloud chemistry is modelled using the scheme of Walcek and Taylor (1986). Wet deposition of trace gases and aerosols is treated in and below clouds. Within clouds the aerosols and trace gases dissolved in the water are collected by rain, graupel, and snow. Below clouds the wet scavenging by precipitation is parameterized using the approach of Easter et al. (2004).
The simulation of the activation, resuspension, and wet scavenging of the aerosol particles requires a prognostic treatment of the cloud droplets. The prognostic treatment of the cloud droplets takes into account the losses due to collision, coalescence, collection, and evaporation, and the source due to nucleation. The Lin and Morrison microphysics schemes in WRF-Chem version 3.4 include the prognostic treatment of the cloud droplet number concentration. The source due to nucleation is parameterized following Ghan et al. (1997). Both microphysical schemes take into account the autoconversion of cloud droplets to rain dependent on the cloud droplet number. Therefore, aerosol activation affects both the rain rate and the liquid water content. The droplet number concentration affects the calculation of the cloud droplet effective radius and cloud optical thickness (COT). The interaction of clouds with the incoming shortwave radiation is done by linking the microphysics to the radiation scheme. The reader should note that the contribution to SOA concentration by cloud chemistry is missing and the interaction of aerosol with ice nuclei is not taken into account in this version of the model.
Panel
The simulations were conducted over three one-way nested domains centred on the Netherlands, as shown in Fig. 1. The coarse domain (D1) has 30 km horizontal resolution, domain 2 (D2) 10 km, and domain 3 (D3) is cloud resolving at 2 km resolution. In our runs we used 67 vertical levels extending up to 50 hPa.
The main physical and chemical parameterizations used are listed in Table 1. The model setup is the same for all three domains, except that no cumulus parameterization is used for D3. Wet scavenging and cloud chemistry from both parameterized updraft and resolved clouds are taken into account in D1 and D2. However, in these domains the sub-grid cloud processes involve only the interstitial aerosol, i.e. the aerosol–cloud coupling is not considered in convective parameterization. Therefore, the indirect effects are well resolved for domains with resolution less than 10 km in the version of WRF-Chem used in this work.
Physical and chemical parameterizations used in the simulation.
As mentioned in Sect. 2.1, two key uncertainties in SOA production are the
deposition velocity and ageing factor of OCVs. The first is assumed to be
25 % (this value is called “deposition factor” in WRF-Chem) of dry
deposition velocity of HNO
We simulated the period from 14 to 30 May 2008. We chose this period because
aerosol and cloud state-of-art measurements were available to evaluate the
model (see Sect. 3). Moreover, during this period anticyclonic and cyclonic
meteorological conditions were observed which allows for the evaluation of
the model under varying conditions. The initial and boundary meteorological
conditions for D1 are provided by the European Centre for Medium range
Weather Forecast (ECMWF) analyses with a horizontal resolution of
0.5
A series of 30 h simulations were performed on each day starting at 00:00 UTC, with the first 6 h discarded as model spin-up for meteorology. Meteorology of D1 is reinitialized from global analysis, while initial and boundary meteorology conditions for D2 and D3 are taken from D1 and D2, respectively. For all three domains, the chemical initial state is restarted from the previous run, while the chemical boundary conditions of D2 and D3 are taken from D1 and D2, respectively. The first 13 days of May 2008 are also simulated to spin-up the chemistry.
Anthropogenic emissions data are taken from the Netherlands Organization for Applied Scientific Research (TNO)
2007 inventory (Kuenen et al., 2014). TNO is a gridded European inventory
with resolution of 0.125
Horizontal and vertical interpolation, temporal disaggregation, NMVOC speciation, and aggregation of emissions into WRF-Chem species is done following Tuccella et al. (2012), with minor updates described in Curci et al. (2015a). In order to prevent spurious concentration of aerosol particles, the distribution of aerosol emissions into WRF-Chem modes is based on the low emission scenario of Elleman and Covert (2010). In all 10 % of the emitted aerosol mass is attributed to Aitken mode, and 90 % to the accumulation mode.
Biogenic emissions are calculated online with Model of Emissions of Gases and Aerosols from Nature (MEGAN) (Guenther et al., 2006). Dust and sea salt emissions from soil and seawater are calculated online in the simulations.
We evaluated model performances in D3 against ground and aircraft-based data collected in May 2008 during the Intensive Cloud Aerosol Measurement Campaign (IMPACT) in the frame of the EUCAARI project (Kulmala et al., 2011). Model results were also evaluated against MODIS satellite data.
An overview of the synoptic conditions of May 2008 over central Europe is given by Hamburger et al. (2011). The first 15 days of May are characterized by an anticyclonic block, while the period from 16 to 31 is dominated by westerly wind and passage of several fronts. The days from 17 to 20 May are referred as “scavenged background situation” (Mensah et al., 2012), because they are dominated by a northerly wind from the North Sea associated with a low aerosol mass loading, due to wet scavenging. The period starting from 23 May is dominated by long-range transport of dust from Sahara desert (Roelofs et al., 2010; Bègue et al., 2015).
Meteorological and aerosol ground-based measurements used in this study are
collected in Cabauw (the Netherlands) at Cabauw Experimental Site for
Atmospheric Research (CESAR) observatory Cabauw (Fig. 1). CESAR observatory
is a tower located at 51
Standard meteorological variables are collected at 2, 10, 20, 40, 80, 140, and 200 m height. Furthermore, in this study we used the measurements of temperature and relative humidity profiles obtained with radiometer, and aerosol speciation from aerosol mass spectrometry (AMS) collected at 60 m (Mensah et al., 2012).
The model is also compared to O
Although Cabauw supersite provides very detailed measurements, it could not
be enough to characterize the model performance near the surface. Therefore,
WRF-Chem is also compared to daily PM
During May 2008, a French ATR-42 research aircraft performed 22 research
flights (RF). In this work we used 14 RF to evaluate the model. Their tracks
are reported in Fig. 1, while flight number and take-off and landing time are
reported in Table S1 in the Supplement. RF50, RF55, RF56, RF57, RF58, RF61,
and RF62 were conducted around the Cabauw supersite, in order to study the origin
and characteristic of air masses collected at Cabauw. Other RFs were aimed at
the study of aerosol properties along a quasi-Lagrangian flight track, with
west–east and north–south transects. ATR-42 was equipped with instrumentation
suitable for aerosol–cloud interaction measurements. We used the measurements
from a condensation particle counter (CPC), the CPC3010 with a cut-off size of
15 nm, a Cloud Condensation Nuclei Counter (CCNC) for CCN number
concentration measurements, and an AMS. During the campaign a scanning
mobility particle sizer (SMPS) was used to measure the number size
distribution of aerosol particles with a diameter in the range of
0.02–0.5
The model was also evaluated with MODIS-Aqua aerosol and cloud data. The Level 2 products used here are MYD04 and MYD06 collection 051 for aerosol and clouds, respectively. For ease of comparison with model output, both satellite and model data were regridded onto a common lat.–long. regular grid. Model output is sampled at same time and location of each MODIS pixel, and then data are averaged in space and time over the same grid. In this study the horizontal spacing of the common grid is 4 km.
Model results are compared to ground-based and aircraft observations, as
detailed in Sect. 3. The statistical indices used are the Pearson's
correlation coefficient (
Statistical indices of the comparison of WRF-Chem to observations
of temperature (
Observed and simulated time series of vertical profile of the
temperature
Figure 2 shows the observed and modelled time series of hourly vertical
profiles of temperature and relative humidity at the Cabauw supersite. WRF-Chem
reproduces the day-to-day variation of temperature, before, after, and during
the wet period. As shown by statistical indices (Table 2), within the first
200 m, the model reproduces the temperature with a correlation of 0.93–0.95
and a mean bias of about
The model reproduces the vertical structure of relative humidity (Fig. 2b)
over the whole period, but it has the tendency to overestimate
(underestimate) the higher (lower) observed values. This behaviour is more
evident during scavenging days, when the relative humidity between
1000 and 2000 m is overestimated on average by 40 %, but sometimes up to
60 %. Errors of this magnitude in simulating the vertical profile of RH
were already found in previous works (Misenis and Zhang, 2011; Luo and Yu,
2011). Nevertheless, the model correlation and mean bias are 0.84 and
The biases of the temperature and relative humidity could be due to a
misrepresentation of soil (and sea) temperature and soil moisture or by
misrepresentation of the clouds and rain. These two problems are tightly
coupled via land surface–atmosphere interaction. The errors in the simulation
of surface moisture and energy budget influences the fluxes of latent heat
and moisture in the atmosphere, affecting the local circulation, convective
available potential energy (CAPE), cloud formation, and rain pattern (Pielke,
2001; Holt et al., 2006). Moreover, WRF-Chem tends to fail simulating the
thermodynamic variables near coastlines, because the uncertainties of land
use data may play an important role (Misenis and Zhang, 2010). Initial and
boundary meteorological conditions may also play an important role. Bao et
al. (2005) demonstrated that meteorological prediction is sensitive to used
input data. They showed that varying the inputs used as initial and boundary
conditions, the mean daily model bias ranges from
Observed (black lines) at Cesar tower and simulated (red lines)
time series of wind speed
In Fig. 3 we compare the time series of observed and predicted wind speed and
direction at several heights of Cabauw tower. WRF-Chem captures the diurnal
trend of wind speed, but it overestimates the wind speed during the night.
Generally, we found the higher correlation at 10 and 200 m (0.78 and 0.76
respectively) and higher NMB between 20 and 40 m (
Figure 4 displays the comparison between the observed and modelled hourly
time series and average diurnal cycles of O
WRF-Chem reproduces the day to day variations of O
WRF-Chem simulates the NO
Observed and simulated time series of gas-phase species
Ammonia is reproduced with a correlation of 0.43. WRF-Chem underestimates the
NH
The nitrous acid concentrations are not well captured by the model and are underestimated by 95 %. This bias could be partly explained by the inefficiency of NO oxidation, the only important reaction known to form HONO. Li et al. (2014), indeed, demonstrated that the major sources of HONO are some unknown reactions that consume nitrogen oxides and hydrogen oxide radicals.
The model reproduces the measured SO
The different uncertainties found for the involved species may depend on the
choice of the chemical mechanism. Knote et al. (2015a) compared several
chemical mechanisms within a box model constrained by the same meteorological
conditions and emissions, and found that the prediction of the O
Figure 5 shows the simulated and observed time series and diurnal cycle of
aerosol sulfate, nitrate, ammonium, and organic matter, at CESAR
observatory. WRF-Chem simulates the measured SO
WRF-Chem captures the daily variations of SO
NO
The behaviour of the simulated NH
As in Fig. 4, but for aerosol mass speciation at Cesar observatory observed at 60 m height.
Similar performances are found in reproducing inorganic aerosols at other
Dutch EMEP sites (see Sect. 3.1). Daily SO
Organic matter is reproduced with a correlation coefficient of 0.75. WRF-Chem
reproduces the right concentration during the dry period (the decrease in the wet
days) and following recovery. The mean bias is negative by about
0.4
The reader should consider that aerosol composition measurements performed
with the AMS are representative of particles with a diameter between roughly
100 and 700 nm, whereas the model is evaluated with aerosol concentration
representative of PM
The model evaluation performed so far is representative of few points in the
domain and does not include other aerosol components like black carbon or
primary PM. This could limit our understanding of the model behaviour. In
order to overcome this shortcoming, WRF-Chem is also compared to daily
PM
The results obtained here are statistically consistent with other modelling
studies over Europe (e.g., Lecœur and Signeur, 2013; Zhang et al., 2013b;
Balzarini et al., 2015). For example, the results of European annual
simulations of Balzarini et al. (2015) exhibited a correlation of 0.48, 0.60,
and 0.56 for surface SO
Observed and modelled mean values, standard deviations, and relative number (expressed as percentage) of aerosol species, number of aerosol particles, cloud condensation nuclei, over the whole period of the aircraft campaign in boundary layer and free troposphere.
Box plot of SO
The comparison of WRF-Chem to aircraft data is done by interpolating the model output point by point along the flight track. Observed and modelled aircraft data are presented by using the box plots for planetary boundary layer (PBL) and free troposphere (FT) (see Fig. 6). The height of the PBL was lower than 1600 m during the whole campaign (Crumeyrolle et al., 2013). Therefore, we considered for PBL and FT concentrations the data below and above 1600 m up to 3000–4000 m, respectively. This rough approximation of PBL height could affect the comparison of the model to data.
Figure 6 displays the observed and modelled box plots of the mass
concentration of SO
The average concentrations of inorganic aerosols show little absolute error
(2–8 %) with respect to the observations in the PBL, while the NO
Although the predicted aerosol mass of each species is within the range of the observed values for most of the flights used in this study (see Fig. 4), the model does not capture the full range of the measured concentrations. This assertion is made quantitative by the standard deviations reported in Table 3. The predicted standard deviations for each species are lower than observed. In the PBL, the observed and modelled standard deviations differ by 4–10 % for inorganic ions and 55 % for OM. In the FT, the differences are higher than in the PBL. The model predicts standard deviations lower than 10–40 % for inorganic particles and lower than 65 % for organic matter with respect to the measurements.
For the purpose of this analysis, it is also interesting to explore how the
model reproduces the relative fraction of aerosol mass species with altitude
(see Table 3). WRF-Chem overpredicts the relative fraction of the SO
Looking at the individual flights, it is possible to note how the model captures the aerosol mass trend as a function of the synoptic frame in both the PBL and FT, during the dry period, scavenging days, and dust period characterized by southerly wind and passage of several fronts. The FT is a layer mainly affected by long-range transport and cloud contamination. Therefore, the relative small bias in simulating aerosol inorganic mass in FT means that the model resolves quite correctly the large-scale transport and processes related to clouds.
Nevertheless, it should be noted that SO
The simulated OM concentration is always at the lower end of the observed
variability. Several factors may explain this systematic bias. First of all,
our simulations do not include the processing of organic compounds in aqueous
chemistry. SOA may be produced in the clouds (Hallquist et al., 2009).
Modelling studies suggest that the contribution of cloud chemistry to SOA
budget is almost as much as the mass formed from the gas phase (Ervens et
al., 2011). OM prediction is also affected by meteorological errors. Bei et
al. (2012) found that the uncertainties in meteorological initial conditions
have a significant impact on the simulations of the peaks, horizontal
distribution, and temporal variation of SOA. The same authors demonstrated
that the spread of the simulated peaks can reach up to
4.0
Same as Fig. 6 but for observed and simulated PM
In order to have a more complete overview of the model skill in reproducing
the upper air aerosol mass concentration, we also compare the observed and
modelled PM
Vertical profiles (shadow) along the flight track of 14 May (RF50)
of modelled PM
Model correlation with observations is high, 0.75 and 0.80 in PBL and FT,
respectively. The absolute mean bias is
Although the box plot and statistical summary (Table 3) provide significant
information on the model performances, the model skills in reproducing
vertical profiles of aerosol properties need to be evaluated. Therefore, we
also look at model vertical profiles along the flight tracks. As an example,
we have chosen the 14 May 2008 (RF50) for two reasons: first, there is a
relatively large contribution of OM, SOA (70–85 % of OM), and CCN (see
Fig. 10), and second, it is a day of high pressure; thus, the interpretation
of the results is not complicated by cloud processes. Figure 8 displays the
comparison of modelled vertical profiles of PM
The comparison of WRF-Chem output with aircraft measurements of the number concentration of condensation nuclei (CN) and of CCN at 0.2 % of supersaturation is done by using the box plots as for aerosol mass. In this case the modelled and measured data are smoothed by using a 10 s running mean.
Figure 9 reports the comparison of observed and modelled CN within PBL and FT. The measured and predicted average, standard deviation, and correlation of the CN number over the whole period of our analysis are reported in Table 3.
Same as Fig. 6 but for observed and simulated condensation nuclei (CN) concentrations. The blue colour represents the observations while the model is displayed in red colour.
The model resolves the decrease of a factor of 5–6 of CN concentration between the PBL and the FT. The differences in simulated concentrations between land and sea (RF51 and 52) are also captured by the modelling system. Nevertheless, WRF-Chem overestimates, on average, the observed CN by a factor of 1.4 within PBL and 1.7 within the FT. The bias is less pronounced above the sea during the RF51 and RF52, where the anthropogenic sources are less important. Moreover, it should be noted that in some cases, for example during the RF56, 57, and 58, the predicted CN are completely outside the range of the observed values. In these cases the predicted CN are biased high by about a factor of 2–3. Predicted CN concentration shows a higher variability than measured, especially in the free troposphere where the difference of modelled standard deviation is biased high by 155 %. Anyway, the modelled CN concentration correlates well with the observed one (0.40 and 0.74 in PBL and FT, respectively). These values are consistent with the 0.61 found by Luo and Yu (2011) studying the new particle formation and its contribution to CN with a version of WRF-Chem including an advanced aerosol microphysical model.
Figure 10 shows the comparison of observed and modelled CCN at 0.2 % of
supersaturation. The measured and predicted average and standard deviation of
CCN are showed in Table 3. The bias of simulated CCN
Same as Fig. 6 but for observed and simulated cloud condensation nuclei (CCN) concentrations at 0.2 % of supersaturation. The blue colour represents the observations, the model is displayed in red colour.
The aerosol particles that mostly contribute to the CCN number are those of
accumulation and coarse modes, and accumulation and coarse-mode particles are
also the major contributor to PM
Aerosol optical thickness at 500 nm from MODIS-Aqua
The analysis of CCN efficiency reveals other interesting features in the
model behaviour. The CCN efficiency is given by the CCN
The so-calculated and modelled CCN efficiencies could be underestimated. In
general, the CCN efficiency should be computed with the aerosol population
with size larger than the minimum activation diameter (Asmi et al., 2012).
The latter depends on the aerosol type and ranges from about 50 to 125 nm.
We calculated the observed CCN
WRF-Chem output was also compared to AOT, cloud microphysical, and optical properties retrieved by MODIS-Aqua.
Figure 11 shows the comparison between the AOT at 550 nm measured by MODIS
and the corresponding AOT predicted by the model during the high-pressure
period on 14 May 2008. WRF-Chem reproduces the spatial distribution of
observed AOT, such as the lowest values in the southern part of the domain or
the highest values around Cabauw, but underestimates the strong gradient
between the eastern boundary and the centre of the domain. The model
overestimates the regional mean of AOT; indeed, the domain averages are
0.38
The 17–19 May 2008 averages of droplet effective radius at cloud top (first row), liquid water path (second row), and liquid cloud optical thickness (third row), retrieved using MODIS-aqua observations (first column), predicted by model in the references run (CTRL, second columns) and sensitivity test without SOA (NOSOA, third column).
As the WRF microphysics scheme accounts for aerosols only within liquid
clouds, comparison among predicted cloud optical and microphysical properties
with MODIS data was done separately for liquid, excluding mixed clouds. Top
liquid cloud
The comparison between the predicted and observed
As shown in Table 4,
The biases found here are quite different from the WRF-Chem study by Yang et
al. (2011) on the modelling of marine stratocumulus in the south-east Pacific.
They have shown a bias of
MODIS and modelled mean values and standard deviations of droplet effective radius at cloud top, liquid cloud water path, and liquid cloud optical thickness, on 17–19, 25–27, and 28–30 May 2008.
The 17–19 May 2008 averages of observed and simulated distribution
function of droplet effective radius at cloud top
Figure 13 displays the distribution functions (DF) of
Now it is interesting to analyse the model behaviour in reproducing the total CWP and COT given by contribution of all cloud phases. Modelled CWP was calculated by vertically integrating all cloud mixing ratios (water, rainwater, ice, snow, and graupel). Predicted COT is given by the contribution of the liquid water and ice. The contribution of the liquid water was calculated as described above for liquid water cloud. The contribution of ice phase to COT was calculated following Ebert and Curry (1992).
MODIS and modelled mean values and standard deviations of cloud water path and cloud optical thickness of clouds in all phases, on 17–19, 25–27, and 28–30 May 2008.
As in Fig. 12, but for clouds in mixed phase.
Figure 14 displays the comparison between observed and predicted CWP and COT in P1, whereas the same figures for P2 and P3 are reported in the Supplement (Figs. S8 and S9). Although for all three cases, the model reproduces with good approximation the shape and localizations of the cloud systems, CWP and COT are systematically overestimated (except COT in P2). As shown in Table 5, the predicted domain average of CWP presents, indeed, a bias of 62, 41, and 80 % for P1, P2, and P3, respectively, whereas the bias of COT is about 15 % in P1 and P3.
At this point of the analysis, although the aerosol–cloud interaction is a very complex non-linear process, we are able to relate the model error in aerosol particles to the uncertainties in cloud prediction. The overestimation of CN leads to overprediction of the CCN. Higher number of CCN means clouds with a higher number of cloud droplets, higher water content, smaller droplets, and clouds optically deeper.
In addition to the uncertainties in aerosol particle simulation, one typical source of error in the prediction of cloud fields is the choices related to the model setup. For example Otkin and Greenwald (2008) found a strong sensitivity of cloud properties while evaluating the response of the WRF model to the permutation of several PBL and cloud microphysical schemes. Moreover, the same authors have shown that the low level clouds are sensitive to PBL parameterization, whereas the upper level clouds are sensitive to both PBL and microphysics schemes.
One element that may affect the model–satellite comparison are the uncertainties associated with the retrieval. For example, in South Pacific stratocumulus, MODIS overestimates the droplet effective radius by 13–20 % with respect to concomitant in situ measurements (Painemal and Zuidema, 2011; King et al., 2013). The overestimation of COT by MODIS results in the overestimation of CWP (King et al., 2013). Henrich et al. (2010) have shown systematic differences between MODIS data and in situ observations. Indeed, analysing a system of thin cumulus clouds during the EUCAARI campaign, they also found that MODIS overestimates the droplet effective radius by a factor of 2–3 and COT is 2–3 times lower than the in situ measurements.
The last part of this study focussed on the evaluation of the impact of SOA on the simulation of cloud fields. We performed sensitivity simulations during P1, P2, and P3 without the SOA (NOSOA), and compared them to the reference run (CTRL) discussed so far. NOSOA runs are carried out only in the higher resolution domain (D3). The simulations of all three periods are initialized at 00:00 UTC with the same meteorological and chemical input data used for CTRL, except chemical initial conditions that are restarted by a previous NOSOA run. Each period is preceded by 30 h of simulation used as spin-up for D3 chemistry. The sensitivity simulation is performed zeroing the arrays pertaining to SOA. Thus, the SOA fields are not affected by incoming SOA from domain boundaries or by local production. We did not perform the sensitivity tests with the SORGAM option because this model produces very little SOA mass concentrations (Grell at al., 2005; McKeen et al., 2007; Tuccella et al., 2012). Therefore, we may assume that simulations with SORGAM and without SOA (in VBS option) are roughly equivalent. The advantage of this assumption is that the model is forced with the same initial meteorological conditions and boundary meteorological and chemical conditions as the CTRL simulation. The use of SORGAM would require running the model on all three domains, leading to different results on D2 which is used to initialize D3. Finally, this would introduce dependencies on the D3 input data making the comparison not directly comparable to the CTRL run.
Maximum dBZ at 06:00 UTC of 17 May for CTRL run
The comparisons of
Figures 14, S8, and S9 report the comparison of modelled CWP and COT of all cloud phases predict in CTRL and sensitivity runs with MODIS data. As well as for liquid phase, including SOA aerosol particles in the runs, the WRF-Chem skills to reproduce the observed pattern of observed CWP and COT increase. As shown in Table 5, domain-averaged CWP and COT are larger up to about 50 % in CTRL with respect to NOSOA.
Vertical profile of PM
As in Fig. 16, but for water (colour), and frozen (black contours) hydrometeors.
Now it is interesting to explore the impact of SOA on the vertical structure
of the cloud fields. As an example we chose the 17 May because around
06:00 UTC a frontal system associated with a trough from the North Sea crossed
the Benelux area (Fig. S10). In both runs, some isolated and shallow clouds
form during the night. When the cold front reaches Benelux around
05:00–06:00 UTC, a low pressure centre forms (Fig. S11). The winds rotate
around the low pressure with speeds up to 14 m s
Secondary organic aerosol particles play an important role in aerosol–cloud–radiation interaction because they contribute to the global budget of radiation and cloud condensation nuclei (CCN). The introduction of SOA particles in numerical simulations has the potential to reduce the uncertainties on the prediction of meteorological fields and air quality. To this aim, a parameterization for SOA production based on the recent VBS approach was coupled with microphysics and radiative schemes in the WRF-Chem community model.
The performance of the updated model at cloud resolving scale (2 km horizontal resolution) was evaluated using ground- and aircraft-based measurements collected during the IMPACT-EUCAARI campaign and the data from the MODIS satellite instrument. The study focuses on the Benelux area, around the supersite of Cabauw, from 14 to 30 May 2008. The analysed period was characterized by a few days of high pressure (14–15 May), followed by a scavenged background situation (17–20 May), and finally by long-range transport of Saharan dust with the passage of southerly fronts (23–31 May).
The model reproduces the variations of meteorological variables as a function
of the synoptic frame. The model broadly captures the inter- and
infra-diurnal variability of O
The analysis of aircraft data reveals that WRF-Chem captures the aerosol mass
trend both in the PBL and the free troposphere (FT). The predicted aloft
aerosol mass of each species is within the observed values range, but the
model does not capture the full range of the measured concentrations; the
modelled standard deviations of aerosol mass are lower than those observed.
Nevertheless, SO
Condensation nuclei (CN) are overestimated by a factor of 1.4 and 1.7 in the PBL and FT, respectively. However, in some cases, the predicted CN are overestimated by a factor of 3. Predicted CN show higher variability than measurements. The model correlation with observed CN is 0.40 and 0.74 in PBL and FT, respectively. These values are consistent with the 0.61 below 10 km of altitude found by Luo and Yu (2011) in the eastern USA with WRF-Chem including an advanced aerosol microphysical model. Model biases in predicting CN are attributable in large part to the uncertainties of primary particle emissions (mostly in the PBL) and to the nucleation rate.
The bias of simulated CCN is more contained with respect to that of CN. The
CCN efficiency (CCN
The bias of simulated CN affects the prediction of droplet
In summary, the model behaviour of this new chemistry option in WRF-Chem in simulating the relationship between aerosol and cloud fields may be summarized in this way. The overestimation of CN results in the overprediction of the CCN. A higher number of CCN leads to clouds with a higher number of cloud droplets, higher water content, smaller droplets, and clouds optically deeper.
As test application of the new chemistry option, we performed a sensitivity
simulation where SOA mass concentration is set to zero. The aim was to answer
two questions:
Does the introduction of SOA particles improve the numerical prediction of cloud fields? The introduction of SOA in the numerical simulations improves the predicted
spatial pattern of microphysical and optical properties of cloud in liquid
and all phases. NOSOA runs show an average What is the impact of SOA particles on cloud development? The analysis was conducted on a convective system. The simulated radar
reflectivity is larger for run with SOA; i.e. the intensity of the storm is
stronger in the CTRL run. The CTRL simulation exhibits a larger number of
hydrometeors and stronger updrafts and downdrafts. The larger differences in
the simulated fields of vertical wind and hydrometeors are associated with the
larger differences of PM
On the basis of the results discussed in this work, the option RACM-MADE-VBS coupled with cloud microphysics and radiation allows the WRF-Chem community to use a powerful tool for the study of the aerosol–cloud interactions, improved in terms of representation of the aerosol processes with respect to previous versions based on the RADM/MADE/SORGAM scheme.
For the future, there is still large space for improvements. For example, a more advanced treatment of deposition of organic condensable vapours is desirable. Moreover, the missing production of SOA in cloud is a gap that should also be filled. Finally, the extension of aerosol–cloud interaction to the ice-phase would lead to a complete representation of the aerosol indirect effects.
The new chemistry option in namelist.input is
The first step is to create a new chemistry option. The package
racm_soa_vbs_aqchem_kpp (chemopt New chemistry package is a KPP option. Therefore, we created a new
subdirectory in
The last step is to update the subroutines in the
Let Obs
The Pearson's correlation ( Mean bias:
Normalized mean bias (NMB):
Normalized mean gross error (NMGE):
The code updated, described, and evaluated here will be incorporated in the
next available release of WRF-Chem. The users will be able to freely download
the code from the WRF website
(
This work was founded by the University of L'Aquila (Italy) and Regione Abruzzo in the frame of the “High Formation Project” (P.O.F.S.E 2007-2013), and the Italian Space Agency in the frame of the PRIMES (contract I/017/11/0) project. Paolo Tuccella is grateful to the National and Oceanic Administration (NOAA) of Boulder (CO, USA) for the hospitality, to Ravan Ahmadov and Stuart McKeen for the precious and profitable discussions about the parameterization for secondary organic aerosol, and to Steven Peckham for the assistance in the implementation of the new chemical option in the repository version of WRF-Chem. We are grateful to the Euro-Mediterranean Center on Climate Change (CMCC) for having made available their supercomputer to perform the numerical simulations. The authors thank Hugo Denier van der Gon for providing the TNO emissions. Finally, the authors are grateful to two anonymous reviewers for their suggestions that helped to improve this paper. Edited by: F. O'Connor