Simulating aerosols over Arabian Peninsula with CHIMERE: Sensitivity to soil, surface parameters and anthropogenic emission inventories

Atmospheric Environment 128 (2016) 185e197

Contents lists available at ScienceDirect

Atmospheric Environment journal homepage: www.elsevier.com/locate/atmosenv

Simulating aerosols over Arabian Peninsula with CHIMERE: Sensitivity to soil, surface parameters and anthropogenic emission inventories S. Naseema Beegum a, *, Imen Gherboudj a, Naira Chaouch a, Florian Couvidat b, Laurent Menut c, Hosni Ghedira a a b c

Earth Observation and Hydro-Climatology Lab, Masdar Institute of Science and Technology, Abu Dhabi, United Arab Emirates Institut National de l’Environnement Industriel et des Risques, Verneuil-en-Halatte, France Institut Pierre-Simon Laplace, Laboratoire de Meteorologie Dynamique, Ecole Polytechnique, Palaiseau, France

h i g h l i g h t s
Modeling of aerosol optical depth using the chemistry transport model CHIMERE.
Adaption of the CHIMERE model for the Arabian peninsula.
Integration of new datasets on soil/surface properties and anthropogenic emissions.
Significant improvement in AOD simulation with the new datasets.

a r t i c l e i n f o

a b s t r a c t

Article history: Received 7 October 2015 Received in revised form 4 January 2016 Accepted 5 January 2016 Available online 7 January 2016

A three dimensional chemistry transport model, CHIMERE, was used to simulate the aerosol optical depths (AOD) over the Arabian Peninsula desert with an offline coupling of Weather Research and Forecasting (WRF) model. The simulations were undertaken with: (i) different horizontal and vertical configurations, (ii) new datasets derived for soil/surface properties, and (iii) EDGAR-HTAP anthropogenic emissions inventories. The model performance evaluations were assessed: (i) qualitatively using MODIS (Moderate-Resolution Imaging Spectroradiometer) deep blue (DB) AOD data for the two local dust events of August 6th and 23rd (2013), and (ii) quantitatively using AERONET (Aerosol Robotic Network) AOD observations, CALIPSO (Cloud-Aerosol Lidar and Infrared Pathfinder Satellite Observation) aerosol extinction profiles, and AOD simulations from various forecast models. The model results were observed to be highly sensitive to erodibility and aerodynamic surface roughness length. The use of new datasets on soil erodibility, derived from the MODIS reflectance, and aerodynamic surface roughness length (z0), derived from the ERA-Interim datasets, significantly improved the simulation results. Simulations with the global EDGAR-HTAP anthropogenic emission inventories brought the simulated AOD values closer to the observations. Performance testing of the adapted model for the Arabian Peninsula domain with improved datasets showed good agreement between AERONET AOD measurements and CHIMERE simulations, where the correlation coefficient (R) is 0.6. Higher values of the correlation coefficients and slopes were observed for the dusty periods compared to the non-dusty periods. © 2016 Elsevier Ltd. All rights reserved.

Keywords: CHIMERE Chemistry transport model Aerosol optical depth EDGAR-HTAP emissions Surface roughness length Soil erodibility

1. Introduction Mineral dust is one of the dominant components of the total atmospheric aerosol loading (Kohfeld and Tegen, 2007). On average, an annual total of ~2000 Tg of dust is emitted in the

* Corresponding author. E-mail address: [email protected] (S.N. Beegum). http://dx.doi.org/10.1016/j.atmosenv.2016.01.010 1352-2310/© 2016 Elsevier Ltd. All rights reserved.

atmosphere, and around 20% of these emissions are from the Middle-East region (Iran, north-eastern Iraq and Syria, and the southern Arabian Peninsula) (Shao et al., 2011). During its lifetime in the atmosphere, mineral dust has great impact on the EarthAtmosphere system (IPCC, 2013), cloud formation (DeMott, 2003), air quality, visibility and human health (Giannadaki et al., 2014), and marine ecosystem (Al-Shehhi et al., 2014). These implications emphasize the importance of understanding the occurrence and intensity of dust events, as well as their spatiotemporal

186

S.N. Beegum et al. / Atmospheric Environment 128 (2016) 185e197

distribution. Though the observations (ground/satellite based) are the key to validating and refining the models; three dimensional dust forecast models with appropriate dust emission schemes and datasets can provide accurate AOD simulations over the domain of interest. During the last few decades, a large number of regional and global models have provided dust forecasts. These include GOCART: Goddard Chemistry Aerosol Radiation and Transport model operated by National Aeronautics and Space Administration-Goddard Space Flight Center (Chin et al., 2000), MACC-ECMWF: Monitoring Atmospheric Composition and Climate model operated by European Centre for Medium-Range Weather Forecasts (Morcrette et al., 2009), MetUM: Unified Model from UK Met Office (Woodward, 2001), GEOS-5: Goddard Earth Observing System Model Version 5 (Colarco et al., 2010), and NGAC: NEMS-GFS Aerosol Component model from National Centers for Environmental Prediction (Lu et al., 2010), BSC-DREAM8b: Dust REgional Atmospheric Model operated by the Barcelona Supercomputing
rez et al., 2006), NMMB/BSC-Dust: non-hydrostatic Center (Pe multi-scale model operated by Barcelona Supercomputing Center
rez et al., 2011), DREAM-NMME-MACC: non-hydrostatic multi(Pe scale model by the coupling between DREAM and MACC models, and CHIMERE: A multi-scale chemistry transport model for air quality and dust forecasts (Menut et al., 2005). These models were coupled online or offline with either short/medium ranges regional weather forecast models or large scale global climate models to simulate the mineral dust cycle. Global models reproduce the seasonal dust features at a coarse spatial resolution (1.5 e5 ), while the regional models are efficient to simulate the mesoscale dust events. In addition, the regional models use more complex dust emission schemes as their resolution is closer to the small-scale dynamical processes. Furthermore, the comparison and validation with observations of dust events are better suited for regional models. CHIMERE is one of the chemistry transport models dedicated to provide dust and air quality forecast at regional scale. Several sensitively analyses were carried out for CHIMERE using different: (i) meteorological data set (De-Meij et al., 2009), (ii) meteorological data grid resolution (Srivastava et al., 2014), (iii) soil and surface data sets (Menut et al., 2013, 2005), (iv) soil size distribution (Menut et al., 2005), (v) soil moisture data (Briant et al., 2014), and (vi) number of aerosol size bins (Foret et al., 2006). The dust emission parameterization scheme (Alfaro and Gomes, 2001) is found to be strongly related with the wind speed and therefore accurate meteorological parameter is an essential factor for realistic aerosol simulations (De-Meij et al., 2009). The spatial interpolation of the meteorological parameters is observed to reduce the accuracy of the simulated aerosol concentrations (Srivastava et al., 2014). In addition, Menut et al. (2013) have analyzed the sensitivity of the dust emission scheme in Northern Africa using different datasets of soil texture (LISA, STATSGO-FAO) and roughness length (LISA, ERS,USGS) and their results indicate that ERS (European Remote Sensing) and STATSGO-FAO data sets provide realistic spatial patterns of dust emission. Even though all these sensitivity studies have improved the performance of CHIMERE model in general, the fine-tuning of the model over the Arabian Peninsula with realistic datasets on soil and surface properties is virtually non-existent. The objective of the study is to forecast the atmospheric dust over the Arabian Peninsula using the CHIMERE model. For this purpose, different configurations with new datasets describing soil/ surface properties and emission inventories are tested to fine-tune the CHIMERE model over the domain. This adaptation process of the model includes: (i) default configuration with the built-in soil/ surface properties at lower (500 hPa) and higher (200 hPa) vertical

levels (ii) MODIS soil erodibility and ERS scatterometer derived surface roughness length (iii) MODIS soil erodibility and ERAInterim surface roughness length (iv) keeping the latest configuration with EDGAR-HTAP anthropogenic emission inventories, and (v) nested runs at 9 km resolution with final configuration. These datasets are the best available database in view of the existing literature survey over the region. Since simulating the background dust-concentration in good accuracy is a challenge for any dust models rather than simulating strong dust storm events, we have selected August (2013), which has moderate dusty and normal days, for the adaption of the model. Qualitative and quantitative validations of the adapted model AOD simulation are performed using different satellites-based products (MODIS DB and CALIPSO vertical profiles), AERONET observations, and other forecast model outputs. 2. Methodology 2.1. CHIMERE model description CHIMERE is an open access multi-scale Eulerian chemistry transport model mainly intended to produce hourly forecasts of several aerosol and pollutant gas species concentrations (Bessagnet et al., 2008; Chimere documentation, 2014). The concentrations are computed by solving the continuity equation for processes such as emissions, transport, deposition, chemical reactions, and aerosol dynamics. The selection of appropriate parameterization schemes for the aforementioned processes is generally based on previous modeling studies, as detailed in Chimere documentation (2014). The 3rd order Parabolic Piecewise method (PPM) is found to be the most accurate scheme for horizontal advection, while the 2nd order Van Leer scheme is mostly considered for vertical advection. A Kdiffusion parameterization scheme is used for vertical turbulent mixing. The deposition module includes both wet (in-cloud and sub-cloud) and dry scavenging processes for aerosols in eight size bins. The aerosol hygroscopic growth factor is taken into account in the deposition processes using a thermodynamic equilibrium model (ISORROPIA) by computing the transition relative humidity between the phases. CHIMERE accounts for five types of emissions: (i) biogenic emissions, (ii) sea-salt emissions, (iii) anthropogenic emissions, (iv) fire emissions, and (v) dust emissions. The biogenic emissions are calculated using the MEGAN model (Guenther et al., 2006), while the marine aerosol production is based on the scheme proposed by Monahan et al. (1986). The anthropogenic emissions are incorporated into the CHIMERE model domain using the EDGAR-HTAP global emission inventories, collected as a part of HTAP (Hemispheric transport of Air Pollutants project) (Janssens-Maenhout et al., 2012). Dust emissions are calculated using the parameterization scheme proposed by Alfaro and Gomes (2001) (referred to hereafter as AG2001). This dust emission scheme is well validated over the deserts of Middle-East and North-African region (Gherboudj et al., 2015; Menut et al., 2013; Mokhtari et al., 2012). The soil and surface parameters, such as land-use, soil composition (percentage of clay, silt and sand), soil erodibility, and surface roughness length (z0), are required for the AG2001 scheme. While the soil composition over the world is provided from the STATSGOFAO database, the soil erodibility and surface roughness can be derived from different sources. 2.2. Model domain The CHIMERE model has been set up in one-way hourly nesting (no feed-back from inner to outer domain) configurations with 27 km and 9 km for the outer and inner domains, respectively. The

S.N. Beegum et al. / Atmospheric Environment 128 (2016) 185e197

outer domain is bounded between 9.5 Ne41.3 N and 36.8 Ee60.5 E, while the inner domain is bounded between 14.5 Ne35.5 N and 33 Ee58.5 E, as shown in Fig. 1. The vertical levels of the model has been changed from 8 (up to 500 hPa) to 15 (up to 200 hPa) from the surface. The model simulations are computed for a period of 30 days, from August 2 to 31 of 2013, and the entire year of 2010. 2.3. Database for the model adaptation 2.3.1. Meteorology and boundary conditions GOCART climatological simulations are used as the lateral boundary condition data, while the meteorological fields are simulated using the Weather Research Forecast Advanced Research (WRF-ARW) mesoscale model version 3.5 (Wang et al., 2014). WRFARW simulations are performed in one way nested run using the global analysis product of the National Center for Environmental Prediction (NCEP-GFS) at 0.5 spatial resolution for the initial and boundary conditions for the outer domain. 2.3.2. Soil erodibility The soil erodibility is one of the critical parameters required for the dust emission estimation. Several methods are described in the literature to derive the erodibility from/by: (i) topography of the terrain (Ginoux et al., 2001), (ii) geomorphic approaches (Jenson and Domingue, 1988), or (iii) surface reflectance (Grini et al., 2005). The erodibility map built-in CHIMERE model over the study domain is shown in Fig. 2a. The figure depicts poor spatial variability of this variable over the Arabian Peninsula desert with values equal to either 0 or 1, which do not represent the real conditions. Therefore, soil erodibility from the MODIS surface reflectance data has been derived, following the algorithm proposed by Grini et al. (2005). The erodibility is estimated from the ratio of the mean surface reflectance, calculated by averaging the monthly mean reflectance acquired for the period from 2010 to 2013, to the maximum surface reflectance (Fig. 2b). As displayed in this figure,

187

the soil erodibility shows high spatial variability, in comparison with the built-in map (Fig. 2a), with high values over the sandy areas because of their high surface reflectivity. The highest values are observed over Saudi Arabia, while moderately high values are observed over Iraq along the Tigris/Euphrates river basin, which are considered as the top most erodible surfaces. 2.3.3. Surface roughness length The surface roughness length (z0), which is used to estimate the threshold friction velocity, constitutes another key factor in mineral dust emission modeling. At synoptic scales, the parameter is referred as aerodynamic roughness length and its magnitude varies from a few centimeters to a meter (Menut et al., 2013). At this spatial scale, the USGS aerodynamic roughness length values are not suitable for the dust emission on relatively smooth arid regions. There were many earlier attempts to map z0 over deserts in mesoscale resolution (called the aeolian roughness length), which are capable for accounting for sparse vegetation and small obstacles, using different satellite observations. Marticorena et al. (2004) mapped aeolian roughness length over North-Africa based on measurements from the spaceborne POLDER (POLarization and Directionality of the Earth's Reflectance) sensor. Based on the measurements from the ERS scatterometer, Prigent et al. (2005) derived the aeolian roughness map over arid/semi-arid areas, which is incorporated into the CHIMERE model as a default map to estimate dust over Middle-east and North-Africa (MENA) region. However, it shows low spatial variability over the Arabian Peninsula, Fig. 3a. In addition, the values greater than 0.08 cm are considered erroneous. Several earlier studies used a constant value of 0.01 cm for locations where the surface roughness length is not available (Astitha et al., 2012). For the present study, aeolian roughness, z0 derived from ECMWF ERA-Interim re-analysis data at spatial resolution of 25 km (Simmons et al., 2007) are used to compute the spatial variability of dust emissions instead of the default map (Fig. 3b). 2.4. Supporting data for model validation The aerosol data used for this study are AERONET AOD, CALIPSO vertical profiles of aerosol and cloud extinction coefficients (Winker et al., 2010) and MODIS DB AOD (Hsu et al., 2013). AERONET AOD and CALIPSO profiles are also considered to: (i) identify two different dusty episodes (dusty and cloudy days from August 4the6th and dusty period from August 22nde24th) for the model adaptation, and (ii) quantitatively validate the AOD simulations carried out for the year 2010.

Fig. 1. Geographical domain for the CHIMERE model nested runs with the locations of AERONET stations.

2.4.1. Ground-based measurements Level 2 AERONET AOD products with an accuracy of ±0.01e0.02 (Holben et al., 2006) are considered for the validation of the model at the stations; namely Solar-Village (24.91 N, 46.40 E), KAUST Campus (22.31 N, 39.10 E), Eilat (29.51 N, 34.92 E) and Sede-Boker (30.86 N, 34.78 E) (Fig. 1). While the Solar-Village is an inland station located in the Arabian Peninsula, far away from the Arabian Gulf and other industrialized areas, the KAUST campus is a coastal location near the red-sea and ~95 km away from the urban center Jeddah. Sede-Boker and Eilat stations are located in Negev desert, where the effect of long-range transport of aerosols from west Asia/ Africa or urban pollution from Europe is significant (Maenhaut et al., 2014). Fig. 4 presents the level 2 hourly averaged AOD at 500 nm (AOD500) and Angstrom exponent (a, estimated using the Angstrom relation t ¼ bla , where b is the turbidity parameter) for August 2013 at the considered AERONET stations. As shown in the figure, the day-to-day variability in AOD as well as a is significant at

188

S.N. Beegum et al. / Atmospheric Environment 128 (2016) 185e197

Fig. 2. Soil erodibility over the Arabian Peninsula: a) built-in map in CHIMERE model, and b), derived from MODIS reflectance.

Fig. 3. Surface roughness length over the Arabian Peninsula derived from: a) ERS scatterometer data, which is built-in map in CHIMERE model, and b) ECMWF ERA-Interim data.

dust storm events. The KAUST station depicts more significant temporal variations in AOD, which might be caused by the longrange transport of aerosol from the surrounding deserts (West Asian and North-African) and urban regions. Similarly, identical temporal variations are observed for Sede-Boker and Eilat stations, but with lower values of AOD (<0.4) and higher values of a (mostly > 1.0). Though Sede-Boker is an arid region with significant amount of local production of dust, the observed lowest values of AOD with highest a values correspond to relatively quiet periods with less amount of dust activity and a significant amount of anthropogenic pollution. In fact, the dust events over this region are more frequent during the transition seasons with maximum dust activities during AprileMay, whereas the summer period is relatively dust-free (Derimian et al., 2006). However, as this station is in the downwind direction between dust from the Sahara/Arabian Peninsula and urban pollution from Europe, the various air-mass advection pathways influence the aerosol variability (Derimian et al., 2006).

Fig. 4. Day-to-day variability of aerosol optical depth at 500 nm (a) and (b) the Angstrom wavelength exponent (a) at four AERONET stations (Solar-Village, KAUST Campus, Eilat, and Sede-Boker).

all the stations. In general, Solar-Village and KAUST Campus depict similar temporal variability, in which AOD (a) varies between ~0.2 (0.05) and ~1.1 (1.2). The two distinct peaks observed in AOD (with lower values of a) at these two stations around the first (3rde6th) and third (22nde24th) week of August are associated with local

2.4.2. Satellite-based data Fig. 5 illustrates the CALIPSO level 2 (King et al., 2004) vertical profiles of aerosol and cloud extinctions (km1) for August 5the6th (descending passes) and August 23rde24th (ascending passes) over the Arabian Peninsula. These profiles were selected based on the maximum available spatial coverage over the domain. The accuracy of these products is ~30% and this is mainly attributed the difference to the low dust lidar ratio (Tesche et al., 2013). From Fig. 5a, it is clear that the August 5the6th period was persistently dusty and cloudy, especially over the southern part of the

S.N. Beegum et al. / Atmospheric Environment 128 (2016) 185e197

189

Fig. 5. Vertical profiles of aerosol and cloud extinction for the two considered dust events: a) August 5th e 6th and b) August 23rd e 24th.

Peninsula. Thicker aerosol layers with extinction coefficients reaching ~1 km1 are observed at the altitude region of 2e3 km from surface (Fig. 5a1). The cloud profiles of August 5th (Fig. 5a2) indicate the presence of dense optically thick clouds, particularly at higher altitudes, and this makes the aerosol detection difficult from space. Conversely, the vertical profiles of August 23rde24th show another dusty event but with no significant amount of clouds (Fig. 5b). Thick aerosol layers reaching up to 5 km altitudes with relatively cloud-free atmosphere are observed during this period. MODIS DB (AOD550), which was developed over visibly bright surfaces, is considered to be more accurate than other satellite products. Overall, high accuracy is observed for both MODIS DB Aqua/Terra products over the Middle East and North Africa region with a root mean-square error (RMSE)  0.24 and determination coefficient ranging between 0.63 and 0.86 depending on the location when validated against AERONET measurements (Hsu et al., 2013; Remer et al., 2013; Shi et al., 2013). These data are used in the subsequent sections for the qualitative validation of the model. 2.5. Model validation To estimate the degree of agreement between the AERONET AODOi observations and CHIMERE AODSi simulations over each station, the statistical parameters such as Mean Bias Error (MBE), Root Mean Square Error (RMSE), Correlation Coefficient (R), and Fractional Gross Error (FGE) are calculated for the hourly time-series data: The equations of the statistical parameters, following Seinfeld and Pandis (1998), are given in Appendix 1. 3. Results and discussions The CHIMERE AOD simulations are carried out over the Arabian Peninsula during August, 2013, for the two dust episodes (cloudy and dusty period of August 4the6th and dusty period of August 22nde24th) with default configuration of CHIMERE and the adapted version of the model. The validation of the AOD simulations are performed, both qualitatively and quantitatively, based on

the inter-comparison with the observations (in-situ and satellites) and various forecast models simulations. For the qualitative comparison, the AOD simulations at specific times during August 2013 are compared with the MODIS DB AOD. For the quantitative comparison, the AOD simulations are compared with: (i) AERONET measurements for the whole year of 2010, (ii) various forecast models for August 2013, and (iii) CALIPSO-derived aerosol extinction profiles for specific times during August 2013.

3.1. Performance evaluation of the default CHIMERE configuration Figs. 6 and 7 represent the simulated CHIMERE AOD600 at 10UT for the two dusty periods of August 4the6th and August 22nde24th, respectively. These simulations are performed using the ERS scatterometer derived surface roughness length (z0) and built-in soil erodibility with 8 (up to 500 hPa) and 15 (up to 200 hPa) vertical levels from the surface. The corresponding MODIS DB AOD550 maps displayed in Figs. 6c and 7c are used for the validation of these simulations. The simulations on August 4the6th showed that spatial pattern of the CHIMERE AOD600 values with 8 vertical levels underestimates the MODIS AOD550 (Fig. 6a). Therefore, the model has been reconfigured with 15 vertical levels and the results are given in Fig. 6b. The CHIMERE AOD600 simulations are found to improve significantly with an increase in vertical levels. This improvement in the results is due to the better representation of the long-range transport of aerosols in the model, especially through higher altitudes. Studies showed that during the summer/spring seasons, the mean dust layer tops around 4 kme8 km over the dust source regions of the MENA region during storm events (Liu et al., 2008). Signatures of long-range transport of dust plumes from the MENA region through higher altitude are observed over other continental locations as well (Prospero, 1999). Even with more vertical levels, the spatial pattern is poorly reproduced in the model. Similarly, the AOD simulations during the dust episode of 22nde24th August show underestimations for both vertical configurations, as displayed in Fig. 7a and b. Therefore, the simulation

190

S.N. Beegum et al. / Atmospheric Environment 128 (2016) 185e197

Fig. 6. CHIMERE simulated AOD (600 nm) over the Arabian Peninsula for the dust episode of August 4the6th using: a) default data of the CHIMERE model at vertical resolution of 8 levels up to 500 hPa, b) default data of the CHIMERE model at vertical resolution of 15 levels up to 200 hPa, and c) MODIS deep blue AOD (550 nm).

results clearly indicate that the default CHIMERE model with builtin soil/surface data is inadequate for the Arabian Peninsula region, even though the configuration of the model is reported to have reasonably good accuracy over the North-Africa in continental scales (Menut et al., 2013).

simulations are carried out to examine the effects of the soil/surface parameters, by considering MODIS derived erodibility and (i) builtin surface roughness length (ii) ECMWF ERA-Interim derived surface roughness length (iii) ECMWF ERA-Interim surface roughness length and EDGAR-HTAP emission inventories, and (iv) keeping the final configuration with a nested run.

3.2. Adaptation of the CHIMERE model for Arabian Peninsula Recent studies, however, have shown that the dust emission scheme is highly sensitive to the soil/surface properties, and therefore realistic datasets on soil/surface parameters are extremely important for simulation accuracy. As such, AOD

3.2.1. CHIMERE model with MODIS derived erodibility and default surface roughness length The AOD simulations are carried out for the full month of August, using the new erodibility map derived from the MODIS reflectance. The simulations on August 6th and 23rd, as

S.N. Beegum et al. / Atmospheric Environment 128 (2016) 185e197

191

Fig. 7. Estimated aerosol optical depth (600 nm) over the Arabian Peninsula for the dust episode of August 22nde24th using: a) default data of the CHIMERE model and at vertical resolution of 8 levels up to 500 hPa, b) default data of the CHIMERE model and at vertical resolution of 15 levels up to 200 hPa, and c) MODIS deep blue AOD (550 nm).

representatives of the two dust events, are shown in Fig. 8a1 and 8a2, respectively. It is interesting to note that these simulations showed significant improvement of the AOD values where the spatial pattern is close to the MODIS DB AODs displayed in Figs. 6c3 and 7c2. In fact, recent research studies have demonstrated that the performance of the dust models is affected by the accuracy of the erodibility representations in models. However, the AOD simulations are still underestimated, which indicates the importance of analyzing the influence of other parameters in simulating dust emissions.

3.2.2. CHIMERE model with MODIS erodibility and ECMWF derived surface roughness length The AOD simulations are carried out using ECMWF ERA-Interim derived surface roughness length (z0) (Fig. 3b), keeping the MODIS derived erodibility. As displayed in Fig. 8b1 and b2, the simulations improved drastically and depict good agreement with the MODIS DB estimates. These results confirm that the logarithmic wind profile method, employed for the estimation of the ERA-Interim surface roughness length from the wind profile data, is well representative for the study domain (Shao, 2008).

192

S.N. Beegum et al. / Atmospheric Environment 128 (2016) 185e197

Fig. 8. CHIMERE simulated AOD (at 600 nm) over the Arabian Peninsula using: a) MODIS erodibility, b) ECMWF ERA-Interim derived surface roughness length, and c) HTAP emission inventories on 6th (top panel) and 23rd (bottom panel) of August 2013 respectively.

3.2.3. CHIMERE model with anthropogenic emission inventories The aforementioned AOD simulations are carried out for the natural emissions, such as dust and sea-salt emissions. However, enhancement in urbanization and industrialization sectors (mining, refineries, manufacturing chemicals and plastics, construction activities) over the past few decades, along with the discovery of oil, has led to a significant increase in local urban pollution (Ghanem, 2001). In addition, the long-range transport of urban pollution from Europe and west-Asia has resulted in high levels of air pollution over the Arabian Peninsula (Gherboudj and Ghedira, 2014; Lelieveld et al., 2009). Previous dust modeling studies over the MENA region have shown that the dust model under-predicts the AOD for cases of high Angstrom exponents showing signifi
rez et al., 2006). cant influence from anthropogenic pollution (Pe Therefore, the AOD simulations are carried out with the incorporation of the global EDGAR-HTAP emission inventories in the latest configuration of the CHIMERE model (Section 3.2.2) and the results are shown in Fig. 8c1,8c2. The improvements in the AOD simulations suggest that the anthropogenic emission inventories are important in increasing the accuracy of model over the Arabian Peninsula. It is interesting to note that even during August, in which the impact of anthropogenic aerosols is reported to be low; the non-dust aerosol contribution is significant. This finding speculates that during winter season, the emission inventories will be extremely important for accurate AOD simulations. 3.2.4. CHIMERE model with the nested domain configurations Though the simulations are found to improve significantly

through the refinements in the model configurations and with the new databases, the values at the domain boundaries need further improvement. This could be improved by forcing the model with more accurate boundary conditions data obtained from the oneway nesting of the CHIMERE model with 27 km (outer) and 9 km (inner) domain resolutions. The simulated AOD for the finer domain for August 6th and 23rd is shown in Fig. 9. As displayed on these figures, the spatial variability of the AOD simulations is in excellent agreement with the MODIS DB AOD (Figs. 6c3 and 7c2). The figures clearly show that the adapted configuration of the CHIMERE model effectively represents the spatial variability, as well as the magnitudes of the AOD values. 3.2.5. Analysis of the adapted model performance using AERONET observations Fig. 10 presents the simulated CHIMERE AOD500 using the aforementioned model configurations along with the AERONET measurements at the four considered stations. The modeled AOD500 values are estimated using the Angstrom relation using AODs at 400 and 600 nm. As displayed in Fig. 10, the model results are improved when MODIS erodibility and ECMWF ERA-Interim surface roughness length datasets are incorporated. The day-today variability of AOD500 is further improved with the incorporation of anthropogenic emissions. Furthermore, the nesting refined the simulations again, particularly those close to the domain boundary, such as Eilat and Sede-Boker. The statistical parameters (MBE, RMSE, R and FGE) describing the correlation between the measurements and simulations using the

S.N. Beegum et al. / Atmospheric Environment 128 (2016) 185e197

193

Fig. 9. Finer nested domain simulations of AOD from the adapted CHIMERE model on: a) August 6th and b) August 23rd.

Fig. 10. Time series of the hourly AOD500 simulations for August 2013 with different configurations of CHIMERE along with the AERONET AOD measurements at a) Eilat, b) SedeBoker, c) KAUST Campus and d) Solar-Village.

aforementioned configurations are calculated for the study period (Table 1) and show large AOD underestimation when the default configuration of the model is used. However, the agreement

improves significantly from the original configuration using the adapted CHIMERE model with new datasets on soil erodibility and surface roughness length. The observed R/RMSE range between

Table 1 The model statistical scores against AERONET observations at different locations. CHIMERE configurations

Statistic parameters KAUST campus

Default Default with high vertical resolution With MODIS erodibility MODIS erodibility and ERA-Interim z0 EGDAR-HTAP emission inventories With nesting

Solar-Village

Eilat

Sede-Boker

MBE

RMSE

R

FGE

MBE

RMSE

R

FGE

MBE

RMSE

R

FGE

MBE

RMSE

R

FGE

0.34 0.34 0.10 0.13 0.02 0.02

0.33 0.3 0.1 0.08 0.02 0.02

0.10 0.22 0.51 0.46 0.54 0.63

0.009 0.009 0.002 0.002 0.001 0.001

0.33 0.31 0.17 0.11 0.08 0.06

0.292 0.2 0.14 0.13 0.06 0.05

0.04 0.15 0.20 0.23 0.34 0.40

0.006 0.006 0.005 0.002 0.001 0.004

0.11 0.09 0.06 0.08 0.06 0.01

0.1 0.08 0.05 0.05 0.03 0.03

0.12 0.18 0.60 0.75 0.76 0.80

0.005 0.004 0.003 0.003 0.002 0.001

0.09 0.06 0.04 0.05 0.01 0.01

0.05 0.04 0.03 0.03 0.02 0.02

0.16 0.13 0.60 0.50 0.61 0.62

0.004 0.003 0.003 0.002 0.001 0.001

194

S.N. Beegum et al. / Atmospheric Environment 128 (2016) 185e197

0.23/0.03 and 0.75/0.08. When anthropogenic emission inventories are taken into account for the simulations, the error between the observations and simulations is found to reduce further. The statistical parameters also show significant improvements with R values ranging from 0.34 (Solar-Village) to 0.76 (Eilat) and RMSE ranging from 0.03 (Eilat) and 0.06 (Solar-Village). Finally, the same configuration with nested run (9 km) is found to fine-tune the simulations. The correlation-coefficient obtained with the nested runs varies from 0.40 (Solar-Village) to 0.80 (Eilat), while the RMSE ranges between 0.03 (Eilat) and 0.05 (Solar-Village). The observed high R values are comparable or even higher than those reported in
rez et al., 2011). For instance, other studies (De La Paz et al., 2013; Pe when compared with the AERONET observations, the NMMB/BSCDust AOD simulations showed high R values over different regions, such as the Subtropical Atlantic (0.7e0.76), the North-Africa and Iberian Peninsula (~0.7), Central and Eastern Mediterranean
rez et al., 2011). In addition, (0.59e0.82), and Sahel (0.65e0.71) (Pe lower R values (0.33e0.48) are obtained when simulating the transported dust from Sahara to the Iberian Peninsula using CMAQ (community multi-scale air quality) model coupled with dust emission module (De La Paz et al., 2013). 3.3. Performance evaluation of the adapted CHIMERE model The performance of the adapted CHIMERE model using MODIS erodibility, ECMWF ERA-Interim surface roughness length and EDGAR-HTAP emission inventories over the Arabian Peninsula desert are evaluated quantitatively using the AERONET measurements, various forecast models simulations, and CALIPSO vertical profiles. 3.3.1. CHIMERE versus AERONET Hourly time-series of simulated CHIMERE AOD500 along with the corresponding AERONET measurements for the two stations of Sede-Boker and Solar-Village for 2010 are presented in Fig. 11. At Solar-Village, the model is able to capture the higher AOD values during the dusty spring/summer seasons (MareAug). However the model is found to underestimate the observations during the autumn/winter seasons (SepeNov). At Sede-Boker, the day-to-day variability in AOD is more or less accurately captured in the model. The scatter plot between the CHIMERE hourly AOD and the corresponding AERONET AOD observations (both at 500 nm) is shown in Fig. 12. The data points are classified into two categories:

a) Solar-Village station

AOD(550nm)

2

AERONET CHIMERE

1.5 1 0.5 0 0

60

AOD(550nm)

2

120

180

240

b) Sede-Boker station

300

360

AERONET CHIMERE

1.5

(i) dusty periods with angstrom component, a less than 1, and (ii) non-dusty periods (anthropogenic) with a greater than or equal to 1. The basic statistical parameters (R, bias, rmse, slope and intercept) are displayed for each category for both the stations. As displayed on this figure, better agreement between the CHIMERE and AERONET values is observed for both stations during the dusty periods. The R values for both stations are consistent with those obtained for August 2013 (~0.6) during the dusty periods. The underestimation is higher (Sede-Boker: R ¼ 0.22 and slope ¼ 0.20; Solar-Village: R ¼ 0.86 and slope ¼ 0.37) at both stations during the non-dusty periods, suggesting that anthropogenic emission inventories need to be improved over the domain. 3.3.2. CHIMERE versus other forecast models A cross validation of AOD simulations between CHIMERE and other forecast models (BSC-DREAM8b_v2, MACC-ECMWF, DREAM8-MACC, NMMB/BSC-Dust, UM MetOffice and NASA/ GEOS-5), obtained from the SDS-WAS (Sand and Dust Storm Warning Advisory and Assessment System, Barcelona, Spain) website, is performed for August (2013) at AERONET stations (Fig. 13). As displayed in these figures, the adapted CHIMERE model over the domain better represents the variability of the AOD at all the sites, similar to the BSC-DREAM and MACC-ECMWF simulations. Furthermore, our simulations with anthropogenic emission are found to have good performance at all the stations, while the earlier studies indicated that the aforementioned dust forecast models underestimate the AOD during non-dusty season (Basart et al., 2009). In addition, unlike the SDS-WAS simulations, our simulations have high spatial resolution and can reliably reproduce the spatial variability of the aerosol products in local scales. 3.3.3. CHIMERE versus CALIPSO The CHIMERE vertical profile of the aerosol mass concentrations (PM10) are validated with the CALIPSO vertical profiles of aerosol extinction coefficient on August 6th and 23rd at ~10UT (Fig. 14). As displayed on these figures, the height of the dust plume and the vertical gradient is more or less accurately depicted in the simulations. The thick dust plume reaching ~5 km in the latitude region 15 Ne25 N on August 06th is clearly observed in both observations and model simulations (Fig. 14a1,14a2). The gradual decrease in the dust plume intensity beyond 25 N is also discernible from the model results. The relatively less intense dust plume with higher vertical extend is fairly clear in the simulations of August 23rd as well (Fig. 14b1,14b2). However, CALIPSO is found to underestimate the dust extinction, particularly when the plume is relatively less intense (Fig. 14b1). This could arise from either the CHIMERE model uncertainties or from the uncertainties in the retrieval of extinction profiles as demonstrated by Tesche et al. (2013), Zong et al. (2015). In addition, it was demonstrated that CALIPSO misclassifies mineral dust in the upper troposphere as thin cirrus clouds and thereby reducing dust extinction values (Yu et al., 2010). Nevertheless, the present study shows moderately close agreement in the vertical structure with the observations, which indicates that the horizontal and vertical advection schemes used in the model are realistic over the Arabian Peninsula desert. 4. Conclusions

1 0.5 0 0

60

120

180

240

Days number (2010)

300

360

Fig. 11. Time-series of hourly AOD500 for the year 2010 from the CHIMERE simulations and AERONET observations at the stations of (a) Sede-Boker and (b) Solar-Village.

The configuration of a three dimensional chemistry transport model CHIMERE has been adapted over the Arabian Peninsula. For this purpose, CHIMERE was run with an offline coupling with WRF model in nested configuration at spatial resolutions of 27 km and 9 km for outer and inner domains, respectively. CHIMERE model simulations were carried out using a variety of configurations: (i) default configuration with the built-in soil and surface properties at

S.N. Beegum et al. / Atmospheric Environment 128 (2016) 185e197

195

Fig. 12. Scatter plot between the hourly AOD500 from the CHIMERE simulations and AERONET observations for the year 2010 at the stations of Sede-Boker and Solar-Village for both dusty and non-dusty periods.

Fig. 13. Model inter-comparison with different operational forecast models under SDS-WAS for August 2013.

lower (500 hPa) and higher (200 hPa) vertical levels (ii) with MODIS soil erodibility and default ERS scatterometer derived surface roughness length (iii) with MODIS soil erodibility and ERAInterim surface roughness length (iv) keeping the latest configuration with EDGAR-HTAP anthropogenic emission inventories, and (v) with nested runs with finer 9 km resolution keeping the final configuration. Finally, the regionally adapted model was used to simulate the AOD for a period of a year. The CHIMERE AOD simulations with the all considered configurations were validated both qualitatively and qualitatively based on AERONET measurements, MODIS DB AOD and CALIPSO profiles. The main findings are summarized below:
CHIMERE simulations with USGS soil and surface parameters yielded higher errors, when compared with in-situ and satellite observations.


Increasing the number of vertical layers reduced the errors as model was able to account for the accurate representation of the long-range transport.
CHIMERE simulations were observed to be highly sensitive to the soil and surface datasets such as soil erodibility and surface roughness length. With the inclusion of MODIS soil erodibility, the CHIMERE simulations were improved significantly, both in the absolute magnitudes and in the spatial pattern of AOD. The results were also improved by using the ECMWF ERA-Interim surface roughness length.
Simulations with the global EDGAR-HTAP anthropogenic emission inventories brought the AOD simulations close to the observations.
Vertical structure of the simulated aerosols was in good agreement with the CALIPSO profiles.
CHIMERE simulations at finer spatial resolution of 9 km with nested domain configuration resulted in accurate and realistic

196

S.N. Beegum et al. / Atmospheric Environment 128 (2016) 185e197

Fig. 14. Vertical Profiles of Aerosol extinction coefficient (km1) from the CALIPSO (left panel) and the profiles of CHIMERE simulated PM10 mass concentrations at 10UT: a) August 06th and b) August 23rd.

simulations of AODs, when compared with AERONET measurements, with correlation coefficients of 0.4e0.8 at the considered stations.
Performance evaluation of the adapted CHIMERE model with the ARONET measurements for a period of 1 year showed good agreement with correlation coefficient of ~0.6 with higher values of slope and correlation coefficients during dusty periods.

vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi u n uX RMSE ¼ t ðAODSi  AODOi Þ2

(2)

i¼1

However, the usefulness of RMSE is questionable and the interpretation becomes more difficult when prominent outliers occur.
The Correlation coefficient (R),

Pn 

Acknowledgments

   AODOi  AODO  AODSi  AODS R ¼ rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 2 P  2 Pn   ni¼1 AODSi  AODS i¼1 AODOi  AODO i¼1

We acknowledge the principal investigators of the AERONET stations Solar-Village, KAUST Campus, Sede-Boker and Eilat for proving AOD data (http://aeronet.gsfc.nasa.gov). We also thank the team of the WMO SDS-WAS for the compared dust product plots (http://sds-was.aemet.es/forecast-products/dust-forecasts). Acknowledgements are also due for NASA-EOS team members for providing CALIOP/CALIPSO datasets. We thank NASA's GiovanniInteractive Visualization and Analysis team for providing MODIS deep blue data and ECMWF ERA-Interim for global reanalysis data. The support and resources from the High Performance Computing Cluster at Masdar Institute are gratefully acknowledged.

where AODOi and AODSi represent the average values of the time series for both observations and model simulations, respectively. The correlation coefficient indicates how efficient is the model to catch up the temporal fluctuations with respect to the observations.
Fractional Gross Error (FGE), which is a measure of model error is,

FGE ¼ Appendix 1. Model statistics
The Mean Bias Error (MBE), which represents the average deviations between the two datasets,

(3)

 i¼n   2X ðAODSi  AODOi Þ  n i¼1 ðAODSi þ AODOi Þ

(4)

The score of FGE varies between 0 and 2 and behaves symmetrically with respect to underestimation and overestimation, without over emphasizing outliers. References

n 1X MBE ¼ ðAODSi  AODOi Þ n

(1)

i¼1

The ideal value of this parameter is zero, while negative and positive values indicate respectively underestimation and overestimation.
Root mean square error (RMSE), combines the spread of individual errors and is strongly dominated by the largest values due to the squaring operation,

Alfaro, S.C., Gomes, L., 2001. Modeling mineral aerosol production by wind erosion: emission intensities and aerosol size distributions in source areas. J. Geophys. Res. 106, 18075e18084. Al-Shehhi, M., Gherboudj, I., Ghedira, H., 2014. An overview of historical harmful algae blooms outbreaks in the Arabian Seas. Mar. Pollut. Bull. 86, 314e324. Astitha, M., Lelieveld, J., Abdel Kader, M., Pozzer, A., de Meij, A., 2012. Parameterization of dust emissions in the global atmospheric chemistry-climate model EMAC: impact of nudging and soil properties. Atmos. Chem. Phys. 12, 11057e11083.
rez, C., Cuevas, E., Baldasano, J.M., Gobbi, G.P., 2009. Aerosol characBasart, S., Pe terization 20 in northern Africa, northeastern Atlantic, Mediterranean Basin and Middle East from directsun aeronet observations. Atmos. Chem. Phys. 9,

S.N. Beegum et al. / Atmospheric Environment 128 (2016) 185e197 8265e8282. Bessagnet, B., Menut, L., Aymoz, G., Chepfer, H., Vautard, R., 2008. Modeling dust emissions and transport within Europe: the Ukraine March 2007 event. J. Geophys. Res. Atmos. 113, 1e13. Briant, R., Menut, L., Siour, G., Prigent, C., 2014. Homogenized modeling of mineral dust emissions over Europe and Africa using the CHIMERE model. Geosci. Model Dev. Discuss. 7, 3441e3480. Chimere documentation, 2014. Documentation of the Chemistry-transport Model CHIMERE. Chin, M., Rood, R.B., Lin, S.-J., Muller, J.-F., Thompson, A.M., 2000. Atmospheric sulfur cycle simulated in the global model GOCART: model description and global properties. J. Geophys. Res. 105, 671e687. Colarco, P., Da Silva, A., Chin, M., Diehl, T., 2010. Online simulations of global aerosol distributions in the NASA GEOS-4 model and comparisons to satellite and ground-based aerosol optical depth. J. Geophys. Res. Atmos. 115.
s, J., Pe
rez, J., De La Paz, D., Vedrenne, M., Borge, R., Lumbreras, J., Manuel de Andre Rodríguez, E., Karanasiou, A., Moreno, T., Boldo, E., Linares, C., 2013. Modelling Saharan dust transport into the Mediterranean basin with CMAQ. Atmos. Environ. 70, 337e350. De-Meij, A., Thunis, P., Bessagnet, B., Cuvelier, C., 2009. The sensitivity of the CHIMERE model to emissions reduction scenarios on air quality in northern Italy. Atmos. Environ. 43, 1897e1907. DeMott, P.J., 2003. African dust aerosols as atmospheric ice nuclei. Geophys. Res. Lett. 30, 26e29. Derimian, Y., Karnieli, A., Kaufman, Y.J., Andreae, M.O., Andreae, T.W., Dubovik, O., Maenhaut, W., Koren, I., Holben, B.N., 2006. Dust and pollution aerosols over the Negev desert, Israel: properties, transport, and radiative effect. J. Geophys. Res. Atmos. 111, 1e14. Foret, G., Bergametti, G., Dulac, F., Menut, L., 2006. An optimized particle size bin scheme for modeling mineral dust aerosols. J. Geophys. Res. D17310. Ghanem, S.M., 2001. Industrialization in the UAE. In: Abed, I., Hellyer, P. (Eds.), United Arab Emirates: New Perspective, pp. 260e276. Gherboudj, I., Beegum, S.N., Marticorena, B., Ghedira, H., 2015. Dust emission parameterization scheme over the MENA region: sensitivity analysis to soil moisture and soil texture. J. Geophys. Res. Atmos. 120 http://dx.doi.org/10.1002/ 2015 JD023338. Gherboudj, I., Ghedira, H., 2014. Spatiotemporal assessment of dust loading over the United Arab Emirates. Int. J. Climatol. 34, 3321e3335. Giannadaki, D., Pozzer, A., Lelieveld, J., 2014. Modeled global effects of airborne desert dust on air quality and premature mortality. Atmos. Chem. Phys. 14, 957e968. Ginoux, P., Chin, M., Tegen, I., Prospero, J.M., Holben, B., Dubovik, O., LIn, S.-J., 2001. Sources and distributions of dust aerosols simulated with the GOCART model. J. Geophys. Res. 106, 20255e20273. Grini, A., Myhre, G., Zender, C., Isaksen, I.S.A., 2005. Model simulations of dust sources and transport in the global atmosphere: effects of soil erodibility and wind speed variability. J. Geophys. Res. 110, D02205. Guenther, A., Karl, T., Harley, P., Wiedinmyer, C., Palmer, P.I., Geron, C., 2006. Estimates of global terrestrial isoprene emissions using MEGAN (Model of emissions of gases and aerosols from nature). Atmos. Chem. Phys. Discuss. 6, 107e173. Holben, B., Eck, T., Slutsker, I., Smirnov, A., Sinyuk, A., Schafer, J., Giles, D., Dubovik, O., 2006. AERONET's version 2.0 quality assurance criteria. Proc. SPIE 6408, Remote Sens. Atmos. Clouds, 64080Q, http://dx.doi.org/10.1117/12. 706524. Hsu, N.C., Jeong, M.J., Bettenhausen, C., Sayer, A.M., Hansell, R., Seftor, C.S., Huang, J., Tsay, S.C., 2013. Enhanced deep blue aerosol retrieval algorithm: the second generation. J. Geophys. Res. Atmos. 118, 9296e9315. IPCC, 2013. Climate Change 2013: the Physical Science Basis. Contribution of Working Group I to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change [Stocker, T.F., D. Qin, G.-K. Plattner, M. Tignor, S.K. Allen, J. Boschung, A. Nauels, Y. Xia]. Janssens-Maenhout, G., Dentener, F., Aardenne, J., Van Monni, S., Pagliari, V., Orlandini, L., Klimont, Z., Kurokawa, J., Akimoto, H., Ohara, T., Wankmüller, R., Battye, B., Grano, D., Zuber, A., Keating, T., 2012. EDGAR-HTAP: a Harmonized Gridded Air Pollution Emission Dataset Based on National Inventories. Office, Ispra (Italy). Jenson, S.K., Domingue, J.O., 1988. Extracting topographic structure from digital elevation data for geographic information system analysis. Engineering 54, 1593e1600. King, M., Closs, J., Spangler, S., Greenstone, R., Wharton, S., Myers, M., 2004. EOS Data Products Handbook. EOS Projec. ed. Kohfeld, K., Tegen, I., 2007. Record of Mineral Aerosols and Their Role in the Earth System. in: Elsevier, 1e26 (Ed.), Treatise on Geochemistry. €ckel, P., Pozzer, A., Hadjinicolaou, P., Cammas, J.-P., Beirle, S., Lelieveld, J., Hoor, P., Jo 2009. Severe ozone air pollution in the Persian Gulf region. Atmos. Chem. Phys. 9, 1393e1406. Liu, Z., Omar, A., Vaughan, M., Hair, J., Kittaka, C., Hu, Y., Powell, K., Trepte, C., Winker, D., Hostetler, C., Ferrare, R., Pierce, R., 2008. CALIPSO lidar observations

197

of the optical properties of Saharan dust: a case study of long-range transport. J. Geophys. Res. Atmos. 113, 1e20. Lu, S., Huang, H., Hou, Y., da Silva, A., Chin, M., Moorthi, S., Wang, J., Juang, H.H., Iredell, M., McQueen, J., Diehl, T.L., 2010. Development of Global Aerosol Forecasting System at NCEP. American Geophysical Union, Fall Meeting. Maenhaut, W., Karnieli, A., Andreae, M.O., 2014. Ten-year study of fine aerosol at Sde Boker, Israel, using PIXE: time trends, seasonal variation, correlations, and source areas for anthropogenic elements. Nucl. Instrum. Methods Phys. Res. Sect. B Beam Interact. Mater. Atoms 318, 119e124. Marticorena, B., Chazetteb, P., Bergamettia, G., Dulacb, F., Legrandc, M., 2004. Mapping the aerodynamic roughness length of desert surfaces from the POLDER/ADEOS bi-directional reflectance product. Int. J. Remote Sens. 25, 603e626.
rez, C., Haustein, K., Bessagnet, B., Prigent, C., Alfaro, S., 2013. Impact of Menut, L., Pe surface roughness and soil texture on mineral dust emission fluxes modeling. J. Geophys. Res. Atmos. 118, 6505e6520. Menut, L., Schmechtig, C., Marticorena, B., 2005. Sensitivity of the sandblasting fluxes calculations to the soil size distribution accuracy. J. Atmos. Ocean. Technol. 22, 1875e1884. Mokhtari, M., Gomes, L., Tulet, P., Rezoug, T., 2012. Importance of the surface size distribution of erodible material: an improvement on the dust entrainment and deposition (DEAD) model. Geosci. Model Dev. 5, 581e598. Monahan, E.C., Spiel, D.E., Davidson, K.L., 1986. A model of marine aerosol generation via whitecaps and wave disruption. Oceanogr. Sci. Libr. 2, 167e174. Morcrette, J.-J., Boucher, O., Jones, L., Salmond, D., Bechtold, P., Beljaars, A., Benedetti, A., Bonet, A., Kaiser, J.W., Razinger, M., Schulz, M., Serrar, S., Simmons, A.J., Sofiev, M., Suttie, M., Tompkins, A.M., Untch, A., 2009. Aerosol analysis and forecast in the European centre for medium-range weather forecasts integrated forecast system: forward modeling. J. Geophys. Res. Atmos. 114, D06206.
rez, C., Haustein, K., Janjic, Z., Jorba, O., Huneeus, N., Baldasano, J.M., Black, T., Pe Basart, S., Nickovic, S., Miller, R.L., Perlwitz, J.P., Schulz, M., Thomson, M., 2011. Atmospheric dust modeling from meso to global scales with the online NMMB/ BSC-dust model e part 1: model description, annual simulations and evaluation. Atmos. Chem. Phys. 11, 13001e13027. €
rez, C., Nickovic, S., Pejanovic, G., Baldasano, J.M., Ozsoy, Pe E., 2006. Interactive dust-radiation modeling: a step to improve weather forecasts. J. Geophys. Res. Atmos. 111. Prigent, C., Tegen, I., Aires, F., Marticorena, B., Zribi, M., 2005. Estimation of the aerodynamic roughness length in arid and semi-arid regions over the globe with the ERS scatterometer. J. Geophys. Res. D Atmos. 110, 1e12. Prospero, J., 1999. Long-term measurements of the transport of African mineral dust to the southeastern United States: implications for regional air quality. J. Geophys. Res. 104, 15,917e15,927. Remer, L.a., Mattoo, S., Levy, R.C., Munchak, L.A., 2013. MODIS 3 km aerosol product: algorithm and global perspective. Atmos. Meas. Tech. 6, 1829e1844. Seinfeld, J.H., Pandis, S.N., 1998. Atmospheric Chemistry and Physics: from Air Pollution to Climate Change, second ed. John Wiley & Sons Ltd, UK, New York. Shao, Y., 2008. Physics and Modelling of Wind Erosion. Springer. Shao, Y., Ishizuka, M., Mikami, M., Leys, J.F., 2011. Parameterization of size-resolved dust emission and validation with measurements. J. Geophys. Res. Atmos. 116, 1e19. Shi, Y., Zhang, J., Reid, J.S., Hyer, E.J., Hsu, N.C., 2013. Critical evaluation of the MODIS deep blue aerosol optical depth product for data assimilation over North Africa. Atmos. Meas. Tech. 6, 949e969. Simmons, A.J., Uppala, S.M., Dee, D.P., Kobayashi, S., 2007. ERA-Interim: new ECMWF reanalysis products from 1989 onwards. ECMWF Newsl. 110, 29e35. Srivastava, N., Satheesh, S.K., Blond, N., 2014. Sensitivity of meteorological input and soil properties in simulating aerosols (dust, PM10, and BC) using CHIMERE chemistry transport model. J. Earth Syst. Sci. 123, 1249e1264. Tesche, M., Wandinger, U., Ansmann, A., Althausen, D., Müller, D., Omar, A.H., 2013. Ground-based validation of CALIPSO observations of dust and smoke in the Cape Verde region. J. Geophys. Res. Atmos. 118, 2889e2902. Wang, W., Bruyere, C., Duda, M., Dudhia, J., Gill, D., Kavulich, M., Keene, K., Lin, H., Michalakes, J., Rizvi, S., Zhang, X., 2014. User's Guide for Advanced Research WRF (ARW) Modeling System Version 3. Winker, D.M., Pelon, J., Coakley, J.A., Ackerman, S.A., Charlson, R.J., Colarco, P.R., Flamant, P., Fu, Q., Hoff, R.M., Kittaka, C., Kubar, T.L., Le Treut, H.,
gie, G., Poole, L., Powell, K., Trepte, C., Vaughan, M.A., McCormick, M.P., Me Wielicki, B.A., 2010. The CALIPSO Mission: a global 3D view of aerosols and clouds. Bull. Am. Meteorol. Soc. 91, 1211e1229. Woodward, S., 2001. Modeling the atmospheric life cycle and radiative impact of mineral dust in the Hadley Centre climate model. J. Geophys. Res. 106, 18155. Yu, H., Chin, M., Winker, D.M., Omar, A.H., Liu, Z., Kittaka, C., Diehl, T., 2010. Global view of aerosol vertical distributions from CALIPSO lidar measurements and GOCART simulations: regional and seasonal variations. J. Geophys. Res. 115, D00H30. Zong, X., Xia, X., Che, H., 2015. Validation of aerosol optical depth and climatology of aerosol vertical distribution in the Taklimakan Desert. Atmos. Pollut. Res. 6, 239e244.