**5. Results**

As described previously, Limassol was the main site for the development of the AIRSPACE methodology for the estimation of the PM levels. The ground based data were used to validate the satellite data. Complementary to the Limassol site, Nicosia's and Larnaca's site observa‐ tions were used to validate the performance of the models. In this section, the major results from the AIRSPACE project are analysed in some detail.

#### **5.1. Dataset validation**

In the AIRSPACE project, both ground based and satellite observations were used to provide aerosol related information for South Eastern Mediterranean region. The first goal of the AIRSPACE project was the validation of the satellite observations in Cyprus, an area affected by aerosol from a variety of sources and surrounded by sea. The ground based observations performed over Limassol and Nicosia were used as the main sites for the validation of the satellite observations.

To incorporate both the spatial and temporal variability of aerosol distribution, the MODIS retrievals at 10 km x 10 km resolution and the AERONET direct Sun measurements at 15 minute intervals (Holben et al., 1998) need to be co-located in space and time.

The AERONET data provide the ground truth for the MODIS validation. The global CUT-TEPAK ground-based AERONET sunphotometer measures aerosol optical thickness in eight channels (340 to 1640 nm). The instrument takes measurements every 15 minutes. From the observations taken within ±30 minutes of MODIS overpass time (Ichoku et al., 2002), mean values of the optical parameters were calculated. Therefore, the maximum number of AERONET observations within the hour of an overpass is 5. Fewer observations within the hour indicate data have been removed by the AERONET Run-Time Cloud Checking procedure.

The study required at least 2 out of possible 5 AERONET measurements to be within ±30 min of MODIS overpasses and at least 5 out of possible 25 MODIS retrievals to be within a 25 km radius centred over the AERONET site. The mean values of the collocated spatial and temporal ensemble were then used in a linear regression analysis and in calculating RMS errors. The AERONET level 1.5 data were cloud screened. Though the level 2.0 data provide final calibration, they are not available for the entire time period of the project. Therefore, the level 1.5 data (instead of level 2.0) were used in the operational MODIS aerosol validation scheme.

A total of 352 points of AERONET site representing the correlated criteria for the MODIS- and AERONET derived AOT were collected in the period from April 2010 to December 2012.

Figure 10 features the correlation of the MODIS AQUA and TERRA sensors and CUT\_TEPAK AERONET measurements. The slope of linear regression in the correlation plot between MODIS and AERONET provides an overview of possible differences. The correlation coeffi‐ cient value of the order of 0.62 for both TERRA and AQUA satellites is due to the coast line of the Limassol site. Limassol's CUT-TEPAK AERONET site is a coastal area, thus the surface inhomogeneity or sub-pixel water contamination has a larger effect than anticipated in continental coastal regions (Nisantzi et al., 2012). The systematic biases overestimations in MODIS retrievals are mainly due to aerosol model assumptions (deviation of 0–20%) andin‐ strument calibration (2–5%).

Thermal Optical Transmitance (TOT) to measure EC-OC particle concentration, gravimetric mass determination and X-Ray fluorescence to determine trace elemental composition of PM2.5 and PM10. Samples up to 19 June 2012 have been completely analyzed. The remaining

As described previously, Limassol was the main site for the development of the AIRSPACE methodology for the estimation of the PM levels. The ground based data were used to validate the satellite data. Complementary to the Limassol site, Nicosia's and Larnaca's site observa‐ tions were used to validate the performance of the models. In this section, the major results

In the AIRSPACE project, both ground based and satellite observations were used to provide aerosol related information for South Eastern Mediterranean region. The first goal of the AIRSPACE project was the validation of the satellite observations in Cyprus, an area affected by aerosol from a variety of sources and surrounded by sea. The ground based observations performed over Limassol and Nicosia were used as the main sites for the validation of the

To incorporate both the spatial and temporal variability of aerosol distribution, the MODIS retrievals at 10 km x 10 km resolution and the AERONET direct Sun measurements at 15-

The AERONET data provide the ground truth for the MODIS validation. The global CUT-TEPAK ground-based AERONET sunphotometer measures aerosol optical thickness in eight channels (340 to 1640 nm). The instrument takes measurements every 15 minutes. From the observations taken within ±30 minutes of MODIS overpass time (Ichoku et al., 2002), mean values of the optical parameters were calculated. Therefore, the maximum number of AERONET observations within the hour of an overpass is 5. Fewer observations within the hour indicate data have been removed by the AERONET Run-Time Cloud Checking

The study required at least 2 out of possible 5 AERONET measurements to be within ±30 min of MODIS overpasses and at least 5 out of possible 25 MODIS retrievals to be within a 25 km radius centred over the AERONET site. The mean values of the collocated spatial and temporal ensemble were then used in a linear regression analysis and in calculating RMS errors. The AERONET level 1.5 data were cloud screened. Though the level 2.0 data provide final calibration, they are not available for the entire time period of the project. Therefore, the level 1.5 data (instead of level 2.0) were used in the operational MODIS aerosol validation scheme. A total of 352 points of AERONET site representing the correlated criteria for the MODIS- and AERONET derived AOT were collected in the period from April 2010 to December 2012.

minute intervals (Holben et al., 1998) need to be co-located in space and time.

samples have undergone chemical process for analysis.

194 Remote Sensing of Environment: Integrated Approaches

from the AIRSPACE project are analysed in some detail.

**5. Results**

**5.1. Dataset validation**

satellite observations.

procedure.

**Figure 10.** Comparisons of MODIS and AERONET derived at 0.50 nm wavelength, encompassing 352 points from CUT-TEPAK AERONET coastal site. The solid line represents the slopes of linear regression both for AQUA and TERRA MODIS sensors

Using the MICROTOPS II AOT, the procedure was duplicated for the validation of the satellite observations in Nicosia. The number of collocated and synchronized ground based and satellite measurements were statistically low in order to provide correlation factor which can represent a reliable validation study.

### **5.2. Satellite climatology**

In the present work, the Level 2, 10x10km, MOD04 aerosol products (Collection 051) were retrieved for the years 2001 to 2011 from NASA's Level 1 and Atmosphere Archive and Distribution System (LAADS). The AOT fields were extracted from the 'Opti‐ cal\_Depth\_Land\_And\_Ocean' parameter which provides the AOT at 550nm derived via the dark-target algorithms and with best quality data (Remer et al., 2005). According to Remer et al. (2009), the AOT fields for this product have been respectively validated to within the error bounds of (0.04+0.05AOT) and ±(0.05+0.15AOT) at 550nm.

intermediate values. The two distinct maxima associated with dust transport phenomena are observed at all sites in May and August. The value for the first peak in May is approximately the same for all urban sites (~0.40) but for August, the levels for Nicosia are higher (~0.45)

Air Pollution from Space http://dx.doi.org/10.5772/39310 197

**Figure 12.** Average monthly AOT. (LE, LA, LM, and AM mark the sites of Nicosia, Larnaca, Limassol and Ag. Marina)

One element of the AIRSPACE program in Cyprus was the measurement of ground level PM

**5.3. PM surface analysis**

concentrations by Harvard Impactors.

compared to the other two urban sites (~0.35 for Larnaca and Limassol).

Based on the above AOT data, subsets for the area of Cyprus were extracted and mean monthly climatology maps were constructed for the period 2001-2011. For the area considered, the number of days with valid TERRA AOT measurements ranged approximately from 1000 to 2300 (which amount to 25%-57% time coverage), as shown in Figure 11. The highest number of valid measurements was observed over the central area of Cyprus (in the vicinity of Troodos Mountain), whereas near the coastline, this number decreased.

**Figure 11.** Number of valid TERA AOT observations for the period 2001-2011

The maps for each month are presented in Figure 12. The seasonal cycle of the aerosol load is well depicted. Minima are observed during winter months and maxima during spring and summer when intense phenomena associated with dust transport from Sahara desert are more frequent. The respective monthly average values for the three urban sites of Nicosia, Larnaca, and Limassol (marked as LE, LA and LM, respectively, on the maps) and the background site of Agia Marina, (marked as AM) have been calculated. In general, the background site is characterised by lower aerosol loads (ranging from 0.1 to 0.28) than those observed at the urban sites. Limassol (the main port city) presents the highest values for the period January-May and Nicosia (the capital city) from June to December. For this latter period, Larnaca presents intermediate values. The two distinct maxima associated with dust transport phenomena are observed at all sites in May and August. The value for the first peak in May is approximately the same for all urban sites (~0.40) but for August, the levels for Nicosia are higher (~0.45) compared to the other two urban sites (~0.35 for Larnaca and Limassol).

**Figure 12.** Average monthly AOT. (LE, LA, LM, and AM mark the sites of Nicosia, Larnaca, Limassol and Ag. Marina)

#### **5.3. PM surface analysis**

**5.2. Satellite climatology**

196 Remote Sensing of Environment: Integrated Approaches

bounds of (0.04+0.05AOT) and ±(0.05+0.15AOT) at 550nm.

Mountain), whereas near the coastline, this number decreased.

**Figure 11.** Number of valid TERA AOT observations for the period 2001-2011

In the present work, the Level 2, 10x10km, MOD04 aerosol products (Collection 051) were retrieved for the years 2001 to 2011 from NASA's Level 1 and Atmosphere Archive and Distribution System (LAADS). The AOT fields were extracted from the 'Opti‐ cal\_Depth\_Land\_And\_Ocean' parameter which provides the AOT at 550nm derived via the dark-target algorithms and with best quality data (Remer et al., 2005). According to Remer et al. (2009), the AOT fields for this product have been respectively validated to within the error

Based on the above AOT data, subsets for the area of Cyprus were extracted and mean monthly climatology maps were constructed for the period 2001-2011. For the area considered, the number of days with valid TERRA AOT measurements ranged approximately from 1000 to 2300 (which amount to 25%-57% time coverage), as shown in Figure 11. The highest number of valid measurements was observed over the central area of Cyprus (in the vicinity of Troodos

The maps for each month are presented in Figure 12. The seasonal cycle of the aerosol load is well depicted. Minima are observed during winter months and maxima during spring and summer when intense phenomena associated with dust transport from Sahara desert are more frequent. The respective monthly average values for the three urban sites of Nicosia, Larnaca, and Limassol (marked as LE, LA and LM, respectively, on the maps) and the background site of Agia Marina, (marked as AM) have been calculated. In general, the background site is characterised by lower aerosol loads (ranging from 0.1 to 0.28) than those observed at the urban sites. Limassol (the main port city) presents the highest values for the period January-May and Nicosia (the capital city) from June to December. For this latter period, Larnaca presents

One element of the AIRSPACE program in Cyprus was the measurement of ground level PM concentrations by Harvard Impactors.

Statistics for the Limassol site show that for the first six months of observations the mean value for PM10 is 32.1 μg/m3 , for PM2.5 13.4 μg/m3 and for total carbon 2.3 μg/m3 with standard deviations of 20.9, 4.6 and 1.1 μg/m3 , respectively.

PM10 and PM2.5 were analysed, for trace elements such as sulfur, magnesium, aluminum, sodium, silicon, chlorine, potassium and calcium. Statistics for some of those trace elements for PM2.5 are shown below, in Table 1.


**Table 1.** Trace elements statistics for the first 6 months sampling, for PM2.5

Figure 13 and 14 indicate the 6 month time series of the PM2.5 and PM10 concentrations, as well as the elemental, organic and total carbon levels from the Limassol filters.

**Figure 13.** Time series of PM10 and PM2.5 at the Limassol site for the first 6 months' samples

Air Pollution from Space http://dx.doi.org/10.5772/39310 199

**Figure 14.** Time series of organic (OC), elemental (EC), and total carbon (TC) at the Limassol site for the first 6 months'

samples.

Analysis of these initial samples revealed evidence of a dust storm event recorded on 12 March 2012, with PM10 and PM2.5 concentrations reaching up to 156.6 μg/m3 and 29.4 μg/m3 , respec‐ tively. These values are several times higher than the typical values shown during the sampling period and well above the 24-hour limit value set by EEA, especially for PM10.

PM10 and PM2.5 concentrations show a small increase from the start of the sampling (Janu‐ ary 2012) until June 2012, indicating a temporal relationship.

#### **5.4. Statistical model**

Based on the data collected a statistical model was established for estimation of PM concen‐ trations from AOT measurements. Using a general linear regression model, the AOT retrieved by MODIS was used to predict ground-level PM10 concentrations in Limassol, Cyprus.

The proposed model by Liu et al. (2007) is given in equation 1:

*Ln*(*PM*10) = *β* <sup>0</sup> <sup>+</sup> *<sup>β</sup>* 1(log*AOT* ) + *<sup>β</sup>* 2(log*AE*) + *<sup>β</sup>* 3 (*WVdep*) + *β* 4(ln(*<sup>T</sup>* )) + *<sup>β</sup>* 5(ln(*RH* )) + *<sup>β</sup>* 6(ln(*WS*)) + *<sup>β</sup>* 7(*Wd*) + *<sup>β</sup>* 8(*<sup>P</sup>*) + *<sup>β</sup>* 9(*PBL* ) (1)

**Figure 13.** Time series of PM10 and PM2.5 at the Limassol site for the first 6 months' samples

Statistics for the Limassol site show that for the first six months of observations the mean value

PM10 and PM2.5 were analysed, for trace elements such as sulfur, magnesium, aluminum, sodium, silicon, chlorine, potassium and calcium. Statistics for some of those trace elements

Sodium 0.27 0.15 0.23 Magnesium 0.06 0.08 0.05 Aluminum 0.16 0.29 0.07 Silicon 0.29 0.55 0.13 Sulfur 1.29 0.73 0.99 Chlorine 0.07 0.17 0.02 Potassium 0.11 0.06 0.10 Calcium 0.20 0.37 0.12

Figure 13 and 14 indicate the 6 month time series of the PM2.5 and PM10 concentrations, as well

Analysis of these initial samples revealed evidence of a dust storm event recorded on 12 March

tively. These values are several times higher than the typical values shown during the sampling

PM10 and PM2.5 concentrations show a small increase from the start of the sampling (Janu‐

Based on the data collected a statistical model was established for estimation of PM concen‐ trations from AOT measurements. Using a general linear regression model, the AOT retrieved by MODIS was used to predict ground-level PM10 concentrations in Limassol, Cyprus.

5(ln(*RH* )) + *<sup>β</sup>*

6(ln(*WS*)) + *<sup>β</sup>*

7 (*Wd*) + *β* 8 (*P*) + *β*

, respectively.

and for total carbon 2.3 μg/m3

**Mean (μg/m3) SD (μg/m3) Median (μg/m3)**

with standard

and 29.4 μg/m3

, respec‐

9(*PBL* ) (1)

, for PM2.5 13.4 μg/m3

**Table 1.** Trace elements statistics for the first 6 months sampling, for PM2.5

as the elemental, organic and total carbon levels from the Limassol filters.

period and well above the 24-hour limit value set by EEA, especially for PM10.

4(ln(*<sup>T</sup>* )) + *<sup>β</sup>*

2012, with PM10 and PM2.5 concentrations reaching up to 156.6 μg/m3

ary 2012) until June 2012, indicating a temporal relationship.

The proposed model by Liu et al. (2007) is given in equation 1:

3(*WVdep*) + *<sup>β</sup>*

**5.4. Statistical model**

*Ln*(*PM*10) =

(log*AOT* ) + *β*

2(log*AE*) + *<sup>β</sup>*

*β* <sup>0</sup> <sup>+</sup> *<sup>β</sup>* 1

for PM10 is 32.1 μg/m3

deviations of 20.9, 4.6 and 1.1 μg/m3

198 Remote Sensing of Environment: Integrated Approaches

for PM2.5 are shown below, in Table 1.

**Figure 14.** Time series of organic (OC), elemental (EC), and total carbon (TC) at the Limassol site for the first 6 months' samples.

Where βi are the regression coefficients, AOT is the Aerosol Optical Thickness, AE is the Ångström Exponent, WV is the Water Vapour, T is the surface temperature, WS is the wind speed, Wd the wind direction, P is the pressure at surface level and PBL is the Planetary boundary layer height.


The available data set in AIRSPACE project are given in Table 3:

**Table 2.** AIRSPACE dataset used for the statistical model

Based on the proposed methodology, the performance of the multi-regression model was examined by introducing one predictor (Xi) at a time, together with the initial predictor, the AOT at 500nm (Xi i=0). For each predictor Xi, four transformations (j) were considered

**CC= 0.853 CC=0.850**

**f 1 =log(aod)** 

**f 2 =f1**

**f 3 =f2**

**f 4 =f3**

**f 5 =f4 +Humidity** 

**f 6 =f5**

**f 7 =f6 +WindDir** 

**f 8 =f7 +Presure** 

**f 9 =f8 +log(PBL)** 

**+log(Angstrom)** 

**+log(Temp)** 

**+log(Wind)** 

**+Wavapour-mean(Wavapour)**

Air Pollution from Space http://dx.doi.org/10.5772/39310 201

Constant Term -4.117 -4.111 Ln(AOT) 0.952 0.943 Ln(AE) 0.299 0.299 Wavapour-mean(Wavapour) -0.393 -0.384 Ln(Temp) 0.643 0.6.23 Humidity 0.008 0.008 Ln(Wind) -0.096 -0.071 WindDir -0.0004

**0.7 0.8 0.9 1**

**Correcation Coefficient (PM10 - Predicted PM10)**

**f1 f2 f3**

**f4 f5 f6**

**Function & Predictors**

**f7 f8 f9**

The results of the above sensitivity analysis indicate the maximum performance of the model

**Figure 15.** The correlation coefficient between the predicted and measured PM10 is presented for 8 different models

<sup>−</sup> 4.11 <sup>+</sup> 0.952(log*AOT* ) <sup>+</sup> 0.299(log*AE*) <sup>−</sup> 0.393(*WVdep*) <sup>+</sup> 0.643(ln(*<sup>T</sup>* )) <sup>+</sup> 0.008(*RH* ) <sup>−</sup> 0.096(ln(*WS*)) (2)

**Table 4.** Best correlation coefficients and regression coefficients

of the order of CC=0.85 as shown in equation 2:

*Ln*(*PM*10) <sup>=</sup>


#### **Table 3.**

From the above options (j=1 to 4), the one with the highest correlation coefficient (CCij) between predicted and measured PM10 was selected. In each iteration step k, the maximum values of the CCij = CCik were compared, in order to select the predictor Xik with the highest positive impact. Due to the limited dataset, no evident seasonal dependence was noted (Cook and Sanford, 1982).

The results are presented below. In Figure 15 the correlation coefficient between the predicted and measured PM10 is presented for 8 different models. The maximum performance of the model is reached by using the following predictors (in strength order), with a correlation coefficient on the order of CC=0.85


**Table 4.** Best correlation coefficients and regression coefficients

Where βi are the regression coefficients, AOT is the Aerosol Optical Thickness, AE is the Ångström Exponent, WV is the Water Vapour, T is the surface temperature, WS is the wind speed, Wd the wind direction, P is the pressure at surface level and PBL is the Planetary

(Akrotiri Air Base, Cyprus)

Based on the proposed methodology, the performance of the multi-regression model was examined by introducing one predictor (Xi) at a time, together with the initial predictor, the AOT at 500nm (Xi i=0). For each predictor Xi, four transformations (j) were considered

From the above options (j=1 to 4), the one with the highest correlation coefficient (CCij) between predicted and measured PM10 was selected. In each iteration step k, the maximum values of the CCij = CCik were compared, in order to select the predictor Xik with the highest positive impact. Due to the limited dataset, no evident seasonal dependence was noted (Cook and

The results are presented below. In Figure 15 the correlation coefficient between the predicted and measured PM10 is presented for 8 different models. The maximum performance of the model is reached by using the following predictors (in strength order), with a correlation

The available data set in AIRSPACE project are given in Table 3:

**Parameters Instrument** Aerosol Optical Depth CIMEL Angstrom Exponent CIMEL Total Column Water Vapour CIMEL

PM 10 Dust Track TSI

Meteorological Data METAR-LCRA

PBL height LIDAR

**Table 2.** AIRSPACE dataset used for the statistical model

**# transformation Type of parameter involved**

3 Departures from mean value of Xi

4 Ratio of mean value of Xi

coefficient on the order of CC=0.85

1 Ln(Xi) 2 Xi

**Table 3.**

Sanford, 1982).

boundary layer height.

200 Remote Sensing of Environment: Integrated Approaches

The results of the above sensitivity analysis indicate the maximum performance of the model of the order of CC=0.85 as shown in equation 2:

$$\begin{aligned} \text{Ln(PM}\_{10}) &= \\ -4.11 + 0.952(\log AOT) + 0.299(\log AE) - 0.393(\text{MVdep}) + 0.643(\text{ln}(T)) + 0.008(\text{RH}) - 0.096(\text{ln(WS)}) \end{aligned} \tag{2}$$

Finally, using formula 2 as the best model and the coefficients derived and shown in Table 4, the relationship between the model's prediction and the measured PM10 concentrations, is shown in Figure 16. The residuals, i.e, the differences between the measured and the predicted values of the PM concentration are shown in Figure 17. The points in the residual plot in Figure 17 are randomly dispersed around the horizontal axis, thus, a linear regression model is appropriate.

**5.5. Chemical model**

Within AIRSPACE project, a high resolution atmospheric Chemistry General Circulation Model (AC-GCM) was used to study the emission, transport and deposition of dust. The Modular Earth Sub-model System (MESSy version 2.41) (Joeckel et al., 2005; 2006; 2010) is an earth system model which is capable of running with multiple representations of processes simultaneously paired to the core atmospheric general circulation model (ECHAM5). The model configuration used in the present study has a spectral resolution of T255L31 (0.5°, 50km) and 31 vertical levels up to 10 hPa. Gleser et al. (2012) emphasized the importance of higher resolution simulations for better dust representation in the model. As this is a global model, no boundary conditions are necessary. All known emission sources are included, while the initial conditions originate from the ERA40 reanalysis data (European Centre for Medium-Range Weather Forecasts - ECMWF) at 0.5-degree resolution. Every 12 hours of operation, the model fields are moved towards the ERA40 data in order to simulate the meteorological conditions, as precised as possible. In order to reduce computational time, the model uses a simplified chemistry module, preserving only the sulfate and NOx interactions which are considered the most important as far as the aerosols are considered. The model output is averaged and stored over 5-hour intervals, which provides an entire diurnal cycle after 5 days. The configuration includes also a simplified sulphate chemistry scheme (Gleser et al., 2012) allowing the production of sulphuric acid and particulate sulphate, which play an important role in transforming dust particles from hydrophobic into hydrophilic, thus affecting their ability to interact with clouds and be removed by precipitation (Astitha et al., 2012). The ammonia (NH3) reaction with sulphate and corresponding coating with dust (Ginoux et al., 2012) is also considered in this study. Due to the focus on dust episodes, a reduced version of the atmospheric chemistry scheme was applied which did not account for secondary inorganic and organic aerosol species associated with air pollution. The model used ECMWF gridded meteorological data to represent the actual meteorological conditions. To ensure adequate representation of the pollutants and dust in the atmosphere, the model runs for 15 days (spinoff) to create from the meteorology and the emissions the current weather conditions. This strategy ensures that the existing pollutants not represented in the model are removed from the atmosphere, while the sources will produce pollutants that will be dispersed in the atmosphere. After the initial spin-off, the atmospheric conditions represented from the model fields and the pollutant concentrations are considered as close to reality as possible. The model

Air Pollution from Space http://dx.doi.org/10.5772/39310 203

simulation was performed over the period of September to October 2011.

provide a clear picture for the evaluation assessment of the model.

The most significant issue for the operational run of a numerical model prediction of the dust is the complete absence of initial conditions for pollutant and dust concentrations. This enforces the utilization of global models to simulate the atmosphere with extremely accurate emission inventories which are absent or not complete for North Africa and Eastern Mediter‐ ranean. The latter is an important source of uncertainty for concentrations. Furthermore, the sparse coverage of measurements for the spatial validation of the model in the region does not

The use of a global model necessitated the utilization of a large grid due to computational limitations. The global grid introduced an adequate representation of the topography of the

**Figure 16.** Comparison between predicted and measured PM10 by TSI DUST Track at Limassol (Red line : linear fit )

**Figure 17.** Differences between the measured and the predicted value of the PM concentration (Residual plots)

#### **5.5. Chemical model**

Finally, using formula 2 as the best model and the coefficients derived and shown in Table 4, the relationship between the model's prediction and the measured PM10 concentrations, is shown in Figure 16. The residuals, i.e, the differences between the measured and the predicted values of the PM concentration are shown in Figure 17. The points in the residual plot in Figure 17 are randomly dispersed around the horizontal axis, thus, a linear regression model is

**Figure 16.** Comparison between predicted and measured PM10 by TSI DUST Track at Limassol (Red line : linear fit )

**Figure 17.** Differences between the measured and the predicted value of the PM concentration (Residual plots)

appropriate.

202 Remote Sensing of Environment: Integrated Approaches

Within AIRSPACE project, a high resolution atmospheric Chemistry General Circulation Model (AC-GCM) was used to study the emission, transport and deposition of dust. The Modular Earth Sub-model System (MESSy version 2.41) (Joeckel et al., 2005; 2006; 2010) is an earth system model which is capable of running with multiple representations of processes simultaneously paired to the core atmospheric general circulation model (ECHAM5). The model configuration used in the present study has a spectral resolution of T255L31 (0.5°, 50km) and 31 vertical levels up to 10 hPa. Gleser et al. (2012) emphasized the importance of higher resolution simulations for better dust representation in the model. As this is a global model, no boundary conditions are necessary. All known emission sources are included, while the initial conditions originate from the ERA40 reanalysis data (European Centre for Medium-Range Weather Forecasts - ECMWF) at 0.5-degree resolution. Every 12 hours of operation, the model fields are moved towards the ERA40 data in order to simulate the meteorological conditions, as precised as possible. In order to reduce computational time, the model uses a simplified chemistry module, preserving only the sulfate and NOx interactions which are considered the most important as far as the aerosols are considered. The model output is averaged and stored over 5-hour intervals, which provides an entire diurnal cycle after 5 days. The configuration includes also a simplified sulphate chemistry scheme (Gleser et al., 2012) allowing the production of sulphuric acid and particulate sulphate, which play an important role in transforming dust particles from hydrophobic into hydrophilic, thus affecting their ability to interact with clouds and be removed by precipitation (Astitha et al., 2012). The ammonia (NH3) reaction with sulphate and corresponding coating with dust (Ginoux et al., 2012) is also considered in this study. Due to the focus on dust episodes, a reduced version of the atmospheric chemistry scheme was applied which did not account for secondary inorganic and organic aerosol species associated with air pollution. The model used ECMWF gridded meteorological data to represent the actual meteorological conditions. To ensure adequate representation of the pollutants and dust in the atmosphere, the model runs for 15 days (spinoff) to create from the meteorology and the emissions the current weather conditions. This strategy ensures that the existing pollutants not represented in the model are removed from the atmosphere, while the sources will produce pollutants that will be dispersed in the atmosphere. After the initial spin-off, the atmospheric conditions represented from the model fields and the pollutant concentrations are considered as close to reality as possible. The model simulation was performed over the period of September to October 2011.

The most significant issue for the operational run of a numerical model prediction of the dust is the complete absence of initial conditions for pollutant and dust concentrations. This enforces the utilization of global models to simulate the atmosphere with extremely accurate emission inventories which are absent or not complete for North Africa and Eastern Mediter‐ ranean. The latter is an important source of uncertainty for concentrations. Furthermore, the sparse coverage of measurements for the spatial validation of the model in the region does not provide a clear picture for the evaluation assessment of the model.

The use of a global model necessitated the utilization of a large grid due to computational limitations. The global grid introduced an adequate representation of the topography of the

**Figure 18.** AERONET stations used to evaluate the model results

models and requires special parameterization of processes that often lead to errors. Another restriction is the simplified chemistry used for the simulation. The computational power necessary for the implementation of a full chemistry scheme is not currently available.

**Figure 19.** Scatter plot between modeled and observed AOT for different AERONET stations

Furthermore, model AOT estimations have been compared with the available AOT measure‐ ments from CUT-TEPAK AERONET site. Figure 20 shows the time evolution of the AOT for the Limassol AERONET station together with the model results. As shown in Figure 20, the model is generally, in agreement with observations in both magnitude and timing for Limassol with respect to the average measured values. The comparison between the modeled and

Air Pollution from Space http://dx.doi.org/10.5772/39310 205

An integrated methodology for assessing and studying air pollution in several areas of Cyprus was presented through the AIRSPACE project. Satellite derived aerosol optical thickness data along with LIDAR, sun-photometric and in-situ (PM) measurements were analyzed. The proposed integration of several tools and technologies provides to the user an alternative way

First, a new multiple linear regression model for estimating PM10 using AOT values and some other auxiliary meteorological atmospheric parameters has been developed for the urban area of Limassol in Cyprus. AOT can be retrieved by satellite sensors and is validated on the ground

observed AOT indicates the ability of the model to simulate the AOT adequately.

**5. Conclusions**

for assessing and monitoring air pollution.

The model results were evaluated using the AOT fields provided by the NASA AERONET available from http://aeronet.gsfc.nasa.gov. The data comparison represents the AOT for all aerosols simulated in the model as well as those observed in the atmosphere at 550nm wavelength. The observed AOT was averaged over the 5-hour output intervals in line with the averaged AOT over the same period from the model. Figure 18 shows the eight AERONET stations which observational data were available during the simulation period and which were used in this study. These stations are not necessarily located in dust-dominated regions but can be more strongly affected by other aerosol types, including air pollution.

The scatter plot between the modeled and observed AOT is shown in Figure 19. Different colors and symbols are used for each station ID (see legend). As shown, the model is capable of simulating the AOT in general. However, at some stations (Leipzig, Palencia, Paris) the model tends to underestimate the observed AOT. This is explained by the use of the reduced atmospheric chemistry scheme in the model that does not fully account for urban air pollution in addition to the unresolved physics at small scales in the global models. However, the comparison of the output of the model for the AOT with the measured values from the AERONET network indicates that the simulated atmosphere is valid in areas with similar climatological and industrial characteristics to Cyprus, while for areas with heavy industry, there is a significant deviation which can be justified from the reduced chemistry module used for the runs.

**Figure 19.** Scatter plot between modeled and observed AOT for different AERONET stations

Furthermore, model AOT estimations have been compared with the available AOT measure‐ ments from CUT-TEPAK AERONET site. Figure 20 shows the time evolution of the AOT for the Limassol AERONET station together with the model results. As shown in Figure 20, the model is generally, in agreement with observations in both magnitude and timing for Limassol with respect to the average measured values. The comparison between the modeled and observed AOT indicates the ability of the model to simulate the AOT adequately.
