Aerosol levels at Mediterranean Basin are significantly affected by desert dust that is eroded in North Africa and is transported northwards. This study aims to assess the performance of the Dust REgional Atmospheric Model (BSC-DREAM8b) in the prediction of dust outbreaks near the surface in Eastern Mediterranean. For this purpose, model PM10 predictions covering a 7-year period and PM10 observations at five surface monitoring sites in Greece are used. A quantitative criterion is set to select the significant dust outbreaks defined as those when the predicted PM10 surface concentration exceeds 12
A large portion of atmospheric Particulate Matter (PM) is derived from arid regions of the Earth (North Africa, Arabian Peninsula, central Asia, Australia, etc.) and is distributed all over the globe. Saharan desert is responsible for up to half of the global mineral dust emissions, thus it is considered as the most important dust source worldwide [
The North African desert dust cycle depends on the synoptic circulations, which control the frequency and extent of transport and on the washout by precipitation which influences the residences time of dust particles in the atmosphere. The bulk of the dust is transported westward into the Atlantic Ocean and an unnegligible part is transported northward across the Mediterranean basin to southern and even central Europe [
Regional modelling is considered as a useful tool to simulate and predict the dust cycle in the atmosphere. The Dust REgional Atmospheric Model (DREAM) [
The aim of this study is to examine the capability of the updated version of DREAM at the Barcelona Supercomputer Center (BSC-DREAM8b, [
PM10 data recorded automatically at five monitoring sites in Greece are exploited in this study (Table
PM10 monitoring sites1.
Monitoring site | Latitude (N) | Longitude (E) | Altitude (m) | Type | Data availability (%) |
---|---|---|---|---|---|
Heraklion | 25.13 | 10 | Urban | 71 | |
Finokalia | 35.33 | 25.67 | 150 | Background | 70 |
Thrakomakedones | 38.14 | 23.76 | 550 | Suburban | 82 |
Volos | 39.37 | 22.94 | 31 | Urban | 82 |
Panorama | 40.59 | 23.03 | 330 | Suburban | 81 |
The updated version of DREAM model, BSC-DREAM8b [
For the present study, simulation is initialized with 24-hourly (at 00UTC) updated NCEP (National Centers for Environmental Prediction)
BSC-DREAM8b’s simulations regarding the contribution of transported dust to surface PM10 concentration at Finokalia, Thrakomakedones, Volos, and Panorama are exploited in this study. As Finokalia station is situated 70 km northeast of Heraklion, and although Heraklion is included in an adjacent model’s grid point, it is assumed that the contribution of transported dust to surface PM10 concentration at Heraklion is the same as at Finokalia.
In this study, the capability of the BSC-DREAM8b to simulate the desert dust transported in the Eastern Mediterranea especially during dust outbreaks is examined. Therefore, model outputs have to be filtered so as to select the most important dust transport events that significantly influence surface PM10 levels. It was decided to define one threshold for all monitoring stations, based on the modelled PM10 surface concentrations. For this purpose, modelled PM10 surface concentrations are classified and the critical part of their distribution is presented in Figure
Frequency distribution of modelled PM10 surface concentrations. Finokalia: solid line; Thrakomakedones: bold solid line; Volos: dashed line; Panorama: dots.
The criterion that is defined above is applied to the time series predicted by BSC-DREAM8b. Quantitative information about the selected events is provided in Table
Statistics for the selected dust outbreaks during the examined 7-year period.
Site | ||||
---|---|---|---|---|
Finokalia | Thrakomakedones | Volos | Panorama | |
Number of days when predicted PM10 surface concentration exceeds the threshold value (12 | 134 | 67 | 49 | 20 |
Number of events identified | 78 | 41 | 32 | 11 |
The duration of the selected events is presented in Figure
(a) Duration and (b) seasonality of the most significant dust outbreaks according to BSC-DREAM8b.
The seasonal distribution of the selected events is presented in Figure
BSC-DREAM8b’s outputs are compared to surface PM10 observations. The efficiency of the model to predict dust transport episodes is discussed and the contribution of transported desert dust to the determination of surface PM10 concentrations at the areas studied is presented. Predicting PM concentrations in time and space is important for air quality and health concerns, weather prediction, and climate studies. As dust transport events influence surface levels for relatively short time periods, it is preferred to compare the model’s results to the short-term component of measured PM10 concentration. Short-term variations are attributed to weather processes. They could be separated from the original time series by using the Kolmogorov-Zurbenko (KZm
In this study, the KZ15,5 filter is applied to the time series of the observed daily average PM10 concentration. Days are separated in two groups. One group includes the days when a dust outbreak is identified by the application of the selection criterion that is presented above, while the other group includes the rest of the days. The short-term component is classified and the frequency of occurrence of each class for every group is presented in Figure
Frequency distribution of the short-term component of the observed PM10 concentrations at (a) Finokalia, (b) Heraklion, (c) Thrakomakedones, (d) Volos, and (e) Panorama; Grey bars correspond to the days when a dust outbreak is identified, while black bars correspond to the rest of the days.
It should be noted though that the increase in PM10 levels, observed during the days when dust events take place according to BSC-DREAM8b, could be attributed not only to the transported dust but also to the local meteorological conditions that are prevailing during dust outbreaks. It is well known that specific meteorological conditions could favour the accumulation of dust as well as the accumulation of pollutants that are locally emitted due to anthropogenic activities. Moreover, anthropogenic as well as natural emissions of PM10 undergo a significant temporal variation that could mask the influence of the transported Saharan dust.
The analysis of the data that refer to the selected events is extended in order to quantify the contribution of the transported dust to the observed near surface PM10 levels. For this reason, model predictions are compared to PM10 observations, as well as to their short-term component. Four cases are examined. Case (a): model predictions are compared to the short-term component of the PM10 observations. Case (b): model predictions are compared to the PM10 observations. Case (c): model predictions are compared to the short-term component of PM10 observations, only when the short-term component of PM10 concentrations is higher than the value forecasted by the model. Case (d): model predictions are compared to the PM10 observations, only when the short-term component of PM10 concentrations is higher than the value forecasted by the model. The results of the cases (a) and (b) include uncertainties as model’s possible mispredictions are not excluded from the analysis. The results of cases (c) and (d) are more indicative as the introduction of the criterion regarding the short-term component cuts off the possible mispredicted events and allows the inclusion in the analysis of only those days when dust transport could have contributed to the increase of the observed PM10 levels.
A regression line is calculated for every site for each of the cases examined, considering the model’s prediction as the independent variable and the PM10 concentration or its short-term component as the dependent one. Additionally, intercept is forced to zero. The results of all regressions are presented in Table
Contribution of transported dust to near surface PM10 levels1.
Site | Case a | Case b | Case c | Case d | ||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|
S | R2 | N | S | R2 | N | S | R2 | N | S | R2 | N | |
Finokalia | 1.48 | 0.44 | 33 | 3.69 | 0.64 | 45 | 2.61 | 0.74 | 17 | 5.32 | 0.82 | 18 |
Thrakomakedones | 2.06 | 0.49 | 40 | 3.96 | 0.68 | 59 | 2.58 | 0.86 | 29 | 5.04 | 0.80 | 30 |
Volos | 1.19 | 0.42 | 25 | 3.55 | 0.83 | 36 | 1.92 | 0.71 | 13 | 4.08 | 0.83 | 13 |
Panorama | 1.80 | 0.85 | 12 | 3.77 | 0.95 | 12 | 1.99 | 0.92 | 10 | 3.94 | 0.97 | 10 |
According to the analysis of case (c), dust contribution accounts for the 38, 39, 52, and 50% of the increase in PM10 concentrations at Finokalia, Thrakomakedones, Volos, and Panorama, respectively, while according to the results of case (d), the 19, 20, 25, and 25% of the measured PM10 concentrations is attributed to dust transport. The percentage differences support the conclusion drawn above regarding the drawback of the application of the KZ filter.
As Finokalia is a background site, it was expected that dust contribution to surface PM10 levels would be relatively higher. This fact indicates that BSC-DREAM8b probably underestimates the dust transported at Finokalia station during the dust outbreaks and that the background PM10 levels at this site could be influenced not only by the transported desert dust but also by particles emitted by other natural sources. Gerasopoulos et al. [
The objective of this study is to assess the efficiency of BSC-DREAM8b model to predict Saharan dust transport episodes in the Eastern Mediterranean. For this reason, model outputs that cover the period 2001–2007 are compared to PM10 data recorded by five automatic monitoring stations in Greece. A quantitative criterion is established in order to select the most important dust outbreaks. When modelled surface concentration becomes higher than 12
The duration of the selected events is longer at the monitoring sites located to south Greece (namely, Finokalia and Thrakomakedones). A significant dust outbreak that lasted more than 5 days is detected at all sites in June 2007. This event coincided with the heat wave episode that influenced the Eastern Mediterranean, contributing to the aggravation of air quality and discomfort conditions.
Dust transport is more favoured during summer and spring at Thrakomakedones, Volos, and Panorama and during winter and spring at Finokalia. Additionally, ~20% of the selected dust outbreaks at Finokalia, Thrakomakedones, and Volos are observed during autumn, therefore autumn could also be considered as a season when important dust transport events occur over the Eastern Mediterranean.
In order to evaluate BSC-DREAM8b outputs, the modelled values are compared to the daily average values of surface PM10 concentration, as well as to their short-term component. Short-term variations are attributed to weather processes and do not include seasonal or long term variations. The short-term component is separated by applying a KZ15,5 filter to the original time series and then by subtracting the filtered time series from the original one. The values of the short-term component are positive (indicating an increase in PM10 levels) and higher than 12
The work has been financed by the MED-APICE project in co financed by the European Regional Development Fund in the framework of the MED Programme.