PREDICTING DAILY MAXIMUM OZONE CONCENTRATION IN AN URBAN AREA Papanastasiou D. K. and Melas D. Laboratory of Atmospheric Physics, Department of Physics, Campus Box 149, Aristotle University of Thessaloniki, 54124 Thessaloniki, Greece E-mail: dkpapan@auth.gr, melas@auth.gr ABSTRACT The objective of this paper is to develop an analytical model, relating daily maximum hourly (DMH) value of ozone concentration in the urban area of Volos in central Greece, with various meteorological variables and pollution, in order to predict the next day’s DMH value of ozone concentration. The evaluation of the developed relationship showed that its degree of successfulness is very promising. KEY WORDS: Ozone, air pollution, air pollution prediction, regression model. INTRODUCTION Ozone is an oxidant chemical compound, which is one of the most significant components of smog, which appears in the atmosphere of a lot of cities around the world. Indeed, ozone is noticed as a smog indicator. Ozone has serious effects on human and animal health, vegetation and materials1,2,3,4,5,6,7. Several authors found that the function of respiration is reduced proportionally to the increasing concentrations of ambient ozone. Respiratory symptoms such as coughing and asthma have been associated with ozone concentrations as low as 150 ppb. A number of studies evaluating animals (rats and monkeys) exposed to ozone for a few hours or days have shown alterations in the respiratory tract, in which the lowest observed effect levels were in the range of 80 – 200 ppb. The symptoms to that exposure were lung infections and inflammation and morphological alteration of lung. Additionally, ozone has a great impact in the agricultural yields and it causes foliage damage. European Community (EC) has established maximum thresholds for ozone concentration in order to avoid its unpleasant consequences. The information and the alert thresholds for ozone that are proposed by EC to European countries with the Directive 2002/3 are 180 μg/m3 and 240 μg/m3 respectively, for a period of 1 hour. The same Directive specifies a target value for the protection of human health and it says that from the beginning of 2010 the daily maximum average value for a period of 8 hours of ozone concentration should not exceed 120 μg/m3 on more than 25 days per year. From those that are mentioned above is obvious that it is very significant to achieve a forecast of ozone concentration. When high concentrations are predicted, the authorities might announce a public warning and an advice to industry, so as people with respiratory problems could take some precautions and industries could lower the emissions of ozone formation precursors. A 24 hours forecast provides the ability for these actions. The relationship between pollutant concentrations and meteorological variables has been the subject of numerous studies during the past few decades. A variety of statistical methods 8,9,10,11,12 have been utilized in order to develop techniques, which will enable qualitative or quantitative short-term forecasts. A common method13,14,15,16,17,18,19,20,21 that it is used -1- widely is to correlate meteorological variables and concentration of other pollutants with the concentration of a certain pollutant. The variables used in analytical modelling are chosen to represent meteorological conditions unfavourable for pollutant dispersion and, to a certain degree, short-term variations in emissions. The purpose of the present study is to investigate the qualitative relation between the ozone concentration levels and meteorological variables and to develop an analytical model, relating DMH value of ozone concentration in the urban area of Volos with various meteorological variables and pollution, in order to predict the next day’s DMH value of ozone concentration. DESCRIPTION OF THE VOLOS AREA AND DATA USED The city of Volos in Greece is located at the north coast of Pagasitikos gulf, where there is a smaller gulf, the gulf of Volos. The urban area covers an approximately 51,4 km2 area and approximately 4 km of the coast of Pagasitikos gulf. At a distance of approximately 3 km to the northeast of the city is mountain Pelion, which extends from the north to the southeast of the city. Its higher peaks are at 1551 m and at 1471 m, which are to the northeast and to the east of the city respectively and at distance of approximately 12 km from the city. Pagasitikos gulf expands to the south of the city and it takes out to the Aegean Sea through a channel of approximately 5,5 km width which is at 30 km southern of the city. To the northwest of the city are some hills with 500 m mean height. From the description above, it is derived that the city has two physical ventilation channels. The one is the gulf and the other is that small valley that is vested by the hills that described above. The direction of the small valley is similar with the direction that sea breeze blows22. In the greater urban area of Volos live 118.564 people according to the 2001’s national inventory, but it is believed that the habitants nowadays are much more. The roads inside the city are generally narrow and traffic is fairly intense during some hours of the day, especially when the shops are opened. Tourism and the port for passengers and for wares are two factors that aggravate traffic. There are two rather small industrial areas to the west of the town and a big cement industry to the east of the town. The Department of Environment of the Prefecture of Magnesia selects data for some air pollutants and also selects data for some meteorological variables. So, all the data that are used in this analysis were taken from that station and they refer to the period from 2001 until 2003. PREDICTORS OF THE CONCENTRATION LEVELS OF POLLUTANTS23 The most important meteorological processes that influence the local pollutant concentrations is dry and wet deposition, advection by the horizontal wind, vertical dilution within the boundary layer accomplished mainly by turbulence and photochemical reactions with other gases, which occur under the effect of solar radiation. It is thus very important that the meteorological variables used in analytical modelling for forecasting pollution concentrations cover the above-mentioned atmospheric processes. These variables are regarded as predictors for the pollutant levels. Unfortunately there weren’t data for some variables that can be regarded as predictors for the pollutant levels, such as the day-to-day air temperature change at 850 hPa, the rainfall and the total radiation flux. Also, there weren’t available data for vehicular traffic. The present analysis is limited to these variables, which were found to correlate significantly with the DMH values of ozone concentration. So, we used the horizontal wind speed -2- and the wind direction. Also, we used the ground temperature and the relative humidity, so as radiation flux could be estimated indirectly. Also, because some variables weren’t correlated linear with ozone concentration, we used decimal logarithms, indexes etc in order to achieve the best correlation between the predictors and the predictand. Persistency of high pollution levels Earlier studies have shown that the possibility of occurrence of pollution episodes is increased if the previous day’s pollution levels were higher than normal. In the present study we used the previous day’s DMH value of ozone concentration, as a parameter indicating the potential of a pollution episode for the next 24 hours. Solar radiation flux Photochemical reactions in the atmosphere occur under the effect of solar radiation, particularly when the magnitude of total solar radiation reaching the earth exceeds a threshold. Also, solar radiation relates with the mixing height, as it relates with the turbulent kinetic energy, which affects the mixing height. When the mixing layer over the urban area is shallow, pollutants tend to accumulate in the surface air. It is thus expected that, among other parameters, ozone concentration will depend upon the intensity of the incoming solar radiation and therefore upon the altitude of the sun and the atmospheric absorption. For these reasons, solar radiation flux should be estimated in calculations. But, since there weren’t data for the radiation flux, the ground temperature and the relative humidity were used, so as radiation flux could be estimated indirectly. It is obvious that when radiation flux increases, ground temperature is higher and relative humidity is smaller. Also, high air temperatures are associated with the slow moving high pressure systems, clear and sunny skies, stagnant circulation and subsiding upper air. All these contribute to the production and accumulation of ozone, so temperature can be regarded as one of the strongest predictors of ozone concentration. Ambient humidity affects the minimum temperature via two mechanisms. Firstly via the absorption of long-wave radiation emitted by the earth that would otherwise under dry and cloudless conditions be lost to sky and secondly via the release of the latent heat of condensation as the sensible temperature falls to the dew point. The wind speed and direction The basic meteorological parameters determining the horizontal transport and dispersion of air pollutants are the mean wind speed and the wind direction. There are not any studies for Volos area from which someone can derive that a specific wind direction causes an increase to the pollutants in area. The only thing that someone can say concerning to ozone, is that the DMH value of ozone concentration usually appears when sea breeze is in total development. This is happening typically at 17:00 local time, when wind blows from south – southeast directions with the highest speeds. Annual variation of a pollutant’s concentration Some pollutants present a significant annual variation. Concerning to ozone, during summer its concentrations will be higher, as the amount of solar radiation is bigger during summer. As it is known, solar radiation affects ozone concentration, as it is necessary in pro- -3- ducing ozone from nitrogen dioxide. In this study we used the following expression to describe the annual variation of ozone, so as to achieve a consequence among the periods. 2 π Mi Y1 cos , where Mi is a number from 1 to 12 that refers to the month. 12 ESTIMATION OF THE NEXT DAY’S DMH VALUE OF OZONE CONCENTRATION The quantitative estimation of next day’s DMH value of ozone concentration is based on an analytical expression derived by performing multiple regression analysis. As independent variables we used only those predictands that are correlated significantly with the dependent variable. The expression of multiple regression analysis has the following form: y m1 x1 m2 x 2 m3 x 3 m4 x 4 m5 x 5 m6 x 6 m7 x 7 b where: y The forecasted DMH value of ozone concentration (μg/m3). x 1 The DMH value of ozone concentration (μg/m3). x2 The value of ground temperature (0C) when the DMH value of ozone concentration appears. x3 The DMH value of ground temperature (0C). The daily minimum hourly value of relative humidity (%). x4 x5 The inverse of the hourly mean value of wind speed (m/s) at 16:00 - 17:00 hours (local time). WD 110 π x6 Υ2 1 sin The value Υ2 of the following expression: 180 where WD is the prevailing wind direction in degrees at 16:00 - 17:00 hours (local time). From the definition of this index, if Υ2 equals to 0 then wind direction is 1700. This is the mean direction from which sea breeze blows at 16:00 - 17:00 hours. The value of Y1 . x7 Variables x 3 , x 4 , x 5 , x 6 , x 7 refer to the day that we want to predict the DMH value of ozone concentration and variables x1, x 2 refer to the day before. VALIDATION OF THE ANALYTICAL MODEL As it is already said, the variables that were used in this study are only those that are correlated significantly with the DMH value of ozone concentration. From the analysis derived that the variable that has the most significant effect to one day’s ozone concentration is ozone concentration of the day before, followed by the ground temperature (table 1). This conclusion is in accordance with the theory of ozone formation and with the results of earlier studies. -4- Table 1: Correlation coefficients of the used variables with y (DMH value of ozone concentration) Correlation Coefficient x1 0,89 V a r i a b l e x3 x5 x6 x2 x4 0,75 0,75 0,63 0,39 0,48 x7 0,79 Data were split in two parts. The 66 % of data was used in multiple regression analysis and the 34 % of data was used for the evaluation of the analytical model. The analytical model that was produced from the 66 % of data was used to predict the DMH values of ozone concentration for the rest 34 % of days. The result was compared with the DMH values of ozone concentration that were observed during the same days. The comparison is shown in Figure 1. From figure 1 it is noticeable that the agreement between observed and predicted DMH values of ozone concentration is perfect and the coefficient of determination R of the analytical expression derived by performing multiple regression analysis is 0,93. Thus, the evaluation of the developed relationship showed that its degree of successfulness is very promising. 0 10 30 40 50 60 70 80 90 100 110 120 130 140 150 160 170 170 170 160 160 R=0,93 150 PREDICTED DMH VALUE OF OZONE CONCENTRATION 20 150 140 140 130 130 120 120 110 110 100 100 90 90 80 80 70 70 60 60 50 50 40 40 30 30 20 20 10 10 0 0 0 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 160 170 OBSERVED DMH VALUE OF OZONE CONCENTRATION -5- Figure 1: Comparison between predicted and observed DMH value of ozone concentration (μg/m3), for the period 2001 – 2003. REFERENCES 1. Cole S. (1996). Lung damage linked to combined fine particle, ozone exposure in new toxicology study. Environ. Sci. Technol., 30, 9328A. 2. Evans L.S., Adamski J.H., Renfro J.R. (1996). Relationships between cellular injury, visible injury of leaves and ozone exposure levels for several dicotyledonous plant species at Great Smoky Mountains National Park. Environ. Exp. Botany 36 (2), 229 – 237. 3. Novak K., Skelly J.M., Schaub M., Krauchi N., Hug C., Landolt W., Bleuler P. (2003). Ozone air pollution and foliar injury development on native plants of Switzerland. Environ. Poll. 125, 41 – 52. 4. Sartor F., Snacken R., Demuth C., Walckiers D. (1995). Temperature, ambient ozone levels and mortality during summer 1994 in Belgium. Environ. Res. 70, 105 – 113. 5. Shan Y, Feng S., Izuta T., Aoki M., Totsuka T. (1996). The individual and combined effects of ozone and simulated acid rain on growth, gas exchange rate water-use efficiency of Pinus Armadi Franch. Environ. Pollut. 91 (3), 355 – 361. 6. Puskás, J., Nowinszky, L., Bozó, L., Ferenczy, Z. (2003a): The Ozone Content of the Air. In: L. Nowinszky (ed.) The Handbook of Light Trapping, Savaria University Press, Szombathely pp. 170-172. 7. Puskás, J., Nowinszky, L., Bozó, L., Ferenczy, Z. (2003b): A levegő ózontartalma. In: Nowinszky L. (szerk.) A fénycsapdázás kézikönyve, Savaria University Press, Szombathely pp. 168-169. 8. Berlyand M.E. (1991). Prediction and regulation of air pollution. Kluwer Academic Publishers. 9. Diem J.E., Comrie A.C. (2002). Predictive mapping of air pollution involving sparse spatial observations. Environmental Pollution 119, 99 – 117. 10. Prybutok V.R., Yi J., Mitchell D. (2000). Comparison of neural network models with ARIMA and regression models for prediction of Houston’s daily maximum ozone concentrations. European Journal of Operational Research, Vol. 122, 31 – 40. 11. Yi J., Prybutok V.R. (1996). A neural network model forecasting for prediction of daily maximum ozone concentration in an industrialized urban area. Environmental Pollution, Vol. 92, No 3, 349 – 357. 12. Robeson S.M., Steyn D.G. (1990). Evaluation and comparison of statistical forecast models for daily maximum ozone concentrations. Atmos. Environ. 24B, 303 – 312. 13. Fuller G.W., Carslaw D.C., Lodge H.W. (2002). An empirical approach for the prediction of daily mean PM10 concentrations. Atmos. Environ., Vol. 36, 1431 – 1441. 14. Hubbard M.C., Cobourn W.G. (1998). Development of a regression model to forecast ground-level ozone concentration in Louisville, KY. Atmos. Environ., Vol. 32, No 14/15, 2637 – 2647. 15. Massart B.G.J., Kvalheim O.M., Stige L., Aasheim R. (1998). Ozone forecasting from meteorological variables, Part II, Daily maximum ground-level O3 concentration from local weather forecasts. Chemometrics and Intelligent Laboratory Systems, Vol. 42, 191 – 197. 16. Papanastasiou D.K., Melas D., Zerefos C.F. (2002). Forecast of ozone levels in the region of Volos. 6th Hellenic Conference in Meteorology, Climatology and Atmospheric Physics. Ioannina, Greece. 17. Papanastasiou D.K., Melas D., Zerefos C.F. (2003). Relationship of meteorological variables and pollution with ozone concentrations in an urban area. 2nd International Conference on Applications of Natural, Technological and Economical Sciences. Szombathely, Hungary. -6- 18. Papanastasiou D.K., Melas D. (2004). Analysis and forecast of PM10 concentration in a medium size city. 3nd International Conference on Applications of Natural, Technological and Economical Sciences. Szombathely, Hungary. 19. Perez P., Reyes J. (2002). Prediction of maximum of 24-h average of PM10 concentrations 30 h in advance in Santiago, Chile. Atmos. Environ., Vol. 36, 4555 – 4561. 20. Ziomas I.C., Melas D., Zerefos C.S., Bais A.F. (1995). On the relationship between peak ozone levels and meteorological variables. Fresenious Envir Bull 4, 53-58. 21. Ziomas I.C., Melas D., Zerefos C.S., Paliatsos A., Bais A.F. (1995). Forecasting peak pollutant levels using meteorological variables and indexes. Atmos. Environ. 29, 3703 – 3711. 22. Papanastasiou D.K. (1995). Sea breeze in the region of Volos. Laboratory of Atmospheric Physics, Department of Physics, Aristotle University of Thessaloniki, Greece. 23. Berlyand M.E. (1991). Prediction and regulation of air pollution. Atmospheric Sciences Library, Kluwer Academic. Dordecht, The Netherlands. -7-