REGIONAL AIR QUALITY FORECASTING OVER GREECE WITHIN PROMOTE Poupkou A. (1), D. Melas (1), I. Kioutsioukis (2), I. Lisaridis (1), P. Symeonidis (1), D. Balis (1), S. Karathanasis (3) and S. Kazadzis (1) (1) Laboratory of Atmospheric Physics, Aristotle University of Thessaloniki, Box 149, 4124 Thessaloniki, Greece (2) Democritus University of Thrace, P.O. Box 447, 67 Xanthi, Greece (3) Region of Central Macedonia, 48 Thessaloniki, Greece ABSTRACT An air quality forecast system is set up in order to predict near surface air pollutants levels over Greece. The forecast system consists of the prognostic meteorological model MM and the photochemical air quality model CAMx and runs operationally on a daily basis to issue 72-hour forecasts. The modeling domain, having spatial resolution equal to km, covers Greece and a part of the neighbouring Balkan countries. MM simulations are initialized and develop lateral boundary conditions using GFS Global Forecast. The boundary concentrations of the photochemical model are predictions of the European scale Chemistry Transport EURAD-CTM. The emission data used as CAMx input data are hourly values of biogenic NMVOCs and anthropogenic NOx, NMVOCs and CO emissions. The forecast system produces maps of daily maximum hourly concentrations of ozone and nitrogen dioxide and daily maximum 8haverage concentrations of carbon monoxide. Evaluation of the air quality forecasting has been performed by Greek environmental authorities. The forecast system is capable of reproducing the main features of air quality over the area where observations were available. ing system performance is better regarding O 3 and NO 2 while CO concentrations are systematically underestimated. More reliable is the forecast issued during days when it is not raining. 1 DESCRIPTION OF THE FORECAST SYSTEM The air quality forecast system consists of two numerical models: 1. the prognostic meteorological model PSU/NCAR mesoscale model (MM version 3.7) and 2. the photochemical air quality model Comprehensive Air quality with extensions (CAMx version 4.). MM is a limited-area, nonhydrostatic, terrain-following sigma-coordinate model designed to simulate or predict mesoscale atmospheric circulation []. CAMx is an Eulerian photochemical dispersion model that allows the integrated assessment of gaseous and particulate air pollution as it simulates the emission, dispersion, chemical reaction and removal of pollutants in the troposphere over many scales ranging from sub-urban to continental [4]. The air quality forecast system is set up at the Laboratory of Atmospheric Physics. It runs operationally on a daily basis to issue 72- hour forecasts of the near surface air pollutants concentrations, mainly ozone, nitrogen dioxide and carbon monoxide. The flow chart of the modeling system is presented in figure 1. Air quality predictions are performed on a regional scale and focus mainly on Greece. MM CAMx ing System Topography Land Use Meteorological Initial and Boundary Conditions MM Mesoscale Meteorological Preprocessor Emission Data CAMx Photochemical Air Quality Chemical Initial and Boundary Conditions Gas Phase Pollutant Concentrations Chemistry Parameters Photolysis Rates Fig. 1. Flow chart of the forecast system.
1.1 ing domain of the forecast system MM is implemented for two nested grids (figure 2). The coarse grid (D1) has x grid points and covers the greater part of the Balkan Region with a horizontal mesh width of km. The fine grid (D2) consists of 121 x 121 grid points having horizontal spatial resolution of km. It zooms in an area that covers Greece and a part of its neighboring countries. Both grids have the same vertical structure of 33 σ levels. The model top is mbar. CAMx modeling domain is a part of the MM fine grid. It is divided into 1 rows by 1 columns with horizontal grid resolution of km. There are vertical layers extending up to approximately 2. km above ground. The vertical layers are unevenly distributed with higher resolution at the near-surface layers. The first layer height is m. 1.2 s set-up and input data description Fig. 2. ing domain of the forecast system. In order to perform the meteorological forecasting, MM needs to acquire the latest predictions of meteorological parameter fields. Therefore, the global 12 UTC forecast of GFS of the National Center for Environmental Protection (NCEP) is transferred from an anonymous ftp-server to the local system. GFS forecast is usually available at around 16 UTC. Initial and boundary conditions, given in standard-pressure-surfaces with a horizontal resolution of one degree, are then interpolated onto the regional grid of the forecast system. Elevation and -category vegetation/land-use data used as input to MM model are derived from the USGS Data Center and have 2 minutes horizontal resolution. MM performs an 84-h forecast and model outputs are available around 3 UTC. In order to perform air quality forecasting, a preprocessor program prepares the MM meteorological fields in a format suitable for being used as input data to CAMx model. The emission data used are typical diurnal biogenic NMVOCs and anthropogenic NOx, NMVOCs and CO emission variations for every month of the year derived from an emission inventory with spatial resolution of km compiled for the modeling area by the Laboratory of Atmospheric Physics [7]. The biogenic emissions were calculated following the methodology of the EMEP/CORINAIR emission inventory guidebook [2]. The anthropogenic emission fields were estimated using data of: a) official national emission inventories (e.g. the transport sector emission inventory for Greece [8]) and b) international emission databases (e.g. EMEP, CORINAIR [3,1]). Examples of hourly anthropogenic and biogenic emission fields are presented in figure 3. The boundary concentrations of the chemical species are predictions of the European scale Chemistry Transport EURAD-CTM. The TUV radiative transfer and photolysis model, developed at the National Center of Atmospheric Research (NCAR), is used as a CAMx preprocessor to provide the air quality model with a multi-dimensional lookup table of photolytic rates [6]. The Carbon Bond Mechanism (CB-IV) is employed for solving chemical kinetics. The 72-h air quality forecast is completed around 4 UTC and is available on the web around UTC.
Anthropogenic NOx Emissions (kgr/h) Anthropogenic NOx Emissions (kgr/h) 1.E-4 to 2.1E+1 2.1E+1 to.e+1.e+1 to 1.E+2 1.E+2 to.e+2 1.E+3 to 2.E+3 1.E-4 to 2.1E+1 2.1E+1 to.e+1.e+1 to 1.E+2 1.E+2 to.e+2 1.E+3 to 2.E+3 Time period: : - 1: ST Month: June Time period: 12: - 13: ST Month: June Biogenic Isoprene Emissions (kgr-c/h) Biogenic Isoprene Emissions (kgr-c/h) 1.E-4 to.e+1.e+1 to 1.E+2 1.E+2 to 3.E+2 3.E+2 to.e+2 1.E+3 to 1.4E+3 1.E-4 to.e+1.e+1 to 1.E+2 1.E+2 to 3.E+2 3.E+2 to.e+2 1.E+3 to 1.4E+3 Time period: 7: - 8: ST Month: June Time period: 12: - 13: ST Month: June Fig. 3. Hourly anthropogenic and biogenic emission fields used as CAMx input data. 2 AIR QUALITY FORECAST RESULTS Examples of 24-h predicted daily maximum hourly concentrations of ozone and nitrogen dioxide and daily maximum 8h-average concentrations of carbon monoxide are presented in figure 4. NOx and CO concentrations have peaks scattered throughout the domain. NOx concentrations are maximum mainly over the large power plants and urban centres. The highest CO levels are found over the major urban agglomerations. Ozone concentrations have a more uniform distribution showing maximum values over maritime areas south of Athens which are influenced by the urban plume. Possible underestimations of pollutant concentrations over some areas of the modeling domain (e.g. Istanbul) are due to probable underestimations of the emission source data.
O 3 Daily Maximum (ppm) NO 2 Daily Maximum (ppm) CO Daily 8h-Average Maximum (ppm) Fig. 4. 24-h air quality forecast maps. 3 EVALUATION OF THE AIR QUALITY FORECAST The 24-h air quality forecast has been evaluated using O 3, NO 2 and CO surface data from the ground-level monitoring network of the Region of Central Macedonia (Greek authority responsible for the forecast system evaluation). The measuring stations are located at the greater area of Thessaloniki which is the second largest urban center of Greece. Averaged hourly observation data of the monitoring network were used for the evaluation conducted for the period of one month (December ). It should be pointed out that the comparison between measurements and model results cannot be straightforward since the former represent point measurements while the latter represent the average concentrations over a considerably larger area. In figure, the observed mean daily concentrations of O 3, NO 2 and CO are compared with forecasted values. The agreement between day to day variation of predicted and measured daily O 3 levels can be considered satisfactory during the second part of December. The first two weeks and more particularly the time period from to of December were characterised by rainy weather. During this period, the forecast system did not perform well regarding O 3 concentrations suggesting that wet removal processes are not well parameterised by the modeling system. Forecasted and observed mean daily NO 2 values are in very good agreement. Both the temporal pattern and the modeled NO 2 levels compare favourably with the measurements. Although the modeling system is capable of reproducing the temporal pattern of CO daily values variation, it systematically underestimates CO concentrations. CO emission data over the greater area of Thessaloniki has probably to be reconsidered. The forecasted and observed mean diurnal variation of O 3, NO 2 and CO is presented in figure 6. Diurnal variations have been calculated excluding modeled and observed data for days of rainy weather.
O 3 measurements reveal the absence of any diurnal signal (mean values in the range of 19 to 27 ppb) indicating limited photochemical O 3 production. The 24-h forecast reproduces quite well O 3 levels, showing however a typical diurnal O 3 variation with higher values reaching 36 ppb during afternoon and O 3 concentrations ranging between 24 ppb and 28 ppb during the rest of the day. According to observation data, NO 2 concentrations are lower during nighttime taking values as low as 12 ppb. There is only a weak variation during daytime when NO 2 mean hourly values range between ppb and 32 ppb. The maximum levels of the NO 2 mean diurnal cycle are satisfactorily captured by the 24-h forecast. However, model results show a strong variation of NO 2 concentrations during daytime characterized by low NO 2 values during afternoon suggesting photochemical activity that reduces NO 2 levels in order secondary pollutants to be produced. In the case of CO, there is only a qualitative agreement between the predicted and observed diurnal cycle according to which CO levels are higher during morning and at late evening hours. The observed maximum CO values reach 1.7 ppm. CO values are underestimated by the model by a factor of 3. O 3 Mean Daily Concentrations NO 2 Mean Daily Concentrations 4 4 1 2 3 4 6 7 8 9 11 12 13 14 16 17 18 19 21 22 23 24 26 27 28 29 31 1 2 3 4 6 7 8 9 11 12 13 14 16 17 18 19 21 22 23 24 26 27 28 29 31 Day Day CO Mean Daily Concentrations Concentration (ppm) 2. 1.8 1.6 1.4 1.2 1..8.6.4.2. 1 2 3 4 6 7 8 9 11 12 13 14 16 17 18 19 21 22 23 24 26 27 28 29 31 Day Fig.. Forecasted and observed mean daily concentrations of O 3, NO 2 and CO over Thessaloniki for December.
O 3 Mean Hourly Concentrations NO 2 Mean Hourly Concentrations 4 1 2 3 4 6 7 8 9 11 12 13 14 16 17 18 19 21 22 23 24 1 2 3 4 6 7 8 9 11 12 13 14 16 17 18 19 21 22 23 24 Hour Hour 1.8 1.6 CO Mean Hourly Concentrations Concentration (ppm) 1.4 1.2 1..8.6.4.2. 1 2 3 4 6 7 8 9 11 12 13 14 16 17 18 19 21 22 23 24 Hour Fig. 6. Forecasted and observed mean diurnal variation of O 3, NO 2 and CO over Thessaloniki for December. 4 CONCLUSIONS An operational air quality forecast system is set up in order to issue 72-hour forecasts of near surface air pollutants levels over Greece. The system has been evaluated by Greek authorities (the Region of Central Macedonia). The forecast system is capable of reproducing the main features of air quality over the area where observations were available. ing system performance is better regarding O 3 and NO 2 while CO concentrations are systematically underestimated. More reliable is the forecast issued during days when it is not raining. Future forecasting activities would greatly benefit from an updated emission inventory (mainly for CO). Finer modeling resolution, mainly over the areas where large urban centers and emission sources are located, is needed in order to reduce the existing discrepancies between forecasted and observed pollutant levels. REFERENCES 1. EEA, 1997. CORINAIR 94 Summary Report, Final Version. European Topic Centre on Air Emissions, European Environmental Agency. 2. EMEP/CORINAIR Emission Inventory Guidebook - 3rd edition, 2. European Environment Agency. 3. EMEP/MSC-W, 2. Emission data reported to UNECE/EMEP: Quality assurance and trend analysis & Presentation of WebDab. MSC-W Status Report 2, EMEP/MSC-W NOTE 1/2. 4. ENVIRON,. User s guide CAMx - Comprehensive Air Quality with extensions, Version 4.. ENVIRON International Corporation, 4.899.7, June.. Grell, G. A., J. Dudhia and D. R. Stauffer, 1994. A description of the fifth-generation Penn State/NCAR mesoscale model (MM). NCAR Technical Note, NCAR/TN-398+STR. 6. Madronich, S. 1993. UV radiation in the natural and perturbed atmosphere, in Environmental Effects of UV (Ultraviolet) Radiation (M. Tevini, ed.), Lewis Publisher, Boca Raton. 7. Poupkou, A., Symeonidis, P., Lisaridis, I., Pouspourika, E., Yay, O.D., Melas, D., Ziomas, I., Balis, D. and Zerefos, C., 4. Compilation of an emission inventory for the purpose of studying the regional photochemical pollution in the Balkan Region. In the Proceedings of the Quadrennial Ozone Symposium 4, Kos, Greece, pp. 92-93. 8. Symeonidis, P., Ziomas, I. and Proyou, A., 4. Development of an emission inventory system from transport in Greece. Environmental ling & Software, Vol. 19, Num 4, 413-421.