Development of a GIS-based decision support system for urban air quality management in the city of Istanbul

Atmospheric Environment 44 (2010) 441e454

Contents lists available at ScienceDirect

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

Development of a GIS-based decision support system for urban air quality management in the city of Istanbul Tolga Elbir a, *, Nizamettin Mangir b, Melik Kara a, Sedef Simsir a, Tuba Eren a, Seda Ozdemir b a b

Dokuz Eylul University, Department of Environmental Engineering, Tinaztepe Campus, Izmir 35160, Turkey Istanbul Metropolitan Municipality, Directorate of Environmental Protection, Istanbul, Turkey

a r t i c l e i n f o

a b s t r a c t

Article history: Received 20 May 2009 Received in revised form 12 October 2009 Accepted 6 November 2009

A decision support system has been developed for urban air quality management in the metropolitan area of Istanbul. The system is based on CALMET/CALPUFF dispersion modeling system, digital maps, and related databases to estimate the emissions and spatial distribution of air pollutants with the help of a GIS software. The system estimates ambient air pollution levels at high temporal and spatial resolutions and enables mapping of emissions and air quality levels. Mapping and scenario results can be compared with air quality limits. Impact assessment of air pollution abatement measures can also be carried out. Ó 2009 Elsevier Ltd. All rights reserved.

Keywords: Air quality management Decision support system Air quality modeling Emission inventory Geographical information system

1. Introduction Many cities around the world, particularly in developing countries, are experiencing rapid growth. Yet, in the absence of adequate environmental policies, this growth is occurring at a considerable, and often increasing, economic and social cost. More people, more industry, and more motor vehicles cause ever-worsening air quality that poses a serious environmental threat in many cities. Aside from its severe local effects, urban air pollution has profound regional and global impacts. Urban emissions are major contributors to the problems of ozone layer depletion and ground level ozone, global warming and climate change (Dyominov and Zadorozhny, 2005; Jenkin and Clemitshaw, 2000; Ramanathan and Feng, 2009; Alcamo et al., 2002; Jacob and Winner, 2009; Vautard and Hauglustaine, 2007; Kinney, 2008). Urban air pollution also causes respiratory diseases (Selgrade, 2000; Devalia et al., 1994; Schwartz, 1993; Lee et al., 2007; Jalaludin et al., 2004; Boezen et al., 1999). To deal with these problems at the regional level, the air quality in cities have to be monitored and managed. An effective environmental planning and management process helps decision makers to formulate and implement realistic and effective strategies and action plans to improve air quality. These strategies and action plans have to systematically address the short

* Corresponding author. Tel.: þ90 232 4127133; fax: þ90 232 4530922. E-mail address: [email protected] (T. Elbir). 1352-2310/$ e see front matter Ó 2009 Elsevier Ltd. All rights reserved. doi:10.1016/j.atmosenv.2009.11.008

and long-term causes of urban air pollution and help the city to achieve a sustainable growth pattern. Many computer based decision support systems for urban air quality management are applied in major cities around the world. The application of decision support systems is an opportunity for improving air quality planning in large cities. Decision support systems generally include emission inventories, air quality monitoring, modeling, mapping, and air quality impact assessment of various control strategies. They support the evaluation of action plans by using information to the public about past and present air quality levels. Examples of decision support systems used by the local authorities in major European cities such as Stockholm, Lisbon, Milano, Berlin, Geneva, Vienna, Paris, Oslo and Athens are the Swedish AirViro (SMHI, 2009), the Austrian AirWare (Fedra and Haurie, 1999), the Norwegian AirQUIS (Bohler et al., 2002) and the Swedish EnviMan (Tarodo, 2003) systems. A few more decision support systems for specific air quality management studies are in operation around the world (Guerrero et al., 2008; Lim et al., 2005; Elbir, 2004; Puliafito et al., 2003; Fine and Ambrosiano, 1996; Schmidt and Schafer, 1998; Jensen et al., 2001; Lin and Lin, 2002; Finzi et al., 1991). Available air quality decision support systems like Airviro, AirWare, AirQuis and EnviMan generally apply simple air quality dispersion models, they have low spatial resolution that does not conform with the high spatial variability of characteristics in urban areas, and they only consider simple exposure assessment if any (Jensen, 1999). Further, they are commercial systems developed to

442

T. Elbir et al. / Atmospheric Environment 44 (2010) 441e454

Fig. 1. Locations of industrial zones, monitoring and meteorological stations in the study area.

be marketed in many countries and they certainly do not take into consideration the local conditions (i.e., specific fuels, industrial activities, and vehicle types/ages) within their administrative databases. Istanbul is one of the world's biggest cities with w12 million population. Air pollution in Istanbul is one of the important problems of modern life due to rapid population growth, dense immigration, improper site selection for industry, usage of poor quality fuels, usage of old combustion technologies in industry, lack of control technologies for stack gases, and traffic emissions (Incecik, 1986; Bozyazi et al., 2000; Unal et al., 2000; Sumen et al., 2005). To identify the causes of this problem, and to improve the air quality, a geographical information system (GIS) based decision support system has been developed for the city of Istanbul. Preparation of a comprehensive emission inventory, air quality

modeling, air quality mapping by GIS, and scenario analysis for air pollution abatement were carried out as the components of this urban air quality management system. 2. Characteristics of the study area The city of Istanbul is located in the north-west Marmara region of Turkey, in the coordinates of 28 100 and 29 400 East longitudes and 40 500 and 41300 North latitudes (Fig. 1). The city with a surface area of 5313 km2 (TUIK, 2008) has three neighboring provinces (i.e., Kocaeli in the east, Bursa in the south and Tekirdag in the west). Marmara Sea at south and Black Sea at north surround the city besides these provinces. The Bosphorus, connects the Black Sea with Marmara Sea and divides the city of Istanbul into two parts, and also separates the Continents of Europe and Asia.

T. Elbir et al. / Atmospheric Environment 44 (2010) 441e454

The Black Sea in the north and Marmara Sea in the south produce a differential heating of surfaces, leading to different meteorological conditions that may play a role in the transport of pollutants (Im et al., 2008). During summer anti-cyclonic conditions, local circulation is completely driven by thermal contrasts between land and Marmara Sea and The Black Sea (Anteplioglu et al., 2007). The complex terrain of the city also influences the circulation systems. The pressure patterns over the city and surrounding land area are known to be controlled by the Eurasiatic high-pressure axis, the subtropical high and its extension over North Africa, and the belt of low pressure over the Mediterranean Sea (Topcu et al., 2003). Istanbul has a typical Mediterranean climate with hot summer and mild-rainy winter. In summer, the weather in Istanbul is hot and humid, the temperature between June and September averaging as 24  C. Summers are relatively dry, but rain occurs all year round. The annual average rainfall is 718 mm for the past 50 years while highest annual rainfall was 943 mm in 1980. The city is subject to moisture laden mid-latitude cyclones during winter months; thus most of the precipitation falls in winter (Ezber et al., 2007). During winter the weather is cold, wet and occasionally snowy. Snowfalls tend to be heavy at times, but temperatures rarely drop below the freezing point. The maximum monthly average temperature in the city is 38.4  C in June, while the minimum monthly temperature is 1.9  C in February. Istanbul also tends to be a windy city. The annual mean wind speed is 3.5 m s1 while the predominant wind directions are: NNE, 34.3%; NE, 24.0%; SSW, 11.1% and N, 10.8% for the year 2007. The wind roses of different stations in the city are given in Supplementary Material (Figure S1). The hourly mixing heights are generally between 50 and 945 m in the city. The monthly variations of average hourly mixing heights estimated in the main station of the city (Goztepe) in this study are given in Supplementary Material (Figure S2). There are twenty-five surface meteorological stations and one upper air station in the city and its surroundings. The main meteorological station (Goztepe) located at the centre of city is an upper air and surface station (Fig. 1). The list of meteorological stations, their coordinates and the source of data is given in Supplementary Material (Table S1). More details about the meteorological data for the months of the year 2007 in twenty-five stations can be found in Supplementary Material (Tables S2eS6). Istanbul is one of largest industrial centers in Turkey. It employs approximately 20% of Turkey's industrial labor and contributes 38% of Turkey's industrial workspace (ITO, 2009). Food processing, textile production, oil products, rubber, metal-ware, leather, chemicals, pharmaceuticals, electronics, glass, machinery, automotive, transport vehicles, paper and paper products, and alcoholic drinks are among the city's major industrial products. Larger industrial facilities are usually agglomerated in the four organized industrial zones (Ikitelli, Dudullu, Beylikduzu and Tuzla). A big public power and cogeneration plant, a big cement plant and several sand and gravel processing plants are present in the city as the major pollutant sources. Fig. 1 shows the map of the study area including the main industrial zones. The city has several neighboring industrial regions, i.e., Kocaeli, Bursa and Tekirdag, that are major pollutant sources in the region (Esen et al., 2005; Karademir, 2006; Cetin et al., 2007). Metropolitan Municipality Directorate of Environmental Protection has monitored the urban air quality in Istanbul since 1998 in 10 monitoring stations that were located at various sites considering the topography. The locations of the current monitoring stations are illustrated in Fig. 1. These stations continuously measure CO, NOX, SO2, PM10, and O3 that are reported on hourly basis. Air pollution experienced in Istanbul has reached to significant levels since 1980s. There have been several studies monitoring the

443

air quality in the different sites of the city (Sen, 1998; Tayanc, 2000; Topcu et al., 2003; Akkoyunlu and Erturk, 2003; Karaca et al., 2005; Ercelebi and Toros, 2009). It was reported that sulfur dioxide concentrations reach to high levels during the residential heating season due to the consumption of low-quality lignites having high sulfur contents (Tayanc, 2000). Results of a previous study indicated that during the 1985e1991 period there was an increasing trend in SO2 concentrations in the city. However, the increased usage of natural gas and high-quality coal in the residential areas have significantly improved the air quality in Istanbul (Akkoyunlu and Erturk, 2003). Sen (1998) have applied an estimation method for SO2 and PM10 concentrations at 16 different sites, using the values recorded at urban/suburban air pollution measurement sites of the city. Another study investigated the relationships between PM2.5, PM2.5e10 and PM10 data, and determined the frequency distribution characteristics of PM2.5 and PM10 (Karaca et al., 2005). Air quality measurements were also evaluated using the climatological conditions to analyze the air pollution episodes and air pollution potential in the city (Topcu et al., 2003). The atmospheric circulation over central Europe was dominated by a cold-core surface anti-cyclone, a climatologically favored feature during cold seasons. Air trajectories associated with this anti-cyclone were capable of transporting high concentrations of pollutants over long distances to Istanbul (Kindap, 2008). Several studies have also examined the transport of pollutants from Europe to Turkey (Kindap et al., 2006; Kindap, 2008). These studies indicate that Eastern Europe has been responsible for some pollution events in Istanbul. 3. System architecture 3.1. System overview Geographical Information System (GIS) is used as the main component of the system for capturing, storing, checking and manipulating data that are spatially referenced. The ArcGIS application developed by ESRI was used because of its relative user friendly structure and its generalized use by local authorities and research institutes. This software is also well suited to develop dynamic environmental models. In this software, a particular display of the different shapes (industries, houses and roads) is called themes and they can be selected in any order, e.g. localization of industries, emission patterns, etc. These themes can be selected or sorted according to the user criteria, highlighting the most relevant features on individual digital maps. It is important to note that GIS is not used only as a map viewer in the system, but rather as an integrated tool to handle data from many sources. Fig. 2 illustrates the principal components and steps for the decision support system developed. This system was developed following the common methodology described and used by several recent studies (Guerrero et al., 2008; Lim et al., 2005; Elbir, 2004; Puliafito et al., 2003; Jensen et al., 2001; Lin and Lin, 2002; Finzi et al., 1991). First step is data collection. It is necessary to collect the industrial source specific information on production capacities, raw materials used, manufacturing processes, fuel consumptions, stack characteristics (i.e., stack height, diameter, flue gas temperature and exit gas velocity); the domestic heating source information like number of inhabitants, types of fuels used, fuel consumptions and population densities; and the traffic source information like numbers, types and ages of vehicles, fuels used and fuel consumption. Next, the emission factors are used to prepare an emission inventory, that is computed and stored in GIS databases. In order to use a generic GIS-based emission estimation methodology, readily available data describing the locations of pollutant sources and associated rates of polluting activities in all

444

T. Elbir et al. / Atmospheric Environment 44 (2010) 441e454

measurements, then it is necessary to return to the first step and check for errors in the estimation of the relevant parameters or perform the necessary corrections in the calculations.

Industry Residential heating Traffic

Input data

3.2. Emission inventory

Emission factors

Emission Inventory

Meteorology

Air Quality Modeling (CALMET/CALPUFF)

Topography

Calibration (Select points coincident with monitoring stations)

Compare

Monitoring sites Not satisfactory Acceptable

Simulation (Prepare output maps, Compute new scenarios)

Fig. 2. Structure of the system developed.

administrative areas in the city are assembled in the form of spreadsheet files. After calculation of emissions in these spreadsheets using appropriate emission factors, the emissions are linked together with required map layers using ArcInfo's Macro Language. The structure of the inventory within the GIS follows the vector data model that uses point, line and area features to represent emission sources. The next step in the system is the use of CALMET/CALPUFF modeling system (Scire et al., 2000) to predict urban air quality levels. Implementing air quality models under GIS environment is one of the strong features of the decision support system. GIS techniques are capable of providing geospatial air quality models, i.e., at anytime and any location anyone can access the air quality in the area (Jensen, 1999). In this system, the connection between the input data (i.e., meteorological data and emission inventory) and the models is provided by several computer programs written in Visual Basic programming language. Besides, the input and output data associated with the modeling studies are linked together with the map layers using ArcInfo's Macro Language. Once the modeling system is calibrated, then different scenarios can be simulated in the decision support system developed (Fig. 2). If no acceptable match is obtained between calibration and

In a systematic way, the emission sources are broadly categorized as point, line and area sources, covering industrial, vehicular and residential sources, respectively. The biogenic and natural emissions could not be included in the study due to the lack of data. Four major pollutants consisting of particulate matter (PM10), sulfur dioxide (SO2), carbon monoxide (CO) and nitrogen oxides (NOX) emitted through these sources were identified. The emissions were assumed as ground based and were calculated using fuel consumption data and appropriate emission factors. Emission factors were taken from European CORINAIR database (CITEPA, 1992) and US Environmental Protection Agency (USEPA) emission factors catalogue (USEPA, 1995). Whenever European emission factors were insufficient to indicate the industrial subcategories, EPA emission factors were used. Detailed information on the emission factors used in the present study can be found elsewhere (Elbir and Muezzinoglu, 2004). The contributions from natural sources, industrial process (fugitive), on-road fugitive and non-road vehicular emissions were not included in these calculations. Stacks with emissions higher than certain amounts were classified as point sources and included in the inventory. In the study area 1025 industrial plants contributing to air pollution were identified and they were included in the emission inventory. Also, many industrial facilities (n ¼ 12 432) including light industry and commercial areas are located in four major organized industrial zones (OIZ) (Ikitelli OIZ, Dudullu OIZ, Beylikduzu OIZ and Tuzla OIZ) in the study area. Two hundred and twenty-five of them are identified as major industrial plants and their emissions were calculated in this study. Emissions from sources, too small and difficult to be surveyed individually, were considered collectively as area sources. Therefore, residential heating sources constitute area sources. They were evaluated and allocated on the grid system with respect to population density in the study area. The collected information for the calculation of residential heating emissions, included mainly number of inhabitants, number of residences, types of fuels used, fuel consumption statistics, and combustion characteristics. The fuel use pattern in the city is generally controlled by population density and income level of the inhabitants. 3.3. Air quality modeling The CALMET/CALPUFF modeling system (Scire et al., 2000) was used to calculate the dispersion of pollutants from their sources. The CALMET/CALPUFF has been adopted by the United States Environmental Protection Agency (EPA) in its Guideline on Air Quality Models (USEPA, 2005) as a preferred model for assessing long range transport of pollutants and their impacts. The CALMET/ CALPUFF modeling system includes three main components: CALMET, CALPUFF and post processing and graphical display programs. CALMET is a diagnostic meteorological model that generates mass consistent wind fields over complex terrain. The CALMET meteorological model in its basic form produces hourly fields of three-dimensional winds and various micrometeorological variables based on the input of routinely available surface and upper air meteorological observations. CALPUFF is a Lagrangian puff model and a multi-layer, gridded non-steady-state puff dispersion model that can simulate the effects of temporally and spatially varying meteorological conditions on pollutant transport, removal by dry and wet deposition processes, and transformation

T. Elbir et al. / Atmospheric Environment 44 (2010) 441e454

445

Fig. 3. (a) Simulated SO2 emission distribution in the city of Istanbul for the year 2007, t y1 (b) Simulated annual average SO2 concentration distribution in the city of Istanbul for the year 2007, mg m3.

446

T. Elbir et al. / Atmospheric Environment 44 (2010) 441e454

Fig. 4. (a) Simulated PM10 emission distribution in the city of Istanbul for the year 2007, t y1 (b) Simulated annual average PM10 concentration distribution in the city of Istanbul for the year 2007, mg m3.

T. Elbir et al. / Atmospheric Environment 44 (2010) 441e454

447

Fig. 5. (a) Simulated NOX emission distribution in the city of Istanbul for the year 2007, t y1 (b) Simulated annual average NOX concentration distribution in the city of Istanbul for the year 2007, mg m3.

448

T. Elbir et al. / Atmospheric Environment 44 (2010) 441e454

Fig. 6. (a) Simulated CO emission distribution in the city of Istanbul for the year 2007, t y1 (b) Simulated annual average CO concentration distribution in the city of Istanbul for the year 2007, mg m3.

T. Elbir et al. / Atmospheric Environment 44 (2010) 441e454

449

calculations, it was assumed that the background concentrations are negligible. Therefore, the results represent the air quality levels originated only from the sources located in the modeling domain. 3.4. Model evaluation

Fig. 7. The emission contributions of each pollutant sector for the year 2007.

through chemical reactions. The model is developed to simulate continuous puffs of pollutants being emitted from a source into the ambient wind flow. As the wind flow changes from hour to hour, the path of each puff takes changes to the new wind flow direction. Puff diffusion is Gaussian and concentrations are based on the contributions of each puff as it passes over or near a receptor point (Scire et al., 2000). CALPUFF model requires each source to be described according to the emission inventory: stack dimensions, stack-gas exit temperature, emission flow and velocity. CALMET model requires local meteorological data such as hourly surface observations of wind speed, wind direction, temperature, cloud cover, ceiling height, surface pressure and relative humidity. The output of the model is then calculated for each grid of the study area. Two different meteorological modeling techniques, i.e., prognostic and diagnostic modeling are mainly used in literature. Both modeling techniques have their advantages and disadvantages. The use of a diagnostic model like CALMET could be more representative for the modeling domains including urban centers like Istanbul. Because, the diagnostic models like CALMET have efficient parameterization schemes that account for the deflection of streamlines over topographical features, blocking of flows by topographical features during stable stratified conditions, and slope flows. In general, these features do a reasonable job of simulating surface air flow features around smaller-scale topography that were not captured by a prognostic simulation. The study area was selected as 170 km  85 km to cover the Istanbul metropolitan area (Fig. 1). CALMET uses an interpolation scheme that allows observed wind data to be heavily weighted in the vicinity of the meteorological stations. Due to the existing meteorological stations (n ¼ 7) outside the metropolitan area, the modeling domain for CALMET was expanded to an area of 250 km  140 km to provide more representative wind field data for CALPUFF. The grid size used for estimating the grid-based emissions and concentrations is 1 km  1 km. This extended modeling domain required a regional scale model like CALPUFF taking into account the three-dimensional wind fields and other boundary layer parameters. CALMET land use categories and associated geophysical parameters based on the U.S. Geological Survey Land Use Classification System were used for the study area. The default land use categories and the default values of several geophysical parameters such as surface roughness length (i.e., 0.001 m for water body and 1.0 m for urban land), albedo, bowen ratio, soil heat flux parameter and heat flux can be found elsewhere (Scire et al., 2000). Using the CALMET/CALPUFF modeling system in the present study, the contribution of pollutant sources to the air quality in Istanbul was investigated in detail. In modeling

CALPUFF model performance was tested by comparing the predicted pollutant concentrations with those measured at ten ambient air quality stations. Mainly two methods were used: root of the mean square error (RMSE) and an index of agreement (d). The index of agreement determines the degree to which magnitudes and signs of the observed value about mean observed value are related to the predicted deviation about mean predicted value, and allows for sensitivity toward difference in observed and predicted values as well as proportionality changes. RMSE summarizes the difference between the observed and predicted concentrations. More information about these methods can be found elsewhere (Elbir, 2003). 4. Results and discussion The emission patterns and distributions of air pollutants in the city of Istanbul were studied for the year of 2007. The results show that industry is the most polluting sector for SO2 contributing to about 83% of total emissions while residential heating is the most polluting sector for PM10 contributing to 52% of total emissions. Traffic is the most polluting sector for NOX and CO emissions with the contributions of 89% and 68%, respectively, in the study area. When the emissions of individual industries were examined, a public power and cogeneration plant was found to be the largest source of air pollution in the study area due to its high fuel oil-6 consumption. It contributed by 84% to the total industrial SO2 emissions and by 70% to the overall SO2 emissions. Several sand and gravel processing plants and one big cement plant (having the highest PM10 emission) are the major contributors to PM10 emissions in the study area. Several different dust control equipments varying from electrostatic precipitators to fabric filters are used in the cement plant depending on the regulatory requirements. Therefore, control technologies and their efficiencies were taken into account in calculating the emissions from large facilities, such as this cement plant. Model predictions show that in terms of SO2 the most polluted areas are generally around the power plant. The maximum annual average SO2 concentration (w175 mg m3) during the year 2007 occurs near the plant in the city center. Fig. 3 shows that 99% of this concentration comes from industrial plants. This highest concentration occurs at the coordinates of (639 500, 4 538 500) in the grid system. The concentrations within the range of 20e50 mg m3 occur around the organized industrial zones. As for the domestic contribution, several districts have a similar highest annual average concentration of approximately 5 mg m3. For high PM10 concentrations in the metropolitan area, the cement plant and sand-gravel processing plants in Gaziosmanpasa region are the main sources. The maximum annual average PM10 concentration of approximately 90 mg m3 during the year 2007 occurs in Gaziosmanpasa region (Fig. 4). South part of Buyukcekmece is found as the second polluted area for PM10 due to presence of the big cement factory in the vicinity. These two activities cover 18% of overall PM10 emissions and 63% of industrial PM10 emissions in the city. The most polluting sector for CO and NOX is the traffic in the city. The maximum annual average NOX and CO concentrations are 1610 and 3140 mg m3 that occur near highways. Figs. 5 and 6 show that 99% of these concentrations are contributed from vehicles. This highest concentration occurs at the coordinates of (664 500, 4 542 500) in the grid system. Several major industries also cause

450

Table 1 The comparison between simulated and monitored daily average concentrations. Stations

Overall

Aksaray Pre SO2

PM10 Number of valid observations Mean (mg m3) Maximum (mg m3) Standard deviation (mg m3) Correlation coeffcient (R) Index of agreement (d) RMSE NOX

CO

Number of valid observations Mean (mg m3) Maximum (mg m3) Standard deviation (mg m3) Correlation coeffcient (R) Index of agreement (d) RMSE Number of valid observations Mean (mg m3) Maximum (mg m3) Standard deviation (mg m3) Correlation coeffcient (R) Index of agreement (d) RMSE

51

Besiktas

Pre

Pre

Obs

287

10.8 34.6 7.8

12.6 27.0 5.2

0.37 0.60 0.08

12.1 119.1 14.2

11.6 55.0 11.4

84.6 151 30.7

0.46 0.38 0.79

22.6 105 18.3

68.1 205 24.0

143 373 328 1177 76.4 232.6

0.62 0.53 1.85

90 485 80.4

0.70 0.75 6.40

() No data, Pre: prediction, Obs: observation.

859 3889 593 0.50 0.65 5.62

62.3 128 18.5

238 1960 302 0.33 0.41 6.62

10.3 69 12.7

102 291 46.1

97 503 84.7

76.5 182 25.4

0.45 0.58 5.22

Uskudar Pre

5.9 60.0 9.3

6.2 50.0 6.1

7.4 35 8.7

16.2 54.6 10.4

7.4 70 11.4 0.22 0.35 0.64

e e e

e e e 224

e e e

0.25 0.41 4.04

136 1138 150 0.57 0.46 4.57

10.3 61 10.4

e e e

e e e

143 526 117.6

64.5 172 28.1

0.39 0.53 7.84

4.4 29.1 6.0

3.4 31 5.2

783 2268 423

337 1232 268 0.58 0.54 7.02

15.3 11.2 55.0 175.0 11.3 19.1

80.6 266 36.8

9.7 83 15.7

11.5 44.7 9.2

e e e

36 198 43.6

60.1 137 23.8

124 503 106 0.62 0.46 8.10

164 847 194 0.81 0.56 6.71

10.5 105 13.9

69.0 266 27.9

1225 59 148 234 1177 36.1 172.5

91 499 65.1

0.36 0.44 1.71

47 819 2751 475

12.1 57.0 10.1

0.22 0.37 0.65

0.69 0.76 0.39

124

8.7 119.1 12.0

1463

0.22 0.37 0.56

e e e

Obs

0.34 0.58 0.13

181

94 e 422 e 72.0 e

Pre 2588

104

e

850 4225 585

Obs

0.52 0.67 0.13

0.47 0.38 0.85

234

Pre 175

277

0.67 0.71 1.00

895 3591 839

Umraniye Obs

0.39 0.53 0.15

235

219 532 2356 275

9.5 43.0 7.9

0.38 0.39 0.60

e e e

282 383 1489 246

60.0 127 21.7

e

e e e

6.7 64.3 9.6

Pre 281

116

0.43 0.50 0.37

e e e

57 465 81

29.9 69 16.8

Kartal Obs

0.43 0.65 0.10

45

e

88 e 499 e 71.3 e

10.1 57.0 10.2

0.31 0.57 0.12

61.5 189 25.1

Pre 258

8.7 56.5 10.4

201 61.9 145 23.3

Kadikoy Obs

278

0.38 0.60 0.14

e

630 2137 382

Obs

12.7 62.8 14.0

0.45 0.39 0.58

342 266 2177 322

Yenibosna Pre 294

123

0.51 0.70 0.78

806 2565 283

Obs

0.12 0.39 0.10

210

333 663 2766 393

13.5 53.0 9.9

0.50 0.35 0.70

0.30 0.50 0.88

277 1040 4777 693

83 397 87.9

9.7 80.7 12.2

Pre 271

184

342

0.36 0.30 3.56

53

10.1 73 12.7

Sariyer Obs

0.40 0.62 0.13

0.13 0.33 0.56

219

265 746 174.3

13.5 50.4 9.6

196

0.46 0.42 0.51

38

7.9 88.7 13.5

Pre 349

0.36 0.56 0.15

195

10.8 24 5.1

Esenler Obs

344

0.27 0.51 0.16

22

648 3126 597

Alibeykoy

2135 698 2651 552

371 3889 508 0.39 0.57 6.10

686 4777 428

T. Elbir et al. / Atmospheric Environment 44 (2010) 441e454

Number of valid observations Mean (mg m3) Maximum (mg m3) Standard deviation (mg m3) Correlation coefficient (R) Index of agreement (d) RMSE

Obs

T. Elbir et al. / Atmospheric Environment 44 (2010) 441e454

451

Fig. 8. Time-series plots of observed and predicted daily a) NOX and b) CO concentrations, mg m3.

an annual average NOX concentration of 50 mg m3 in Buyukcekmece region. The contribution of residential heating sector is very small (w4 mg m3). Figs. 3e6 demonstrate the distribution maps for the estimated emissions and the concentration distributions simulated by CALPUFF for PM, SO2, CO and NOX. According to the estimated results of the grid-based emissions mentioned above, the emission contributions of each source sector were summarized in Fig. 7. Data from a recent study (Unal et al., 2009) conducted by Istanbul Metropolitan Municipality in collaboration with the World Resources Institute Center for Sustainable Transport (EMBARQ) in 2007 was used to estimate the traffic emissions in the metropolitan area. The technology distribution of vehicles was developed using a combination of two approaches. Vehicles were videotaped on a various streets. Then, the videotapes were examined to count the numbers of the various types of vehicles driving in Istanbul streets. Parking areas were also surveyed to collect specific technology information about vehicles operating in Istanbul. Comparison of average predicted and monitored concentrations show that overall accuracy of the predictions for all pollutants was high except PM10. The overall accuracies of NOX, CO and SO2 concentration predictions (index of agreement) were about 44%, 57% and 58% respectively while the accuracy was 37% for PM10 concentration prediction. When each monitoring station was evaluated separately, the prediction accuracies at Umraniye,

Aksaray, and Uskudar stations was better than overall accuracy with values of 76% for NOX, 75% for CO, 67% for SO2, and 50% for PM10. However, the accuracy in Sariyer station was found as 39% for PM10 and SO2, 41% for CO. The relevant analysis of the root of mean square error (RMSE) indicated that there was a normal error in the model prediction. Furthermore, the R values (max. 0.81) suggested a good model behavior in terms of calculating daily NOX concentrations. Table 1 shows the results of the statistical analysis. It should be noted that the uncertainty of the predictions might arise from two different sources: the emission calculations and dispersion modeling. The uncertainties of measurement techniques at the stations as well as improper site selection for the stations might also be important factors for getting a relatively lower relevance between predictions and actual measurements. Meteorological modeling might be another reason for poor agreement between the observed and predicted values by dispersion model. Gilliam et al. (2005) have evaluated the performance of CALMET in New York city (especially at the lower Manhattan area). It was reported that CALMET derived wind speeds had a bias as high as 1.5 m s1. Independent wind observations in Lower Manhattan have also suggested that in some cases the wind direction estimates of CALMET may be significantly different than the observed values because of the urban influence. However, overall the CALMET model was found to provide meteorology that is adequate for driving a plume model most of the time (Gilliam et al., 2005). In the present study, to test the meteorological model (CALMET)

452

T. Elbir et al. / Atmospheric Environment 44 (2010) 441e454

Fig. 9. Seasonal average SO2 concentrations in the winter for the scenario, mg m3.

T. Elbir et al. / Atmospheric Environment 44 (2010) 441e454

performance, the statistical analysis suggested by Gilliam et al. (2005) was applied and the differences between the wind speed and direction observations in Goztepe station and the CALMET model simulations were assessed. For this analysis, CALMET model was re-run excluding all surface meteorological data from Goztepe station. The statistics were separated according to the observed flow direction at the site. Each of four flow classifications covers a 90 sector that is rotated 30 with respect to north. The flow classifications are northerly (330 e60 ), easterly (60 e150 ), southerly (150 e240 ) and westerly (240 e330 ). The mean bias and mean absolute error were used as the indicators. Biases for wind directions were negative indicating that the CALMET wind direction is not rotated clockwise relative to the observed wind. This suggested that the urban area of Istanbul might have an important influence to the sea-breeze flow in the city. The average bias overall directions was about 0.3 m s1, significantly lower than the value reported by Gilliam et al. (2005) (1.5 m s1) indicating a good wind speed predictability for CALMET (Table S7). From these evaluations it can be concluded that for Istanbul, the CALMET satisfactorily predicts the meteorology for CALPUFF model. To further examine the CALPUFF model performance, two timeseries plots of observed and predicted daily NOX and CO concentrations for Kadikoy and Sariyer stations, respectively, are shown in Fig. 8 as the examples for a good and a poor agreement between the measured and predicted concentrations. The similar time series of pollutant concentrations can be easily obtained for user-selected receptor points from the model runs in the system. The model performed well in predicting the diurnal variations of the concentrations for NOx and CO. However, the model underestimated CO concentrations generally, whereas NOX concentrations were overpredicted. The substantial under-prediction of the concentrations can be attributed in part to the pollutant sources that could not be included in the emission inventory (i.e., marine transport, natural sources, and pollutant sources located outside the city, such as neighboring industrial zones). On the other hand, the overprediction can be attributed to some extent by the fact that the location of the station was very close to the pollutant sources like a street that was not modeled as a line source, since traffic emissions other than highways were modeled as area sources. The system developed in the present study can also be used to assess the impacts of different air quality management policies. For demonstration, this paper presents one scenario. As it is related to the fuel usage changes in residential areas of Istanbul, the main theme in the scenario is the reduction of natural gas consumption in houses. Turkey imports natural gas from its neighboring countries such as Russia, Azerbaijan, and Iran. Turkey has experienced natural gas outages several times in the past, as the supplier countries tend to severely reduce the amount of the gas they export when the domestic demand increases under specific conditions (i.e., periods of very cold weather). The aim of this scenario is to switch the fuel types in residential areas when a natural gas crisis occurs. The scenario assumes that lignite with a sulfur content of 1% will be used instead of natural gas in houses during the winter season. Shifting the fuel type in residential areas results in 4.1 times higher SO2 emissions from residential heating compared to the current emissions (10 983 t y1) for the winter season. The results of the scenario show that the overall seasonal SO2 concentrations are increased almost in all regions of the city during the winter season (Fig. 9). Maximum concentration increases from 20 mg m3 to 70 mg m3. This means that the concentrations after the fuel type shift scenario are approximately four times higher than the current ones. For reducing future air pollution in the city, industrial plants should take some precautions such as using cleaner fuels, using new pollution prevention technologies for production processes,

453

and improving their combustion technologies. Appropriate control technologies for harmful stack gases should be used in the industrial plants which are the highest polluters in the study area. First of all, the control technologies should be applied for the big power plant. Because, the model results show that almost half of annual average SO2 concentrations predicted at the metropolitan area comes from this plant alone. The use of cleaner fossil fuels such as natural gas should be extended in residential areas. Due to the use of lignite with high sulfur content in residential areas, critical air quality levels have occurred in winter season. The results presented in this paper are preliminary outcomes of ongoing studies. The system will be further developed, e.g. the determination of effects of air pollution on human health, vegetation, forest and materials, the modeling of other components such as ozone, other greenhouse gases and heavy metals. 5. Conclusions Decision support systems suitable to urban air quality management are not routinely applied in Turkey, although such systems are in operation in other European middle-sized and larger cities. However, present studies in Turkey often have a low spatial resolution and do not take the full advantage of GIS and administrative databases. A decision support system was developed for urban air quality management in the city of Istanbul. Calculation of a comprehensive emission inventory, air quality modeling, air quality mapping by GIS and scenario analysis for air pollution abatement were carried out as the components of this system. Air quality modeling was the main component of the system and CALPUFF, a Langrangian puff dispersion model was mainly used for air quality predictions. CALMET meteorological preprocessors were used to produce meteorological data for CALPUFF model in the system. The decision support system developed in this study provides easy access to the softwares to determine the air quality anywhere and anytime in the study area. The system has been also developed as a tool to have an idea about the results of a case causing changes on air quality within a reasonable run time. Acknowledgement This research was carried out with financial support from the LIFE06-TCY/TR/000283 project, funded by the European Commission within LIFE Third Countries program. Appendix. Supplementary material Supplementary data associated with this article can be found, in the online version, at doi:10.1016/j.atmosenv.2009.11.008. References _ Akkoyunlu, A., Erturk, F., 2003. Evaluation of air pollution trends in Istanbul. International Journal of Environment and Pollution 18, 388e398. Alcamo, J., Mayerhofer, P., Guardans, R., Harmelen, T., Minnen, J., Onigkeit, J., Posch, M., Vries, B., 2002. An integrated assessment of regional air pollution and climate change in Europe: findings of the AIR-CLIM project. Environmental Science & Policy 4, 257e272. Anteplioglu, U., Incecik, S., Topcu, S., 2007. An Investigation of local anthropogenic effects on photochemical air pollution in Istanbul with model study. In: Borrego, C., Norman, A.L. (Eds.), Air Pollution Modeling and Its Application, vol. XVII. Springer, US, pp. 20e28. Boezen, H.M., Van der Zee, C.S., Postma, D.S., Vonk, J.M., Gerritsen, J., Hoek, G., Brunekreef, B., Rijcken, B., Schouten, J.P., 1999. Effects of ambient air pollution on upper and lower respiratory symptoms and peak expiratory flow in children. The Lancet 353, 874e878. Bohler, T., Karatzas, K., Peinel, G., Rose, T., San Jose, R., 2002. Providing multi-modal access to environmental data-customizable information services for

454

T. Elbir et al. / Atmospheric Environment 44 (2010) 441e454

disseminating urban air quality information in APNEE. Computers, Environment and Urban Systems 26, 39e61. Bozyazi, E., Incecik, S., Mannaerts, C., Brussel, M., 2000. Analysis and mapping of air pollution using a GIS approach: a case study of Istanbul. In: Longhurst, J.W.S., Brebbia, C.A., Power, H. (Eds.), Air Pollution, vol. VIIII. Wessex Institute Technology, Cambridge, England, pp. 431e440. Cetin, S., Karademir, A., Pekey, B., Ayberk, A., 2007. Inventory of emissions of primary air pollutants in the city of Kocaeli, Turkey. Environmental Monitoring and Assessment 128, 165e175. CITEPA (Centre Interprofessional Technique de la Pollution Atmospherique), 1992. Corinair Inventory-Default Emission Factors Handbook. CITEPA, Paris, France. Devalia, J.L., Rusznak, C., Davies, R.J., 1994. Air pollution in the 1990s-cause of increased respiratory disease. Respiratory Medicine 88, 241e244. Dyominov, I.G., Zadorozhny, A.M., 2005. Greenhouse gases and recovery of the Earth's ozone layer. Advances in Space Research 35, 1369e1374. Elbir, T., 2003. Comparison of model predictions with the data of an urban air quality monitoring network in Izmir, Turkey. Atmospheric Environment 38, 2149e2157. Elbir, T., 2004. A GIS based decision support system for estimation, visualization and analysis of air pollution for large Turkish cities. Atmospheric Environment 38, 4509e4517. Elbir, T., Muezzinoglu, A., 2004. Estimation of emission strengths of primary air pollutants in the city of Izmir, Turkey. Atmospheric Environment 38, 1851e1857. Ercelebi, S.G., Toros, H., 2009. Extreme value analysis of Istanbul air pollution data. Clean 37, 122e131. Esen, F., Tasdemir, Y., Cindoruk, S.S., 2005. Evaluation of NOx and O-3 concentrations in the atmosphere of Bursa, Turkey. Environmental Forensics 6, 311e317. Ezber, Y., Sen, O.L., Kindap, T., Karaca, M., 2007. Climatic effects of urbanization in Istanbul: a statistical and modeling analysis. International Journal of Climatology 27, 667e679. Fedra, K., Haurie, A., 1999. A decision support system for air quality management combining GIS and optimization techniques. International Journal of Environment and Pollution 12, 125e146. Fine, S.S., Ambrosiano, J., 1996. The environmental decision support system: overview and air quality application. Environmental Applications, 152e157. Finzi, G., Zani, L., Zecchini, M., 1991. A decision support system for the design of air quality monitoring networks. Environmental Software 6, 55e60. Gilliam, R.C., Childs, P.P., Hubar, A.H., Raman, S., 2005. Metropolitan-scale transport and dispersion from the New York World Trade Center following September 11, 2001. Part I: an evaluation of the CALMET meteorological model. Pure and Applied Geophysics 162, 1981e2003. Guerrero, P.J., Jorba, O., Baldasano, J.M., Gasso, S., 2008. The use of a modelling system as a tool for air quality management: annual high-resolution simulations and evaluation. Science of the Total Environment 390, 323e340. Im, U., Tayanc, M., Yenigun, O., 2008. Interaction patterns of major photochemical pollutants in Istanbul, Turkey. Atmospheric Research 89, 382e390. Incecik, S., 1986. Investigation of atmospheric conditions in Istanbul leading to air pollution episodes. Atmospheric Environment 30, 2739e2749. Istanbul Chamber of Commerce (ITO), 2009. Web site of ITO. http://www.ito.org.tr. Jacob, D.J., Winner, D.A., 2009. Effect of climate change on air quality. Atmospheric Environment 43, 51e63. Jalaludin, B.B., O'Toole, B.I., Leeder, S.R., 2004. Acute effects of urban ambient air pollution on respiratory symptoms, asthma medication use, and doctor visits for asthma in a cohort of Australian children. Environmental Research 95, 32e42. Jenkin, M.E., Clemitshaw, K.C., 2000. Ozone and other secondary photochemical pollutants: chemical processes governing their formation in the planetary boundary layer. Atmospheric Environment 34, 2499e2527. Jensen, S.S., 1999. A Geographic Approach to Modelling Human Exposure to Traffic Air Pollution Using GIS. Ph.D. thesis, National Environmental Research Institute, Roskilde, Denmark, 165p. Jensen, S.S., Berkowicz, R., Hansen, H.S., Hertel, O., 2001. A Danish decision-support GIS tool for management of urban air quality and human exposures. Transportation Research Part D: Transport and Environment 6, 229e241.

Karaca, F., Alagha, O., Erturk, F., 2005. Statistical characterization of atmospheric PM10 and PM2.5 concentrations at a non-impacted suburban site of Istanbul, Turkey. Chemosphere 59, 1183e1190. Karademir, A., 2006. Evaluation of the potential air pollution from fuel combustion in industrial boilers in Kocaeli, Turkey. Fuel 85, 1894e1903. Kindap, T., Unal, A., Chen, S.H., Hu, Y., Odman, M.T., Karaca, M., 2006. Long-range aerosol transport from Europe to Istanbul, Turkey. Atmospheric Environment 40, 3536e3547. Kindap, T., 2008. Identifying the trans-boundary transport of air pollutants to the city of Istanbul under specific weather conditions. Water, Air & Soil Pollution 189, 279e289. Kinney, P.L., 2008. Climate change, air quality, and human health. American Journal of Preventive Medicine 35, 459e467. Lee, J., Son, J., Cho, Y., 2007. The adverse effects of fine particle air pollution on respiratory function in the elderly. Science of the Total Environment 385, 28e36. Lim, L.L., Hughes, S.J., Hellawell, E.E., 2005. Integrated decision support system for urban air quality assessment. Environmental Modelling & Software 20, 947e954. Lin, M.D., Lin, Y.C., 2002. The application of GIS to air quality analysis in Taichung city, Taiwan, ROC. Environmental Modelling and Software 17, 11e19. Puliafito, E., Guevara, M., Puliafito, C., 2003. Characterization of urban air quality using GIS as a management system. Environmental Pollution 122, 105e117. Ramanathan, V., Feng, Y., 2009. Air pollution, greenhouse gases and climate change: global and regional perspectives. Atmospheric Environment 43, 37e50. Schmidt, M., Schafer, R.P., 1998. An integrated simulation system for traffic induced air pollution. Environmental Modelling and Software 13, 295e303. Schwartz, J., 1993. Particulate air pollution and chronic respiratory disease. Environmental Research 62, 7e13. Scire, J.S., Strimaitis, D.G., Yamartino, R.J., 2000. A User's Guide for the CALPUFF Dispersion Model. Earth Tech, Inc., Concord, MA. Selgrade, M.K., 2000. Air pollution and respiratory disease: extrapolating from animal models to human health effects. Immunopharmacology 48, 319e324. Sen, Z., 1998. An application of a regional air pollution estimation model over Istanbul urban area. Atmospheric Environment 32, 3425e3433. Sumen, M., Oktar, A., Mangir, N., Ozcelik, R., 2005. Actions taken in Istanbul towards the aim of preventing air pollution. In: Proceedings of the Third International Symposium on Air Quality Management at Urban, Regional and Global Scales. 26e30 September 2005, Istanbul, pp. 1418e1429. Swedish Meteorological and Hydrological Institute (SMHI), 2009. Web site of SMHI. http://www.smhi.se. Tarodo, J., 2003. Continuous emission monitoring. World Cement 34, 67e72. Tayanc, M., 2000. An assessment of spatial and temporal variation of sulfur dioxide levels over Istanbul, Turkey. Environmental Pollution 107, 61e69. Topcu, S., Incecik, S., Unal, Y.S., 2003. The influence of meteorological conditions and stringent emission control on high TSP episodes in Istanbul. Environmental Science & Pollution Research 10, 24e32. Turkish Statistical Institute (TUIK), 2008. Statistical Yearbook of Turkey. TUIK, p. 4. Unal, Y.S., Incecik, S., Borhan, Y., Mentes, S., 2000. Factors influencing the variability of SO2 concentrations in Istanbul. Journal of the Air & Waste Management Association 50, 75e84. Unal, A., Bulay, S., Barth, M., Lents, J., Mangir, N., Davis, N., Osses, M., Tolvett, S., Demir, I., 2009. The roadmap for clean air: a mobile source emissions inventory for Istanbul, Turkey. Journal of Air and Waste Management Under Review. USEPA (The United States Environmental Protection Agency), 1995. Compilation of Air Pollutant Emission Factors e Stationary Point and Area Sources. USEPA, Research Triangle Park, NC, AP-42. USEPA (The United States Environmental Protection Agency), 2005. Guideline on Air Quality Models. 40 CFR Part 51 Appendix W. USEPA. Vautard, R., Hauglustaine, D., 2007. Impact of global climate change on regional air quality: introduction to the thematic issue. Comptes Rendus Geosciences 339, 703e708.