THREE-DIMENSIONAL GROUND-BASED MEASUREMENTS OF URBAN AIR QUALITY TO EVALUATE SATELLITE DERIVED INTERPRETATIONS FOR URBAN AIR POLLUTION K. SCHÄFER1∗ , G. FÖMMEL1 , H. HOFFMANN1 , S. BRIZ1 , W. JUNKERMANN1 , S. EMEIS1 , C. JAHN1 , S. LEIPOLD1 , A. SEDLMAIER1 , S. DINEV2 , G. REISHOFER2 , L. WINDHOLZ2 , N. SOULAKELLIS3, N. SIFAKIS4 and D. SARIGIANNIS5 1 Institut für Meteorologie und Klimaforschung, Bereich Atmosphärische Umweltforschung,
Forschungszentrum Karlsruhe GmbH (former Fraunhofer-Institut für Atmosphärische Umweltforschung IFU), Garmisch-Partenkirchen, Germany; 2 Institut für Experimentalphysik, Technische Universität Graz, Graz, Austria; 3 Department of Geography, University of the Aegean, Mytilini, Greece; 4 Institute for Space Applications and Remote Sensing, National Observatory of Athens, Greece; 5 Institute for Health and Consumer Protection, EC-DG Joint Research Centre, Ispra (VA), Italy (∗ author for correspondence, e-mail:
[email protected], Fax: +49 8821 73573)
Abstract. Urban air quality and meteorological measurements were carried out in the region of Brescia (Italy) simultaneously to the acquisition of satellite data during winter and summer smog conditions in 1999. The main objectives of the campaigns were: delivery of data for the validation of air pollution interpretations based on satellite imagery, and determination of the aerosol optical thickness in spectral ranges similar to those used by satellites. During the winter campaign the ground-based network was complemented by local stations and by SODAR, DOAS, and FTIR remote sensing measurements. Size distributions of aerosol particles up to 4,000 m a.s.l. were measured by means of an ultra-light aircraft, which was also equipped with meteorological sensors and an ozone sensor. During the summer campaign an interference filter actinometer, an integrating nephelometer and an ozone LIDAR were operated additionally. The satellite images acquired and processed were taken from SPOT. Optical thickness retrieved from interference filter actinometer measurements were compared with the retrievals from the satellite imagery in the same spectral intervals. It is concluded that remaining aerosols in the reference image yield an off-set in the satellite retrieval data and that information about the vertical structure of the boundary layer is very important. Keywords: aircraft measurements, DOAS, FTIR spectrometry, in situ measurements, interference filter actinometer, LIDAR, remote sensing, satellite, smog, SODAR
1. Introduction Satellite images of high spatial resolution (i.e., horizontal resolution of tens of metres), such as taken by Landsat or SPOT, contain useful environmental information not only on land cover and vegetation but also on tropospheric pollution (Sifakis, 1992). This information, in terms of aerosol optical thickness (AOT), can be used as an indicator of air pollution’s density and spatial distribution over extended areas (e.g., Sifakis et al., 1998): The AOT is correlated to the particle mass Water, Air, and Soil Pollution: Focus 2: 91–102, 2002. © 2002 Kluwer Academic Publishers. Printed in the Netherlands.
92
K. SCHÄFER ET AL.
loading (Cachorro and Tanre, 1997) which, in the case of small-sized particles in urban and industrial areas, is due to ammonium sulphate and/or nitrate salts. These, in turn, are coupled with sulphuric acid and nitric acid droplets and consequently with their gaseous precursors NO2 and SO2 (Colls, 1997). Furthermore, aerosols are necessary reaction surfaces in the air for ozone production processes. The work presented here is part of the investigations aiming at defining the exact correlation between the atmospheric information retrieved by satellite (in terms of optical thickness) and air quality (in terms of pollutant concentrations). The algorithms and results are presented in Sifakis et al. (1999) and Sarigiannis et al. (2001). More specifically, this research is intended to find out if reliable information on air pollution distribution can be extracted on the basis of columnar atmospheric AOT values retrieved from high spatial resolution satellite data during cloud-free days. To examine this, two ground-based measurement campaigns were performed simultaneously with the satellite overpasses. The objectives of these campaigns were: 1. Delivery of necessary input data for modelling air pollution transport; 2. Delivery of data for the evaluation of air pollution interpretations based on satellite; 3. Determination of the AOT in spectral ranges similar to those used by the satellites; 4. Elaboration of a proposal for the optimisation of the existing air-pollution monitoring network in the city of Brescia on the basis of satellite imagery. The present article will concentrate on items 2 and 3. The background for these investigations is the EU Council Directive on ambient air quality assessment and management (96/62/EC), which recommends the use of operational instruments such as air pollution monitoring sensors, networks and modelling. The greater area of the city of Brescia in the region of Lombardy (Northern Italy) was selected as investigation area because it is a city in a high industrialised region with an air pollution monitoring network. The air quality in Brescia is influenced by advection of air from industrial areas north of Brescia, and sometimes by polluted air from Milano and Bergamo during west wind episodes an area with frequent smog episodes. Air pollution episodes in Alpine valleys in Northern Italy had already been investigated (Dosio et al., 2001). Brescia has approximately 200,000 inhabitants and a complex topography dominated by three main features: (1) in the south of the city there is flat terrain (the Po plain with heights around 120 m a.s.l.), (2) the city itself is situated at the end of the ‘Val Trompia’, a valley, which runs from north to south, (3) five kilometres north from Brescia the Nave valley branches off from the Val Trompia to the east. East and west of Brescia there are mountains (first mountains are Monte Maddalena (874 m) in the east and Monte Picastello (383 m) in the west). The altitude of the mountains in the northern part of Val Trompia is about 1,000 m a.s.l.
THREE-DIMENSIONAL GROUND-BASED MEASUREMENTS
93
2. Methodology Air pollution investigations were performed simultaneously with the acquisition of the satellite images in the greater Brescia area during winter (18 January–19 February, 1999) and summer (30 August–23 September, 1999). During winter several days with high air pollution due to stable atmospheric stratification occurred and during summer photochemical smog conditions happened. In addition to the standard measurements carried out systematically by the existing air pollution monitoring network of the city of Brescia operated by ASM Brescia, the following devices were used: • A Fourier Transform Infrared Spectrometer system (FTIR) addressing pathintegrated concentrations measurements of CO, CH4 , N2 O, CO2 , NO and O3 (Haus et al., 1994). • A system for Differential Optical Absorption Spectroscopy (DOAS) addressing path-integrated concentrations measurements of NO, NO2 , SO2 and O3 (Grant et al., 1992). • A Sound Detection and Ranging system (SODAR) measuring the vertical profiles of wind vector (Reitebuch and Emeis, 1998). • In situ CO (IFU: Horiba, ASM: Thermo Environmental Instruments), NO (IFU: winter Horiba, summer Monitor Labs, ASM: Thermo Environmental Instruments), NO2 (IFU: winter Horiba, summer Monitor Labs, ASM: Thermo Environmental Instruments), SO2 (IFU: Horiba, ASM: Thermo Environmental Instruments), O3 , (IFU: winter Horiba, summer Thermo Environmental Instruments, ASM: Thermo Environmental Instruments), CH4 (Horiba), dust (FAG Kugelfischer), PM10 (Sierra Andersen) and meteorological (wind, temperature, humidity, pressure, solar radiation) measurement instruments (Friedrichs, Thies, Ammonit, Zeno). • An interference filter actinometer measuring the AOT from the ground (Wiegner and Rabus, personal communication). • An integrating nephelometer measuring the aerosol scattering coefficient in sampled air (SCI). • A scanning Light Detection and Ranging system (LIDAR) measuring the vertical profiles of ozone concentration (Löscher and Windholz, 2001). • Finally, an ultra-light aircraft was operated to measure in situ ozone, meteorological parameters (temperature, solar radiation, dew point, pressure) and total extinction in 15 channels (Junkermann, 2001). The positions of all previous instruments at the ground were selected according to the criterion that as little as possible, mobile and fixed emission sources are in the surroundings of the measurement sites (see Figure 1). Further the instruments were installed at sites which accomplished the existing monitoring network.
94
K. SCHÄFER ET AL.
Figure 1. Survey of measurement sites during the campaigns in the region of Brescia in January/February 1999 and August/September 1999. 1: Waste water treatment Verziano; 2: Scuola Media Statale di Concesio; 3: water plant Fonte di Cogozzo; 4: ENEL substation between Bovezzo and Nave; 5: fountain ‘Pozzo Roma’, Collebeato.
THREE-DIMENSIONAL GROUND-BASED MEASUREMENTS
95
E.g. in the Nave valley there is no measurement station, therefore the devices measuring meteorological and air pollution data were installed in the Val Trompia (Pos. 2, 3 and 5 in Figure 1) as well as in the valley near Nave (Pos. 4 in Figure 1). Extensive information of the wind system over Brescia was obtained by a longrange SODAR (up to 800 m, vertical resolution 30 m), which was situated 5 km south of Brescia in the imaginary prolongation of the Val Trompia axis (Pos. 1 in Figure 1). During the summer campaign at the same site as the long-range SODAR in Verziano the dominant air pollutant ozone was characterised by a LIDAR giving vertical profiles of ozone concentrations and aerosol optical parameters from 100 m up to 2,000 m altitude (vertical resolution 100 m). Additionally, the AOT in 10 spectral ranges was measured from solar radiation by an interference filter actinometer at that site which was available during this campaign only to determine the total atmospheric aerosol content. The determination of the optical thickness of the atmosphere from the ground is done in similar spectral ranges as from satellite images. This kind of ground-truthing of space-based sensors was discussed by King et al. (1999) and is one of the goals of the AERONET (AErosol RObotic NETwork) program (internet http://aeronet.gsfc.nasa.gov:8080/) as well as the Multi-Filter Rotating Shadow-band Radiometer (MFRSR) network (Alexandrov et al., 2001). The FTIR, LIDAR and interference filter actinometer measurements were performed during appropriate weather conditions from the morning until the evening. All the other instruments were running. During each campaign an inter-comparison of all instruments at a place of well-mixed air was performed. If necessary, mean values among the measurement results of the different devices (in situ analyses of trace gas concentrations from air sampling and open-path spectroscopic measurements by FTIR and DOAS) were calculated to correct each data set. High spatial resolution SPOT satellite images, which were taken during the two campaigns at different days at about 10:00 a.m. local time were acquired and processed (see Sarigiannis et al., 2001). These satellite images covered an area of 60 km × 60 km.
3. Results and Discussion 3.1. S YNOPTIC SITUATION AND BOUNDARY- LAYER STRUCTURE The first days of the winter campaign were influenced by high pressure over Italy and surface inversions. Low temperatures in the evening and night (lower than 5 ◦ C) resulted in formation of strong fog. In the south of Brescia the air was highly polluted, resulting in bad visibility. A periodic wind system dominated in the Val Trompia and also in the valley of Nave: winds coming down-valley during the evening, night and morning during the coldest hours of the day (mountain breeze) and winds coming from southern directions during the warm hours of the day (valley breeze). So the northern parts of Brescia in the Val Trompia and valley
96
K. SCHÄFER ET AL.
Figure 2. Vertical profiles measured by the ultra-light aircraft at 18 February, 1999, 09:12–10:00 UTC: ozone concentration (ppb), extinction (1/100 km), potential temperature (◦ C), spread (i.e. temperature minus dewpoint temperature) (◦ C). Please note: The origin for spread has been shifted to 30 (vertical line). ‘n/s’ denotes neutral/stable, ‘s’ stable, and ‘vs’ very stable thermal stratification.
of Nave were influenced by air pollution coming from Brescia and the Po plain during the day. In the night, the air quality in Brescia was influenced by cold air flux coming from the Val Trompia. The Val Trompia and particularly the ancillary valleys (Lumezzane) are characterised by strong traffic and industry (metal fusion and treatment), so the cold air flux during the night was ‘injected’ with polluted air. The complex orographic structure results in a difficult vertical structure of the boundary layer with up to 5 sub-layers characterised by different thermal stability and different degrees of pollution as shown in Figures 2 and 3 containing the measured data of the ultra-light aircraft. The structure of the boundary layer above Brescia was more complex and extended up to greater heights during the summer campaign than during the winter campaign. The top of the boundary layer – indicated by less than 1 particle per cm3 (thin leftmost curve in Figures 2 and 3) – was around 1,500 m in February and 3,500 m in September. The long-range SODAR gave the valley/mountain wind regime structure in the boundary layer up to 800 m above ground as well as information about the layering of the atmosphere in this altitude range. During the aircraft flights the boundary layer height could be found on 18 February at about 1,250 m (above layer 2 in Figure 2) and on 19 February at 1,300 m. There was a frequent peak in the dust concentration between 08:00 and 10:00 a.m. At 4 February peak values in near-ground air pollution (3.5 mg m−3
THREE-DIMENSIONAL GROUND-BASED MEASUREMENTS
97
Figure 3. Vertical profiles measured by the ultra-light aircraft at 9 September 1999, 08:38–09:13 UTC: ozone concentration (ppb), extinction (1/100 km), potential temperature (◦ C), spread (◦ C). Please note: The origin for spread has been shifted to 30 (vertical line). ‘n/s’ denotes neutral/stable, ‘s’ stable, and ‘n’ neutral thermal stratification.
CO, 140 µg m−3 NO2 and 45 µg m−3 SO2 ) were reached nearby the city. Peak values of near-ground ozone concentration of about 100 ppb were found from 7 September until 15 September. During the aircraft flights from 5 until 9 September the boundary layer height could be found at about 3,200 m (5 September), higher than 3,200 m (6 and 7 September) and at 3,500 m (9 September) near the top of layer 5 shown in Figure 3. The scanning LIDAR measurements were processed to give vertical profiles of ozone concentration up to 1,700 m height with a vertical resolution of better than 100 m from the morning until the evening in the period 2 September until 21 September, i.e. delivering a temporal extension of the data from the aircraft flights and showing the spatial development of ozone concentration during the day. A comparison of ozone height profiles from measurements of the LIDAR and the ultra-light aircraft on 5 September is showing a typical deviation of ±10% (see Figure 4). 3.2. A EROSOL OPTICAL THICKNESS The AOT was measured together with the solar radiation by the ground-based interference filter actinometer. As the AOT is an integral quantity it is not effected
98
K. SCHÄFER ET AL.
Figure 4. Comparison of ozone profiles measured by LIDAR and ultra-light aircraft at 5 September, 1999.
by the complex structure of the boundary-layer. But during comparisons of AOT with near-surface aerosol concentrations the boundary-layer structure is decisive. The comparison with the AOT retrieved from satellite images was performed for the days 9, 12 and 13 September. High-quality satellite images were available for these days only. A channel, corresponding best to the 2nd spectral band of Landsat and to the 1st spectral band of SPOT, was the wavelength range 545.5–565.2 nm. Measurements in the time frame from 09:30 until 10:30 a.m. local time, i.e., the time of satellite passes, were used for comparison. The analysis of the direct-beam extinction measurements was performed after calibration by the Langley regression which is used for the retrieval of the MFRSR instruments as well (Harrison and Michalsky, 1994). The AOT measured from the ground at those days was 1.03, 0.500 and 0.516. These values were compared with the following AOT values extracted by SPOT satellite at the average wavelength of 550 nm at the location
THREE-DIMENSIONAL GROUND-BASED MEASUREMENTS
99
Figure 5. Quadratic dots show the actual aerosol optical thickness from satellite retrieval (dta) versus ground-based interference filter actinometer measurements (ground) at 9, 12 and 13 September, 1999. The triangular dots are with dta values shifted by 0.25 because the reference image was not totally free of aerosols. The deviation between these values and the ground values is less than ±10%.
of the ground-based instrument: 0.653, 0.271, 0.281. The algorithms are described in detail in Sarigiannis et al. (2001) as well as retrieval results are given in the last one. It seems that there is a shift of 0.25 in these values in comparison to the ground-based values giving values of 0.903, 0.521, 0.531 after summation. The deviation between these values and the interference filter actinometer data is less than ±10%, which is a satisfactory evaluation of the values retrieved from satellite images by these ground-based measurements of optical thickness (see Figure 5). The main reason for the AOT shift is that the reference image which was used for the retrieval of the AOT from the actual satellite images was not totally free of aerosols.
100
K. SCHÄFER ET AL.
4. Conclusions The most important input parameter for the extraction of information about air pollutants from the AOT retrieval of satellite images is the vertical structure of the boundary layer (PBL). The pollution level in the PBL above Brescia is influenced by local as well as regional emission sources. The dynamic structure of the PBL is dominated by the near-by Alps and the mountain breezes. Elevated heating and cooling surfaces in the nearby mountains modify the atmosphere. By advection, layers from the mountains are brought over Brescia resulting in a very complicated PBL structure and quite large PBL heights. Future ground-truthing of this kind of satellite data retrievals for determination of air pollution is necessary and should include: • Information about structure and height of boundary layer during satellite passes by remote sensing (backscatter LIDAR and/or DIAL ozone LIDAR up to 4,000 m altitude and/or long-range SODAR in homogeneous terrain); • Evaluation of aerosol column density retrievals from satellite images with data from the regional or global atmospheric aerosol ground-based measurement networks (AERONET, MFRSR) which can be directly compared with high spatial resolution satellite data as e.g. from the satellite IKONOS. The methods developed within this work will be applied at additional test sites and developed further in the framework of a new research and development project. Due to the high horizontal resolution of satellite imagery and the information content about aerosol loading and air pollution satellite imagery will be a useful tool to study the effectiveness of ground-based air pollution networks. These satellite data will be used in the future to optimise the spatial and technical configuration of air pollution networks to deliver data with most information for the objectives for their operation. The data described in this article were used as input to calculate secondary aerosol by an appropriate model. The modelled AOT were correlated with the AOT retrieved from satellite images during the same time (Sarigiannis et al., 2001). This procedure was an important step towards the use of satellite images to characterise air pollution.
Acknowledgements The financial support by the Commission of the European Community under Grant ENV4-CT97-0417 (project Integrated Computational Assessment via Remote Observation System (ICAROS): internet http://mara.jrc.it/icaros.html, including reports and publications) is gratefully acknowledged.
THREE-DIMENSIONAL GROUND-BASED MEASUREMENTS
101
The authors like to thank Mr. Percesepe, Mr. Bissolati and Mr. Bonetti from ASM Brescia for the operation and data management of the ground-based air pollution monitoring network in Brescia and the in situ monitoring station in the valley near Nave (including PM10) during the campaigns as well as Dr. Wiegner and Mr. Rabus from the Meteorologisches Institut der Universität München for the possibility to work with their interference filter actinometer and for their support during data analysis.
References Alexandrov, M. D., Lacis, A. A., Carlson, B. E. and Cairns, B.: 2001, ‘MFRSR-based climatologies of atmospheric aerosols, trace gases and water vapor’, in J. E. Russell, K. Schäfer and O. LadoBordowsky (eds), Remote Sensing of Clouds and the Atmosphere V, Proc. SPIE 4168, pp. 256– 264. Cachorro, V. and Tanre, D.: 1997, ‘The correlation between particle mass loading and extinction: Application to desert dust aerosol content estimation’, Remote Sensing Environ. 60, 187–194. Colls, J.: 1997, Air Pollution: An Introduction, E&FN SPON, London, 341 pp. Dosio, A., Emeis, S., Graziani, G., Junkermann, W. and Levy, A.: 2001, ‘Assessing the meteorological conditions of a deep Italian valley system by means of a measuring campaign and simulations with two models during a summer smog episode’, J. Atmosph. Environ. 35, 5441–5454. Grant, W. B., Kagann, R. H. and McClenny, W. A.: 1992, ‘Optical remote measurement of toxic gases’, J. Air Waste Manage. Assoc. 42, 18–30. Harrison, L. and Michalsky, J.: 1994, ‘Objective algorithms for the retrieval of optical depths from ground-based measurements’, Appl. Optics 33, 5126–5131. Haus, R., Schäfer, K., Bautzer, W., Heland, J., Mosebach, H., Bittner, H. and Eisenmann, T.: 1994, ‘Mobile FTIS-monitoring of air pollution’, Appl. Optics 33, 5682–5689. Junkermann, W.: 2001, ‘An ultra-light aircraft as platform for research in the lower troposphere: system performance and first results from radiation transfer studies in stratiform aerosol layers and broken cloud conditions’, J. Atmosph. Ocean Technol. 18, 934–946. King, M. D., Kaufman, Y. J., Tanré, D. and Nakajima, T.: 1999, ‘Remote sensing of tropospheric aerosols from space: Past, present and future’, Bull. Amer. Meteorol. Soc. 80, 2229–2259. Löscher, A. and Windholz, L.: 2001, Analyse von LIDAR-Ozon-Messungen der OLAK Messkampagne im August 1999 in Berlin auf die Möglichkeit einer Optimierung der Auswerteparameter, um die Schwankungsbreite der Messwerte zu reduzieren und die Ergebnisse zu verbessern, Interne Berichte 37, Institut für Experimentalphysik der Technischen Universität Graz, Petersgasse 16, A-8010 Graz, 85 pp. Reitebuch, O. and Emeis, S.: 1998, ‘SODAR-measurements for atmospheric research and environmental monitoring’, Meteorologische Zeitschrift, N. F. 7, 11–14. Sarigiannis, D., Soulakellis, N., Schäfer, K., Tombrou, M., Sifakis, N., Assimakopoulos, D., Lointier, M., Bossioli, E., Dantou, A. and Saisana, M.: 2001, ‘ICAROS: Integrated Computational Assessment via Remote Observation System’, Final Report for the European Community funded Project ENV4-CT97-0417, Institute for Health and Consumer Protection, EC–DG Joint Research Centre, Ispra (VA), I-21020, Italy. Sifakis, N.: 1992, ‘Potentiality of the High Spatial Resolution Satellite Imagery for Tracking of Air Pollution in the Lower Troposphere; The Athens Case Study’, Ph.D. Thesis, University Paris-7, 296 pp.
102
K. SCHÄFER ET AL.
Sifakis, N., Soulakellis, N. and Paronis, D.: 1998, ‘Quantitative mapping of air pollution density using Earth observations: A new processing method and application on an urban area’, Internat. J. Remote Sens. 19, 3289–3300. Sifakis, N. I., Sarigiannis, D., Assimakopoulos, D., Bonetti, A., Nicoloyanni, E., Lointier, M., Schäfer, K., Soulakellis, N. and Tombrou, M.: 1999, ‘ICAROS: Integrated Computational Assessment via Remote Observation System using Satellite Earth Observations’, in Proceedings of the HELECO’99 International Exhibition and Conference, Thessaloniki, 3–6 June 1999, pp. 486–495.