Elsevier

Computers & Geosciences

Volume 62, January 2014, Pages 142-149
Computers & Geosciences

CHRISTINE Code for High ResolutIon Satellite mapping of optical ThIckness and ÅNgstrom Exponent. Part II: First application to the urban area of Athens, Greece and comparison to results from previous contrast-reduction codes

https://doi.org/10.1016/j.cageo.2013.05.011Get rights and content

Highlights

  • Development of new algorithm and code for high resolution satellite mapping of AOT.

  • Novel retrieval method of relative AOT values not based on look-up tables.

  • Alleviation of the previous codes of the “invariant ground reflectance” assumption.

  • AOT retrieval with higher levels of confidence.

  • Ångstrom exponent and consequent aerosol size distribution mapping.

Abstract

There is an increasing demand for exploiting satellite data in urban air quality assessment. High spatial resolution satellite data can be used to retrieve the aerosol optical thickness (AOT), as an air quality indicator, over urban areas. One of the methods to achieve this applies the contrast-reduction principle to a set of two satellite images, one of which has minimum aerosol content and is used as a reference. Previous satellite image processing codes that followed this approach were subject to surface changes which may have occurred in the time interval between the processed images acquisition. In order to eliminate this potential source of AOT miscalculation the CHRISTINE Code for High Resolution Satellite Mapping of Optical Thickness and Ångstrom Exponent was developed. This new code takes into consideration contrast reduction in more than one spectral band, and applies the Ångstrom's law to isolate atmospheric and surface components. The code underwent its first testing using Landsat satellite data acquired before 2001 (when air pollution was at its peak) over the study area of Athens (Greece). Results showed that CHRISTINE can effectively separate contrast modifications attributed to atmospheric changes from those due to surface changes. Comparison against the previous SMA Satellite Mapping of Aerosols code showed an average improvement of 21% in terms of area over which AOT could be retrieved with high confidence. CHRISTINE also approximates the aerosol size distribution over the studied area. These preliminary findings show that the new code can be used to counteract for spatial deficiencies in urban monitoring networks. In the case of Athens the application to archived satellite data also allowed hindcasts for the period prior to ground based aerosol measurements.

Introduction

Urban air-quality measurements are heavily based on in-situ measurements of pollutants by ground based stations, which lack spatial continuity and may not be readily available in remote areas and developing countries (Gupta et al., 2006). This leads to an increasing demand for exploitation of satellite data. Unfortunately, while satellite remote sensing has become a valuable tool for assessing atmospheric pollutants at a global level (Borowiak and Dentener, 2006), there has been little effort to map air quality at detailed level, namely at the urban scale (Nichol and Wong, 2009). A poor collaboration between air pollution and remote sensing scientists as well as limited resources (both financially and in trained personnel) of the urban air quality monitoring sector are at the origin of this deficiency (Engel-Cox et al., 2004). Another reason for this is that no satellite mission aims at monitoring urban air pollution with fine spatial resolution.

A few satellite sensors, such as TOMS and GOME, and more recently MOPITT, AIRS and SCIAMACHY, gather systematically data on atmospheric species including aerosols and selected gases. The low spatial resolution (i.e., tens of kilometers) of these spectrometers is relevant to the spatial domain of global scale studies. On the other hand, optical satellite sensors estimate the aerosol load in terms of columnar aerosol optical thickness (AOT) which, when divided by the atmospheric mixing height (Dandou et al., 2002), can be converted to scattering coefficient values (kscat) predicting particle concentrations at ground level (Jiang et al., 2007, Liu et al., 2007) particularly PM2.5 mainly during winter conditions (Schaefer et al., 2008). AOT may be retrieved at various spatial resolutions by assessing the influence of the aerosols to satellite radiometry. More specifically, the polar AVHRR and ATSR-2, and the geostationary GOES-8 and SEVIRI may retrieve AOT with moderate spatial resolution (i.e., few kilometers), but they address large scale pollution phenomena (e.g., Popp et al., 2007). MODIS and MERIS provide on a daily basis standardized products on aerosol load and properties at moderate-to-high spatial resolution (i.e., hundred meters) and they can cover regional scale phenomena (e.g., Chu et al., 2003, Von Hoyningen-Huene et al., 2003, Tang et al., 2005, Liang et al., 2006, Kokhanovsky et al., 2007). The last generation of very high spatial resolution sensors (ground sampling distances of less than a meter), such as IKONOS-2, Quickbird, GeoEye-1 and WorldView-1, may provide circumstantial information on air pollution (e.g., visible pollution plumes) or on ancillary parameters influencing emissions and dispersion (e.g., land cover, traffic load, terrain roughness) (Eikvil et al., 2009). Finally, the latest active satellite lidar instruments, such as CALIPSO, allow to accurately assessing aerosol properties and load, but they focus on the vertical not the horizontal distribution.

The current study focuses on the use of high spatial resolution (HSR) sensors (i.e., resolution from tens to a few hundreds of meters), such as Landsat TM/ETM+, SPOT HRV/HRVIR and IRS LISS-III, that pre-date most of the optical sensors previously mentioned but address land and sea rather than atmospheric observations. Nonetheless, qualitative atmospheric observations using HSR sensors were cited as early as in the 70s; they concerned smoke emitted from industrial/urban sites or from forest fires (Short et al., 1976), assessment of pollution associated to statistical indicators (Potter and Medlowitz, 1975) and aerosol estimations over water (Griggs, 1975). The first quantitative pollution assessments appeared in the early 80s (Fraser et al., 1984), and attempts for pollution mapping started in the late 80s (Sifakis,, Tanré et al., 1988) and early 90s (Sifakis and Deschamps, 1992). The use of HSR satellites to map over urban areas progressively received further attention from various researchers who have developed a range of techniques with respective limitations (e.g., Kocifaj and Horvath, 2005). For example, the “clear water” method can be used only above water surfaces (Gordon and Clark, 1981) while the dense dark vegetation (DDV) method, requires the presence of vegetated areas (Kaufman and Sendra, 1988), and the “deep blue” method, proposed by Hsu et al. (2004), has shown satisfactory results over bright targets but is applicable only to satellite sensors with bands sensitive in the blue spectral area. Finally, the “contrast-reduction” principle, namely the apparent or observed-at-satellite reduction of contrast between distinct surface targets, engendered by the scattering mechanism of aerosols in the visible and infrared spectral areas, gives satisfactory results over urban areas, composed by heterogeneous land parcels. This part of the study (Part II) aims at applying the new Code for High Resolution Satellite Mapping of Optical Thickness and Ångstrom Exponent (CHRISTINE), developed in Part I of this article, to HSR data of the Athens urban area and for the years prior to 2001 when no ground measurement on aerosol concentrations was yet available.

The CHRISTINE code applies the contrast reduction principle to a set of two satellite images, one of which has minimum aerosol content and is used as a reference. Contrast reduction was first described by Middleton (1952) through the Koschmieder's equation:C/Co=exp(AOT)where C is the observed and Co the real contrast of a target. Eq. (1) gives an idea of the sensitive dependence of AOT on contrast reduction. Contrast reduction may also be sensitive to other than atmospheric variations; previous contrast reduction codes were subject to surface changes which may have occurred in the time interval between the processed images acquisition. In order to eliminate this potential source of AOT miscalculation CHRISTINE takes into consideration contrast reduction in more than one spectral band, allowing to separate contrast modifications attributed to atmospheric changes from those due to surface changes. This increases the area over which AOT can be retrieved with higher level of confidence. The new code furthermore provides an approximation of the Ångstrom coefficient and creates maps depicting the aerosol size distribution.

Section snippets

Study area

The geographic area selected for testing the new code covers the Greater Athens Area (i.e., the Athens basin). This region is well known for air pollution problems and the prior existence of a large body of associated studies (e.g., Kambezids et al., 1998). Athens is a city with a history of environmental, particularly air quality, problems (Kambezidis and Sifakis, 2004). It accommodates a population of about 3.5 million inhabitants (census 2001). Most of the population live within the Athens

Results

The primary outcome of CHRISTINE is a map depicting the horizontal distribution of AOT. Fig. 3 depicts 12 of the 13 AOT maps resulted from the application of the code to the selected images. These maps depict the spatial distribution of urban aerosols over the Athens basin during typical pollution days between 1986 and 2001. The first thing to notice is that the pollution cloud is confined, in all but two cases inside the basin around the city center. In two acquisitions (i.e., 1986 and 1987)

Summary and conclusions

A new image processing code (CHRISTINE) for high resolution satellite mapping of aerosol optical thickness (AOT) and Ångstrom coefficient approximation underwent its first testing. The code is based on the contrast reduction principle between a reference satellite image (with minimum aerosol content) and an examined image. The new code intends to overcome an important limitation of the previous (DTA and SMA) codes: their sensitivity to artifacts attributed to ground reflectance variations

References (41)

  • A. Retalis et al.

    Urban aerosol mapping over Athens using the differential textural analysis (DTA) algorithm on MERIS-ENVISAT data

    ISPRS Journal of Photogrammetry and Remote Sensing

    (2010)
  • P.A. Scheff et al.

    Characterization and source identification of respirable particulate matter in Athens, Greece

    Atmospheric Environment

    (1990)
  • J. Tang et al.

    Aerosol optical thickness determination by exploiting the synergy of TERRA and AQUA MODIS

    Remote Sensing of Environment

    (2005)
  • D.A. Chu et al.

    Global monitoring of air pollution over land from the Earth Observing System-Terra Moderate Resolution Imaging Spectroradiometer (MODIS)

    Journal of Geophysical Research

    (2003)
  • J. Colls

    Air Pollution: An Introduction

    (1997)
  • T. Cvitaš et al.

    Characteristics of air pollution during the summer in Athens, Greece

    Staubtechnik—Reinhaltung der Luft

    (1985)
  • A. Dandou et al.

    The importance of mixing height in characterising pollution levels from aerosol optical thickness derived by satellite

    Water, Air, and Soil Pollution: Focus

    (2002)
  • J. Engel-Cox et al.

    Recommendations on the use of satellite remote-sensing data for urban air quality

    Journal of the Air & Waste Management Association

    (2004)
  • R.S. Fraser et al.

    Interaction mechanisms within the atmosphere

  • Cited by (0)

    View full text