Home Articles Estimating atmospheric turbidity from SPOT and GMS-5 data

Estimating atmospheric turbidity from SPOT and GMS-5 data

Gin-Rong Liu, Tang-Huang Lin, A.J. Chen
Center for Space and Remote Sensing Research
National Central University, Chung-Lu, Taiwan 320
Tel:(886)-3-4227151 ext 7620, FaxL996)-3-4255535
E-mail: [email protected]

The atmospheric turbidity is one very important factor in the air pollution measurements and monitoring with remotely sensed data, especially in visible bands. The scattering effects of atmospheric molecular and aerosols in varying atmospheric turbidity conditions can influent the original spectral information of remotely sensed data strongly. From another point of view, the atmospheric turbidity can be estimate by evaluating the information variation induced by the scattering effects. Tanre et al. proposed the Structure Function (SF) to estimate the atmospheric optical depth in 1988. Their study result showed that the Aerosol Optical Depth (AOD) can be assessed with Landsat TM data by assuming the landcovers are same in the set of multi-temporal TM images. In this study, the SF method is improved for appliying in Taiwan area. Owing to the rather rough terrain and complex landuse properties in Taiwan area, this study used higher spatial resolution SPOT data and hourly GMS-5 data, to derive the AOD. The result shows the improvements in this study can get satisfying result. The result reveals we can derive these satellite data for the monitoring of hourly air pollution and air quality variation.

The air pollution index (PSI) contains the effects of SO2, NO2, CO, O3 and turbidity, etc. Since turbidity is one of the main effects of the air pollution, effective monitoring of the aerosol and suspend sediments (S.S.) of the atmosphere becomes very important for the air pollution control. Because of having the advantages of winder Range, higher temporal resolution and data consistent property than the traditional measured data, applying satellite observed data to monitor the air pollution condition becomes one alternative way. In this study the aerosol optical depth (AOD) had been estimated from kinds of satellite data around Taiwan area, and demonstrated to show its application in air pollution monitoring.

Because aerosol parameters are not easy to estimate, methods of aerosol parameter retrievals has been discussed by many research terms. For example, an image-based of retrieved aerosol characteristics by the Dense Dark Vegetation (DDV) method to calibrate atmospheric effect of image of image itself was present by Liu et al. in 1996. But the DDV method applied in this study showed larger error when there are few or none DDV pixels in the test images. By assuming the ground reflectance is constant, variations of satellite signal may be attributed to variations of the atmospheric optical properties. Based upon this, another method of multi-temporal aerosol parameter retrievals by a reference ground measurement was accomplished by Tanre et al. in 1988. The single-directional structure function (SF) is defined for deriving AOD from Landsat TM data. Holben also got good results by applying SF method to NOAA AVHRR data in 1992.

In this study, an improvement of SF method is applied to SPOT and GMS-5 data. It also showed reasonable results can be gotten. The high spatial resolution SPOT data can provide more detailed information of atmosphere on local area, like industrial or manufacturer locations, and the GMS-5 data can hourly provide the distribution and variation information of atmosphere around the wise area. If these kinds of satellite data can be routinely applied, effective monitoring of the air pollution from satellite observation can be accomplished easily. This is the main aim of this study.

The method of AOD retrieval from satellite data developed by Tanre et al. in 1988 was improved by Liu et al. in 1997. The improvements of SF method contain : (1) The single- directional SF is replaced by multi-directional SF to describe the surface characteristics more completely 92) Introduce an “optimum number” to reduce the characteristics effects of local landcovers and terrain. Inthis study the improved algrithm of AOD retrieval is applied for the high spatial resolution SPOT data and hourly GMS-5 data. The flow chart of using satellite data to monitor the air pollution is showed in figure 1. In the flow chart, two points are very important : one is the AOD retrieval and the other is to establish the relationship between AOD and S.S. Basically, the relationship between AOD and turbidity or air quality index is more complex. The variation of surface boundary layer thickness and ground observation range seems to be unclear, and need more analysis to find the relationship between AOD and S.S. This research need to be studied continuously because it is practical in the air pollution monitoring in the future.

Figure 1. The flow chart of using satellite data to monitor the air pollution.
1.AOD retrieved from SPOT data
Seven SPOT images, Mar. 7, Apr. 3, Jun. 28, Aug. 10, Aug. 31, Sep. 16, and Sep. 26 in 1994, are used to retrieve the AOD in this paper.The reference AOD is obtained by DDV method (Liu et. Al., 1996) for lack of ground measurements during the SPOT images observation. The results of AOD retrieval of 3 test area in each channel of SPOT are shown in table 1 and seems to be very reasonable when compared with the results of DDV method, especially for few or few or none DDV pixels in the images on 7 Mar and 26 Sep, 1994. The distribution of AOD retrieval from XSI data, 31 Aug, 1994, is shown in figure 2. The AOD on the right-down and left-down area of the figure are larger than other area because of one industrial area or some manufacture locations and a little urban.

Data DDV method SF method
Area1 Area2 Area3 Mean Std
1994/03/07 1.59580 .74165 .75668 .76108 .75314 .00823
1994/04/03 .65540 .61667 .62257 .60847 .61590 .00578
1994/06/28 .44530 .52885 .42064 .45748 .46889 .04779
1994/08/10 .40840(ref.) .43395 .43395 .43395 .43395 0
1994/09/26 .12990 .57785 .54803 .42580 .51723 06578
1994/08/31 .45690 .58989 .54566 .50144 .54566 .03611
1994/09/19 .43140 .63786 .56861 .48750 .56466 .06144
Data DDV method SF method
Area1 Area2 Area3 Mean Std
1994/03/07 1.34590 .68829 .71965 .74514 .71769 .02325
1994/04/03 .55160 .52445 .55613 .57545 .55201 .02102
1994/06/28 .37270 .47250 .43114 .36419 .44261 .04462
1994/08/10 .53640(ref.) .42779 .42779 .42779 .42779 0
1994/09/26 .08430 .42004 .54969 .38529 .45167 .07075
1994/08/31 .37810 .42274 .51481 .45723 .46493 .03798
1994/09/19 .34020 .43374 .49018 .37193 .43195 .04829
Data DDV method SF method
Area1 Area2 Area3 Mean Std
1994/03/07 1.03050 .64990 .65513 .67113 .65872 .00903
1994/04/03 .42080 .52758 .44424 .51858 .49680 .03735
1994/06/28 .28190(ref.) .28543 .28543 .28543 .28543 0
1994/08/10 .82270 .45917 .46117 .48701 .46911 .01268
1994/09/26 .04280 .30467 .27128 .36234 .31276 .03761
1994/08/31 .28100 .47419 .40316 .46395 .44710 .03135
1994/09/19 .23440 .33094 .32229 .33291 .32871 .00461

Table 1. The comparison of AOD retrieved from SPOT data by SF and DDV methods.

Figure 2. The contours of AOD (x 0.1) estimated from SPOT data on 31 Aug, 1994.

2.AOD retrieved from GMS-5 data
The GMS-5 visible data, 01Z 27 June, 1998, was used to retrieve AOD . The reference AOD is obtained by Ce-318 sumphotometer ground measurements during the images observation. Good result of AOD retrieved from GMS-5 visible band (550 ~900 mm) in Chung-Li area is 0.230 when compared with the ground measurement, 0.225. The retrieved AOD of whole image is illuminated in figure 3. At also showed that a reasonable distribution of AOD around Taiwan area which the AOD near cities area are larger than oceanic or mountain area.

Figure 3. the distribution of AOD estimated from GMS-5 data on 01Z 27 Jun, 1998.
3. The relationship between AOD and S.S.
The sunphotometer observation is located on the National Central University, Chung-Li, Taiwan to observe AOD, and the S.S. data is provided from the Environmental Protection Agency (EPA) of Taiwan measured at EPA Chung-Li station. The relationship between these two ground observations is showed in figure 4. It seems no correlation can be found between these two measurements from the figure 4. It may be caused by the sunphotometer observation mode. The measurement of sunphotometer includes the total column AOD of atmosphere, but the measurement of S.S. may be just near ground surface. So, when the boundary of ground layer is increased , the measurements of S.S. on the ground will be decreased. But the total quantity of S.S. in the atmosphere does not change. So, the thickness variation of boundary layer and ground observation range should be considered further to find the relationship between AOD and S.S. Also, it needs long-term data for advanced analysis.

Figure 4. The relationship between AOD and S.S.
The results of applying SPOT and GMS-5 data to estimate the local AOD shows high practical. Once the relationship between AOD and S.S. is established, the satellite observation can be applied to monitor the air pollution efficiently. So, to construct the relationship between AOD and S.S. is most important thing for monitoring air pollution satellite observation in the future research.


  • Holben, e. Vermot, Y.J. Kaufman, D. Tanre, and V.Kalb, 1992: Aerosol Retrieval over Land from AVHRR Data-Application for Atmospheric Correction. IEEE Trans. on Geoscience and Remote Sensing, Vol. 30, NO. 2 pp. 212-222.
  • Liu G.R, T.H. Lin, and A.J. Chen, 1997: An Improved Method to Determine Aerosol Optical Depth from SPOT Data. COAA ’97-First International Ocean-Atmosphere Conference, 19-19 Oct 1997, Washington, D.C., USA.
  • Liu C.H, A.J. Chen, and g.R. Liu, 1996: An Image-based Retrieval Algorithm of aerosol Characteristics and Surface Reflectance for Satellite Images. INT. J. Remote Sensing, Vol. 17, No. 17, 3477-3500.
  • Tanre D., P.Y Deschamps, C. Devaus, and M. Herman, 1988: Estimation of Saharan Aerosol Optical Depth from Blurring Effects in Thematic Mapper Data. J. Geophys Res., pp Vol. 93, No. D12, 15955-15964.