Discrimination of cloud and snow covered regions using IRS-P3 MOS data

The Indian region is geographically divided into the Himalayan, Indo-Gangetic basin and peninsular regions. The Himalayan region has paramount importance in the economic development of the country. All kinds of natural resources are found here. The high topography of the Himalayan region has direct influence on the dynamics of the weather conditions. The major rivers originating from the Himalayan region are the sources of surface and ground water specially for people living in the Indo-Gangetic basin. Monitoring and mapping of the Himalayan region on a day-to-day basis is possible using the broad-band remote sensing sensors. The spatial resolution of microwave sensors is poor compared to visible optical sensors. The visible sensors suffer from the problem of cloud cover, specially in high altitude regions such as the Himalayan region. The spectral reflectance of snow, ice, cloud, vegetation and bare ground have distinct behaviour. However, due to the presence of cloud, the information from the snow, ice, vegetation and bare ground are masked by the high reflectance of the cloud. The effect of the cloud1 can be distinguished in the Short Wave Infrared (SWIR).

The Institute of Space Sensor Technology of the German Aerospace Centre (DLR) has developed a Visible/Near Infrared (VIS/NIR) imaging Spectrometer named Modular Optoelectronic Scanner (MOS)2,3 launched on 21 March 1996 on-board the Indian Remote Sensing satellite IRS-P3 (ref. 4) into a sun-synchronous polar orbit at 817 km. MOS sensors consist of two separate imaging spectro-radiometers MOS-A (4 bands) and MOS-B (13 bands) in the VIS and NIR and a CCD-line camera MOS-C (1 band) in the SWIR. The details of MOS-A, B and C sensors5 are given in Table 1.

505.gif (11303 bytes)

506.jpg (66028 bytes)

Data recorded by MOS are stored as 16-bit digital numbers. The radiance Lij measured by the MOS sensor for a pixel i in channel j is given by Neumann
et al.5 as

Lij = (Uij – Uoij)*fij*kj, (1)

where i is the pixel number in scan line, 1:140 for MOS-A, 1..384 for MOS-B and 1..300 for MOS-C; j is the channel number (wavelength), 1..4 for MOS-A, 1..13 for MOS-B and 1 for MOS-C; Uij is the readout pixel value (in digital count); Uoij is the dark value (in digital units, measured during calibration phase); fij is the sensitivity correction factor; and kj is the calibration factor.

In level-1B file, the dark-value corrected pixel values multiplied by the sensitivity correction factor [(Uij –Uoij)*fij] is stored as 16-bit digital number (unsigned short format). The calibration factor kj for each channel of MOS-A, B and C are written in the file header provided by NRSA, Hyderabad behind the keywords ACAL_VAL and BALC_VAL in the order of the channel numbers. These coefficients are constant up to now from the beginning of the mission. So, multiplying the DN values with the calibration coefficient, one can easily obtain the radiance values. From the radiance values, reflectance is computed using the following equation5

Rij = [p *Lij]/[cos q s*eoij], (2)

where Rij is the reflectance; Lij is the radiance; q s is the solar zenith angle; and e0ij is the solar irradiance.

Values of q s and e0ij of central pixel of each sub-scene have been provided with the data. The values of q s and e0ij for each pixel are assumed to be constant. In Figure 1, colour composite image (Figure 1 a) of MOS-B (channels 11, 8 and 7) has been compared with MOS-C (Figure 1 b) image (channel 1–1600 nm). After various combinations of channels, we have found colour composite image of channels 11, 8 and 7 giving sharp features to study bare ground, vegetation, snow and cloud surfaces. The image shown in Figure 1 a, b represents an area in the Himalayan region, the coordinates of the four corners of the image are shown in Figure 1 a. This image is extracted from the full take scene along the path 95 taken on 11 May 1998. The reflectances and radiances at various sites covered with snow, vegetation, cloud, and bare ground have been extracted and are shown in Figure 1 c, d, respectively. From the texture of the two images, it is clearly seen that the area is covered by the mountain. In the colour composite image (Figure 1 a), the red tone reflects vegetation cover. The white patches seen in Figure 1 a do not differentiate snow cover from cloud cover. The bare ground is seen with dark tone, which is easily mapped using colour composite image if the cloud cover is not present. However, in the MOS-C image at 1600 nm, snow reflectance is lower than those for land, therefore snow cover is seen darker in the MOS-C image (Figure 1 b) than that of cloud cover which has high reflectance, almost nearly the same as in the colour composite image (Figure 1 a). Figure 1 c, d shows the characteristic behaviour of spectral reflectance of snow, cloud, vegetation and bare ground. Higher radiances and higher reflectances over cloud and snow are seen compared to those over vegetation and bare ground in the VIS ranges. In the SWIR (1600 nm) the higher reflectance contract between cloud and the rest of the surfaces (snow, cloud and bare ground) is seen. In Figure 2, we have shown a classified image of the area. The image has been classified using Maximum Likelihood Supervised Classification method from MOS-B data. In the classified image, four classes (snow, cloud, vegetation and bareground) are clearly seen. The reflectance values shown in Figure 1 d also confirm the presence of different classes in the image. The present study shows that using MOS-B colour composite image, MOS-C image and spectral reflectance values, reliable mapping of snow cover region can be carried out.

 

  1. Zimmermann, G. and Neumann, A, Proceedings of 32nd COSPAR, Nagoya, PSRDC1-0004, 1998.
  2. Zimmermann, G., Proceedings of COSPAR colloquium, Space RS of Subtropical Oceans (SRSSO), Taipei, 13–16 September 1995.
  3. Zimmermann, G. and Neumann, A., Proceedings of 1st International Workshop on MOS-IRS and Ocean Color, DLR, Institute of Space Sensor Technology, Berlin, 1997, pp. 1–9.
  4. Thyagarajan, D., Neumann, A. and Zimmermann, G., Acta Astron., 1996, 39, 9–12.
  5. Neumann, A., Walzel, T., Tschentscher, B., Gerasch, B. and Krawczyk, H., MOS-IRS Data Processing, Software and data products, Pre-Launch Release 24.01.96, 1996.

 

 

ACKNOWLEDGEMENT.  We are grateful to Dr G. Zimmermann, Dr A. Neumann and Mr T. Walzel for their help in displaying the MOS data. R.P.S. is grateful to former SAC Director, Dr George Joseph and Dr R. R. Navalgund who have provided an opportunity to familiarize with MOS data. We thank the anonymous referees for their useful comments and suggestions. P.D. is grateful to CSIR, New Delhi for the award of Junior Research Fellowship. The data was made available to R.P.S. by ISRO, Bangalore free of charge through NRSA, Hyderabad. We are grateful to Dr K. Kasturirangan, Chairman, ISRO and Mr V. Jayraman, Director, EOS, ISRO, Bangalore for data.

 

 

RAMESH P. SINGH

SUDIPA ROY

PRASANJIT DASH

 

 

Department of Civil Engineering,

Indian Institute of Technology,

Kanpur 208 016, India