The column water vapour content over ocean can presently be derived from space-borne microwave sensor data^1,2. Over land surfaces microwave sensors fail due to highly varying microwave emissivity and water vapour amounts can be determined from radiosonde measurements, infrared space-borne data and back-scattered measurements of solar radiation. The radiosonde measurements have the problem of being only point measurements, whereas, the back-scattered solar radiation measurements do not reach a high accuracy³. Various methods have been used to determine column water vapour from back-scattered solar radiation using the highly resolved spectral data of imaging spectrometers^3–5.

Atmospheric water vapour is given in centimeter precipitable water which is the height of the liquid water column that would emerge when condensing all water vapour in the zenith direction on to one unit area. One cm corresponds to 1 g/cm² of column water vapour. The method to deduce atmospheric water vapour is discussed in detail by Tahl and Schonermark⁶ which is given by the Continuum Interpolated Band Ratio (CIBR):

 

CIBR = L(l_v)/c₁L(l_w1) + c₂L(l_w2)
                                                                 (1)

 

where L is the radiance at the top of atmosphere. The index v indicates the water vapour channel and w1, w2 are the two neighbouring window channels. The coefficients c₁ and c₂ are defined as:

 

c₁ = (l_w2 – l_v)/(l_w2 – l_w1),                     (2)

 

c₂ = (l_v – l_w1)/(l_w2 – l_w1).                     (3)

 

The 940 nm absorption band is the most sensitive to variations in atmospheric water vapour content³. The channel 12 (940 nm) of the MOS-B spectrometer has been selected for the water vapour channel and correspondingly channel 11 (867 nm) and 13 (1009 nm) have been used as the window channels. The water vapour path values are converted to column water vapour (V_c) by using the zenith angles of the sun and sensor according to

 

V_c = V_p (1/cos q_i + 1/cos q_s)^–1,              (4)

 

where V_c is column water vapour (cm) and q_i is the viewing angle equal to 0° for MOS-B and q_s is the solar zenith angle, q_s is taken for the center pixel of the scene. These information are available with the MOS data from the Data Center, National Remote Sensing Agency, Hyderabad.

The water vapour path is the amount of the gas, which is penetrated by the radiation on its sun surface-sensor path. Tahl and Schonermark⁶ have found that for non-vegetation and vegetation cover the water vapour path values (V_p) are different. So there is a need to distinguish between non-vegetation and vegetation cover. In order to determine the surface cover of a pixel NDVI has been used as the criteria. The NDVI has been computed using following formula:

 

NDVI = [L(867) – L(683)]/

[L(867) + L(683)],                (5)

 

where L(867) corresponds to channel 11 and L(683) corresponds to channel 8 of IRS-P3 MOS-B sensor.

The threshold between non-vegetation and vegetation cover has been set to 0.5. Tahl and Schonermark⁶ have given the following equations for water vapour path values in the case of both non-vegetation and vegetation cover.

 

V_p = (–ln (CIBR)/0.592)^1/0.568,             (6)

 

and

 

V_p = (–ln (CIBR)/0.599)^1/0.575.             (7)

 

 

Taking radiation at the top of the atmosphere, the portion of the path radiance compared to ground reflected radiance is larger over dark surfaces compared to bright ones. As the path radiance does not reach the earth’s surface, it has a shorter way through the atmosphere other than the surface reflected radiance and therefore, it has less penetration through water vapour. Therefore, to correct for the underestimation of water vapour content for low reflecting surface cover. Tahl and Schonermark⁶ have given a correction for V_p. The V_p corrected equations for both non-vegetation and vegetation cover are:

 

V_p,corr = V_p/0.464 + 0.130*ln[c₁L(867)

+ c₂L(1009)/cosq_s],                   (8)

 

and

 

V_p,corr = V_p/0.587 + 0.092*ln[c₁L(867)

+ c₂L(1009)/cosq_s].                   (9)

 

The water vapour path values are different over vegetation and non-vegetation cover. Therefore, the estimate of vegetation cover in the area is essential before determination of water vapour.

A subscene consisting of 800 scan lines by 384 pixels has been extracted from MOS-B data path 95 of 11 January 1989 and 11 May 1998. Detailed characteristics⁷ of IRS-P3 MOS sensors are given in Table 1. This subscene covers an area of 416 km by 200 km on the earth’s surface. Using eq. (5) NDVI values have been computed and colour-coded NDVI images have been generated. Figure 1 a and b show the NDVI images of 11 January 1998 and 11 May 1998, respectively. Figure 2 a and b show the colour composite (band 11, 8 and 7) images of 11 January 1998 and 11 May 1998, respectively. The latitude and longitude of the four corners of the subscene are shown in Figure 2 b). NDVI values obtained vary in the range between –1 and +1. These values have been rescaled from 0 to +1. Water bodies, both inland and the sea with 0 NDVI values have been shown with blue colour. The non-vegetation covered areas are shown with dark red colour and areas with highest amount of vegetation cover are shown in light green colour. In the 11 January 1998 image, low NDVI values are seen in areas where cloud cover is present. The vegetated areas generally yield high NDVI values because of their relatively high reflectance in infrared and low visible reflectance. In the colour composite images of Figure 2 a and b, the areas with red colour reflect vegetated areas. In band 11 (867 nm, infrared region) vegetated areas have high reflectance values and appear red. In this region there are moist tropical evergreen forests, which show high NDVI values (green colour) both in the 11 January 1998 and 11 May 1998 images. The spatial features in the NDVI map match very well with the spatial features of the these colour composite images.

Spectral reflectance plots at points marked on the colour composite images of 11 January 1998 and 11 May 1998 are shown in Figure 3 a and b, respectively. In Figure 2 b, we have selected points 1 to 6 which are lying over vegetated region, point 7 lying over inland water body and point 8 lying over the sea intruding into land. The spectral reflectance plots (Figure 3 b) for points 1 to 6 manifest the ‘peak and valley’ configuration of green healthy vegetation. The valleys in the visible portion of the spectrum are attributed to the pigments in the plant leaves. Chlorophyll strongly absorbs energy in the wavelength band centered at 670 nm, which can be seen as an absorption peak at band 7 (650 nm) and band 8 (685 nm). The reflectance increases dramatically from the visible to infrared portion of the spectrum at about 700 nm. Plant reflectance in the range 700 nm to 1200 nm is primarily due to the internal structure of the plant leaves. Near infrared reflectance increases with the number of layers of leaves in a canopy. A minor absorption peak at band 10 (814 nm) and a major absorption peak at band 12 (942 nm) are seen (Figure 3 b). These absorption peaks are attributed to the strong absorption at these wavelengths (water absorption bands) due to the presence of water in the leaves. Reflectance peaks can be seen at bands 9 (750 nm), 11 (867 nm) and 13 (1010 nm). Vegetation curves also have an absorption peak at 1400 nm and a small reflection peak at 1600 nm. However, as MOS has no band at 1400 nm this feature cannot be seen in the reflectance curves. From Figures 2 b and 3 b, it can be seen that in the infrared region (700 nm to 1200 nm) points lying on regions which have darker shades of red (highly vegetated areas) in the colour composite image have high reflectance values and those lying on regions with lighter shades of red (less vegetated areas) have low reflectance values. The reflectance values are found to be highest in band 11 and lowest in band 8 (Figure 3 b). Due to this high contrast, these bands are used for NDVI calculation. In Figure 2 a, points 5 and 6 lie over cloud-covered regions, so at these points high reflectance (Figure 3 a) values for all bands are seen. The reflectance plots confirm the presence of vegetated and non-vegetated regions.

Figure 4 a and b show the column water vapour (cm) images of 11 January 1998 and 11 May 1998, respectively. The column water vapour values range from 0.22 to 0.60 cm in Figure 4 a and b. We have found that the column water vapour is higher in 11 January 1998 than 11 May 1998. This observation is also consistent with the presence of clouds found at various places in the 11 January 1998 image. This seems to be plausible due to contrast in the surface temperature of ocean and land as a result of which water evaporates from the ocean and moves in to the atmosphere over the land. The dark values are those with smaller amounts of water vapour and high values are marked white. Water surfaces have been masked and are painted black. Due to the very low reflectance of water in the spectral region of MOS bands 11 (868 nm), 12 (942 nm) and 13 (1011 nm), the signal received over water surfaces mainly consists of path radiance; therefore the estimate of water vapour column cannot be made. As the algorithm used is based upon the surface reflected radiance, which is attenuated by water vapour on its way through the atmosphere, it cannot be applied over water surfaces.

In the water vapour image of 11 January 1998 and especially of 11 May 1998, the Western Ghats are clearly visible running parallel to the coastline. The Western Ghats have higher amounts of water vapour compared to the adjoining coastal plains. The white patches in the image of 11 January 1998 is due to cloud cover. In the corresponding water vapour image, higher amount of water vapour represent the clouds which obviously contain a lot of water vapour. When compared with the NDVI images (Figure 1 a and b) of the same area it is seen that lower water vapour amounts are found over highly vegetated regions surrounded by higher amounts of water vapour over less vegetated areas. In the absence of ground truth data, it is difficult to validate the present results. However, the present results show a great potentiality of IRS-P3 MOS data in the determination of water vapour over land region.

1.   Prabhakara, C., Chang, H. D. and Chang, A. T. C., J. Appl. Meteorol., 1982, 21, 59–68.

2.   Schluessel, P. and Emery, W. J., Int. J. Remote Sensing, 1990, 11, 757–766.

3.   Gao, B. C. and Goetz, A. F. H., J. Geophys. Res., 1990, 95, 3549–3564.

4.   Gao, B. C., Westwater, E. R., Stankov,
B. B., Birkenheuer, D. and Goetz, A. F. H., J. Appl. Meteorol., 1992, 31, 1193–1200.

5.   Frouin, R., Deschamps, P. Y. and Lecomte, P., J. Appl. Meteorol., 1990, 29, 448–460.

6.   Tahl, S. and Schonermark, M. V., Int. J. Remote Sensing, 1998, 19, 3223–3236.

7.   Zimmermann, G. and Neumann, A., Proceedings of the 1st International Workshop on MOS-IRS and Ocean Colour, DLR, Berlin, 1997, pp. 1–9.

 ACKNOWLEDGEMENT.  We are grateful to Drs G. Zimmermann, A. Neumann and Mr Thomas Walzel, DLR, Germany for their help regarding IRS-P3 MOS data processing and Mr Robert Scheu for his help in preparing NDVI image.

 

RAMESH P. SINGH

SUDIPA ROY

 Department of Civil Engineering,

Indian Institute of Technology,

Kanpur 208 016, India