completed. But, the velocity of water flow becomes slower and 
the capability of purification decreased, which increases the 
possibility of sand sediment, organic matter deterioration and 
algae growth. Therefore, how to monitor the pollution of the 
Three Gorges Dam and protect the water quality effectively 
will be an important question. 
2. DATA AND IMAGE DISPOSAL 
2.1 Data Information 
The CHRIS/PROBA data in this study is acquired on April 1 st ; 
2006.The quality of images is good. In order to retrieve the 
chlorophyll concentration with images using semi- experiential 
method, one campaign was carried out for 18 samples on 6-7 th , 
April, 2006.The samplings were distributed along the 
reservoir where is far away from the dam above 5 kilometres. 
The geography coordinates of the samplings were recorded by 
GPS device. Simultaneously, remote sensing reflectance was 
measured with Field Spec HandHeld2500. Water samples were 
collected from the surface above 30 cm below in the vertical 
direction with a standard 21 polyethylene water-fetching 
instrument immediately after measuring spectra. Then they 
were held on the deepfreeze half with ice bags to reserve for 
approximately 4h every afternoon, then returned to the 
laboratory for concentration measure. Additionally, 
transparency was measured with a secchi disk at the shade side 
of the boat. 
2.2 Image Processing 
2.2.1 Destriping: Push-broom images equipped with a matrix 
detector, like CHRIS sensor, are heavily affected by coherent 
disturbance spatially and spectrally due to sensitivity changes 
between the detector elements (Dong Guang-xiang, 2006).So 
noises affected imaging are more complicated and there are 
horizontal and vertical strips noises on images. 
More scholars have done researches to remove the destrpings. 
Sandra Mannheim (2004), Alessandro Barducci (2005), Dong 
Guang-xiang (2006) deal with the images with emendation 
factor, ratio and iterative methods respectively. The values of 
the resultant images that they obtained are consistent with the 
original values nearly, and most noises have been removed. 
However, those methods, which are applicable for professional 
peoples, are difficult for curbstone persons. Taking one thing 
with another, we use HDFclean of ESA to remove the strips 
(See http://www.earth.esa.int/proba/HDFcleanv2.html).This 
method is easier than those methods and the result is not worse 
either. The same result has been educed in other researches 
(Ted Miltion, 2006) 
Fig 2. The contrast of the original and the resultant images 
Fig 2 shows the contrast between the original and the resultant 
images. The value does not change obviously, while the noises 
have removed, not only the wide strips (shown as red rectangle 
mark) but also the slender ones (shown as blue rectangle mark). 
2.2.2 Radiation Correction: In order to compare and analyze 
quantificationally, the data that satellite sensor transmits needs 
to be transformed to radiation lightness or reflectivity (Wu 
Yun-zhao, 2004). As the data that the CHRIS sensor accepts is 
atmospheric radiation lightness, we only need to transform the 
atmospheric radiation lightness to reflectivity. 
The reflectivity (r ) is calculated by equation (1). 
r= 7tLid ]-< (!) 
E> eosift) 
Where L*= the atmospheric radiation lightness; 
d= the distance between the sun and the earth 
(astronomical unit); 
Ex= Mean solar atmospheric irradiances; 
9s= solar zenith angle. 
2.2.3 Atmospheric Correction: The process of imaging is 
very complex and the reflected electromagnetic signals 
transmitting form the sun to the sensor is affected by 
atmosphere inevitably. The main atmospheric components 
affecting the electromagnetic radiation in the visible and NIR 
spectral region are ozone, aerosols, and water vapour. Those 
cause the loss or the corruption of part of the carried 
information above the observed target. When the radiation 
transmits in the atmosphere, it is absorbed by gas molecules 
and scattered by aerosol particles, and at the same time, the 
scattered signals are accepts directly or reflected into sensor. In 
those visible and NIR wavelengths, the atmospheric influence 
is strong enough to modify the reflected electromagnetic signal. 
As a result, the contrast of images is reduced and the 
readability is debased, which increases the difficulties to 
interpretation (Wang Zhao, 2006). Thus, any set of 
CHRIS/PROBA data needs for a cautious removal of the 
atmospheric effects in the initial processing steps, to assure a 
maximal accuracy and reliability in the results inferred by the 
latter exploitation of the data. 
Fig3. Histogram of reflectivity distributing 
At present, the best method to correct the atmosphere effect is 
radiation transfer model, which is based on the radiation 
transfer theory and yields satisfactory results. The methods that 
applied widely are 6S model, LOWTRAN model, MORTRAN