A New Approach on Operational
Offshore Suspended Sediment Algorithm for Satellite Images
Ming Deng1, Yan
Li1, Jujie Yang2, Jin Li2, Shujing
Li1 Key words: remote sensing, suspended
sediment concentration, slope algorithm. 1 the Second Institute of Oceanography, Hangzhou, 310012 2 Peking University, Beijing, 100871 Abstract Remote sensing in detecting offshore-suspended sediment is one of the important tasks of ocean research. Currently most achievements in this field focus on the low SSC (Suspended Sediment Concentration). However, in China the offshore environment is characterized turbid water with high SSC. This background makes the research in high SSC becomes a special and urgent task of China. In marine remote sensing, how to remove atmosphere effect is a crucial and necessary step because only about 30% of the single detected by sensors is valid (Hovis and Leung, 1997). And the most difficult point is high SSC remote sensing is it too. Recently a new method, slope algorithm, is put forward by LiYan and Lijin (1998, 1999) to solve this problem. To put it into operational system and test this algorithm, we select the coast of East China Sea, including Chandjiang Estuary and Hangzhou Bay which are famous for their high SSC, as our study area, NOAA14/AVHRR CH1 and CH2 as data source and realize this algorithm through computer program. Preliminary long time series of SSC images have been acquired successfully. In this paper we will illustrate the following aspects in detail: general construction, solution to the difficult and key points, discussion about its feasibility, stability and application. 1 Fundamental about the slope algorithm The water-leaving reflectance of sea surface changes greatly in the transmitting process. If consider two different bands, it can be proved that the change in the slope of their relation curve is linear in the transmitting process, namely the relation curve of water-leaving reflectance is similar with that of the reflectance detected by remote sensor. It is just through the relation between slopes, atmospheric effects, including high SSC areas, can be corrected. And the next, through the relation between water-leaving reflectance of sea surface and SSC to deduce the relation between slope and SSC, thus SSC can be calculated out (Li Yan, Li Jin, 1998; 1999). 2 Basis for programming algorithm design The key step in carrying out this algorithm on computer is how to get the relationship curve of CH1/CH2 and corresponding slope from NOAA/AVHRR CH1 and CH2 images. The following basic facts are the basis for the computer algorithm design:
In practice, we take two methods, simulation method and max method. Fig.1 CH1_CH2 relation curve (1997/10/16) Fig. 2 the influence of atmospheric asymmetry. 3. Simulation 3.1 Basic steps of program This method directly uses spatial local area and gray level local area to acquire slope K. The basic steps are;
See fig.3. It is the SSC result image on 16th, Oct., 1997. Our results show that the stability of slope is under the control of both spatial local area and gray level local area. To spatial local area, the smaller the block is, the better the result is. Because in small block the state of atmosphere could be considered as symmetry, which better ensures the accuracy of slope. But to gray level local area, the larger the block, the better. Because to simulate curve need enough statistic data. The conflict between the two local areas is the key and difficult point of the simulation method. Presently, the following measures are taken:
4 Maximum method 4.1 Basic steps of program Maximum method (LiYan, Lijing, 1998; 1999 ) is an in directed method to get slope. It can be simply proved: The slope (K) of CH1~CH2 relation curve can be expressed as: K=dCH2/dCH1 Namely K-dCH2/dCH1=0 Because K is a constant, there has: d(K*CH1-CH2)/dCH1=0 This equation is just the discriminate function of the K*CH1-CH2 extremum. It can also prove this extremum is a maximum. Thus to get slope K could convert to get the max of K*CH1-CH2. Namely the distribution of K is equivalent to that of the max of K*CH1-CH2. What we do in the practice is similar to that in the simulation method only different at the way we get slope. In every block, through calculating the max of K*CH1-CH2 to build the relation among K~CH1, CH2. Here K is preset (e.g., K= 0.02, 0.04 …….5 a arismetic series with step of 0.02). 4.2 Example and discussion See fig.4. It is the SSC result image using maximum method on 16th , Oct., 1997. Through the comparison of a number of result images, we find that:
5. Error Comparison A resulting the Changjiang Estuary shows that the maximum method is better than simulation method (See Fig.5). and we can see that to high SSC, the error of simulation method is apparent mainly caused by the fact 4 that we have stated before. And it means that the difference of two methods will enlarge within certain area. Fig 5. Comparison of calculating data with true data 6. Conclusion Close to one hundred NOAA14/AVHRR CH1 and CH2 images, receiving 1997 and 1998, have been dealt with the Simulation Method and Maximum Method for the Slope algorithm. Compared with our accumulated historical data of the past experience (notice that it is difficult to get synchronous in situ data) most results are satisfying. Of course this new algorithm need much more in situ data to firmly confirm it. We also notice the following problems: how to decrease the influence effectively when thin clouds or fog cover most of the research area; how to increase the sensitivity tot the very high SSC water. References:
|