A red tide detecting method using spectral angle mapping technology based on MODIS data

摘要

Remote sensing technique has become one of the most important means of red tide detection. Now, there have been a number of successful applications of red tide detections using remote sensing technology in the world. The present detecting technology basically based on the true color images of the ocean, chlorophyll a (CHL-a)maps and sea surface temperature (SST)maps, however, these methods all have some limitations. Red tides can be cased by numerous kinds of phytoplankton or zooplankton. Some of the red tides will case an apparent color change in the ocean, but some will not. So, in the first method, if there is no change in the ocean color pictures, the red tide will not be detected. CHL-a index will increase when the red tide is cased by phytoplankton, that is also the theoretical basis of the method 2, but it will not work when the red tide is cased by zooplankton. Many researches indicated that red tides can arose a rising in SST, but ocean currents and so many other factors can cause the phenomenon as well. It seems that method 3 maybe not the perfect way too. In this paper, the author suggested a process based on the spectral signature and spectral angle mapping (SAM)method to detect red tides. Spectral signature of red tides has three features. Firstly, there are 2 peaks in the wavelength rage of 500nm-800nm, and the second reflection peak in the wavelength rage of 680nm-710nm generally, is the sign to distinguish a red tide from normal sea water. Secondly, the height of the second reflection peak varies with the dominant species which cause the red tide. Last but not the least, the shape of the spectral signature of a certain kind red tide is almost the same, differences only exist in the modular of spectral vector, and that is the theoretical basis of the SAMbased red tides detection process. AS preprocess, minimum noise transform (MNF)and pixel purity index (PPI)should be done to collect the end-members. Then, we can select the end-members in the n-dimensional visualization of ENVI software. By examining the endmembers' spectral signature, people can tell whether it is the red tide pixel or not, even which dominant species are as well with a complete spectral libraries. Finally, with the SAM method, red tide class can be highlighted. It is also very convenient to do the area calculation and change detection after this kind of classification. In conclusion, the method discussed above can be used to detect all kinds of red tides, and with the high temporal resolution of MODIS, it is possible for us to analyze the relationship of the probable influencing factors of red tides and the change of red tide area based on the change detection. That will be helpful to realize the red tide prediction which could minimize the losses caused by red tides.