Fluorescence Discrimination Technique Based On Weighted Nonnegative Least-squares For The Phytoplankton In The Coastal Waters

Recently, more and more red tides have been occurred in the China Sea, which has attracted more attentions of the society as an important marine environmental pollution and natural disasters. The key to predicting the red tides and reducing its influence is to find out the mechanism how the red tide happens. And the premise of knowing the mechanism how the red tides happens and dies out is the observing of the variability of the phytoplankton community before and after the red tides happens, including the phytoplankton groups, quantity and the species, etc. The in vivo fluorescence methods can provide abundant information and avoid using complicated pre-dispose procedure. With the high sensitivity and selectivity, it can satisfy the requirements of speediness and real time for the red tides survey.In this paper, twenty-five common algae which belong to Dinophyta, Bacillariophyta, Chrysophyta, Cyanophyta, Cryptophyta and Chlorophyta were chosen to culture. The algae were cultured in the laboratory and the 3D fluorescence spectra were measured. The determining technique of the compositions of the phytoplankton community was established using the fluorescence spectra on the level of division (especially for the Dinophyta and Bacillariophyta) using the multi-variable regression method and the weighted nonnegative least-squares. And the discrimination technique on the level of genus when the red tide was happened was established by support vector machines.(1) The establishment of the determining technique of the compositions of the phytoplankton community basing on the fluorescence excitation spectra. Firstly, the Rayleigh scatter in the 3D fluorescence spectra measured was corrected, and the noise in the spectra was wiped off using the S-G smoothing methods. According the report of the literature, the fluorescence excitation spectra at the emission wavelength of 680nm was selected to establish the technique. The precision and reproduction of the spectra were analyzed by calculating the relative standard deviation. And then the Fisher's linear discriminant analysis (FLDA) has been done to test whether the spectra of algae species belonging to different divisions can be differentiated. The mean spectra of every division were used as the classified criterion (norm spectra) according to the results of the precision of the spectra. The linear independence of the norm spectra and influence of irradiance on the norm spectra were also discussed subsequently. The determining technique of the compositions of the phytoplankton community was established with the weighted nonnegative least-squares and multi-variable linear regression methods. The single algae samples, the mixed sample and the in situ samples were test on the level of division using the technique in this chapter, and the right discrimination ratio was 76.8% for Dinophyta, 97.4% for Bacillariophyta, 91.7% for Chrysophyta, 100% for Cyanophyta, 100% for Cryptophyta and 100% for Chlorophyta. The right ratio was 83.6% for the mixed samples, 54.2% for the Maidao enclosure samples and 66.7% for the Jiaozhou Bay samples.(2) The establishment of the determining technique of the compositions of the phytoplankton community basing on the disperse fluorescence excitation spectra. According the derivative analyzes and the development of the LED, the fluorescence excitation spectra at the emission wavelength of 680nm consisted of eleven excitation wavelength was selected to establish the technique. The precision and reproduction of the spectra were analyzed by calculating the relative standard deviation. And then the Fisher's linear discriminant analysis (FLDA) has been done to test whether the spectra of algae species belonging to different divisions can be differentiated intuitively. The mean spectra of every division were used as the classified criterion (norm spectra) according to the results of the precision of the spectra. The linear independence of the norm spectra and influence of irradiance on the norm spectra were also discussed subsequently. The determining technique of the compositions of the phytoplankton community was established with the weighted nonnegative least-squares and multi-variable linear regression methods. The single algae samples, the mixed sample and the in situ samples were test on the level of division using the technique in this chapter, and the right discrimination ratio was 81.6% for Dinophyta, 92.2% for Bacillariophyta, 86.1% for Chrysophyta, 100% for Cyanophyta, 100% for Cryptophyta and 100% for Chlorophyta. The right ratio was 85.2% for the mixed samples, 58.3% for the Maidao enclosure samples and 83.3% for the Jiaozhou Bay samples. It offered the theoretic foundation for the manufacture of fluorescence spectrophotometer using the LED as the light source and can satisfy the requirements of marine supervise station for the red tide.(3) The establishment of the determining technique of the compositions of the phytoplankton community basing on the multi-excitation wavelength fluorescence emission spectra. Firstly, the Rayleigh scatter in the 3D fluorescence spectra measured was corrected, and the noise in the spectra was wiped off using the S-G smoothing methods. According the report of the literature, four excitation wavelengths were selected and the range of emission wavelength is 600-750nm. The multi-excitation wavelength fluorescence emission spectra were constructed in the way of end to end. The precision and reproduction of the spectra were analyzed by calculating the relative standard deviation. And then the Fisher's linear discriminant analysis (FLDA) has been done to test whether the spectra of algae species belonging to different divisions can be differentiated intuitively. The mean spectra of every division were used as the classified criterion (norm spectra) according to the results of the precision of the spectra. The linear independence of the norm spectra and influence of irradiance on the norm spectra were also discussed subsequently. The determining technique of the compositions of the phytoplankton community was established with the weighted nonnegative least-squares and multi-variable linear regression methods. The single algae samples, the mixed sample and the in situ samples were test on the level of division using the technique in this chapter, and the right discrimination ratio was 81.6% for Dinophyta, 96.9% for Bacillariophyta, 72.2% for Chrysophyta, 100% for Cyanophyta, 100% for Cryptophyta and 100% for Chlorophyta. The right ratio was 83.6% for the mixed samples, 62.5% for the Maidao enclosure samples and 66.7% for the Jiaozhou Bay samples. This technique support the usage of marine laser lidar system for the red tide supervise(4) The discrimination methods of red tide algae on the level of species were established using the support vector machines with the 3D fluorescence spectra. Firstly, the Rayleigh scatter in the 3D fluorescence spectra measured was corrected, and the noise in the spectra was wiped off using the S-G smoothing methods. According the report of the literature, the fluorescence excitation spectra at the emission wavelength of 680nm was selected to establish the technique. The precision and reproduction of the spectra were analyzed by calculating the relative standard deviation. And then the Fisher's linear discriminant analysis (FLDA) has been done to test whether the spectra of algae species belonging to different divisions can be differentiated intuitively.When the support vector machines were used, the radial kernel function was selected according the 3D fluorescence spectra. The radius and punish factor were set, and the QP function parameters were defined. Then the quadratic loss function was solved using the function of quadprog in Matlab, and the Lagrange multiplier was calculated and the other parameters were also calculated subsequently. Calculating the symbol function and then judge which species the tested sample should be. And the right discrimination ratio was 96.9%, but when the level of noise was up to 5%, the right discrimination ratio was declined to 75%. The right discrimination ratio was 80.1% for the mixed samples on the level of genus.The determining technique of the compositions of the phytoplankton community basing on the fluorescence excitation spectra, the disperse fluorescence excitation spectra and the multi-excitation wavelength fluorescence emission spectra were established. And the discrimination methods of red tide algae on the level of species using the support vector machines with the 3D fluorescence spectra offered another method for the discrimination of red tide on the level of genus.