Fluorescent Analytical Method Of Phytoplankton By Genetic Algorithms

Because of the importance of phytoplankton in the marine environment, monitoring phytoplankton classes and their abundances is a routine task in marine scientific research. With the frequent occurrence of red-tide, there is an urgent need for rapid analysis methods that can provide qualitative and quantitative information of phytoplankton in nature aquatic systems. Three-dimensional fluorescence spectra can provide all the fluorescence fingerprint with high sensitivity. In this paper, three-dimensional fluorescence spectra of 26 species in vivo phytoplankton were studied. These phytoplankton are belonging to five divisions, including eight Dinophyta species, nine Bacillariophyta species, five Chromophyta species, two Chlorophyta species and two Cyanophyta species and they are cultivated in lab under a temperature level of 20℃and three illumination intensities(15000 lux, 10000 lux, 6000 lux). Through extracting a number of fingerprint characteristics with Genetic Algorithms (GA), the classification of phytoplankton is refined and a new method, which is based on fluorescence fingerprint characteristics, for rapid classification and identification of phytoplankton is established. The main research results are as follows:1. The Layer-by-layer classification method of three-dimensional fluorescence spectra of in vivo phytoplankton is presented. Layer-by-layer classification method is based on the idea of classifying step-by-step. Characteristics of three-dimensional fluorescence spectra are extracted with Principle Components Analysis, and the first layer spectra characteristics library plus the second layer spectra characteristics library are established. Single species and mixed samples are identified with Non-Negative Least-squares Regression method (NNLS) using this Layer-by-layer classification method. Results show that this method can identify phytoplankton samples in the level of divisions and spectral categories. Since coastal waters and lakes usually have dominating species, this method can be applied to meet the actual needs. This method has many advantages, such as simple, with small spectral libraries, being able to identify fast and choose the classification levels according to the actual requirements.2. The GA classification method of three-dimensional fluorescence spectra of in vivo phytoplankton is presented. GA is used to extract combined characteristics from three directions of the three-dimensional fluorescence spectra, including the horizontal direction (excitation spectra), vertical direction (emission spectra), oblique direction (synchronous spectra). Results show that the combined characteristics of excitation spectra are fingerprint characteristics of three-dimensional fluorescence spectra of in vivo phytoplankton, which can not only effectively distinguish phytoplankton of different divisions, but also present more differences among species. According to the differences of the fingerprint characteristics, the 26 phytoplankton species belonging to five devisions in this paper can be divided into 18 spectral categories, which is much better than former studies. 228 spectra of single species and 405 spectra of mixed samples are identified using the fingerprint characteristics and NNLS. For single-species, recognition accuracies are high, which respectively are 98.5%, 99.6% and 100% on the level of species, spectral categories and divisions. For spectra of mixed samples, the recognition accuracies have markedly increased both on the level of spectral categories and divisions when using the fingerprint characteristics, which are 86.4% and 97.8%. The recognition accuracies have respectively increased 30 percentage points and 10 percentage points than previous studies. They are also much better than using the layer-by-layer method. Therefore, it can be concluded that the fingerprint characteristics obtained by using GA contain effective classification information. It can effectively refined the classification and improve the identification accuracy. Meanwhile, the rusults also verifies the effectiveness of GA in extracting characteristics of three-dimensional fluorescence spectra of phytoplankton.3 A method for quantitative analysis of in vivo phytoplankton, which is base on the chlorophyll a concentration, is set up. In order to assess phytoplankton class aboudance, the relationship between fluorescence spectra intensities of in vivo phytoplankton and chlorophyll a concentration of corresponding sample are studied. A good linear correlation between them is showed(R =0.9146 ~0.9935). According to the obtained linear correlation equations and in vivo phytoplankton fluorescence spectra, chlorophyll a concentrations of all the correct identified samples can be obtained. Relative errors of the chlorophyll a concentrations between the calculated values by the in vivo method and measured values by the in vitro method are calculated. 90% of these relative errors fell between -0.6-0.2, which prove that this method can be used for rapid quantitative analysis of phytoplankton, with a fixed range of relative errors. The detection limit of phytoplankton species by this kind of in vivo method is 0.08μg / l~4.38μg / l, changing along with different phytoplankton species.In sum, this work first time uses GA to extract fingerprint characteristics of three-dimensional fluorescence spectral of in vivo phytoplankton, define the spectral categories and establish a new method for rapid classification and identification of phytoplankton. This method have the prospective application values.