A Technique For Assessment Of Phytoplankton Class Abundance Developed By 3-d Fluorescence Spectra And Multiwavelet

Red tides occur frequently around the world which can be harmful to both fish and humans.It is reported that in recent 60 years, there are about 265 times red tide accidents in marine coastal area of China Sea, especially in East China Sea. For that reason, it is urgent to establish a fast, highly sensitive and low-cost technique for determination of phytoplankton community composition. The in vivo 3-D fluorescence spectra can give the "fingerprint" information specific to each class of phytoplankton, and the fluorometric method can be sensitive and simple, which make it possible to be used as an in situ detection technique for phytoplankton population.Based on the research of the composition of phytoplankton community in marine coastal area of China, the thesis chose 43 phytoplankton species which occurred as predominant ones or frequently caused harmful algae blooms in recent years in the East China Sea and cultured them in laboratory. Those phytoplankton species belong to 32 genera of 7 divisions,among which 14 species of 9 genera are from Bacillariophyta,6 species of 6 genera from Chrysophyta,4 species of 4 genera from Chlorophyta,11 species of 6 genera from Dinophyta,2 species of 2 genera from Xanthophyta,2 species of 1 genus from Cryptophyta, and4 species of 4 genera from Cyanophyta.In the lab, those phytoplankton species were cultured under different conditions and in-vivo 3-D fluorescence spectra were obtained, then multiwavelet analysis and NNLS technique were utilized to develop a fluorometric technique which not only can differentiate the phytoplankton populations at division level, but also can identify the species causing harmful algae blooms (HAB) at genus level when HAB happens.The results are as follows:1.The thesis selected two multiwavelet functions (GHM multiwavelet and SA4 multiwavelet) to decompose the fluorescence spectrum data, thus characteristic scale vectors and wavelet vectors were given. Bayesian discriminant analysis was applied to choose the proper vectors for discrimination among those vectors,and the Ca2, Ca3 and Ca2-Ca3 were selected as primary characteristic spectra. 2.The database of reference spectra was established by application of the systematic cluster to the Ca2,Ca3 and Ca2-Ca3 of GHM multiwavelet and SA4 multiwavelet at the division level and the genus level respectively. Then the method for differentiation of phytoplankton populations was developed with multivariate linear regression resolved by nonnegative least squares.The reference spectra of two multiwavelets were applied to identify the samples composed of one phytoplankton species and simulative mixed samples both at division level and genus level, and the optimal reference spectra was selected based on the results of identification. For GHM multiwavelet, the Ca2-Ca3 reference spectra gave the best results,94.4% and 92.7% of the samples composed of one phytoplankton species were discriminated correctly at the division level and genus level respectively. And for the SA4 multiwavelet, the Ca2 reference spectra was the best, for the samples composed of one species,94.6% were recognized at the division level and 92.7% at the genus level.In order to test the performance of the fluorometric technique, lots of simulative mixed samples were made and analyzed:the fluorescence spectra from different phytoplankton classes which scattering has been removed from and were normalized were mixed at four different ratios:3:1,4:1,6:1,9:1.The results indicated that when used GHM multiwavelet, the average recognition rates of 97.6% to 97.9% at division level and 66.8% to 90.7% at genus level were achived respectively, for SA4 multiwavelet, the average recognition rates of 98.3% to 99.1% at division level and 63.1% to 91.6% at genus level were given respectively under different proportions.When the proportion of the dominant group is no less than 4:1,the average recognition rates of 95% at division and 70% at genus level were achived at least respectively;and when the ratio of the dominant group is up to 6:1,at least the average recognition rates of 80% can be achived at genus level.3.As to the samples composed of two phytoplankton species, the average recognition rates of 90.1% at division level and 76.4% at genus level were given based on the Ca2-Ca3 reference spectra, and as for the SA4 multiwavelet, they could be recognized by 90.9% at division level and 74.2% at genus level respectively based on the Ca2 reference spectra. Besides, for the samples form the Jiaozhou Bay and the seawater of mesocosm,90% at division level and at least 80% at genus level can be identified based on the reference spectra of the two multiwavelets.The innovation of this thesis is to utilize multiwavelet to extract the characteristics specific to each phytoplankton class from the 3-D fluorescence spectra and develop a fluorometric technique which not only can differentiate phytoplankton populations at division level, but also can identify the phytoplankton species causing harmful algae blooms (HAB)at genus level when HAB happens.