| Harmful algal blooms are ubiquitous natural phenomena caused by the excessive growthof phytoplankton. In recent years, more and more red tides have occurred in the marine areasof China and the key bio-species here have rapidly increased. This has led to severe economicinjury, serious risks to the local marine culture and fishing industry, and increased potentialhuman health risks. These problems have drawn significant attention from the scientificcommunity, and call for a rapid and effective method for emergency monitoring ofphytoplankton communities. The three-dimensional fluorescence spectra can give all the‘fingerprint’ information of the algae within the wavelength coverage and is a potentialtechnique for the algae identification, which has drawn more attention for determining thephytoplankton community. Current researches conducted aiming at in vivo fluorescence ofphytoplankton, and large amounts of membrane samples can’t be measured by thesetechniques. It urgently needs to establish the method for the distant waters and analyzing thefilter samples and discriminate the algae population composition on the basis of it. Also, thefluorescence discrimination technique was established based on wavelet in our researchgroup, and there existed some problems such as the poor discrimination of some algae(especially some red tide algae). Here, some techniques were used to extract the featurespectra, and the complementary simultaneous feature spectra database was to be establishedand to perfect the in vivo fluorescence discrimination techniquePart Ι: Based on the study of the composition of phytoplankton population in coastalarea of China,39algae species(most were dominant species or red tide algae) were selectedfor the experiment and cultured in different conditions for the3-D fluorescencemeasurements. Three kinds of mathematical method of wavelet, wavelet packet,2-D waveletanalysis, and some chemometrics methods such as Bayesian analysis, MLR(Multiple Linear Regression) were used together. The pigment extract fluorescence discrimination techniquewas established finally and it had better capabilities of yielding differentiated assessments ofalgae population distributions at the division level and the genus level(when the HABsbroken). Specific research results were as follows:1. The relative standard deviation (RSD) was used as a criteria to analysis the stability of thefluorescence spectra, and the variance (between-column variance and interclassvariance)was used to analyze the otherness between different division and genus level.2. Wavelets analysis(db7, coif2), wavelet packet analysis,2-D wavelet analysis were used toextract the feature spectra of algae species, and the optimal feature spectra were selectedby Bayesian analysis to compose the reference spectra database. Based on the database,three different discrimination fluorescence techniques were established by non-negativeleast squares. Above all techniques were tested: For single samples, the correctlydiscriminating ratios(CDRs) were95.5%，95.4%，95.5%and94.3%，with the relativecontent were89.4%，87.6%，88.8%，87.3%at the division level, respectively；And theCDRs were89.3%，89.1%，90.1%and87.8%at the genus level, respectively. Forsimulative mixtures, when the mixed proportion was50%, the CDRs could be greaterthan90%except Xanthophyta by the wavelet and the wavelet packet techniques; and thediscrimination was a little poor by the2-D technique for the Xanthophyta andCyanophyta. When the mixed proportion was60%, the CDRs could be greater than85%except Xanthophyta by the three techniques. For the discrimination at the genus level,when the proportion of the dominant species reached80%, the dominant algae speciescould be recognized and the CDRs was greater than75%. When the proportion of thedominant species reached90%, the CDRs was greater than80%and20red tide algaecould be recognized correctly with the CDRs of90%.3. The db7fluorescence technique was used to analyse the filter samples, the results wereconsistent with the results of HPLC-CHEMTAX for the discrimination of Diatom.However, it couldn’t discriminate the Cryptomonas correctly. When the feature spectra database was expanded by adding the field spectra and to analyze the sample, thediscrimination results were improved especially for the discrimination of Cryptomonas. Itclearly seen that the fluorescence technique could be used to analyze the filter samples ofthe phytoplankton, and if more and more field sample spectra were obtained to update thefeature spectra database, the technique would be gain better discrimination results.Part Π:53algae species(most were dominant species or red tide algae) were selected forthe experiment and cultured in different conditions for the3-D fluorescence measurements.Three kinds of mathematical methods of wavelet, wavelet packet,2-D wavelet analysis, andsome chemometrics methods such as Bayesian analysis, Cluster analysis and MLR(MultipleLinear Regression) were used together, three in vivo fluorescence discrimination techniquewere established, and the complementarities between the different fluorescence techniquewas discussed to establish the simultaneous discrimination technique which had bettercapabilities of yielding differentiated assessments of algae population distributions at thedivision level and the genus level(when the HABs broken). Specific research results were asfollows:1. Wavelets analysis(db7, coif2), wavelet packet analysis,2-D wavelet analysis were usedto extract the feature spectra of algae species, and three different discriminationfluorescence techniques were established by non-negative least squares. And thecomplementarities were discussed. And the simultaneous discrimination technique wasestablished finally and the db7norm spectra database was the first database of thediscrimination and other feature spectra were construct the second data database.2. The simultaneousness spectra database technique was tested: for the single samples, theaverage CDRs was96.0%at the division level, the CDRs was87.4at the genus level.The algae species such as Oc，Db，Ld，Rh，Ks，Pl，As, the CDRs of which were increase16.7,10，6.8，11.7，14.2，16.7，33.3percentage point. For the simulative samples whenthe algae dominance were60%,70%,80%,90%, the CDRs were89.3%,93.8%,96.1%and96.9%at the division level, with the average relative contents of58.4%,68.7%, 77.5%, and86.1%, respectively; the CDRs were64.7%,92.6%,93.8%and93.9%,respectively at the genus level. In which, Cf (60%dominance) and (60%dominance) Dbwere increase31.2and35.2percentage point, respectively.3. For24particular field samples from Jiaozhou Bay and Enclosure experiment, the resultsof23samples were consistent with the microscopic examination at the division level;and5samples from8samples were successfully recognized at the genus level which thealgae dominance achieved80%of the total biomass.4. For particular applications in Bohai sea and Jiaozhou Bay, it was possible to estimate thephytoplankton community composition and relative abundance of different classes, andthe whole results agree with that of published papers.The innovation of this paper is: Based on3-D fluorescence spectra of the phytoplanktonpigment extract, three kinds of mathematical method of wavelet, wavelet packet and2-Dwavelet analysis were synthetically applied to establish the pigment extract fluorescencediscrimination technique. The technique could meet the demand of the distant waters andanalyzing large amounts of membrane samples. Another innovation was based on the in vivofluorescence spectra of phytoplankton, three kinds of mathematical method of wavelet,wavelet packet and2-D wavelet analysis were synthetically applied to extract the featurespectra and the complementarities were studied. The complementary simultaneousdiscrimination technique was established finally and the CDRs were improved for somespecies especially for some red tide algae. The technique not only can discriminate algaepopulations at division level, but also can identify the algae species causing harmful algaeblooms(HAB) at genus level when HAB happens. |
PDF Full Text Request