Resumes & CVs

A fluorometric method for the discrimination of harmful algal bloom species developed by wavelet analysis

Description
An in vivo 3D fluorescence discrimination technique for ten harmful algal bloom (HAB) species that belong to eight genera of four divisions was developed by wavelet analysis. Daubechies-7 (db7) was employed as the mother wavelet. The fifth scale
Categories
Published
of 7
All materials on our website are shared by users. If you have any questions about copyright issues, please report us to resolve them. We are always happy to assist you.
Related Documents
Share
Transcript
  A  󿬂 uorometric method for the discrimination of harmful algal bloom speciesdeveloped by wavelet analysis Fang Zhang a,b , Rongguo Su a , Xiulin Wang a, ⁎ , Liang Wang c , Jianfeng He b ,Minghong Cai b , Wei Luo b , Zhixia Zheng d a College of Chemistry and Chemical Engineering, Ocean University of China, Qingdao 266100, PR China b SOA Key Laboratory for Polar Science, Polar Research Institute of China, Shanghai 200136, PR China c College of Information Science and Engineering, Ocean University of China, Qingdao 266100, PR China d Laboratory of Environmental Engineering, School of Chemistry and Materials, University of Science and Technology of China, Hefei 230026, PR China a b s t r a c ta r t i c l e i n f o  Article history: Received 17 January 2008Received in revised form 29 September 2008Accepted 1 October 2008 Keywords: Classi 󿬁 actionDiscriminationFluorometric methodHarmful algal bloomScale vectorsWavelet analysis An in vivo 3D 󿬂 uorescence discrimination technique for ten harmful algal bloom (HAB) species that belong toeight genera of four divisions was developed by wavelet analysis. Daubechies-7 (db7) was employed as themother wavelet. The  󿬁 fth scale domains were selected as the discriminant characteristic spectra (DCS). Basedon the DCS, The phytoplankton species at different growth stages were classi 󿬁 ed correctly at both thedivision and the genus level by Bayesian discriminant analysis (BDA). Based on the reference spectra of theDCS, the discrimination method of the phytoplankton species was established by the nonnegative leastsquares (NNLS) method. The correct discrimination ratios (CDRs) for samples of the single species were 96.1%with 0% blank noise and 93.3% with 10% noise at the genus level, while the CDRs were both 100% with 0% or10% blank noise at the division level. When blank noise was up to 20%, the CDRs were down to 85.2% at thegenus level and 98.0% at the division level. For the mixture samples, the CDRs of the dominant species were98.3% and 96.3%, respectively, at the division level and at the genus level. As dominant species,  Prorocentrumminimum  ( Pm ),  Gymnodinium simplex  ( Gs ),  Scrippsiella trochoidea  ( Sc  ),  Skeletonema costatuma  ( Sk ),  Chaeto-ceros  ( Cu  and  De ),  Phaeocystis globosa  ( Cg  ) and  Chlorella pynnoidosa  ( Ch ) can be correctly discriminated atboth the division level and the genus level.© 2008 Elsevier B.V. All rights reserved. 1. Introduction Phytoplankton is the main source of primary production in bothmarine and freshwater habitats (Lee et al.,1995). Some of the speciesoccasionally grow very fast or  “ bloom ”  and accumulate into dense,visible patches near the surface of water. They destroy the zoologicalbalance of the water. Worse than that, some of the species producepotent neurotoxins that can be transferred through the food webwhere they affect and even kill the higher forms of life such aszooplankton,shell 󿬁 sh, 󿬁 sh,birds,marinemammalsandevenhumans.This phenomenon is called harmful algal blooms (HABs). Most HABsare caused by only one algal species. When HABs break out, thebiomass of a certain causative species of HABs is more than 90% of thephytoplankton populations (Millie,1997; Johnsen et al.,1994).In recent years, HABs have occurred frequently. In order todiscriminate and determine the algal species causing HABs in situ,rapidly and reliably, many techniques such as absorption spectra(Laurion et al., 2003; Briacud et al., 2004), high performance liquidchromatography (Wong and Wong, 2003; Riegman and Kraay, 2001),confocal laser scanning microscopy (Larson and Passy, 2005) and 󿬂 uorescence spectra (Lee et al., 1995; Beutler et al., 2003; Liu et al.,2005; Zhang et al., 2006, 2007) have been developed. Among thesemethods,  󿬂 uorometric methods become more and more common forthey can work in vivo and in situ (Poryvkina et al., 2000).It is known that phytoplankton belonging to different divisionscontain different kinds of pigments, which are bound to differentproteins in a living cell. Various pigment-protein units in a cell havedifferentstructures andplay differentroles inphotosynthesis (Beutleretal.,2003).Thepigmentcompositionisonemajorbasisoftaxonomy.The shape of   in vivo  󿬂 uorescence spectra of phytoplankton speciesdepends on the pigment composition to a great extent. According tothe ratio of   󿬂 uorescence intensity belonging to chlorophyll  a  andaccessory pigments, Yentsch and Phinney (1985) divided phytoplank-ton species into four groups, i.e., diatoms/dino 󿬂 agellates/coccolitho-phores, green unicells, crytomonads/redsandcyanobacteria.Basedonthe  󿬂 uorescence excitation spectra with  󿬁 ve excitation wavelengths,the Algae Analyser produced by the bbe-Moldaenke Company candifferentiate phytoplankton species into four classes, i.e., green algae,blue green algae, cryptophyceae and brown algae. The content of chlorophyll  a  of these four classes can be determinated. Liu et al.  Journal of Experimental Marine Biology and Ecology 368 (2009) 37 – 43 ⁎  Corresponding author. Tel.: +86 532 66782143; fax: +86 532 82031799. E-mail address:  surongguo@ouc.edu.cn (X. Wang).0022-0981/$  –  see front matter © 2008 Elsevier B.V. All rights reserved.doi:10.1016/j.jembe.2008.10.004 Contents lists available at ScienceDirect  Journal of Experimental Marine Biology and Ecology  journal homepage: www.elsevier.com/locate/jembe  (2005) developed a  󿬂 uorometric discrimination technique for a freshwater alga,  Chlorella vulgaris , by synchronous  󿬂 uorescence spectra.Zhang et al. (2006) utilized principal component analysis (PCA) toextract the characteristics from 3D  󿬂 uorescence spectra of Cyano-phyta, Chrysophyta, Chlorophyta, Dinophyta and Bacillariophyta. Themajor pigments are common to the phytoplankton species belongingto the same division, while the minor ones are always speci 󿬁 c. Also,the bonding modes of the pigments to proteins are unique (Millie,1997). 3D  󿬂 uorescence spectra can provide more information thaneither emission spectra or excitation spectra, including pigmentcontent and the bonding modes of the pigments (Zhang et al.,2007). This makes it possible todifferentiate phytoplankton speciesatthe genus, or even the species level.Wavelet analysis can decompose signals to different time domainsandscaledomains.Itusesabasisfunctionasthemotherwavelet,witha certain scale (the width of the analyzing function window) toinvestigate the time-scale properties of an incoming signal. Bychanging the width of the functionwindow, both speci 󿬁 c and generalproperties of the signal are captured. Probing a signal according toscale is a very useful approach to get de-noised characteristic featuresRamos et al., 2007). Besides the utilization in processing  󿬂 uorescencesignals(Erikssonetal.,2000;Olivo-Marin,2002),waveletanalysis hasalso been utilized in the data treatment of high performance liquidchromatography (Dinç et al., 2005; Beltrán et al., 2006), Ramanspectra (Ramos et al., 2007), infrared spectra (Cocchi et al., 2005; Vannucci et al., 2005) and mass spectra (Vannucci et al., 2005) as well as in other analytical methods. However, to our knowledge, thephytoplankton has not been classi 󿬁 ed or discriminated to the genus(species) level by 3D  󿬂 uorescence spectra, nor has wavelet analysisbeen utilized in the discrimination of phytoplankton species.In the present work, a  󿬂 uorometric method was developed for thediscrimination of causative species of HABs by wavelet analysis andmultiple linear regression resolved by nonnegative least squares(NNLS) method (Groenen et al., 2000; Haimovich et al., 2006). 2. Material and methods  2.1. Phytoplankton cultures Ten phytoplankton species belonging to eight genera of fourdivisions were obtained fromthe Marine Pollution Eco-chemistry KeyLaboratory in the Ocean University of China. These species, recordedastheHABcausativespeciesintheEastChinaSea,arelistedinTable1.Each species was cultured with growth medium f/2 in 1 L conical 󿬂 asks. They all have three replicative cultures and were kept in acultivating machine. The culturing temperature was 20 °C, and theirradiances were 87.6, 116.8, and 146 Wm -2 . The light:dark cycle was12:12 h. The culture period was 15 days for all ten species, except for Ps , which was cultured for 10 days, because it has a shorter growthperiod. Laboratory mixtures were made up of these species. Thebiomass of the dominant species accounted for 75% or 80% that of themixture.  2.2. Fluorescence measurements The 3D  󿬂 uorescence spectra of each culture were determinedevery three days (every two days for  Ps ) using a F4500  󿬂 uorescencespectrophotometer (Hitachi, Japan) at room temperature whichchanged from 18-22 °C. Both the slit width and the wavelengthintervals were 5 nm; and the scanning speed was 200 nm/s. Theranges of the excitation and emission wavelengths of the sampleswere 400 – 600 and 500 – 750 nm, respectively. All 3D  󿬂 uorescencespectra were obtained under such measuring conditions. All 3D 󿬂 uorescence spectra were obtained under these measuring condi-tions. Each sample had three replicates. The concentration of chlorophyll  a wasdeterminedusingabbeAlgaeAnalyser(Moldaenke,Germany).  2.3. Data processing  2.3.1. Data pre-processing  All the 3D  󿬂 uorescence spectra were processed using MATLAB 6.5software.Eachofthespectrawaschangedintoamatrixwith31rowsand41 columns. The Rayleigh and Raman scattering were removed by linearinterpolation (Zepp et al., 2004). Each row of the 3D  󿬂 uorescencespectrum was connected  ‘ end to end ’  to form a big emission spectrum(BEms). In the same way, a big excitation spectrum (BExs) was obtainedby connecting each column of the 3D  󿬂 uorescence spectrum. All thespectra were standardized by the calculation  x ⁎ =  x −  x min  x max −  x min ;  in which  x i  is every data point in the spectra ;  x max  =  max i  x i ð Þ  and  x min  =  min i  x i ð Þ .  2.3.2. Analysis of the stability of the 3D  󿬂 uorescence spectra The stability of the 3D  󿬂 uorescence spectra of each phytoplanktonspecies, cultured under different lights or at different growth stages,was analyzed according tothe Included Angle Cosine (IAC), which canbe utilized as a criterion of the similarity between different spectraThe results are shown in Fig.1.When the stability of the  󿬂 uorescence spectra of the phytoplank-tonspeciesculturedunderdifferentlightconditionswasanalyzed,theIAC was calculated between the  󿬂 uorescence spectra of onephytoplankton species at the same growth stage cultured at146 Wm -2 and 87.6 Wm -2 , or of this species at one growth stagecultured at 146 Wm -2 and one of the other species at thecorresponding growth stage cultured at 87.6 Wm -2 . The averages of the IAC of one species or of one species with one of the other specieswere obtained. The results are shown in Fig. 2.  2.3.3. Decomposition of the 3D  󿬂 uorescence spectra by wavelet analysis Daubechies-7 (db7), an orthogonal wavelet function, wasemployed as the mother wavelet. The scale vectors and time vectorswere obtained.The process can be described by the following mathematicexpressions.In L2(R), an orthogonal wavelet basis is obtained by dilating andtranslating a mother wavelet  ψ  as  ψ  jk (  x )=2  j /2 ψ (2  j  x - k ) with  j ,  k integers. A function  f   can then be represented by the wavelet series  f  (  x )= Σ   j , k ∈  Z  d  jk ψ  jk (  x ), with wavelet coef  󿬁 cients  d  jk =  ∫   f  (  x ) ψ  jk (  x ) dx  describ-ing features of the function  f   at the spatial locations indexed by  k  andscalesindexedby  j .Onespectrumof  Pm istakenforanexample;someof the scale vectors of both the BExs and BEms are shown in Fig. 3.  2.3.4. Selection of the discriminant characteristic spectra (DCS) The DCS were selected from the scale vectors by Bayesiandiscriminant analysis (BDA) (Han and Tang, 2003; Vannucci et al.,2005; Gao and Davis, 2006). The results of the differentiation of thephytoplankton species DCS were obtained among different culturingrepliccates, different growth stages and different culturing lights. Thedifferentiating results of all phytoplankton species based on the DCSbyBDAareshowninTable2.ThestandardizedDCSareshowninFig.4.  Table 1 Phytoplankton species included in the investigationSpecies Abbreviation Genus Division Prorocentrum dentatum Pr Prorocentrum  Dinophyta Prorocentrum minimum PmGymnodinium simplex Gs GymnodiniumScrippsiella trochoidea Sc ScrippsiellaPseudo-nitzschia pungens Ps Pseudo-nitzschia  Bacillariophyta Skeletonema costatuma Sk SkeletonemaChaetoceros curvisetus Cu ChaetocerosChaetoceros Debilis DePhaeocystis globosa Cg Phaeocystis  Chrysophyta Chlorella pynnoidosa Ch Chlorella  Chlorophyta38  F. Zhang et al. / Journal of Experimental Marine Biology and Ecology 368 (2009) 37  – 43  The stability of the DCS was analyzed in the same way as theanalysis described in 2.2.2, the results are shown in Figs. 5 and 6.  2.3.5. Discrimination of the phytoplankton species Based on the DCS, cluster analysis (Vassiliou et al., 1989) wereutilized to extract reference spectra of each phytoplankton species.Thesereferencespectrawereutilizedtoestablishadatabase.Basedonit, the discrimination of both the single species and the laboratorymixtures was done by multiple linear regression resolved by NNLSmethod.This is a problem to calculate the nonnegative coef  󿬁 cient  x i  of aknown variables  a i  in the function of   b =  x 0 + a 1  x 1 + a 2  x 2 + … + a i  x i + e . Inthe systems studied in the present work,  b  is the spectral vector of anunknown phytoplankton system;  a i  is the reference spectral vector of the  i th phytoplanktonspecies;  e  is the residual between the realvalueand the estimated one of   b ;  x i  is the quotient of the reference spectralvectorof the  i th phytoplanktonspecies. The discrimination results areshown in Table 3.NNLS was used to solve the multiple regression models by thecalculation of   min  x z 0  q x ð Þ =  12  kj  Ax − b kj 22 , where  A ∈ IR m × n ,  b ∈ IR m ,  m  aregiven and  m ≥ n . 3. Results  3.1. Analysis of the stability of the 3D  󿬂 uorescence spectra AsshowninFig.1,forthesephytoplanktonspecies,theaverageIACbetween the 3D  󿬂 uorescence spectra of one species cultured underdifferent light conditions are in the range of 0.992 to 0.998. It alsoshows that when cultured under different lights, the IAC between the3D 󿬂 uorescencespectraofdifferentspeciesareintherangeof0.842to0.997. All of them are less than those of the same species.The average IAC between the 3D  󿬂 uorescence spectra of thephytoplanktonspecies,exceptfor Pm ,atdifferentgrowthstagesareinthe range of 0.956 to 0.990 (Fig. 2). For  Pm , the shape of the spectraobtained at the growth stage of the 3rd day was found to be differentfrom that on other growth stages. Without the consideration of thisgrowth stage, the average IAC of the 3D  󿬂 uorescence spectra of   Pm  is0.999. The average IAC between the 3D  󿬂 uorescence spectra of different species are in the range of 0.852 to 0.998. They are still lessthan those between the 3D  󿬂 uorescence spectra of one species atdifferent growth stages, except for the ones between  Cu  and  De . Thereis little difference between the IAC of them.  3.2. The establishment of the DCS  3.3. The classi  󿬁 cation and the stability of the DCS  The results of BDA (data not given) show that the scale vectors of both the BEms and BExs (BExs-Ca5 and BEms-Ca5) are useful forclassifying phytoplankton species. Consequently, the utilizations of themwerechosenastheDCS(Fig.4).AsshowninTable2,toallthe10 phytoplankton species, except for  Cu  and  De , the CCRs are more than93.0% under the  󿬁 rst three classifying conditions. Their CCRs are even100%totheclassi 󿬁 cationofonespeciesatdifferentgrowthstages.TheCCRs are more than 99.0% at the genus level and no less than 99.9% atdivision level.  Cu  and  De , which belong to the same genus, are easilyrecognized as each other. They would be treated as one species( Chaetoceros ) in this method.For these phytoplankton species, the average IAC between the DCSof one species cultured under different light conditions are in therange of 0.995 to 0.999 (Fig. 5). When cultured under different lights,the IAC between the DCS of different species are in the range of 0.991to 0.998, which is less than that of the same species.The average IAC between the DCS of the phytoplankton species atdifferent growth stages arein the range of 0.996 to 1.000 between theDCS of one species, except for  Pm  (Fig. 6). The growth stages also havesome in 󿬂 uence on the shape of the DCS of   Pm . Still, without theconsideration of the 3rd day, the average IAC of the DCS of   Pm  atdifferent growth stages is 0.999. The average IACs between the DCS of different species are in the range of 0.919 to 0.998. Just like that of  Fig. 2.  The IAC between the 3D  󿬂 uorescence spectra of one phytoplankton species or different species at different growth stages. Fig.1.  The IAC between the 3D  󿬂 uorescence spectra of one phytoplankton species or different species cultured under different light conditions.39 F. Zhang et al. / Journal of Experimental Marine Biology and Ecology 368 (2009) 37  – 43  different light conditions, the value is still less than that from the DCSof the same species. However, the IACs between  Cu  and  De  are stillalmost the same.  3.4. The discrimination of phytoplankton species 33 reference spectra were obtained in all to these 10 species. Tothe samples of single species, the CDRs are more than 90% with 0% or10% blank noise for all the 9 species except for  Ps , and the CDRs of   Pm , Gs ,  Sc, Cg   and  Ch  are 100%. For  Ps,  the CDRs are more than 80% with0% or 10% blank noise. When the blank noise reaches 20%, many algalsamples cannot be discriminated correctly at the genus level. Thismakes the average CDR of the genus level down to 85%. At thedivision level, the CDRs are more than 98% with 0%,10% or 20% blanknoise. Fig. 3.  The decomposition of the  󿬂 uorescence spectra by db7: (a) BExs: BExs-Ca3, BExs-Ca4, BExs-Ca5, and BExs-Ca6 present for the third, fourth,  󿬁 fth and sixth scale vectors of theBExs, respectively; (b) BEms: BEms-Ca3, BEms-Ca4, BEms-Ca5, and BEms-Ca6 present for the third, fourth,  󿬁 fth and sixth scale vectors of the BEms, respectively.40  F. Zhang et al. / Journal of Experimental Marine Biology and Ecology 368 (2009) 37  – 43  For mixture samples, almost all the dominant species of thesamples can be discriminated (Table 3(b)). The CDRs of   Pm ,  Gs ,  Sc  ,  Sk , Chaetoceros ,  Cg   and  Ch  are 100%. The CDRs at the division and thegenus level are 98.1% and 96.3%, respectively. 4. Discussion 4.1. The analysis of the stability of the 3D  󿬂 uorescence spectra Phytoplankton  in vivo  󿬂 uorescence depends not only on thetaxonomic position, but also on the photoadaptation state. Both theculture light conditions and the growth stages of the phytoplanktonhave an in 󿬂 uence on the excitation  󿬂 uorescence spectra (Poryvkinaetal.,2000). Inordertotestthein 󿬂 uenceof thetwoconditionsonthe3D  󿬂 uorescence spectra, the recurrence of the spectra of thephytoplankton under different light conditions (87.6 and 146 Wm -2 ),as well as at different growth stages (the 3rd-15th days under theculture light of 146 Wm -2 ) was analyzed. All the 10 phytoplanktonspeciesseemtohavehightolerance tothechangeof theexperimentallight conditions. The in 󿬂 uence of the light condition cannot overstepthe difference between different species. The 3D  󿬂 uorescence spectraof different phytoplankton species show different stability on thechange of growth stage under high culture light. Most of them showgood stability. Like the comparison between different culturing lights,growth stages under our experimental conditions cannot confuse the3D 󿬂 uorescencespectraatthespeciesorgeneralevel.Theonesbelongtothesamegenusareprobablymoresimilar.Alongwiththechangeof culturinglightandgrowthstages,sometimesitmaybemoresimilartothe one belonging to the same genus than to itself, such as  De  and  Cu .Consequently, most of the 3D  󿬂 uorescence spectra of the phytoplank-ton species show high light and growth stability at genus or specieslevel under our experimental conditions. This makes it possible todiscriminate phytoplankton at genus or even species level.The characteristics exist in the original spectra, but not veryobvious (Fig. 3). A method is needed to extract and enlarge thesecharacteristics. Wavelet called  ‘ the microscope in mathematics ’  isexactly the one. It can decompose the srcinal signals (spectra) todifferent space and make different characteristics emerge. Scalevectors have the trait of stable and de-noise. Hence, the DCS wereselected from them. BDA is a classifying method. It can be used to test  Table 2 The BDA results of the ten phytoplankton species by DCSCCRs (%)  Pr Pm Gs Sc Ps Sk Cu De Cg Ch  Gen. Div.1 100 94.4 100 98.9 100 100 68.5 76.4 100 100 99.4 1002 100 93.3 98.9 100 98.9 100 66.3 73.0 100 100 99.1 99.93 100 100 100 100 100 100 2.78 66.7 100 100 100 1004 100 100 65.2 100 62.2 100 92.1 14.6 100 100 92.7 99.9 ⁎ 1standsfortheclassi 󿬁 cationofdifferentmeasuringreplicatesfromoneculturingreplicates.2standsfortheclassi 󿬁 cationofdifferentculturingreplicatesfromonespecies.3standsfor the classi 󿬁 cation of one species at different growth stages. 4 stands for the classi 󿬁 cation of one species under different culture lights. Gen. and Div. stands for the genus and thedivision level, respectively. Fig. 4.  The DCS of the 10 phytoplankton species. Fig. 5.  The IAC between the DCS of one species or different phytoplankton species cultured under different conditions.41 F. Zhang et al. / Journal of Experimental Marine Biology and Ecology 368 (2009) 37  – 43
Search
Similar documents
View more...
Related Search
We Need Your Support
Thank you for visiting our website and your interest in our free products and services. We are nonprofit website to share and download documents. To the running of this website, we need your help to support us.

Thanks to everyone for your continued support.

No, Thanks