On-line analysis of algae in water by discrete three-dimensional fluorescence spectroscopy

Nanjing Zhao; Xiaoling Zhang; Gaofang Yin; Ruifang Yang; Li Hu; Shuang Chen; Jianguo Liu; Wenqing Liu

doi:10.1364/OE.26.00A251

1. Introduction

The eutrophication of reservoirs, rivers, lakes and alongshore sea areas in China has resulted in the excessive proliferation of phytoplankton, possibly causing red tides and cyanobacteria blooms in large area, affecting the ecological environment and threatening human health and drinking water safety. The structures of phytoplankton population vary with water qualities and seasons, and meanwhile the fluorescence efficiencies are significantly different among different species of phytoplankton, so the total chlorophyll a concentration does not fully reflect the phytoplankton biomass in natural water body. In contrast, the classification monitoring is more accurate in indicating the phytoplankton biomass, and is therefore more reasonably used to on-line measure and real-time analyze phytoplankton. The conventional laboratory instruments are difficult to achieve on-line classification of phytoplankton because of the varying structures of phytoplankton population for different water qualities and seasons. With high sensitivity, small interference and good identification performance, in vivo fluorescence spectroscopy technique has been widely used to monitor phytoplankton in recent years. The contents of chlorophyll a are calculated by measuring the in vivo fluorescence spectra of chlorophyll a in molecules of cells. For example, Hydrolab multi-parameter water quality analyzer produced by HACH Company (USA) utilizes single-point fluorescence to realize the classification of Cyanophyta and Chlorophyta in natural water [1-2]. FluoroProbe developed by the BBE [3-4] Company (Germany) can determines the concentrations of chlorophyll a in four algae groups quickly by using multi-point excitation spectra, but it is still impossible to distinguish Diatom and Pyrrophyta [5]. In practice, with relatively little information, two-dimensional excitation spectra between some different categories are highly similar, bringing great error to classification of living algae [6-7]. So improving accuracy of measurement and analysis of algae becomes a key problem.

In this manuscript, the classification and concentration prediction of twelve species of algae belonging to five categories are realized by discrete three-dimensional fluorescence spectra coupled with non-negative weighted least squares linear regression method. It is also demonstrated that more characteristics are obtained by discrete three-dimensional fluorescence spectra than two-dimensional excitation spectra. Our research provided the method for fast and accurate measurement of phytoplankton concentration.

2. Methods

2.1 Algal cultures

Twelve algae belonging to Cyanophyta, Chlorophyta, Diatom, Pyrrophyta and Cryptophyta were cultivated in incubator (25 ± 1°C) for 14days under 12/12 h light/ dark conditions (light, 120µmol photons/m²/s; dark, 0µmol photons/m²/s). The algae species and the culture medium were listed in Table 1. All species were provided by Freshwater Algae Culture Collection at the Institute of Hydrobiology, FACHB-collection.

Table 1. The algae species and the culture medium of twelve algae belonging to Cyanophyta, Chlorophyta, Diatom, Pyrrophyta and Cryptophyta

View Table | View all tables in this article

2.2 Estimation of Chl a by the spectrophotometric method and dilutions

The chlorophyll a standard concentration of initial sample of each pure alga was measured by UV2550 spectrophotometer (HITACH, Japan) [8]. A series of gradient samples of pure algae were prepared by diluting initial samples with demonized water and mixed samples according to different volume ratios of pure algae samples.

2.3 Discrete three-dimensional fluorescence spectra measurements

The discrete three-dimensional fluorescence spectra were measured by the On-line analyzer for algal fluorescence spectrum (AFA) developed by Anhui Institute of Optics and Fine Mechanics, Chinese Academy of Sciences. The discrete excitation wavelengths were 380, 420, 440, 460, 470, 480, 500, 520, 570, 580, 600, 620 and 660 nm, and the emission 620, 640, 660, 675, 685, 695, 705 and 715 nm [9]. In order to eliminate the effect of scattered light, the fluorescence intensities in the corresponding points were set to zero.

2.4 Standard fluorescence spectra

All the discrete three-dimensional fluorescence spectra of pure algae measured by AFA were normalized by its chlorophyll a concentration. The standard three-dimensional spectrum of each category is the average of all the normalized spectra of pure algae belonging to the same category [10].

F_{i} = (\begin{matrix} ε_{11} & \dots & ε_{1 n} \\ ⋮ & ⋱ & ⋮ \\ ε_{m 1} & \dots & ε_{m n} \end{matrix})

F_i is the standard three-dimensional spectrum matrix of the ith category, and ε_mn is the spectra intensity of the ith category at the excitation wavelength n and the emission wavelength m.

2.5 Non-negative weighted least squares linear regression

The non-negative weighted least squares linear regression analysis can be formulated as:

M = \sum_{i = 1}^{5} f_{i} \cdot a_{i} + χ

M is the unfolded spectra of the measured three-dimensional spectrum matrix of the mixed mixture; f_i represents the unfolded spectrum of the standard spectral matrix of the ith category; a_i is the chlorophyll a concentration of the ith category, and χ is the error. However, to account for different reliabilities from the f_i of the standard spectra depending on wavelength and on species, the weighting factor ω_i are introduced in to the error sum:

χ^{2} = \sum_{λ = 1}^{m n} {(\frac{M - \sum_{i = 1}^{5} f_{i} \cdot a_{i}}{\sum_{i = 1}^{5} ω_{i} \cdot a_{i}})}^{2}

ω_i is the standard deviation of all the phytoplankton species of the ith category.

3. Discussion

3.1 Fluorescence spectrum

The two-dimensional excitation spectra (chlorophyll a: emission at 685nm) and the discrete three-dimensional fluorescence spectra of all experimental algae are shown in Fig. 1 and Fig. 2, respectively. Figure 1 represents the normalized fluorescence excitation spectra of different algae. It shows that the chlorophyll a excitation spectra of different algae are very similar to each other except cyanophyta.

Fig. 1 The chlorophyll a (685nm) fluorescence excitation spectrum.

Download Full Size | PDF

Fig. 2 Discrete three-dimensional fluorescence spectra.

Download Full Size | PDF

Figure 2 shows the discrete three-dimensional fluorescence spectra of different phytoplankton. In order to eliminate the influence of algal concentrations on the spectra, the spectral intensities of all samples are normalized by standard chlorophyll a concentrations. The discrete three-dimensional fluorescence spectra of algae give the fingerprint information of different algae clearly. Besides the Cyanobacteria, Cryptomonas with the presence of phycoerythrin are distinct from other algae, which are not clearly achieved by chlorophyll a fluorescence excitation spectra in Fig. 1.

3.2 Spectral similarity

The differences of spectra of different categories and the similarities in the same categories are the basis for the classification of algae. In this work, the vector cosine method is introduced to quantify the similarity between the two spectra. Katty et al. [10] defines the similarity index as:

S I = 1 - [\frac{2}{π} \times arc \cos (\frac{A_{1} \cdot A_{2}}{| A_{1} | \cdot | A_{2} |})] \begin{matrix}  \end{matrix} 0 \leq S I \leq 1

A₁ and A₂ represent the spectral vectors of the two algae samples, respectively. The closer the SI is to 1, the higher the spectral similarity is; the closer the SI is to 0, the lower the spectral similarity is.

According to the definition, SI is a diagonally symmetric array, and the diagonal value must be 1. In order to compare the similarity of discrete three-dimensional fluorescence spectra and excitation fluorescence spectra between different algae, Table 2 combines the result of the two methods, the left part in normal type represents the similarity of the excitation fluorescence spectra of phytoplankton, and the right part in italic type represents the similarity of discrete three-dimensional fluorescence spectra of phytoplankton. The similarity among Chlorophyta, Diatom and Pyrrophyta are extremely high, the similarity indexes between Chlorophyta and Diatom, Chlorophyta and Pyrrophyta, Diatom and Pyrrophyta by excitation fluorescence spectra are 0.94, 0.93 and 0.97 respectively, even higher than that of the same category in some cases, and reduced to 0.88, 0.85 and 0.92 by discrete three-dimensional fluorescence spectra, respectively.

Table 2. Similarity of fluorescence spectra of different phytoplankton

View Table | View all tables in this article

3.3 Comparison of classification results

Table 3 and Table 4 are the classification results of 12 species of phytoplankton, which are expressed in the concentration of chlorophyll a. There are 72 samples of pure culture, 50 samples of mixed cultures with two categories in each sample, and 3 samples of mixed cultures with five categories in each sample. The results of recognition, classification and concentration prediction by the two analytical methods are compared. The accuracies of recognition, classification and concentration prediction are evaluated by recognition rate (RR), classification accuracy (CA) and recovery rate (RC).

Table 3. Discrimination results for samples of single division

View Table | View all tables in this article

Table 4. Discrimination results for mixed samples

View Table | View all tables in this article

Table 3 shows the accuracy of the discrete three-dimensional spectra method for the identification of pure culture samples is 94.44% with 4 times of error identification and that of excitation fluorescence spectra method 93.06% with 5 times of error identification. Except for diatoms, the recall rate was more than 90.65% and the classification accuracy was less than 9.6% for other algae by discrete three-dimensional fluorescence spectra, yet the recall rate was higher than 69.48%, and the classification accuracy was less than 11.5% when the samples were measured by excitation fluorescence spectra method. The recovery rates of three species of diatoms by discrete three-dimensional fluorescence spectra were 111.37%, 56.34%, 65.06% increased by 28.7%, 34.1%, 24.4% than that by the excitation spectrum method; the classification accuracy of three species of diatoms were 15.5%, 28.2%, 22.6%, increased by 16.8%, 46.8%, 33.5% than that by the excitation spectrum method, respectively.

The recognition accuracy of 30 mixed culture samples (Excluding Diatom samples) by the discrete three-dimensional fluorescence spectra is 100% without error identification, but recognition accuracy by the excitation fluorescence spectra method is only 76.7% with 7 times of error identification. The error identification of 23 mixed samples (mixed culture samples with Diatom) by the discrete three-dimensional fluorescence spectra and the excitation fluorescence spectra method was 6 and 12 resulting in recognition accuracy 73.9% and 47.8% respectively. The average recovery rates of Cyanophyta, Chlorophyta, Diatom, Pyrrophyta and Cryptophyta in the mixed samples by discrete three-dimensional fluorescence spectra were 80.9%, 80.2%, 25.8%, 73.6% and 85.7%, slightly larger than that by the excitation fluorescence spectra method. The absolute value of classification accuracies of Cyanophyta, Chlorophyta, Diatom, Pyrrophyta and Cryptophyta by discrete three-dimensional fluorescence spectra are less than 11.2%, 5.7%, 83.1%, 66.4% and17.6% respectively, increased by 11.5%, 46.7%, 16.9%, 2.7% and 2.2% respectively than that by the excitation fluorescence spectra method.

4. Conclusions

In this work, the classification and concentration prediction of twelve algae belonging to five categories were studied by the discrete three-dimensional fluorescence spectra, and the results were compared with that of the chlorophyll a fluorescence excitation spectra method. Results show that the recognition accuracy rate in pure samples by discrete three-dimensional fluorescence spectra is improved 1.38%, and the recovery rate and classification accuracy in pure diatom samples 34.1% and 46.8%, respectively; the recognition accuracy rate of mixed samples by discrete three-dimensional fluorescence spectra is enhanced by 26.1%, the recovery rate of mixed samples with Chlorophyta 37.8%, and the classification accuracy of mixed samples with diatoms 54.6%. It is indicated that the measurement accuracy and classification accuracy of the phytoplankton concentration can be significantly improved by the discrete three-dimensional fluorescence spectra.

Funding

The National Key Research and Development Program of China (2016YFC1400600); the Open fund of Qingdao National Laboratory for Marine Science and Technology (QNLM2016ORP0312); National “863” Program of China (2014AA06A509); Natural National Science Foundation of China (31400317); the Open Fund of Key Laboratory of Environment Optics and Technology (2005OP173065-2017-01).

References and links

1. J. Hilton, E. Rigg, and G. Jaworski, “Algal identification using in vivo fluorescence spectra,” J. Plankton Res. 11(1), 65–74 (1989). [CrossRef]

2. J. Gregor and M. Blahoslav, “A Simple In Vivo Fluorescence Method for the Selective Detection and Quantification of Freshwater Cyanobacteria and Eukaryotic Algae,” Clean-Soil Air Water 33(2), 142–148 (2005).

3. J. Kolbowski and U. Schreiber, “Computer-controlled phytoplankton analyzer based on 4-wavelengths PAM chlorophyll fluorometer,” in Photosynthesis: From Light to Biosphere, Volume V (Kluwer Academic Publishers, 1995), pp. 825–828.

4. C. S. Yentsch and D. A. Phinney, “Spectral fluorescence: an ataxonomic tool for studying the structure of phytoplankton populations,” J. Plankton Res. 7(5), 617–632 (1985). [CrossRef]

5. M. Beutler, K. H. Wiltshire, B. Meyer, C. Moldaenke, C. Lüring, M. Meyerhöfer, U.-P. Hansen, and H. Dau, “A fluorometric method for the differentiation of algal populations in vivo and in situ,” Photosynth. Res. 72(1), 39–53 (2002). [CrossRef] [PubMed]

6. J. W. Harrison, E. T. Howel, S. B. Watson, R. E. H. Smith, “Improved estimates of phytoplankton community composition based on in situ spectral fluorescence: use of ordination and field-derived norm spectra for the bbe FluoroProbe,” Canadian Journal of Fisheries & Aquatic Sciences 73, 10 (2016). [CrossRef]

7. N. Escoffier, C. Bernard, S. Hamlaou, and A. Groleau, “Quantifying phytoplankton communities using spectral fluorescence: the effects of species composition and physiological state,” J. Plankton Res. 37(1), 1–15 (2014).

8. National Environmental Protection Bureau, “Monitoring Analysis Methods for Water and Wastewater,” Zhongguo Huanjing Kexue (EPA, 1997).

9. G. Yin, N. Zhao, L. Hu, X. Yu, C. Shi, X. Xiao, L. Fang, J. Duan, T. Gan, Y. Zhang, J. Liu, and W. Liu, “Classified Measurement of Phytoplankton Based on Characteristic Fluorescence of Photosynthetic Pigments,” Acta Opt. Sin. 34(09), 0930005 (2014). [CrossRef]

10. K. X. Wan, I. Vidavsky, and M. L. Gross, “Comparing similar spectra: From similarity index to spectral contrast angle,” J. Am. Soc. Mass Spectrom. 13(1), 85–88 (2002). [CrossRef] [PubMed]

Categories	Genus	Species	Cultures media	Abbreviatin
Cyanophyta	Microcystis	Microcystis aeruginosa	BG11	Ama
	Anabaena	Anabaena sp.	BG11	Aaf
	Merismopedia	Merismopedia	BG11	Amp
	Oscillatoria	Oscillatoria lutea	BG11	Aol
Chlorophyta	Chlorella	Chlorella pyrenoidosa	BG11	Bcp
	Chlorella	Chlorella vulgaris	BG11	Bcv
	Scenedesmus	Scenedesmus obliquus	BG11	Bso
Diatom	Synedra	Synedra acusvar	CSI	Her
	Navicula	Navicula sp.	CSI	Hna
	Nitzschia	Nitzschia sp.	CSI	Hni
Pyrrophyta	Peridinium	Peridinium umbonatum var. inaequale	119	Kpe
Cryptomonas	Cryptomonas	Cryptomonas obovata.	AF-6	Mco

SI	Similarity of discrete three-dimensional fluorescence spectra							SI
Similarity of excitation fluorescence spectra		A	B	H	K	M		Similarity of discrete three-dimensional fluorescence spectra
	A	1	0.17	0.15	0.15	0.26	A
	B	0.17	1	0.88	0.85	0.81	B
	H	0.12	0.94	1	0.92	0.78	H
	K	0.12	0.93	0.97	1	0.72	K
	M	0.32	0.80	0.74	0.74	1	M
		A	B	H	K	M
SI	Similarity of excitation fluorescence spectra							SI

No	Components	SC	Discrete three-dimensional fluorescence spectroscopy				Excitation fluorescence spectroscopy
No	Components	SC	MC	RC	CA	F	MC	RC	CA	F
1	Aaf	55.4	54.6	100.6	0.8	0	56.9	103.4	1.0	0
2	Ama	57.6	54.9	95.0	0.0	0	52.4	90.3	0.0	0
3	Amp	54.6	55.3	102.2	0.1	0	55.7	102.0	0.0	0
4	Aol	2.4	3.3	142.2	3.4	0	3.4	157.3	3.4	0
5	Bcp	71.0	71.0	100.3	0.0	0	62.4	88.9	4.5	0
6	Bcv	59.5	53.12	90.7	1.6	0	42.8	69.5	11.5	0
7	Bso	73.4	72.7	99.9	0.7	0	68.1	94.4	4.5	0
8	Her	28.5	30.8	111.4	15.5	1	21.9	82.7	32.3	1
9	Hna	33.6	17.9	56.3	28.2	2	11.8	22.3	75.0	1
10	Hni	45.7	26.4	65.1	22.6	1	12.4	40.7	56.1	3
11	Kpe	28.3	28.6	100.7	9.6	0	28.3	101.7	9.7	0
12	Mco	113.6	114.1	101.6	0.2	0	113.8	101.5	0.4	0

No		Components	SC	Discrete three-dimensional fluorescence spectroscopy				Excitation fluorescence spectroscopy
No		Components	SC	MC	RC	CA	F	MC	RC	CA	F
1	A	A/B	22.6	16.1	71.5	9.4	0	16.2	71.8	−16.3	0
2	A	A/H	22.6	17.6	78.0	−8.1	0	17.5	77.5	22.7	0
3	A	A/K	22.6	18.1	80.0	−5.2	0	17.7	78.4	−5.5	0
4	A	A/M	22.6	20.0	88.5	−11.2	0	20.3	90.2	−11.8	0
5	A	A/B/H/K/M	3.8	3.2	86.3	−9.6	0	3.2	84.0	−14.6	0
6	B	A/B	48.7	37.6	77.3	0.6	0	24.5	50.3	52.5	2
7	B	B/H	48.7	36.4	74.9	−4.2	0	34.9	71.8	13.6	0
8	B	B/K	48.7	41.2	84.7	−4.7	0	45.4	93.3	6.7	0
9	B	B/M	48.7	38.7	79.4	−5.7	0	25.3	51.9	32.1	0
10	B	A/B/H/K/M	18.9	16.0	84.8	2.7	0	8.9	46.9	43.5	0
11	H	A/H	28.9	11.9	41.2	34.9	0	11.0	38.0	58.3	1
12	H	B/H	28.9	2.9	10.1	79.4	2	0.0	0.0	100.0	5
13	H	H/K	28.9	3.4	11.7	72.2	2	16.4	56.8	72.7	2
14	H	H/M	32.9	5.5	16.8	83.1	2	5.0	15.1	72.1	3
15	H	A/B/H/K/M	11.2	5.5	49.2	24.0	0	8.8	78.4	−78.6	1
16	K	A/K	21.6	13.9	64.3	25.1%	0	14.0	64.9	36.3	1
17	K	B/K	21.6	13.3	61.5	15.6	0	11.6	53.6	18.4	0
18	K	H/K	21.6	26.5	12.4	−37.1	0	20.2	93.2	−69.1	1
19	K	K/M	24.6	6.2	25.0	66.4	0	7.2	29.2	64.1	1
20	K	A/B/H/K/M	8.4	8.0	94.9	−14.6	0	7.0	83.0	24.7	2
21	M	A/M	101.1	84.0	83.1	0.4	0	85.8	84.8	0.8	0
22	M	B/M	78.6	70.0	89.0	−0.5	0	71.2	90.5	−12.9	0
23	M	H/M	78.6	70.9	90.2	−17.6	0	70.4	89.6	−14.5	0
24	M	K/M	78.6	64.9	82.5	−13.6	0	64.5	82.0	−15.4	0
25	M	A/B/H/K/M	16.9	14.1	83.7	−6.0	0	14.4	85.3	−19.8	0

Categories	Genus	Species	Cultures media	Abbreviatin
Cyanophyta	Microcystis	Microcystis aeruginosa	BG11	Ama
	Anabaena	Anabaena sp.	BG11	Aaf
	Merismopedia	Merismopedia	BG11	Amp
	Oscillatoria	Oscillatoria lutea	BG11	Aol
Chlorophyta	Chlorella	Chlorella pyrenoidosa	BG11	Bcp
	Chlorella	Chlorella vulgaris	BG11	Bcv
	Scenedesmus	Scenedesmus obliquus	BG11	Bso
Diatom	Synedra	Synedra acusvar	CSI	Her
	Navicula	Navicula sp.	CSI	Hna
	Nitzschia	Nitzschia sp.	CSI	Hni
Pyrrophyta	Peridinium	Peridinium umbonatum var. inaequale	119	Kpe
Cryptomonas	Cryptomonas	Cryptomonas obovata.	AF-6	Mco

On-line analysis of algae in water by discrete three-dimensional fluorescence spectroscopy

Abstract

1. Introduction

2. Methods

2.1 Algal cultures

2.2 Estimation of Chl a by the spectrophotometric method and dilutions

2.3 Discrete three-dimensional fluorescence spectra measurements

2.4 Standard fluorescence spectra

2.5 Non-negative weighted least squares linear regression

3. Discussion

3.1 Fluorescence spectrum

3.2 Spectral similarity

3.3 Comparison of classification results

4. Conclusions

Funding

References and links

Cited By

Figures (2)

Tables (4)

Equations (4)

Optics Express