Estimation of Phytoplankton Pigment Concentration in the South China Sea from Hyperspectral Absorption Data

Guifen Wang; Yinxue Zhang; Wenlong Xu; Wen Zhou; Hualian Wu; Zhantang Xu; Wenxi Cao

doi:10.3788/AOS202141.0601002

Journals >Acta Optica Sinica >Volume 41 >Issue 6 >Page 0601002 > Article

Acta Optica Sinica
Vol. 41, Issue 6, 0601002 (2021)

Estimation of Phytoplankton Pigment Concentration in the South China Sea from Hyperspectral Absorption Data

Guifen Wang^1、2、*, Yinxue Zhang^1、2, Wenlong Xu^1、2, Wen Zhou³, Hualian Wu⁴, Zhantang Xu³, and Wenxi Cao³

Author Affiliations

¹Key Laboratory of Marine Hazards Forecasting, Ministry of Natural Resources, Hohai University, Nanjing, Jiangsu 210098, China

²College of Oceanography, Hohai University, Nanjing, Jiangsu 210098, China

³State Key Laboratory of Tropical Oceanography, South China Sea Institute of Oceanology, Chinese Academy of Sciences, Guangzhou, Guangdong 510301, China

⁴CAS Key Laboratory of Tropical Marine Bio-Resources and Ecology, South China Sea Institute of Oceanology, Chinese Academy of Sciences, Guangzhou, Guangdong 510301, China

show less

DOI: 10.3788/AOS202141.0601002 Cite this Article Set citation alerts

Guifen Wang, Yinxue Zhang, Wenlong Xu, Wen Zhou, Hualian Wu, Zhantang Xu, Wenxi Cao. Estimation of Phytoplankton Pigment Concentration in the South China Sea from Hyperspectral Absorption Data[J]. Acta Optica Sinica, 2021, 41(6): 0601002 Copy Citation Text

show less

Fig. 1. Regression relationship between concentration of TChl a and concentration of accessory pigments (AP)

Download full size

Representative spectral absorption coefficients and their derivatives. (a) 4th derivative spectrum; (b) 2nd derivative spectrum; (c)1st derivative spectrum; (d) absorption coefficient

Fig. 2. Representative spectral absorption coefficients and their derivatives. (a) 4^th derivative spectrum; (b) 2^nd derivative spectrum; (c)1^st derivative spectrum; (d) absorption coefficient

Download full size

Fig. 3. Correlation coefficients between selected phytoplankton pigments concentration and derivative spectra. (a)(b) 1^st derivative spectra; (c)(d) 2^nd derivative spectra; (e)(f) 4^th derivative spectra

Download full size

Fig. 4. Comparison between the predicted value from the PLS model based on 2^nd derivative spectrum and the measured pigment concentrations for the validation data set, in which the 1∶1 ratio is shown as a solid line. (a) TChl a; (b) PSC; (c) PPC; (d) 19But; (e) Fuco; (f) 19Hex; (g) Diadino; (h) Zea

Download full size

Fig. 5. Comparison between the predicted value from the PLS model based on 4^th derivative spectrum and the measured pigment concentrations for the validation data set, in which the 1∶1 ratio is shown as a solid line. (a) TChl a; (b) PSC; (c) PPC; (d) 19But; (e) Fuco; (f) 19Hex; (g) Diadino; (h) Zea

Download full size

Fig. 6. Comparison of the validation results between the PLS regression model based on the 2^nd and 4^th derivative spectra and the empirical model based on total TChl a concentration. (a) Determination coefficient R² for linear regression between the predicted and measured pigment concentrations; (b) RMSE; (c) median percent difference; (d) mean percent difference

Download full size

Fig. 7. Validation results of the PLS regression model based on the 2^nd derivative spectrum and data collected in different months. (a) Determination coefficient R² for linear regression between the predicted and measured pigment concentrations; (b) RMSE; (c) median percent difference; (d) mean percent difference

Download full size

Cruise	Period	Number of samples
200609 KFHC	2006.09	24
200708 KFHC	2007.08	27
200808 KFHC	2008.08	14
200909 KFHC	2009.09	48
201004 NSFC	2010.04	89
201212 NSFC	2012.12	97
201308 NSFC	2013.08	37
201506 NSFC	2015.06	95

Table 1. Cruise information in the South China Sea between 2006—2015

Pigment name	Abbreviation
Chlorophyll a	Chl a
Chlorophyll b	Chl b
Chlorophyll c₁,c₂	Chl c₁₊₂
Chlorophyll c₃	Chl c₃
Divinyl Chlorophyll a	DV Chla
Fucoxanthin	Fuco
Peridinin	Perid
19'-hexanoyloxyfucoxanthin	19Hex
Zeaxanthin	Zea
19'-butanoyloxyfucoxanthin	19But
Alloxanthin	Allo
Diadinoxanthin	Diadino
β-carotene	β caro

Table 2. Used pigments and their abbreviations

Pigment	Range	Mean	Standard deviation	Correlation coefficient R
TChl a	[0.027,2.317]	0.262	0.261	1.00
PSC	[0.007,1.029]	0.087	0.115	0.87
PPC	[0.011,0.442]	0.091	0.047	0.58
19But	[0.001,0.205]	0.020	0.029	0.78
Fuco	[0.001,0.888]	0.027	0.072	0.78
19Hex	[0.004,0.209]	0.035	0.036	0.84
Diadino	[0.001,0.070]	0.007	0.008	0.72
Zea	[0.001,0.355]	0.062	0.034	0.03

Table 3. Statistical distribution of different pigments, and the correlation coefficients R of log-transformed concentrations between accessory pigments and total Chlorophyll a (N=431)

PLS model	Pigment	N	${R_{MSE}}_{}$ /(mg·m^-3)	${η_{x}}_{}$ /%	${η_{y}}_{}$ /%	R²	a	b
Based on 2^ndderivative spectrum	TChl a	7	0.129	87.92	79.27	0.72	0.761	0.060
	PSC	4	0.054	82.69	79.11	0.74	0.748	0.019
	PPC	4	0.034	82.23	47.33	0.38	0.413	0.052
	19But	3	0.017	80.15	66.52	0.63	0.646	0.006
	Fuco	5	0.035	84.62	83.76	0.72	0.725	0.006
	19Hex	3	0.023	79.49	63.27	0.57	0.617	0.013
	Diadino	8	0.004	89.86	80.58	0.72	0.745	0.002
	Zea	4	0.027	82.22	39.18	0.28	0.315	0.044
Based on 4^thderivative spectrum	TChl a	6	0.139	75.27	75.94	0.68	0.716	0.071
	PSC	5	0.057	72.36	77.62	0.70	0.718	0.022
	PPC	12	0.036	89.42	52.55	0.30	0.392	0.054
	19But	5	0.018	73.88	66.35	0.59	0.626	0.006
	Fuco	5	0.040	69.85	78.20	0.64	0.647	0.008
	19Hex	2	0.022	58.22	61.50	0.58	0.604	0.013
	Diadino	6	0.004	75.04	76.45	0.67	0.711	0.002
	Zea	12	0.028	89.44	43.45	0.20	0.274	0.046

Table 4. PLS parameters of 2^nd derivative spectrum and 4^th derivative spectrum models

PLS model	Pigment	R²	a	b	${R_{MSE}}_{}$ /(mg·m^-3)	N	${Δ_{MDPD}}_{}$ /%	${Δ_{MPD}}_{}$ /%
PLS model based on 2^ndderivative spectrum	TChl a	0.75	0.615	0.074	0.169	0	24.99	37.83
	PSC	0.97	0.714	0.011	0.049	0	25.83	36.00
	PPC	0.54	0.296	0.056	0.045	0	21.85	31.41
	19But	0.83	0.704	0.007	0.014	0	30.95	62.03
	Fuco	0.88	0.558	0.006	0.045	31	75.01	118.31
	19Hex	0.82	0.699	0.012	0.017	0	27.03	42.87
	Diadino	0.73	0.543	0.002	0.006	1	30.72	37.83
	Zea	0.59	0.283	0.040	0.033	0	28.69	114.26
PLS model based on 4^thderivative spectrum	TChl a	0.72	0.563	0.089	0.180	1	27.29	43.65
	PSC	0.89	0.636	0.020	0.063	3	33.22	51.14
	PPC	0.38	0.289	0.056	0.048	0	23.79	37.18
	19But	0.80	0.688	0.007	0.015	1	32.13	70.45
	Fuco	0.67	0.415	0.010	0.060	33	80.94	153.48
	19Hex	0.83	0.669	0.009	0.018	1	21.65	33.50
	Diadino	0.71	0.461	0.003	0.006	0	24.08	35.88
	Zea	0.43	0.290	0.038	0.034	2	35.64	103.47

Table 5. Validation results for the PLS model based on the 2^nd and 4^th derivative spectra

Pigment	A	B	R²	a	b	${R_{MSE}}_{}$ /(mg·m^-3)	${Δ_{MDPD}}_{}$ /%	${Δ_{MPD}}_{}$ /%
PSC	0.372	1.144	0.83	0.808	0.013	0.057	34.91	41.82
PPC	0.149	0.310	0.28	0.236	0.075	0.049	20.62	42.17
19But	0.063	0.805	0.44	0.366	0.013	0.025	63.85	118.41
Fuco	0.172	1.912	0.59	0.656	0.009	0.058	48.25	63.89
19Hex	0.099	0.701	0.75	0.583	0.016	0.021	33.35	56.19
Diadino	0.026	0.977	0.84	0.742	0.002	0.004	24.17	28.44
Zea	0.079	0.128	0.00	0.009	0.066	0.044	26.78	212.20

Table 6. Empirical model's coefficients(A,B) based on TChl a and the validation results

Download Citation

Set citation alerts for the article

Tools

Set citation alerts for the article

Save the article for my favorites

Paper Information