Algae classification algorithm based on chlorophyll fluorescence dynamics

Abstract

This study presents a rapid and accurate algorithm for classifying algae based on chlorophyll fluorescence dynamics. A comprehensive dataset of five representative algal species was collected using a photosynthesis meter. A Bayesian-optimized extreme gradient boosting (XGBoost)-based prediction model was developed and evaluated against three machine-learning methods. The experimental results demonstrate that the proposed method achieves the highest overall classification performance, with an accuracy rate of 0.96. The feature importance analysis further indicated that fluorescence wavelengths contributed significantly to classification, with the highest importance at 440 nm and the lowest at 563 nm. These findings indicate that fluorescence dynamics can effectively capture species-specific spectral characteristics and provide a robust framework for rapid identification of diverse algal taxa. This work contributes to a better understanding of marine ecosystem dynamics. It provides technical support for marine environmental monitoring and the early detection of ecological events, such as harmful algal blooms.

Keywords

Chlorophyll fluorescence dynamics / Machine learning / Classification forecast / Marine ecological monitoring / Red tide warning

References

[1] Barsanti L, Birindelli L, Gualtieri P. Water monitoring by means of digital microscopy identification and classification of microalgae. Environ Sci-Proc Impacts, 2021, 23(10): 1443-1457

[2] Bartold M, Kluczek M. A machine learning approach for mapping chlorophyll fluorescence at inland wetlands. Remote Sens, 2023, 15(9): 2392

[3] Bercel TL, Kranz SA. Effects of spectral light quality on the growth, productivity, and elemental ratios in differently pigmented marine phytoplankton species. J Appl Phycol, 2021, 34: 185-202

[4] Beutler M, Wiltshire KH, Meyer B, Moldaenke C, Lüring C, Meyerhöfer M, et al. . A fluorometric method for the differentiation of algal populations in vivo and in situ. Photosynth Res, 2002, 72(1): 39-53

[5] Brumen B, Černezel A, Bošnjak L. Overview of machine learning process modelling. Entropy, 2021, 23(9): 1123

[6] Cai JX, Lee ZJ, Lin ZH, Hsu CH, Lin Y. An integrated algorithm with feature selection, data augmentation, and XGBoost for ovarian cancer. Mathematics, 2024, 12(24): 4041

[7] Choi W, Yang W, Na J, Lee G, Nam W. Feature optimization for gait phase estimation with a genetic algorithm and Bayesian optimization. Appl Sci, 2021, 11(19): 8940

[8] Dong M, Yin GF, Zhao NJ, Wang X, Zhang XL, Ma MJ, et al. . Accurate acquisition of initial fluorescence in phytoplankton primary productivity measurement using fluorescence kinetics method. Acta Opt Sin, 2024, 44(12): 19-25

[9] Fujibayashi M, Furuta S, Inoue E, Ichise S, Takei N. Dominance of harmful algae, Microcystis spp. and Micrasterias hardyi, has negative consequences for bivalves in a freshwater lake. Harmful Algae, 2021, 101 101967

[10] Gaur A, Pant G, Jalal AS. Comparative assessment of artificial intelligence (AI)-based algorithms for detection of harmful bloom-forming algae: an eco-environmental approach toward sustainability. Appl Water Sci, 2023, 13(5): 115

[11] Gong XF, Wan ZY, Jin P, Jin SH, Li XQ. Drought-driven divergence in photosynthetic performance between two Cunninghamia lanceolata provenances: insights from gas exchange and chlorophyll fluorescence dynamics. Plants Basel, 2025, 14(10): 1487

[12] Gou T, Liang RC, Guo QW, Chen SL, Zhao XM, Ma QL. Study on water environmental quality and algal bloom risk after the heavy metal pollution accident in Xiannv Lake. Ecol Sci, 2024, 43(5): 121-130

[13] Hastie T, Tibshirani R, Friedman J. The elements of statistical learning: data mining, inference, and prediction, 2009, New York. Springer

[14] Hattenrath-Lehmann TK, Zhen Y, Wallace RB, Tang YZ, Gobler CJ. Mapping the distribution of cysts from the toxic dinoflagellate Cochlodinium polykrikoides in bloom-prone estuaries by a novel fluorescence in situ hybridization assay. Appl Environ Microbiol, 2016, 82(4): 1114-1125

[15] Hologa R, Scheffczyk K, Dreiser C, Gärtner S. Tree species classification in a temperate mixed mountain forest landscape using random forest and multiple datasets. Remote Sens, 2021, 13(22): 4657

[16] Hu XP, Su RG, Zou WM, Ren SJ, Wang HT, Chai XP, et al. . Research on the discrimination methods of algae based on the fluorescence excitation spectra. Acta Oceanol Sin, 2010, 29(4): 116-128

[17] Hu YF, Guo B, Sun JN. Collaborative spectrum sensing algorithm based on Bayesian optimization XGBoost. Comput Appl Softw, 2024, 41(7): 121-127

[18] James G, Witten D, Hastie T, Tibshirani R. An introduction to statistical learning: with applications in R, 2013, New York. Springer

[19] Joshi A, Pradhan B, Chakraborty S, Behera MD. Winter wheat yield prediction in the conterminous United States using solar-induced chlorophyll fluorescence data and XGBoost and random forest algorithm. Ecol Inform, 2023, 77 102194

[20] Kim KH, Cho Y, Park H, Shin DW, Park J, Jung IJ et al (2018) Cultivation of microalgae in floating algal culture systems deployed in the ocean. In: Proceedings of the Korean Society of Industrial and Engineering Chemistry. KSIEC, p 199. https://kiss.kstudy.com/Detail/Ar?key=3602671

[21] Lan BY, Chen YM, Wu KS. A granular XGBoost classification algorithm. Appl Intell, 2025, 55(13): 895

[22] Lei X, Ye Y. Study on an adaptive thermal comfort model with K-nearest-neighbors (KNN) algorithm. Build Environ, 2021, 202 108026

[23] Li RZ, Gao LM, Wu GJ, Dong J. Multiple marine algae identification based on three-dimensional fluorescence spectroscopy and multi-label convolutional neural network. Spectrochim Acta A Mol Biomol Spectrosc, 2024, 311 123938

[24] Linda KM, Wiebe HCFK. Methods to estimate the diversity in the marine photosynthetic protist community with illustrations from case studies: a review. Diversity, 2010, 2(7): 973-1014

[25] Liu JJ, Yin GF, Zhao NJ, Zhang XL, Gan TT, Chen M, et al. . Experimental study on the relationship between chlorophyll fluorescence yield and photosynthetic activity in phytoplankton. Acta Opt Sin, 2023, 43(23): 2330001

[26] Ma LY, Yu SL, Zhao SY, Sun JM. Superheated steam temperature prediction models based on XGBoost optimized with random search algorithm. J North China Electr Power Univ (Nat Sci Ed), 2021, 48(4): 99-105

[27] Ma Y, Kong DL, Ye XY, Ding YL. Application and comparison of type 2 diabetes with comorbid hypertension classification prediction models based on random forest and XGBoost algorithms. J Guangdong Med Univ, 2024, 42(5): 523-534

[28] Markulj KA, Sudarić A, Matoša Kočar M, Duvnjak T, Liović I, Mijić A, et al. . Detailed insight into the behaviour of chlorophyll a fluorescence transient curves and parameters during different times of dark adaptation in sunflower leaves. Agronomy, 2024, 14(5): 954

[29] Murphy KP. Machine learning: a probabilistic perspective, 2012, Cambridge. MIT Press

[30] Park TG, Park YT. Detection of Cochlodinium polykrikoides and Gymnodinium impudicum (Dinophyceae) in sediment samples from Korea using real-time PCR. Harmful Algae, 2010, 9(1): 59-65

[31] Plyusnina TY, Chervitsov RN, Khrushchev SS, Kiseleva DG, Drozdenko T, Tikhomirova EI et al (2023) Identification of the toxic effects of heavy metals on phytoplankton by the analysis of chlorophyll fluorescence induction curves using machine learning. Theor Appl Ecol 2:126–134. https://doi.org/10.25750/1995-4301-2023-2-126-134

[32] Reimer J, Stöcklin S, Comella LM, Woias P, Werner C, Reindl L, et al. . An autonomous and wireless pulse-amplitude modulated chlorophyll fluorometer. Tm - Technisches Messen, 2021, 88(12): 773-784

[33] Schlie TP, Dierend W, Köpcke D, Rath T. Recording of low-oxygen stress response using chlorophyll fluorescence kinetics in apple fruit. Food Bioprocess Technol, 2023, 17(4): 1004-1016

[34] Shan SH, Xu L, Chen K, Tong MM, Wang XP. A rapid fluorescence approach on differentiation of typical dinoflagellate of East China Sea. Spectrochim Acta A Mol Biomol Spectrosc, 2022, 276 121216

[35] Shan XH, Ji YY, Yu YH, Gu S. Automatic identification of submarine cable with multi-beam sonar data based on XGBoost. J Ocean Technol, 2025, 44(2): 45-53

[36] Shi XJ, Wang DH, Li L, Wang Y, Ning RS, Yu SL, et al. . Algal classification and chlorophyll-a concentration determination using convolutional neural networks and three-dimensional fluorescence data matrices. Environ Res, 2025, 266 120500

[37] Sokolova M, Lapalme G. A systematic analysis of performance measures for classification tasks. Inf Process Manage, 2009, 45(4): 427-437

[38] Song Y. Forecasting short-term chlorophyll a concentration in Lake Erie using the machine learning XGBoost algorithm. Environ Res Lett, 2025, 20(6): 064029

[39] Su RG, Liang SK, Hu XP, Wang XL. Discrimination of 6 toxic red tide algae occurred in East China Sea by 3D fluorescence spectra. Mar Environ Res, 2008, 27(3): 265-268

[40] Tehrani NA, Janalipour M, Babaei H. Estimating water surface Chlorophyll-a concentration by big remote sensing data in the Persian Gulf, Bushehr. Remote Sens Earth Syst Sci, 2021, 4: 87-95

[41] Trainer VL, Hardy FJ. Integrative monitoring of marine and freshwater harmful algae in Washington State for public health protection. Toxins, 2015, 7(4): 1206-1234

[42] Tsai DDW, Chen PH, Ramaraj R. The potential of carbon dioxide capture and sequestration with algae. Ecol Eng, 2017, 98: 17-23

[43] Van Heukelem L, Thomas CS. Computer-assisted high-performance liquid chromatography method development with applications to the isolation and analysis of phytoplankton pigments. J Chromatogr A, 2001, 910(1): 31-49

[44] Wang SQ, Zhang XL, Yin GF, Zhao NJ, Dong M, Wang X, et al. . High-precision and rapid analysis method for three-dimensional fluorescence spectra of phytoplankton based on an interpolated reference spectral library. Acta Opt Sin, 2025, 45(10): 354-364

[45] Wong CK, Wong CK. HPLC pigment analysis of marine phytoplankton during a red tide occurrence in Tolo Harbour, Hong Kong. Chemosphere, 2003, 52(9): 1633-1640

[46] Wu H, Wu JW, Zheng DF, Zhong QM, Nian TK (2025a) Study on the prediction model of peak flood discharge of non-homogeneous landslide dam breach based on machine learning. Earth Sci 1–12. https://link.cnki.net/urlid/42.1874.p.20251111.1342.009

[47] Wu XY, Khan R, Gao HJ, Liu HB, Zhang J, Ma X. Low light alters the photosynthesis process in cigar tobacco via modulation of the chlorophyll content, chlorophyll fluorescence, and gene expression. Agriculture, 2021, 11(8): 755

[48] Wu YD, Xian B, Xiang XW, Fang F, Chu FH, Deng XK, et al. . Identification of key feature variables and prediction of harmful algal blooms in a water diversion lake based on interpretable machine learning. Environ Res, 2025, 276 121491

[49] Xu WP, Li WD, Zhang Y, Zhang TH, Chen HW (2021) Bagging-gradient boosting decision tree based milling cutter wear status prediction modelling. SN Appl Sci 3:879. https://doi.org/10.1007/S42452-021-04856-2

[50] Yan XY, Gu HM, Xiao YF, Ren H, Ni J. XGBoost algorithm applied in the interpretation of tight-sand gas reservoir on well logging data. Oil Geophys Prospect, 2019, 54(2): 447-455

[51] Yvon-Durocher G, Allen AP, Cellamare M, Dossena M, Gaston KJ, Leitao M, et al. . Five years of experimental warming increases the biodiversity and productivity of phytoplankton. PLoS Biol, 2015, 13(12): e1002324

[52] Zhang QL, Huang Y, An S. Quantification of phytoplankton groups using in-situ multi-excitation chlorophyll fluorescence measurements and machine learning (mf-ML). Algal Res, 2025, 90 104155

[53] Zimmermann J, Glöckner G, Jahn R, Enke N, Gemeinholzer B. Metabarcoding vs. morphological identification to assess diatom diversity in environmental studies. Mol Ecol Resour, 2015, 15(3): 526-542

Funding

National Key Research and Development Program(2022YFC3104104)

Taishan Scholars Program(tsqn202408288)

Key R&D Program of Shandong Province(2025CXGC010711)

Major Innovation Project for the Science Education Industry Integration Pilot Project of Qilu University of Technology (Shandong Academy of Sciences)(2023JBZ03)

RIGHTS & PERMISSIONS

The Author(s)