Decoding environmental memory and spatiotemporal heterogeneity of algal blooms: an explainable machine learning framework with ecological thresholds

Abstract

Algal blooms in eutrophic lakes are shaped by non-linear responses to both current and antecedent environmental conditions, while their dominant drivers can vary across space and bloom stage. This study developed an explainable machine-learning framework for predicting algal density using high-frequency monitoring data collected from six stations in Lake Taihu, China, during 2021–2024. The lake was divided into eastern and western functional zones, and separate models were developed for stable, outbreak, and decline stages. Two predictor schemes were compared: W1 contained contemporaneous environmental variables, whereas W2 additionally incorporated 3–15 d lagged, moving-average, and cumulative predictors. Across four algorithms and seven data subsets, W2 outperformed W1 in all 28 matched comparisons: mean test-set R² increased by 0.140, while MSE, RMSE, and MAE decreased by 41.2%, 24.0%, and 27.8%, respectively. XGBoost achieved the best overall performance, with a maximum test-set R² of 0.901. An integrated ranking based on model-specific importance, permutation importance, and SHapley Additive exPlanations (SHAP) showed strong context dependence. Solar radiation, total phosphorus (TP), and the 15-d mean N:P ratio dominated the global model; TP and total nitrogen were prominent during eastern functional-zone outbreak and decline stages, whereas antecedent pH, permanganate index, precipitation, turbidity, conductivity, and N:P were more influential in western functional-zone stage-specific models. BIC-supported segmented SHAP analysis with bootstrap confidence intervals identified reproducible model-derived transition ranges, including stage-dependent TP transitions on the eastern functional zone, pH transitions near 7.8 on the western shore, and N:P transitions during bloom decline. These results demonstrate that antecedent environmental information improves prediction and that spatial-stage stratification reveals local response regimes that are obscured in a lake-wide model.

Graphical abstract

Keywords

Harmful Algal Blooms (HABs) / Explainable machine learning / Spatiotemporal heterogeneity / Ecological thresholds / Lag effects

Highlight

	● Explainable ML reveals environmental memory and HAB heterogeneity in Taihu.
	● Lag features (3–15 d) lift XGBoost R 2 to 0.90 and cut MSE by ~41%.
	● Global + six zone–stage submodels capture regime shifts across bloom phases.
	● East is TP-controlled (0.025–0.070 mg/L), while the west is pH/hydrometeorology-driven.
	● N:P ratio acts as a decline-phase switch, informing recurrence and restoration.

References

[1] Ai H P , Zhang K , Sun J C , Zhang H C . (2023). Short-term Lake Erie algal bloom prediction by classification, regression models. Water Research, 232: 119710

[2] Almela P , Elser J J , Giersch J J , Hotaling S , Hamilton T L . (2024). Influence of snow cover on albedo reduction by snow algae. mBio, 16(2): e03630–24

[3] Ao Y Y , Fan J T , Guo F , Li M Y , Li A P , Shi Y , Wei J . (2025). Machine learning-based early warning of algal blooms: a case study of key environmental factors in the Anzhaoxin River Basin. Water, 17(5): 725

[4] Biau G , Scornet E . (2016). A random forest guided tour. Test, 25(2): 197–227

[5] Chen T Q, Guestrin C (2016). XGBoost: a scalable tree boosting system. arXiv preprint arXiv: 1603.02754

[6] Cui H , Tao Y W , Li J , Zhang J H , Xiao H , Milne R . (2024). Predicting and analyzing the algal population dynamics of a grass-type lake with explainable machine learning. Journal of Environmental Management, 354: 120394

[7] Cuker B E , Gama P T , Burkholder J M . (1990). Type of suspended clay influences lake productivity and phytoplankton community response to phosphorus loading. Limnology and Oceanography, 35(4): 830–839

[8] Demiray B Z , Mermer O , Baydaroğlu Ö , Demir I . (2025). Predicting harmful algal blooms using explainable deep learning models: a comparative study. Water, 17(5): 676

[9] Ding S M , Chen M S , Gong M D , Fan X F , Qin B Q , Xu H , Gao S S , Jin Z F , Tsang D C W , Zhang C S . (2018). Internal phosphorus loading from sediments causes seasonal nitrogen limitation for harmful algal blooms. Science of the Total Environment, 625: 872–884

[10] Fang Y , Huang R T , Zhang Y Y , Zhang J , Xi W N , Shi X Y . (2025). Utilizing machine learning models to grasp water quality dynamic changes in lake eutrophication through phytoplankton parameters. Frontiers of Environmental Science & Engineering, 19(2): 14

[11] Girdner S , Sprague J . (2025). Seasonal patterns of phytoplankton photoacclimation in a dimictic oligotrophic lake over 5 yr: implications for chlorophyll as an estimate of biomass. Limnology and Oceanography, 70(S1): S28–S39

[12] Gui J , Sun J . (2024). Phytoplankton carbon to chlorophyll a model development: a review. Frontiers in Marine Science, 11: 1466072

[13] Hajihosseinlou M , Maghsoudi A , Ghezelbash R . (2023). A novel scheme for mapping of MVT-type Pb–Zn prospectivity: LightGBM, a highly efficient gradient boosting decision tree machine learning algorithm. Natural Resources Research, 32(6): 2417–2438

[14] Hao R , Sun X , Zhao Y C , Zhang R , Li H W , Shang J H . (2025). Advancing floating photovoltaic systems: trends, challenges, and future directions in sustainable energy development. International Journal of Green Energy, 22(4): 688–709

[15] Jeong B , Chapeta M R , Kim M , Kim J , Shin J , Cha Y . (2022). Machine learning-based prediction of harmful algal blooms in water supply reservoirs. Water Quality Research Journal, 57(4): 304–318

[16] Jin L L , Huang H , Ren H Q . (2025). AI-driven transformation of water treatment technology and industry: toward a new era of comprehensive innovation. Frontiers of Environmental Science & Engineering, 19(8): 114

[17] Lintz H E , McCune B , Gray A N , McCulloh K A . (2011). Quantifying ecological thresholds from response surfaces. Ecological Modelling, 222(3): 427–436

[18] Liu C J , Chen Y , Zou L , Cheng B F , Huang T H . (2022). Time-lag effect: river algal blooms on multiple driving factors. Frontiers in Earth Science, 9: 813287

[19] Ma J R , Qin B Q , Wu P , Zhou J , Niu C , Deng J M , Niu H L . (2015). Controlling cyanobacterial blooms by managing nutrient ratio and limitation in a large hyper-eutrophic lake: lake Taihu, China. Journal of Environmental Sciences, 27: 80–86

[20] Ogle K , Barber J J , Barron-Gafford G A , Bentley L P , Young J M , Huxman T E , Loik M E , Tissue D T . (2015). Quantifying ecological memory in plant and ecosystem processes. Ecology Letters, 18(3): 221–235

[21] Paerl H W , Xu H , McCarthy M J , Zhu G W , Qin B Q , Li Y P , Gardner W S . (2011). Controlling harmful cyanobacterial blooms in a hyper-eutrophic lake (Lake Taihu, China): the need for a dual nutrient (N & P) management strategy. Water Research, 45(5): 1973–1983

[22] Pal M , Yesankar P J , Dwivedi A , Qureshi A . (2020). Biotic control of harmful algal blooms (HABs): a brief review. Journal of Environmental Management, 268: 110687

[23] Poikane S , Kelly M G , Várbíró G , Borics G , Erős T , Hellsten S , Kolada A , Lukács B A , Solheim A L , López J P . et al. (2022). Estimating nutrient thresholds for eutrophication management: novel insights from understudied lake types. Science of the Total Environment, 827: 154242

[24] Power N (2025). NASA POWER Data Access Viewer. Washington, DC: National Aeronautics and Space Administration

[25] Qin B Q , Deng J M , Shi K , Wang J , Brookes J , Zhou J , Zhang Y L , Zhu G W , Paerl H W , Wu L . (2021). Extreme climate anomalies enhancing cyanobacterial blooms in Eutrophic Lake Taihu, China. Water Resources Research, 57(7): e2020WR029371

[26] Qin B Q , Xu P Z , Wu Q L , Luo L C , Zhang Y L . (2007). Environmental issues of Lake Taihu, China. Hydrobiologia, 581(1): 3–14

[27] Qin B Q , Zhang Y L , Zhu G W , Gao G . (2023). Eutrophication control of large shallow lakes in China. Science of the Total Environment, 881: 163494

[28] Qin B Q , Zhu G W , Gao G , Zhang Y L , Li W , Paerl H W , Carmichael W W . (2010). A drinking water crisis in Lake Taihu, China: linkage to climatic variability and lake management. Environmental management, 45(1): 105–112

[29] Sandrini G , Tann R P , Schuurmans J M , Van Beusekom S A M , Matthijs H C P , Huisman J . (2016). Diel variation in gene expression of the CO₂-concentrating mechanism during a harmful cyanobacterial bloom. Frontiers in Microbiology, 7: 551

[30] Spake R , Barajas-Barbosa M P , Blowes S A , Bowler D E , Callaghan C T , Garbowski M , Jurburg S D , Van Klink R , Korell L , Ladouceur E . et al. (2022). Detecting thresholds of ecological change in the anthropocene. Annual Review of Environment and Resources, 47(1): 797–821

[31] Stief P , Schauberger C , Lund M B , Greve A , Abed R M M , Al-Najjar M A A , Attard K , Bonaglia S , Deutzmann J S , Franco-Cisterna B . et al. (2022). Intracellular nitrate storage by diatoms can be an important nitrogen pool in freshwater and marine ecosystems. Communications Earth & Environment, 3(1): 154

[32] Sun G J , Zhu W T , Qian X Y , Wei C L , Xie P F , Shi Y , Cao X Y , He Y . (2025). Machine learning models for chlorophyll-a forecasting in a freshwater lake: case study of lake Taihu. Water, 17(8): 1219

[33] Tao Y W , Ren J L , Zhu H P , Li J , Cui H . (2024). Exploring spatiotemporal patterns of algal cell density in lake Dianchi with explainable machine learning. Environmental Pollution, 356: 124395

[34] Wang L , Shan K , Yi Y , Yang H , Zhang Y Y , Xie M J , Zhou Q C , Shang M S . (2024). Employing hybrid deep learning for near-real-time forecasts of sensor-based algal parameters in a Microcystis bloom-dominated lake. Science of the Total Environment, 922: 171009

[35] Wang Q , Sun L , Zhu Y , Wang S B , Duan C Y , Yang C J , Zhang Y M , Liu D J , Zhao L , Tang J L . (2022). Hysteresis effects of meteorological variation-induced algal blooms: a case study based on satellite-observed data from Dianchi Lake, China (1988–2020). Science of the Total Environment, 812: 152558

[36] Wang Q , Wang T , Zhao S P , Yang K , Wen X Y , Zhao M H , Luo F , Jiang B L , Jin Y , Zhang B Y . (2023). Comprehensive meteorological factors analysis and lag correlation study for cyanobacterial blooms in shallow plateau lake. Ecological Indicators, 153: 110394

[37] Wang Y T , Khan S , Milham P J , Lin Z W , Qi X X , Gao H M , Duan Y K , Shi J L , Oduro C , Ullah H . et al. (2025). Sediment-associated ammonium-N increased benthic chlorophyll-a in streams: Evidence from SHAP integrated machine learning analysis. Ecological Indicators, 178: 113933

[38] Wu T F , Qin B Q , Brookes J D , Yan W M , Ji X Y , Feng J . (2019). Spatial distribution of sediment nitrogen and phosphorus in Lake Taihu from a hydrodynamics-induced transport perspective. Science of the Total Environment, 650: 1554–1565

[39] Xu Y , Zhang D , Lin J Q , Peng Q D , Lei X H , Jin T T , Wang J , Yuan R F . (2024). Prediction of phytoplankton biomass and identification of key influencing factors using interpretable machine learning models. Ecological Indicators, 158: 111320

[40] Yajima H , Derot J . (2018). Application of the Random Forest model for chlorophyll-a forecasts in fresh and brackish water bodies in Japan, using multivariate long-term databases. Journal of Hydroinformatics, 20(1): 206–220

[41] Yang Z J, Wei C Y, Liu D F, Lin Q C, Huang Y L, Wang C F, Ji D B, Ma J, Yang H (2022). The influence of hydraulic characteristics on algal bloom in three gorges reservoir, China: a combination of cultural experiments and field monitoring. Water Research, 211, 118030

[42] Zepernick B N , Gann E R , Martin R M , Pound H L , Krausfeldt L E , Chaffin J D , Wilhelm S W . (2021). Elevated pH conditions associated with Microcystis spp. Blooms decrease viability of the cultured diatom Fragilaria crotonensis and natural diatoms in Lake Erie. Frontiers in Microbiology, 12: 598736

[43] Zhang M L , Zhou Z H. . (2007). ML-KNN: a lazy learning approach to multi-label learning. Pattern Recognition, 40(7): 2038–2048

[44] Zhang Y L , Qin B Q , Shi K , Zhang Y B , Deng J M , Wild M , Li L , Zhou Y Q , Yao X L , Liu M . et al. (2020). Radiation dimming and decreasing water clarity fuel underwater darkening in lakes. Science Bulletin, 65(19): 1675–1684

[45] Zou W , Cao Z G , Wang X L , Huang Q F , Xu H , Zhu M Y , Zhang Y L , Qin B Q , Zhu G W . (2025). The hydrological and nutrient conditions mediate algal biomass response to compound heatwave and drought events in large shallow lakes. Freshwater Biology, 70(6): e70061

Rights & permissions

Higher Education Press 2027