Deep forecasting architectures mirror ecological time scales in water quality prediction
6 affiliations
- 1Fujian Provincial Key Laboratory for Coastal Ecology and Environmental Studies, College of the Environment and Ecology, Xiamen University, Xiamen, 361102, China; State Key Laboratory of Marine Environmental Science, Xiamen University, Xiamen, 361102, China
- 2Fujian Provincial Key Laboratory for Coastal Ecology and Environmental Studies, College of the Environment and Ecology, Xiamen University, Xiamen, 361102, China; Corresponding author.
- 3College of Life and Environmental Science, Wenzhou University, Wenzhou, 325035, China; National and Local Joint Engineering Research Center of Ecological Treatment Technology for Urban Water Pollution, Wenzhou University, Wenzhou, 325035, China
- 4Department of Earth and Planetary Sciences, The University of Hong Kong, 999077, Hong Kong, China
- 5State Key Laboratory of Lake and Watershed Science for Water Security, Nanjing Institute of Geography and Limnology, Chinese Academy of Sciences, Nanjing, 211135, China; University of Chinese Academy of Sciences, Beijing, 100049, China; Helmholtz-Centre for Environmental Research-UFZ, Department Lake Research, Magdeburg, 39114, Germany
- 6Helmholtz-Centre for Environmental Research-UFZ, Department Lake Research, Magdeburg, 39114, Germany; Brandenburg University of Technology, Faculty Environment and Natural Sciences, Cottbus, 03046, Germany
Abstract
Accurate water quality forecasting—specifically tracking phytoplankton dynamics like chlorophyll a —is crucial for safeguarding aquatic ecosystems and global water security. Deep temporal architectures offer powerful non-linear modeling capabilities, promising to overcome the parameterization bottlenecks of traditional process-based models. However, current applications treat these models as black boxes, leaving it obscure how specific architectural modules interact with the intrinsic ecological time scales governing aquatic environments. Here, we reveal a fundamental architectural–ecological duality in water quality forecasting by systematically dissecting nine prediction architectures—spanning classic baselines to modern deep models—across two ecologically contrasting reservoirs. Using chlorophyll a dynamics to represent integrated system responses, we show that patch embedding serves as a universal temporal operator, mitigating noise and enhancing forecasting accuracy by up to 10.1% when integrated into standard baselines. Crucially, inter-variable modeling strategies dictate the effective forecast horizon: channel-independent architectures dominate short-term (1–7 days) chlorophyll a forecasting (Nash–Sutcliffe efficiency up to 0.88) by capturing biomass self-persistence, whereas cross-variable attention architectures excel in medium-term (8–15 days) prediction (efficiency up to 0.83) by uncoupling delayed nutrient–temperature interactions. Explainable AI further confirms a temporal shift in feature reliance from current biomass to lagged drivers over extended horizons. Beyond water quality, this mechanistic alignment between deep learning modules and ecological time scales establishes a scalable blueprint for building interpretable, horizon-adaptive AI frameworks across complex climate and environmental systems.
Related
- Operational chemical weather forecasting with the ECCC online Regional Air Quality Deterministic Prediction System version 023 (RAQDPS023) – Part 1: system description
- Integrating local ecological knowledge from professional fishers with environmental observations can improve the understanding, forecasting, and management of climate-sensitive jellyfish hazards
- Cost-aware evaluation of time-series foundation models for urban air-quality and temperature forecasting
- Physically constrained Mamba forecasting supports multiobjective reservoir water quality decisions
- Advancing Near-Term Water Quality Forecasting with Explainable Machine Learning and Probabilistic Uncertainty Quantification
BibTeX
@article{yu2026deep,
title = {{Deep forecasting architectures mirror ecological time scales in water quality prediction}},
author = {Yiqi Yu and Mingzhen Zhang and Zhongyao Liang and Fan Qu and Yao Wang and Xiangzhen Kong and Karsten Rinke and Nengwang Chen},
journal = {Environmental Science and Ecotechnology},
year = {2026},
month = sep,
doi = {10.1016/j.ese.2026.100755},
url = {https://doi.org/10.1016/j.ese.2026.100755},
note = {CC BY},
}
RIS
TY - JOUR
TI - Deep forecasting architectures mirror ecological time scales in water quality prediction
AU - Yiqi Yu
AU - Mingzhen Zhang
AU - Zhongyao Liang
AU - Fan Qu
AU - Yao Wang
AU - Xiangzhen Kong
AU - Karsten Rinke
AU - Nengwang Chen
PY - 2026
DA - 2026/09/01
JO - Environmental Science and Ecotechnology
DO - 10.1016/j.ese.2026.100755
UR - https://doi.org/10.1016/j.ese.2026.100755
AB - Accurate water quality forecasting—specifically tracking phytoplankton dynamics like chlorophyll a —is crucial for safeguarding aquatic ecosystems and global water security. Deep temporal architectures offer powerful non-linear modeling capabilities, promising to overcome the parameterization bottlenecks of traditional process-based models. However, current applications treat these models as black boxes, leaving it obscure how specific architectural modules interact with the intrinsic ecological time scales governing aquatic environments. Here, we reveal a fundamental architectural–ecological duality in water quality forecasting by systematically dissecting nine prediction architectures—spanning classic baselines to modern deep models—across two ecologically contrasting reservoirs. Using chlorophyll a dynamics to represent integrated system responses, we show that patch embedding serves as a universal temporal operator, mitigating noise and enhancing forecasting accuracy by up to 10.1% when integrated into standard baselines. Crucially, inter-variable modeling strategies dictate the effective forecast horizon: channel-independent architectures dominate short-term (1–7 days) chlorophyll a forecasting (Nash–Sutcliffe efficiency up to 0.88) by capturing biomass self-persistence, whereas cross-variable attention architectures excel in medium-term (8–15 days) prediction (efficiency up to 0.83) by uncoupling delayed nutrient–temperature interactions. Explainable AI further confirms a temporal shift in feature reliance from current biomass to lagged drivers over extended horizons. Beyond water quality, this mechanistic alignment between deep learning modules and ecological time scales establishes a scalable blueprint for building interpretable, horizon-adaptive AI frameworks across complex climate and environmental systems.
SN - 2666-4984
ER -