Dataset of Systematic Review on Water Quality Prediction Using Machine Learning

Research contextEstuarine environments face strong environmental pressure and human use; water quality there matters for ecosystems and management decisions. At the same time, there is growing literature using machine learning, deep learning, and artificial intelligence to predict, forecast, or model water-quality variables. That body of work is spread across many journals and databases; a systematic synthesis helps map the state of the art, gaps, and methodological patterns. ObjectiveTo characterize and compile in a transparent, replicable way the set of peer-reviewed articles that link AI / machine learning / deep learning to prediction or modeling of water quality in estuarine settings—supporting qualitative and quantitative analysis of the review and open sharing of metadata via a repository. MethodologyA systematic literature review was conducted. The structured search (Boolean blocks for technology, water-quality task, and “estuary/estuaries”) was run on indexed databases (Scopus, Web of Science, ScienceDirect), with time ranges and field coverage as allowed by each platform. Records were managed in Zotero, with duplicate removal and screening (title/abstract and full text), disagreements resolved by consensus, and explicit inclusion criteria (e.g. estuarine focus; ML/DL/AI used to predict or model water-quality parameters; English language; sufficient methodological detail). Selection is documented (e.g. PRISMA flowchart). The main empirical product in this deposit is the metadata spreadsheet for the final included studies, plus supporting files (search strategy, criteria, review manuscript). Use of the dataThe data support: (1) citing and archiving the review’s study set on Zenodo or similar; (2) bibliometric or secondary analyses (by year, country, journal, keywords); (3) integrating DOIs and abstracts into synthesis tools or reference managers; (4) reproducibility (others can verify or extend the selection); (5) figures, narrative reviews, or future work on gaps (predicted variables, algorithms, spatial/temporal scales). The manuscript PDF complements academic use; the spreadsheet is the core for tabular analysis of included records.研究背景：河口环境面临严峻的环境压力与人类活动干扰，其水质状况对生态系统及管理决策至关重要。与此同时，越来越多的研究采用机器学习（Machine Learning, ML）、深度学习（Deep Learning, DL）与人工智能（Artificial Intelligence, AI）技术开展水质变量的预测、预报或建模工作。此类研究成果分散于众多期刊与数据库中，通过系统性综述能够梳理该领域的研究现状、现存缺口与方法学范式。 研究目标：以透明、可复现的方式梳理并汇编已同行评议的相关文献，这些文献将人工智能/机器学习/深度学习技术与河口地区水质预测或建模相结合，旨在支持该综述的定性与定量分析，并通过知识库开放共享元数据（metadata）。 研究方法：本研究开展了系统性文献综述。针对技术、水质任务与“河口/河口群”构建布尔逻辑检索式，在Scopus、Web of Science、ScienceDirect等索引数据库中进行结构化检索，检索时间范围与字段覆盖范围遵循各平台的限定规则。检索记录通过Zotero进行管理，完成去重与筛选（包括题名/摘要筛选与全文筛选），通过共识协商解决意见分歧，并明确纳入标准：例如以河口为研究对象、采用机器学习/深度学习/人工智能技术预测或建模水质参数、英文文献、具备充分的方法学细节。文献筛选流程已进行文档化记录（例如采用PRISMA流程图）。本数据集的主要实证产出为最终纳入研究的元数据电子表格，以及辅助文件（检索策略、纳入标准、综述手稿）。 数据用途：本数据集可用于：(1) 在Zenodo或同类平台上对本综述的研究集进行引用与归档；(2) 开展文献计量学或二次分析（例如按年份、国家、期刊、关键词维度）；(3) 将文献DOI与摘要整合至综述工具或参考文献管理软件中；(4) 保障研究可复现性（其他研究者可验证或拓展筛选流程）；(5) 为相关研究图表制作、叙述性综述或针对现存缺口的后续工作（例如预测变量、算法、时空尺度）提供支撑。综述手稿PDF可辅助学术应用，而电子表格则是对纳入记录进行表格化分析的核心载体。