High-frequency water quality monitoring systems provide valuable measurements for predicting the trend of water quality, warning of abnormal activities or operating hydrological models. However, missing values are prevalent due to network miscommunication, device replacement or failure. Applying datasets with missing values can lead to biased results in statistical analysis or hydrological modelling work. We develop a cloud-based data processing system combining advanced algorithms to impute monitoring data in near real-time. The system provides high compatibility for supporting different water quality variables, imputation algorithms and extensive scalability to support numerous data streams. Based on the proposed approach, we review various imputation methods which can be applied to water quality data. Overall, this work provides a systematic design of a water quality data imputation system, explores the advantages and limitations of selected data imputation methods and analyses the imputation performance of two real-time water quality monitoring systems located in both the USA and Australia. The results provide practical guidelines for data imputation applications in water quality data.

高频水质监测系统为预测水质趋势、异常活动预警或运行水文模型提供了有价值的测量。然而，由于网络通信错误、设备更换或故障，缺失值很普遍。应用具有缺失值的数据集可能会导致统计分析或水文建模工作中的结果有偏差。我们开发了一个基于云的数据处理系统，结合了先进的算法来近乎实时地估算监测数据。该系统为支持不同的水质变量、插补算法和广泛的可扩展性提供了高兼容性，以支持众多数据流。基于所提出的方法，我们回顾了可应用于水质数据的各种插补方法。总体，这项工作提供了水质数据插补系统的系统设计，探讨了所选数据插补方法的优点和局限性，并分析了位于美国和澳大利亚的两个实时水质监测系统的插补性能。结果为水质数据中的数据插补应用提供了实用指南。