Beijing University of Technology’s Bi Jing Team

From “Seeing” to “Predicting”: How Does a Digital Water Management Simulation Platform Understand a River?

Editor's Note:

By the time a monitoring station issues an alert for excessive pollution levels, contamination may have already occurred. By the time manual inspections identify the source, the pollutant plume may have spread downstream. Faced with water pollution challenges that span regions, change rapidly, and involve complex data sources, water management authorities need more than just data. They need the ability to turn data into predictions, early warnings, and informed decisions.

In this edition of Outstanding Simulation Cases, we explore “Digital Water Management: An Intelligent Water Environment Sensing, Monitoring, and Early Warning System”. Built on multi-source sensing and data integration, the platform focuses on water quality prediction and intelligent early warning. It brings future changes in water quality to the screen in advance, giving water management teams more time to respond.

Abstract

Key Highlights:

The project team at Beijing University of Technology developed a water environment simulation platform that integrates intelligent sensing, data fusion, water quality prediction, anomaly alerts, pollution source tracing, and decision support. The platform establishes a closed-loop management process: Sensing → Prediction → Early Warning → Source Tracing → Decision-Making.

1. The Challenge of Water Management: Data Is Visible, but Risks Are Detected Too Late

Changes in the water environment rarely follow a smooth curve. Heavy rainfall, changes in upstream discharges, or river network operations can all cause water quality to fluctuate within a short period. Traditional monitoring relies mainly on automated stations at fixed monitoring points. Although these stations provide highly accurate data, they may fail to capture conditions across an entire water body. Meanwhile, hydrological, water quality, remote sensing, pollution source, and social data are often scattered across different departments, making comprehensive assessment more difficult.

Another challenge is that fixed-threshold systems usually issue alerts only after indicators exceed set limits. Complex mechanistic models also face limitations, including numerous parameters, difficult calibration, and long computation times. The real challenge is not simply detecting abnormal readings, but predicting how water quality will change and where risks may originate before pollution spreads further.

2. Tackling the Challenge: Building an Intelligent Water Environment Simulation Platform

To address these challenges, the team developed an Intelligent Water Environment Sensing, Monitoring, and Early Warning System. The platform consists of six layers: resource, computing, model, service, presentation, and management. It uses Docker, Kubernetes, and a microservices architecture to organize computing resources and business modules, enabling unified access to multi-source data, model deployment as services, and on-demand scaling.

Figure 1. Overall Architecture of the Intelligent Water Environment Sensing, Monitoring, and Early Warning System

The platform brings together data from ground-based automatic monitoring stations, satellite remote sensing, meteorological and hydrological systems, pollution sources, and operational files. After data cleaning, standardization, and integration, these sources form a unified water environment data foundation. Satellites provide wide-area monitoring, drones focus on key areas, and ground stations deliver high-frequency monitoring and calibration data. This space–air–ground integrated sensing approach helps fill the spatial gaps left by fixed monitoring stations.

Built on this unified data foundation, the platform packages algorithmic models as callable services. Monitoring data continuously flows into the system, where models predict trends and identify anomalies. The results are then displayed through maps, charts, and specialized analysis interfaces. Here, simulation is not just a static result on a screen; it is an operational capability that supports the entire process of sensing, analysis, assessment, and response.

Figure 2. Integrated Intelligent Sensing Interface: Bringing Together Regional Water Quality, River Network, and Pollution Source Information

3. Platform Capabilities: From Comprehensive Sensing to Early Prediction

The platform provides a range of functions, including intelligent sensing, data fusion, integrated GIS visualization, remote sensing identification, water quality prediction, anomaly alerts, pollution source tracing, risk assessment, and decision support. These modules work together rather than in isolation. Sensing and data management provide reliable inputs, prediction and early warning identify risks in advance, and source tracing and decision support turn analytical results into a basis for on-site investigation and response.

Water quality prediction is one of the platform’s key capabilities. By integrating high-frequency water quality data, upstream and downstream spatial relationships, meteorological and rainfall data, and remote sensing information, the system uses multi-factor models to predict changes over the next few hours to several days. Managers can query results by region, monitoring section, time, and indicator. They can also compare measured and predicted values on the same chart to track key indicators and turning points in trends, moving from monitoring current conditions to anticipating what comes next.

Figure 3. Water Quality Trend Simulation: Comparing Measured and Predicted Values

Intelligent early warning brings these predictions directly into the water management process. By combining water quality prediction models with anomaly detection models, the platform identifies exceedances of water quality standards, sudden changes in indicators, and signs of abnormal trends. It also dynamically adjusts warning criteria to account for seasonal and hydrological differences, helping reduce false alarms and missed detections caused by fixed thresholds. Anomalous monitoring sections are marked on the map according to risk level, with details such as the time of exceedance, affected indicators, anomaly type, and proportion displayed simultaneously. This helps managers quickly identify areas requiring investigation.

Figure 4. Water Quality Early Warning Interface: Locating Anomalous Monitoring Sections and Displaying Risk Status

Once an alert is triggered, the system can combine river network topology, upstream and downstream monitoring data, and pollution source information to conduct probabilistic analysis and rank potential sources and pathways of pollutant spread. In this way, the platform establishes a continuous process of “monitoring current conditions → predicting trends → issuing alerts → tracing sources”. This helps shift water environment management from reviewing reports after problems occur toward proactive assessment and early intervention.

4. Real-World Applications: Three Typical Use Cases

Application 1 | National Water Ecology and Environment Planning

The platform supported planning research conducted by the Department of Water Ecology and Environment of China’s Ministry of Ecology and Environment. It integrated data from 195 automatic monitoring stations and included a WeChat mini-program for planning research. The mini-program supports data visualization, public questionnaires, and on-site records, helping identify water ecology and environmental issues and supporting research into mechanisms for addressing them.

Application 2 | Integrated Supervision of the Haihe River Basin

In 2021, the platform was deployed locally at the Ecological Environment Supervision and Administration Bureau of the Beihai Sea Area for the Haihe River Basin under the Ministry of Ecology and Environment. It incorporated data from online pollution source monitoring and monitoring of key drinking water sources into a local database. The team also improved the process for identifying water pollution through remote sensing. These capabilities support integrated visualization of the basin’s water environment, water quality prediction and early warning, and intelligent analysis.

Application 3 | Water Environment Management in Langfang

The platform’s results have been applied to ecological and environmental management in Langfang, supporting the preparation of local water ecology and environmental protection plans through problem analysis and assessments of whether targets are achievable. Remote sensing monitoring and empirical comparisons have also been conducted in selected areas, helping improve the digital and intelligent capabilities of local water environment management.

These three applications cover national planning, river basin supervision, and local environmental management, demonstrating the platform’s ability to move from research models into real-world operations.

From “monitoring a river” to “understanding a river”, the real goal of digital water management is not simply to collect more data. It is to turn data into earlier predictions, clearer leads, and more reliable evidence for action.