Automated Water Quality Survey and Evaluation Using an IoT Platform with Mobile Sensor Nodes

Teng Li, Min Xia, Jiahong Chen, Yuanjie Zhao, Clarence W. de Silva

TL;DR

This paper introduces a novel IoT platform for automated, high-resolution water quality monitoring. It features a hexagonal grid-based survey planner for efficient mobile sensor node deployment and an Online Water Quality Index (OLWQI) that effectively processes large volumes of real-time data, overcoming limitations of traditional off-line indices.

What Problem Does This Paper Address?

Traditional water quality monitoring relies on time-consuming, labor-intensive field measurements, leading to limited spatial and temporal coverage. While existing automated systems with static sensors exist, they lack the flexibility for comprehensive spatiotemporal analysis. Furthermore, interpreting the vast amounts of data generated by automated systems requires efficient indexing methods that can handle real-time, high-frequency data, a challenge not fully addressed by existing off-line Water Quality Indices (WQIs) like the CCME WQI. This paper aims to bridge this gap by developing a rapidly deployable IoT platform that combines efficient survey planning with an effective online data indexing method.

Methods

The core of this work lies in two main innovations:

  • Hexagonal Grid-Based Survey Planner (HGSP): This planner addresses the challenge of efficiently covering a study area with mobile sensor nodes (MSNs) under energy and time constraints.
  • Cellular Decomposition: The study area is divided into uniform hexagonal cells.
  • Sampling Locations of Interest (SLoIs): The centers of these cells are designated as SLoIs.
  • Minimum Spanning Tree (MST): An MST is constructed based on these SLoIs to create a base path.
  • Circumnavigation Path: A path ring is generated by circumnavigating the MST, ensuring all SLoIs are visited.
  • Sub-path Division: For multiple MSNs, the path ring is divided into sub-paths to meet time interval requirements. This approach aims for optimal coverage and efficient energy usage.
  • Online Water Quality Index (OLWQI): This index is designed to interpret large volumes of online water quality data, addressing the limitations of the CCME WQI for real-time applications.
  • Modification of CCME WQI: The OLWQI is derived from the CCME WQI but incorporates modifications to handle high-frequency data and mitigate biases.
  • Scope Factor Modification: The Scope Factor is modified to use the average of all measurements for a parameter, preventing bias from a few failed tests.
  • Amplitude Factor Modification: The Amplitude Factor is normalized to ensure an unbiased index score across parameters with different data ranges.
  • Analytical Formulation: The OLWQI is expressed in an analytical form, enabling efficient automated execution on devices.

The platform architecture includes Mobile Sensor Nodes (MSNs) equipped with multiple sensors, a Base Station (BS) for local data aggregation and processing, and a Remote Server (RS) for overall system management and user access.

Datasets

The paper describes experiments using two primary datasets:

  • Yosef Wosk Reflecting Pool, University of British Columbia, Canada: This is a real-world deployment site used to demonstrate the platform's performance in terms of path planning and initial water quality evaluation.
  • Chesapeake Bay Interpretive Buoy System (CBIBS) dataset: A large, realistic dataset collected from June 1 to July 1, 2017, at 10 monitoring stations. This dataset, containing measurements of Chlorophyll A, Dissolved Oxygen, Turbidity, Conductivity, Salinity, and Temperature, was used for further validation and comparison of the OLWQI with the CCME WQI.

Key Insights and Ideas

  • Bridging the Gap: The paper successfully integrates efficient spatiotemporal survey planning with a robust online data indexing method, creating a comprehensive solution for automated water quality monitoring.
  • Hexagonal Grid Advantage: The hexagonal grid decomposition offers a spatially balanced distribution of sampling locations, ensuring consistent coverage and efficient path planning.
  • OLWQI's Superiority for Online Data: The modified OLWQI demonstrates significantly better performance than the traditional CCME WQI when dealing with large volumes of online data, particularly in preventing bias from a few failed tests or parameters with wide data ranges.
  • Cost-Effectiveness and Deployability: The platform emphasizes low-cost components and a design that allows for rapid deployment and easy maintenance, making it practical for real-world applications.
  • Automation for Environmental Monitoring: The entire system is designed for automation, from mission generation to data processing and quality indexing, reducing human intervention and enabling more frequent and detailed monitoring.

Key Results

  • HGSP Performance: The HGSP algorithm demonstrated comparable or better path lengths than the Traveling Salesman Problem (TSP) approach for covering sampling locations, while significantly reducing computation time. For instance, at r=2.4m, HGSP achieved a total path length of 366m in 0.60s, whereas TSP took 3.17s for the same path.
  • OLWQI vs. CCME WQI: In field tests, the OLWQI consistently provided higher scores (e.g., 98/100) compared to the CCME WQI (e.g., 77/100) for the same data, indicating a more favorable and less biased evaluation. The OLWQI's Scope Factor was not unduly degraded by a small number of failed tests, unlike the CCME WQI.
  • Factor Sensitivity: The OLWQI exhibits more balanced factor sensitivities across its components (Scope, Frequency, Amplitude) compared to the CCME WQI, which is heavily influenced by the Scope Factor.
  • MSN Deployment: The system successfully deployed two MSNs to cover 88 sampling locations within a 15-minute time interval requirement, demonstrating the effectiveness of the multi-MSN strategy.

Limitations and Future Work

  • Environmental Model Adaptation: The authors suggest extending the platform to adaptively modify objective sampling locations over time by incorporating environmental models. This would allow for more intelligent and targeted data collection.
  • Initial Deployment for Model Knowledge: The proposed approach can serve as an initial deployment to gather prior knowledge for potential environmental models.
  • Further Validation: While the CBIBS dataset provided valuable validation, continued testing with diverse datasets and environmental conditions would further strengthen the platform's robustness.
  • Energy Budget Optimization: While the HGSP considers energy budgets, further research into more sophisticated energy optimization strategies for MSNs could extend their operational lifespan.

Summary updated

BibTeX
@inproceedings{li2017automated,
  author = {Teng Li and Min Xia and Jiahong Chen and Yuanjie Zhao and Clarence W. de Silva},
  title = {{Automated Water Quality Survey and Evaluation Using an IoT Platform with Mobile Sensor Nodes}},
  booktitle = {Sensors},
  year = {2017},
  url = {https://openreview.net/forum?id=4c7Ul-i2Po},
}
Figure