Low-Cost Synchronous Monitoring of Agricultural Water Quality and Greenhouse Gas Emissions Using IoT and Intelligent Algorithms

First Author: Zhang Huazhan PhD student (Nanjing Agricultural University, College of Resources and Environmental Sciences)

Corresponding Author: Gao Xiang Associate Professor (Nanjing Agricultural University, College of Resources and Environmental Sciences)

Paper DOI: 10.1016/j.watres.2024.122663

Low-Cost Synchronous Monitoring of Agricultural Water Quality and Greenhouse Gas Emissions Using IoT and Intelligent Algorithms

Graphic Abstract

Summary of Results

Recently, Nanjing Agricultural University published a research paper titled “Synchronous monitoring of agricultural water qualities and greenhouse gas emissions based on low-cost Internet of Things and intelligent algorithms” in Water Research (DOI: 10.1016/j.watres.2024.122663). The research team developed a synchronous monitoring system for water quality and greenhouse gas emissions based on the Internet of Things (IoT) and intelligent algorithms (WG-IoT-MS). This system is equipped with low-cost sensors and integrates intelligent algorithms to monitor the concentration of dissolved N₂O in real-time. By combining air-water gas exchange models, the system effectively monitors and simulates CO₂ and N₂O emissions from agricultural water bodies, reducing monitoring costs by about 60%. The research results provide a new idea and technology for the synchronous monitoring of agricultural water environments and greenhouse gases..

Overview of the Full Text

This study presents an innovative and cost-effective IoT monitoring system (WG-IoT-MS), aimed at real-time synchronous monitoring of water quality and greenhouse gas emissions from small agricultural water bodies, showing significant improvements over traditional methods:

1. Six intelligent algorithms were constructed for predicting dissolved N₂O concentration to select the best algorithm. Among them, the support vector regression algorithm (SVR-rbf) model for dissolved N₂O concentration performed excellently (R² > 0.89), with good tolerance for missing values. 2. The system integrates various low-cost sensors for pH, water temperature, CO₂, etc., achieving remote data transmission through a GPRS module, powered alternately by solar energy and batteries, ensuring long-term stable operation in the field. 3. By combining gas-water interface exchange models, it can simultaneously monitor the emission flux of CO₂ and N₂O, showing good consistency with traditional floating box measurement results (R² > 0.70). 4. The system is characterized by high portability and low cost, with monitoring costs only 40% of traditional methods, making it suitable for large-scale promotion and application. This work provides a new technical solution for monitoring agricultural water environments and has considerable application prospects in resource-limited areas.

Introduction

Traditional methods of monitoring water quality and greenhouse gases face challenges such as high costs, poor portability, and high labor demand. The WG-IoT-MS system developed in this work provides a new approach to address these challenges by integrating IoT technology and intelligent algorithms.

Illustrated Guide

System Design

WG-IoT-MS system integrates various low-cost sensors including pH, water temperature, air temperature, water depth, flow rate, nitrate nitrogen, CO₂ sensors. The system uses Arduino Mega2560 as the microcontroller, communicates with OneNET via a GPRS module, and stores data in a cloud database. Additionally, the monitoring system is powered alternately by a 12V battery and an 18W solar panel, ensuring reliable operation in remote areas.。

Optimal Model Selection

Fig. 2. Comparison of the observed and modeled dissolved N₂O concentration on different model test sets: (a) RFR, (b) ETR, (c) GBR, (d) SVR-rbf, (e) KNN, (f) BPNN. The blue line shows the linear regression through the data, and the gray line indicates a 1:1 fit. Copyright 2024, Elsevier Inc.

This study trained six intelligent algorithms using 2023 observational data to evaluate model performance based on prediction accuracy from test set data. In terms of prediction accuracy metrics, the extra trees regression (ETR) and support vector regression with radial basis function (SVR-rbf) algorithms showed high prediction accuracy (Fig. 2b and d), followed by random forest regression (RFR), gradient boosting regression (GBR), and backpropagation neural network (BPNN) algorithms (Fig. 2a, c, and f), with K-nearest neighbors (KNN) showing the lowest prediction accuracy (Fig. 2e)。

Missing Value Tolerance

To explore the impact of missing model input parameters on the prediction accuracy of dissolved N₂O concentration, this study further compared the performance of ETR and SVR-rbf under six different missing scenarios and four missing proportions (MP = 25%, 50%, 75%, 100%). As the number of missing variables and the missing proportion (MP) increased, differences in prediction accuracy were observed. Although the SVR-rbf showed slightly lower prediction accuracy than ETR on the overall data, its accuracy decreased at a relatively slower rate. Overall, the SVR-rbf outperformed ETR in overall performance, demonstrating stronger prediction stability and robustness against disturbances. Therefore, this study selected SVR-rbf as the optimal algorithm for dissolved N₂O concentration.

Applicability of Scenarios

When the test data came from lakes ( Fig. 4c), the model performed well (R²  = 0.82, MAPE = 13％). When the test data came from paddy fields (Fig. 4a), R² slightly decreased, but the prediction performance remained acceptable (R²  = 0.80, MAPE = 13％). However, when the test data came from pond scenarios (Fig. 4b), the prediction performance was relatively poor (R²  = 0.62, MAPE = 15％). This difference may stem from the different aquatic species cultivated in the aquaculture ponds, leading to specific water temperature and pH values.。

Validation of Emission Flux

Based on WG-IoT-MS, using the dissolved N₂O concentration model and the air-water gas exchange model to monitor/simulate CO₂ and N₂O emissions in the study area and comparing with floating box method results. The results show that the emission flux measurement results of both methods are relatively consistent (R²> 0.7). At different time points, the average absolute errors (MAE) of the simulated values of CO₂ and N₂O emission flux from the two paddy field monitoring sites were 7.16 mg C/(m²·h) and 2.01 g N/(m²·h). This study successfully achieved low-cost, portable monitoring of CO₂ and N₂O emission flux, demonstrating acceptable performance.。

Conclusion

This work reports the development and application of an innovative synchronous monitoring system for water quality and greenhouse gas emissions. The research integrates low-cost IoT sensing technology and intelligent algorithms, providing a new approach and technology to overcome the cost and portability issues of traditional monitoring equipment/methods for water quality and greenhouse gases. With the continuous accumulation, updating, and optimization of real measurement data in the later stage, the monitoring accuracy and applicability of this system will continue to improve.

Acknowledgments: PhD student Zhang Huazhan is the first author of the paper, and Associate Professor Gao Xiang is the corresponding author. Professors Zou Jianwen, Ding Yanfeng, Jiang Xiaosan, Xiong Zhengqin, Wang Songhan, Wang Jinyang, Researcher Wang Xuelai from the Remote Sensing Application Center of the Ministry of Ecology and Environment, and Associate Professor Li Wentao from Nanjing University also participated in this research. This research was funded by the National Natural Science Foundation, Jiangsu Province Carbon Peak Carbon Neutrality Technology Innovation Project, Jiangsu Province Agricultural Science and Technology Independent Innovation Project, Nanjing Agricultural University High-Level Talent Introduction Start-up Fund, and Jiangsu Province Graduate Scientific Research Practice Innovation Program.

Zhao Weihua’s Research Directions:

(1) Statistical analysis, modeling, and machine learning (AI) in the fields of environment and dual carbon;

(2) Processes and microorganisms in wastewater treatment

Welcome to cooperate on government science and technology projects, research topics, science and technology awards, paper patents, achievement transformation, engineering projects, etc.

Zhao Weihua personal homepage: https://hjxy.qut.edu.cn/info/1074/3436.htm (Click “Read More”)

Academic Positions:

Editorial Board Member of HydroResearch (2022.11-2024.11) (Indexed)

Editorial Board Member of Science for Energy and Environment (2023.9-2024.9)

Editorial Board Member of Green Carbon (2023.9-2026.5)

Editorial Board Member of Engineering Microbiology (2024.1)

Editorial Board Member of Journal of Ecology and Rural Environment (2023.4-2025.3) (Core Chinese)

Editorial Board Member of Journal of Aquatic Biology (2023.7-2025.7) (Core Chinese)

Editorial Board Member of Journal of North China University of Water Resources and Electric Power (Natural Science Edition) (2023.10-2026.9) (Core Chinese)

Editorial Board Member of Tianjin University of Science and Technology Journal (2023.11) (Core Chinese)

Editorial Board Member of Civil Technology (2022.12-2024.12) (National General Publication)

Editorial Board Member of Wetland Science and Management (2023.9-2025.9) (Science and Technology Core)

Editorial Board Member of Southwest Agricultural Journal (2023.12) (Core Chinese)

Editorial Board Member of Ningbo University Journal (Science and Engineering Edition) (2024.1-2025.12) (Science and Technology Core)

Editorial Board Member of Three Gorges Ecological Environment Monitoring (2023.12)

Editorial Board Member of Journal of Xinjiang Environmental Science (2024.5-2025.5)

Editor Dr. Zhao Weihu’s WeChat Communication:

Related posts:

Leave a Comment Cancel reply