物联网传感器融合监测塔纳湖水质：《科学报告》192 次观测中无迹卡尔曼融合对叶绿素 a 方差降幅达 93.72%
《科学报告》2026 年 11 月 11 日刊出 Haileyesus Amssaya、Fisseha Mekuria 与 Tesfa Tegegne 的研究，以埃塞俄比亚塔纳湖为案例，研究基于物联网的多传感器数据融合在淡水生态系统监测中的应用。论文指出，入侵水生杂草在淡水生态系统中快速蔓延，需要稳健的长期监测和高效的数据采集技术。
研究提出一套物联网传感器融合框架，以提高环境监测的准确性与可靠性，实时采集 pH、总氮、总磷和叶绿素 a 等关键水体参数。团队比较了加权、卡尔曼、扩展卡尔曼、无迹卡尔曼、贝叶斯和粒子滤波等多种传感器融合算法。实验数据集由部署在塔纳湖的多个传感器在三个月内采集的 192 次传感器观测组成。
结果显示，无迹卡尔曼传感器融合显著提高了测量精度与稳定性：在降低方差方面，相比无融合方法，总磷改善 81.46%、总氮改善 84.10%、叶绿素 a 改善 93.72%。加权融合与粒子滤波在 pH 上取得 96.24% 的改善。
论文说明该研究未获资助，于 2026 年 5 月 30 日投稿、11 月 9 日接受、11 月 11 日刊出。本稿数字均以论文摘要为准。
IoT Sensor Fusion Monitors Lake Tana Water Quality: In 192 Observations, the Unscented Kalman Fusion Cut Chlorophyll-a Variance by 93.72 Percent, Scientific Reports Finds
Scientific Reports published a study by Haileyesus Amssaya, Fisseha Mekuria and Tesfa Tegegne on November 11, 2026, presenting an IoT-based multi-sensor data fusion framework for freshwater ecosystem monitoring, with Lake Tana in Ethiopia as the case study. The paper states that the rapid proliferation of invasive aquatic weeds in freshwater ecosystems necessitates robust long-term monitoring and efficient data acquisition techniques.
The framework enables real-time acquisition of key aquatic parameters such as pH, total nitrogen, total phosphorus and chlorophyll-a. The team used sensor fusion algorithms including weighted, Kalman, extended Kalman, unscented Kalman, Bayesian and particle filter methods to enhance the precision of the collected data. The experimental dataset consisted of 192 sensor observations collected over three months using multiple sensors deployed in Lake Tana.
Experimental results demonstrated that the unscented Kalman sensor fusion significantly enhanced measurement accuracy and stability, achieving improvements over the no-fusion approach in reducing variance of 81.46 percent for total phosphorus, 84.10 percent for total nitrogen and 93.72 percent for chlorophyll-a. The weighted and particle filters achieved a 96.24 percent improvement for pH.
The paper states that no funding was received for the research. It was received on May 30, 2026, accepted on November 9, 2026 and published on November 11, 2026. All figures follow the abstract of the paper.
来源：《科学报告》
发布：GSM·美国（发布链接）
图片来源：图：图中是埃塞俄比亚塔纳湖的湖面与岸线。图中是塔纳湖本尊的实物照片，本研究案例湖泊即该湖。Radosław Botev，CC BY 3.0 pl，经Wikimedia Commons。 / Photo: file photo. Radosław Botev, CC BY 3.0 pl, via Wikimedia Commons.
GSM Global · 本篇英文版 English edition of this story