KALMAN FILTER APPLICATION IN WATER LEVEL FLOOD DETECTION

Abstract

Flooding is one of the most frequent natural disasters and often causes substantial damage to infrastructure and communities. Reliable flood monitoring and Early Warning Systems (FEWS) are therefore essential to mitigate risks and enable timely response. However, measurement noise caused by environmental disturbances such as water ripples can result in unstable sensor readings and false flood alerts. This study aims to improve measurement stability in an IoT-based FEWS by integrating a Kalman Filter (KF) into an ESP32-based embedded platform. The proposed system employs a resistive water-level sensor for real-time data acquisition, KF-based signal processing, web-based monitoring, and automated WhatsApp notifications via the Fonnte API. Experimental evaluation was conducted using 91 observations across three water-level scenarios representing conditions below, crossing, and above the flood-warning threshold. The KF improved measurement accuracy to 97.94%, 97.64%, and 97.75% across the three scenarios, while reducing measurement fluctuations caused by environmental disturbances. Furthermore, the proposed approach eliminated three false flood alerts generated by unfiltered measurements during the below-threshold scenario. These results demonstrate that integrating KF-based filtering into an IoT-based FEWS improves measurement reliability and supports a real-time flood monitoring and early warning.