Efficient River Water Level Sensing Using Cellular CSI and Joint Space-Time Processing

Abstract

Accurate and timely water level monitoring is critical for flood prevention, environmental management, and emerging smart infrastructure systems. Traditional water sensing methods often rely on dedicated sensors, which can be costly to deploy and difficult to maintain and are vulnerable to damage during floods.In this work, we propose a novel cellular signalbased sensing scheme that passively estimates water level changes using downlink mobile signals from existing communication infrastructure. By capturing subtle variations in channel state information (CSI), the proposed method estimates the length changes of the water-reflected signal path, which correspond to water level variations. A space-time processing framework is developed to jointly estimate the angle of arrival and Doppler shift, enabling isolation and enhancement of the water-reflected path via beamforming, while effectively suppressing environmental noise. The phase evolution of the beamformed signal is then extracted to infer water level changes. To address clock asynchronism between the transmitter and receiver inherent in bistatic systems, we introduce a beamforming-based compensation technique for removing time-varying random phase offsets in CSI. Field experiments conducted across a river demonstrate that the proposed method enables accurate and reliable water level estimation, achieving a mean accuracy ranging from 1.5 cm to 3.05 cm across different receiver configurations and deployments.