The limited penetration of smart water meters in current water distribution systems (WDS) compromises the accuracy and reliability of WDS simulations. Beyond the effectiveness of parameter calibration algorithms, the precision of demand allocation at nodes, especially for non-online metered demands, is pivotal to simulation fidelity. Traditionally, this task is addressed through top-down (TD) and bottom-up (BU) approaches. However, these methods are often constrained by outdated population data, assumptions of uniform demand distribution, and substantial initial workload. To overcome these limitations, this study proposes a novel node demand allocation method based on real-time cellular signaling (CS) data, which leverages dynamic population proportions to more accurately capture the relationship between real-time population distribution and water consumption in urban areas. A real-world WDS with three water use scenarios (weekdays, weekends, and holidays) demonstrates the utility of the proposed method, and results show that: (1) the proposed CS method achieves the lowest WDS simulation errors, with a maximum absolute error of 1.15 m at pressure sensors and a maximum absolute relative error of 3.71 % at flow sensors, outperforming the BU and TD methods
(2) the incorporation of real-time cellular signaling data allows for precise demand allocation by accurately capturing population fluctuations, particularly in high-mobility areas
and (3) the method's high accuracy and adaptability make it well-suited for complex urban water distribution systems, with seamless integration into GIS and WDS platforms, enhancing its practical implementation.