A novel Bayesian Hierarchical Network Model (BHNM) is designed for ensemble predictions of daily river stage, leveraging the spatial interdependence of river networks and hydrometeorological variables from the upstream catchment area between gauge stations. The model allows parameters to dynamically vary over time, influenced by chosen covariates specific to each day. By utilizing the river network's structure to integrate flow and stage data from upstream gauges, along with precipitation data from the immediate surrounding area, the model efficiently captures the spatial correlation of stage. An empirical application focusing on the daily monsoon period (July–August) stage is demonstrated at three-gauge stations on the Narmada basin in central India, spanning from 1978 to 2014. The optimal set of covariates comprises daily streamflow and stage readings from upstream gauges based on travel times, alongside daily precipitation data from the region between two stations. Model validation of the forecasts and uncertainty estimates, show skillful and reliable predictions compared to a basic linear model. This has potential applications for flood warning and mitigation.