Stream gages track data vital for water resources management. Frequently collected streamflow data are invaluable to many hydrological analyses, such as the assessment of the magnitude and frequency of floods or droughts, evaluation of water quality for implementing the Total Maximum Daily Loads (TMDL). The United States Geological Survey (USGS) operates about 85% of all stream gages in US. According to USGS, average operation cost was about $14,000 annually per typical continuous stream gage (Norris et al., 2007) with most cost associated with site inspections and field experiments. Due to the cost limitation, most gages are deployed to large rivers, and the number of gages is rapidly decreasing over the past decades as a result of the shrinking budget allocation. Meanwhile the need for high quality streamflow data is ever increasing in order to meet the demand for better water resource management and resilience to climate change. For example, there are increasing demands for river stage data from smaller tributaries, which are very few but can greatly improve the performance of flood-forecasting models and warning systems (Kruger et al., 2016). To overcome these challenges, innovative solutions are demanded to substantially reduce the cost and hazards associated with current stream gauging operations.Despite great advances in flow measurements technologies, most USGS stream gages apply the identical method that has been used for more than 100 years. Specifically, discharge is not measured directly, it is calculated from the height of water surface, that is, the "stage," which can be measured more conveniently. Discharge is directly measured only periodically following the velocity-area method (Turnipseed & Sauer, 2010). An empirical relation between discharge and stage is established with repeated field experiments. The relation, or the "rating-curve," can then be applied to report discharge continuously based on stage measurements. Since it is often difficult, sometimes dangerous, to conduct field experiments during high flows, extrapolation of the rating curve is unavoidable to estimate discharge of extreme flood flows. Substantial errors can be introduced with rating curve extrapolation, which causes great uncertainties to flood frequency analysis (Lang et al., 2010;Le Coz et al., 2014).Non-contact or remote sensing solutions may solve difficulties of standard river gaging processes. Kruger et al. (2016) showed that bridge mounted ultrasonic sensors can be a low-cost alternative to traditional river gaging methods. It has been demonstrated that radars can be applied to detect both water depth and surface