Flow and sediment transport dynamics in fluvial systems play critical roles in shaping river morphology, in the design and use of riverine infrastructure, and in the broader management of watersheds. However, these properties are often difficult to measure comprehensively. Previous work has suggested the use of proximal seismic signals resulting from flow and bed load transport to construct more complete records of these fluvial processes. We investigate a small (184 km2; < 20 m3/s), snowmelt‐fed mountain river in the Northern Colorado Rocky Mountains during May–August 2015 to capture peak runoff with colocated measurements of discharge and seismic noise. Three‐component seismometers were placed in close proximity to the channel bank (~1 m) within the hyporheic zone (at times submerged beneath the water table). We recorded a broad spectrum of seismic signals excited by discharge, including novel, low‐frequency (0.1–2 Hz) signals observed predominantly on the horizontal components. The characteristics of these low‐frequency signals are not consistent with an elastically propagating seismic wave. We instead infer that they likely arise from the sensor tilting in response to viscoelastic deformation occurring near the channel and propose large‐scale turbulent eddies as a forcing mechanism. Calibrating horizontal seismic power with hydrograph flow rates over the course of a rainstorm for individual sensors, we demonstrate that these unique signals can be used to accurately estimate river discharge with simple regressions. This technique shows promise for augmenting seismic monitoring of rivers by enabling discharge rates to be estimated from outside the channel using easily deployed and noninvasive seismic instrumentation.