Riparian buffer strips (RBS) may protect surface water and groundwater in agricultural settings, although their effectiveness, observed in field-scale studies, may not extend to a watershed scale. Hydrologically-controlled leaching plots have often shown RBS to be effective at buffering nutrients and pesticides, but uncontrolled field studies have sometimes suggested limited effectiveness. The limited RBS effectiveness may be explained by the spatiotemporal hydrological heterogeneity near non-irrigated fields. This hypothesis was tested in conventional corn and soy fields in the St. Lawrence Lowlands of southern Quebec (Canada), where spring melt brings heavy and rapid runoff, while summer months are hot and dry. One field with a mineral soil (Saint-Rochde-l'Achigan) and another with an organic-rich soil (Boisbriand) were equipped with passive runoff collectors, suction cup lysimeters, and piezometers placed before and after a 3 m-wide RBS, and monitored from 2011 to 2014. Soil topography of the RBS was mapped to a 1 cm vertical precision and a 50 cm sampling grid. On average, surface runoff intersects the RBS perpendicularly, but is subject to substantial local heterogeneity. Groundwater saturates the root zones, but flows little at the time of snowmelt. Groundwater flow is not consistently perpendicular to the RBS, and may reverse, flowing from stream to field under low water flow regimes with stream-aquifer connectivity, thus affecting RBS effectiveness calculations. Groundwater flow direction can be influenced by stratigraphy, local soil hydraulic properties, and historical modification of the agricultural stream beds. Understanding the spatiotemporal heterogeneity of surface and groundwater flows is essential to correctly assess the effectiveness of RBS in intercepting agro-chemical pollution. The implicit assumption that water flows across vegetated RBS, from the field to the stream, should always be verified.