Separating flow components in tile drainage and streamflow in a headwater agricultural catchment

This study was conducted in a headwater, tile-drained, agricultural catchment in southern Ontario, under a typical cropping system in the Great Lakes Region. Hydrograph separation has become a common approach to infer water sources and nutrient pathways to stream discharge but is less frequently used to separate tile drain effluent. The identification of hydrologic and nutrient pathways into both streams and tiles is important in agricultural management and modelling. We separated ten event hydrographs for one contributing tile and stream outlet site over a one-year period using two graphically-based approaches and a mass balance approach using a variety of common geochemical tracers. Few studies to date have examined the application of various hydrological separation techniques in tile discharge across a range of hydrometric conditions. Although methods differed in their operational definition of baseflow, we observed similar baseflow estimates across differing methods during wet, spring conditions and diverging estimates were found for snowmelt periods. When calibrating a graphically-based method with tracers, the resulting hydrographs became less flashy with smaller peak baseflow amounts compared to the tracer-based separation alone and is therefore not recommended when intra-event dynamics are of importance. Findings show that electrical conductivity can be successfully used in headwater agricultural streams and tile drainage discharge to separate flow components, but sampling additional tracers during critical periods, such as snowmelt or following fertilizer application, is recommended to ensure the most appropriate tracers and adequate end-member representation is acquired. • Variability in baseflow estimates is similar across methods for tile and streamflow. • Lowest variability in results across methods occurred during wet, spring events. • There is no “one size fits all” tracer. • Combining methods through calibration not recommended for sub-event timescales.