Streamflow generation on a small agricultural catchment during autumn recharge: II. Stormflow periods

To increase understanding of hydrologic flow paths, two- and three-component tracer models were used to determine stormflow components on a 19.8 ha agricultural catchment in central Pennsylvania. Three events were studied during the 1989 autumn recharge period with the first and largest storm having a 5–10 year return period, 80 mm rainfall. Continuous precipitation and streamflow records were augmented by observations and sampling of spring and seepage discharge along with shallow wells and soil water. During the largest storm event, a two-tracer, three-component model separation using 18O and Si indicated that 42% of total flow was derived from shallow subsurface stormflow, 11% from surface event water, and 47% from deep subsurface groundwater flow. Similar analysis using Si and NO 3 for a 34 mm rainfall event showed 17% of total flow to be from shallow subsurface flow, 10% from surface event water and 73% from deep subsurface groundwater flow. Based upon analysis of tracer concentration variations in soil water, shallow well water and springflow during these events, shallow subsurface stormflow was found to be a transient mixture of infiltrated event water (28% average maximum contribution) and stored pre-event water (72% average minimum contribution). Thus, two major pathways for transferring event water to streamflow existed in these two events: (1) overland flow and channel precipitation (10–11% of total flow); (2) shallow subsurface stormflow (5–12% of total flow). In the third event analyzed, 9 mm of rainfall was insufficient to generate shallow subsurface stormflow, and two-component models using 18O, Si and NO 3 indicated 10%, 4% and 3% event water contributions, respectively. Major early autumn storms, occurring before subsurface water becomes well mixed in this region, offer an opportunity to differentiate shallow and deep subsurface pathways for streamflow generation using two-tracer, three-component models.