Using long-term data sets to understand transit times in contrasting headwater catchments

摘要

Long-term tracer data collected over an 8 year period were analyzed to explore the transit times of two small (-1 km~2), contrasting headwater catchments in the uplands of Scotland. At Loch Ard, the catchment was characterized by low permeability gleyed soils overlying metamorphic geology. At Sourhope, more freely draining podzolic soils were dominant, which mantled fractured and faulted volcanic rocks. Hydrometric data and chemically-based hydrograph separations indicated that Loch Ard was a flashy catchment dominated by runoff processes in the upper soil horizons. In contrast, around 77% of annual flow at Sourhope was sustained by well-buffered groundwater sources. Weekly Cl~- time series in precipitation and stream flow revealed similar variability in inputs at both sites, but much greater damping in outputs at Sourhope. Despite this, both catchments filtered white noise frequencies in precipitation inputs into 1/f outputs. These input-output relationships were modeled with a range of transit time distributions (TTD). At the responsive Loch Ard catchment, mean transit times (MB') for the study period were estimated at 135-202 days. Models based on a gamma distribution or two parallel linear reservoirs were best able to capture the short- and long-term fluctuations in stream water in response to input variations. At Sour-hope, the highly damped tracer signal in stream waters was poorly captured by all the TTDs used. Resulting MIT estimates of 1830-1970 days are based on weak model fits and poorly identifiable parameter sets, indicating that natural tracers such as Cl~- are inadequate for catchments where MTTs are greater than a few years. At both sites, estimates of MTT using moving windows over the 8 year data sets revealed sensitivity to precipitation amounts and the length of monitoring period. It is concluded that time series of around 4 years are required to adequately constrain MTT estimates.