Solute Transport Application

The Melton Branch Watershed is a 0.67 ha subcatchment on the Oak Ridge Reservation of the Department of Energy and is covered with deciduous forest and located in the ridge and valley southern Appalachian area. The top soils of the site can be roughly categorized into A, B, and C horizons that are weathered or partially weathered from the limestone-shale interbedded bedrocks. The A horizon is very rich in organic matters with little water capacity. The B horizon is mostly silt and clay within 1 meter of depth from the surface. Both A and B horizons contain a fair amount of plant roots. Below the B horizon, the C horizon could go as deep as 3 to 5 meters. The saprolite in the C horizon is in fact partially weathered shale with fracture density around 200 per meter. The fracture walls are usually coated with iron and manganese oxides and the fractures are filled with translocated clay particles.

The site is fairly well instrumented, with 15 banks of piezomenter, solution samplers and tensiometers at three depth. An additional 15 piezometers, including one groundwater well, are located at two transects. Each of these wells is interfaced with an on-site computer. A line source buried at 50 cm deep is located on the top of the hill that has been used for Cl and Br tracer release. A tipping bucket rain gage is sitting nearby to record rainfall rates. Down at the bottom of the hill, there is a subsurface trench measured 16 meters wide and 3 meters deep. Six stainless steel pans were pressed into the soil horizons to collect subsurface storm flow from the hillslopes. The water from the top three pans was piped into two tipping bucket rain gages sitting on one end of an H-flume. The water level in the H-flume was recorded by an ultrasonic water level detector that was interfaced with the on-site computer. The water from the lower three pans was similarly routed and the flux rates were similar recorded.

During the storm of February 17-23, 1991, we released a batch of CaCl₂ and MgBr₂ through the buried line source on top of the hill. The release rate stayed fairly constant at 0.194 m³/h for 16.5 hours. The concentration of Br as MgBr₂ was kept constant at 8000 mg Br/L. We monitored the plume of Br for this and the subsequent storms for more than three months, taking solution samples from the tipping buckets and H-flumes in the trench frequently during storms. The chart on the left shows the initial breakthrough of the Br tracer, in red, and the subsurface flux rate of the C horizon in broken pink lines. The Br concentration in the water appears to fluctuate with the amount of recharge from the rain. Our model prediction, using a three-pore-region conceptual model and the multiregion flow and transport codes, 3DMURF and 3DMURT, respectively, is depicted in blue. Most of the hydraulic parameters of this model were taken without modification from laboratory and field data. We slightly tweaked the mass transfer coefficients in a previous calibration run and the macrodispersion coefficients were upscaled from those obtained using laboratory soil column injection data. The initial arrival of the prediction is about five hours off the observation and we have seen a lot more variation in the predicted concentration. This variation, in our three-region model results shown here, can be seen largely caused by the fast flow in the macropores that may channel most of the rainwater in heavy storms. From the correlation between the observed Br concentration and the flux rates, we have enough evidence that this interpretation by the three-region model is correct. We also observed similar responses to the recharging water in the mesopores, which are believed to have caused preferential flow under partially saturated conditions. The uncertainty here is very likely a bit higher than the macropores, but from the contrast between concentrations of the mesopores and the micropores in which Br concentration increased gradually with the coming and going of storms, we have to suggest that preferential flow under partially saturated conditions is not unlikely. Overall, given very minimal amount of geological information at the time of the modeling, we have a prediction of roughly a factor of two of the observed Br concentration. Looking back, I wished I had more time and data to characterize the uncertainty of our prediction, but the project was about to be wrapped up in a hurry in a month or two.