UNCERTAINTY IN DETERMINISTIC GROUNDWATER TRANSPORT MODELS DUE TO THE ASSUMPTION OF MACRODISPERSIVE MIXING - EVIDENCE FROM THE CAPE-COD (MASSACHUSETTS, USA) AND BORDEN (ONTARIO, CANADA) TRACER TESTS

Deterministic transport models based on the advection-dispersion equat ion are widely used to simulate groundwater contaminant transport. Onl y the largest heterogeneities and velocity field variations are explic itly modeled by the advection part of such models because subsurface e xplorations allow limited understanding of the distribution of heterog eneity and velocities. Smaller heterogeneities and associated velocity field variations are not incorporated in the modeled velocity field, but their overall mixing effect is represented implicitly as macrodisp ersion. As a result, such models do not replicate the complex small-sc ale variation of actual concentration distributions, but instead simul ate a smoother concentration distribution. This discrepancy causes sig nificant uncertainty in modeled concentrations. In this paper, such un certainty is quantified for the detailed concentration distribution da ta sets of the Cape Cod and Borden natural-gradient tracer tests. Mode ls of these tests could be made with relatively little uncertainty abo ut the source distribution, large-scale flow field, and apparent macro dispersivities. As earlier moment analyses reveal, the ensemble-averag e bromide migration in both tests was approximately consistent with cl assical advection-dispersion theory. Therefore, the reported uncertain ties are primarily due to the use of macrodispersivity to represent mi xing caused by small-scale velocity field variations. Analytic three-d imensional transport models were used to simulate the migration of bro mide, a non-reactive tracer. The distribution of log(c(a)/c(m)), where c(a) is actual concentration and c(m) is modeled concentration at the same point, had a standard deviation of similar to 0.70 for both test s. The distribution of vertically-averaged concentration predictions, log(Sigma c(a)/Sigma Zc(m)), where the summation is over each multi-le vel sampler, had a standard deviation of similar to 0.45 for both test s. Comparing the peak actual concentration to the peak modeled concent ration at any given time results in a standard deviation of similar to 0.12 in the statistic log(c(a(max))/c(m(max))) for both tests. Althou gh the uncertainties listed above pertain to the scales of un-modeled velocity variation in these models at these sites. the reported uncert ainties could serve as lower bound estimates for most deterministic mo del applications. Uncertainty due to the assumption of macrodispersive mixing tends to increase as the plume scale decreases or as the scale of un-modeled velocity field variations increases.