Particle Tracking-Based Strategies For Simulating Transport in a Transient Groundwater Flow Field at Yucca Flat, Nevada Test Site, USA

Developing probabilistic-based calculations of contaminant concentrations over the next 1000 years at Yucca Flat, Nevada Test site, require tremendous computational effort in this highly complex hydrogeologic surface environment. The sources of contamination, underground nuclear tests conducted between 1951 and 1992, not only released radionuclides to the subsurface but also created abrupt, significant changes in rock properties and caused large transients in the measured hydraulic gradients. To efficiently model contaminant migration from these sources we use a particle-based approach within a transient flow field. Here, we present results using two methods; first, an explicit representation of time-varying sources using large numbers of particles introduced at source-specific rates over time, each representing a unique mass of solute. This method provides good results, but is computationally expensive since sensitivity to uncertainty in source term and transport parameters can only be explored with discrete process-model runs. The second method employs a convolution method (PLUMECALC) which can efficiently consider a large number of variations in the source terms and in certain transport parameters with a single process-model run. Implementation of this second approach required extension of the existing methodology to conditions of transient flow. We find very good comparison between the two methods on small test problems and excellent computational advantages when applying the convolution method in the NTS application