MODEL COMPLEXITY: A PERSONAL ODYSSEY

All models are a simplification of reality; some are necessarily more complex than others. The level of complexity for any model should be dictated by the purpose for which it is intended to serve. This point is illuminated through three numerical models of varying complexity developed over the past 20 years. The first model is an application to evaluate the relative merit of several remedial measures based on the cleanup time estimate for a contaminated aquifer. A simple linear reservoir model was used to compute the number of pore volumes of clean water that must be flushed through the contaminated aquifer to achieve a desired concentration target. Then a particle tracking program was used to calculate the advective travel time for a single pore volume, which, multiplied by the number of pore volumes, yielded the estimate of cleanup time. While the model was highly simplified, it was consistent with the availability of data and served the intended purpose well. The second model is an application to analyze a natural-gradient tracer test conducted in the Columbus Air Force Base in Mississippi (the MADE site). In spite of over 3000 hydraulic conductivity measurements and over 6000 concentration data within a small test “cube” of 100 m by 250 m by 10 m, a conventional advection-dispersion model was found to incapable of reproducing the observed tracer plume. Some added complexity, i.e., a mass transfer process between connected preferential flow paths and less permeable matrix, had to be introduced to achieve an improved match with the field observation. The third model is an application to explore uranium fate and transport in a physically and chemically heterogeneous site in Hanford, Washington. Previous transport modeling efforts based on the equilibrium-controlled retardation factor (K_d) approach had failed to predict the dynamics of a uranium plume adjacent to Columbia River and the mass flux discharged into the river. Consequently, multi-component, multi-rate surface complexation reactions were shown to be key to understanding uranium fate and transport at the study site. The resulting model was highly complex, but it was necessitated by the complex nature of the physical and chemical processes that govern uranium fate and transport.