THE EFFECTS OF FLOOD DYNAMICS ON CHEMICAL FATE 
AND TRANSPORT: IMPLICATIONS FOR AGRICULTURAL 
MANAGEMENT PRACTICES

Timothy R. Ellsworth, Stefan Mayer, Charles W. Boast, and Miguel Restrepo

University of Illinois at Urbana-Champaign
 

Abstract

Chemical transport in soil has rarely been simulated with accuracy, as a consequence of subsurface heterogeneity. For instance, classical single continuum approaches have failed in describing preferential transport and this has recently resulted in a plethora of multiple continuum modeling approaches, in which continua are coupled via transfer coefficients and gradients for mass and energy transfers. These approaches have proven useful in many contexts, yet the ability to use such approaches in inverse modeling has been limited due to model complexity and also due to the fact that such models have a discrete scale of applicability. The goal of our work has been to develop a discrete three-dimensional network flow and transport model that can be used in an inverse mode for estimating transport properties from observed solute concentration data, as well as a framework for developing upscaled transport descriptions. 

We have developed a computer model that simulates flow and transport in a 3-D heterogeneous flow network. This approach employs an efficient solution method, which is based on the assumption that iso-pressure surfaces are known a priori (for instance, steady-state, gravity-driven vertical flows in unsaturated soil with horizontal iso-pressure surfaces). The flow and transport heterogeneity in this system is assumed to be a result of conductance heterogeneity. We have simulated flow and transport processes under widely contrasting conductance distributions. In generating these conductance distributions, we have varied the spatial correlation structure (both isotropic and anisotropic structures), as well as the heterogeneity of these distributions. The flow and transport solution methods readily allow for highly variable flow fields (conductance values range from 10² to 10¹² in different simulations). Particle tracking is employed to characterize solute transport in the network. The model allows solute input as either a flux boundary condition or an arbitrary initial soil distribution. Detection/measurement methods include spatial resident concentrations as well as exit boundary flux concentrations. 

These simulation methods allow us to quantify the internal variability structure of spatial plume evolution, as well as detailed breakthrough curve estimation. The model results are very sensitive to conductance spatial structure. In generating the conductance distribution, we first employ a geostatistical simulation using either sequential indicator simulation or sequential gaussian simulation. Then, we modify the simulated distribution to be consistent with the constraints imposed on the flow field. We have employed several alternative methods for ensuring this consistency. The iso-pressure assumption requires that the sum of incoming segment conductance values for a given node is proportional to the sum of all outgoing segment conductance values for that node. In addition, it further requires that this proportionality ratio be the same for all nodes on a given iso-pressure contour. Of the several methods we have evaluated to satisfy this constraint, the one that maintains the greatest similarity to the original conductance distribution is based on a "local" optimization at the node scale.