Abstract

TOUGH2 and iTOUGH2 are used to conduct forward and inverse simulations of a large-scale infiltration experiment at the Maricopa Agricultural Center (MAC) near Phoenix, Arizona. Three site representations are considered: uniform horizontal layers, layers composed of uniform segments, and layers having randomly varying properties. Due to a paucity of hydraulic parameter measurements at the MAC, these are inferred from soil compositional data using generic data bases and pedotransfer functions. Variogram analyses of these data support the laterally nonuniform and randomly variable site representations. To reproduce observed water contents it is necessary to modify the inferred hydraulic parameters through inverse modeling based on preliminary sensitivity and error analyses. Model discrimination criteria are used to rank the three calibrated site models. Layers composed of uniform segments are ranked highest due to their superior performance and relative simplicity. The choice representation is validated by using it to reproduce water contents during another infiltration experiment.

Citations