Heterogeneity Preserving Inversion Methods for Vadose Zone Flow Modeling
Natural recharge to groundwater in semi-arid regions relies on unsaturated flow through an often deep vadose zone. Understanding and modeling such flow requires intimate knowledge of surface-atmosphere interactions (which includes complex patterns caused by seasonally active vegetation) but also the subsurface heterogeneity of hydraulic properties. It is expensive and labor intensive to measure such properties, especially for the extensive and deep vadose zones which are prevalent in the western United States and other densely populated semi-arid regions in the world. Pedotransfer functions (PTFs) offer a cheap means to estimate hydraulic properties from soil or sediment texture, but suffer from inaccuracies that would likely bias modeled deep vadose zone flow.
In pursuit of the question “How much subsurface heterogeneity must be accounted for in vadose zone flow modeling?” we quantify the effectiveness of several methods for generating “fields” of subsurface hydraulic properties as needed for 1D to 3D numerical simulations. The main line of the presentation will deal with a detailed study carried out at Maricopa, Arizona. The study site is a 50×50 meter and 15 meter deep vadose zone at which a 28-day constant-rate infiltration experiment was conducted in 2001. Moisture content at this site was measured with neutron thermalization at 400 locations daily during the infiltration period, and at irregular intervals 100 and 200 days prior to and after infiltration, respectively. Our research shows that direct simulations based on PTF-estimated heterogeneous fields of hydraulic properties poorly represent the measured infiltration plume. However, several types of model inversions using different generalizations of heterogeneity present in the PTF estimates yield acceptable results and offer a potential means to limit the collection of site-specific data. In addition to the Maricopa site, we will also briefly outline surface soil-climate-deep vadose zone interactions for managed systems (a hypothetical golf course) and rangelands for long (10-25 year) time series. Much of this work is still ongoing, but presently these simulations indicate that the type of surface soil and the amount of winter rain strongly affects the amount of deep vadose zone infiltration. Conversely, surface soils as well as the amount of summer rain (the North American monsoon) have very limited effect on long-term nonriparian infiltration.