Uncertainty in Runoff and Recharge for Surface-Water/Groundwater Models in the Arid and Semiarid Edwards Plateau, Central Texas

The volume and distribution of recharge is arguably the most critical source of uncertainty in both surface-water and groundwater models, as recharge estimates often carry high potential for error. This uncertainty is magnified when attempting to quantify recharge in arid and semi-arid environments, where precipitation may be insufficient to instigate distributed recharge. Precipitation is often intensely concentrated in space and time, and recharge occurs in pulses through paths of preferential flow such as sinks or stream beds. If karstic conduit flow is well developed in the groundwater system, the boundary between the flashy baseflow elements and direct runoff can be blurred. Often, integrated surface-water/groundwater models are developed in an attempt to gain insight on recharge mechanisms and rates. Unless this uncertainty is adequately constrained, models cannot be trusted to generate meaningfully accurate insights for use in water-resource management planning. Furthermore, most surface-water modeling methods are built on assumptions and conceptual models developed for engineering purposes in humid areas, few of which can be applied without alteration to arid and semi-arid environments. Integrated surface-water/groundwater models have been used to explore recharge, runoff, and the related sources of uncertainty for both processes in the arid and semi-arid Edwards Plateau in central Texas. This experience provides insight on how to constrain conceptual and numerical models used to simulate surface-water and groundwater flow under variable precipitation and recharge conditions. Elements of the surface-water/groundwater models that constitute the greatest sources of uncertainty include (1) channel transmission losses to groundwater; (2) magnitude of precipitation; (3) the presence and size of a “tension zone” soil storage element in which water is vulnerable to evapotranspiration but not infiltration; (4) the calculation method for actual evapotranspiration; and (5) the level of discretization compared to the temporal and spatial variability of the system.