Calibration of a Deterministic Net Infiltration Model for the Death Valley Region Using Measured Daily Stream Flows

Recharge estimates are needed to define the upper boundary conditions for groundwater flow models used to analyze water contamination and availability in the Death Valley region of southeastern California and southern Nevada. Estimated net infiltration, where net infiltration is defined as percolation below the root zone, was used as an indicator of potential recharge. Spatially distributed net infiltration was estimated using a deterministic soil-moisture model that incorporates distributed parameters to define basin characteristics, including topography, vegetation, soil properties, and bedrock geology. Daily soil moisture, evapotranspiration, runoff, surface-water discharge (stream flow), and net infiltration (based on a maximum root zone depth of 6 meters) are simulated using the defined basin characteristics and climate input consisting of continuous records of daily precipitation and air temperature from a network of monitoring sites. The deterministic model was calibrated using a trial-and-error approach of matching simulated and measured daily stream flow records at 26 stream-gaging sites in the Death Valley region. Model parameters were selected for optimization based on high uncertainty in parameter values, and were adjusted through a range of values during the calibration process. The selected parameters were bedrock hydraulic conductivity, a root-zone available water storage term for bedrock, root density coefficients for bedrock and soils, estimates of summer and winter storm durations, soil hydraulic conductivity for stream channels, and parameters defining an empirical channel flow area function. An optimized parameter set for the Death Valley region was manually created on the basis of the collective calibration results for the majority of the 26 stream flow records. Calibration of the regional-scale net infiltration model was difficult owing to: (1) available daily climate records did not accurately define the occurrence, magnitude, and spatial distribution of localized summer thunderstorms, and did not accurately define precipitation intensity for both winter and summer storms; (2) uncertainty in stream channel characteristics, such as channel geometry and soil properties; (3) divergent surface-water flow on alluvial fans and across normally dry playa lakes. Although model calibration proved to be difficult for some storms and for some of the stream-gaging sites, the net infiltration model was found to be satisfactory in predicting the general timing and magnitude of the larger, more widespread winter storms recorded for the Death Valley region. The final set of optimized parameters was developed using only winter storm events and also using comparisons of simulated monthly and annual stream discharge with recorded discharge. The calibrated net infiltration model was applied using daily climate records from 1950 through 1999 to develop an estimate of modern climate net infiltration for the Death Valley region. The net infiltration model was evaluated by comparing simulated net infiltration with published basin-wide recharge estimates from previous studies using alternative techniques, such as empirical precipitation-recharge models and the chloride mass balance method. Results from model evaluation indicate a need to expand model calibration using additional sources of data, such as snow cover and soil moisture measurements.