Simulation of Groundwater Flow and Effects of Groundwater Irrigation on Stream Base Flow in the Elkhorn and Loup River Basins, Nebraska

This report presents the phase-two groundwater-flow simulations and predicted effects of groundwater irrigation on stream base flow in the Elkhorn and Loup River Basins of central Nebraska. 

Simulation of Groundwater Flow and Effects of Groundwater Irrigation

This report describes the construction and calibration of the simulations and the methods used to predict changes to stream base flow that result from changes to groundwater irrigation. Effects of groundwater irrigation were evaluated using three distinct approaches: (1) a base-flow depletion analysis, derived from results of the model simulation, mapped the spatial distribution of the percentage of pumped water that causes base-flow depletion at the end of a 50-year period; (2) groundwater-flow simulations were used to predict changes to stream base flow that resulted from changes to the amount of irrigated area during a 25-year period; and (3) a simulation-optimization model determined the minimum reduction of groundwater pumpage that would be necessary in the Elkhorn River Basin to maintain various hypothetical levels of base flow in the Elkhorn River. The climate, land use, water use and management, hydrogeology, and general description of the conceptual model were described by Peterson and others (2008) and are not presented herein.

Simulation of Groundwater Flow 

This section describes the topical background, methods, and results for developing the phase-two simulation of groundwater flow. The simulation, or model, was developed to simulate groundwater flow, groundwater withdrawals, and stream-aquifer interactions for the Elkhorn and Loup River Basins, Nebraska. To simulate those processes, large amounts of hydrogeologic data from numerous sources were needed to describe aquifer properties and hydrologic stresses. These data were compiled as spatially referenced data layers within a geographic information system (GIS) and then assigned to the simulation at discrete intervals in space and time. Simulations were built for this study using MODFLOW–2005 (Harbaugh, 2005), with assistance from Groundwater Vistas Version 5 software (Environmental Simulations, Inc., 2009).
The hydrogeologic data (simulation parameters) describing the study area were assigned to the simulation directly and through calibration. For the direct case, characteristics such as recharge, land use, streambed properties, and hydraulic conductivity were introduced into the simulation using the best available information (appendix 1) and used as compiled. Once all available information was compiled and entered into the simulation, the results from the simulation were compared to measured groundwater levels, decadal groundwater-level changes, and estimated groundwater discharge to streams (hereinafter referred to as base flow). Differences between simulation results and values were used to guide calibration, which is the process of obtaining parameter values to construct a framework useful for describing the hydro-geologic characteristics of the study area (Reilly and Harbaugh, 2004). Simulations were calibrated by adjusting selected parameters until simulated groundwater levels, decadal groundwater level changes, and base flow best reproduced measured values
(see “Calibration” section of this report). Calibration proceeded in two stages. In the first stage, manual trial-and-error calibration techniques were used to adjust average recharge from precipitation, additional recharge beneath irrigated and non irrigated cropland, horizontal hydraulic conductivity (KH), streamed hydraulic conductivity (KS ), and maximum evapotranspiration (ET) rates from groundwater to achieve the best match with measured groundwater levels, decadal groundwater-level changes, and base-flow data. Recharge from precipitation was calibrated as a constant, average rate throughout the simulation period rather than a time variable rate during the manual trial-and-error calibration stage. The second stage of calibration used automated calibration techniques and incorporated recharge from precipitation as a temporally changing value. The automated, or inverse modeling, calibration stage used the Parameter Estimation software (PEST) (Doherty, 2008a, 2008b) (appendix 2). Adjustable parameters for the automated calibration were recharge from precipitation and KH.

Spatial and Temporal Discretization 

To simulate flow using MODFLOW, the study area is divided into a grid of discrete cells. Hydrogeologic properties, initial conditions, and simulation results are assigned to each grid cell. The actual hydrogeologic system is continuous rather than discrete; therefore, groundwater-flow simulations are always an approximation of the actual system. Simulations with a smaller grid-cell size generally yield more accurate approximations of the actual system because less averaging occurs as spatially variable properties are assigned to grid cells, especially where large changes take place over small distances. The study area was simulated using a uniformly spaced grid of 162 rows and 248 columns of 1-mile (mi) by 1-mi cells, covering an area of 40,176 mi2 (fig. 2). This is a refinement of the phase-one simulation, which used grid cells 2 mi by 2 mi in extent. The active simulation area, which is smaller than the extent of the model domain, encompasses 29,707 mi2 and includes areas with an estimated aquifer saturated thickness of at least 10 feet (ft). Similar to the phase-one simulations, a single unconfined layer was used to simulate the aquifer.

If a simulation is used to evaluate the aquifer system as a function of time, it is referred to as a transient simulation and is divided into discrete time intervals called stress periods. Hydrologic stresses, such as recharge and pumping, are held constant within each stress period. In the ELM study area, major changes in land-use practices occurred from 1895 to 1940 and from 1940 through 2005. Starting in 1895, irrigation canals were constructed, and water was diverted from streams for agriculture. Simulation of conditions from 1895 through 1939 used two stress periods (1895 to 1929 and 1929 through 1939) to represent the two time periods when new canal systems became operational and caused a change to recharge from canal seepage (see “Additional Recharge from Canal Seepage”). From 1940 through 2005, irrigated agriculture expanded to include wells and additional canals. The 1940 through 2005 period was simulated using 66 stress periods, one stress period for each year. Simulated hydrologic stresses were updated during each of those annual stress periods so that changes to land use and irrigation development with time are represented in the simulation. Groundwater levels were needed to represent 1895 conditions at the beginning of the simulation (Reilly and Harbaugh, 2004) and measured groundwater levels were unavailable during 1895; therefore, a pre-1895 period was simulated to represent the system in long-term equilibrium, or steady-state conditions. When a steady-state simulation is used to define starting conditions for a transient simulation, the steady-state simulation uses the same aquifer properties and hydrologic stresses, with the exception of stresses such as pumping. This period was simulated using a single transient stress period that was 1,000 years long. It was determined that 1,000 years was a sufficient amount of time to reproduce long-term equilibrium conditions because simulated change to groundwater storage was close to zero. This approach was used in place of a true steady-state stress period, a single stress period having a single time step and a storage term set to zero, because it helped prevent numerical instability in the far northeast corner of the study area and resulted in fewer dry cells. Dry cells are cells that become inactive when calculated interim groundwater levels drop below the simulated base of
the aquifer during the iterative approximations of groundwaterflow equations (Harbaugh, 2005). During calibration, simulated results of the pre-1895 period were not compared to calibration targets; however, 1895 simulated groundwater levels were used as starting groundwater levels for the 1895 through 1939 simulation, and 1939 simulation results were compared to measured groundwater levels and estimated base flows (see “Calibration” section of this report). This was considered appropriate because water development from 1895 through 1939 only occurred in a relatively small area along the southern boundary of the simulation.

Download Link

https://on.doi.gov/2Il2IQc