Summary


The permeable volcanic bedrock of the upper Klamath Basin hosts a substantial regional groundwater flow system that provides much of the flow to major streams and lakes. These streams and lakes, in turn, provide water for wildlife habitat and are the principal source of irrigation water for the basin’s agricultural economy. Increased allocation of surface water for aquatic wildlife in the past decade has resulted in increased reliance on groundwater for irrigation. The potential effects of increased groundwater pumping on groundwater levels and discharge to springs and streams has caused concern among irrigators dependent on groundwater, resource managers, wildlife biologists, and other stakeholders. In order to better understand the groundwater hydrology of the basin, to provide information on the potential impacts of increased groundwater development, and to aid in the development of groundwater management strategies, the U.S. Geological Survey (USGS), in collaboration with the Oregon Water Resources Department and the U.S. Bureau of Reclamation, developed a groundwater flow model that can simulate the response of the hydrologic system to these new stresses. 


The flow model, which is described in this report, was developed using the USGS MODFLOW finite-difference modeling code. Model cells have lateral dimensions of 2,500 feet (ft) by 2,500 ft and are aligned in a grid consisting of 285 east-west trending rows and 210 north-south trending columns covering the entire upper Klamath Basin. In the vertical dimension, the model consists of three layers of varying thicknesses ranging from about 5 ft to 3,600 ft, depending on topography and proximity to the edge of the model. Hydraulic characteristics of subsurface materials are represented in 18 hydraulic parameter zones reflecting large-scale geologic conditions. Hydraulic parameter zonation is simpler at depth due to the lack of detailed geologic information.


Boundary conditions include specified-flux boundaries and head-dependent flux boundaries. Most boundaries with adjacent basins, as well as the contact with underlying low-permeability early Tertiary strata, are formulated with specified fluxes of zero. Groundwater recharge and pumping are simulated as specified fluxes varying each quarterly stress period. Head-dependent flux boundaries include streams, lakes and reservoirs, agricultural drains, evapotranspiration directly from aquifers in areas of shallow groundwater, and boundaries with adjacent basins in selected areas. All major streams and most major tributaries with substantial groundwater discharge are included in the model. 


The model was calibrated using inverse methods to transient conditions from 1989 to 2004. Calibration data included 5,636 head measurements from 663 wells. Of these, 444 wells had time series consisting of 2 to 64 observations. Estimates of average groundwater discharge were available for 52 stream reaches or spring complexes. Time series of estimated groundwater discharge were available for 10 stream reaches or springs. The calibration data show that the groundwater system in the upper Klamath Basin responds to decadal climate cycles, with groundwater levels and spring flows rising and falling in response to wet and dry periods. Groundwater levels also show seasonal and year-to-year fluctuations in response to groundwater pumping. 


Calibrated hydraulic conductivity values span nearly four orders of magnitude, ranging from 5.9×10^–6 feet per second (ft/s) for Quaternary volcanic rocks in the southern part of the model area to 1.2×10^–2 ft/s for Mazama tephra deposits. Late Tertiary volcanic deposits range from 1.0×10^–5 ft/s to 9.3×10^–4 ft/s. Quaternary volcanic deposits (other than Mazama tephra deposits) range from 5.9×10^–6 ft/s to 4.0×10^–5 ft/s. Late Tertiary sedimentary strata range from 2.9×10^–4 ft/s to 3.5×10^–3 ft/s, and the calibrated hydraulic conductivity for Quaternary sediments is 5.8×10^–3 ft/s. Calibrated specific storage values range from 7.5×10^–7 ft^–1 to 1.0×10^–3 ft^–1, with the smallest values more common with increasing depth. Vertical anisotropy (the ratio of horizontal hydraulic conductivity to vertical hydraulic conductivity) ranges from 10 to 1,000.


Model fit is evaluated by looking at the magnitude and distribution of the differences between field observations of heads and fluxes and their simulated equivalents (known as residuals). Fitted error statistics indicate that simulated heads are on average within about 30 ft of field measurements. Head residuals, which should ideally be random, show some geographical clustering. This is probably an artifact caused by the lack of detailed subsurface geologic information in many areas, and by the representation of spatially variable hydraulic properties in broad, uniform zones. Heads show a slight negative bias, meaning that simulated values have a tendency to be higher rather than lower compared to measurements. This is likely an artifact of the coarse vertical discretization. Visual comparisons of time series of simulated and measured heads show that the model simulates observed climate-driven water‑level fluctuations over most of the model area. The model also simulates pumping-caused water-level changes in heavily pumped areas around the Klamath Reclamation Project. Visual comparison of time series of simulated and measured groundwater discharge to stream reaches shows that the model captures both the overall volumes and climate‑driven fluctuations of groundwater discharge to major streams. Pumping effects are generally not visually detectable in streamflow or groundwater-discharge records.


The model has the ability to simulate the effects of external stresses, such as pumping or climate variations, on the water levels and groundwater discharge to streams, lakes, drains, and other boundaries. Example model simulations show that the timing and location of the effects of groundwater pumping vary markedly depending on the pumping location. Pumping from wells close (within a few miles) to groundwater-discharge features, such as springs, drains, and certain streams, can affect those features within weeks or months of the onset of pumping, and the impacts can be essentially fully manifest in several years. However, simulations indicate that responses to seasonal variations in pumping rates are buffered by the groundwater system, and peak impacts are closer to mean annual pumping rates than to instantaneous pumping rates. In other words, pumping effects are spread out over the entire year. When pumping locations are distant (more than several miles) from discharge features, the effects take many years or decades to fully impact those features, and much of the pumped water comes from groundwater storage over a broad geographic area even after two decades. Moreover, because the effects are spread out over a broad area, the impacts to individual features are much smaller than in the case of nearby pumping. Simulations show that the discharge features most affected by pumping in the area of the Klamath Reclamation Project are agricultural drains, and impacts to other surface-water features are small in comparison. Reductions in discharge to agricultural drains could potentially have operational considerations for Reclamation Project managers; reductions could also have ramifications with regard to refuge water supplies.


Developing a groundwater management strategy in the upper Klamath Basin requires understanding the effects of a wide range of possible pumping scenarios on groundwater levels and discharge, and identifying the best pumping strategy to meet water-user needs while not resulting in unacceptable impacts. To meet this need, a groundwater management model was developed that uses techniques of constrained optimization along with the groundwater flow model to identify the optimal strategy to meet water-user needs while honoring defined constraints on impacts to groundwater levels or streams. The coupled models are referred to as groundwater simulation-optimization models.


Example groundwater simulation-optimization models were formulated to demonstrate their utility in developing strategies to meet water demand in the upper Klamath Basin. The models maximize groundwater pumping while simultaneously avoiding the detrimental impacts of pumping on groundwater levels and discharge. Total groundwater withdrawals were calculated under alternative constraints for drawdown, reductions in groundwater discharge to surface water, and for water demand to understand the potential benefits and limitations for groundwater development in the upper Klamath Basin. 


The initial application of the simulation-optimization model was made with the base-case constraint definitions that limit seasonal, year-to-year, and long-term drawdowns, limit reductions in groundwater discharge to selected streams, and limit reduction in groundwater discharge to the Klamath Project drain system. Given the example constraints and current well configuration, the optimization analysis identified approximately 56,000 acre-ft per year of groundwater that can be pumped on an annual basis in addition to the background pumping fixed at the 2000 pumping rate, with 64 percent of the total pumping occurring in the third quarter of the water year and 80 percent occurring in model layers 1 and 2. Subsequent model applications indicated that changes in the groundwater-discharge, drawdown, and water-demand constraint limits could result in substantial changes in optimal allowable groundwater withdrawal. It is important to note that the demonstration exercise does not include historic climate variability or off-project (but nearby) supplemental irrigation pumping, both of which will affect results.


The sensitivity of the optimal solution to the model constraints was tested by modifying their limits. The sensitivity of the solution to the drain-discharge constraints was tested by varying the upper bound on the allowable reduction in groundwater discharge to the drain system. Total withdrawal calculated by the optimization model ranged from approximately 33,000 acre-ft for a 10-percent constraint to approximately 77,000 acre-ft for a 40-percent limit. The sensitivity of the solution to the seasonal and year-to-year drawdown constraints’ limits was also tested. Varying the seasonal drawdown limit from 10 to 30 ft resulted in total withdrawal increasing from approximately 52,000 to approximately 57,000 acre-ft; varying the year-to-year drawdown limit from 2 to 8 ft results in total withdrawal varying from approximately 53,000 to approximately 57,000 acre-ft. Increasing the minimum amount of withdrawal in the fourth quarter also affected the optimization results. Varying the fourth-quarter water demand from about 23,000 to about 41,000 acre-ft resulted in total withdrawal decreasing from about 56,000 to about 49,000 acre-ft; when the seasonal‑demand constraint was increased to about 45,000 acre-ft, the optimization model was infeasible, indicating that volume cannot be pumped without violating one of the constraints. Finally, the optimization model was modified to test the impact of including groundwater‑discharge constraints for the upper Lost River. The limit of these constraints was adjusted from 6 to 50 percent of baseline groundwater discharge, resulting in total withdrawal increasing from about 56,000 to about 60,000 acre-ft. For all constraint types tested in the sensitivity analyses, the optimal solution varied in a nonlinear manner over the range of constraint bounds tested. 


The simulation-optimization model and its applications for the upper Klamath Basin provide an improved understanding of how the groundwater and surface-water system responds to sustained groundwater pumping within the Bureau of Reclamation’s Klamath Project. Optimization model results indicate that additional pumping within the project area could be managed to minimize impact on the groundwater discharge that supports wildlife habitat in the upper Klamath Basin. For all scenarios tested, the reduction in groundwater discharge resulting from increased pumping was less than 0.2 percent, which is well within the 6-percent limit defined in the Klamath Basin Restoration Agreement. The results of the different applications of the model demonstrate the importance of identifying constraint limits in order to better define the amount and distribution of groundwater withdrawal that is sustainable. The analyses in the demonstration case presented in this report are limited by the assumption of steady average climate conditions. It is critical to note that optimal groundwater pumping volumes and patterns will change when historic hydrologic variability and the effects of nearby off-project supplemental irrigation pumping are included in the simulation-optimization model. Because these factors are not included, the pumping volumes presented may overestimate true optimal values and are not intended to be used for management decisions.


Next steps in the application of groundwater modeling in the upper Klamath Basin could include refinement of the groundwater flow model to better simulate processes and conditions in key areas of management concern. Actual application of groundwater management models will require refinement of groundwater management objectives and constraints in consultation with water users and resource management agencies, and incorporation of realistic climate variability and background supplemental pumping.


First posted May 5, 2012