Hydrological Modeling Tools
Inyo/Los Angeles Cooperative Study

Cooperative Study Proposal
Approved by the Inyo/Los Angeles Standing Committee May 11, 2000

Title: Development of hydrological modeling tools

Principal investigators:
Robert Harrington, Inyo County Water Department
Saeed Jorat, Los Angeles Department of Water and Power

Purpose

The purpose of this study is to develop hydrological modeling tools that can be used to evaluate the impact of groundwater pumping, climatic variations, surface water management, and other hydrologic changes on groundwater levels. It is desirable for Inyo Co. and LADWP to jointly develop a set of groundwater modeling tools so that analyses by either party are based on methods and models that are understood and accessible by the other party. In section III.D of the Inyo/LA water agreement, it is agreed that water levels in deep and shallow aquifers will be monitored. In as much as groundwater modeling is the only available tool for internally consistent interpretation of such data, development of groundwater models is essential to fulfilling the monitoring goals of the agreement.

Background

The Green Book (Section IV) alludes to three types of quantitative hydrological analysis that are useful for determining cause and effect relationships between hydrological variables: analytical modeling, regression analysis, and numerical modeling. Analytical modeling is primarily used for determination of aquifer material parameters from aquifer tests and is the subject of another proposed cooperative study (#4, Characterization of Confining Layer Hydraulic Conductivity and Storage Properties in Owens Valley, Inyo County, California). The methods proposed for the present study are regression and numerical modeling. Both methods have been applied in the past to groundwater management problems in the Owens Valley, each having advantages and disadvantages.

Regression models are empirical in nature, and methods for assessing the uncertainty in model predictions are well established and quantitative (Haan, 1977; Holder, 1985). However, because regression models rely on statistical relationships between observed variables, they cannot be used to simulate conditions outside the range of the observations from which the model was derived. For example, a regression model should not be used to assess the effects of a drought more severe than exists in the record from which the model was developed. They are also limited spatially to the region near the well for which the model was derived, and the distance from that well that the results from the model apply is unknown. In order to acquire a spatially extensive regression modeling capability, it is necessary to have models for many wells. Nevertheless, regression models are an attractive way of developing predictive tools because they are much simpler to parameterize than physically based numerical models, and thus development is relatively quick and inexpensive. Their performance and utility depend in part on how many data are available from which to develop the model.

Numerical models are well suited to exploring hypothetical scenarios, because they are based on approximate representations of the physical principles of groundwater flow, but they are relatively time consuming and expensive to calibrate, require much information to characterize the system, and the uncertainty of the model predictions is hard to quantify. Numerical models generate a spatially complete hydrologic prediction, but it is inevitable that assumptions are made in parameterizing the model, which renders error analysis difficult. An extensive discussion of numerical groundwater modeling can be found in Anderson and Woessner (1992).

In general, at those discrete locations where sufficient data are available, regression models yield more accurate predictions and a better quantified level of uncertainty than numerical models, whereas numerical models provide a broader, regional picture of hydrologic response to a larger variety of hydrologic variables than addressed by regression models. In this study, regression and numerical models will be developed in parallel, enabling them to be used in a complementary fashion. Both regression models and numerical models have been routinely used in the Owens Valley to address water resource questions, as summarized in the following section.

Regression Models.

The earliest application of regression methods to Owens Valley groundwater problems was by Williams (1978), who used valley-wide averages of change in depth to water, pumping, and runoff to predict change in depth to water. Because most shallow test wells date from no earlier than the early 1970’s, a short period of record hampered his analysis. Regression methods were revisited in the 1980’s by Hutchison (1986a; 1986b), who explored the relation between pumping and runoff, and determined baseflow to the gaining reach of the Owens River. Later, regression models were adopted as indicators of water table response to pumping and runoff for the purpose of evaluating LADWP’s annual pumping plans (Jackson, 1992; Kavounas, 1993; Woodward-Clyde, 1995; Jackson, 1996). Most recently, focus has been on quantifying the uncertainty the regression predictions, and developing probabilistic approaches to assessing the impacts of groundwater pumping (Woodward-Clyde, 1997; Harrington, 1998). The copious data available have resulted in well-quantified confidence intervals for the model predictions, and regression models have proven a robust tool for evaluation of annual pumping plans.

Numerical Models.

The earliest use of a numerical model to evaluate Owens Valley groundwater resources was by Williams (1969), to model the Independence area. In the late 1980’s, several parallel efforts involving LADWP, ICWD, USGS, and other researchers were undertaken to develop both valley-wide and finer scale numerical models of Owens Valley groundwater flow. These efforts included models of the entire Owens Valley (Danskin, 1988; Danskin, 1998) and models of smaller portions of the Owens Valley (Blevins, 1988a; Blevins, 1988b; Blevins, 1988c; Guymon and Yen, 1988; Luhdorff and Scalmanini, 1988; LADWP, 1988; Radell and Hutchison, 1988; Radell, 1989). Groundwater models have also been developed in the Owens Valley to evaluate resources and impacts for specific projects such as the Owens Lake Soda Ash Company (Woodward-Clyde, 1993; Luhdorff and Scalmanini, 1994), water exportation from Cabin Bar Ranch (LADWP, 1993), and dust abatement on Owens Lake (Wirganowicz, 1997; Schumer, 1997). Most recent efforts have used USGS’s MODFLOW as the numerical basis for developing groundwater models (McDonald and Harbaugh, 1988). The culmination of the USGS’s efforts was a basin-wide MODFLOW application (Danskin, 1998). This model will be the starting point for further development of numerical modeling tools.

Tasks

Prepare necessary data files for running various forms of regression models. Data files will be prepared in Excel spreadsheet platform to be usable by both agencies. The Regression models should be updated whenever additional data become available. This is critical, because the period during which management has been directed by the Drought Recovery Policy contains years during which both pumping and runoff were low, and including these data results in better differentiation between the effects of pumping and recharge.

Currently, the error in LADWP’s spring runoff forecast is not included in the error analysis of the regression models. This should be included in the regression models by obtaining the necessary records from LADWP, and developing the statistics of the errors for inclusion into the regression models.

The regression models developed by Inyo County use valley-wide runoff, whereas the models used by LADWP are based on runoff by subbasin. The efficacy of using subbasin runoff should be determined by comparative validation of models using subbasin runoff versus models using valley-wide runoff.

Additionally, the regression models operate on a yearly time step. For some applications such as soil water balance modeling (see study #2, Development of a model for predicting phreatophyte water use and soil water replenishment), a shorter time step is desirable. This change requires determination of feasibility and then identification of sites where this modification is feasible, assembly of the requisite data sets, recalibrating the regression models, and evaluation of the results.

Model development – numerical models.

The numerical modeling tasks proposed for this cooperative study are the first phase of groundwater modeling effort in the Owens Valley. Scope of subsequent groundwater modeling work will depend on the results of the first phase, as described in the second task and will be proposed under a future cooperative study.

To attain mutual confidence in the USGS model, a post audit should be done prior to any use of the model for operational or management evaluation purposes. The purpose of this task is to evaluate the performance of the model and develop model input and output for the period after the calibration/validation intervals for each model. The USGS model was calibrated for steady state conditions using water year 1963 and transient conditions using water years 1963-1984. Water years 1985-1988 were then used for verification. The ensuing period 1989-1999 should be used for post auditing of the model. The post audit should consist of both a simulation of observed conditions, and an evaluation of the management alternatives presented in Danskin (1998). Actions necessary to complete the post audit are:

Installation of the USGS input data on Inyo and LADWP computers;

Identification and assembly of input data necessary for running the post audit (e.g., runoff, recharge, evapotranspiration, surface water flows, and extraction rates)

Identification and assembly of data necessary for evaluating post audit results (e.g., well hydrographs and outflows from the groundwater system)

Preprocessing of input data into a form ingestible by the model

Execution of post audit and archival of post audit products. A time series of hydrologic variables produced by the model should be archived for use in relating hydrologic changes to vegetation changes.

Evaluation of post audit results. This will consist of comparison between observed and modeled well hydrographs and outflows from the groundwater flow system.

Based on the results of the post audit, provide direction for the following actions:

Recalibration and updating of the model. The post audit should determine whether the model would benefit from recalibration.

Determination of the necessity or desirability of refining spatial discretization of the model. The spatial resolution of the model can be refined by either developing sub-models of the existing grid, or refining the entire existing grid. The post audit results should be examined for the possibility and practicality of refining the entire model grid, which would avoid the effort of development of multiple submodels.

Determination of necessary or desirable additional capabilities that should be incorporated in the model. It is likely that the Horizontal Flow Barrier package should replace the current method of treating faults as low conductivity cells. Other refinements that should be considered are an integrated surface water – groundwater water budget, an evaluation of the distribution of extraction from the two model layers, and use of MODFLOW Stream packages.

Determine the need and feasibility of refining the model temporally by the use of shorter stress periods, quarterly or monthly.

As with any modeling effort, it is highly desirable to quantify the uncertainty in model results. Following the post audit, a recommendation should be made for how to attain this capability.

Products

A functional installation of the USGS groundwater model on ICWD and LADWP systems.

A joint report documenting the modifications to the regression models to (1) define the criteria for choosing indicator wells, (2) incorporate uncertainty in the runoff forecasts into the regression models, and (3) evaluation of feasibility and possible implementation of the regression models on a shorter time step for use in the soil water balance model.

A joint report documenting a post audit of the USGS model, including documentation of the input and evaluation data, archival of the model products, and recommendations for the second phase of Owens Valley groundwater modeling efforts.

Time-table

Work will be performed in-house by Inyo and LADWP staff according to the following schedule:

All phases of this cooperative study will be undertaken jointly; however, because of the availability of certain resources in the respective offices of LA and Inyo, it is desirable that each entity take the lead in certain phases of the work. LA and Inyo will each be responsible for model installation in their respective systems; LA will take the lead on data assembly for post auditing the numerical model; Inyo will take the lead on regression model revisions; evaluation of results and preparation of reports will be undertaken jointly.

References

Anderson, M. P., and W. W. Woessner, Applied Groundwater Modeling, Academic, 1992.

Blevins, M. L., Review of Owens Valley groundwater flow models, letter to Bill Hutchison, on file at ICWD, July 12, 1988, 1988a.

Blevins, M. L., Big Pine model report, LADWP report, 1988b (?, no date).

Blevins, M. L., ISB model report, LADWP report, 1988c (?, no date).

Danskin, W. R., Preliminary evaluation of the hydrogeologic system in Owens Valley, California, USGS WRI 88-4003, 1988.

Danskin, W. R., Evaluation of the hydrologic system and selected water management alternatives in the Owens Valley, California, USGS WSP 2370-H, in press.

Guymon, G. L., and C. C. Yen, An efficient deterministic-probabilistic approach to modeling regional ground-water flow: application to Owens Valley, California, USGS OFR 88-91, 1988.

Haan, C. T., Statistical Methods in Hydrology, Iowa State Univ. Press, 1977

Harbaugh, A. W., computer program for calculating subregional water budgets using results from the U.S.Geological Survey modular three-dimensional ground-water flow model, OFR 90-392, 1990.

Harrington, R. F., Multiple regression modeling of water table response to pumping and runoff, report for Inyo County Water Dept. and Los Angeles Department of Water and Power, 1998.

Holder, R. L., Multiple Regression in Hydrology, Institute of Hydrology, 1985.

Hutchison, W. R., Linear regression analysis of Owens Valley pumping April 1972 – March 1986a, ICWD report 86-1, 1986.

Hutchison, W. R., Estimation of baseflow: Owens River at Keeler Bridge, ICWD report 86-4, 1986b.

Hutchison, W. R., and M. J. Radell, Preliminary analysis of groundwater flow in the Owens Valley with the use of finite-difference models, Inyo Co. Water Dept., Report 88-1, 1988.

Jackson, R., Water level predicting multiple linear regression models developed for the eighteen indicator shallow test wells in Owens Valley, California, ICWD report 92-3, 1992.

Jackson, R., Water level predicting multiple linear regression models developed for shallow groundwater observation wells in the Bishop well field in Owens Valley, California, Report 96-2 (draft), Inyo Co. Water Dept., 1996.

Kavounas, P., Water level predicting multiple linear regression models developed for the eighteen indicator shallow test wells in Owens Valley, California, memorandum to Randy Jackson, ICWD, June 13, 1993.

Luhdorff & Scalmanini Consulting Engineers, Analysis of groundwater flow in the Bishop Basin with the use of a finite difference model, report prepared by William R. Hutchinson, 1988.

Los Angeles Department of Water and Power, Development of a mathematical groundwater flow model of the Owens Lake Basin area, California, 1988.

Los Angeles Department of Water and Power, Revised draft environmental impact report for the Anheuser-Busch Cpmpanies LasAngeles Brewery water supply study, prepared by Montgomery Watson, state clearinghouse no. 89021513, 1993.

Luhdorff & Scalmanini Consulting Engineers, OLSAC ground-water model: recalibration and simulation of pumping impacts, prepared for Owens Lake Soda Ash Co., 1994.

McDonald, M. G., and A. W. Harbaugh, A modular three-dimensional finite-difference ground-water flow model, USGS Tech. WRI, Book 6, Chapter A1, 1988.

Radell, M. J., Three-dimensional groundwater flow model use and application – Bishop Basin, Owens Valley, California, Masters thesis, Univ. of Arizona, Tucson, AZ, 1989.

Radell, M. J., and W. R. Hutchison, Compare and contrast analysis of four groundwater flow models covering portions of the Owens Valley, ICWD report 88-2, 1988.

Schumer, R., Extension and refinement of the Owens Lake groundwater basin numerical simulation, Masters thesis, University of Nevada-Reno, 1997.

Williams, D. E., Geohydrologic study of the Owens Valley ground-water reservoir, PhD. Thesis, New Mexico Inst. of mining and Tech., 1969.

Williams, P. B., Changes in the Owens Valley shallow groundwater levels from 1970 to 1978, report prepared for the Inyo Co. Board of Supervisors, 1978.

Wirganowicz, M., Numerical simulation of the Owens Lake Groundwater Basin, California, Masters thesis, University of Nevada-Reno, 1997.

Woodward-Clyde Consultants, Hydrologic evaluation of proposed Owens Lake Soda Ash Company pumping near Cottonwood Springs, report prepared for Inyo County Water Dept., 1993.

Woodward-Clyde Consultants, Multiple Regression Model Development, Report prepared for Inyo Co. Water Dept., Ref. No. 95A130, 1995.

Woodward-Clyde Consultants, Alternative Monte Carlo analyses of well 419T – Taboose-Aberdeen wellfield, Report prepared for Inyo Co. Water Dept., 1997.

Inyo/Los Angeles Cooperative Studies