Water Court

Weld County, State of Colorado

Court Address:  901 9th Avenue, Greeley, CO  80631


Phone Number :(970)351-7300         



















Case Number: 96 CW 14


Water Division No. 1





C.R.C.P. Rule 41(b)(1)



I.  Applicable Standard for ruling on Rule 41(1)(b) motions.

Where, as here, the court is the trier of fact, the court must weigh all the evidence, determine questions of credibility, and draw all reasonable inferences regardless of which party such inferences may favor. The standard is not whether the plaintiff established a prima facie case, but whether judgment in favor of defendant is justified on the evidence presented. Teodonno v. Bachman, 158 Colo. 1(1965); Campbell vs. Commercial Credit Plan, Inc., 670 P.2d 813 (Colo. App. 1983).


II.  Theory of the Case.

The Applicant (PCSR) proposes to create storage by withdrawing 140,000 acre-feet (AF) from the South Park Formation, a saturated aquifer consisting to two strata that are separated, for most of the length of the formation, by a relatively impermeable stratum.  The resulting cone of depletion will constitute the “storage vessel.”  The withdrawn water will be discharged into the South Platte River stream system for delivery to the City of Aurora under an existing contract between the City of Aurora and PCSR. 

PCSR proposes to store water, during times that the South Platte River system is free, by diverting excess stream flows into recharge reservoirs and ditch fields that will be located upon the upper surface of the South Park Formation.  The water thus diverted and stored will percolate into the underlying aquifer, recharging it and reducing the volume of the depletion cone that has resulted from PCSR’s pumping.  PCSR claims the amount of this recharge as the actual volume of the water it will store.

The South Park Formation is tributary to the South Platte River system.  PCSR has no water rights in the system, other than the 1996 rights that it seeks in this application; hence, its pumping from the aquifer constitutes an out-of-priority diversion that will eventually result in depletions to the stream system.  PCSR intends to augment these depletions, at times when there is no water in the surface reservoirs, by pumping additional water from the aquifer for direct discharge to the stream.  It also intends to claim, as part of its recharge to storage, irrigation run-off, precipitation and water salvaged as a consequence of the loss of the surface vegetation. This vegetation will be destroyed as the direct result of PCSR’s lowering of the water table by its pumping regime.

PCSR will determine the quantity and timing of depletions by use of the ModFlow computer model to determine aquifer characteristics that govern the timing and amount of recharge, the fluctuation in the size of the depletion cone, and timing, location and amount of depletions to the stream.

The depletion cone will exist beyond the life of the project, until it is eliminated by recharge from natural sources, fluctuating in size in tandem with fluctuating precipitation patterns.   Quantification and timing of recharge, like quantification and timing of stream depletions, depends on both the validity and reliability of PCSR’s ModFlow computer model. [1]


III.             ModFlow Model

The admission of scientific evidence and expert opinion is controlled by C.R.E. Rule 702.  People v. Schreck , __ P.2d __, (Colo. 2001); Case No. 00 SA105; 2001 Colo., April 23, 2001.  In this case, the focus of Rule 702 is upon the reliability of the Modflow model output and the expert opinion based on model output.  The relevance of such output and opinion turns upon the usefulness of such evidence to the court.

The court finds that the model, itself, is widely used to model aquifer parameters, among other uses, and that it is capable of producing reliable, relevant results.  However, the court concludes that, in order for computer modeling results to be reliable, and hence relevant, for predicting the timing and amount of both depletions and recharge, the model must be operated in a manner that is consistent with accepted modeling techniques.  If the model is operated in some other manner, there must be sufficient evidence that such other method produces valid and reliable results. 

In this case the techniques applicable to the operation of a ground-water flow model such as ModFlow are set forth in the American Society For Testing And Materials (ATSM) Standard Guide for Application of a Ground-Water Flow Model to a Site-Specific Problem (ATSM Guide D-5447)[2], and Calibrating a Ground-Water Flow Model Problem (ATSM Guide D-5981)[3].  The court’s paraphrase of the relevant techniques follows. 

The guide covers the application of a ground-water model to a particular site or problem, and refers to the application of a mathematical technique to the solution of a site-specific ground-water problem.  The guide also illustrates the major steps taken in developing a ground-water flow model to simulate an aquifer system that has been studied in the field.  The guide cautions that each ground-water model is unique and may require additional procedures in its development and application.  One of the first major steps is calibration.

Calibration.  The ATSM Guide defines calibration as the process of re­fining a model representation of the hydrogeologic frame­work to achieve a desired degree of correspondence between the model simulation and field observations.  In practice, model calibration is frequently accom­plished through trial-and-error adjustment of the mode1's input data to match field observations, and continues until the degree of correspondence between the model simulation and the observed physical characteristics is consistent with project objectives. Calibration is evaluated through analysis of residuals, i.e., the difference between the observed value and the simulated value of a given variable, and is conducted to bring the mean of the residuals close to zero, and to minimize the standard error[4] of the residuals.  Most site-specific ground-water flow models must be calibrated prior to use in predictions. In these cases, calibration is a necessary, but not sufficient, condition, that must be obtained to have confidence in the model's predictions.

If it becomes apparent during calibration that there are no realistic values for the hydraulic properties of the aquifer that will allow the model to reproduce the calibration targets, then the conceptual model of the site may need to be revisited, or the construction of the model may need to be revised.  In addition, the source and quality of the data used to establish the calibration targets may need to be reexamined. Calibration often necessitates reconstruction of por­tions of the model, resulting in changes or refinements in the conceptual model. The modeler then revisits previous steps to achieve a better representation of the physical system.

Verification.  Calibration of a ground-water model to a single set of field measurements does not guarantee a unique solution. In order to reduce the problem of nonuniqueness, the model calculations may be compared to another set of field observations that represent a different set of boundary conditions or stresses. This process is referred to in the ATSM Guide as verification.  In model verification, the calibrated model is used to simulate a different set of aquifer stresses for which field measurements have been made. The model results are then compared to the field measurements to assess the degree of correspondence. If the comparison is not favorable, addi­tional calibration or data collection is required. Successful verification of the ground-water flow model results in a higher degree of confidence in model predictions.

Sensitivity Analysis.  The ATSM is not an inflexible description of techniques for calibrating a ground-water flow model.  Other techniques may be applied as appropriate and, after due consideration, some of the techniques herein may be omitted, altered, or enhanced.  As an example, a ca1i­brated but unverified model may still be used to perform predictive simulations when coupled with a careful sensi­tivity analysis.  Sensitivity analysis is a quantitative method of deter­mining the effect of parameter variation on model results. The purpose of a sensitivity analysis is to quantify the uncertainty in the calibrated model caused by uncertainty in the modeled values of aquifer parameters.  It is a means to identify the model inputs that have the most influence on model calibration and predic­tions.  Sensitivity analysis provides users with an understanding of the level of confidence in model results and identifies data deficiencies.  Sensitivity of a model parameter is often expressed as the relative rate of change of a selected model calculation with respect to a given parameter. If a small change in the input parameter causes a significant change in the output, the model is sensitive to that parameter.

            The court finds that, as applied in this case, the model generated information that is not sufficiently reliable to support the experts’ or the court’s reliance on modeling results, because:

1.      The model was not calibrated in accordance with accepted standards.

2.      No sensitivity analysis was conducted on the model.

3.      The model produced anomalous results that PCSR’s experts were unable to explain.

4.      The residual error in water level prediction, in two successive runs on each of 12 wells, ranged between 6’ and 198’.

5.      The residual error between observed and model-predicted levels for 15 of 33 flowing springs exceeds 1 foot, and ranges between +46 feet and –26feet.  The absolute residual mean is 14 feet.

6.      PCSR’s expert and model designer, Dr. Harvey Eastman, recognized and reported the need for additional data, the need for additional model calibration, the need for explanation of the anomalous results and the need for further evaluation of the model target data, before he could defend the admissibility of the model results at trial.  PCSR declined to follow Dr. Eastman’s advice.

7.      An independent peer review was begun, but not completed.


Accordingly, the court finds that the Modflow model, as operated in this case, did not produce results of sufficient reliability to permit a reasonably accurate determination of either the timing, amount and location of stream depletion, or timing and amount of aquifer recharge.


IV.  Augmentation

            PCSR proposes to divert from tributaries of the South Platte River, in times of heavy precipitation when the river is free, and store these diversions in recharge reservoirs and ditch fields for the purpose of augmenting its out-of-priority pumping.

            As a preliminary matter, a conditional right to pump water that would injure senior appropriators may not be decreed, except in conjunction with a plan for augmentation. Fox v. Division Engineer, 810 P.2d 644, 645 (Colo. 1991);          Bohn v. Kuiper, 195 Colo. 17, 19, 575 P.2d 402, 403 (1978).  The South Platte river basin is substantially over-appropriated.  Thornton v. Bijou Irr. Co., 926 P.2d 1, 71,  n. 66; Cache La Poudre Water Users Ass'n v. Glacier View Meadows, 191 Colo. 53, 58, 550 P.2d 288, 292 (1976).  Hence, out-of-priority pumping is presumed to be injurious, and without an approved augmentation plan, PCSR’s pumping plan cannot be decreed.

In approving an augmentation plan, the court must consider the depletions from an applicant's proposed use of water, in quantity and in time; the amount and timing of augmentation water that will be provided by the applicant; and the existence of injury to vested users. The plan for augmentation must be sufficient to permit the continuation of PCSR’s pumping, when curtailment would otherwise be required to meet a valid senior call.  PCSR must provide replacement water necessary to meet the lawful requirements of a senior diverter in time and location, and to the extent that the applicant's diversion would deprive a senior of its lawful entitlement.   C.R.S. § 37-92-305(8).

With respect to predicted depletions, it has been established that 140,000 AF will be withdrawn and must be replaced, beginning at some indefinite time following the commencement of pumping, and will vary in an indefinite amount until depletions are setoff by recharge, creating an equilibrium.  As applied in this case,  Modflow does not establish any of the above variables with specificity, let alone reliability.  PCSR’s alternative plan, to establish depletion and recharge in quantity and time by way of a monitoring program, is prospective and incapable of predicting depletions until after they have occurred.  It is unrefuted that depletions to the stream will be lagged with respect to PCSR’s pumping; however, the timing of the lag has not been determined.  The court assumes that the delay between the commencement of pumping and its effect on the stream system will result in year-round losses to the stream, whether pumping is seasonal or continual.  This assumption is subject to debate, but is the least troublesome to PCSR in terms of augmenting its depletions.  In sum, the court is unable to predict injurious depletions in either location, quantity or time.  On this basis alone, PCSR has not met its burden.

Availability of water for augmentation.  PCSR proposes to augment stream depletions with water stored in the aquifer by recharge.

An essential component of an augmentation plan is the provision for adequate replacement water. Replacement water may be provided from any water source legally available for use in an augmentation plan. Williams v. Midway Ranches Property Owners Ass'n, Inc., 938 P.2d 515, 524 (Colo. 1997).


The predictions of average annual rainfall, and percentage of years when rainfall exceeds senior calls, are the subject of controversy both as to methodology and as to resulting data.

PCSR’s most recent position is that its system for the collection and storage of water, available when the river is free, is unnecessary.  PCSR argues that sufficient aquifer recharge will result from precipitation, irrigation run-off and water salvaged as a consequence of the loss of the surface vegetation from lands owned by PCSR. 

§ 37-92-103(10.5) provides that:

“Waters in underground aquifers are not in storage or stored except to the extent waters in such aquifers are placed there by other than natural means with water to which the person placing such water in the underground aquifer has a conditional or decreed right.”


These three above sources recharge the aquifer a) by natural means, and b) with water that belongs to the stream system, and for which PCSR has no decreed right.  Therefore, PCSR’s most recent position is untenable, and the viability of its augmentation plan depends upon the adequacy of its North Branch collection and reservoir recharge systems.

  PCSR’s augmentation plan thus depends upon the availability of water that it can divert in priority, which in turn depends upon stream flow dynamics of the various tributaries of the South Platte river that feed PCSR’s collection system. As a threshold requirement, PCSR must prove that water is available “in priority, in sufficient quantities, and on sufficiently frequent occasions to enable it to complete the appropriation with diligence and within a reasonable time.” In Re Water Rights Of The Bd. of County Comm’rs, Arapahoe County, 891 P.2d 952, 962 (Colo. 1995).  Here, the timing, frequency and quantity of water available from PCSR’s collection system is critical to the adequacy of the augmentation plan. 

PCSR, through its expert, Ross Bethel, claims that its collection system will yield an average of 4255 AF/year over and above total senior calling rights.  In reaching his opinion on average yield, Bethel employed a computer modeling program knows as RIBSIM (RIver Basin SImulation Model) which he developed and has applied in estimating yields in seven different river basins over a twenty-year period.  The validity and reliability of the model is not disputed.  However, like any computer model, the accuracy of the results depends upon the accuracy of the data and the validity of the assumptions used by the modeler.

            The Tarryall Creek basin, which supplies PCSR’s North Branch collection system, consists of Tarryall Creek, and its tributaries Michigan, Jefferson and Snyder Creeks, and Packer Gulch.  Smaller tributaries include French Creek, a tributary of Michigan Creek; and, Deadman Gulch and Guernsey Creeks, tributaries to Snyder Creek. Flow records exist for only Tarryall, Michigan, Jefferson and French creeks, although there are a number of gages at certain confluences.  The existing flow records are of limited duration, generally for different time periods, and present an insufficient basis in themselves to estimate average stream flow for the 47-year period utilized by Bethel and Eastman.  For Tarryall Creek, the records span a 100-month period and were recorded at its upstream-most gage at Como (Tarcomo gage).

            In order to estimate average yield, Bethel first relied on flow records for the north fork of the South Platte river (North Fork), which drains a basin adjacent to the Tarryall basin.  These records date from 1943 to the present, and are based upon records from a gage on Geneva Creek at Grant.  Bethel first compared the North Fork flow data with data from the same 100-month period for which Tarryall Creek recorded flows.  From his analysis, Bethel concluded that the flow at Tarryall Creek averages 28% of the flow in the North Fork. 

He then used existing gage records to correlate Tarryall Creek flows with other streams in basin, e.g. Michigan and Jefferson Creeks. There is some inherent error in this comparison, because there are no diversions above the Tarcomo gage, while there are diversions above the Michigan and Jefferson gages that are not regularly recorded.  Thus, no adjustment for diversions could be made to the Michigan and Jefferson flows.

Bethel then compared precipitation records to stream flows within each basin and his hypothesized precipitation zones within each basin, and derived a factor for the relationship between stream flow and precipitation in these various zones.  For precipitation data, Bethel utilized the Colorado Climate Center’s climatological map, which shows annual average precipitation during the period 1950-1980.  He then adjusted the size and location of the precipitation zones to conform to known flow data, and derived a unit runoff figure of 650 AF/mi.2.

Bethel determined the area for each stream basin from digitized USGS topographical maps, and calculated the Tarryall Creek basin to be 23.5 mi.2.  Working backwards, he found that the gaged stream flow yielded a unit runoff figure of 553 AF/mi.2, as opposed to 650 AF/mi.2 that he had derived from reported average rainfall.  He then adjusted the reported precipitation to yield model results that were consistent with the results obtained from measured runoff.  In order to estimate average stream flow in the ungaged basins of Deadman and  Park Gulches, and Guernsey and Ohler Creeks, he applied the precipitation-runoff relationship.           

Bethel first applied climatological data from adjacent basins, then adjusted this data to reflect measured flows.  For example, he used data recorded in the French Creek basin, which had been gathered only during the period 1986-90. He compared estimated flows with recorded flows for that period, and then adjusted precipitation values so that the estimated flows matched the measured flows.  He applied the adjusted rainfall figures and the unit runoff figure of 553 AF/mi.2  to the areas of the ungaged basins, and in this way generated estimates of the average annual flow in each of the ungaged basins.  He then totalled the flows from each basin to determine total average annual stream flow in the Tarryall basin.

            Based upon Bethel’s experience, and his testimony regarding the acceptability of the methods he employed, the court concludes that Bethel’s method for determining average annual stream flows produces reasonable threshold estimates.  However, Bethel did not validate his results by deriving a precipitation-runoff relationship in the North Fork basin, or deriving a unit runoff-per-acre figure for the North Fork basin, for comparison to the results he generated in the Tarryall Creek basin.  Inherent in Bethel’s assumptions is that precipitation, alone, is responsible for runoff.  This assumption ignores other variables that can affect runoff, such as length-width ratio of a basin, the mean slope of the basin, the subsurface geology and the nature of the vegetative cover.  At a minimum, the relative influence of these other variables should be assessed in order to determine the degree to which eliminating them from the model will alter the results.

            In addition, Bethel’s calculation of legally available flows did not factor out irrigation runoff.  Nor did he adjust for the changing call regime in the South Platte river basin.  For example, calls were placed for first time between 1982 and 1995 including Chatfield Reservoir storage rights, and two Cheesman storage rights: its first-fill (1993) and second-fill (1995) rights, four Denver direct flow rights, and four 1976 CWCB minimum stream flow rights on Tarryall Creek.  Hence, Bethel’s model overestimates average stream flow that is legally available for diversion by PCSR.

Bethel testified that no range of error determinations for his model results was necessary, and therefore he performed none.   As an example of  model error, average simulated flows in the Tarryall basin are more than 1100 AF higher than gaged flows.  The range in disparity between simulated and gaged flows is about 22,000 AF at its extremes. Hence the court is without sufficient information to estimate best- and worst-case scenarios for Bethel’s predictions. This case appears to be the first instance in which RIBSIM surface flow estimates have been used as input to a ground water model such as ModFlow.

            Storage.  PCSR’s plan further depends upon sufficient surface storage to contain the legally available water that is diverted and applied to aquifer recharge.  PCSR proposes to create five recharge reservoirs with a total capacity of 25,000 AF.  17,000 AF of this proposed storage is contained in Recharge Reservoir #1, which lies upon BLM lands.  This proposed reservoir would occupy a basin, the bed of which lies directly upon the upper surface of the Upper South Park formation; hence, it would be an ideal location for a recharge reservoir. 

The United States claims that, within the BLM holdings,  there are thousands of acres of wetlands, including fens; that the plant habitat includes both threatened and endangered plant species; and that there is no substantial probability that the BLM will grant PCSR access to this parcel.  The BLM was prevented from presenting evidence on these points because of its untimely Rule 26 disclosures. 

Nevertheless, ignoring the U.S.’s claims as unsupported by evidence, these recharge vessels will serve no purpose until the South Park Formation is depleted by PCSR’s pumping, and the cone of depression reaches the vessels[5]. Until then, the stored water will simply increase the natural discharge of the aquifer into the stream system. The volume of the water that will be stored in the vessels is at this time so problematic as to be speculative.  The amount of water available for recharge depends upon the relationship between the stream conditions that immediately preceded the conditions permitting recharge, and the rate of reservoir loss to the stream[6].  That is, the reservoir volume, at any given time during which the aquifer remains saturated, can be determined only if both the rate of discharge to the stream is known, and the volume diverted from PCSR’s collection system is known.  Hence, the volume of water available for recharge at the point in time when the cone of depression reaches the recharge vessels requires results generated by the ModFlow and RIBSIM models.  

PCSR assumes that, on the average, more water will be collected and placed in recharge vessels, under its 1996 priority, than will be withdrawn annually for release to the river.  The court accepts, for the limited purpose of exploring the practicality of this assumption, that 4320 AF will be annually available to PCSR in the long run, and that PCSR will annually pump only 2500 AF from the aquifer to meet its contract obligation to Aurora.  On average, then, North Branch collection system will generate about 1800 AF more water that it will withdraw from the South Park Formation. 

The average 1800 AF surplus is misleading, however.  For example, the court considers three scenarios: a) aquifer recharge is less than 2500 AF/year; b) aquifer recharge equals 2500 AF/year; and, aquifer recharge exceeds AF/year.

If the recharge rate is less than the pumping rate, then the South Park Formation depletes at an annual rate equal to the difference between the two rates.  In this scenario, the aquifer will continue to deplete over the period of pumping, and will not recover until an indefinite time after pumping ceases.

If the recharge rate equals or exceeds the pumping rate, then aquifer recharge will vary depending upon the volume of water legally available to the PCSR collection system.  Where the rate of recharge equals the rate of depletion, in those time periods that depletion caused by pumping exceeds the volume of available recharge water, depletion to the aquifer will continue and will not be replaced until after pumping ceases. In sum, the court concludes that the recharge operation is hydrologically – as opposed to legally – feasible if and only if the recharge rate provided by the PCSR recharge facilities exceeds the depletion rate caused by pumping.  The certainty that recharge to the aquifer will be be sufficient to satisfy the legal requirement of PCSR’s plan depends upon the reliability of both the Modflow and RIBSIM models.

PCSR proposes to withdraw additional water from the South Park Formation during those periods when sufficient augmentation water is not available to offset stream depletions.  The court concludes that this proposal is no more than a scheme to augment out-of-priority depletions with additional out-of-priority pumping, and under the above analysis, exacerbates depletions to the aquifer and to the South Platte river system. 


In summary, the court concludes that: 


1.      The results produced by the ModFlow groundwater model are insufficiently reliable as a basis to determine the timing, amount and location of stream depletions, or to determine the rate of aquifer recharge resulting from the PCSR recharge facilities. 


2.      The results produced by the RIBSIM surface flow model are insufficiently reliable as a basis for determining either average stream flow or legal availability of water in the Tarryall Creek basin.  Further, these results overestimate stream flows that are available to PCSR.


3.      PCSR’s revised Exhibit Z, a set of proposed terms and conditions that are wholly prospective in their application, does not meet the statutory requirement that injurious out-of-priority depletions must be quantified in time, place and location prior to the approval of an augmentation plan.


Therefore, PCSR has not met its burden of proof.


The court concludes that recharging a cone of depletion is not “storage”.  PCSR’s out-of-priority pumping will create a cone of depletion that must be refilled.  In the court’s view, replacing such depletions constitutes augmentation imposed by law, rather than storage.  Storage is the introduction of water into a vessel that was empty to begin with, for later removal by the storing party.  By analogy, borrowing water, like borrowing money, creates a deficit, not an asset.  A payment on an installment loan cannot be equated with a deposit to savings. PCSR’s scheme begins by creating a water deficit, adding to that deficit by supplemental pumping when rainfall does not offset stream depletions. PCSR gambles that sufficient rain will fall, in the long run, to repay both the deficit created by the initial and supplemental pumping, thereby restoring the stream system to its normal condition.

Finally, the beneficial uses ascribed to PCSR’s claims are interdependent and no claim can survive the failure of another.


            Accordingly, the Application is dismissed.


            Ordered by the court, June 1, 2001:




Jonathan W. Hays, Water Judge

Water Division 1          


[1] It appears that PCSR has abandoned reliance on the ModFlow results for predicting depletion and recharge and now proposes to assess them, after they occur, through a monitoring program.


[2] Exhibit P-566

[3] Exhibit P-569

[4] The standard error is a statistical measure of the variance in the expected values of an unknown parameter, e.g., the hydraulic conductivity estimates that are generated by ModFlow.

[5] For the purpose of this analysis, the court discounts the fact that PCSR cannot legally pump without a court-approved augmentation plan.

[6] The court recognizes that the timing and extent of the cone of depression will be influenced by the volume of stored recharge water at times when the South Park Formation is saturated, but has omitted this additional uncertainty for the sake of simplicity.