Supplementary scientific findings drawn from Principia's ongoing analysis of the Applicant's modelling of the South Park ground water system are presented below. As before, they are grouped as numbered paragraphs with clarifying sub-paragraphs wherever appropriate. The sequence in which they are presented follows those that were contained in Section 8.0 of Principia's scientific report dated February 3, 2000 and, once again, does not bear any special significance.

25. Analysis indicates that the response of the ground water system to well-pumping stimuli, continued for substantial periods of time, is predicted by the model to be essentially non-linear. The real-life response of portions of this system to such stimuli will clearly be non-linear. Non-linearity in predicted results is due to water-table conditions, both prescribed within the model and that created by continued well pumping causing initially confined model layers to become unconfined, wherein transmissivity is directly proportional to potentiometric head. Such non-linearity is significant, for, it is not mathematically correct to draw inferences regarding the stream depletion impacts predicted with the model for one set of conditions and then to draw inferences, simply by algebraic proportionality, regarding the impacts due to a different set of conditions, such as for example those caused by either doubling or halving the proposed pumping rates.
26. Analysis of time-dependent water budgets, or mass-balance checks, based upon the results predicted by the ground water model illustrates some of the complex mechanisms that influence the South Park ground water system. It also identifies the significance of obtaining valid basin-specific data, of the need for obtaining complete sets of data in significant categories and of the need to minimize making numbers of unverified assumptions. It thus indicates the need and importance: for a model to represent, faithfully, the behavior of complex ground water systems; of developing a proper modelling framework; of properly calibrating and verifying a model to relevant components of basin-specific data; of conducting systematic parameter-sensitivity analyses; and, of properly applying a model to make predictions that can be relied upon. In each of these respects, the model is flawed.
27. Analysis of the model's calibration status and especially the spatial distribution of residuals within an envelope of one mile from the chosen calibration target locations, indicates that they are too large to represent satisfactory model calibration. Comparisons of supplementary figures prepared to illustrate this feature with those based upon earlier analysis by Principia, confirms this finding. The model cannot be viewed as satisfactorily calibrated during any time span within the period considered by the Applicant.
28. Analysis indicates that the ground water inflow of approximately 1,500 acre-feet, represented as entering the ground water model domain through the scattered general-head boundary grid cells, and recharging the ground water system, does not represent an independent prediction made by the model. It is the consequence of postulations about the areas of such inflow and assumed values of the general-head, aquifer thickness and aquifer conductance prescribed in the appropriate boundary grid cells. Prescriptions of unverified values to such variables does have a quantitative influence upon the model predictions. For instance, the calculated boundary inflow rate is large enough to potentially mask the influences of the proposed project and thereby to skew the water budget for the ground water system. Measurement data are vital to substantiate the very existence of real general head boundaries and to allow the model to calculate the inflow rates with reliability. This is particularly true in view of the significant magnitude of the postulated boundary inflow rates.
29. Supplementary analysis of the model framework and predictions obtained as a consequence of it, indicates that several implausible and unverified representations have been made in the model. For instance, calculations of evapotranspiration rates involving areas represented as either containing irrigated lands and/or natural vegetation, as embedded in the model, are in error. Other instances are identified in the following paragraphs.
30. Analysis indicates that the calculational procedure used to account for the presence of more than one geological material types within a grid cell of the model, is erroneous. For instance, even when both the geologic materials identified by the Applicant by Qa or alluvium and Qb or weathered bedrock designations are present within a cell, the procedure used only allows the properties of one material type to be taken into account. Consequently, the calculated prescriptions of precipitation recharge rates assigned to model cells, which have been expressed as functions of the grid-cell proportions of Qa and Qb, are in error. The consequences of such an error, alone, cascades into errors in predicted quantities including stream depletion estimates.
31. Analysis indicates that the choice of the ground surface elevation distribution as utilized by the Applicant for purposes of modelling ground water recharge rates as well as vegetative consumptive uses, is flawed. Figure 4-20l has illustrated the difference between the ET ground surface elevation distribution assigned to model grid cells and the ground surface elevation distribution as used for making recharge calculations. The illustrated recharge ground surface values are higher than the ET ground surface values in some areas of the model domain, and lower in others, by several feet. These differences occur in a random pattern throughout the model domain. If any model grid cell contains both low lying and higher areas, it is usually the case that more ET will occur in the low lying areas and more precipitation will occur in higher areas. However, as Fig 4-20l illustrates, the opposite actually occurs with equal frequency in the model. Therefore, the conclusion to be drawn is that the Applicant has either incorrectly assigned values to the ET ground surface elevation, or to the recharge-calculation ground surface elevation, or to both.
32. Analysis indicates that values prescribed in the model for specific yield and storage coefficient distributions have been interchanged in the eastern segment of the model domain. When the confining pressure is positive in the model, such as in the vicinity of the stream channels, this interchange leads to double accounting since changes in potentiometric head causes the specific yield to be extracted from both model layers 1 and 2. The consequence is a reduction in water-level variations near those streams and thereby a reduction in potential impacts on the streams as predicted by the model. In other areas on the east side of the domain where the confining pressure is negative in the model, water-level fluctuations are overestimated since the storage coefficient for layer 2 is applied to the modelled water table. The consequence is clearly an underestimation of the change in storage for a given change in predicted potentiometric head.
33. The stage of streams, i.e. water-level elevation in a stream, is a critical parameter used in calculating stream-aquifer interactions, critical to this case since they represent gains and losses to streams under different conditions. In this calculation, stage is compared to the aquifer water level in its immediate vicinity, i.e. within the same model grid cell, to compute the direction and magnitude of water exchange between the stream and the aquifer. The prescription of stream stage values is entirely erroneous in the model. The Applicant chose to assign a fixed value of stage to all stream segments, which neither varies with time nor as a consequence of changes to stream flow rate. The MODFLOW computer program in fact provides for the calculation of stream stage, to account for just such variations, via the so-called Prudic Stream Package, instead of simply being prescribed. This feature permits the stream stage to be calculated by a well-known algebraic procedure involving the flow rate in the stream, the stream-bed elevation, the local slope of the stream, the stream width and a Manning's roughness coefficient. The Applicant chose not to invoke this feature. As a consequence, the stage in model stream segments remain at the prescribed value regardless of the flow rate in the stream. Indeed, this prescribed stage remains unaltered irrespective of whether the flow rate is 0.001 cubic feet per second (cfs) or 1,000 cfs, which is clearly implausible. Only when the stream flow rate drops precisely to 0.0 cfs, indicating that the stream has dried up, does the prescribed stage value change in the model. With the stream dry, the stream bed elevation is used in the model as the stage value to allow the stream to behave as a drain. Furthermore, Figure 4-25a depicts the prescribed stream stage value relative to the ET ground surface. In those grid cells identified by magenta and red colors, the prescribed stream stage values actually lie above the ET ground surface. Where a stream runs along a surface elevation contour line and the majority of the grid cell lies over a low lying portion, this prescription may be appropriate when accurately identified. However, since the relevant model calculation compares the potentiometric head calculated at the center of the grid cell with the prescribed stage in the stream segment in that cell, the calculation is clearly erroneous.
34. Analysis indicates that prescriptions of values to stream-bed parameters for purposes of calculating stream-aquifer interactions are also flawed. The assignments of width, length, thickness and hydraulic conductivity to each model grid cell through which stream flows are represented materially affect exchanges of flows between these streams and the underlying aquifer. In particular, the assignments of stream-bed thickness and its hydraulic conductivity which exert significant influences upon the magnitudes of stream losses and gains have not been selected based upon any measurements made within South Park. Parameter-sensitivity analysis of such assignments have not been reported by the Applicant. The choices of values so assigned, even had the prescribed stream stage been correct, has influenced the stream flows predicted by the model. Errors in prescribed stream-bed thickness or its hydraulic conductivity or both will prevent the model from predicting historical stream flows correctly. Indeed, as constructed, the model may be calibrated to the identical degree by prescribing different combinations of stream stage and stream conductance. However, when stresses are imposed upon the ground water system, such as the pumping proposed by the Applicant, significantly different quantities of stream depletion will be predicted by the model, depending upon which combination is chosen. Thus, the model has been denied the capability of predicting stream depletions correctly.
35. Analysis indicates that the prescription of a 10-foot stream-bed thickness to those model grid cells through which stream flow is represented, is improper. It has not been demonstrated with respect to measurement data that such a thickness is indeed encountered for the major streams flowing through the domain chosen for modelling, let alone whether such a thickness is uniform through all reaches of such streams. This prescription quantitatively affects the calculated values of stream-aquifer interactions. For example, halving the prescribed stream-bed thickness can double the calculated stream loss in a losing reach, all other parameters being unaltered in the calculation. Hence, calculations of stream depletions by the model are in error for this reason alone.
36. Analysis indicates that the NOCUP model run contains a significant routing error. The stream flow in Tarryall Creek from its confluence with Park Gulch is depicted in Figure 6-25z. The predicted flow rate in the NOCUP model run is one order of magnitude lower than gauged flow rates as a result of this routing error. However, when this routing error alone is corrected, it has no effect on the predicted heads and stream leakage in this NOCUP model run. This is because the stage in Tarryall Creek and the resulting leakage from it, are pre-determined by the fixed value prescribed for stream stage. A similar routing error causes flow from the North Branch Collection System ditch to be diverted to Tarryall Creek where it crosses the highway US285. This error increases the predicted stream flow in Tarryall Creek as can be observed in Figure 6-25z. This error also appears to have no effect on the potentiometric heads and stream leakage predicted by the model under either the NOCUP or SPCUP conditions, because the stream stage in these creeks were prescribed as constant.
37. Analysis indicates that the representation of proposed recharge trenches in the model is flawed. Figures 4-22a through 4-22d depict modelled streams at locations within the model domain where the Applicant has proposed to construct recharge trenches. In the model, these streams are used to predict the movements of water to, and leakages from, these recharge trenches. However, as can be observed in the figures, the flow of water from one model grid cell to the next is frequently directed uphill. Since this cannot occur in reality, the operation of the recharge trenches in reality will clearly be different from that represented in the model as occurring. Figure 6-26a and 6-26b have illustrated areas where ET occur in the model during both NOCUP and SPCUP simulations. In the areas of the recharge trenches, ET occur in the NOCUP model run but not in the SPCUP model run. The presence of surface water in these trenches is likely to lead to some ET in their vicinity. However, the ground water model as constructed and used is incapable of accounting for these losses. The amount of water available for recharge is thereby overestimated and, hence, the impacts caused by proposed project are underestimated. As a consequence, the model is incapable of predicting the actual behavior of the proposed recharge trenches.
38. It is essential to verify the uniqueness of the model calibrations to known hydrogeological conditions in the South Park ground water system. Only in this way can it be established with reliability whether, or not, the calibrated model is capable of predicting the consequences of proposed stressing of the ground water system in ways that are different from that which may have occurred historically. Even though the Applicant has claimed to have implemented two transient calibration steps with the model, analysis indicates that these are neither true transient calibration runs nor do they constitute model verifications. In the so-called Transient-1 run segment, the two decades of the 1960s and the 1970s are represented by temporally invariant conditions for aquifer stresses. Similarly the Transient-2 run segment is represented by different temporally-constant aquifer stresses throughout each year from 1980 through 1996. This is neither a steady-state calibration nor a true transient calibration, but rather a sequence of postulated quasi-steady state calibrations. Only in the NOCUP and SPCUP simulations was the model run with variations in prescribed stresses lasting time spans of less than one year. In a true transient calibration, the aquifer stresses should also be varied, in accordance with real-life observations, throughout the year. It is unprecedented to adopt stress periods for transient model calibration runs that are significantly larger than stress periods used in predictive runs involving the same time spans. As a consequence, the model was never truly calibrated to transient conditions. It was never verified that the model calibration, as achieved, is unique and reliable.
39. Analysis indicates that the representation of evaporative losses from the Applicant's proposed storage reservoir is flawed. Prescribing a constant evaporation rate combined with constant reservoir area results in calculated evaporative losses being invariant from year to year whatever the conditions of water availability for storage in the proposed reservoir are. Previously, Figure 6-28i has illustrated that during part of the model simulation run, ET actually occurs from only six of the eleven reservoir cells since model layer 1 has actually dried out. By contrast, Figure 6-42l illustrates the area covered by the reservoir in the model's stream package where infiltration will occur. During the months corresponding to the predicted result shown in Figure 6-28i, a total of 2,204 acre-feet of water was placed' in the proposed reservoir. This quantity is in addition to the 3,131 acre-feet that was also placed' in the reservoir during the previous month. Despite such large quantities supposedly being present in this reservoir, evaporation from it is represented as occurring only within the six grid cells shown in red in Figure 6-28i.
40. Since the proposed storage reservoir is represented in the model with the ground water ET package, parameter values corresponding to vegetative consumptive use have been supplied to it. The extinction depth for the area covered by the reservoir is prescribed to be 6 feet, i.e. identical to areas where vegetative consumptive use is represented in the model. Analysis indicates that when there is no water in the reservoir, but the ground water table is within six feet of the ground surface prescribed as the bottom of the reservoir, the model will calculate ET as occurring in that area. The technique for representing evaporation from the proposed storage reservoir is thus improper since vegetative consumptive use cannot normally occur in an area that is proposed to be subjected to frequent inundation.
41. Analysis indicates that the model does not faithfully represent the locations, magnitudes and timing of vegetative consumptive use or ET near streams located within the domain chosen for modelling. Figures 6-26a and 6-26b have previously also illustrated that along many of these streams, particularly Tarryall Creek north of the Sportsmen's Ranch, as well as other creeks located throughout the domain, the model does not account for ET occurring near their banks. This is counter to expectations and observations that ET does and will occur near streams. No explanations have been presented for ignoring such ET near streams which exert influences on the quantities and timing of stream-aquifer interactions. For this reason alone, the model is incapable of predicting stream depletions reliably.
42. Analysis indicates that the model does not faithfully represent the locations, magnitudes and timing of vegetative consumptive use elsewhere within the model domain. Figures 6-26a and 6-26b have also indicated that along the edge of model domain, ET is represented as occurring in a number of grid cells, especially in the southern half of the domain. Yet, Figure 4-32a, which displays ET cover assumed by the Applicant for modelling, identifies that no ET occurs in these areas. In fact, the Applicant's acknowledgment that attempts to calibrate the model were unsuccessful contained the following statement, quoted verbatim here, in the context of prescribing a non-zero ET rate in those very grid cells: "(Row + Col/3 + 460)/530 for cells with known evapotranspiration. (Row + Col/3)*1e-7 + 1.6e-5 for cells with no known evapotranspiration. The latter formula acknowledges that ET will occur from any cell with a ground water level near the surface, but for cells with no known ET, the area of ET in that cell will be relatively small. If the water level in that cell is near the surface, we have a calibration problem." Such an acknowledgment confirms that the Applicant made conscious choices and was aware of their consequences.
43. In addition to quantitative errors and improprieties in ET predicted by the model, qualitative errors in the predicted ET have also been analyzed. A model that is not qualitatively correct in representing the locations and timing of vegetative consumptive use cannot possibly be capable of quantifying ET within the chosen model domain. Since ET is an important factor in predicting near-surface ground water levels in a model, it is not only an important contributor to the ground water system, but also exerts significant influences upon stream-aquifer interactions. Without an accurate representation of ET, the model is incapable of predicting stream-aquifer interactions.
44. As a consequence of numerous instances where unverified values have been prescribed to model variables and parameters, and unsuccessful calibration attempts, the predictions made by the model are implausible. Figures 5-28, 5-29, 6-19 and 6-20 demonstrate this implausibility by illustrating the frequent occurrence of water-level mounding above the local ground surface. The artificial creation of such "lakes", especially in the southern half of the model domain, has no known real-life counterpart. In these figures, such artificial lakes indicated by dark blue and purple colors identify those locations where the extent of water table mounding has been predicted to occur in excess of 20 and 100 feet, respectively. Water level predictions by the model are thus in error over significant portions of the chosen model domain. For this reason alone, the model is incapable of predicting impacts caused by the proposed project.
45. The assigned stream flow rates in the model of the South Park ground water system prevent it from predicting impacts to some surface water bodies. For example, Link Ditch, identified as model Segment 83, is represented as diverting a constant flow of 0.02 cfs flow under the Quasi-Steady State, Transient-1 and Transient-2 model calibration runs. However, in the NOCUP and SPCUP simulation runs, the diversions of this ditch are treated as identically zero throughout the simulation run. By contrast, Table IV-4 of the Applicant's Surface Water Availability Report does list the Link Ditch as having an active irrigation water right of 19 cfs. The electronic spreadsheet file, called moddivert.wb3, as produced by the Applicant identifies the historical annual-average diversion rate to the Link Ditch as 0.828 cfs. As a consequence, the modelled stream flow rate in Tarryall Creek, after the ditch diversion point is erroneous, the leakage from and impacts to the Link Ditch predicted by the NOCUP and SPCUP model runs are identically zero, and the timing and location of corresponding recharges to the aquifer are in error. As a consequence, the model is incapable of predicting impacts in general and stream depletions in particular, with any reliability.
46. The Applicant's surface water model adopts, as noble, the stream leakage and loss values as fluxes predicted by the ground water model. A computer program named BINSUMD and written by the Applicant's experts has been used to calculate these fluxes. This program utilizes the instantaneous values of fluxes, as predicted by the ground water model at the end of each stress period, i.e. a month, and multiplies this quantity by the number of days in the month to obtain the monthly flow rate used in its further calculations. This procedure is clearly incorrect. During time spans of either rapidly rising or rapidly declining ground water levels or relatively rapid changes to aquifer stresses such as consumption of ground water by vegetation, the instantaneous flux at the end of the month cannot reliably represent the cumulative flux for the month. The consequence is that the timing, magnitude and even the sign of stream-aquifer interactions used in the surface water model are different from the corresponding values predicted by the ground water model. Such contradictions cannot both be correct.
47. In the model of the South Park ground water system, natural springs are represented by using the stream and drain packages available in the computer program MODFLOW. Some springs located within the system, such as the King Mine Springs, are represented using this program's stream package. The ground water source of this spring lies outside the chosen model domain and the flow appears as a prescribed stream flow rate in the model. As treated by the model, such stream flows can of course then seep into the ground water system thereby recharging it. Other springs such as the Link Springs, are represented using the computer program's drain package. This causes a predicted ground water withdrawal to occur from the aquifer when the potentiometric head is sufficiently high, i.e. above the assigned value of spring elevation. However, this withdrawn water is pre-determined never to re-enter the ground water system. Both these methods of representing the behavior of springs involved unsubstantiated assumptions and are flawed.
48. The representation of streams in the model was further analyzed. Such analysis showed, as illustrated in Figure 6-24c for example, that more than 60 percent of the stream cells used to represent the Garcia and Seven Mile Creeks are described as having broken connection with the aquifer already in the NOCUP model run. Substantiation of this hypothesis has not been offered by the Applicant. Thus, the model run made under SPCUP conditions will predict that no impacts to these stream cells will occur as a result of pumping from the Applicant's project, an outcome that can be viewed as pre-determined. Such a pre-determination is particularly significant since these streams are tributary to the Middle Fork of the South Platte river and, relying apparently entirely upon results predicted by the model, no plan for augmentation has been proposed for these creeks.
49. Analysis indicates that the representation of well pumping in the model of the South Park ground water system, is flawed. During the Quasi-Steady State time span, i.e. prior to 1960, no well pumping is represented as occurring. Through the Transient-1 and Transient-2 calibration time spans, well pumping is represented as increasing from 1970 through 1996. The total rate of well pumping rate assigned to the model in 1996, the starting year of the NOCUP and SPCUP model simulation runs is double the value assigned for the identical wells in the same year, at the end of the Transient-2 calibration time span. Figure 5-73 illustrates well pumping representations in non-SPCUP wells for the entire model simulation period. Since both assignments for the identical year cannot be simultaneously correct, either the calibration attempt or the NOCUP prediction run is in error, for this reason alone. It is yet another example of the Applicant adjusting assigned aquifer parameters that were used to seek calibration of the model, in order to make predictive runs with it.
50. The Applicant's Proposed Decree Exhibit Z identifies two stream gages named as "UpPark" and "TarComCo" as being suitable for gaging stream depletion impacts caused by its proposed project pumping. Further, the Applicant has proposed that flow rates measured in the future at these gages be compared to the flow rates measured at downstream gages in order to establish the magnitude of such depletions to streams. In order that such gaged records be accurate for this purpose, it is necessary that these gages be demonstrated as located truly outside the zone of influence of the proposed project pumping. Figure 6-47a identifies the locations of these gages and the potentiometric heads predicted by the model at those locations. At the UpPark location, the model, flawed as it is, does predict significant drawdowns defined as the predicted NOCUP results subtracted from the predicted SPCUP results. Clearly, the UpPark gage is not located outside the zone of influence of proposed pumping. On the other hand, the flawed model does indeed predict that the TarComCo gage location lies outside this zone of influence. This is due to the fact that the potentiometric heads predicted by the model under both NOCUP and SPCUP conditions are essentially the same. However, as Figure 6-47b demonstrates, this outcome was pre-determined by the model framework choices made by the Applicant. Figure 6-47b depicts an east-west cross section along Tarryall Creek. At the west end, ground water enters the domain through a prescribed general head boundary. The volume of water so entering the model domain is determined by the parameter values assigned by the Applicant to this general head boundary. These values remain unsubstantiated. Ground water then flows eastward along the alluvium, past the TarComCo gage, meeting the prescribed lower formation geometry approximately at 2,000 feet downstream of this gage. Explicitly excluded from the model are the contacts which formations such as the Laramie-Fox Hills make with the alluvium. In reality, impacts to such formations will, in turn, be transmitted further west. However, the model precludes such impacts from being predicted due to these formations having been explicitly excluded from the model domain. Thus, neither of these two gages can provide the information sought by the Applicant.