HomeMy WebLinkAboutReport - Groundwater Modeling - 10-29-20 Part 1
Butte | Anaconda | Great Falls | Bozeman | Kalispell · www.waterenvtech.com 1
MEMORANDUM
To: Morrison-Maierle, Inc.
From: Brad Rutherford, Michael E. Nicklin, PhD, PE
Date: October 29, 2020
Re: Groundwater flow model for managing groundwater levels at Northwest Crossing
Introduction
The focus of this report is to summarize groundwater modeling effort by Water & Environmental
Technologies, Inc. (WET) addressing shallow groundwater issues at the Northwest Crossing
Subdivision. The purpose of the modeling was to manage groundwater levels in the vicinity of
that proposed development. This development is located between West Oak Street and Baxter
Lane and west of Flanders Mille Road in Bozeman, MT (Figure 1).
Site Background Information
Presently, agricultural drains are in place at the footprint of the proposed development which
acted to control shallow groundwater for land cultivation purposes (Figure 1). The approximate
depth of those drains is about 3 feet below the ground surface. A recent drain rupture had
occurred leading to a surface discharge from 50 to 100 GPM until a repair was completed.
Given those drains are relatively shallow, and, given that there are questions about their
integrity, it was determined that it would be appropriate to construct an engineered drain system
as part of the overall development plans. Hence, a model effort was conducted to provide
information for design purposes.
A trench drain (Located at MW-1) and four new monitoring wells (MW-2, MW-3, MW-4, & MW-5)
were installed in July 2020 (Figure 1). The trench drain system consists of a vertical 24 inch
culvert with a plate welded to the bottom emplaced into the ground at a depth of 14.5 feet. 15-
foot long, 6-inch plastic corrugated pipes were placed laterally out from the culvert perpendicular
to the groundwater flow direction at a depth of 13.75 feet. A constant rate (45 GPM) pumping
test was conducted on July 21, 2020 and water levels were recorded in the new monitoring well
network.
General Hydrogeologic Setting
MEMORANDUM · Northwest Crossing Groundwater Model
Butte | Anaconda | Great Falls | Bozeman | Kalispell · www.waterenvtech.com 2
The Gallatin Valley is a topographic and structural intermontane basin bounded by folded and
faulted sedimentary, metamorphic, and igneous rocks ranging from Precambrian to Cretaceous
age. The Gallatin valley is located in the Three Forks structural basin which includes the Gallatin
Valley, the Madison Valley, and the lower Jefferson River Valley (Hackett et al. 1960).
The Gallatin valley today is a remnant of a combination of early Tertiary Laramide
compressional uplift and mid to late Tertiary Basin and Range extensional movement. The
Laramide uplift folded and faulted the crust forming the ancestral Rocky Mountains. Subsequent
extensional movement reactivated Laramide faults and pulled the ancestral Rockies apart,
leaving a series of basins separated by north to south trending mountain ranges. The Gallatin
Valley, along with the Madison and lower Jefferson Valleys of today are each a portion of one of
these large basins; the Three Forks structural basin.
The Three Forks Basin continued to gradually subside throughout mid to late Tertiary times and
includes a large volume of erosional debris from the bounding highlands and a considerable
amount of volcanically derived sediments as well. Sediments within the basin vary
compositionally from Archean gneisses and schists to Paleozoic carbonates and sandstones to
reworked early to mid-Tertiary mudstones, sandstones and conglomerates (Hackett, et al.
1960).
More details on the geology of the Gallatin Valley area can be obtained from Vuke et al. (2014)
(see Figure 3).
Near surface lithologies at the project area are predominantly Quaternary alluvial fan gravels
(Qafo) and Quaternary braid-plain alluvium. These are, in turn, underlain by reworked Tertiary
sediments.
Modelling Objectives
The primary modeling objectives were as follows:
• Create a calibrated model that provides a reasonable representation of the hydrologic
system in the vicinity of the Northwest Crossing Subdivision using both steady state
representative groundwater levels and transient pump test observations.
• Design a drain system and predict the resulting groundwater elevation following
installation of the system.
Model Development and Application
MEMORANDUM · Northwest Crossing Groundwater Model
Butte | Anaconda | Great Falls | Bozeman | Kalispell · www.waterenvtech.com 3
MODFLOW consists of a Main Program which utilizes a series of highly independent
subroutines called “modules.” These modules are grouped into packages. The packages may
be combined to simulate a variety of groundwater flow conditions. In a sense, the manner that
these packages may be pieced together is akin to “à la carte” dependent upon the situation.
Some of these packages are required while others may be used for site-specific circumstances.
The following MODFLOW packages employed in the valley-wide model included the following
(as appropriate):
• Basic Package (BAS)
• Block Centered Flow (BCF)
• Solver Package (PCG2)
• Recharge Package (RCH)
• Drain Package (DRN)
• River Package (RIV)
• Well Package (WEL)
• General-Head Boundary Package (GHB)
The boundary conditions employed in the local subdivision model were the following (see Figure
1):
• Upgradient and downgradient ends of the model were represented using GHB.
• Edges of the model domain parallel to the dominant groundwater flow direction were
represented by no-flow boundary conditions.
• Streams and ditches were represented using RIV.
• Existing and proposed drains were simulated using DRN.
• Lateral corrugated pipes at the location of MW-1 were represented using WEL for the
trench pumping test simulation.
The primary steps used to construct the groundwater model were the following:
• Groundwater Vistas graphical user interface was used to assist in defining the following
components of the model structure:
o Model domain:
▪ The model domain was rotated 10° west to align the long edge of th
rectangular domain parallel to the regional groundwater flow direction
o Model grid structure:
▪ The domain is discretized into 436 rows, 409 columns, and 2 layers.
o Model grid spacing:
▪ The grid spacing ranges from 5 feet to 80 feet. Refined discretization was
applied within the model focus area (10 feet) and within the vicinity of the
new monitoring wells (5 feet).
MEMORANDUM · Northwest Crossing Groundwater Model
Butte | Anaconda | Great Falls | Bozeman | Kalispell · www.waterenvtech.com 4
o Model Rivers and Drains
▪ Line feature representations of streams and ditches (rivers and drains)
were manually digitized in ArcGIS. The Intersect tool in ArcGIS was used
to discretize the line features within the grid-cells. For each discretized
line feature, the stage was extracted from Montana State Library Lidar
data at the midpoint of the line and the river conductance was calculated
from the linear length of each discretized river segment.
▪ Montana State Library Lidar data were used to define the top elevation of
the model domain and assign the proposed drain stages.
• Initial parameters were assigned as follows (as appropriate):
o Hydraulic conductivity values were initially assigned utilizing transmissivity/
hydraulic conductivity information from Hackett, et al. (1960), transmissivity data
obtained by consultants, and an analogous groundwater modelling effort at the
Norton Ranch.
o Recharge was defined as follows:
▪ Initial recharge distribution was assigned using Farnes Precipitation
Interpolated Contour Lines.
Model Calibration
A total of 12 targets within the Northwest Crossing Subdivision were employed for the steady-
state calibration and they are presented in Table 1 and Figure 2.
The calibration effort included the following aspects:
• Modification of the following parameters:
o Hydraulic conductivity
o Recharge
o Storativity – Specific yield
• Use of the auto-sensitivity run feature of GV which allowed multiple simulations to be run
automatically, thus accelerating the calibration process.
• Use of transient simulation to match the drawdown and recovery data recorded by
Morrison-Maierle in MW-1, MW-2, MW-3, MW-4, and MW-5 during the trench test
(Figure 3). Hydraulic conductivity and specific yield were manipulated for multiple
transient simulations to iteratively converge towards an approximate fit to the recorded
drawdown/recovery curves.
Many simulations were required to develop a calibrated model that was both statistically
satisfactory and produced parameters (recharge, hydraulic conductivity, and specific yield) that
were considered reasonably consistent with the hydrogeology and groundwater hydraulics at
the local scale.
MEMORANDUM · Northwest Crossing Groundwater Model
Butte | Anaconda | Great Falls | Bozeman | Kalispell · www.waterenvtech.com 5
Steady State Model Calibration Results
The steady-state calibration statistics are presented in Table 1. The absolute residual mean is
0.47 feet. The absolute residual mean represents the magnitude of the difference between the
measured water levels and the simulated water levels for the simulated area. The absolute
residual mean was employed during the model calibration process and the goal was to minimize
the value of this statistic until no further significant reduction could be attained.
The steady state residual mean is 0.07 feet. A residual is the difference between the observed
potentiometric head (well water level) and the computed potentiometric head. The residual
mean is the average of the residuals for all observation points. This mean represents the
tendency of the model to either over-predict or under-predict water levels. There is very limited
tendency to either over-predict or under-predict observed well water levels.
The residual standard deviation is 0.61 feet. This statistic is simply the standard deviation
representing the dispersion of model residuals computed. The amount of dispersion from the
targeted values is deemed to be low.
The sum of squares is 72.6. The sum of squares is the sum of each residual squared. In effect,
the smaller this value the better the model fit. The size of this value is also dependent upon the
number of observation points. The root-mean square error (RMS), 0.61, is computed using
square root of the sum of squares divided by the number of observations. Minimizing the value
of the RMS error was a goal during model calibration.
The minimum residual was -0.74 feet (negative implies over prediction) at EMW-3. In effect, it
states that the 2019 geometric mean water level is 0.74 feet lower than the simulated value.
Note that the simulated water level for EMW-3 is within the range of observed water levels for
2019 (Figure 2).
The maximum residual was 1.41 feet at EMW-1. This is the equivalent of stating that the target
water level is 1.41 feet higher than the simulated value. Note that the simulated water level is
outside the range of water levels recorded in 2019. EMW-1 is the only well that produced a
simulated value outside the observed water level range from 2019.
The scaled absolute residual mean is 0.004. This statistic is often used to determine if an
adequate calibration is achieved. According to James Rumbaugh, developer of Groundwater
Vistas, a typical modeling effort should obtain a value less than 0.1. Based upon this criterion,
the model provides an excellent fit to the observation data.
Plots showing the statistical fits are given in Figure 4. The simulated head vs. observed head
and residual head vs. observed head plots show the fit of the simulated values with respect to
observed head. The plots are used to show how accurately the model simulates water levels
upgradient, centrally, and downgradient within the model domain.
MEMORANDUM · Northwest Crossing Groundwater Model
Butte | Anaconda | Great Falls | Bozeman | Kalispell · www.waterenvtech.com 6
Calibrated model parameters for this simulation effort match well with the analogous effort at the
Norton Ranch.
Transient Trench Test Model Calibration Results
Transient simulation results can be seen in Figure 3. For the new monitoring well array
drawdown and recovery data were matched within 0.1 feet excluding the pumping Well (MW-1)
which had drawdown 2.6 feet greater than was simulated.
The transient simulation effort underpredicted small scale ( ~0.03 - 0.10 feet, 0.73 feet in EMW-
8) drawdown observed in relatively distant existing monitoring wells EMW-7, EMW-8, EMW-10,
and EMW-14 during the trench pumping test. The transient simulation did not simultaneously
match drawdown in the new monitoring wells and existing monitoring wells. This is deemed to
be due to the aquifer switching from confined to unconfined locally during the pumping test.
Sensitivity Analysis
Due to the local scale and lack of spatially distributed parameter data, parameter zonation was
not defined within the model domain. A single value was assigned for hydraulic conductivity and
recharge over the entire domain. Sensitivity analysis was performed on hydraulic conductivity
and recharge parameters (Table 2). The steady state model calibration was not sensitive to
changes in hydraulic conductivity or recharge.
Application Efforts
The goal of the modelling effort was to match observed head and gradient within the Northwest
Crossing Subdivision to predict the effect of adding in additional drains for the purpose of
increasing the vertical separation between groundwater and the topographic surface. Existing
drain tiles are known to exist at the site (Figure 1) to accomplish a similar objective. Elevation
data were not available for the existing drains, so elevations were assigned at ~3 feet below the
topographic surface. Two north-south existing drain laterals are assumed to exist at the site,
however inclusion of the laterals into the simulation negatively impacted the calibration. It was
assumed that the two laterals are not significantly impacting the groundwater head because
they are damaged or possibly the 3-dimensional location of the laterals was not properly
implemented in the simulation. The two north-south laterals were removed from the simulation.
Contrarily, the main stem of the existing drain at the site improved the model calibration.
Implementation of a new/deeper drain system will dewater this existing system.
Three proposed drains were aligned from south to north approximately parallel to groundwater
flow beneath planned sidewalks in areas where ground water levels were closest to the
topographic surface (Figure 5). Drain elevations were assigned based on Lidar elevation data
and assigned ~6 feet below the topographic surface apart from the western most drain near
Baxter Creek. The reach of Baxter Creek that extends through the subdivision gains
groundwater, so the western most drain was designed to keep Baxter Creek from becoming a
MEMORANDUM · Northwest Crossing Groundwater Model
Butte | Anaconda | Great Falls | Bozeman | Kalispell · www.waterenvtech.com 7
losing reach and to minimize impacts to the surrounding wetlands. Additionally, the drain system
was designed to be gravity driven with a daylight at the northwest corner of the subdivision.
Captured groundwater was designed to return to the existing channel of Baxter Creek, so the
drain elevation at the northwest corner was limited by the existing channel base. Based on the
simulation results, Baxter Ditch loses water to the groundwater under current conditions so
impacts to the ditch and surrounding wetlands should be minimal.
To test the effectiveness of the new drains, steady state simulations were completed under
conditions of average groundwater levels and high groundwater levels. For average water levels
the steady state calibrated model was used. For the high water level simulations the stages of
the streams and ditches (RIV cells) were increased by 1.5 feet, the head at the upgradient and
down gradient edges of the model (GHB cells) were increased by 1.5 feet, and the recharge
was increased from 0.00061 feet/day (2.67 inches/year) to 0.001 feet/day (4.38 inches/year).
The resulting steady state high water level can be seen in Figure 2 on the hydrographs for the
existing monitoring network (EMW wells).
Results for the average and high groundwater simulations are shown in Figures 6 & 7 Page 1
and Figures 6 & 7 Page 2, respectively. Figure 6 exhibits 1-foot groundwater contour maps of
the site before (left) and after (right) installation of the proposed drain system. Figure 7 shows
depth to groundwater before and after installation of the drains. Based on the simulation results
discharge to Baxter Creek from the drain system at average groundwater level is 1.46 CFS and
at high groundwater level is 2.38 CFS.
Proposed drains were designed to minimize simulated groundwater drawdown within wetland
areas near Baxter Creek and Baxter Ditch. Figure 8 shows that simulated groundwater
drawdown at average water level following installation of the proposed drain system are less
than 0.5 feet within wetland areas surrounding Baxter Creek. Simulated drawdown values in
delineated Wetlands bordering Baxter Ditch are less than 1.0 feet for most of the subdivision.
Small areas of delineated wetlands on the east side of Baxter Ditch show a simulated drawdown
between 1.0 and 3.0 feet. Additionally, the primary areas of simulated groundwater drawdown
are within the subdivision footprint and the majority of drawdown outside the subdivision
footprint is less than 1.0 feet. Relatively small areas outside the footprint, along the southern
and eastern boundary, exceed 1.0 feet simulated drawdown. Groundwater collected by the
proposed drain tile system will be returned to Baxter Creek, which will have little impact on the
downgradient (north) groundwater levels. Minor simulated impacts to downgradient water levels
suggests the volume collected in the proposed drain tile system enters Baxter Creek under
natural/current flow conditions.
Summary/Recommendations
This groundwater simulation effort proved to effectively match the observation data at the
location of the Northwest Crossing Subdivision. Model calibration using both transient pump test
data and steady state representative groundwater elevations proved effective to confidently
predict ground water flow parameters at the site.
MEMORANDUM · Northwest Crossing Groundwater Model
Butte | Anaconda | Great Falls | Bozeman | Kalispell · www.waterenvtech.com 8
This effort presents one drain design scenario with a gravity drainage system that discharges to
Baxter Creek at the northwest corner of the subdivision. Additional scenarios could be applied to
address tradeoffs between wetland issues and desired groundwater levels in the subdivision
lots. Additional focused dewatering could be applied within individual lots as well. The
presented scenario is believed to provide a neutral solution with equal weight on lot dewatering
and wetland preservation.
References
Hackett, O.M., F.N. Visher, R.G. McMurtrey and W.L. Steinhilber (1960). Geology and Ground-
Water Resources of the Gallatin Valley Gallatin County Montana. Geological Survey Water-
Suppply Paper 1482.
McDonald, M.G. and A.W. Harbaugh (1988 and 1996). A Modular Three-Dimensional Finite-
Difference Groundwater Flow Model. U.S. Geological Survey Open-File Report, 1988.
Vuke, S.M., J.D. Lonn, R.B. Berg and C.J. Schmidt (2014). Geologic Map of the Bozeman 30' x
60' Quadrangle Southwestern Montana. Montana Bureau of Mines and Geology Open-File
Report 648.
Well ID X (ft) Y (ft) Date Target Computed Residual
MW-1 365377.1 126021.1 2020-07-21 4725.69 4725.90 -0.22
MW-2 365428.3 126022.6 2020-07-21 4725.58 4726.02 -0.44
MW-3 365378.4 126095.8 2020-07-21 4724.46 4725.07 -0.61
MW-4 365328.1 126020.3 2020-07-21 4725.72 4725.78 -0.06
MW-5 365377.1 125965.1 2020-07-21 4726.38 4726.53 -0.15
EMW-1 363464.2 125415.1 2019 Geometric Mean 4729.94 4728.53 1.41
EMW-3 363538.7 126364.6 2020 Geometric Mean 4717.82 4718.56 -0.74
EMW-4 363561.2 126832.6 2021 Geometric Mean 4713.41 4713.42 -0.01
EMW-7 364741.7 126637.6 2022 Geometric Mean 4717.75 4717.35 0.40
EMW-8 364739.3 126185.8 2023 Geometric Mean 4722.68 4722.87 -0.19
EMW-10 365209.3 125739.0 2024 Geometric Mean 4729.35 4728.75 0.60
EMW-14 364152.0 127066.7 2025 Geometric Mean 4713.04 4712.18 0.86
Residual Mean 0.07
Absolute Residual Mean 0.47
Residual Std. Deviation 0.61
Sum of Squares 4.47
RMS Error 0.61
Min. Residual -0.74
Max. Residual 1.41
Number of Observations 12
Range in Observations 16.90
Scaled Residual Std. Deviation 0.036
Scaled Absolute Residual Mean 0.028
Scaled RMS Error 0.036
Scaled Residual Mean 0.004
Value Absolute
Residual Mean RMS Error
Hydraulic Conductivity (ft/d)10 Test Value 0.63 0.70
20 Test Value 0.54 0.64
30 Test Value 0.51 0.62
40 Test Value 0.49 0.61
50 Base Value 0.47 0.61
60 Test Value 0.46 0.61
70 Test Value 0.44 0.61
80 Test Value 0.43 0.61
90 Test Value 0.43 0.62
100 Test Value 0.43 0.63
150 Test Value 0.59 0.77
Recharge (ft/d)0.00001 Test Value 0.45 0.64
0.00005 Test Value 0.45 0.63
0.00010 Test Value 0.44 0.63
0.00061 Base Value 0.47 0.61
0.00100 Test Value 0.52 0.62
0.00500 Test Value 1.23 1.41
Parameter
Table 2
Sensitivity Analysis Results
Table 1
Steady State Targets & Calibration Results