Study Area

1
Multivariate Statistical Analysis of Groundwater
Chemistry Arturo Woocay (awoocay_at_miners.utep.edu)
and John C. Walton (walton_at_utep.edu) Environmental
Science and Engineering The University of Texas
at El Paso, 500 W. University Ave. El Paso, Texas
79968
Abstract The multivariate statistical
methods of Principal Component Factor Analysis
(PCFA) and k-Means Cluster Analysis (KMCA) are
sequentially used on groundwater chemistry form
the Amargosa Desert region to estimate possible
hydrochemical processes and facies along with
probable groundwater flow paths and evolution in
the region. PCFA is applied to major ion data and
results are rotated, thus reducing the number of
variables describing the system and allowing for
better interpretation of the systems variation
in terms of hydrochemical processes. KMCA is
applied to factor scores derived from the rotated
PCFA to allow the grouping of sampling locations
with similar water chemistries into hydrochemical
facies. The resulting rotated factor loadings and
scores are presented as biplots, demonstrating
relationships between and among variables and
sampling-locations. Derived factor scores and
hydrochemical facies overlaid on a digital
elevation map of the region provide a visual
picture of hydrochemical evolutions, localized
recharge centers and potential groundwater
interactions with geologic and topographic
features in the region.
Factor Analysis A Principal Component Factor
Analysis (PCFA) of the data was performed Using
Statistica 7 11. The first four factors were
extracted to reduce the number of variables from
seven to four and to find relationships among the
original variables 12. Original variables are
expressed as a linear combination of the
underlying common factors, 13,14, a high
loading indicates a high degree of correlation. A
normalized varimax rotation of factors is
performed and rotated factor loadings for the
major ion chemistry are generated along with
factor scores for each sampling location. Rotated
factor loadings are presented on Table 1.
Cluster Analysis The first four rotated PCFA
scores were subjected to a k-Means Cluster
Analysis (KMCA) to group observations with a
similar genesis into separate hydrochemical
facies. The KMCA is nonhierarchical supervised
partitioning method that tries to group data that
is similar and separate data that is not based on
Euclidean distances of variables 13. From
previous analysis it was determined to group data
into seven (k) clusters. Although the seven
hydrochemical facies are derived independently of
lithological data, they are found to be in good
agreement with their respective lithology and
thus these groups are validated as hydrochemical
facies 17.
Rotated Factor Contours on DEMs Contour plots of
each of the resulting factors were overlaid on a
digital elevation model (DEM) of the region in
order to reveal ground water signatures and
potential flowpaths. A contour plot of a factor
would be equivalent to a contour plot of a
hydrochemical process indicating its direction of
evolution and delineating areas influenced by
that process 18.
Table 1 Rotated Factor Loadings for Major Ions Table 1 Rotated Factor Loadings for Major Ions Table 1 Rotated Factor Loadings for Major Ions Table 1 Rotated Factor Loadings for Major Ions Table 1 Rotated Factor Loadings for Major Ions
Parameter Factor 1 Factor 2 Factor 3 Factor 4
Mg2 Ca2 SO42- Cl- Na Alk K 0.922 0.866 0.541 0.202 -0.004 0.514 0.299 0.055 0.322 0.667 0.924 0.727 0.232 0.265 0.289 0.036 0.260 0.129 0.659 0.769 0.199 0.135 0.315 0.318 0.195 0.148 0.246 0.893
Variation Percentage 2.287 32.6 2.059 29.4 1.235 17.6 1.137 16.2
Alk Alkalinity in (CaCO3) Alk Alkalinity in (CaCO3) Alk Alkalinity in (CaCO3) Alk Alkalinity in (CaCO3) Alk Alkalinity in (CaCO3)
Biplot A biplot is simultaneous bivariate (factor
loadings and factor scores) scatter plots that
provides a visual picture of the relationships
between and among different ions and sampling
locations, in addition, it shows objective
sampling-location groupings, and thus provides
more insight than Piper diagrams 15,16. Each
factor, with a certain chemical composition,
implies a dominating hydrochemical process, and a
clustered group implies a hydrochemical facies
with similar genesis, evolution and/or
composition 17 indicated by the underlying
factors.
Figure 2 The biplots presented here have two
scales one for factor scores of sampling
locations (i.e., bottom and left), and the other
for factor loadings of ions (i.e., top and
right). Each ion vector indicates the direction
of increasing ion content in the samples, and
their projection onto the factor axis is their
correlation to that factor. This biplot is a
diagram customized to the dominant hydrochemical
processes (i.e. the factors), showing the
hydrochemical facies and demonstrating the
chemical composition of the processes and facies
of the system.
Figure 3 a) Rotated Factor 1 is dominated by Mg2
and Ca2 ions, which are typically associated
with the dissolution of carbonates, and is
interpreted as an indication of the degree of
influence of, or mixing with, the carbonate
aquifer. High values are found at Crater Flat,
Amargosa Flat and Ash Meadows, which are down
gradient of outcrops of the underlying carbonate
aquifer (Bare Mountain Specter Range, Stripped
Hills, and Skeleton Hills). b) Rotated Factor 2
is primarily composed of Cl-, Na, and SO42-
high levels of these ions are generally
associated with elevated amounts of the water
evaporation that caused their concentration, and
is perceived as a measure of the degree of
evolution through evaporation. Low values form a
trough surrounding Fortymile Wash. c) Rotated
Factor 3 is dominated by alkalinity and Na, and
is most likely related to the weathering of
silicate minerals with the generation of
alkalinity and the concomitant release of Na.
These values present a clear separation between
groundwater west and east of Yucca Mountain. d)
Rotated Factor 4 is mostly composed of K and
suggests that silicate weathering is significant
in this system. These values appear to create a
faint pathway originating in the Oasis Valley and
following the Amargosa River.
Figure 1 Static ground water elevation contours
in meters above sea level based on 1,088 wells
1, (342 wells within the area are shown)
overlaid on a satellite image of the Amargosa
Desert region. Contour intervals are reduced from
100 to 20 meters (m) between the 800 and 660 m
levels.
Introduction Study Area The Amargosa Desert is in
the southern portion of Nye County in south
central Nevada, within the Great Basin, and is
part of the Death Valley ground water basin. The
ephemeral Amargosa River begins in the Oasis
Valley, turns southeast to cross through the
length of the Amargosa Desert, then continues
south until it bends west, and finally enters
Death Valley from the southeast. Yucca Mountain,
north of the Amargosa Desert, is a group of
north-trending block-faulted ridges of volcanic
rocks (ash-flow and ash-fall tuffs) 2. Yucca
Mountain has been chosen as the site of a
high-level nuclear waste repository expected to
hold approximately 77,000 metric tons of
radioactive waste. Fortymile Wash is an ephemeral
drainage that originates in the uplands between
Timber Mountain and Shoshone Mountain, it flows
southward along the east of Yucca Mountain, and
fans out in the northern part of the Amargosa
Desert before reaching the Amargosa River.
Highway 95 fault is named so as it lies below and
approximately along the highway. Furthermore, a
deep carbonate aquifer that is locally up to 4600
meters (m) thick 3 and composed mainly of
Paleozoic limestones and dolomites 4, underlies
most of the tuff volcanic rocks and alluvial fill
5. The present climate in the Amargosa Desert
region is considered arid to semiarid, with
average annual precipitation ranging from less
than 130 millimeters (mm) at lower elevations to
more than 280 mm at higher elevations 6. In
contrast, the climate at the end of the Tioga
glacial maximum of Wisconsin glaciation in North
America, at approximately 11,500 years before
present (yr BP), was wetter and colder than the
present 7,8. Contemporary potentiometric water
levels of the region are presented on Figure
1. Data Set Ground water chemistry major ion data
were obtained from the Nye County Nuclear Waste
Repository Project Office (NWRPO) website as of
March 2003 9 and a Los Alamos National
Laboratory report 10. Data were compiled into a
single database, covering the Amargosa Desert
region. Sampling locations are mostly wells, some
of which have multiple screened depths, while the
remaining are fresh springs springs containing
high levels of evaporites were excluded from
analysis.
Conclusions Three common trends are observed from
contours of the rotated factors First, a large
trough of more dilute waters follows along the
path of Fortymile Wash and turns to the
southeast, where the wash joins the Amargosa
River. Second, the presence of noise in the
contours, apparent along the Highway 95 Fault.
Third, a gradual increase in Ca2, Mg2 and Cl-
along the pathway of the Amargosa River coming
out of the Oasis Valley. The geochemical data
presented herein suggests that groundwater
beneath Fortymile Wash follows the surface of the
wash until it appears to merge and mix with
groundwater beneath the Amargosa River. The
signature from Fortymile Wash is believed to
represent the relic of focused infiltration of
surface runoff along the course of the wash
during past pluvial periods, when the climate was
colder and wetter than the present and the amount
of runoff in the wash was significantly greater.
The results from the multivariate statistic
analyses provide objective grouping of major
ions, into hydrochemical processes and
sampling-locations, into hydrochemical facies.
Specifically, evidence is presented of
past-focused recharge around Fortymile Wash
climate-induced changes surrounding the wash, and
some potential interaction of ground water with
Highway 95 fault. The work herein demonstrates
how the use multivariate methods of statistical
analyses of water chemistry provide further
understanding of ground water flow and evolution
in the Amargosa Desert region.

• References
• Buqo, T., Personal communication by email.
  buqo_at_aol.com, with attachment Master Regional
  Water Levels 2004.xls. August 27, 2004.
• Kelkar S., P. Tseng, T. Miller, R. Pawar, A.
  Meijer, B. Robinson, G. Zyvoloski, E. Kwicklis,
  A.A. Eddebbarh, and B. Arnold. 2003. Site/subsite
  scale saturated-zone flow-transport models for
  Yucca Mountain. IHLRWM 2003, Las Vegas, NV
  114-121.
• Stetzenbach K.J., V.F. Hodge, C. Guoa, I.M.
  Farnhama, K.H. Johannessonc. 2001. Geochemical
  and statistical evidence of deep carbonate ground
  water within overlying volcanic rock
  aquifers/aquitards of southern Nevada. Journal of
  Hydrology 243 254-271.
• Flint A.L., L.E. Flint, E.M. Kwicklis, G.S.
  Bodvarsson, and J.T. Fabryka-Martin. 2001.
  Hydrology of Yucca Mountain, Nevada. Review of
  Geophysics, 39, no. 4 447-470.
• Kwicklis E.M., A. Meijer, and J.T.
  Fabryka-Martin. 2003. Geochemical inverse models
  of ground water mixing and chemical evolution in
  the Yucca Mountain area. IHLRWM 2003, Las Vegas,
  NV 328-339.
• Flint A.L., L.E. Flint, G.S. Bodvarsson, E.M.
  Kwicklis, and J.T. Fabryka-Martin. 2001.
  Evolution of the conceptual model of unsaturated
  zone hydrology at Yucca Mountain, Nevada. Journal
  of Hydrology, 247 1-30.
• Benson, L., M. Kashgarian, R. Rye, S. Lund, F.
  Paillet, J. Smoot, C. Kester, S. Mensing, D.
  Meko, and S. Lindström. 2002. Holocene
  multidecadal and multicentennial droughts
  affecting northern California and Nevada.
  Quaternary Science Reviews 21 659-682.
• Harvey, A.M., P.E. Wigand, and S.G. Wells. 1999.
  Response of alluvial fan systems to the late
  Pleistocene to Holocene climatic transitions
  contrasts between the margins of pluvial Lakes
  Lahontan and Mojave, Nevada and California, USA.
  Catena 36 255-281.

• NWRPO (Nuclear Waste Repository Project Office).
  2003. Geochemistry data files. Nye County, Nevada
  website. Accessed March 2003. http//www.nyecouty.
  com)
• LANL (Los Alamos National Laboratory), 2003.
  Regional ground water hydrochemical data in the
  Yucca Mountain are used as direct input to
  ANL-NBS-00021, Revision 01. LA0309RR831233.001,
  May, 2003.
• StatSoft, Inc. 1984-2006. SATISTICA Computer
  program manual for Windows. Tulsa, OK. Kernel
  release 5.5. www.statsoft.com
• Mellinger M., 1987. Multivariate data analysis
  its methods. Chemometrics and Intelligent
  Laboratory Systems 2 29-36. Rencher, A.C.,
  Methods of Multivariate Analysis, Second edition,
  Wiley-Interscience, USA, 2002.
• Rencher, A.C., Methods of Multivariate Analysis,
  Second edition, Wiley-Interscience, USA, 2002.
• Grande J.A., A. González, R. Beltrán, and D.
  Sánchez-Rodas. 1996. Application of factor
  analysis in the aquifer system of Ayamonte-Hueva
  (Spain). Ground Water 34, no. 1 155-161.
• Güler, C., G. Thyne, J.E. McCray, and A.K.
  Turner. 2002. Evaluation of graphical and
  multivariate statistical methods for
  classification of water chemistry data.
  Hydrogeology Journal 10 455474.
• Dalton, M.G., and S.B. Upchurch. 1978.
  Interpretation of hydrochemical facies by factor
  analysis. Ground Water 16, no. 4 228233.
• Thyne, G., C. Güler, and E. Poeter. 2004.
  Sequential analysis of hydrochemical data for
  watershed characterization. Ground Water 42, no.
  5 711723.
• Lawrence F.W., and S.B. Upchurch. 1982.
  Identification of recharge areas using
  geochemical factor analysis. Ground Water 20, no.
  6 680-687.

Acknowledgments This work was funded by the Nye
County, Nevada, Nuclear Waste Repository Project
Office through cooperative research grant
DE-FC28-02RW12163 from the U.S. Department of
Energy, Office of Civilian Radioactive Waste
Management