Multiphysics Modeling

Series Editors
Jochen Bundschuh
University of Southern Queensland (USQ), Toowoomba, Australia
Royal Institute of Technology (KTH), Stockholm, Sweden

Mario Csar Surez Arriaga

Department of Applied Mathematics and Earth Sciences,
Faculty of Physics and Mathematical Sciences, Michoacn University UMSNH,
Morelia, Michoacn, Mexico

ISSN: 1877-0274

Volume 5

2012 by Taylor & Francis Group, LLC

Geochemical Modeling
of Groundwater, Vadose
and Geothermal Systems

Editors
Jochen Bundschuh
University of Southern Queensland (USQ), Toowoomba, Australia
Royal Institute of Technology (KTH), Stockholm, Sweden

Michael Zilberbrand
Hydrological Service of Israel, Research Division, Jerusalem, Israel

2012 by Taylor & Francis Group, LLC

CRC Press
Taylor & Francis Group
6000 Broken Sound Parkway NW, Suite 300
Boca Raton, FL 33487-2742
2012 by Taylor & Francis Group, LLC
CRC Press is an imprint of Taylor & Francis Group, an Informa business
No claim to original U.S. Government works
Version Date: 20120117
International Standard Book Number-13: 978-0-203-12298-3 (eBook - PDF)
This book contains information obtained from authentic and highly regarded sources. Reasonable efforts have been
made to publish reliable data and information, but the author and publisher cannot assume responsibility for the validity of all materials or the consequences of their use. The authors and publishers have attempted to trace the copyright
holders of all material reproduced in this publication and apologize to copyright holders if permission to publish in this
form has not been obtained. If any copyright material has not been acknowledged please write and let us know so we may
rectify in any future reprint.
Except as permitted under U.S. Copyright Law, no part of this book may be reprinted, reproduced, transmitted, or utilized in any form by any electronic, mechanical, or other means, now known or hereafter invented, including photocopying, microfilming, and recording, or in any information storage or retrieval system, without written permission from the
publishers.
For permission to photocopy or use material electronically from this work, please access www.copyright.com (http://
www.copyright.com/) or contact the Copyright Clearance Center, Inc. (CCC), 222 Rosewood Drive, Danvers, MA 01923,
978-750-8400. CCC is a not-for-profit organization that provides licenses and registration for a variety of users. For
organizations that have been granted a photocopy license by the CCC, a separate system of payment has been arranged.
Trademark Notice: Product or corporate names may be trademarks or registered trademarks, and are used only for
identification and explanation without intent to infringe.
Visit the Taylor & Francis Web site at
http://www.taylorandfrancis.com
and the CRC Press Web site at
http://www.crcpress.com

2012 by Taylor & Francis Group, LLC

About the book series

Numerical modeling is the process of obtaining approximate solutions to complex problems

of scientific and/or engineering interest. The book series addresses novel mathematical and
numerical techniques with an interdisciplinary emphasis that cuts across all fields of science,
engineering and technology. It focuses on breakthrough research in a richly varied range
of applications in physical, chemical, biological, geoscientific, medical and other fields in
response to the explosively growing interest in numerical modeling in general and its expansion to ever more sophisticated physics. The goal of this series is to bridge the knowledge gap
among engineers, scientists, and software developers trained in a variety of disciplines and to
improve knowledge transfer among these groups involved in research, development and/or
education.
This book series offers a unique collection of worked problems in different fields of engineering and applied mathematics and science, with a welcome emphasis on coupling techniques. The book series fills a need for up-to-date information on numerical modeling. Faster
computers and newly developed or improved numerical methods such as boundary element
and meshless methods or genetic codes have made numerical modeling the most efficient
state-of-art tool for integrating scientific and technological knowledge in the description of
phenomena and processes in engineered and natural systems. In general, these challenging
problems are fundamentally coupled processes that involve dynamically evolving fluid flow,
mass transport, heat transfer, deformation of solids, and chemical and biological reactions.
This series provides an understanding of complicated coupled phenomena and processes,
its forecasting, and approaches in problem solving for a diverse group of applications, including natural resources exploration and exploitation (e.g., water resources and geothermal and
petroleum reservoirs), natural disaster risk reduction (earthquakes, volcanic eruptions, tsunamis), evaluation and mitigation of human-induced phenomena as climate change), and
optimization of engineering systems (e.g., construction design, manufacturing processes).
Jochen Bundschuh
Mario Csar Surez-Arriaga
(Series Editors)

vii
2012 by Taylor & Francis Group, LLC

Editorial board of the book series

Iouri Balachov

Jacob Bear

Angelika Bunse-Gerstner

Chun-Jung Chen

Alexander H.D. Cheng

Martn A. Daz Viera

Hans J. Diersch

Jesus A. Dominguez
Donald Estep

Ed Fontes
Edward Furlani

Pierre Glynn

Ismael Herrera

Advanced Power Generation, Physical Sciences Division,

SRI International, Menlo Park, CA 94025, USA
E-mail: [email protected]
Dept. of Civil and Environmental Eng., Technion,
Israel Inst. of Technology, Haifa 32000, Israel
E-mail: [email protected]
Center of Industrial Mathematics, Faculty of
Mathematics and Computer Science, University of Bremen,
Bremen, Germany
Email: [email protected]
Chen Life Science Group, Research Division,
National Synchrotron Radiation Research Center,
and Department of Physics,National Tsing
Hua University, Hsinchu 30076, Taiwan
Email: [email protected]
Department of Civil Engineering, University of
Mississippi, MS 38677-1848
E-mail: [email protected]
Instituto Mexicano del Petrleo (IMP), Mexico City, Mexico
E-mail: [email protected]
Groundwater Modelling Centre, DHI-WASY GmbH,
12526 Berlin, Germany
E-mail: [email protected]
ASRC Aerospace Corporation, Kennedy Space Center, FL, USA
E-mail: [email protected]
Department of Mathematics, Department of Statistics,
Program for Interdisciplinary Mathematics,
Ecology, & Statistics; Director, Center for
Interdisciplinary Mathematics and Statistics,
Colorado State University, Fort Collins, CO 80523, USA
E-mail: [email protected]
COMSOL, SE-111 40, Stockholm, Sweden
E-mail: [email protected]
Device Physics and Simulation, OCTO,
Device Science & Technology Research,
Fluidics Dept., Eastman Kodak Company,
NY 14650-2216, USA
E-mail: [email protected]
National Research Program/Eastern Branch, U.S. Geological
Survey, 432 National Center, Reston, VA 20192, USA
Email: [email protected]
Institute of Geophysics, National University
of Mexico (UNAM), 14000, Mexico D.F., Mexico
E-mail: [email protected]
ix

2012 by Taylor & Francis Group, LLC

Editorial board of the book series

Rafid al Khoury

Jim Knox

William Layton

Kewen Li

Jen-Fin Lin

Rainald Lhner

Emily Nelson

Enrico Nobile

Jennifer Ryan

Rosalind Sadleir

Fernando Samaniego V.

Peter Schtzl

Xinpu Shen

Roger Thunvik

Clifford I. Voss
Thomas Westermann

Michael Zilberbrand

2012 by Taylor & Francis Group, LLC

Computational Mechanics, Faculty of Civil Engineering

and Geosciences, Delft University of Technology,
2628 CN Delft, The Netherlands
E-mail: [email protected]
Life Support Systems Development Team,
NASA Marshall Space Flight Center, Huntsville,
AL 35812, USA
E-mail: [email protected]
Department of Mathematics, University of Pittsburgh,
Pittsburgh, PA 15260, USA
E-mail: [email protected]
Stanford University, Department of Energy
Resources Engineering, Stanford, CA 94305-2220, USA
E-mail: [email protected]
Center for Micro/Nano Science and Technology,
National Cheng Kung University, Tainan, Taiwan
E-mail: [email protected]
School of Computational Sciences,
George Mason University, MS 6A2, USA
E-mail: [email protected]
Bio Science and Technology Branch, NASA Glenn
Research Center, Cleveland, OH 44135, USA
E-mail: [email protected]
Department of Naval Architecture,
Ocean and Environmental Engineering (DINMA),
University of Trieste, Trieste, Italy
E-mail: [email protected]
Delft Institute of Applied Mathematics, Delft University
of Technology, 2628 CD Delft, The Netherlands
E-mail: [email protected]
Department of Biomedical Engineering,
University of Florida, Gainesville, FL 32611-6131, USA
E-mail: [email protected]
Faculty of Engineering, Institute of Geophysics, National
University of Mexico (UNAM), 14000, Mexico City, Mexico
E-mail: [email protected]
Groundwater Modelling Centre, DHI-WASY GmbH,
12526 Berlin, Germany
E-mail: [email protected]
Landmark Graphics Corporation, Houston,
TX 77042-3021, USA
E-mail: [email protected]
Dept. Land &Water Resources Engineering, Royal Institute
of Technology (KTH), SE-100 44 Stockholm, Sweden
E-mail: [email protected]
U.S. Geological Survey, Reston, VA 20192, USA
E-mail: [email protected]
Karlsruhe University of Applied Sciences,
76133 Karlsruhe, Germany
E-mail: [email protected]
Hydrological Service of Israel, Jerusalem 91360, Israel
E-mail: [email protected]

Table of contents

About the book series

vii

Editorial board of the book series

ix

Contributors

xvii

Foreword

xix

Editors preface

xxi

About the editors

xxv

Acknowledgements

xxvii

Section 1: Introduction to groundwater geochemistry and fundamentals

of hydrogeochemical modeling
1

Hydrogeochemistry principles for geochemical modeling

(J. Bundschuh & O. Sracek)
1.1 Sampling and analysis of water, solids and gases
1.1.1 Measurement of field parameters
1.1.2 Filtration and preservation of water samples
1.1.3 Sampling of solid materials
1.1.4 Sampling of gases
1.2 Introduction to thermodynamics
1.3 Chemical composition of precipitation
1.4 Hydrochemical processes
1.4.1 Introduction
1.4.2 Oxidation-reduction reactions
1.4.3 Organic matter decomposition, photosynthesis and aerobic respiration
1.4.4 Nitrification and denitrification
1.4.5 Sorption
1.5 Kinetics

3
3
5
7
8
9
10
15
16
16
16
17
17
18
22

Thermodynamics of gas and mineral solubility in the unsaturated-zone water

(L. Mercury & M. Zilberbrand)
2.1 Introduction
2.2 Background
2.2.1 Capillary water
2.2.2 Capillarizing the water by the dryness of the soil atmosphere
2.2.3 Capillarity and size of pores
2.2.4 Capillary water: stable or metastable?
2.3 Capillary thermodynamics
2.3.1 Capillary solutions and the gas-solutions equilibria
2.3.2 Solids in capillary situations

27
27
27
27
30
31
32
33
33
34

xi
2012 by Taylor & Francis Group, LLC

xii

Table of contents
2.3.3 Thermodynamic modeling of reactions in capillary systems
2.3.4 Simplified modeling of salt solubility in capillary systems
Illustrations in natural settings
2.4.1 Capillarity and mineralogy of desert roses
2.4.2 Capillarity and the dissolution of gases
Hydrogeochemical modeling in the unsaturated zone
Conclusions

34
35
36
36
38
39
40

Governing equations and solution algorithms for geochemical modeling

(C. Ayora, M.W. Saaltink & J. Carrera)
3.1 The formulation of reactions
3.1.1 Species, reactions and stoichiometric coefficients
3.1.2 Equilibrium reactions in terms of the stoichiometric matrix
3.1.3 Primary and secondary species
3.1.4 Components and component matrix
3.1.4.1 Method 1 (aqueous components)
3.1.4.2 Method 2 (eliminate constant activity species)
3.1.4.3 Other methods
3.2 Homogeneous reactions
3.2.1 Speciation calculations
3.2.1.1 Algorithm 1
3.2.1.2 Algorithm 2
3.3 Heterogeneous reactions
3.3.1 Surface complexation reactions
3.3.2 Cation exchange reactions
3.3.3 Reactions with a solid phase
3.3.4 Reactions with a gas phase
3.4 Reaction paths
3.5 Formulation of kinetic reactions

45
45
45
47
49
52
53
57
57
58
59
60
61
63
63
68
71
71
73
76

Fluid flow, solute and heat transport equations

(M.W. Saaltink, A. Yakirevich, J. Carrera & C. Ayora)
4.1 Introduction
4.2 Groundwater flow equations
4.2.1 Single phase flow
4.2.1.1 The conservation mass for the fluid
4.2.1.2 The momentum mass balance equations for the fluid
4.2.1.3 Flow equations
4.2.2 Multiphase flow
4.2.2.1 Multiphase system
4.3 Transport of conservative solutes
4.3.1 Advection, diffusion and dispersion
4.3.1.1 Advection
4.3.1.2 Diffusion
4.3.1.3 Dispersion
4.3.2 Transport equations of conservative solutes
4.4 Heat transport equations
4.4.1 Conduction and convection
4.4.1.1 Heat conduction
4.4.1.2 Heat convection
4.4.2 Heat transport in single fluid phase systems
4.4.3 Heat transport in multiple fluid phases systems

83
83
83
84
84
84
87
90
90
92
92
92
93
94
96
97
97
97
98
98
99

2.4

2.5
2.6
3

2012 by Taylor & Francis Group, LLC

Table of contents xiii

4.5

4.6

Reactive transport
4.5.1 The need for reactive transport: calcite dissolution in the fresh-salt
water mixing zone
4.5.2 Mass balance equations
4.5.3 Constant activity species
4.5.4 Analytical solution for a binary system: equilibrium reaction rates
4.5.4.1 Problem statement
4.5.4.2 Methodology of solution
4.5.4.3 An analytical solution: pulse injection in a binary system
The effect of heterogeneity and non-local formulations
4.6.1 The limitations of traditional formulations and the need
for upscaling
4.6.2 Solution of reactive transport in MRMT formulations

99
99
102
106
108
108
109
112
115
116
119

Numerical solutions of reactive transport equations

(M.W. Saaltink, J. Carrera & C. Ayora)
5.1 Introduction
5.2 Methods for discretizing space and time
5.2.1 Finite differences
5.2.1.1 Fundamentals
5.2.1.2 Application to conservative transport
5.2.2 Finite elements
5.2.3 Instability and numerical dispersion
5.3 Methods for solving reactive transport equations
5.3.1 Sequential Iteration Approach (SIA)
5.3.2 Direct Substitution Approach (DSA)
5.3.3 Comparison between SIA and DSA

127
127
127
127
127
129
131
134
135
136
138
140

Elaboration of a geochemical model

(M. Zilberbrand)
6.1 Introduction
6.2 Model types and the most popular existing software packages
6.2.1 Speciation-solubility models
6.2.2 Reaction-path models
6.2.3 Inverse (mass-balance) models
6.2.4 Reactive transport models
6.3 Data required for geochemical modeling
6.3.1 Data for speciation-solubility models
6.3.2 Data for reaction-path models
6.3.3 Data for inverse (mass-balance) models
6.3.4 Data for reactive transport models
6.4 Schematization and choice of thermodynamic database
6.5 Modeling and interpretation of its results
6.6 Possible errors and misconceptions in model elaboration

143
143
143
143
145
145
145
145
145
147
147
147
147
149
150

Advances in geochemical modeling for geothermal applications

(P. Birkle)
7.1 Introduction
7.2 Development of geothermal reservoir tools
7.3 Types of geochemical models for geothermal systems
7.4 Requirements for geochemical simulations of geothermal reservoirs
7.5 Popular computer software for geothermal system modeling

153
153
153
155
156
156

2012 by Taylor & Francis Group, LLC

xiv Table of contents

7.6
7.7

7.8

Flow and geochemical model calibration

Selection of recent applications (20002010)Case studies
7.7.1 General applications
7.7.2 Conceptual reservoir models
7.7.3 Lumped parameter models
7.7.4 Advanced numerical modeling
7.7.4.1 Reservoir design and magnitudeReconstruction
of reservoir parameters
7.7.4.2 Origin of acidity for reservoir fluids
7.7.4.3 Mineral-fluid equilibria
7.7.4.4 Fluid reinjectionScaling effects
7.7.4.5 Hot-Dry Rock (HDR) systems
(Soultz-sous-Forts, France)
7.7.4.6 CO2 injection into geothermal reservoirs
ConclusionsFuture challenges

159
160
160
160
164
165
165
165
165
165
168
169
170

Section 2: Cases studies

8

Integrating field observations and inverse and forward modeling: application

at a site with acidic, heavy-metal-contaminated groundwater
(P. Glynn & J. Brown)
8.1 Introduction
8.2 Geochemical modeling: computer codes, theory and assumptions
8.2.1 Inverse geochemical modeling
8.2.1.1 Principles, codes and theory
8.2.1.2 Assumptions used in inverse modeling
8.2.2 Forward geochemical modeling
8.2.2.1 Principles and codes
8.3 The Pinal Creek basin site: brief description
8.3.1 Geology
8.3.2 Hydrology and groundwater flow
8.4 Inverse geochemical modeling at the Pinal Creek site
8.4.1 Examination of end-member waters and
their conservative constituents
8.4.2 The thermodynamic state of the end-member waters
8.4.3 NETPATH inverse modeling: simulation results
8.4.4 Inverse geochemical modeling with PHREEQC
8.5 Reactive-transport modeling at the Pinal Creek site
8.5.1 Summary of previous reactive-transport modeling
8.5.2 A reactive-transport sensitivity analysis on the movement
of pH and pe-controlling mineral fronts
8.5.2.1 A simple model for advective transport of a reactive
front: the MnO2 dissolution front
8.5.2.2 Determination of the initial MnO2,s and carbonate
mineral concentrations
8.5.2.3 Setup of the 1-D reactive-transport simulations
8.5.2.4 Simulation results: movement of the Fe(II)-rich waters
and of the MnO2 dissolution front
8.5.2.5 Simulation results: evolution of the low-pH waters
8.5.2.6 The effect of the initial carbonate to initial MnO2 ratio
on the evolution of the low-pH waters

2012 by Taylor & Francis Group, LLC

181
181
182
182
182
183
186
186
188
189
190
190
191
192
194
200
203
205
206
206
207
209
211
212
213

Table of contents

8.6
8.7
9

10

11

12

8.5.2.7 Influence of the aluminum mineral allowed to precipitate

on the evolution of the low-pH waters
8.5.2.8 Effects of the irreversible dissolution of Ca and Mg
silicates on the evolution of low-pH Fe(II)-rich waters
8.5.2.9 The effect of not allowing rhodochrosite precipitation
8.5.2.10 The CO2 open system simulations
8.5.2.11 The effect of longitudinal dispersion
8.5.2.12 The influence of cation exchange and surface-complexation
sorption processes
8.5.2.13 Other minor effects on the evolution of the low-pH waters
8.5.2.14 Comparison of the reactive transport simulation results
with observations at the Pinal Creek site
Conclusions
The Senior Authors fifteen year perspective on the Glynn
and Brown (1996) paper

xv

215
217
218
218
219
220
221
221
224
226

Models and measurements of porosity and permeability evolution

in a sandstone formation
(S. Emmanuel, J.J. Ague & O. Walderhaug)
9.1 Introduction
9.2 Porosity measurements in mineralized rock
9.3 Theory and numerical modeling of porosity evolution
9.3.1 Conceptual model of the porous medium
9.3.2 Reaction kinetics
9.3.3 Reactive transport equations
9.3.4 Numerical solution and model optimization
9.4 Comparison between numerical models and measurements
9.5 Implications for bulk reaction rates
9.6 Implications for permeability evolution in aquifers
9.7 Concluding remarks

235
235
236
238
238
240
243
244
245
247
248
249

Geochemical modeling of water chemistry evolution in the Guarani

Aquifer System in So Paulo, Brazil
(O. Sracek & R. Hirata)

253

Modeling of reactive transport at a site contaminated by petroleum

hydrocarbons at Hnevice, Czech Republic
(O. Sracek & Z. Vencelides)
11.1 Site characterization and conceptual model
11.2 Speciation and inverse geochemical modeling
11.3 Modeling of reactive transport

259
259
261
263

Numerical modeling for preliminary assessment of natural remediation of phosphorus

in variably saturated soil in a peri-urban settlement in Kampala, Uganda
(R.N. Kulabako, R. Thunvik, M. Nalubega & L.A. Soutter)
12.1 Introduction
12.2 Setting
12.3 Numerical model
12.3.1 Flow model
12.3.2 Solute model
12.3.2.1 Soil phosphorus sorption
12.3.2.2 Solute transport model

267
267
267
269
269
274
274
275

2012 by Taylor & Francis Group, LLC

xvi

Table of contents
12.4 Simulations
12.5 Results and discussion
12.5.1 Field measurements
12.5.2 Pollution and remediation simulation scenarios
12.5.3 Sensitivity analyses
12.5.3.1 Impact of change of sorption coefficients
(KL and Kplin) on pollution time
12.5.3.2 Impact of change of the pore size distribution values
on pollution time
12.5.3.3 Impact of change of the air entry values on pollution time
12.6 Conclusions

276
277
277
278
279
279
279
281
281

Subject index

287

Book series page

305

2012 by Taylor & Francis Group, LLC

Contributors

Jay J. Ague
(Chapter 9)

Department of Geology and Geophysics, Yale

University, New Haven, Connecticut, USA and Yale
Peabody Museum of Natural History, Yale University, New Haven, CT 06511, USA

Carlos Ayora
(Chapters 3, 4 and 5)

Department of Geosciences, Institute of Environmental Assessment and Water Research (IDA-CSIC),

c/Jordi Girona, 1826, 08034 Barcelona, Spain

Peter Birkle
(Chapter 7)

Instituto de Investigaciones Elctricas (IIE), Gerencia de Geotermia, Reforma 113, Col. Palmira,
Cuernavaca, Morelos, 62490 Mexico

Jochen Bundschuh
(Chapter 1)

University of Southern Queensland, Toowoomba,

Queensland 4350, Australia & Department of Land
and Water Resources Engineering, Royal Institute of
Technology, SE-10044 Stockholm, Sweden

James G. Brown
(Chapter 8)

Formerly with the US Geological Survey in Tucson,

Arizona

Jesus Carrera Ramirez

(Chapters 3, 4 and 5)

Institute of Environmental Assessment and Water

Research (IDAEA), Spanish Council for Scientific
Research, Jordi Girona 18, 08034 Barcelona, Spain

Simon Emmanuel
(Chapter 9)

Institute of Earth Sciences, The Hebrew University

of Jerusalem, Givat Ram, Jerusalem, 91904, Israel

Pierre Glynn
(Chapter 8)

National Research Program/Eastern Branch, U.S.

Geological Survey, 432 National Center, Reston, VA
20192, USA

Ricardo Hirata
(Chapter 10)

Department of Sedimentary and Environmental

Geology, Institute of Geosciences, University of So
Paulo, Rua do Lago 562, Cidade Universitria, CEP
05508-900, So Paulo, SP, Brazil

Robinah Nakawunde Kulabako

(Chapter 12)

Department of Civil and Environmental Engineering, Makerere University, P.O.Box 7062 Kampala,
Uganda

Lionel Mercury
(Chapter 2)

Institut des Sciences de la Terre dOrlans, UMR 6113

CNRS/Universit dOrlans, 1 A rue 8 de la Frolerie,
45071 Orlans Cedex, France

xvii
2012 by Taylor & Francis Group, LLC

xviii Contributors
Maimuna Nalubega
(Chapter 12)

Department of Civil and Environmental Engineering, Makerere University, P.O. Box 7062 Kampala,
Uganda

Maarten W. Saaltink
(Chapters 3, 4 and 5)

GHS, Department of Geotechnical Engineering and

Geosciences, Universitat Politecnica de Catalunya,
UPC-BarcelonaTech, Jordi Girona 13, 08034
Barcelona, Spain

Ondra Sracek
(Chapters 1, 10 and 11)

Department of Geology, Faculty of Science, Palack

University, 17. listopadu 12, 771 46 Olomouc, Czech
Republic & OPV s.r.o. (Protection of Groundwater
Ltd), Blohorsk 31, 169 00 Praha 6, Czech Republic

Leigh A. Soutter
(Chapter 12)

Physics Logic LLC, 6333E Mockingbird Lane, Suite

147PMB 722, Dallas, TX 75214, USA

Roger Thunvik
(Chapter 12)

Department of Land and Water Resources Engineering, Royal Institute of Technology, SE-10044 Stockholm, Sweden

Zbynek Vencelides
(Chapter 11)

OPV, Blohorsk 31, 169 00 Praha 6, Czech

Republic

Olav Walderhaug
(Chapter 9)

Statoil ASA, N-4035 Stavanger, Norway

Alexander Yakirevich
(Chapter 4)

Zuckerberg Institute for Water Research (ZIWR),

The Jacob Blaustein Institute for Desert Research,
Ben-Gurion University of the Negev, Sede-Boqer
Campus, 84990, Israel

Michael Zilberbrand
(Chapters 2 and 6)

Hydrological Service of Israel, Yaffo Str. 234,

Jerusalem 91063, Israel

2012 by Taylor & Francis Group, LLC

Foreword

The world beneath our feet often remains unseen, and yet it provides us with water that feeds
our streams, our ecosystems, and our water wells. Its shallowest parts, the soil and vadose
zones, breathe gases in and out, help regulate moisture, fix nutrients, breakdown minerals,
decompose and recycle organic debris and other waste, and sustain a rich microbial ecosystem
that is vital to the sun-fed plants living on the surface. A bit deeper, in the saturated zone, permeable environments (aquifers) provide us with naturally filtered drinking water, and water
for many other uses (e.g., irrigation). Deeper yet, permeability and storativity decrease, fluid
residence times increase, dissolved salts increase, and selected groundwater environments provide an opportunity to store wastes (e.g., brines, hazardous chemicals, nuclear waste). Geochemical processes affect water quality as well as the mineral fabric and geologic framework
of the subsurface, including its permeability and capacity to store water.
The subsurface also provides mineral and energy resources that are essential to human
society. Water is often essential in the formation and development of those resources (e.g.,
hydrothermal mineral deposits, geothermal energy). Groundwater is also often affected by
the presence of mineral or energy resources (e.g., oil, coal) or by their extraction (e.g., acid
mine drainage). Elevated levels of toxic chemicals can occur either naturally (e.g., arsenic in
shallow Bengal basin groundwaters) or as the result of human action (e.g., gasoline and oil
spills). Geochemical and biogeochemical reactions drive the transformation, mobility, and
often the relative toxicity of constituents of concern in groundwater environments.
Because of their large relevant volumes, relative stability, and distribution of residence
times, groundwater environments also serve as an archive of climate and land-use change,
including human-driven change. Dissolved gases, i.e., noble and nitrogen gas concentrations,
in groundwaters around the world have recorded a shift of generally at least 4 to 5C in
recharge temperatures as the world became warmer after the last glacial maximum about
20,000 years ago. Secondary calcite precipitated in aquifers, such as found in Devils Hole (NV,
USA), has recorded the 18O (i.e., temperature) and 13C (i.e., vegetation distribution) history
of recharging water over the last 500,000 years. On shorter timescales, shallow groundwaters
have also recorded the history of land-use change (e.g., increased fertilizers and pesticides)
as well as the diversity of anthropogenic constituents (e.g., chlorofluorocarbon compounds,
methyl tert-butyl ether) introduced in the environment over the last 50 years.
Groundwater provides a ubiquitous control on water availability and water quality in ecosystems, often in more ways than the lay public realizes. The relatively stable temperatures of
groundwaters and their often dominant contribution to streamflow helps moderate temperature and flow variability and helps stabilize aquatic habitats. The lagged response of groundwater systems means that human actions that affect surface conditions have a delayed effect
on the water quality and availability of groundwater, and therefore on the surface waters
and environment that it feeds. Conversely, the stability, lagged response and heterogeneity of
groundwater systems make them difficult and time consuming to remediate, once contaminated or otherwise altered.
Pervasive heterogeneity, from the scale of pores and mineral grains to the regional scale of
geologic formations and facies, is a key characteristic of groundwater environments. In turn,
characterizing and modeling that physical, geochemical, and microbiological heterogeneity,
and its effects on the transport and transformation of groundwater solutes, are major challenges for hydrogeologists and geochemists. The challenges are great enough to warrant the
xix
2012 by Taylor & Francis Group, LLC

xx

Foreword

application of a wide variety of geophysical, geochemical, and numerical modeling tools and
tracers (e.g., heat and solutes).
So, why do we need geochemical and biogeochemical modeling of groundwater systems?
A first answer is that geochemical and microbial processes are active processes that help
trace the origin, flowpaths, residence times, and evolution of waters in the subsurface; and
these processes also often affect the permeability and physical properties of geologic materials controlling flow and transport. Given the complexity of groundwater systems, their
heterogeneity and inaccessibility, the path to enhanced understanding of subsurface systems
lies in making use of all information and simulation tools availablenot just physical flow
and conservative transport models. Modeling, in all its forms, helps organize and provide a
framework for the information that is available, helps recognize information that is missing
and needed to answer specific questions, helps explore (and sometimes forecast) a diversity of
future scenarios, and generally helps test hypotheses and gain understanding.
A second answer is that geochemical and biogeochemical processes transform not only
geologic materials and their interfaces with subsurface fluids, but also affect the nature,
transport, and fate of solutes, colloids, and microbes in the subsurface. Modeling and
understanding these processes and their temperature dependence can potentially result in
better management and regulatory decisions regarding (i) human activities on the land surface, (ii) appropriate extraction and use of subsurface resources (water, energy, minerals),
(iii) suitable practices for waste disposal, and (iv) cost-effective contaminant remediation.
Additionally, modeling and understanding these processes can help explore and document
the archive of climate, environmental, and land-surface changes recorded in groundwater
systemsfrom the distant past when human activities were not a major influence, to the
present when they often are.
The twelve chapters in this handbook provide an excellent introduction and a highly useful
reference on the methods and challenges of groundwater modeling, with an emphasis on the
modeling of geochemical and biogeochemical processes. The principles of aqueous chemistry and the basic theories describing chemical reactions and the physics of fluid, solute, and
heat transport are discussed, together with the algorithms used in numerical simulations of
groundwater geochemical modeling. Practical issues, such as useful sampling methods and
analytical characterization techniques, are covered. A broad array of applications and case
studies of geochemical modeling are also discussed, including geochemical evolution in the
Guarani regional aquifer (in Brazil), porosity development, geothermal applications, and a
diversity of modeling approaches and studies relating to reactive transport at sites contaminated by petroleum hydrocarbons, excess phosphorus, and acidic heavy-metal-contaminated
waters.
The diversity of relevant topics, and the basic and advanced methods and approaches to
geochemical and biogeochemical modeling covered in this handbook, make it a useful reference for practicing hydrogeologists and geochemists around the world. The importance of
discovering, studying, utilizing, preserving, and/or remediating the world beneath our feet
has never been greater for societys welfare.
Pierre Glynn
Chief, National Research Program/Eastern Branch
U.S. Geological Survey, Reston, VA

2012 by Taylor & Francis Group, LLC

Editors preface

The grand aim of all science is to cover the greatest number of empirical facts by
logical deduction from the smallest number of hypotheses or axioms.
Albert Einstein
The purpose of science is not to analyze or describe but to make useful models of the
world. A model is useful if it allows us to get use out of it.
Edward de Bono
Geochemical modeling is an important tool in environmental studies, and in the areas of subsurface and surface hydrology, pedology, water resources management, mining geology, geothermics, hydrocarbon geology, and related areas dealing with the exploration and extraction
of natural resources. Geochemical modeling simulates the chemical and physical processes
affecting the distribution of chemical species in liquid, gas, and solid phases. The reactions
and processes, and their coupled interactions, are dependent on a number of environmental
variables (e.g., temperature, pressure, ionic strength), and are also affected by the dynamics
of matter and energy flows, including fluid, solute and heat flow.
This book seeks to make easily available to a broad readership some basic knowledge and
fundamental concepts regarding the mathematical modeling of geochemical and groundwater processes, from the shallow vadose zone to deep geothermal systems. Since it is impossible
to cover this subject given the space limitations, a selection of essential topics and case studies has been made. However, almost all the developments described herein are discussed in
detail. Fundamental concepts, and the physical laws and equations needed to model different
processes are described and presented in a simple and logical manner.
The book explains in a didactic manner the different applications of geochemical modeling, the existing conceptual methods, and the mathematical and numerical tools that can
provide useful solutions. The reader will also receive a thorough understanding of (i) the
physical laws describing the flow of mass and energy and the transport of solutes, (ii) the partial differential or algebraic equations representing these laws, and (iii) the principal numerical methods that allow approximate solutions of these equations and their corresponding
mathematical models.
In the last 20 years, significant progress has been made in the use of different computational methods for geochemical and groundwater modeling. New concepts, methods and
important findings have been presented in numerous publications. However, these often cover
either specific issues or are restricted to specific systems or applications. Other publications
related to the topics presented in this book may be restricted to describing some modeling
software, or alternatively, the knowledge contained may not be applicable or transferable to
groundwater and geochemical modeling without major modifications. Many of the publications do not necessarily describe the basic physical, chemical, mathematical and numerical
theories and principles used. We saw a need for a synoptic compendium on the fundamentals of groundwater and geochemical modeling that would also have broad applicability to
a diversity of environments. Our compendium clearly reveals the need for further research
and development, that to be most useful, should be informed by additional field studies and
practical applications of numerical modeling. Hopefully, this handbook will stimulate its
readership to address some of these needs.
xxi
2012 by Taylor & Francis Group, LLC

xxii

Editors preface

This book is meant to help overcome some of the barriers that hinder the use (or the
correct use) of geochemical modeling. It presents applications of geochemical modeling in
real case studies. Additionally, we emphasize the need for critical consideration and review of
geochemical modeling and simulation results. A well-designed application will (i) start with
a judicious examination of all available primary information, including the methods and
techniques used for extraction and analysis of water, gas and solid samples, (ii) will carefully
consider the applicable processes and domain boundaries of the simulations, and (iii) will
test the simulation results using sensitivity analyses and a variety of validation/verification
techniques. The reader will learn that the quality of the simulation results depends on the
preparation of the model, the quality of the input data, the knowledge of the geological or
pedological situation, the knowledge of the hydrogeochemical parameters and data, and the
knowledge of the initial values and boundary conditions for the given subsurface system.
Our handbook provides a review of physical, chemical, mathematical, and numerical theory, and describes the correct use of computational methods in simulating chemical reaction
processes in low- and high-temperature aqueous systems such as groundwater, petroleum
and geothermal systems. The book also presents new and stimulating ideas for other possible
simulations and applications.
Chapter 1 gives an introduction to the best practices in sampling and sample preservation, which together with accurate analyses play a vital role and are a precondition for geochemical modeling. The chapter also provides a summary of the basic principles of aqueous
geochemistry and thermodynamics; this section is kept purposely short since more detailed
information is given in several excellent text books to which we refer. In Chapter 2, the thermodynamics of gas and mineral solubility in the unsaturated-zone are treated in more detail
since these aspects are not fully dealt with in the previously mentioned textbooks. Chapter 3
describes the governing equations and solution algorithms for geochemical modeling; it contains the basic concepts for the mathematical formulation of homogeneous and heterogeneous chemical reactions, including their kinetic simulation, and for the calculation of species
concentrations. Chapter 4 provides the physical laws and mathematical equations describing fluid heat, and reactive solute transport, and presents the relevant computational solution algorithms for geochemical modeling; the numerical solution methods for the reactive
transport equations are separately discussed in Chapter 5. Chapter 6 discusses how a conceptual model is implemented into a mathematical/numerical model, and indicates the processes, model parameters and data needs relevant for individual field scenarios and modeling
tasks. Limitations and problems of existing thermodynamic databases are discussed. The
principle types of geochemical models (speciation, reaction-path or forward, inverse- and
reactive-transport models) are described together with examples of the most common codes.
Chapter 7 deals with the advances in geochemical modeling for geothermal applications. The
second part of the book (Chapters 812) discusses the application of geochemical models
in different scientific areas and environmental settings. The focus is on the practical aspects
of modeling, by the use of case studies of real-world environmental problems, including
(i) inverse and forward modeling of heavy metal transport in an aquifer under acidic conditions, (ii) modeling and measurements of porosity and permeability evolution in a sandstone
aquifer, (iii) geochemical modeling of water chemistry evolution along groundwater flow
paths, (iv) modeling of reactive solute transport at a site contaminated by petroleum hydrocarbons, and (v) modeling for preliminary assessment of natural remediation of phosphorus
in variably saturated soil.
This book is addressed to students, teachers, other professionals, and to the institutions
involved in water resources and environmental management. We hope that it will give them
an introduction on the practical use of geochemical modeling in their fields, and that it will
beneficially contribute to policy-making. We also hope that this book will provide a reference used by educational and scientific institutions. The book should prove useful to senior
undergraduate and graduate students, postgraduates, professional geologists and geophysicists,

2012 by Taylor & Francis Group, LLC

Editors preface xxiii

engineers, environmental scientists, soil scientists, hydrochemists, and others interested in
groundwater and geochemistry.
The book fills a gap in the literature through its presentation of insights in geochemical
modeling as applied to a diversity of subsurface systems from close to the Earths surface,
down to deep-seated geothermal reservoirs. It also encourages a broad community of environmental scientists, geologists, chemists, pedologists, hydrologists, engineers and applied
mathematicians to join together to help better protect and manage our freshwaters and our
environment.
Finally, through its 2012 publication date, this book reminds us of the 50th anniversary of
a key milestone in geochemical modeling. In 1962, R.M. Garrels and M.E. Thompson of
Harvard University published the first paper that used thermodynamic modeling to predict
the speciation of seawater.
Jochen Bundschuh
Michael Zilberbrand
January 2012

2012 by Taylor & Francis Group, LLC

About the editors

Jochen Bundschuh (1960, Germany) received his PhD from the University of Tbingen in
1990 after completing a thesis on the numerical modeling of heat transport in aquifers in
1990. He works in the areas of geothermics, subsurface and surface hydrology, integrated
water resources management, and associated disciplines. From 1993 to 1999 he served as an
expert for the German Agency of Technical Cooperation (GTZ), and as a long-term professor for the DAAD (German Academic Exchange Service) in Argentine. From 2001 to 2008,
he served as an adviser to the Costa Rican government at the Instituto Costarricense de
Electricidad (ICE) within the framework of the German governmental cooperation program
(Integrated Expert Program of CIM; GTZ/BA). He assisted Costa Rica in the evaluation and
development of its huge low-enthalpy geothermal resources for power generation. Starting in
2005, he was appointed Affiliate Professor at the Royal Institute of Technology, Stockholm,
Sweden. In 2006, he was elected Vice-President of the International Society of Groundwater
for Sustainable Development ISGSD. From 2009 through 2011, he was Visiting Professor at
the Department of Earth Sciences at the National Cheng Kung University, Tainan, Taiwan.
At the end of 2011, he was appointed as Professor of Hydrogeology at the University of
Southern Queensland, Toowoomba, Australia.
Dr. Bundschuh has authored the books Low-Enthalpy Geothermal Resources for Power
Generation (2008) and Introduction to the Numerical Modeling of Groundwater and
Geothermal Systems: Fundamentals of Mass, Energy and Solute Transport in Poroelastic
Rocks (2010) (both CRC Press/Balkema). He also edited the books Geothermal Energy
Resources for Developing Countries (2002), Natural Arsenic in Groundwater (2005), and
the two-volume monograph Central America: Geology, Resources and Hazards (2007),
Groundwater for Sustainable Development (2008), Natural Arsenic in Groundwater of
Latin America (2008). Dr. Bundschuh also serves as an editor for several book series: Multiphysics Modeling, Arsenic in the Environment, and Sustainable Energy Developments
(all CRC Press/Balkema).

xxv
2012 by Taylor & Francis Group, LLC

xxvi

About the editors

Michael Zilberbrand was born in 1949 in the Ukraine. He received his PhD in 1987 from
the Institute of Geological Sciences of the Academy of Sciences of the Ukraine. His thesis
focused on physical and numerical modeling of water and salt transport in the unsaturated
zone. He has taught a course on unsaturated zone processes at the Hebrew University of
Jerusalem. Since 1996, he has worked as a research scientist in the Hydrological Service of
Israel, applying his knowledge of hydrogeological and hydrogeochemical modeling. He is a
member of the Israeli Geological Society.
Dr. Zilberbrand is a co-author of an approved patent on a device for estimating the rate
and parameters of phase transition and vapor transfer in the air. He has conducted numerous field and laboratory studies and modeling of geochemical processes in the vadose zone
and in groundwater systems, and has also conducted fundamental research of aqueous speciation in the vadose zone.
Dr. Zilberbrand has published 17 papers and 6 book chapters, including a publication in
the Encyclopaedia of Surface and Colloid Science. He has reviewed numerous manuscripts
for the United StateIsrael Binational Science Foundation (BSF), Geochimica et Cosmochimica Acta, Journal of Hydrology, Chemical Geology and other international scientific
journals.

2012 by Taylor & Francis Group, LLC

Acknowledgements

This book would be incomplete without an expression of our sincere and deep sense of
gratitude to Pierre Glynn at US Geological Survey (Reston, VA), for his careful reading; his
valuable comments and suggestions greatly improved chapters 1, 3, 4 and 5.
We thank David Parkhurst of the USGS (U.S. Geological Survey, Denver Federal Center,
Denver, CO, USA) for his valuable remarks on Chapter 2. We express our gratitude to Mario
Csar Arriaga Suarez (Dept. of Applied Mathematics and Earth Sciences, Michoacn University UMSNH, Morelia, Mexico) who provided important inputs to improve the mathematical descriptions in Chapters 3 to 5.
We thank the following scientists who reviewed the other chapters and made a number
of very useful suggestions for its improvement: Nicolas F. Spycher (Earth Sciences Division, Lawrence Berkeley National Laboratory, Berkeley, CA; Chapter 7); David Parkhurst
(U.S. Geological Survey, Denver, CO; Chapter 8); Mark Fuhrman (Office of Nuclear Regulatory Research, U.S. Nuclear Regulatory Commission, Rockville, MD; Chapters 8 and 9);
Alexander Yakirevich (Jacob Blaustein Institute for Desert Research, Ben-Gurion University
of the Negev, Sede-Boqer Campus, Israel; Chapters 10 and 11); Simon Emmanuel (Institute of Earth Sciences, The Hebrew University of Jerusalem, Jerusalem, Israel; Chapters 10
and 11); Maarten W. Saaltink (Dept. Geotechnical Engineering and Geosciences, Universitat Politecnica de Catalunya, Barcelona, Spain; Chapter 12); and Micl Mastrocicco (Earth
Sciences Department, University of Ferrara, Ferrara, Italy). We wish to express our sincere
thanks to them, whose efforts contributed to the high quality of the book.
The editors thank also the technical people of Taylor & Francis Group, for the excellent
typesetting of the manuscript. Finally, we thank the board members of the book series Multiphysics Modeling for their editorial suggestions.
Jochen Bundschuh
Michael Zilberbrand
January, 2012

xxvii
2012 by Taylor & Francis Group, LLC

CHAPTER 8
Integrating field observations and inverse and forward
modeling: application at a site with acidic, heavy-metalcontaminated groundwater
Pierre Glynn & James Brown
I had, said he, come to an entirely erroneous conclusion which shows, my dear
Watson, how dangerous it always is to reason from insufficient data.
Sir Arthur Conan Doyle (1891)
Remember that all models are wrong; the practical question is how wrong do they
have to be to not be useful.
George Edward Pelham Box (1987)

8.1

INTRODUCTION

Historical note: This chapter presents a revised version of Glynn and Brown (1996), a study that
discussed in detail the theory and application of inverse and forward geochemical modeling
with the computer codes PHREEQC, PHREEQM, and NETPATH, and the application
of these codes in modeling contaminated groundwaters in the Pinal Creek basin (Arizona,
USA). The modeling study focuses on the evolution and transport of acidic, reducing, waters
because those conditions control the transport of metal contaminants in the groundwaters
of the Pinal Creek basin. In addition to presenting the essential results and conclusions of
Glynn and Brown (1996), this chapter provides a 15-year perspective on the original study. It
discusses follow-up work done in the Pinal Creek basin, comments on lessons learned from
investigations at the site, and provides suggestions that may be useful to geochemical and
modeling investigations at other sites.
The construction of a multispecies reactive transport model used to predict the future
evolution and movement of groundwater contaminants requires, at a minimum, three separate
but related elements: (i) an understanding of the groundwater flow system and its possible
transients, (ii) an understanding of dispersion and other processes causing observed dilution
or mixing of different water types, and (iii) an understanding of the primary reactions
controlling the distribution of various contaminants, not only among solid and phases, but
also within the groundwater itself. The degree of understanding of all three of these elements,
and perhaps more importantly an appreciation for the remaining knowledge gaps, will be
essential in determining the usefulness of the constructed model. Indeed, even though a
groundwater model may not adequately predict the future evolution of a contaminant plume,
the process of constructing and using a model often results in an improved understanding of
contaminant transport at the site.

Sources: *Sir Arthur Conan Doyle: The Adventure of the Speckled Band, The Adventures of Sherlock
Holmes, 1891; **George Edward Pelham Box; In: George E.P. Box and Norman R. Draper: Empirical
Model-Building and Response Surfaces, 1987.

181

CH08.indd 181

11/30/2011 10:36:30 AM

182

Geochemical modeling of groundwater, vadose and geothermal systems

Most groundwater contamination sites have less geochemical and hydrogeologic

information known about them than is desirable for predictive modeling of reactive
contaminant transport. Detailed hydrogeologic and geochemical studies are usually much
too expensive to consider.1 The resulting lack of knowledge on the operative chemical and
hydrologic processes at a given site means that investigators should try to use, as efficiently
as possible, all tools and knowledge available. We believe that a combination of inverse and
forward modeling of groundwater flow, contaminant transport, and geochemical evolution,
may often provide the greatest knowledge gains for the least amount of money and time. In
particular, geochemical inverse modeling should be used first, prior to forward geochemical
modeling, both to explain the currently observed groundwater chemistry in the aquifer
system, and to make predictions on the future chemical evolution of the groundwaters.
This chapter focuses on geochemical modeling and will show how both inverse and forward
geochemical modeling approaches were used to better understand the evolution of acidic,
heavy-metal-contaminated groundwaters in the Pinal Creek basin, near Globe, Arizona. The
Pinal Creek basin is a site with sparse spatial information (30 wells distributed in a 15 km
long and 102 to 103 m wide sulfate plume) and with significant temporal variations in both
chemical and hydrological characteristics (water-table movements of more than 15 m during a three month period, groundwater velocities on the order of 3 to 5 m day1; Brown and
Harvey, 1994). The Pinal Creek site is well suited to test our modeling philosophy.

8.2

GEOCHEMICAL MODELING: COMPUTER CODES, THEORY

AND ASSUMPTIONS

8.2.1

Inverse geochemical modeling

8.2.1.1 Principles, codes and theory

Inverse geochemical modeling uses existing groundwater chemical and isotopic analyses, which
are assumed to be representative of the groundwater along a given groundwater flowpath,
and attempts to identify and quantify the reactions that are responsible for the chemical and
isotopic evolution of the groundwater. Although an aqueous speciation code may be used to
identify thermodynamically possible (or impossible) reactions and to determine the dissolved
inorganic carbon content and the redox state (RS) of the groundwaters, the inverse modeling
approach does not require that reactions proceed to thermodynamic equilibrium. Indeed,
mass-balance constraints and the judgment of the user concerning the possible reactions are
the only constraints posed in the inverse modeling approach.
Inverse geochemical modeling codes (BALANCE, Parkhurst et al., 1982; NETPATH,
Plummer et al., 1991, 1994; PHREEQC,2 Parkhurst, 1995, 1997, Parkhurst and Appelo,
1999; PHREEQCi,3 Charlton et al., 1997 and Charlton and Parkhurst, 1999) essentially
solve a system of algebraic mass-balance equations. These relate the masses of elements,
isotopes, redox oxidation states, and water found in a final sampled water along a flowpath
to the masses found in contributing initial waters along the same flowpath, and to the

Studies at the Cape Cod (LeBlanc, 1984) and Borden sites (Mackay et al., 1986) are examples of what
we would consider detailed studies. On the order of 103 to 104 sampling points were installed to study
plumes on the order of 102 meters to a few kilometers long. However, even at these sites, after three
decades, many questions remain regarding the operative geochemical and hydrogeologic processes, and
studies continue to refine and improve existing knowledge.
2
The PHREEQC series of numerical codes (PHREEQCi, PHREEQC versions 1 and 2) have both
inverse and forward geochemical modeling capabilities.
3
PHREEQCi is an interactive version of PHREEQC with a windows-based graphical user interface.

BUNDSCH_Book.indb 182

11/29/2011 12:31:50 PM

Integrating field observations and inverse and forward modeling 183

masses lost or gained through heterogeneous reactions between the aqueous phase and
other phases (solid, gaseous, exchange or surface sorption phases). Each inverse model
calculated by an inverse geochemical modeling code consists of a set of reaction mass transfers and initial solution fractions that satisfies the algebraic mass balance constraints for
the observed initial and final solution compositions. The list of possible reactions and
contributing initial solutions is postulated by the user. Usually, more than one model, or
set of mass transfers, can be calculated for a given problem, and the user then tries to add
additional constraints, for example additional information regarding the chemical and/or
isotopic compositions of the initial and final solutions, to limit the number of models.
Mass-action equations and thermodynamic equilibrium speciation of the initial and
final aqueous solutions are also solved in the NETPATH and PHREEQC codes (but not in
the earlier BALANCE code). These speciation calculations primarily serve to inform the user
as to the reactions that might be thermodynamically feasible. The calculations are also used
to establish the redox and charge balance states and to determine the total dissolved inorganic carbon balance for the various aqueous solutions. Significant differences exist between
the NETPATH and the PHREEQC inverse modeling capabilities, for example in the treatment of redox balances and solution charge balances. Perhaps, the most important difference
between NETPATH and PHREEQC is that PHREEQC allows each analytical input datum
for each of the aqueous solutions to be adjusted within an uncertainty range specified by
the user. PHREEQC then calculates sets of phase mass transfers, solution mixing fractions,
and adjustments to the analytical data that satisfy the mass-balance constraints, are consistent with the specified uncertainty ranges, and minimize the sum of the adjustments to
the analytical data. The mathematical descriptions of the NETPATH and PHREEQC (and
PHREEQCi) codes will not be discussed here. The reader is referred to Glynn and Brown
(1996), Parkhurst (1995, 1997), Parkhurst and Appelo (1999), and Plummer et al. (1994) for
further information. More recently, in their guide on radiocarbon dating of groundwater
systems, Plummer and Glynn (in press) review the unique capabilities of the NETPATH code
to simulate Rayleigh fractionation processes and assess groundwater residence times through
inverse geochemical modeling.
8.2.1.2 Assumptions used in inverse modeling
By definition, a model is a construct of assumptions that is meant to help understand some
facet(s) of reality. Inverse geochemical modeling of groundwaters requires the user to make
assumptions concerning (i) the types of geochemical reactions postulated to be present,
(ii) the rates of reaction relative to the movement of the water and its mobile constituents,
and (iii) the present distribution of chemical constituents in the aquifer system studied and
the prior evolution of this distribution. The last 2 sets of assumptions require that the user
have some presumptive knowledge of the groundwater flow and transport system and of its
prior evolution.
Knowledge of flowpaths and the assumption of a steady-state groundwater flow field
These are the most important and possibly the most tenuous assumptions used in inverse
modeling of the chemical and isotopic evolution in a groundwater system. The user often does
not have enough hydrologic knowledge to precisely determine the flowpaths in a groundwater
system. Furthermore, even if there is sufficient knowledge of the hydrogeologic system,
existing wells must often be used. One rarely has the luxury of emplacing new sampling wells.
However, when analyses are available from several wells, the spatial array of chemical and
isotopic information may itself be used to decide the most likely flowpath. In most cases, the
user will pick a direction that shows the least amount of dilution for the most conservative
solutes to deduce the direction of flow.
Most groundwater systems are likely to experience some seasonal and interannual
fluctuations in hydraulic heads. Therefore, flowlines and groundwater velocities are likely to
change at least seasonally, and steady-state conditions may not apply during the time scale

BUNDSCH_Book.indb 183

11/29/2011 12:31:51 PM

184

Geochemical modeling of groundwater, vadose and geothermal systems

of interest. The time scale of interest will normally be the time required for groundwater
flow between the wells used in the inverse modeling simulation. The user typically assumes
a steady-state groundwater flow field over the time scale of interest, or at the least assumes
that the groundwater flow field observed at the time of sampling represents the long-term
average velocity field.
The assumption of chemical steady-state
The groundwater analyses used in an inverse model usually represent samples taken concurrently
or near-concurrently. The inverse modeling approach generally assumes that the parcel of
water sampled from a final, downgradient, well (well B in Fig. 8.1) used to have the same
composition as that of the water sampled concurrently at an initial, upgradient, well (well
A in Fig. 8.1). This assumption will certainly be reasonable if the groundwater system has
remained in chemical (and isotopic) steady-state at least during the travel time required for the
water to move from the initial well to the final well. The assumption of chemical and isotopic
steady-state simply states that although chemical and isotopic compositions may vary spatially,
they do not vary in time at any given point in the groundwater system. In groundwater systems
with spatially varying chemical and isotopic compositions, the assumption of chemical steady
state implies a steady-state groundwater flow field (i.e., hydrologic steady-state), that is, flow
lines and groundwater velocities that have not varied in time.
Most groundwater contamination cases involve dynamically evolving contaminant
plumes, for which there can be no assumption of chemical steady-state. (Steady-state plumes,
in which the rate of diffusive/dispersive and reactive loss of solutes balances the rate of their
influx and/or production are uncommon). Fortunately, the assumption of chemical and

Figure 8.1. Two map views of a groundwater contaminant plume. Left: actual layout of the plume,
drawn with a single concentration contour of concern. A is a well emplaced near or in the source of
the contamination (stippled) and B and C are wells further downgradient. Right: results of a transport
model for the same groundwater contaminant plume based on a fit of concentration data obtained
from several observation wells. Additional concentration contours are drawn. The large transverse and
longitudinal dispersion of the modeled plume results not only from the mixing that actually occurs
in the ground but also from mixing that occurs during pumping at the observation wells. Uncertainty
in simulation results is also caused by the inability to obtain a sufficiently detailed time-dependent
representation of the contaminant plume and of the transient groundwater velocity field.

BUNDSCH_Book.indb 184

11/29/2011 12:31:51 PM

Integrating field observations and inverse and forward modeling 185

isotopic steady-state is not required for inverse geochemical modeling to be meaningful. The
less stringent constraint required for inverse modeling is that the initial, upgradient, water
composition used in inverse modeling (well A, sampled ideally at time t0, or more commonly
at time ts) should be representative of the composition that the final water sampled downgradient at well B (at time ts) used to have, prior to its chemical evolution and movement
from A to B. This less stringent constraint allows the groundwater compositions at points
in between the initial and final wells (A and B) to have varied with time, as long as the
chemical composition of waters from the initial well (A) has remained invariant, or the
appropriate sample from well A at time t0 can be obtained. Strictly from a mass-balance point
of view, it could be argued that the constraint could be reduced further to require only that
the changes in composition between waters from well A (at time t0) and B (at time ts), rather
than the actual compositions of waters from wells A (at time t0) and B (at time ts) should
have remained constant. (A uniform dilution, or concentration, of the waters sampled at
the initial and final wells, however, could lead the user to conclude from his inspection
of the mineral saturation indices and general speciation of the waters that some other set of
reactions was responsible for the evolution of water A into water B.) Finally, although the
locations of the initial and final waters used in inverse geochemical modeling do not
enter in the calculation of the reaction sets and reaction extents responsible for the overall
evolution of water A into B, location information is important in assessing whether the reactions are reasonable from a kinetic, mineralogic and hydrologic perspective.
How does mixing occur in groundwater systems?
The U.S. Geological Survey inverse geochemical modeling codes (BALANCE, NETPATH,
and the general geochemical code PHREEQC) calculate mixing fractions of initial waters
and extent of reactions with solid and gas phases that account for the chemical composition of
the final water. Clearly in most groundwater environments, mixing of groundwaters should
be modeled as a continuous process rather than as a discrete process where a small number
of specified water compositions are mixed together. Unfortunately, the inverse geochemical
codes presented here cannot replicate a continuous mixing process. Forward transport
modeling codes can replicate a continuous mixing process such as dispersion, but even then,
their results and the very basis of their conceptual models are usually fraught with uncertainty.
In using a set of discrete initial water compositions, inverse geochemical models inherently
assume that the initial waters chosen encompass the range of intermediate waters that are
actually involved in the real, continuous mixing process.
Although the location, timing, and sequence of the mixing and reaction processes is of no
mathematical significance in the solution of the mass-balance equations, the user should try
to determine where, when, and why such mixing processes have occurred in the groundwater
environment. The mixing of initial waters by dispersion, for example, may well have led
to heterogeneous mass transfers in areas that are not directly on the flowpath between a
principal initial water and the final water. The inverse models will, nevertheless, implicitly
incorporate those mass transfers in their solution of net mass transfer amounts.
The premise of inverse geochemical modeling is that the final and initial
groundwaters used in a model should be related to each other. Ideally, they represent very small
volumes of water sampled from a unique flowline or pathline. If it were indeed possible to do
so (it is not), then the final groundwater sampled could only have experienced mixing as
the result of the following processes:
1. Diffusion of chemical and isotopic constituents (and possibly of water) to or from the
flowline (or pathline) to neighboring flow lines or to stagnant water zones.
2. Sampling from the pathline for more than an infinitesimally small amount of time, in the
case of a system not in chemical steady-state. (Although most groundwater analyses do
not require large samples, and the samples are therefore typically collected over small time
periods, the length of sampling time may be an important consideration for some analytes.)
If the system is not in chemical steady-state, the concentration of various constituents

BUNDSCH_Book.indb 185

11/29/2011 12:31:51 PM

186

Geochemical modeling of groundwater, vadose and geothermal systems

may be changing as a function of time at the final well and therefore a sample may in fact
represent some groundwater composition averaged out in time and therefore in space.
3. The sampling of multiple flow lines that have undergone different chemical and isotopic
evolution. Sampling-induced mixing of a diversity of solution chemistries may occur
particularly in regions of converging flow, and may also occur when sampling from wells
screened across large or multiple intervals. Flow convergence may occur naturally or may
be the direct result of pumping.
Using the mixing option in an inverse geochemical model may also be needed because
the initial and final groundwaters may not be chemically related despite the best efforts
of the user. Just as excessively high values of dispersivity are often used in groundwater transport modeling because of a lack of precise spatial and temporal information (see
Fig. 8.1), the use of the mixing option in NETPATH or in PHREEQC can often be the
result of insufficient information on a groundwater system. For example, in Figure 8.1, if
the final well used in the inverse model was off to the side of the path line of heaviest contamination (well C for example) and if well A was used as the initial well, the inverse model
defined by the user would probably require a significant contribution of background water
to explain the extent of dilution between well A and the final well. Similarly pumping
a large amount of water from the final well chosen (well B or C) and using the average
composition of this water as the final water composition in the inverse model could also
lead to a serious misrepresentation of the amount of mixing. An error in the mixing fractions
of initial waters could result in significant errors in the reaction mass transfers calculated
by the inverse model. Furthermore, using water compositions averaged over a large volume
by the sampling process could mislead the modeler into thinking that certain reactions were
thermodynamically impossible, when in fact proper sampling, and location, of the initial
and final waters would have indicated that these reactions were in fact possible.
8.2.2

Forward geochemical modeling

8.2.2.1 Principles and codes

Unlike inverse geochemical modeling, which attempts to explain observations and quantify
the reactive mass transfers and mixing processes responsible for the observed chemical
evolution along a flowpath in a groundwater system, forward modeling and reactive-transport
modeling attempt to predict the evolution of a groundwater given certain postulated
reactions and hydrodynamic processes. In their study, Glynn and Brown (1996) used the
PHREEQM (Appelo and Willemsen, 1987) code for their reactive-transport simulations,
rather than PHREEQC, because version 1 of PHREEQC (Parkhurst, 1995), the version
that was available at the time, only simulated advective transport and reactions. PHREEQC
version 2 (Parkhurst and Appelo, 1999) implemented advective-dispersive transport, by using
an algorithm similar to the one used in PHREEQM. Because the concepts of forward geochemical modeling and reactive-transport modeling are widely known, this discussion will be
limited to a brief description of the reactive transport capabilities of the PHREEQM and
PHREEQC geochemical codes.
The PHREEQM code, and later the PHREEQC code, added significant capabilities to
earlier geochemical speciation and mass-transfer codes, such as the U.S Geological Survey code
PHREEQE (Parkhurst et al., 1980). PHREEQM and PHREEQC simulated the transport
of aqueous solutions by advection, dispersion, and diffusion in a 1-dimensional column
(made up of a series of cells) and simulated the reaction of those solutions with minerals
and surfaces inside the column. The codes also allowed simulation of cation-exchange processes. PHREEQM typically used the local equilibrium assumption in its modeling of reactive
transport. PHREEQC, starting with its second version (Parkhurst and Appelo, 1999), implemented a wide range of capabilities that were not available in PHREEQM. These included
modeling of kinetically limited reactions, modeling of additional surface speciation and

BUNDSCH_Book.indb 186

11/29/2011 12:31:51 PM

Integrating field observations and inverse and forward modeling 187

sorption reactions, modeling of solid-solution aqueous-solution reactions as presented in
Glynn (1991a, 2000), transverse diffusion, and many other capabilities.
The transport algorithms in PHREEQM and PHREEQC use an operator splitting technique, where advection is modeled by shifting cell contents from one cell to the next at every
time step or shift. Dispersion and/or diffusion is simulated by mixing the aqueous contents
of each cell with that of adjacent cells. This algorithm gives the codes the advantage (over
most typical finite-difference and finite-element codes) of being able to simulate not only
an advective-dispersive transport process or a diffusive transport process, but also a purely
advective, albeit one-dimensional, transport process. The mixing factors f calculated are a
function of the aquifer dispersivity and the molecular diffusivity D*:
fi =

i + i +1
t
+ 4 D*
2
li li +1
(li li +1
+1 )

(8.1)

where i is the cell number, l is the cell length, and t is the time step.
The above equation can be derived from a finite difference approximation (ignoring advection) centered in space and forward in time. Because the simulation of dispersion is centered in
space, PHREEQM and PHREEQC show no numerical dispersion error for conservative constituents when simulating advection-dispersion processes. Numerical dispersion does occur,
however, in dispersion of non-conservative constituents and is dependent on the amount of
retardation experienced by each constituent and the cell lengths (the maximum numerical
dispersivity equals 1/2 the cell length; cf. Herzer and Kinzelbach, 1989). The codes do not
show any numerical dispersion in simulations with only diffusion as a transport process. The
lack of sequential iterations between the solution of the chemical equilibrium equations and
the simulation of the transport processes at every time step can theoretically generate errors,
although comparisons (Glynn et al., 1991; or Figs. 9 and 10 in Glynn and Brown, 1996) of
PHREEQM with the sequential iteration finite difference code MSTID (Engesgaard and
Kipp, 1992) suggest that the error is typically small as long as an appropriate discretization is
used. Operator splitting can also generate error, although, again, comparisons of PHREEQM
with the MSTID code and results by Steefel and MacQuarrie (1996) suggest that this error is
usually minor. A much more complete description of the PHREEQM and PHREEQC codes
and their capabilities can be found in Appelo and Postma (1993 and 2005, respectively).
Unless mentioned otherwise, all PHREEQC simulations referred to in this chapter were performed with the first version of the code (Parkhurst, 1995).
Forward geochemical modeling is conceptually different from inverse geochemical modeling.
Indeed, forward modeling seeks to predict the chemical composition of an aqueous solution,
given the composition of an initial solution and given certain specified reactions, some of which
are usually considered to reach thermodynamic equilibrium (or have a prescribed kinetic progression). Forward modeling is most suitable and most useful, when the amount of chemical
and isotopic data available for a given groundwater system is limited, and when the objective is
to predict the future evolution of the system. In contrast, inverse modeling is most useful when
abundant chemical, isotopic, hydrologic and mineralogic data are present and all that is desired
is an explanation of the past chemical evolution of the groundwater system. Of course, just as
understanding the past is a key to understanding the future, inverse modeling can provide insight
regarding the reactions that control the future chemical evolution of a groundwater system.
Forward modeling codes can also be used for the purpose of inverse modeling. A series of
trial and error simulations or automated parameter estimation can be used to adjust reaction extents or reaction rates to align forward simulation results with actual observations
(VanCappellen and Gaillard, 1996; Steefel and MacQuarrie, 1996). Although this latest
approach can be time consuming, it does have the advantage over simpler inverse geochemical codes of a more accurate representation of groundwater mixing as a continuous, rather
than discrete, process. This parameter-estimation approach does not necessarily require the

BUNDSCH_Book.indb 187

11/29/2011 12:31:51 PM

188

Geochemical modeling of groundwater, vadose and geothermal systems

assumption of chemical steady-state, although it is subject to the same constraints as simpler

inverse geochemical codes in choosing initial and final water compositions (see discussion in
previous section), their location and time of sampling. The disadvantages of this approach,
relative to non-transport-oriented inverse geochemical modeling, are the computer time
requirements, the significantly greater number of adjustable parameters, including flow and
transport parameters, and the consequently greater number of solutions that may explain the
observations. Further references in this chapter to inverse modeling will generally not refer
to the use of forward codes as part of an inverse modeling approach, although many of the
statements made may apply equally well to this latter more sophisticated approach.
The Pinal Creek Toxics Program investigation site, a site of groundwater contamination by
acidic, metal-laden, and sulfate-rich wastewater near Globe, Arizona provides a good example of the improved understanding of the chemical reaction and transport processes that
may be gained through the combined use of both inverse modeling and forward modeling
approaches. The primary purpose of this chapter is to describe some of the insights that can
be gained through this combined approach, and through the examination of available field
observations. A brief description of the site is first needed.

8.3

THE PINAL CREEK BASIN SITE: BRIEF DESCRIPTION

The Pinal Creek basin is located in central Arizona, about 100 km east of Phoenix (Fig. 8.2). The
surface drainage area of the basin occupies 516 km2, of which 170 km2 is covered by alluvium
and basin fill, which form the regional aquifer; 27 km2 of the basins area is covered by mine

Figure 8.2. Location and generalized geology of the Pinal Creek basin study area. Several wells screened
at different depths exist at each well group site. Well names are related to site names by their first, or first
and second, digits. For example: site 400 includes well 402; site 500 includes wells 503 and 504; site 50
includes well 51; site 450 includes wells 451 and 452. Dashed dotted stream lines represent intermittent
streams. All stream reaches are intermittent, except Pinal Creek itself to the north of well group 500.

BUNDSCH_Book.indb 188

11/29/2011 12:31:52 PM

Integrating field observations and inverse and forward modeling 189

tailings. Mining, mainly for copper, began in the late 1870s and has been the largest economic
activity in the basin. Because of the long history of mining, there are several potential sources of
contamination to the regional aquifer (Eychaner, 1991; Brown and Harvey, 1994). For example,
pyrite is thought to be the most abundant sulfide mineral in the tailings. Following significant
rainfall, oxidation of pyrite in the tailings and subsequent transport of this contaminated water
into the alluvial aquifer could represent a significant source of acidity, iron, and sulfate in the
aquifer. A similar phenomenon could also occur in the relatively undisturbed naturally mineralized areas in the basin, although the amount of this flow is expected to be small in relation to
flow from the disturbed mining areas and tailings. Finally, the unlined water impoundments that
were used during copper refining and mineral processing operations are also a likely source of
groundwater contamination in the basin. The water in the impoundments is rich in sulfuric acid
(used in copper refining) and heavy metals. The largest suspected single source of contamination
in the basin was Webster Lake, an unlined surface-water impoundment that existed from 1940
to 1988. In 1988 it was drained at the order of the U.S. Environmental Protection Agency (Tolle
and Arthur, 1991). Maximum volume of the lake was more than 7 million m3. In 1988, a sample of water from the lake had a pH of 2.7; concentrations of iron and sulfate were 6 g L1 and
20 g L1, respectively; and aluminum, copper, cobalt, nickel, and zinc were present at concentrations greater than 20 mg L1.
8.3.1

Geology

Peterson (1962) describes the complex and highly diverse geology and mineral deposits of
the Pinal Creek basin area. Older Precambrian age rocks in the basin include schist, diorite,
granite, conglomerate, quartzite, limestone, and basalt. Paleozoic age rocks include quartzite,
limestone and shale. Younger Mesozoic- and Cenozoic-age rocks are mainly intrusive and
include granite, granodiorite, diabase, and monzonite. These are exposed in the hills north
of Globe and Miami. Finally and most importantly from an economic perspective, the
igneous and metamorphic rocks include a major body of copper porphyry ore. Chalcocite,
chalcopyrite, and pyrite predominate in the deeper parts of the ore body, while chrysocolla,
malachite, and azurite predominate in its upper, oxidized zone.
The Pinal Creek basins configuration was created by high-angle block faulting associated with
basin subsidence that began 19 to 15 million years ago and continued until about 8 million years
ago. The basin fill, which is derived from rocks of the surrounding mountains and forms the
deeper aquifer in the basin, has a highly variable lithology, ranging from completely unsorted
and unconsolidated rubble of angular blocks as much as 4.5 m in diameter, to well-stratified
deposits of firmly cemented sand, silt, and gravel containing well-rounded pebbles and cobbles (Peterson, 1962). Carbonate content of the basin fill is about 1.5% (Eychaner, 1989).
Unconsolidated stream alluvium overlies the basin fill along Miami Wash, Pinal Creek,
and other drainages. The alluvium is from 300 to 800 m wide and is less than 50 m thick.
The alluvium contains cobble- to clay-sized material, although sand- to gravel-sized material
is most abundant. Sand-sized particles contain quartz, feldspar, and lesser amounts of mica
and a variety of rock fragments. Gravel-sized materials consist mainly of rock fragments of
granite, volcanic rocks, and schist. The alluvium also contains interbedded clays and lenticular clay layers that were found to be as much as 12 m thick at Nugget Wash (Hydro Geo
Chem, Inc., 1989). A sample of alluvium collected in 1985 (Eychaner and Stollenwerk, 1985)
contained 0.34 percent calcite by weight. This concentration is equivalent to 0.18 moles of
carbonate material per liter of water, using the bulk density of 1.65 g cm3 and porosity of
0.316 determined for alluvium used in a column experiment (Stollenwerk, 1994).
Manganese cycling and transport is of interest in our study. Indeed, secondary manganese
oxide deposits are highly visible in the perennial Pinal Creek streambed, frequently forming
a black hardpan layer. Estimates of the concentration of primary manganese oxide minerals were based on samples of alluvium not affected by acidic contamination. At well site
500, in the neutralized part of the plume, the depth-averaged content of manganese oxides

BUNDSCH_Book.indb 189

11/29/2011 12:31:55 PM

190

Geochemical modeling of groundwater, vadose and geothermal systems

was 0.079 mol L1, based on sequential extractions done by Ficklin et al.(1991). Stollenwerk
(1994) estimated that 0.0449 mol L1 of manganese could be available for reaction in a sample
of alluvium obtained from a gravel quarry just east of well site 200.
8.3.2 Hydrology and groundwater flow
Groundwater flow in the Pinal Creek basin is the result of past geologic events, the past and
present climate, and human activities. Because the area climate is semiarid, most of the drainages in the basin are usually dry but convey large amounts of runoff during and after severe
storms. Streams that drain the Pinal Mountains also flow during and following snowmelt
in late winter and early spring. Most of the groundwater recharge in the basin occurs in the
winter and spring, with lesser recharge events during the summer monsoonal period. The
amount and the distribution of precipitation controls the size, frequency, and duration of
streamflow, and therefore controls the quantity and distribution of water that infiltrates the
permeable stream beds and recharges the unconsolidated alluvial and consolidated basin-fill
aquifers.
Groundwater in the basin fill flows generally northward from the flanks of the Pinal
Mountains, and westward from the Apache Peak alluvial fan. Most groundwater in the
basin fill eventually flows upward into the alluvium and then flows generally north to the
perennial reach of Pinal Creek. A greater quantity of water, however, recharges the alluvium
directly and flows north, mixing with the water from the basin fill. In the northern part of
the basin, the aquifer is constricted by the impermeable basement rocks. This constriction
forces groundwater to the land surface, generating perennial flow from about 6 km above
Inspiration Dam to the Salt River, which is a major source of drinking water for the Phoenix
Metropolitan area.

8.4

INVERSE GEOCHEMICAL MODELING AT THE PINAL CREEK SITE

Like most sites with point-source groundwater contamination, the chemistry of groundwaters in the Pinal Creek basin exhibits both spatial and temporal variation. The most heavily
contaminated groundwaters are typically found near the base of the unconsolidated alluvial
aquifer, where a zone of coarser (and possibly less carbonate-rich) material is suspected to
be present. The wells with the most contaminated waters at each well site are 51, 101, 302,
402, 503, 601 and 702 in a down-gradient direction (Fig. 8.2). Although other wells at each
site also show the presence of contaminated water, wells with the most contaminated waters
(as measured by total dissolved solids, chloride, or any other relatively conservative constituent) present the most logical choice for inverse modeling. To further narrow the scope of the
inverse-modeling study, we focus on the significant changes in the chemical characteristics
between samples from well 402 and 503. The two water samples chosen represent an acidic
contaminated water sampled from well 402 in January 1989 and a neutralized contaminated
water sampled from well 503 in November 1991. The two wells are 5.6 kilometers apart.
From the difference in sampling times and the distance between wells, we calculate that a
parcel of water leaving well 402 would have to travel at a linear groundwater velocity of about
5.2 m day1. This velocity is in the range of 4.2 to 5.6 m day1 estimated by using Darcys law,
assuming an effective porosity of 0.3 and a hydraulic conductivity of 200 m day1 (Brown
et al., 1995; Brown, 1996).
Glynn and Brown (1996) initially used the NETPATH inverse modeling code. Unlike the
first version of PHREEQC, the only one available at the time, the NETPATH code was
interactive and allowed the authors to quickly determine primary sources and sinks in their
inverse simulations. The NETPATH code (i) helped identify some of the possible reaction
mechanisms responsible for the chemical evolution of the groundwaters between wells 402
and 503, and (ii) quantified some of the reaction mass-transfers involved.

BUNDSCH_Book.indb 190

11/29/2011 12:31:55 PM

Integrating field observations and inverse and forward modeling 191

8.4.1

Examination of end-member waters and their conservative constituents

The first step in an inverse modeling study is to examine the chemical composition and
thermodynamic state of the waters that will be used in the model. The chemical analyses for
the waters chosen for the inverse modeling study are given below in Table 8.1. The charge
imbalances calculated by the speciation code WATEQFP (embedded in NETPATH) for the
well 402, 504 and 503 samples were 3.1, 1.5 and 4.4%, respectively.
As can be seen, the most significant differences in groundwater chemistry between the
well 402 and 503 samples are the increase in pH from 3.9 to 5.6, the 25% increase in calcium
and magnesium, the complete removal of 590 mg L1 of dissolved iron, the 90% removal of
18.4 mg L1 of dissolved aluminum, the 60% increase in manganese, and the 30% decrease in
sulfate. Silica concentrations are nearly constant. As is the case in any geochemical modeling
analysis, however, conservative (i.e., nonreactive) constituents are perhaps the most important constituents to examine because they give information on physical flow and transport
processes. Any groundwater sampling and analysis program should ensure measurement of
at least one, but preferably two or more, relatively non-reactive tracers, such as chloride, bromide, 18O and 2H. Sodium may also be relatively conservative although it may increase in solution due to cation exchange, feldspar dissolution, or evaporite dissolution processes. Sodium
is rarely taken out of solution by reaction processes other than cation exchange. In the case
of the Pinal Creek groundwaters, the high Ca/Na ratio in the acidic waters (Ca/Na = 2.4 mol/
mol in well 402) and the even higher ratio in the neutralized waters (Ca/Na = 4.2 mol/mol in
well 503) suggest that removal of sodium by cation exchange is not likely.

Table 8.1. Chemical composition of three groundwaters from the Pinal Creek basin: an acidic
contaminated water (well 402), a background uncontaminated water (well 504), and a neutralized
contaminated water (well 503). Concentrations in mg L1. Concentration changes are expressed relative
to well 402. ND: not determined. TDIC: total dissolved inorganic carbon.
Concentration or
value

Well 402
1989/1/12

Well 504
1991/11/22

Well 503
1991/11/22

pH
Eh (in mV)
Temperature
Dissolved oxygen
Calcium
Magnesium
Sodium
Potassium
Iron
Manganese
Aluminum
Strontium
Silica (as SiO2)
Chloride
TDIC (as C)
Alkalinity (as HCO3)
Sulfate
Fluoride

4.13
420 est.2
18C
0.3
502
161
121
7 est.2
591
71.6
18.4
2.29
85.6
140
50
ND
3260
10 est.2

7.05
350
20.5C
6.64
44.63
15.63
19.83
2.1
0.004
<0.001
<0.01
0.335
273
9.73
ND
227
14.23
0.3

5.59
410
18.2C
<0.1
634
200
86
5 est.2
<0.1
116
2.3 est. 2
2.7
91.8
112
ND
66
2350
1.5 est.2

Change:
402 to 503

Change:
Reaction
only1

1.1%
26%
24%
29%
29%
100%
62%
88%
18%
7.2%
20%
ND
ND
28%
85%

57%
54%
13%
16%
100%
106%
84%
44%
26%
0%
ND
ND
8%
81%

1
Assumes chloride is conservative. The relative change expressed represents the relative difference in
concentration between a mixture of waters from wells 402 and 504, determined on the basis of chloride
concentrations, and well 503 water. 2 Values were estimated by inspecting earlier and later analyses.
3
Average of two analyses.

BUNDSCH_Book.indb 191

11/29/2011 12:31:55 PM

192

Geochemical modeling of groundwater, vadose and geothermal systems

The decrease in both Cl and Na between wells 402 and 503 suggests that dilution is
occurring. This dilution may be caused either by longitudinal and transverse dispersion
along the flowpath or by the well-sampling process. It is also important to recognize that
the groundwater sample taken from well 503 in November 1991 was probably not exactly on
the pathline originating from well 402 in January 1989. Well 503 may be off (to the side
or above) the most contaminated pathline, and the well 503 sample might therefore be more
diluted than a hypothetical sample taken from the pathline. In recognition of the difficulty
in determining the causes and the exact proportions of the various groundwaters responsible
for the dilution of the well 503 water relative to the well 402 water, an uncontaminated water
sampled in November 1991 from below the plume at well 504 was used as the source of diluting water in our inverse geochemical model.
Although chloride undergoes a 20% decrease between wells 402 and 503, sodium undergoes an even greater decrease of about 29%. If the decrease in chloride is used to calculate the
fraction (0.21) of water from well 504 diluting the water from well 402, the observed sodium
concentration in well 503 is still l3% lower than the calculated diluted sodium concentration
(Table 8.1, last column). This greater observed decrease in sodium may be at least partly due
to a greater Cl/Na ratio in the average diluting water relative to that of the background water
(well 504) used in the calculation. Indeed, although the average Cl/Na ratio in the uncontaminated waters below the plume (wells 404, 504) or upgradient (well 010) from the plume
is 0.44 mg/mg (0.10), the average Cl/Na ratio for the most contaminated waters along the
flowpath is close to three times higher (well 51:1.48 0.74, well 101: 1.35 0.44, well 302:
1.32 0.36, well 402: 1.29 0.44, well 503: 1.17 0.43; all ratios in mg/mg). [Note the decreasing Cl/Na ratio with distance downgradient, i.e., with increasing neutralization and dilution
of the contaminated waters.] Dilute (only slightly contaminated) groundwaters sampled from
wells on the side of the plume (wells 201, 202) also have a much higher average Cl/Na ratio
(0.91 0.30) than that of the uncontaminated groundwaters. An argument can therefore be
made that these slightly contaminated waters should have been used as diluting waters in the
NETPATH modeling, instead of (or in addition to) the uncontaminated water chosen here.
The discrepancy in the chloride and sodium dilution factors is, however, a measure of the
uncertainty inherent in trying to model the groundwater mixing process with a simple inverse
geochemical model.
8.4.2 The thermodynamic state of the end-member waters
After examining the conservative constituent concentrations of the groundwaters, the next
step is to examine the aqueous-speciation results, in particular the mineral saturation indices (Table 8.2) calculated for the three end-member waters chosen in our model. The speciation calculations were performed with the WATEQFP code incorporated in the database
management code DB distributed with the NETPATH code. The thermodynamic database
used in WATEQFP is a subset of the database described for WATEQ4F (Nordstrom et al.,
1990), and is also the basis for the phreeqc.dat and wateq4f.dat databases included with the
PHREEQC code (Parkhurst and Appelo, 1999).
The background water (well 504) is predominantly rich in CaHCO3, and is typical
of the uncontaminated groundwaters in the Pinal Creek basin. These waters are usually near saturation with calcite, dolomite and chalcedony, have near to slightly above
neutral pH values, and have equilibrium CO2 partial pressures between 101.5 and 102.0.
The uncontaminated groundwaters are typically rich in dissolved oxygen and other
dissolved atmospheric gases (Glynn and Busenberg, 1994a; Robertson, 1991, Winograd
and Robertson, 1982).
In comparison, the acidic water from well 402 is highly undersaturated with respect to
calcite, dolomite, siderite (FeCO3, SI: 2.6) and rhodochrosite (MnCO3, SI: 3.3) and is
near saturation with amorphous silica, kaolinite, and gypsum. The water is undersaturated

BUNDSCH_Book.indb 192

11/29/2011 12:31:55 PM

Integrating field observations and inverse and forward modeling 193

Table 8.2. Saturation indices and carbon dioxide equilibrium partial pressures for
an acidic groundwater (well 402), an uncontaminated groundwater (well 504) and a
neutralized contaminated groundwater (well 503) from the Pinal Creek basin alluvial
and basin-fill aquifers. NC: could not be calculated.

Mineral

Well 402
(acidic)
1989/1/12

Well 504
(background)
1991/11/22

Well 503
(neutralized)
1991/11/22

Calcite
Dolomite
Siderite
Rhodochrosite
Gypsum
Fluorite
SiO2 (am)
Chalcedony
Al(OH)3(am)
Gibbsite
Kaolinite
Alunite
Fe(OH)3(am)
Goethite
K-Jarosite
Log pCO (in atm)

5.0
10.2
2.6
3.3
0.0
3.2
0.1
0.8
4.0
1.3
0.7
1.8
1.0
4.9
0.0
1.0

0.5
1.1
11.5
NC
2.6
2.2
0.6
0.3
NC
NC
NC
NC
0.5
6.4
8.5
1.7

1.8
3.9
NC
0.0
0.1
2.5
0.0
0.8
0.8
1.9
7.2
6.6
NC
NC
NC(<3)
0.9

with respect to amorphous Fe(OH)3 but supersaturated with respect to goethite, and highly
undersaturated with respect to all manganese oxides in the thermodynamic database used by
NETPATH: pyrolusite (MnO2), hausmanite (Mn3O4), manganite (MnOOH) and pyrochroite
(Mn(OH)2). These speciation results, based on the relatively high measured Eh (420 mV), are
consistent with the high Fe and Mn contents of the water and the lack of any evidence of
sulfate reduction. Surprisingly, the calculated equilibrium CO2 partial pressure, 101.0 for well
402, is close to that of the neutralized water (well 503).
In comparison to the acidic water from well 402, the partially neutralized water from
well 503 is not as highly undersaturated with respect to calcite and dolomite, and remains
close to saturation with respect to both gypsum and amorphous silica. Unlike its more
acidic precursor, the water is highly supersaturated with respect to kaolinite and is likely
near saturation with respect to some Al(OH)3 phase: it is undersaturated with respect to
amorphous Al(OH)3 but supersaturated with respect to the more stable form of Al(OH)3,
gibbsite. Although most of the dissolved iron has been removed, manganese has increased,
and the water is at saturation with rhodochrosite. The water is still undersaturated with
respect to several manganese oxides (pyrolusite SI: 9.1, hausmanite SI: 13.4, manganite
SI: 4.7, pyrochroite SI: 7.2), although the uncertainty in these saturation indices is
high, given the poor knowledge of manganese oxide thermodynamics and the dependence
of the calcu1ated saturation indices on the measured Eh. Indeed, lack of data on the
vanishingly small dissolved Mn(IV) and Mn(III) concentrations makes any saturation
index calculations for the Mn(IV) and Mn(III) minerals (pyrolusite, hausmanite, manganite) almost meaningless, because the calculations assume that the measured Eh values
are representative of the Mn(IV)/Mn(II) and Mn(III)/Mn(II) aqueous activity ratios.
Finally, the equilibrium CO2 partial pressure (100.9) is close to that of the acidic water
from well 402, and is more than an order of magnitude higher than are expected from
equilibrium with unsaturated zone CO2 partial pressures (Glynn and Busenberg, 1994b).

BUNDSCH_Book.indb 193

11/29/2011 12:31:56 PM

194

Geochemical modeling of groundwater, vadose and geothermal systems

8.4.3

NETPATH inverse modeling: simulation results

The first NETPATH inverse modeling simulation considered the following 11 mass balance
constraints: Cl, Ca, Mg, Na, Al, Si, RS (redox state), Fe, Mn, C, and S. The simulation
also considered 14 phases for possible reactions. In addition, some phases were forced
to be included (NETPATH model results without the phase were excluded from consideration), and some phase reactions were specified either as dissolution only, or as precipitation/
exsolution only. The phases specified are listed below along with any associated constraints:
Calcite (forced inclusion; dissolution only),
Goethite (forced inclusion; precipitation only),
Gypsum (forced inclusion; precipitation only),
Kaolinite (precipitation only),
SiO2,
Dolomite (dissolution only),
MnO2 (dissolution only),
Rhodochrosite (MnCO3),
Anorthite (CaAl2Si2O8; dissolution only),
Gibbsite,
Mn(OH)3 (precipitation only),
O2 gas (dissolution only),
CO2 gas (exsolution only),
Pure Na phase.
The last phase in the list was added to keep track of the Na imbalance. In the first NETPATH
simulation, the mixing fractions of well 402 and well 504 waters were determined through the
chloride concentrations because no Cl phases were specified. The 11 element mass-balance
constraints allow one mixing fraction and 10 phase mole transfers to be calculated. Because
of the many models that include 10 of the 14 possible phases, additional mass-transfer
limitations were necessary to minimize the number of possible models. NETPATH checked
330 models or possible solutions and found 12 that did not violate the limitations placed (i.e.,
whether a phase was forced to be included in all models, and whether it was allowed to dissolve
only, precipitate/exsolve only, or both). Of the 12 models, 6 are given here (Table 8.3). They
adequately represent the range of possible solutions given by the NETPATH code and will be
further discussed. It is noted that linear combinations of models also form possible models.
The phases in the simulation were chosen, based on knowledge of the mineralogy of the
basin-fill and alluvial aquifer materials, and the examination of saturation indices of the well
402 and well 503 water. In addition to selecting plausible phases, the saturation indices were
used to determine which reactions might be thermodynamically feasible (dissolution-only
and precipitation-only constraints). Although gypsum is not present in the uncontaminated
aquifer, the contaminated waters are consistently slightly supersaturated with respect to
gypsum. In fact, samples brought back from the field precipitate gypsum over the course of
several months. Calcite and dolomite are known to be present in the aquifer materials and were
therefore included in the model. Similarly, there is no lack of manganese oxides in the alluvial
materials. Manganese oxides form at the contact between the Mn(II) rich-groundwaters and
oxygenated groundwaters, and are also widely disseminated in the uncontaminated sand and
gravel (Ficklin et al., 1991). Lind and Stollenwerk (1994) conducted an elution experiment
reacting acidic iron- and manganese-rich groundwater from well 101 with alluvial material
from well 601, which is downgradient from the manganese-contaminated groundwaters. Based
on X-ray diffraction results, Lind and Stollenwerk (1994) found that pyrolusite (-MnO2) and
a solid resembling kutnahorite (CaMn(CO3)2) were present before, but not after the elution of
the alluvial materials.
Although goethite was the Fe(III) phase chosen (for precipitation only), choosing any other
Fe(III) oxide would have resulted in the same Fe mass-transfer values. Similarly, we could

BUNDSCH_Book.indb 194

11/29/2011 12:31:56 PM

Integrating field observations and inverse and forward modeling 195

Table 8.3. Models from the first NETPATH simulation. Results in millimoles per kilogram of H2O.
Positive numbers indicate dissolution, negative numbers precipitation or degassing.

Reaction or
addition
Well 504
Pure Na
Dolomite +
Gypsum F
Goethite F
Calcite +F
Anorthite
Kaolinite
Gibbsite
SiO2
Rhodochrosite
MnO2 +
Mn(OH)3
O2 gas +
CO2 gas
Net protons
consumed

Model 6
O2 gas
rhodo.
diss.

Model 1
anorthite
gibbsite

Model 2
gibbsite
SiO2

Model 3
kaolinite
SiO2

Model 4
electron
transfer

Model 5
O2 gas

0.216
0.579
2.899
2.219
8.339
4.929
0.157

0.216
0.579
2.899
2.219
8.339
5.086

0.216
0.579
2.899
2.219
8.339
5.086

0.216
0.579
2.899
2.219
8.339
4.929
0.157

0.216
0.579
2.899
2.219
8.339
4.929
0.157

0.216
0.579
2.899
2.219
8.339
4.929
0.157

0.766

0.766

0.766

0.766

2.972
4.064

0.452
0.314
2.972
4.064

7.036
5.944

1.092

6.033

6.190

6.190

9.005

1.486
9.005

2.032
10.097

5.707

5.707

5.707

5.707

5.707

5.707

0.226
0.766
2.972
4.064

1.092

Notes: In the first column, F indicates a forced phase, + a dissolution only phase, a precipitation only
or an exsolution only phase.

have picked amorphous Al(OH)3 instead of gibbsite. Whereas thermodynamic stability might
distinguish among the potential iron phases, thermodynamic stability is not incorporated
explicitly into NETPATH calculations. In fact, Fe and Al are most likely precipitating as
fairly amorphous precipitates, that may recrystallize to more stable crystalline forms with
time. The WATEQFP speciation results suggest that the waters near well 503 may be precipitating some Al(OH)3 phase. The speciation results also suggest that kaolinite may be forming
near well 402, but probably is not forming quickly near well 503 (as evidenced by the very high
supersaturation with respect to kaolinite). The precipitation of amorphous forms of Al and
Fe(III) minerals upon reaction of the alluvial sediments with acidic waters is also suggested
by the elution experiments of Lind and Stollenwerk (1994) and by the selective extractions
performed by Ficklin et al. (1991) on core materials from wells 107 (acidic), 451 (partially
neutralized) and 505 (neutralized). Ficklin et al. (1991) also report no visible association
between Al and SO4 and argue against the formation of an AlOHSO4 phase. Stollenwerk
and Eychaner (1987) had earlier argued that this phase controlled aluminum concentrations
in the acidic groundwaters. Furthermore, in his column elution studies, Stollenwerk (1994)
found that he could best simulate the behavior of dissolved aluminum by using amorphous
Al(OH)3 as the solubility-limiting phase at pH values above 4.7 and AlOHSO4 at lower pH
values. Stollenwerk, however, changed the solubility product of the AlOHSO4 phase to best
fit his experimental results (from log K = 3.23 to log K = 2.2).
Considering the available evidence, Glynn and Brown (1996) considered that the issue
of AlOHSO4 precipitation was not resolved and required further research. More recently,
Kirk Nordstrom (U.S. Geological Survey, personal communication, 2011), and a review by
Bigham and Nordstrom (2000), have provided good arguments suggesting that an AlOHSO4
phase probably did not occur in groundwaters at the Pinal Creek site. Because the water
from well 402 is close to saturation with kaolinite, and because kaolinite is known to form

BUNDSCH_Book.indb 195

11/29/2011 12:31:56 PM

196

Geochemical modeling of groundwater, vadose and geothermal systems

in acidic waters with high dissolved silica (Blair Jones, U.S. Geological Survey, personal
communication, 1996), Glynn and Brown (1996) preferred the hypothesis of Al control by
kaolinite in the more acidic waters from the site. However, these authors also investigated the
effect of possible AlOHSO4 reactions in their inverse and forward geochemical modeling for
the site. The results were interesting and therefore will be presented here, despite the recent
knowledge arguing further against AlOHSO4 reaction control at the Pinal Creek site.
Because well 503 water is close to saturation with respect to rhodochrosite, we chose this mineral as a possible Mn sink. We believe that reductive dissolution of MnO2 is the primary process
causing dissolved Fe(II) to oxidize and precipitate from solution. The only problem with this
reaction mechanism is that the increase in dissolved Mn(II) is too small relative to the decrease
in Fe(II). Several other possible reactions could explain the missing dissolved manganese:
1. Mn(II) may be precipitating out as rhodochrosite;
2. An electron transfer process may be taking place during which the oxidation state of the
Mn oxide simply decreases while only partially releasing Mn into the solution.
3. Mn(II) may be sorbing onto the freshly precipitated Fe-oxyhydroxides.
4. Oxygen is known to be diffusing through the unsaturated zone into the groundwaters near
the water table. However, because the sampling depths for wells 402 and 503 are substantially
below the water table, the addition of significant amounts of oxygen is considered unlikely.
Although O2 ingassing was considered in the first NETPATH simulation, this reaction will be discarded in the second simulation. For similar reasons, the possibility of CO2
exsolution from a deep flowpath is unlikely. Glynn and Busenberg (1994b) estimated, based
on their measurements of dissolved gases in the Pinal Creek basin groundwaters, that only
groundwaters within 2 m of the water table could be exsolving dissolved gases and CO2. Significant CO2 exsolution would also cause exsolution of other dissolved gases such as N2 and
Ar. For example, exsolution of CH4 and CO2 from an hydrocarbon contaminant plume has
been held responsible for the low dissolved Ar and N2 concentrations measured in groundwaters from the U.S. Geological Survey Bemidji Toxics site (Revesz et al., 1995). Instead,
groundwaters from the Pinal Creek site show high concentrations of both dissolved N2 and
Ar because of the large amounts (often above 20 mL L1) of excess air entrained during
groundwater recharge (Glynn and Busenberg, 1994b).
Interestingly, the results of the first NETPATH simulation suggest that other Ca and
Mg sources (in addition to calcite and dolomite) are needed if CO2 is disallowed as a sink
for the excess carbon provided by the dissolution of the carbonates. We initially thought
that rhodochrosite (MnCO3) would provide an additional carbon sink, but found that,
given the Mn mass-balance constraints, the rhodochrosite sink would not be sufficient to
account for the excess carbon. The presence of another Mn sink (such as Mn(II) sorption)
instead or in addition to rhodochrosite precipitation would only exacerbate this problem.
Therefore, because no other carbon sinks are likely to be present (the waters are undersaturated with respect to siderite), the next approach was to incorporate another Ca source,
specifically a Ca silicate, so as to reduce the amount of carbon coming into solution from
calcite. Although anorthite was chosen, it is likely that any silicate mineral dissolution
accelerated by the acidic groundwaters would also act as a source of Mg, Na, and K (and
probably Fe and Mn) to the solution. However, the dissolution of Ca-rich silicates (and
perhaps Mg-rich silicates) is expected to be faster than that of Na- and K-rich silicates.
On the basis of their observations and elution experiments, Lind and Stollenwerk (1994)
suggest that tremolite (Ca2Mg5Si8O22(OH)2) dissolution may be a source of both Ca and
Mg to the Pinal Creek groundwaters. Indeed, amphiboles, such as tremolite, and pyroxenes
are expected to have faster reaction rates than feldspar minerals, although their abundance
in the alluvial materials is minor compared to that of the feldspar minerals. The presence of CO2 degassing in all the models found by the first NETPATH simulation suggests
that some Mg-silicate phase (such as tremolite) must be included if models without CO2
degassing are to be found.

BUNDSCH_Book.indb 196

11/29/2011 12:31:56 PM

Integrating field observations and inverse and forward modeling 197

The last row in Table 8.3 gives an estimate of the net number of millimoles of protons
consumed by the various reaction models. The number of protons consumed in each reaction
model was calculated by estimating the number of protons consumed by the dissolution of one
millimole of each solid or gaseous phase multiplied by the mole transfer in the reaction model.
The number of protons consumed is dependent on the degree of protonation or hydroxylation
of the various aqueous species produced by the dissolution reactions. For example, a calcite
dissolution reaction will show consumption of two protons per mole of calcite dissolved if the
reaction is written to produce H2CO30 (or equivalently aqueous CO2; henceforth, H2CO30 refers
to the sum of the H2CO30 and the much more dominant aqueous CO2 species), but will show
consumption of only one proton if the reaction is written to produce HCO3. The proton consumption calculations shown here assume that the reaction byproducts are the dominant aqueous species determined from the speciation of the well 402 water, such as: AlF2+, AlF2+, AlSO4+,
Al(SO4)2 and Al+3 for Al; Mn2+and MnSO40 for Mn; Fe2+ and FeSO40 for Fe; H2CO30 for TDIC
(species listed in order of decreasing predominance). Using this assumption, the number of
moles of protons consumed per mole of phase dissolved are: 14 for tremolite; 8 for anorthite;
6 for kaolinite; 4 for dolomite, MnO2 and O2 gas; 3 for goethite (or Fe(OH)3), gibbsite (or
Al(OH)3) and Mn(OH)3; 2 for calcite and rhodochrosite; 1 for AlOHSO4. All other phases
mentioned in Tables 8.3 and 8.4 are assumed not to consume protons upon dissolution. The
consumption of protons by the heterogeneous mass-transfer reactions must necessarily be
matched by an increase in solution pH and also by the release of protons from homogeneous deprotonation reactions (e.g., H2CO30  HCO3 + H+; HSO4  SO42 + H+). Given (i) that
the increase in pH between wells 402 and 503 corresponds to approximately a 0.1 millimole
decrease in H+ concentration, (ii) that the difference in H2CO30 concentrations in well 402 and
503 waters is less than 1 millimolal (and HCO3 is always at least 5 times lower than the H2CO30
concentration), (iii) that the concentration of HSO4 in well 402 water is close to 0.1 millimolal,
and (iv) that there are no other major homogeneous deprotonation reactions, it appears that
the 5.7 millimoles of proton consumption calculated for the various reaction models presented in Table 8.3 are at least 5 times too high. Unaccounted surface deprotonation or proton
exchange reactions offer one possible reason for this discrepancy. Erroneous reaction models
and analytical uncertainty in the basic data collected are other possible reasons.
The most interesting results of the first NETPATH inverse modeling simulation are the
following. Gas exsolution or dissolution were found necessary in all models, even though
anorthite dissolution and rhodochrosite precipitation were included. Of all the models found
by the first NETPATH simulation, we prefer the 3 models that considered MnO2 dissolution and
rhodochrosite precipitation, rather than an electron-transfer mechanism (MnO2  Mn(OH)3)
or O2 ingassing (with or without accompanying rhodochrosite dissolution). Of those three
models, we also prefer the two models (models 1 and 2 in Table 8.3) that did not involve
kaolinite precipitation. Although possible, the very high supersaturation of well 503 water
with respect to kaolinite suggests that the mineral does not undergo fast precipitation, at least
at pH values > 4. Instead, we favor aluminum control by Al(OH)3 precipitation (with possible
recrystallization to gibbsite).
The second and third NETPATH simulations
A second NETPATH simulation used Na as the conservative constituent, instead of
chloride, and resulted only in invalid models that required the dissolution, rather than the
precipitation, of 2.22 millimoles of gypsum per kg of water. Because both the acidic and
neutralized groundwaters at Pinal Creek are supersaturated with respect to gypsum, slightly
but consistently, a model with gypsum dissolution was not plausible. A pure chloride source
(0.484 millimoles) was used in this second simulation. The calculated mixing fraction of
background water from well 504 was 0.347 (instead of 0.216). A third NETPATH simulation
used an intermediate mixing fraction of 0.281 (instead of 0.216 or 0.347) and resulted in
12 models that were similar to those of the first NETPATH simulation, but had different
mass-transfer amounts. Gypsum precipitation was small (0.003 millimoles). All models

BUNDSCH_Book.indb 197

11/29/2011 12:31:56 PM

198

Geochemical modeling of groundwater, vadose and geothermal systems

required CO2 mass transfer, but in somewhat smaller amounts (e.g., 5.4 instead of 6.0
millimoles for Model 1, Table 8.3).
The fourth NETPATH simulation
A fourth NETPATH simulation was used to explore the effects of including tremolite
[Ca2Mg5Si8O22(OH)2], biotite [KMg1.5Fe1.5AlSi3O10(OH)2], forsterite [Mg2SiO4], a pure Mn
sink (to simulate Mn(II) sorption) and AlOHSO4, while excluding some of the reactions
considered unrealistic in the previous simulations, namely CO2 exsolution, O2 dissolution,
and kaolinite precipitation. Forsterite was included for numerical exploration because it is a
pure Mg-silicate with a high Mg/Si ratio. Tremolite and biotite were included because those
minerals are commonly found in the Pinal Creek basin sediments. Although K is present in
biotite, no mass-balance for K was included in the NETPATH simulations because of the
large uncertainties in our estimated K data. The fourth NETPATH simulation resulted in 19
possible models. Of the 19 models, 6 included more than 1 Mg-silicate phase and are not presented here (Table 8.4) for reasons of space and simplicity. Five models included tremolite as
the only Mg-silicate phase, and differed from each other in their treatment of the Mn and Al
mass-balances (Models 7 and 10 through 13 in Table 8.4). Five other similar models included
biotite instead of tremolite (Model 8 in Table 8.4) and 3 remaining models included forsterite
but did not include AlOHSO4.

Table 8.4. Models from the fourth NETPATH simulation. Same phases included as in first three simulations, except for the following changes: (i) tremolite, biotite, forsterite, AlOHSO4, and a pure Mn sink
included as possible phases; (ii) kaolinite and gases excluded.

Model 9
forsterite

Model 12
tremolite
AlOHSO4

0.216
0.579

0.216
0.579

0.216
0.579

2.219
8.339
1.724
5.102

2.219
8.339
1.724
5.102

0.089
8.339
4.696

10.656
14.530

10.656
14.530

0.648
8.339
4.696
0.559
1.571

7.036
5.944

Model 11
tremolite
Mn sink

Reaction or
addition

Model 7
tremolite

Well 504
Pure Na
Dolomite +
Gypsum F
Goethite F
Calcite +F
Anorthite
AlOHSO4
Gibbsite
SiO2
Rhodochrosite
MnO2 +
Mn(OH)3
Mn sink
Tremolite
Biotite
Forsterite

0.216
0.579

0.216
0.579

0.216
0.579

0.216
0.579

2.219
8.339
4.696
2.130

2.219
11.238
6.146
1.840

2.219
8.339
4.696
3.290

4.712
8.587
2.972
4.064

6.066
9.166
4.422
5.514

7.032
7.717
2.972
4.064

Net protons
consumed

Model 8
biotite

Model 13
tremolite
no
anorth.
AlOHSO4

Model 10
tremolite
electron
transfer

0.580

0.580

4.064

5.445
2.972
4.064

2.130
1.678
4.326
2.972
4.064

2.972
0.580

0.580

0.580

5.708

5.711

1.933
1.450
5.711

7.642

5.711

5.711

11.655

Notes: Results in millimoles per kilogram of H2O. Positive numbers indicate dissolution, negative
numbers precipitation or degassing. In the first column, F indicates a forced phase, + a dissolution only
phase, a precipitation/exsolution only phase.

BUNDSCH_Book.indb 198

11/29/2011 12:31:56 PM

Integrating field observations and inverse and forward modeling 199

None of the models found included dolomite, a mineral present in the alluvium as rock
fragments and in dolomite formations in the surrounding hills. Dolomite should certainly be
as reactive with the acidic waters as some of the silicate minerals. The problem appears to be
that dolomite dissolution requires a carbon sink in addition to rhodochrosite. This carbon
sink has not yet been identified. Of all the models presented in Table 8.4, we prefer the models (Models 7 through 9) that considered MnO2 dissolution and rhodochrosite precipitation
and that did not include AlOHSO4. The models with electron transfer combined with any of
the Mg-silicate phases are also plausible. In reality, the reactions occurring and responsible
for the evolution of well 402 water to well 503 water are likely to be some linear combination that may include, but will not be restricted to, the mass-transfer models that we found
using the NETPATH code. Many more models could have been found had we included other
phases (silicates in particular), but their description and classification would not add substantive insight into the predominant reactions occurring in the basin. Table 8.4 also gives the
millimoles of protons consumed for each reaction, using the assumptions discussed earlier
(see discussion of Table 8.3). Once again it appears that the millimoles of proton consumption calculated for the various reaction models presented in Table 8.4 are about 3 to 10 times
too high. Unaccounted surface deprotonation and proton-exchange reactions and uncertainty in the analytical data offer possible reasons for the discrepancy. This problem will be
circumvented in the PHREEQC inverse modeling demonstration discussed later, because
PHREEQC always includes alkalinity mass-balance and charge-balance equations and also
considers possible uncertainties in the analytical data.
Fifth and sixth NETPATH simulations used the phases considered in the fourth simulation, but explored the effects of using either (i) Na (instead of Cl) as the conservative
constituent to determine the mixing fractions of the waters from wells 402 and 504, or
(ii) an average mixing fraction of well 504 water (similar to the third NETPATH simulation). The results of these simulations are discussed in Glynn and Brown (1996), but the
main results of the simulations are that they either produce unrealistic models (e.g., requiring
gypsum dissolution rather than precipitation), or they require a Na sink. Loss of Na to an
ion-exchange mechanism is unlikely given the relatively high and increasing Ca/Na ratio along
the flowpath. It is possible that accelerated weathering of silicate minerals by the acidic waters
may be causing a significant increase in the cation exchange capacity of minerals exposed to
the groundwaters and could thereby be responsible for a net removal of Na (and other cations) from solution (Blair Jones, U.S. Geological Survey, verbal communication). However, it
is more likely that analytical error or unrepresentative concentrations in the background water
chosen to simulate the downgradient dilution process are responsible for the inconsistency in
the Cl and Na dilution results.
Conclusions from the NETPATH inverse modeling simulations
The most important conclusion provided by the NETPATH simulations is that Ca- and
Mg-silicate mineral dissolution must be a significant process. Many researchers at the Pinal
Creek site originally believed that calcite and dolomite dissolution was responsible for the
most of the acid neutralization. However, Glynn (1991b) demonstrated that the increase in
Sr concentrations between wells 51 and 402 must have been caused by silicate mineral dissolutionthe amount of Sr present in limestone and dolomite formations contributing to
the carbonate content of the alluvial materials is too small relative to the amount of Sr that
precipitates as an impurity in gypsum. Sr is a significant impurity in Ca-silicate minerals and
is released to the solution during their dissolution. Similarly, dissolved inorganic carbon 13C
data (Glynn, Busenberg and Brown, unpublished data collected in June 1993) ranges from
9.15 to 12.90 per mil for the acidic or neutralized contaminated groundwaters, and from
10.95 to 14.00 per mil for the uncontaminated waters, suggesting that neutralization of the
acidic plume by silicate minerals must indeed be important. [All 13C values are expressed relative to the Vienna PDB standard.] If the calcite and dolomite 13C values are near 0 per mil,

BUNDSCH_Book.indb 199

11/29/2011 12:31:57 PM

200

Geochemical modeling of groundwater, vadose and geothermal systems

as is reasonable for marine carbonates, closed system dissolution of those carbonates caused
by acid neutralization reactions deep within the aquifer should result in higher 13C values.
8.4.4

Inverse geochemical modeling with PHREEQC

PHREEQC provides additional capabilities for modeling the chemical evolution between
the well 402 and well 503 waters because it considers uncertainties associated with individual
element analyses and also solves alkalinity-balance, water mass-balance, and charge-balance
equations. PHREEQC allows each analytical datum for each aqueous solution to be adjusted
within an uncertainty range that is specified by the user. PHREEQC determines sets of phase
mass transfers, solution mixing fractions, and adjustments to the analytical data that satisfy
mass-balance constraints and are consistent with the specified uncertainties. As an option,
PHREEQC will also determine mass-transfer models (later referred to as minimal models)
that minimize the number of phases involved. Most constraints used by PHREEQC inverse
modeling are automatically specified by providing a list of potentially reactive phases. For
example, if tremolite is identified as a potential reactant, PHREEQC will automatically include
mass-balance constraints on Ca, Mg, and Si (NETPATH does not do this). In addition to the
mass-balance constraints defined by specifying a list of potential reactants, PHREEQC also
lets the user specify additional mass-balance constraints that may be used in determining the
mixing fractions for two or more solutions that mix to form a final solution.
Unlike NETPATH, PHREEQC includes a charge-balance constraint, which specifies that
the sum of the deviations from the analytical data for a given solution must equal the charge
imbalance present in that solution. PHREEQC also uses a water mass-balance constraint to
account for mixing, water derived from mineral reactions, and water evaporation or dilution
(PHREEQC and NETPATH are not limited to groundwater problems). The charge-balance
and water mass-balance constraints used by PHREEQC are equivalent to including a mass
balance on hydrogen or oxygen. During the inverse modeling simulation, PHREEQC will
adjust not only the analytical element concentrations, it will also adjust the pH of the waters.
The adjustment to total dissolved inorganic carbon is constrained to be consistent with the
adjustments to pH and alkalinity. Finally, in addition to including a general electron balance
constraint (such as is done in NETPATH), PHREEQC also includes mole-balance equations
for individual valence states of redox-active elements.
A PHREEQC inverse-modeling simulation was constructed by including all the phases
used in the NETPATH simulations, except for kaolinite and CO2 and O2 gases, which were
excluded. Including biotite [KMgl.5Fel.5AlSi3O10(OH)2] in PHREEQC forced the code to
account for the K mass-balance. However, K is not expected to accumulate in the Pinal Creek
basin groundwaters and should also not be used to determine the mixing fractions of well 402
and well 504 waters. Therefore, a pure potassium-montmorillonite [K0.33Al2.33Si3.67O10(OH)2]
was also included in the simulations with biotite and was only allowed to precipitate. Both
sodium and chloride were specified as mass-balance constraints. Pure Na, pure Cl and
pure Mn sinks were not specified, because PHREEQC requires charge-balanced phases.
Adding charge imbalance would prevent PHREEQC from correctly adjusting the analytical
data within the user-specified uncertainties. A 5% relative uncertainty was chosen for all
elements, except for K (20%) and for Cl for which an uncertainty of 10% was initially
chosen but later was reduced to 5%. The lower uncertainty for Cl did not affect the number
of mass transfer models calculated by PHREEQC, minimal or otherwise, did not significantly
affect the calculated mixing fractions of well 504 water, and did not result in any changes in
the phases included in the mass-transfer models. The uncertainty in the pH of the three
waters from wells 402, 504 and 503 was 0.05 pH units. The minimal option was initially
chosen to reduce the number of possible models to those minimizing the number of phases
involved. Some additional precipitation-only and dissolution-only constraints were added
for the phases. The models shown in Table 8.5 are a representative selection of all the models
found by PHREEQC. Table 8.5 shows most of the tremolite-containing models, but other

BUNDSCH_Book.indb 200

11/29/2011 12:31:57 PM

Integrating field observations and inverse and forward modeling 201

Table 8.5. PHREEQC inverse modeling simulation results. Amounts of mass transfer and net proton
consumption are reported in millimoles per kilogram of H2O. Only mass-transfer sets (models) with the
minimum significant number of phases are shown.

Reaction or
addition
504mf.
504 mf. min.
504 mf. max.
Dolomite +
Gypsum
Goethite
Calcite +
Anorthite +
AlOHSO4
Gibbsite
SiO2
Rhodochrosite
MnO2 +
Mn(OH)3
Tremolite +
Biotite +
K-mont.

Model 1
tremolite
non-min.

Model 2
tremolite

Model 3
tremolite
no calcite

Model 4
tremolite
e transf.

Model 5
tremolite
AlOHSO4

Model 6
tremolite
AlOHSO4
e transf.

Model 7
biotite

0.258
0.258
0.277
0.398

0.258
0.258
0.277

0.258
0.258
0.277
2.191

0.258
0.258
0.277

0.258
0.258
0.264
0.290

0.258
0.258
0.264

0.258
0.258
0.263
2.291

8.292
3.588

8.292
4.383

8.292

8.291
1.626
2.202

8.292
3.817

8.292
3.991

9.074

0.423

0.423

3.165
2.769
3.903

3.629
2.364
4.309
0.811
0.498

2.512
0.423
3.649
2.757
3.903

0.423
3.649
2.757
3.903

5.448
5.803
2.756
3.903

0.500

0.500

0.142

4.827
8.687
6.659
5.512
0.580

2.691

0.440

2.167
3.253
4.294

0.521
1.796

Net protons
consumed

0.279

0.281

0.272

0.266

0.271

0.270

0.298

Alk. change
Difference

0.214
0.493

0.213
0.494

0.213
0.485

0.213
0.479

0.213
0.484

0.213
0.483

0.190
0.487

Abbreviations: 504 mf., well 504 mixing fraction; +, dissolution only; , precipitation only; Alk., alkalinity
change; non-min., not a minimal model; K-mont., K-montmorillonite; e transf., electron transfer.

models found (with biotite/K-montmorillonite, forsterite or tremolite) were essentially variations on the combination of reactions shown in Table 8.5. It should be remembered that,
similar to the NETPATH models, linear combinations of PHREEQC inverse models also
represent possible models. With the exception of Model 14, all models reported in Table 8.5
are minimal models, i.e., had the minimum number of phases.
PHREEQC calculates, as an option, the minimum and maximum mass transfers associated
with any given phase in any given model. These minimum and maximum mass transfers are
constrained by the uncertainty ranges specified for the various element and pH analyses. Only
the minimum and maximum mixing fractions of well 504 water are given in Table 8.5. The
optimal mixing fractions of well 504 water reported by PHREEQC are close to the minimum possible values and do not change significantly between the various models. The mixing
fraction of 0.258 reported in Tab1e 8.5 is between the values of 0.216 and 0.347 determined
by NETPATH respectively assuming conservative dissolved Cl or conservative dissolved Na.
The absence of gypsum precipitation in all of the minimal or non-minimal models is one
of the most important results of the PHREEQC inverse modeling. PHREEQC revealed
that SO4 could be considered a conservative entity given the 5% uncertainty associated
with the analyses. Although most of the reasonable NETPATH mass-transfer models
4

Model 1 can be compared with its equivalent minimal model, Model 2.

BUNDSCH_Book.indb 201

11/29/2011 12:31:57 PM

202

Geochemical modeling of groundwater, vadose and geothermal systems

based on chloride conservation precipitated 2 millimoles of gypsum, this mass-transfer was

insignificant given the 5% relative uncertainty on the SO4 concentrations and the high SO4
concentrations in the well 402 and 503 waters. All of the mass-transfer models reported by
NETPATH included 10 phases. In contrast, given the uncertainties in the analytical data,
inverse modeling with PHREEQC showed that only 7 or 8 phases were required.
The second major conclusion, which confirmed the earlier NETPATH simulation results,
is that Ca- and Mg-silicate phases are needed in addition to calcite and dolomite to explain
the Ca, Mg and C mass balances. Another conclusion, not revealed by the previous NETPATH simulations, was the presence of dolomite dissolution in some of the minimal models.
This conclusion was satisfying, because dolomite is present in the aquifer and should react
in acidic environments. The PHREEQC results presented in Table 8.5 show a net production
of close to 0.3 millimoles of protons. This proton balance was calculated, as previously done
for the NETPATH model results (see discussion of Table 8.3), from the mass transfers in
the PHREEQC models. The net proton consumption calculated for the PHREEQC models
is an order of magnitude smaller than calculated for the NETPATH models. Because of
alkalinity-balance and charge-balance equations are included in PHREEQC calculations, the
net amount of protons released and consumed by heterogeneous and homogeneous reactions
is consistent with the pH values of the initial and final waters, given the uncertainties specified
by the user. The proton consumption can be checked against the change in alkalinity between
the final well 503 water and the mixture of the well 402 and 504 waters. The alkalinity changes
reported in Table 8.5 use the adjusted alkalinities calculated by PHREEQC for each inverse
model. The difference between the net proton consumption and the net alkalinity change is
caused by the lack of accounting for the consumption of dissolved oxygen (6.64 mg L1 in
well 504) in our proton consumption calculations. The difference of 0.49 millimoles is close
to 4 times the difference between the oxygen content of the well 503 water and a mixture of
the well 504 and well 402 waters (4 0.121 = 0.484 millimoles). The reaction of the dissolved
O2 present in well 504 water reduces the amount of MnO2 that undergoes reductive dissolution. The fact that 4 protons are consumed rather than 2 for each mole of oxygen consumed
appears to be an error in the PHREEQC version 1 code (David Parkhurst, personal communication). Because the MnO2 mass transfer is 30 times greater than the oxygen mass-transfer,
this error does not significantly affect our results.
Out of the 6 minimal models presented in Table 8.5, our favorite models are Models 7, 2, 4,
and 3. These models do not involve AlOHSO4 precipitation. As previously mentioned, Glynn
and Browns (1996) view that the evidence for possible Al control by AlOHSO4 precipitation
during the evolution from well 402 water to well 503 water was weak is supported by Bigham
and Nordstroms (2000) review of the environmental conditions needed for AlOHSO4 occurrence. Consequently, our preference for amorphous Al(OH)3 as the controlling Al phase is
stronger than it was in 1996. Tremolite is present in the alluvial materials and its reaction with
acidic water from the Pinal Creek site has been documented by Lind and Stollenwerk (1994).
Nevertheless, tremolite dissolution probably contributes much less Ca than dissolution of
Ca-rich plagioclase feldspars (such as anorthite and labradorite) during the neutralization
of the acidic groundwaters. The accompanying release of Na during feldspar dissolution
could be a problem, however, because the PHREEQC models consider Na as a conservative
constituent, within the uncertainty of the analytical data. Too much Na dissolution would
require a Na sink, which remains elusive. As mentioned earlier, additional cation exchange
capacity resulting from the transformation of the feldspars and other aluminosilicates into
secondary clay minerals is the only plausible Na sink. Sorption on freshly precipitated iron
oxyhydroxides could represent another Na sink, but the affinity of Na for these precipitates
is considerably weaker than the affinity of the other cations in the solution.
Two additional PHREEQC inverse modeling simulations were conducted allowing cation
exchange of Na for Ca in one simulation and exchange of Na for H in another. The results
are discussed in some detail in Glynn and Brown (1996). The cation exchange capacity of
the Pinal Creek alluvial sediments is generally low, probably less than 1 meq/100 g, given

BUNDSCH_Book.indb 202

11/29/2011 12:31:57 PM

Integrating field observations and inverse and forward modeling 203

the coarseness of the sediment and the low organic carbon content (less than 1%). Lacking
further information regarding ion-exchange reactions at the Pinal Creek site, Models 7, 4, 2
and 3 from the earlier PHREEQC inverse modeling simulation, which do not include cation
exchange (Table 8.5), remain our preferred models. Simulating cation exchange in an inverse
geochemical model presumes that the user has knowledge of the thermodynamically preferred directions of exchange. Although we feel that Na replacement for Ca on exchange
sites should generally not occur given the preferential dissolution of Ca relative to Na in
the mineral dissolution reactions, and the greater charge density (and sorption affinity) of
Ca over Na, the direction of exchange for other ion-exchange reactions has much greater
uncertainty. Proton release from exchange sites during the neutralization of well 402 water is
possible. Our inverse geochemical modeling simulations point out the need for further experiments to determine cation exchange capacities and directions of exchange. The ability of
inverse modeling to highlight knowledge gaps is perhaps one of its greatest benefits. As will
be demonstrated in the next section, forward geochemical modeling may also provide insight
into cation exchange reactions.
Although some mass-transfer processes are likely to occur continuously throughout the
flowpath used in inverse modeling, some mass-transfer processes (such as cation exchange
reactions) will affect the groundwater chemistry only in narrow portions of the flow system.
In the case of continuous processes, an overall rate of reaction (expressed for example in
moles per kg of H2O per travel time or per travelled volume of aquifer) may be provided by
the inverse modeling results. In the case of a non-continuous process, however, such rates will
have little meaning. Unfortunately, inverse geochemical modeling cannot provide information on the heterogeneous mass-transfer reactions occurring at specific points along a flowpath, but provides, instead, only the net amounts of mass transfer between an initial and a
final endpoint. Forward geochemical modeling can, however, provide insight on the evolution through time of chemical compositions at specific points along a postulated flowpath.
Field corroboration or confirmation of forward modeling simulations, however, may require
significantly more spatial (and temporal) information than may be available.
8.5

REACTIVE-TRANSPORT MODELING AT THE PINAL CREEK SITE

Inverse modeling is a valuable tool that can be used to gain an improved understanding of the
geochemical processes that occur, or have previously occurred, in an aquifer. By itself however, inverse modeling cannot be used to make predictions on the future chemical evolution
of a groundwater system, or in the case of a contaminated groundwater, on the movement
of contaminants. Forward reactive-transport modeling is needed to make such predictions.
Inverse modeling results, nevertheless, can be used to identify potential reactions that should
be considered by a reactive transport model.
In metal-contaminated acidic groundwaters, such as those present at the Pinal Creek site,
pH and Eh conditions are the primary chemical variables controlling the transport of metals and determining the quality of the groundwaters. The partially-neutralized, Fe(II)-poor
groundwaters (such as the well 503 water used in our inverse modeling exercise) have significantly lower metal concentrations than the more acidic Fe(II)-rich waters (such as the well
402 water). The partially neutralized waters are still contaminated and have high SO4, Ca,
and Mn concentrations that make them unsuitable for most beneficial uses, but nevertheless
offer a significant improvement in water quality. Therefore, the ability to predict the movement and reaction of the low-pH and Fe(II)-rich groundwater zones is desirable.
Inverse modeling can help identify the possible reactions affecting the neutralization and
oxidation of the low-pH and Fe(II)-rich groundwaters. Generally however, the movement of
the low-pH and Fe(II)-rich waters will be controlled by the following factors:
1. the groundwater velocity field,
2. the dilution of the contaminated groundwaters by longitudinal and transverse dispersion,

BUNDSCH_Book.indb 203

11/29/2011 12:31:57 PM

204

Geochemical modeling of groundwater, vadose and geothermal systems

3. the extents of heterogeneous mass-transfer reactions affecting the pH and Fe(II)

concentrations in the groundwaters (causing the low-pH Fe(II)-rich waters to evolve into
higher-pH Fe(II)-poor waters), and
4. the initial concentration and composition of mineral and gas phases contacting the
groundwaters and responsible for their chemical evolution.
The following sections briefly discuss ongoing research efforts aimed at a better understanding of the movement and chemical evolution of acidic, metal-laden groundwaters at
the Pinal Creek site. The research findings from the Pinal Creek site will hopefully provide
information, not only on the processes affecting acidic, metal-laden groundwaters in semiarid alluvial basins, but also on the most efficient techniques to characterize and model the
spread of contaminated waters at sites with similarly sparse spatial information.
The groundwater velocity field
Groundwater velocities can be estimated through the construction of a groundwater
flow model. Calculations using Darcys law on observed heads and estimated hydraulic
conductivities have provided estimates of groundwater velocities in the Pinal Creek basin
(Brown, 1996; Neaville and Brown, 1993). However, USGS efforts (in the mid to late 1990s)
to construct a general flow model for both the alluvial and basin-fill aquifers were not
successful, in part because of a limited number of piezometric observations, but mainly
because the high hydraulic conductivity of the unconsolidated alluvium and its focused
distribution along stream channels resulted in wetting and drying oscillations that caused
numerical convergence problems. However, geochemical tracers (18O, 2H, 13C, Cl, Ar, N2) were
used successfully by Glynn et al. (1999) to understand many characteristics of the origin of
groundwaters in the alluvial and basin fill aquifers, including recharge temperature, recharge
elevation, the amount of excess air entrained during recharge, and the extent of evaporative
losses. Additionally, repeated measurements, from field campaigns in 1991, 1993, 1996, and
1998, of chlorofluorocarbon tracers CFC-11, CFC-12, and CFC-113, and to a lesser extent
of SF6 and 3H/3He) concentrations, provided estimates of travel times for waters in both the
alluvial and basin-fill aquifers. Despite extreme fluctuations in the water table during that
period, the independent travel-time estimates obtained from different tracers were generally consistent; and furthermore, the travel times obtained for the deeper waters in both the
alluvial and basin-fill aquifers were found to be nearly invariant with time (Glynn et al.,
1999). Because of difficulties in the construction of a MODFLOW groundwater-flow model
for the Pinal Creek aquifers, a plan to use all the available hydrologic, chemical and isotopic data to calibrate a three-dimensional reactive-transport model for the basin was never
achieved. Such a model, and its supporting flow model, would have been useful in refining
the understanding the movement of contaminated waters and the impact of anthropogenic
or natural remediation processes at the Pinal Creek site, and should generally be considered
essential in the investigation of highly dynamic contaminated groundwater systems.
Transport processes and contaminant dilution
The dilution of the acidic metal-contaminated groundwaters is certainly one of the
most important processes responsible for the downgradient decrease in dissolved-metal
concentrations in the Pinal Creek basin. Although this dilution process was already evident in the dry to normal recharge years (19841991), further dilution occurred as a result
of the greater than normal recharge events that started in the spring of 1991, continued in
1992, and culminated in a l00-year-magnitude flooding event in spring 1993 (groundwater
levels rose as much as 16 meters). Advanced modeling techniques are not needed to demonstrate dilution as a result of these events: plots of metal concentrations as a function of
chloride clearly illustrate the process (e.g., Fig. 8.3). However, a 2- or 3-dimensional transient transport model would clearly be useful in determining the relative contributions of
transverse dispersion, longitudinal dispersion, flow convergence, and transient high-intensity

BUNDSCH_Book.indb 204

11/29/2011 12:31:57 PM

Integrating field observations and inverse and forward modeling 205

450

120

400
Manganese (mg L1)

Magnesium (mg L1)

100
350
300
250
200
150
Acidic
Neutralized

100
50

60
40
Acidic
Neutralized

20
0

0
0

100

200
300
400
Chloride (mg L1)

500

600

100

200
300
400
Chloride (mg L1)

500

600

220

800

200

700

180

600

160

Copper (mg L1)

Calcium (mg L1)

80

500
400
300
200

Acidic
Neutralized

100

140
120
100
80
60
Acidic
Neutralized

40
20

0
0

100

200
300
400
Chloride (mg L1)

500

600

100

200
300
400
Chloride (mg L1)

500

600

Figure 8.3. Magnesium, manganese, calcium and copper concentrations plotted against chloride
concentration in acidic (upgradient) and neutralized (downgradient) groundwaters from the Pinal Creek
basin.

recharge in the dilution process. Figure 8.3 also shows the relative importance, for different
elements, of dilution and reaction (oxidation, neutralization, precipitation/sorption) processes in controlling groundwater chemical evolution.
8.5.1

Summary of previous reactive-transport modeling

Previous reactive-transport modeling conducted for the Pinal Creek site (see, Glynn and
Brown, 1996, for more information) is briefly discussed here. Brown (1996) used the
PHREEQC (version 1) code to construct a 1-D reactive transport model of the site based on
the groundwater chemical data collected by the USGS between 1984 and 1994. The author
used the 1984 chloride concentrations to back-fit the observed dilution processes as a function
of time and distance along a flowpath that extended from the most acidic waters (well 51) to
the most down-gradient waters (well 702). Remarkably, given the dynamic hydrology of the
site, only a few adjustments were needed to fit the Cl concentration profiles for subsequent
years. Adjustments were made to fit observed concentrations in 1985, 1988 and more importantly in 1993, when the equivalent of a hundred-year magnitude flood occurred in the
valley. The results of the modeling investigation showed that dilution, rather than sorption or
other reaction processes, could account for the decrease of Cu, Zn, Ni, and Co in the acidic
ground-water as water flowed from well 51 to well 451. Generally, the simulated Fe(II) and

BUNDSCH_Book.indb 205

11/29/2011 12:31:57 PM

206

Geochemical modeling of groundwater, vadose and geothermal systems

Mn(II) concentrations did not match the observed concentrations, especially downgradient
from well 451. This mismatch was partly caused by a lack of knowledge at the time regarding
the reactions controlling Mn and Fe; for example, rhodochrosite precipitation was not considered, and the decrease in Mn(II) concentrations was ascribed to oxidation by in gassing
O2 downgradient from well 503. The mismatch was also partly caused by lack of knowledge
on rates of reactions, not only for redox processes, but also for carbonate and silicate mineral
reactions and other reactions.
Glynn and Brown (1996) summarized the results of Glynn et al. (1991), who simulated the
1-dimensional transport and reactions of a highly acidic water sampled in 1987 from well 51
with a mineral assemblage thought to be reasonable for the Pinal Creek alluvial aquifer. One of
the primary purposes of the simulation was to compare the results obtained from two different
codes: PHREEQM and MST1D. Despite significant differences in their algorithms, the two
codes gave almost identical results. The well 51 water used in the simulations was significantly
more contaminated than the well 402 water that was used in the inverse modeling example presented here and in reactive-transport simulations described in this chapter. Relative to the well 51
water, conservative constituents (Cl, SO4, and Mg) in the well 402 water were diluted by a factor
of approximately 2.4; Na decreased only by about 1.7, suggesting the existence of a Na source
between wells 51 and 402; Fe and Al decreased significantly more than could be explained by
dilution (by factors of 4.7 and 13.5 respectively); SiO2, TDIC and Mn did not change significantly, and Ca and Sr concentrations actually increased by 1.15 and 1.79 times, respectively. The
acidity represented by the potential oxidation and precipitation of the dissolved Fe(II), Mn(II)
and Al in the water from well 51 was 1.31 101 moles of protons compared to 2.60 102 for
the well 402 water, or about 5 times greater. Adding the acidity represented by TDIC does not
significantly change the potential acidity of well 51 water, but does increase the potential acidity
of well 402 water to a proton molality of 3.01 102, a value still 4.5 times lower than that of
well 51 water. [The calculation of potential acidity assumes that (i) each mole of Fe2+ oxidation
and precipitation (by reductive dissolution of MnO2, or by reduction of dissolved O2) produces
2 moles of H+, (ii) oxidation of dissolved Mn2+ and precipitation as MnO2 (or precipitation of
MnCO3) produces 2 moles of H+, (iii) precipitation of Al3+ produces 3 moles of protons, and
(iv) each mole of H2CO3 produces 1 mole of H+.] The Glynn et al. (1991) simulations for the
reactive transport evolution of the well 51 highly acidic water provide an interesting comparison for the reactive-transport simulations presented in the next section.
8.5.2

A reactive-transport sensitivity analysis on the movement

of pH and pe-controlling mineral fronts

Some of the reactions identified by inverse modeling of the chemical evolution of groundwaters
between wells 402 and 503 were used in 1-D PHREEQM and PHREEQC reactive-transport
simulations to determine their effect on the movement of the low-pH and high-Fe(II) groundwaters in the Pinal Creek basin. Many results of this study will also be applicable to other sites
where groundwater is contaminated by sulfuric acid and heavy metals. Before discussing the
results of the PHREEQM and PHREEQC transport simulations, however, we will show how
the movement of a single mineral dissolution front, critical in controlling the redox state of the
Pinal Creek groundwaters, can be modeled without the use of a computer code.
8.5.2.1 A simple model for advective transport of a reactive front:
the MnO2 dissolution front
If O2 ingassing is assumed not to affect the concentrations of Fe(II) at significant depths
below the water table (11 m for well 402, 21 m for well 503), the movement of dissolved Fe(II)
will probably be determined by the reduction of manganese oxides, as exemplified by the
following reaction:
2 Fe2+

BUNDSCH_Book.indb 206

M O2, + 4H2O
MnO

2 ((O )3,,am
2Fe(OH)

M 2 + + 2H +
Mn

(8.2)

11/29/2011 12:32:01 PM

Integrating field observations and inverse and forward modeling 207

The free energy change associated with this reaction is highly negative, and the reaction can
be considered irreversible, regardless of which crystal structure is used for MnO2, s (birnessite,
pyrolusite) or for the precipitated iron oxyhydroxide. As a result, if the kinetics of the reductive dissolution reaction are fast relative to the movement of the water (5 m day1), the velocity of the dissolved Fe(II) front, VFe can be related to the velocity, VH O of the water through
2
an apparent retardation factor R,
(8.3)

VFe VH2 O R

where R is related to the amount, Mini, of MnO2, s initially present in the aquifer (expressed in
moles kg1 H2O) and to the amount, M, of MnO2, s dissolved by a unit mass (1 kg of H2O)
of Fe(II)-rich water:
R = 1+

Mini
M

(8.4)

Equation (8.4) can be related to the more general equation describing the retardation of
sharp reaction fronts in systems with advective, but no dispersive, transport (Dria et al., 1987;
similar expressions for the traveling wave approximation can also be found in Lichtner,
1988, 1985; Ortoleva et al., 1986):

k =1 gik (mkD mUk )

K

R = 1+

j =1 hij c Dj cUj
J

(8.5)

where gik and hij are the stoichiometric coefficients of element i in mineral k and aqueous species j. mk and cj are the mineral and aqueous-species concentrations, respectively. Superscripts
D and U indicate downstream and upstream concentrations, respectively.
8.5.2.2 Determination of the initial MnO2,s and carbonate mineral concentrations
Although inverse modeling can be used to determine the mass-transfer amounts (M) for
various heterogeneous reactions, inverse modeling cannot usually reveal the initial amounts
(Mini) of minerals present in an aquifer. Estimates of average initial mineral contents must be
made on the basis of: batch or column experiments on unaffected aquifer materials, selective
mineral extraction analyses, X-ray evidence, or observation of retardation of reactive fronts
in the field.
By determining the net mineral mass transfer (M) experienced by a packet of water
between two points along a flowpath, an inverse model may be used to set a lower bound on
the initial concentration of that mineral (Mini M), but only if the mineral mass transfer
occurs in a unique and localized part of the flowpath, such as a single reaction front. Indeed,
although expressed in mol kg1 of H2O, the mineral mass transfer determined by an inverse
model integrates reactions occurring in the volume of aquifer traversed by a packet of water.
In the case of a slow mass-transfer reaction (relative to the movement of the water), the
value of M determined by an inverse model will change with increasing flowpath length
until some equilibrium or steady state is reached. The initial mineral concentration, Mini, in
equation (8.4) is also expressed in terms of mol kg1 H2O, but refers to a static mass of water
and therefore a localized aquifer volume. In contrast, the value of M determined by inverse
modeling has a Lagrangian frame of reference and its units refer to a dynamic mass of H2O
that has travelled along a specific flowpath length. In the case where the entire mass transfer
occurs at a sharp reaction front, the difference in units may be moot, at least if the mineral concentration Mini was initially uniform between the initial and final end points of the

BUNDSCH_Book.indb 207

11/29/2011 12:32:02 PM

208

Geochemical modeling of groundwater, vadose and geothermal systems

flowpath. In any case, equations (8.4) and (8.5) will be of limited use for a slow reaction, one
that does not result in the development of a sharp reaction front.5
Glynn and Brown (1996) used a value of 2 102 mol MnO2 per kg H2O in their PHREEQC
and PHREEQM transport simulations, primarily to stay consistent with the simulations
conducted by Glynn et al. (1991). A higher value may be more appropriate. Indeed, on the basis
of his column elution experiments, Stollenwerk (1994) suggested a value of 7.1 millimoles of
MnO2 per kg of sediment. Depending on values of porosity and bulk density assumed, this
solid concentration is equivalent to a value between 3.2 102 (our estimate) and 4.49 102
mol kg1 of H2O (estimate based on Brown, 1996). Based on the selective extraction results
of Ficklin et al. (1991), Brown (1996) used a value of 7.9 102 mol kg1 H2O in his 1-D
simulation. The maximum and minimum amounts of MnO2, dissolved between wells 402
and 503 in inverse modeling simulations were 9.2 103 (Table 6 in Glynn and Brown, 1996)
and 3.9 103 mol kg1 H2O (Table 8.5), respectively. These values may be considered lower
bounds for the initial amount of MnO2 present in the aquifer, if it is assumed that the reductive dissolution of MnO2 is fast relative to the movement of the water; this assumption is
used in all the PHREEQM and PHREEQC forward simulations presented in this paper. The
lack of accurate groundwater flow and transport models, the large distances between the well
sites, and the suggestion that the reductive dissolution of MnO2 may be slow compared to the
groundwater velocities make it difficult to calculate accurately the Fe(II) retardation factor,
and therefore the appropriate initial MnO2 content. In any case, the results of our simulations, and in particular the MnO2 dissolution front retardation factors that we determine, can
easily be extrapolated to other initial MnO2 concentrations.
Determining appropriate initial carbonate (calcite and dolomite) concentrations for the
reactive-transport simulations was also a problem. Brown (1996) used a concentration of
0.18 mol kg1 H2O in his simulation, a value consistent with the carbonate content determined by Eychaner and Stollenwerk (1985) for a sample of alluvium collected in 1985. Brown
(1996) noted that the buffering capacity measurements conducted by Hydro Geo Chem, Inc.
(1989) on alluvial samples collected from three different locations in the Pinal Creek basin
could be translated into equivalent carbonate concentrations of 0.12 mol L1 for sand and
gravel and 0.76 mol L1 for calcareous clay. These measurements, however, did not correct for
possible proton adsorption and silicate dissolution reactions. Based on a description of the
Pinal Creek site by Eychaner (1989, and pers. comm.), Glynn et al. (1991) used an initial carbonate concentration of 0.084 mol kg1 H2O (0.042 mol kg1 H2O calcite and 0.021 mol kg1
H2O dolomite). The resulting pH front retardation factor of 5, for the acidic well 51 water
used in the simulation, approximately matched the relative rate of advance of the low-pH
waters at the Pinal Creek site over the last 50 years. The inverse modeling results discussed
in the present paper generally show a net dissolution between wells 402 and 503 of 1.5 103
to 4.6 103 mol kg1 H2O of primary carbonate minerals (Table 8.5). If the questionable
assumption is made that the carbonate mineral dissolution rates are relatively fast, resulting in localized dissolution, a value of 5 103 mol kg1 H2O may be considered a reasonable lower bound on the initial carbonate concentration representative for the Pinal Creek
alluvial sediments. In any case, because the initial carbonate mineral concentration chosen
for the simulations was the most important adjustable parameter determining movement of
the low-pH waters in our simulations (and is also the most important factor in the chemical
evolution of Pinal Creek groundwaters), a set of 8 different initial carbonate concentrations
ranging from 5.25 103 to 3.32 101 mol kg1 H2O were used to test each reaction-model
modification.

Lichtner (1988) and Ortoleva et al. (1986) have shown that reaction kinetics will not affect the rate
of front propagation given enough time and distance. The front may not be as sharp but it will still
propagate at the same rate as a front resulting from a simulation that uses the Local Equilibrium
Assumption.

BUNDSCH_Book.indb 208

11/29/2011 12:32:04 PM

Integrating field observations and inverse and forward modeling 209

8.5.2.3 Setup of the 1-D reactive-transport simulations
Unless specified otherwise, most of our simulations were conducted with the local equilibrium advective transport code PHREEQC. Dispersion was not usually simulated. The
reactive-transport simulations investigated the effect of the following model variations on the
retardation of the low-pH and high Fe(II) fronts:
1. Changing solid-carbonate concentrations. The initial concentrations chosen were:
5.25 103, 1.05 102, 2.1 102, 3.05 102, 4.2 102, 8.4 102, 1.68 101, and
3.32 101 mol kg1 H2O.
2. Using a longitudinal dispersivity of 560 m, or no dispersivity. The dispersivity of 560 m
represented 10% of the total simulation length, a rule of thumb often applied in groundwater transport modeling. Most simulations, however, used a dispersivity of 0 m, primarily to limit execution times. The PHREEQM and PHREEQC v. 2 (1996 beta version)
geochemical transport codes were used for all simulations with a dispersivity of 560 m.
3. Including or excluding dolomite. Most simulations did not include dolomite.
4. Including or excluding rhodochrosite precipitation.
5. Including or excluding Al(OH)3, kaolinite, or AlOHSO4 precipitation. Two different solubility products were used for AlOHSO4 (103.2 and 102.2). The higher solubility product
is the value adopted by Stollenwerk (1994) in fitting the results of laboratory column
experiments that reacted Pinal Creek acidic groundwater with Pinal Creek sediments.
6. Including or excluding equilibrium with an infinite reservoir of CO2 at partial pressures
of either 100.9865, a value based on unsaturated zone CO2 gas measurements at the Pinal
Creek site (Glynn and Busenberg, 1994b), or 101.33, the value used in Brown (1996). This
latter value was based on the dissolved CO2 concentration at well 503 (November 1991)
reported by Glynn and Busenberg (1994a).
7. Including or excluding cation exchange. Two different cation exchange capacities, 1 meq/
l00 g and 10 meq/100 g, were tested.
8. Including or excluding diffuse double-layer surface-complexation sorption, which was
based on thermodynamic data compiled in Dzombak and Morel (1990) for sorption onto
hydrous ferric oxide.
9. Including or excluding the irreversible dissolution of Ca and Mg silicates to match the
amounts calculated by two of our inverse modeling simulations (PHREEQC Models 2
and 7 in Table 8.5). These simulations assumed a zero-order kinetic dissolution process for the silicate minerals with an inexhaustible supply of silicate minerals. Two of the
simulations were also conducted assuming a zero-order kinetic dissolution process for
MnO2. In all cases, the zero-order kinetic dissolution processes were specified so that the
acidic water would receive, during the course of its evolution through the transport column, exactly the silicate mineral mass transfers (and the MnO2 mass-transfer when applicable) determined by inverse Models 2 and 7 in Table 8.5. The dissolution/precipitation of
all other minerals was allowed to proceed to thermodynamic equilibrium at each time step
and in each cell.
More than 160 reactive-transport simulations were conducted. All simulations used the
water from well 402 (89/1/12) as the infilling solution. The water from well 504 (91/11/22) was
used as the background water initially present in the 1-dimensional column. The 5.6 km-long
column was subdivided into 10 cells of equal length and with initially homogenous mineral,
surface, and aqueous concentrations. (Because mineral concentrations in PHREEQC and
PHREEQM are expressed in terms of mol kg1 H2O, the porosity and bulk density of the
sediments are not defined for the programs). A time step of 112 days was used, thereby
simulating an average linear groundwater velocity of 5 m day1 (representative of the average
groundwater velocity between wells 402 and 503). Up to a maximum of 5000 time steps
(1534 years) were simulated.
Numerical dispersion affected the transport of non-conservative constituents in the
simulations, and depended on the cell-length discretization, x, and on the extent of

BUNDSCH_Book.indb 209

11/29/2011 12:32:04 PM

210

Geochemical modeling of groundwater, vadose and geothermal systems

retardation, R. The corresponding numerical dispersivity, num can be calculated as follows

(Herzer and Kinzelbach, 1989):

num =

x vt

2
2R

(8.6)

or, in the case of our simulations with uniform time and space discretization:

num =

x
1
1
2 R

(8.7)

The maximum numerical dispersivity is half the cell length or 280 m. Retardation factors
of 1, 2, 5 and 10 result in numerical dispersivities of 0, 140, 224 and 252 m, respectively. Ideally, the simulations would have been run with a greater number of cells (and consequently
a greater number of time steps and a smaller cell length). Execution times and the number
of simulation runs, however, made this impractical. In any case, the effects of numerical
dispersion on calculated front retardation factors was small (less than 5%), and decreased
as the retardation (and the number of pore volumes needed to observe that retardation)
increased: the uncertainty in the exact position of a sharp front within the last cell of a
column matters less and less in calculating the retardation factor for a moving front as the
number of pore volumes needed to observe the passing of the front increases. The effects
of numerical dispersion were also tested in the simulations that assigned a longitudinal dispersivity of 560 m to simulate hydrodynamic dispersion. Hydrodynamic dispersion affects
the transport of all constituents, conservative or non-conservative (and therefore retarded).
Numerical dispersion has similar characteristics to hydrodynamic dispersion but, in the case
of the 1D PHREEQC and PHREEQM transport simulations, it only affects the transport
of non-conservative constituents. The results of these simulations are discussed later in this
chapter, but support our conclusions regarding the minor effects of numerical dispersion in
the Pinal Creek analysis.
Amorphous Fe(OH)3, and one aluminum phase (amorphous Al(OH)3, kaolinite, or
AlOHSO4) were allowed to precipitate in all the simulations. Gypsum was allowed to
precipitate in all the simulations except when irreversible zero-order kinetic dissolution
of Ca and Mg-silicate minerals was included (following Models 2 and 7 in Table 8.5).
Rhodochrosite was allowed to precipitate in all sets of simulations, except one. An
essentially infinite amount of chalcedony was present in all cells. An initial concentration of
2 102 mol kg1 H2O of MnO2 was specified in all the simulations (except for the simulations that allowed irreversible dissolution of Ca and Mg silicates). Differing initial concentrations of carbonate minerals (calcite and/or dolomite) were specified in all simulations, but
dolomite was included together with calcite in only a few of those simulations. Dolomite,
however, was the only carbonate mineral present in simulations that attempted to emulate
Model 7 in Table 8.5.
A subset of the simulations presented here are referred to as the Basic Reaction Model
(BRM). The BRM allows one of four possible Al-bearing phases to precipitate, along with
secondary gypsum and amorphous Fe(OH)3 precipitation. The BRM specifies an essentially
infinite amount of chalcedony, an initial MnO2 concentration of 2 102 mol kg1 H2O,
and one of eight fixed initial concentrations of carbonate minerals. Additionally the BRM
assumes local equilibrium, no sorption or cation exchange reactions, no CO2 gas exchange,
and zero longitudinal dispersivity.
Graphical presentation of the complete suite of results is limited to a few representative
pH-breakthrough curves shown in Figures 8.4 and 8.5 for the midpoint of the last cell (cell
10). The fronts shown in Figures 8.4 and 8.5 are generally sharp and correspond to spatial
transitions between the absence and presence of given minerals. The essential findings of the
study are summarized in the sections below.

BUNDSCH_Book.indb 210

11/29/2011 12:32:04 PM

Integrating field observations and inverse and forward modeling 211

Figure 8.4. pH breakthrough curves for four PHREEQC advection-reaction simulations that use the
Basic Reaction Model (BRM) with amorphous Al(OH)3 as the Al-bearing phase allowed to precipitate.
No dolomite was included. Initial calcite concentrations for each of the simulations shown are: a) 2.1 102
mol kg1 H2O, b) 3.0 102 mol kg1 H2O, c) 4.2 102 mol kg1 H2O, d) 8.4 102 mol kg1 H2O.

Figure 8.5. pH breakthrough curves for two PHREEQC advection-reaction simulations including
tremolite and excluding gypsum as determined by inverse Model 2 in Table 8.5. The irreversible dissolution of tremolite (5.00 105 moles per cell per time step) was simulated as a continuous zero-order reaction process matching over the length of the simulation column the amount of tremolite mass-transfer
specified in inverse Model 2. The two simulations shown differed in their consideration of MnO2: (a) an
initial concentration of MnO2 of 2 102 mol kg1 H2O was specified and allowed to react to equilibrium
at each time step; (b) MnO2 was added as a continuous irreversible dissolution process at the rate of
3.903 104 moles per cell per time step, so as to match over the length of the simulation column the
amount of MnO2 mass transfer specified in inverse Model 2.

Simulation results: movement of the Fe(II)-rich waters and of the MnO2

dissolution front
As can be seen from Figures 8.4 and 8.5, the presence, extent, and movement of pH fronts in
the simulations conducted is dependent on a number of reaction processes, including redox
processes and the presence or complete dissolution of primary MnO2 and secondary rhodochrosite. The movement of the redox front is significantly easier to understand. Indeed, the

8.5.2.4

BUNDSCH_Book.indb 211

11/29/2011 12:32:06 PM

212

Geochemical modeling of groundwater, vadose and geothermal systems

movement of the low-pe Fe(II)-rich groundwater zone was primarily dependent on the initial
amount of MnO2 specified for each simulation and was usually little affected by any other
factors. A retardation factor (R) of 4.74 was determined for the BRM simulations containing
2 102 mol MnO2 per kg of H2O. As mentioned earlier, the movement of the MnO2 dissolution front, and of the attendant Fe(II) and low-pe fronts, can easily be calculated using equation
(8.4). Indeed, given the Fe(II) concentration of the infilling solution (591 mg L1, Table 8.1 or
10.58 millimolar) and the stoichiometry (2 moles of Fe(II) to reduce 1 mole of MnO2), M is
5.29 103 mol L1 and R is 4.78. The slight difference between this calculated value of R and
the value of 4.74 determined from the PHREEQM simulations results from the use of molar
concentration units for M in our calculation instead of the molal units used in PHREEQM.
Apart from the initial concentration of MnO2 specified, the retardation of the MnO2 dissolution front is only affected by processes that affect the concentration of the reductant [dissolved
Fe(II)] in the infilling water: namely, dispersion and cation exchange in our simulations. Indeed,
the simulations that included dispersion showed a slight decrease in the retardation factor of
the MnO2 dissolution front (an R value of 4.6 for an MnO2 concentration of 2 102); the simulations that included cation exchange, however, resulted in substantially greater MnO2 retardation factors: 5.2 for the simulations with a cation-exchange capacity (CEC) of 1 meq/100 g and
9.2 for the simulations with CEC of 10 meq/100 g. (The lower CEC is more realistic for the
Pinal Creek alluvial sediments). The increase in retardation factor is due to cation-exchange
reactions that decrease the dissolved Fe(II) concentrations by, for example, Fe/Ca exchange:
Fe2 + + Ca
C

 Ca
C 2 + + FeX 2

(8.8)

Because Fe(II) surface complexation on hydrous ferric oxide was not simulated, the simulations that included surface-complexation sorption reactions did not result in increased
retardation factors for the MnO2 dissolution front.
8.5.2.5 Simulation results: evolution of the low-pH waters
Unlike the movement of the low-pe Fe(II)-rich front, the movement of the low-pH groundwaters is much more difficult to predict and it is not controlled by a unique mineral dissolution/precipitation reaction. Indeed, the PHREEQM and PHREEQC simulation results show
the development of several step changes in pH that are controlled by mineral dissolution and
precipitation fronts. The consumption of the initial calcite (and dolomite, if present) in the
column is typically followed by the complete dissolution of secondary rhodochrosite and,
usually much later (if at all) by the dissolution of the secondary Al-bearing phase included in
the simulation (Al(OH)3, AlOHSO4 or kaolinite). The complete dissolution of these minerals
and the loss of their proton-consuming capacity leads to abrupt pH decreases. In contrast,
the disappearance of MnO2 and the cessation of Fe(II) oxidation and Fe(OH)3 precipitation
results in a sharp pH increase.
The pH breakthrough curves shown in Figure 8.4 for the last cell in the column are typical of the pH breakthrough curves for simulations without silicate dissolution: pH remains
high at 7.34, until the first pore volume (at shift 10) flushes through the column, and neutralized infilling water with a pH of about 6.5 reaches the last cell. If MnO2 is the first mineral
consumed and Al(OH)3 is allowed to precipitate, the pH will increase from 6.5 to 6.8 as the
MnO2 is completely consumed. Later, the complete consumption of calcite will result in a
pH decrease from about 6.8 to about 5.6, and the ensuing consumption of rhodochrosite will
result in another pH drop to near 5.3. Finally, the disappearance of the secondary aluminum
phase will result in yet another pH drop to about 4.06 (given a sufficiently long simulation
time), which is close to the initial infilling water pH of 4.13. Although the final pH values in
simulations with silicate mineral dissolution differ somewhat from the pH values mentioned
above, the general pattern of pH evolution in those simulations is also strongly dependent on
the disappearance of calcite (or calcite and dolomite), MnO2, secondary rhodochrosite, and
the secondary aluminum phase (Fig. 8.5).

BUNDSCH_Book.indb 212

11/29/2011 12:32:07 PM

Integrating field observations and inverse and forward modeling 213

Table 8.6. pH values computed with the PHREEQC geochemical code (version 1; Parkhurst, 1995) by
equilibrating well 402 water with various mineral phases, including chalcedony, gypsum and amorphous
Fe(OH)3 and one of four possible Al bearing phases (cf. table note). The last pH column indicates the
pH value in the absence of any aluminum phase.
Presence of additional minerals?

Computed equilibrium pH as a function of the presence

of an Al phase

MnO2

Calcite

Rhodocrosite

Yes
Yes
Yes
No
No
No

Yes
No
No
Yes
No
No

Yes
Yes
No
Yes
Yes
No

6.453
5.174
5.011
6.788
5.633
5.330

6.442
5.071
3.810
6.772
5.576
4.190

6.462
5.095
3.364
6.748
5.666
4.554

6.467
5.090
3.186
6.818
5.659
3.946

no Al phase

3.175
4.056

Note: The identity of the Al bearing phase is indexed as follows: (1) Al(OH)3, (2) kaolinite, (3) AlOHSO4
with log Ksp = 2.2, (4) AlOHSO4 with log Ksp = 3.23).

Many of the pH values that are plateaus in Figures 8.4 and 8.5 can be simulated by equilibrating the infilling solution (well 402 water) with various assemblages of coexisting mineral phases found during the transport simulations. pH values resulting from PHREEQC
equilibrations of well 402 water with several mineral assemblages are given in Table 8.6. The
resulting pH values usually closely agree with the pH values obtained during the transport
simulations (and are even within half a pH unit of the values reached in the simulations
that incorporated CO2 exsolution and surface-complexation). The PHREEQC equilibration
results clearly indicate that, although the loss of calcite results in a pH decrease to the range
of 5.1 to 5.7, the subsequent loss of rhodochrosite results in a further pH decrease to values
well below 5 if Al(OH)3 is not present. Table 8.6 indicates that the lowest pH value reached
during a simulation, after the complete dissolution of rhodochrosite, depends on the aluminum phase included in the simulation. Invariably, simulations that allowed amorphous
Al(OH)3 precipitation resulted in higher pH values than those simulations that allowed kaolinite, or especially, AlOHSO4 precipitation. In the presence of rhodochrosite (and possibly
calcite), however, the choice of aluminum phase allowed to precipitate had little effect on
the resulting pH values. Finally, Table 8.6 confirms the finding that the lowest pH values
(near 3.1) were invariably attained when MnO2 was still present but all other carbonate and
Al-phases were absent.
8.5.2.6 The effect of the initial carbonate to initial MnO2 ratio on the evolution
of the low-pH waters
The initial carbonate mineral concentration specified in each simulation was the most important control on the movement of the calcite, rhodochrosite, and Al-phase dissolution fronts,
and of the associated pH fronts (Tables 8.7, 8.8 and 8.9). The initial carbonate to MnO2 ratio
(CMR) was the second most important factor affecting the retardation of dissolution fronts
for rhodochrosite and the Al-bearing phase (the rhodochrosite dissolution front is important because of the decrease in pH associated with the front, particularly in the absence of
an Al(OH)3 phase). The set of simulations that allowed irreversible dissolution of Ca- and
Mg-silicate minerals gave very different results from the set of simulations that did not allow
Ca- and Mg-silicate dissolution. Unless mentioned otherwise, the following discussions apply
only to the set of simulations that did not allow silicate mineral dissolution, although many of
our conclusions will also be relevant to the simulations with Ca- and Mg-silicate dissolution.
The CMR ratio specified in a given simulation determines the timing of the complete
dissolution of MnO2 in relation to the complete dissolution of the carbonate phases. The
lowest pH values in the transport simulations usually occurred when MnO2 remained in a

BUNDSCH_Book.indb 213

11/29/2011 12:32:08 PM

214

Geochemical modeling of groundwater, vadose and geothermal systems

Table 8.7. Rhodochrosite dissolution-front retardation factors for different reaction models and
initial carbonate concentrations (in mol kg1 H2O). All models except one used a dispersivity of 0 meters.
All models allowed precipitation of rhodochrosite and of one of four Al-bearing phases. The specified
reaction models have an index that refers to the allowed Al-bearing phase. These indices and the
meanings of the various reaction models are discussed in the table notes and in the text.
Initial carbonate mineral concentrations ( 102)
Reaction model

0.525

1.05

2.1

3.0

4.2

8.4

16.8

33.2

Basic Reaction Model, (1)

Dispersivity, (1), PHREEQM
Dispersivity, (1), PHREEQC
Low CEC, (1)
High CEC, (1)
CO2, log pCO = 0.9865, (1)
2
CO2, log pCO = 1.33, (1)
2
Basic Reaction Model, (3)
Basic Reaction Model, (4)
Low CEC, (3)
High CEC, (3)
Surf. Comp., (3), no Cu, Zn, Co, Ni
Surf. Comp., (3), with Cu, Zn, Co, Ni

1.6
1.5
1.5
1.6
0.0
1.4
1.5
1.6
1.6
1.6
0.0

3.7
3.7
3.6
3.8
5.1
3.5
3.7
3.6
3.6
3.7
5.0
4.7
4.6

5.0
4.5
4.5
5.1
6.5
5.0
5.9
4.9
4.8
5.0
6.4
12.3
7.8

12.6
12.7
12.5
11.8
8.8
12.2
15.2
14.7
14.5
13.6
8.6
22.5
19.8

34.1
35.1

65.6
67.7

128.9
133.0

2.7

2.3
2.3
2.2
2.3
3.5
2.1
2.2
2.3
2.3
2.3
3.5
3.4
3.4

33.7
31.3
32.5
37.5
42.3
41.3
41.8
38.0
47.4
46.4

65.4
63.0
56.5
68.8
81.4
80.9
80.9
77.2
86.8
79.5

128.6
126.3
104.5
131.4
159.4
160.2
158.9
155.4
165.4
140.8

RA
7.3
RA
12.1
RA

RA
10.9
RA
17.9
RA

RA
18.2
RA
29.8
RA

RA
24.3
RA
40.1
RA

RA
32.6
RA
53.6
RA

RA
61.6
RA
101.7
RA

RA
119.4
RA
197.5
RA

RA
235.3
RA
388.9
RA

Tremolite, Model 2, (1)

Tremolite, MnO2, Model 2, (1)
Biotite/K-mont/An, Model 7, (1)
Biotite/K-mont/An, MnO2, Model 7, (1)
Biotite/K-mont/An, Model 7, (3)

Note: The identity of the Al bearing phase is indexed as follows: (1) Al(OH)3, (2) kaolinite, (3) AlOHSO4
with log Ksp = 2.2, (4) AlOHSO4 with log Ksp = 3 .23).
Reaction model abbreviations and meanings:
RA:
rhodochrosite accumulates.
Dispersivity:
basic reaction model with dispersivity of 560 m;
column open to a specified fixed pCO ;
CO2:
2

Low CEC:
High CEC:
Surf. Comp.:
Tremolite, Model 2:
Tremolite, MnO2, Model 2:
Biotite/K-mont/An, Model 7:
Biotite/K-mont/An, MnO2, Model 7:

low cation exchange capacity (1 meq/100 g);

high cation exchange capacity (10 meq/100 g);
surface complexation model;
irreversible tremolite dissolution (Model 2 in Table 8.5);
as above but with zero-order MnO2 dissolution;
irreversible biotite and anorthite dissolution and
K-montmorillonite precipitation (Model 7 in Table 8.5);
same as above but with zero-order MnO2 dissolution.

cell, but all the initial and secondary carbonate (rhodochrosite) phases had completely dissolved. (If additionally the Al-bearing phase had been completely dissolved, the pH was
even lower). This transient remnant of MnO2 (third row in Table 8.6) occurred in all simulations with a CMR ratio less than 1.5 (5.25 103, 1.05 102 and 2.1 102 initial carbonate
concentrations and 2 102 initial MnO2 concentration). Simulations with CMR less than
1.5 resulted in lower pH values and more rapid dissolution of carbonates, and therefore, in
relatively lower retardation factors for the rhodochrosite and Al-phase dissolution fronts
(Tables 8.8 and 8.9; Fig. 8.4a).
In simulations with CMR ratios near 1.5, the complete dissolution of MnO2 and rhodochrosite occurred nearly simultaneously, but was preceded by the complete dissolution of
calcite (Fig. 8.4b). For most of these simulations, the rhodochrosite dissolution front had a
retardation factor (R) close to 4.7, the same as the MnO2 front. Disregarding the Ca and Mg

BUNDSCH_Book.indb 214

11/29/2011 12:32:08 PM

Integrating field observations and inverse and forward modeling 215

Table 8.8. Calcite dissolution-front retardation factors. See Table 8.7 caption and notes for explanation of headings and abbreviations. One additional set of simulations reported here used a CEC of
1 meq/100 g but no rhodochrosite was allowed to precipitate (low CEC, no rhodo).
Initial carbonate mineral concentrations ( 102)
Reaction model

0.525

1.05

2.1

3.0

4.2

8.4

16.8

33.2

Basic reaction model, (1)

Dispersivity, (1), PHREEQM
Dispersivity, (1), PHREEQC
Low CEC, (1)
High CEC, (1)
CO2, log pCO2 = 0.9865, (1)
CO2, log pCO2 = 1.33, (1)
Low CEC, no rhodo., (1)
Basic reaction model, (3)
Basic reaction model, (4)
Low CEC, (3)
High CEC, (3)
Surf. Comp., (3), no Cu, Zn, Co, Ni
Surf. Comp., (3), with Cu, Zn, Co, Ni

1.2
1.0
1.0
1.5
2.5
1.1
1.2
1.5
1.2
1.3
1.5
2.5

2.2
2.0
2.0
2.5
4.9
2.1
2.2
3.0
2.2
2.2
2.6
4.9
2.7
2.6

2.7
2.5
2.5
3.1
5.9
2.6
2.8
3.9
2.7
2.7
3.1
5.9
3.2
3.1

3.4
3.3
3.3
3.8
6.9
3.3
3.6
5.0
3.4
3.5
3.8
7.0
3.9
3.8

8.0
8.0

22.7
23.0

52.0
53.0

1.7

1.5
1.3
1.3
1.9
3.5
1.4
1.5
2.1
1.6
1.6
1.9
3.5
2.1
2.0

8.2
10.7
7.3
8.8
13.8
8.7
8.9
8.9
11.4
10.2
8.9

22.9
25.4
18.4
23.0
32.4
24.7
25.6
25.0
27.4
26.3
22.2

52.3
54.8
40.8
51.2
69.5
56.7
59.1
57.0
59.5
58.4
48.2

Tremolite, Mod. 2, (1)

Tremolite, MnO2, Model 2, (1)
Biotite/K-mont/An, Model 7, (1)
Biotite/K-mont/An, MnO2, Model 7, (1)
Biotite/K-mont/An, Model 7, (3)

1.5
1.6
DA
1.6*
DA

1.9
2.2
DA
2.2*
DA

2.8
3.3
DA
3.2*
DA

3.5
4.3
DA
4.2*
DA

4.5
5.5
DA
5.4*
DA

113.7
10.0
DA
9.7*
DA

353.5
18.9
DA
18.4*
DA

>500
36.8
DA
35.7*
DA

All simulations referring to Model 7 in Table 8.5 have initial dolomite instead of calcite. DA: no calcite
specified and dolomite accumulates.

silicate dissolution simulations, the highest rhodochrosite R values were obtained for simulations that specified a high ion-exchange capacity (CEC), a surface-complexation model, or
equilibrium with a low fixed pCO (101.33). In the high CEC simulations, however, the retar2
dation factor of the rhodochrosite front was lower than that of the MnO2 dissolution front
(R = 9.2), i.e., the rhodochrosite dissolution front preceded the MnO2 dissolution front.
Finally, in the simulations with a CMR ratio greater than 1.5 (4.2 102, 8.4 102,
1.68 101 and 3.32 101 initial carbonate concentrations and 2 102 initial MnO2 concentration), complete dissolution of MnO2 generally occurred before the complete dissolution
of secondary rhodochrosite (with the exception of the two high CEC simulations). In simulations with an initial carbonate concentration of 8.4 102 or greater, exhaustion of the initial
MnO2 also occurred before the complete dissolution of the primary calcite (and/or dolomite).
The faster movement of the MnO2 dissolution front relative to the carbonate mineral dissolution fronts resulted in higher pH waters contacting the carbonate minerals and Al-phase
minerals, and therefore in slower dissolution and higher retardation factors.
8.5.2.7

Influence of the aluminum mineral allowed to precipitate on the evolution

of the low-pH waters
The choice of Al-bearing phase allowed to precipitate was the third major factor controlling
the relative retardation of the rhodochrosite and Al-phase dissolution fronts in the simulations without Ca and Mg silicate dissolution. The simulations that allowed AlOHSO4 to precipitate, instead of amorphous Al(OH)3, exhibited greater R values for the rhodochrosite and
Al-phase dissolution fronts. Remarkably, the simulations that used the more stable AlOHSO4
(solubility product 103.23) compared to simulations that used the less stable AlOHSO4 phase

BUNDSCH_Book.indb 215

11/29/2011 12:32:08 PM

216

Geochemical modeling of groundwater, vadose and geothermal systems

Table 8.9. Retardation factors for the final aluminum phase dissolution-front. See Table 8.7 caption and
notes for explanation of headings and abbreviations. One additional set of simulations reported here used
a CEC of 1 meq/100 g but no rhodochrosite was allowed to precipitate (low CEC, no rhodo.). In some
cases, the aluminum phase was not exhausted during the 5000 time steps allowed for a simulation.
Initial carbonate mineral concentrations ( 102)
Reaction model

0.525

1.05

2.1

3.0

4.2

8.4

16.8

33.2

Basic reaction model, (1)

Dispersivity, (1), PHREEQM
Dispersivity, (1), PHREEQC
Low CEC, (1)
High CEC, (1)
CO2, log pCO = 0.9865, (1)
2
CO2, log pCO = 1.33, (1)
2
Low CEC, no rhodo., (1)
Basic reaction model, (3)
Basic reaction model, (4)
Low CEC, (3)
High CEC, (3)
Surf. Comp., (3), no Cu, Zn, Co, Ni
Surf. Comp., (3), with Cu, Zn, Co, Ni

1.4
1.3
1.2
1.7
2.7
1.3
1.5
1.6
0.0
1.8
1.6
0.0

2.6
2.7
2.7
3.2
5.3
3.2
3.9
3.3
3.6
4.5
3.7
0.0
7.4
8.8

6.5
8.4
7.6
3.9
6.9
5.3
13.9
4.2
7.9
>500
5.0
6.4
24.7
31.6

22.3
25.3
23.3
20.6
9.1
20.8
39.6
8.6
30.3
>500
27.5
8.8
45.9
48.1

69.6
78.9

143.3
162.7

290.8
330.3

2.9

1.8
1.8
1.7
2.2
3.7
1.9
2.3
2.2
2.3
2.7
2.3
0.0
3.5
3.5

68.8
62.6
61.9
106.4
37.7
89.1
>500
87.6
76.0
96.8
102.8

142.6
136.8
113.6
203.9
91.4
171.3
>500
169.9
158.5
176.6
172.3

290.1
284.5
217.3
399.0
198.6
335.5
>500
334.1
323.3
335.8
300.8

Tremolite, Model 2, (1)

Tremolite, MnO2, Model 2, (1)
Biotite/K-mont/An, Model 7, (1)
Biotite/K-mont/An, MnO2, Model 7, (1)
Biotite/K-mont/An, Model 7, (3)

AA
8.2
AA
AA
9.2

AA
12.5
AA
AA
8.0

AA
21.1
AA
AA
6.9

AA
28.4
AA
AA
6.5

AA
38.2
AA
AA
5.0

AA
72.6
AA
AA
5.8

AA
141.1
AA
AA
6.6

AA
278.4
AA
AA
7.3

All simulations referring to Model 7 in Table 8.5 have initial dolomite instead of calcite. AA: Al(OH)3
accumulates!

(solubility product 102.2) resulted in higher R values for the AlOHSO4 dissolution front, but
did not result in significantly different R values for the rhodochrosite dissolution front. More
generally, substantial differences in retardation factors (for a given initial carbonate concentration) between the Al(OH)3 and the AlOHSO4 simulations occurred only in simulations
with CMR ratios above 1.5. The reason for this behavior is not clear, but is probably related
to the faster movement of the MnO2 dissolution front relative to the rhodochrosite dissolution front. In any case, the precipitation of Al(OH)3 generates three times more protons than
AlOHSO4, as demonstrated in the following equations:
Al3+ + SO24 + H2O  AlOHSO 4,s + H+

(8.9)

Al3+ + 3H2O  Al(OH)3,am + 3H+

(8.10)

Precipitation of Al(OH)3 results in substantially lower pH waters and faster carbonate

mineral dissolution, which explains the generally smaller retardation factors found in simulations with Al(OH)3.
The precipitation of Al(OH)3 (or AlOHSO4) generally occurred because dissolution of
carbonate minerals caused an increase in pH. Reaction with the more acidic infilling waters
eventually caused redissolution of the Al-bearing phases. In the case of the simulations with
low initial carbonate concentrations and with either AlOHSO4 (solubility product: 102.2)
or with Al(OH)3 precipitation allowed, complete consumption of the Al-bearing phase
occurred at least twice during each simulation. The retardation factors for the Al-phase
dissolution fronts given in Table 8.9 refer to the last dissolution front in each simulation.

BUNDSCH_Book.indb 216

11/29/2011 12:32:08 PM

Integrating field observations and inverse and forward modeling 217

Effects of the irreversible dissolution of Ca and Mg silicates on the evolution
of low-pH Fe(II)-rich waters
Several simulations were conducted by using reactions specified in the PHREEQC inverse
Models 2 and 7 (Table 8.5). In addition to allowing secondary Al(OH)3, Fe(OH)3, and rhodochrosite to precipitate and specifying initial amounts of calcite (or dolomite) to also react
to equilibrium, these simulations forced the dissolution of a fixed number of moles of Ca
and Mg-silicate minerals into each cell at each time step, thereby simulating a zero-order
kinetic dissolution process that matched (over the 10-cell column) the net amount of Ca- and
Mg-silicate minerals dissolved according to PHREEQC inverse Models 2 and 7 in Table 8.5.
Irreversible dissolution of tremolite was specified in the simulations based on Model 2
(Fig. 8.5). Irreversible dissolution of biotite and anorthite and removal from solution of
K-montmorillonite was specified in the simulations based on Model 7. The silicate dissolution processes would have been more correctly simulated with a pH-dependent rate law, but
this capability was not present in the PHREEQC code used. An inexhaustible supply of
silicate minerals was assumed available to all cells throughout the course of the simulations.
Despite the crudeness of these assumptions, it was hoped that these simulations would contribute some insights on the importance of silicate mineral dissolution in the pH evolution of
acidic contaminated waters.
The Ca- and Mg-silicate dissolution simulations specified an essentially infinite amount
of chalcedony that was available to react to equilibrium. Following the results of Models
2 and 7 in Table 8.5, gypsum precipitation was excluded from the simulations. (A few simulations were run allowing gypsum precipitation, but the results were not substantially different
and are not presented here.) Although an initial fixed amount of MnO2 (2 102 mol kg1
H2O) was specified in most simulations, two sets of simulations specified instead a zero-order
dissolution process for MnO2, according to the amounts determined by Models 2 and 7 in
Table 8.5. Finally, one set of simulations was based on Model 7 (Table 8.5), but allowed
AlOHSO4 to precipitate instead of Al(OH)3.
All the simulations that specified a fixed initial MnO2 concentration resulted in rhodochrosite precipitation and accumulation (Table 8.7, Fig. 8.5a). A rhodochrosite dissolution
front never formed; the continuous proton consumption caused by dissolution of the silicate minerals and the high initial Mn and TDIC concentrations in the infilling water (were
sufficient to ensure that the infilling water never became undersaturated with respect to
rhodochrosite. In contrast, the simulations that allowed irreversible zero-order dissolution
of MnO2 produced rhodochrosite dissolution fronts (Fig. 8.5b) because the oxidation of
dissolved Fe(II) caused the precipitation of Fe(OH)3 and constantly generated acidity that
dissolved rhodochrosite. The R values for the rhodochrosite dissolution fronts nearly doubled for each doubling of the initial carbonate concentration if carbonate concentrations
were above 4.2 102 mol kg1 H2O; at lower initial carbonate concentrations, rhodochrosite
R values increased only by 50% or less for each doubling in initial carbonate concentration.
The two sets of simulations with irreversible tremolite dissolution (with and without fixed
initial MnO2 concentrations) generated calcite dissolution fronts (Table 8.8, Fig. 8.5). In contrast, of the three sets of simulations based on inverse Model 7 (and irreversible biotite and
anorthite dissolution rather than tremolite dissolution), only the simulations that allowed
continuous zero-order dissolution of MnO2 (and consequent production of acidity) generated calcite dissolution fronts. In the simulations with tremolite dissolution and fixed initial
MnO2, the calcite dissolution R values jumped by a factor of 25, from 4.5 to 113.7, as the initial
calcite concentration used in each simulation increased from 4.2 102 to 8.4 102 mol kg1
H2O (and then tripled with each further doubling of the initial calcite concentrations above
that). The jump by a factor of 25 in the R values occurred when, as a result of the higher
initial calcite concentration, the calcite dissolution front moved more slowly than the MnO2
dissolution front (R = 4.7). In simulations that also used a fixed initial MnO2 concentration
(2 102 mol kg1 H2O) but attempted to emulate inverse Model 7, any initial dolomite dissolution was eventually followed by dolomite precipitation and accumulation. Because of the
8.5.2.8

BUNDSCH_Book.indb 217

11/29/2011 12:32:10 PM

218

Geochemical modeling of groundwater, vadose and geothermal systems

additional contribution of Fe(II)-containing biotite, the retardation factor was 4.5 for the
MnO2 dissolution front in these simulations.
Most of the simulations with Ca- and Mg-silicate dissolution did not produce an Al-phase
dissolution front (or produced only a temporary one). The proton consumption caused by
the silicate dissolution reactions caused the continuous precipitation of Al(OH)3 from the
infilling water (Table 8.9). The set of simulations with irreversible tremolite dissolution and
zero-order MnO2 dissolution did generate an Al(OH)3 dissolution front because of the acidity generated by the continuous addition of MnO2 by zero-order kinetics, its oxidation of dissolved Fe(II), and the consequent precipitation of Fe(OH)3. The set of simulations emulating
Model 7, but allowing AlOHSO4 to precipitate instead of Al(OH)3, produced a AlOHSO4
dissolution front. Surprisingly, the R value for that front decreased as the initial carbonate
concentration increased to 4.2 102 mol kg1 H2O and then continued to increase with higher
initial carbonate concentrations.
The pH values obtained during, the evolution of the well 402 water during the simulations with irreversible Ca- and Mg-silicate dissolution (Fig. 8.5) were close to the pH values
predicted for the various mineral assemblages in Table 8.6, but only when carbonate minerals
(rhodochrosite or calcite) were still present and MnO2 was either being added or was still
present. The first 10 shifts of the simulations show a large increase in the pH of cell 10
because of the constant dissolution of Ca- and Mg-silicates into the background water that
initially filled the column; after 10 shifts the water in the column has been replaced by reacted
infilling water.
8.5.2.9 The effect of not allowing rhodochrosite precipitation
One set of simulations, without Ca- and Mg-silicate dissolution, used a low cation-exchange
capacity (1 meq/100 g), but disallowed rhodochrosite precipitation. For those simulations,
the calcite dissolution front was the most important pH front to consider, because it was
either the only, or the slowest (for simulations with initial dolomite and calcite), carbonate
mineral dissolution front that needed to be considered. (The rhodochrosite dissolution front
was the slowest carbonate mineral dissolution front in all other simulations.) The R values
for the calcite dissolution front were greater in the simulations without rhodochrosite precipitation than in the simulations that allowed rhodochrosite precipitation. However, the calcite
R values in the simulations without rhodochrosite precipitation were up to two times smaller
than the rhodochrosite R values in the runs that allowed rhodochrosite precipitation. The
pH values after complete dissolution of calcite in the simulations without rhodochrosite precipitation match the low-pH values after complete rhodochrosite dissolution in the simulations that allowed precipitation of that mineral (Table 8.6).
In simulations with low initial carbonate (2.1 102 mol kg1 H2O and lower), the faster
movement of the Al(OH)3 dissolution front relative to the rhodochrosite dissolution front
meant that the complete dissolution of rhodochrosite led to a large pH decrease (Fig. 8.4a).
In simulations with higher initial carbonate concentrations, the second Al(OH)3 dissolution
front travelled more slowly than the rhodochrosite front and therefore the disappearance of
rhodochrosite resulted in a relatively small decrease in pH (Figs. 8.4b, 8.4c, 8.4d). The R values for the Al(OH)3 final dissolution-front were smaller (by as much as 1/3 in the runs with
high initial carbonate concentrations) in the simulations without rhodochrosite precipitation
relative to the Al(OH)3 R values obtained in the simulations that allowed rhodochrosite to
precipitate. For simulations with the three lowest initial carbonate concentrations (5.25 103,
1.05 102 and 2.1 102), however, differences in the R values between the simulations that
did or did not include rhodochrosite precipitation were relatively small.
8.5.2.10 The CO2 open system simulations
A few reactive-transport simulations allowed the l-dimensional column to equilibrate with an
infinite gaseous CO2 reservoir at a fixed partial pressure of either 101.33 (the calculated pCO
2
for water at well 503, Brown, 1996) or 100.9865 (the measured pCO of the unsaturated zone at
2

BUNDSCH_Book.indb 218

11/29/2011 12:32:10 PM

Integrating field observations and inverse and forward modeling 219

site 500 in November 1991). Opening the system up to CO2 affected the retardation of the
calcite, rhodochrosite and Al(OH)3 dissolution fronts, particularly in the simulations with
at least 3 102 mol kg1 of initial calcite (i.e., with CMR ratios of 1.5 and above). The
simulations conducted under the higher fixed pCO value (1009865) generally resulted in lower
2
R values for the calcite, rhodochrosite, and Al(OH)3 dissolution fronts, compared to the values determined in the BRM simulations. In contrast, the simulations conducted under the
lower fixed pCO value (101.33) resulted in higher R values for all three mineral dissolution
2
fronts, compared to the values determined in the BRM simulations. The relevance of these
simulations to the migration of the acidic ground waters at Pinal Creek field site is questionable. It is unlikely that groundwaters at the site maintain contact with the unsaturated zone
during their evolution. However, the CO2 equilibration simulations have large effects on the
movement of the fronts, and further study is needed to confirm that CO2 ingassing or outgassing is not significant at the field site.
8.5.2.11 The effect of longitudinal dispersion
When the longitudinal dispersivity () is increased from 0 to 560 m (which represents 10%
of the distance between wells 402 and 503), the retardation of the calcite or rhodochrosite
dissolution fronts are not significantly affected, except for the simulations with the lowest
initial carbonate concentrations (3 102 mol kg1 H2O and lower, corresponding to CMR
ratios below 1.5). Dispersion typically has little effect on the propagation of sharp fronts
caused by simple mineral dissolution reactions, especially if the initial mineral concentrations are sufficient to significantly retard the propagation of the fronts. At the limit, dispersion would have no effect on the retardation of a mineral dissolution front for which an
infinite initial mineral concentration had been specified. Of course, using a dispersivity of
560 m has a large effect on the spreading of non-reacting solutes, and is also expected to
affect the mineral dissolution fronts that propagate at nearly the same speed as the water.
Longitudinal dispersion implies that portions of a tracer or solute of interest move both
faster and slower than the average groundwater velocity. The spreading of a conservative
solute after a given travel time, or travel distance, can be calculated from the following equation (Appelo and Postma, 1993):

2vt = 2 x

(8.11)

If = 560 m and x = 5600 m, the distance between wells 402 and 503, x = 2504 m.
x represents a distance between two specific points in a 1-D homogeneous column. At those
two points, a conservative tracer injected continuously at the beginning of the column (i.e.,
at well 402), with a relative concentration c/c0 of 1.0, would have achieved relative concentrations of 0.5 and 0.16, respectively. When the relative concentration of a conservative tracer
reaches 0.5 at well 503, 5600 m downgradient from well 402, the 0.16 relative concentration
level has already moved ahead by 2504 m. In contrast, the 0.84 relative concentration level
would be 2504 m upgradient from well 503. The above analysis assumes steady-state flow
through a porous medium with homogeneous physical properties and also assumes that there
are no dead-water zones into which the conservative tracer can diffuse.
The simulations that included longitudinal dispersion were conducted with both the
PHREEQM code and with an unpublished version of the PHREEQC v. 2 code obtained
in 1996 (David Parkhurst and Tony Appelo, written commun.). Unlike its first 1995 version,
PHREEQC v. 2 simulated longitudinal dispersion by using an algorithm similar to the one
used in PHREEQM. Efforts were made to ensure that the thermodynamic databases for
PHREEQM and PHREEQC were identical, particularly for Al species. Although we are
not aware of any differences, some minor differences may have still remained. The database used in the PHREEQC v. 2 runs was identical to that used in all the other PHREEQC
runs (the PHREEQC v. 1 runs, without dispersion). Unfortunately, we were unable to make
PHREEQC v. 2 converge for the entire set of simulations. Although current versions of

BUNDSCH_Book.indb 219

11/29/2011 12:32:10 PM

220

Geochemical modeling of groundwater, vadose and geothermal systems

PHREEQC are more stable and have more features than the earlier pre-publication version
that was used, modeling comparisons with PHREEQM, and with later versions of PHREEQC
v. 2, leave us confident that the results obtained with the earlier version would match those
obtained with more recent versions.
8.5.2.12 The influence of cation exchange and surface-complexation sorption processes
The reader is referred to Appelo (1996), Appelo and Postma (1993, 2005), Dzombak and
Morel (1990) and Davis and Kent (1990) for excellent descriptions of the theories of cation
exchange and surface-complexation sorption processes. The cation exchange conventions
used in the PHREEQM and PHREEQC codes are described in Appelo and Postma (1993).
PHREEQCs simulation of surface complexation sorption processes implements the diffuse
double-layer surface-complexation model presented by Dzombak and Morel (1990) and is
fully described in Parkhurst (1995) and Parkhurst and Appelo (1999).
Allowing cation exchange with a cation-exchange capacity (CEC) of 1 meq/100 g did
not substantially affect rhodochrosite and calcite retardation factors (Tables 8.7, 8.8). The
ions allowed to exchange were Al3+, Fe2+, Mn2+, Ca2+, Mg2+, Sr2+, Ba2+, Na+, and K+. Proton
exchange was also simulated. The selectivity coefficients used were the default values present
in the PHREEQC thermodynamic database (which are nearly identical to values given in
Appelo and Postma, 1993). A CEC of 1 meq/100 g [equivalent to 52.2 meq kg1 H2O appears
to be a reasonable estimate for the Pinal Creek sediments, given the low organic carbon content (less than 1%) and relative coarseness of the sediments]. The empirical formula given
by Breeuwsma et al. (1986) relating the CEC to the <2 m clay fraction and to the organic
carbon content is assumed applicable (cited in Appelo and Postma, 1993):
CEC (meq/100g) = 0.7 (% clay) + 3.5 (% C)

(8.12)

Direct measurements of the CEC of Pinal Creek sediments, and of the exchangeable ion
composition of the sediments, would be preferable to an empirical formula. The purpose
of our simulations, however, was to determine the effect of ion-exchange processes on the
evolution of the low-pH and high-Fe(II) contaminated groundwaters at the site. Therefore,
simulations were also conducted using an unrealistically high CEC of 10 meq/100 g, a CEC
that would be applicable to sediments with more than 10% clay content (and 1% organic
carbon). As discussed earlier, these high CEC simulations resulted in a doubling of the retardation of the MnO2 dissolution front (from 4.7 to 9.2 for an initial MnO2 concentration
of 2 102 mol kg1 H2O). The high CEC ion-exchange simulations generally increased the
retardation factors for the calcite dissolution front but did not result in a uniform increase
in the retardation of the rhodochrosite or of the Al-phase dissolution fronts (as a function
of initial carbonate). At the lowest initial carbonate concentrations (0.525 102), the high
CEC simulations did not result in rhodochrosite precipitation. At higher initial carbonate
concentrations, between 2.1 102 and 3.0 102 mol kg1 H2O, the high CEC simulations had
higher R values for the rhodochrosite dissolution front than in the model runs that did not
simulate cation exchange (or those that assumed a low CEC). At even higher initial carbonate concentrations (4.2 102 and above), however, this behavior was reversed: the R values
for rhodochrosite dissolution for the high CEC simulations were lower than the R values
observed in the simulations with lower or no ion exchange capacity. The retardation factors for the Al(OH)3 dissolution front also exhibited complex behavior with increasing initial
carbonate concentration.
The PHREEQC simulations that included a diffuse double-layer surface-complexationsorption model resulted in retardation factors for the rhodochrosite, calcite and AlOHSO4
dissolution fronts that were generally higher than the R values determined in simulations
without sorption. Surface protonation and deprotonation reactions were incorporated into
the sorption model, along with surface complexation of Mn, Ca, Mg, Cu, Zn, Co, Ni and
SO4 at weak and strong surface sites.

BUNDSCH_Book.indb 220

11/29/2011 12:32:11 PM

Integrating field observations and inverse and forward modeling 221

The thermodynamic sorption model was based on the compilation of intrinsic constants
for hydrous ferric oxide published by Dzombak and Morel (1990) and used by default in
PHREEQC. The number of weak and strong sorption sites used (1.5 102 and 3.8 104
respectively) and the amount of surface area per kg of H2O used (4032 m2) were identical to
the values used by Brown (1996) in his 1-D simulation of reactive transport at the Pinal Creek
site, and also in his PHREEQC-simulation fit of the column experiment conducted by Stollenwerk (1994). Using a higher number of sorption sites in our simulations would probably have
resulted in more significant effects on the retardation of the mineral dissolution fronts.
Cu (36.2 mg L1), Zn (4.97 mg L1), Co (4.14 mg L1) and Ni (1.57 mg L1) concentrations
measured in the well 402 water were added to one set of simulations. Another set of simulations was conducted without those elements. This second set of simulations resulted in higher
R values for the rhodochrosite and calcite fronts, but in slightly lower R values for the AlOHSO4
front except in the simulations with the two highest initial carbonate concentrations.
8.5.2.13 Other minor effects on the evolution of the low-pH waters
Several simulations included an initial concentration of dolomite, in addition to calcite, or
allowed kaolinite to precipitate (instead of Al(OH)3 or AlOHSO4). The simulations otherwise
used the BRM setup, but all specified the same initial carbonate mineral concentration
(8.4 102 mol kg1 H2O). The resulting retardation factors for the calcite, dolomite and
rhodochrosite dissolution fronts can be found in Glynn and Brown (1996, Table 12). Including or excluding dolomite did not, by itself, affect the retardation of the low-pH fronts. The
important variable was the total moles of solid carbonate, whether in the form of calcite
or dolomite. Allowing kaolinite to precipitate instead of Al(OH)3 resulted in slightly lower
retardation factors for the calcite and rhodochrosite dissolution fronts (Table 12 in Glynn
and Brown). In contrast, the R values obtained for the final kaolinite dissolution front were
generally much greater than those obtained for the final Al(OH)3 dissolution front. These
effects were due to the greater stability of kaolinite, relative to Al(OH)3, in the presence of
acidic water.
8.5.2.14

Comparison of the reactive transport simulation results

with observations at the Pinal Creek site
Although groundwater samples at the Pinal Creek site have been collected and analysed by
the U.S. Geological Survey since 1984, limited resources, and the large size of the basin and
of the contaminant plume, prevented the emplacement of a high number of wells. The sparseness of the available spatial information have made it difficult to determine the location of
the low-pH and high-Fe(II) groundwaters (Fig. 8.6) through time. The longitudinal spreading
of the fronts (and especially of the Fe(II) front), which might give valuable information on
reaction kinetics, has also been difficult to determine.
Nevertheless some estimates of the velocity of the low-pH front can be made, primarily
because of the breakthrough of low-pH waters (from a pH of 4.96 in March 1989 to a pH of
4.24 in August 1989) that was observed at well 451 only a few months after its emplacement
(December 1988). If the creation of Webster Lake in 1940, approximately 18.5 km upgradient
from well 451, provided the principal source of acidic contaminated waters, then an effective velocity of about 1 m day1 can be estimated for the low-pH front over that section of
the aquifer. This velocity can be compared to an estimated average groundwater velocity
of 8.4 m day1 between Webster Lake and well 451, giving an estimated retardation factor of
about 8.4 for the movement of the low-pH waters. This retardation factor estimate is thought
to be a maximum estimate, because the applicable groundwater velocity could be as low as
5 m day1. Eychaner (1991) and Glynn et al. (1991) estimated that a lower R value of 5 would
be reasonable for the movement of the pH front from Webster Lake to well 451.
Similarly, a low-pH (4.0 to 4.5) high-Fe(II) water was found to be present during drilling
(in February 1995) of a group of wells (LPC wells) emplaced by the Pinal Creek Group
(a consortium of copper companies) slightly west of well 503 (on the other side of the

BUNDSCH_Book.indb 221

11/29/2011 12:32:12 PM

222

Geochemical modeling of groundwater, vadose and geothermal systems

creek bed). Given the distance of the LPC wells from well 451 (1.3 km), the minimum velocity
of the low-pH front is estimated to be greater than 0.65 m day1. Using this velocity with an
estimated average groundwater velocity of about 5 m day1 (between 451 and 503) results
in a maximum R value of about 7.7 for the low-pH front. The acidity of the well 451 water
in August 1989 was 2 times lower than that of the well 402 water used in our simulations,
although the potential acidity (cf. discussion of Table 7 in Glynn and Brown, 1996) of the
well 451 water did increase with time until a maximum acidity, about 70% of that for well
402, was reached in November 1991.
In comparison to this field evidence, results of a column elution experiment conducted
by Stollenwerk (1994), using water from well 51 (of composition similar to that presented in
Glynn and Brown, 1996, Table 7) and an 80-cm column containing uncontaminated alluvium
from the Pinal Creek site, indicated an R value for the low-pH front of about 2.5. Many factors can explain the lower retardation found in the column experiment relative to the field
values. The low-pH front R values estimated from the field evidence incorporate significant
effects of dilution (Fig. 8.3). The Webster Lake waters were more acidic than the well 51
water used in Stollenwerks column experiment. The potential acidity of the well 51 water was
4.5 times greater than that of the well 402 water used in the Glynn and Brown (1996) simulations shown here. As a result, the retardation factor of 2.5 determined by Stollenwerk (1994)
should translate to a retardation factor of approximately 7.8, (2.5 1) 4.5 + 1 = 7.75, had
well 402 water been used in his column experiments. Equation (8.4) is used to normalize
these results, assuming that M is proportional to the acidity of the low-pH solution and
that the field and laboratory experiments both had identical homogeneous mineral concentrations and reactions. Similarly, considering that the water from well 45l in August 1989
was about 2 times less acidic than the water used in the simulations shown here, the R value
of 7.7 determined between 451 and the LPC site would correspond to an R value of 4.35, if
normalized to the acidity of the well 402 water (89/1/12). These normalized retardation factors (4.4 and 7.8) can be compared with the R values obtained in our numerical simulations
of the rhodochrosite dissolution front (the best analog for the low-pH front observed in the
field). The simulation results in Table 8.7, for runs without irreversible dissolution of Ca
and Mg-silicates, indicate that initial carbonate mineral concentrations between 2.1 102
and 4.2 102 mol kg1 H2O and certainly no lower than 1.05 102mol kg1 H2O would
give reasonable simulated retardation factors for the low-pH waters. This assumes that an
MnO2 concentration of 2 102 mol kg1 H2O was also reasonable. If not, two different situations can be considered. A lower initial MnO2 concentration would result in a higher CMR
value and therefore (cf. section 8.5.2.6) could increase the simulated retardation factors for
the low-pH front. Consequently, initial carbonate concentrations would have to be adjusted
slightly downward in order to match the estimated lab and field retardation factors. Alternatively, a higher initial MnO2 concentration would probably not have a substantial effect on
the simulated low-pH front R values (for initial carbonate mineral concentrations between
1.05 102 and 3.0 102 mol kg1 H2O).
In the simulations with Ca- and Mg-silicate dissolution, the retardation factors for the
rhodochrosite front are so high that an initial carbonate concentration of 5.25 103 or lower
is needed to match the observed retardation of the low-pH front in Stollenwerks experiments
and in the field, if we assume that the disappearance of rhodochrosite controls the low-pH
front. Such a low initial carbonate concentration is not realistic given our knowledge of the
alluvial sediments at the site, and we conclude that the simulations with irreversible dissolution of Ca- and Mg-silicate should be conducted by using a finite (as opposed to an inexhaustible) source of these silicates. Further field information on the concentration and rate
of dissolution of silicate minerals would however be needed. The calcite R values obtained in
our forward simulations of inverse Model 2 (Table 8.5) are perhaps reasonable, but only for
initial calcite concentrations below 4.2 102 mol kg1 H2O. The disappearance of calcite did
not by itself, however, result in a sufficient pH decrease to match the pH values of 4.0 to 4.5
that we associate with the low-pH front in the field (Fig. 8.6).

BUNDSCH_Book.indb 222

11/29/2011 12:32:12 PM

Integrating field observations and inverse and forward modeling 223

Figure 8.6. pH profiles along an aquifer flowpath based on the most contaminated wells in the Pinal
Creek alluvial aquifer. Well 451 was drilled in December 1988 and destroyed during the floods of spring
1993. Therefore only January 1989 and November 1991 pH data are presented for that well. Water
samples taken from 451 show a large pH decrease between those dates (most of the decrease occurred
between March and August 1989).

The identity of the secondary aluminum phase that precipitates and the rate of redissolution of that phase are also important to consider. Table 8.6 shows that the most acidic waters
(below pH 5.0) can only be obtained if Al(OH)3 is not present: i.e., it either never precipitates
out, or if it does, it has completely redissolved, or its rate of dissolution is so slow that it can
be considered unreactive. If we assume that secondary Al(OH)3 does precipitate out and
that the Local Equilibrium Assumption controls its precipitation and later dissolution, then
our simulations of the Al(OH)3 dissolution front show (Table 8.9) that initial carbonate concentrations would have to be lower than 3.0 102 mol kg1 H2O to match the retardation of
the low-pH front observed in the field and in Stollenwerks laboratory columns.
As of 1996, there had been no direct observation of the breakthrough of the highFe(II) waters at the Pinal Creek site. Waters from wells 451 and the LPC site (near well
503) already had high Fe(II) concentrations during the emplacement of those wells (in
November 1991 and February 1995 respectively). The fact that dissolved Fe(II) was present
at well 451 before the low-pH breakthrough suggests that the Fe(II) front moves faster
than the low-pH front, or at least precedes the low-pH front at that point in time. If we
use the creation of Webster Lake as an initial condition, retardation factors between 5 and
8.4 can be estimated for the high Fe(II) front as it moved between Webster Lake and well
451. Similarly, the evidence based on data from well 451 and from the LPC wells suggests
a maximum retardation factor of 7.7 over that section of the aquifer. The latter factor is
equivalent to a maximum retardation factor of about 3.2 in our simulations, after correcting for the three times higher Fe(II) concentration of well 402 relative to that of well 451
(in August 1989).
In his laboratory column experiments, Stollenwerk (1994) observed retardation factors of
approximately 2 for the high-Mn(II) spike, and 2.5 for the high-Fe(II) front. The lag in the
Fe(II) front suggests that the rate of Fe(OH)3 precipitation was delayed relative to the rate of
MnO2 dissolution. Normalizing those retardation factors to an infilling solution with 4.7 times
less dissolved Fe(II) (for well 402) results in an equivalent retardation factor of between 5.7
and 8.1.
After normalization to the infilling water used in our simulations, the field and laboratory determined retardation factors for the high-Fe(II) front encompass a range from 3.2 to
8.1. This range compares favorably with the retardation factor of 4.7 determined for most
of the simulations. Furthermore, according to field and laboratory observations the MnO2

BUNDSCH_Book.indb 223

11/29/2011 12:32:12 PM

224

Geochemical modeling of groundwater, vadose and geothermal systems

dissolution front moves slightly faster than the low-pH front, but it is not moving much
faster, and therefore the CMR ratio is probably close to 1.5.
The pH values determined for the various pH-plateaus in our simulations (Table 8.6,
Figs. 8.4 and 8.5) are within the range of the values observed in the field (Fig. 8.6). Therefore
the hypothesis that the complete dissolution of secondary rhodochrosite occurs after the
dissolution of calcite and is accompanied by a significant decrease in pH (in the absence of
Al(OH)3) is consistent with the field evidence.
8.6

CONCLUSIONS

The basic theory and assumptions of inverse geochemical modeling were presented. Although
much less commonly used than reactive-transport modeling, particularly in investigations
of groundwater contamination, inverse geochemical modeling provides a powerful tool
to identify the reactions that affect the chemical and isotopic evolution of contaminated
groundwaters. The reaction models identified by inverse modeling can be incorporated into a
geochemical reactive-transport model to investigate the effects of the various reactions on the
transport of reactive contaminants. This approach was demonstrated for the case of acidic
groundwater with heavy-metal contamination in the Pinal Creek basin, Arizona.
The interactive inverse geochemical modeling code NETPATH was used to construct
a series of inverse models that quantified observed differences in chemical composition
between an initial acidic Fe(II)-rich water and an evolutionary, partially-neutralized Fe(II)poor water, according to a postulated set of reactions. Each inverse model was evaluated
according to existing knowledge of the geochemistry of the aquifer and the thermodynamic
and kinetic feasibility of the model reactions. Once a plausible set of inverse models was
identified, the PHREEQC inverse modeling code was used to further evaluate the set of
inverse models. Unlike NETPATH, PHREEQC accounts for uncertainties in the analytical
data and, additionally, includes alkalinity-balance, charge-balance and water-balance constraints in solving inverse models.
Inverse modeling with NETPATH and with PHREEQC quantified the reaction processes
responsible for the evolution of an acidic Fe(II)-rich groundwater into a partially neutralized Fe(II)-poor water at the Pinal Creek site. The principal reaction processes appear to be
the reductive dissolution of solid MnO2 by aqueous Fe(II), the consequent precipitation of
Fe(OH)3, the dissolution of calcite and/or dolomite, the precipitation of an aluminum phase,
probably Al(OH)3(rather than AlOHSO4), and, possibly, the precipitation of chalcedony.
Results of the inverse modeling simulations also led to the following conclusions:
1. Ca- and Mg-silicate dissolution must be an important process during the neutralization of
the low-pH waters, given that CO2 exsolution probably does not occur.
2. Dilution of the acidic groundwaters occurs, and Cl, Na, and SO4 can be considered
conservative constituents that quantify the extent of the dilution. The PHREEQC models
revealed that SO4 could be considered a conservative constituent given its associated
analytical uncertainty.
3. Rhodochrosite precipitation (or possibly an MnO2-Mn(OH)3 electron-transfer or Mn2+
sorption mechanism) is responsible for the lower than expected increase in dissolved
Mn(II) concentrations caused by the aqueous Fe(II) reduction of MnO2 solids.
After a brief review of previous reactive-transport modeling at the Pinal Creek site, the
results of some new forward simulations were presented. These geochemical transport
simulations explored the effects of reactionsidentified by inverse modeling and cited by
other researcherson the chemical evolution of an acidic Fe(II) rich water from the Pinal
Creek site. The purpose of the PHREEQC and PHREEQM transport simulations was to
determine the effect of the reactions on the relative rates of movement of the Fe(II)-rich and
low-pH groundwaters.

BUNDSCH_Book.indb 224

11/29/2011 12:32:12 PM

Integrating field observations and inverse and forward modeling 225

The only factors affecting the retardation of the Fe(II)-rich waters and the propagation
of the MnO2 dissolution front in our simulations were the initial specified concentration of
MnO2 and the concentration of Fe(II) in the inflowing contaminated water. As demonstrated
in the text, the rate of movement of the MnO2 front could be easily calculated by hand,
but only in the absence of ion-exchange or other processes substantially affecting the Fe(II)
concentration.
The propagation of various pH fronts caused by complete dissolution of carbonate and
aluminum minerals, however, could not have been easily predicted. The use of a geochemical
transport code such as PHREEQM or PHREEQC is necessary to determine the retardation
factors applicable to interacting mineral dissolution fronts. The initial amount of carbonate
(calcite and/or dolomite) in each simulation was the primary factor determining the movement of the low-pH waters. Other important factors were the ratio of initial carbonate to
initial MnO2, the type of Al phase allowed to precipitate, and whether or not rhodochrosite
was allowed to precipitate. High initial carbonate to MnO2 ratios, allowing rhodochrosite precipitation, and allowing AlOHSO4 precipitation (instead of Al(OH)3 precipitation), resulted
in higher retardation factors for the movement of the low-pH waters. [As mentioned earlier
though, Bigham and Nordstroms (2000) review of environmental conditions required for the
occurrence of AlOHSO4 phases reinforces Glynn and Browns (1996) view that an Al(OH)3
phase was probably a more reasonable phase to consider in modeling the evolution of Pinal
Creek groundwaters.] Allowing the irreversible dissolution of Ca- and Mg-silicates, so as
to match the mass transfers determined in a few inverse models, resulted in unrealistically
high retardation factors for the rhodochrosite and Al(OH)3 dissolution fronts, although the
retardation factors for the calcite dissolution front were reasonable. More field and laboratory
information is required on the rates and extents of reaction for these silicate minerals to
further constrain reactive-transportsimulations.
Inclusion of ion-exchange processes did not substantially affect the retardation of the
various pH fronts when a reasonable low (1 meq/100 g) CEC was used. Because of surface-protonation, allowing surface-complexation reactions resulted in higher retardation
factors for the carbonate and AlOHSO4 dissolution fronts. Allowing equilibrium with a pCO2
of 1009865 resulted in lower retardation factors for the carbonate and Al(OH)3 dissolution
fronts, particularly at initial carbonate concentrations greater than 3 102 mol kg1 H2O.
Allowing equilibrium with a lower pCO of 101.33 resulted in higher retardation factors for
2
the carbonate and Al(OH)3 dissolution fronts. Simulation of longitudinal dispersion was not
an important factor controlling the movement of the calcite and rhodochrosite dissolution
fronts except at low initial carbonate concentrations. Longitudinal dispersion also did not
affect the simulated rate of movement of the Fe(II)-rich waters. Including dolomite in addition to calcite in the background aquifer and allowing kaolinite to precipitate instead of
Al(OH)3 did not significantly affect the propagation of the low-pH fronts associated with the
dissolution of calcite and rhodochrosite.
A comparison of the retardation factors for the low-pH and high-Fe(II) fronts determined
in our local equilibrium simulations with retardation factors estimated from field evidence
and Stollenwerks (1994) laboratory column elution tests suggests that an initial carbonate
concentration between 2.1 102 and 4.2 102 mol kg1 H2O and the initial MnO2 concentration of 2 102 mol kg1 H2O used in our simulations are reasonable. The pH values
obtained in our local equilibrium simulations are reasonable given the pH values observed
in the field.
The retardation factors determined for the mineral dissolution fronts in various simulations are useful not only in estimating the rate of movement of the low-pH and high-Fe(II)
groundwaters at the Pinal Creek site, but also in identifying the most important chemical
parameters controlling the movement of these contaminated waters. Our simulation results
may also provide information on the rate of movement of acidic, metal-laden waters at other
groundwater contamination sites, and our study illustrates how to adjust for pH, dissolved
metal (Al, Fe, Mn) concentrations, and mineralogical characteristics of a given site.

BUNDSCH_Book.indb 225

11/29/2011 12:32:12 PM

226

Geochemical modeling of groundwater, vadose and geothermal systems

This paper presents and demonstrates an approach for the investigation of the
evolution and movement of contaminated groundwaters: the use of inverse geochemical modeling to identify important reaction processes, followed by geochemical reactivetransport simulations that incorporate the inverse-model-derived reactions and account for a
range of aquifer characteristics. Such an approach results in an improved understanding of
the processes that control the past and future evolution of contaminated groundwaters. This
information may then lead to better predictions for the transport of reactive contaminants,
which, in turn, may result in more effective remediation of groundwater contamination at
sites with sparse spatial information.
8.7

THE SENIOR AUTHORS FIFTEEN YEAR PERSPECTIVE ON THE GLYNN

AND BROWN (1996) PAPER

The geochemical modeling conducted by Glynn and Brown (1996), while confirming and
refining knowledge about chemical reactions in the Pinal Creek groundwater system, also
highlighted significant knowledge gaps. The Glynn and Brown (1996) study provided a basis
for further groundwater geochemical investigations at the Pinal Creek site, including several
reactive-transport modeling studies. For example, Brown et al. (1998) modified and refined
the simulation of groundwater evolution presented in Brown (1996) for 1-D flow (with lateral
dilutions as needed) from well 51 to well 701 from 1984 to 1994. Some of the reactions were
changed in the modified model as well as the spatial distribution of those reactions. For
example, rhodochrosite was allowed to precipitate and CO2 was allowed to equilibrate with a
pCO of 101.33 (typical of the unsaturated zone) in the neutralized zone downgradient from well
2
503. Sensitivity analyses examined, for example, surface complexation and sorption of metals
on iron oxyhydroxide. Sorption was relatively less important than initially considered by the
authors in controlling pH and dissolved metal concentrations relative to other processes (such
as redox processes, carbonate dissolution, and dilution). Longitudinal hydrodynamic
dispersion was found to be relatively unimportant (a beta version of PHREEQC v.2 was
used). Generally, good fits were obtained for the simulation of heavy metals such as Cu, Zn,
and Ni. However, the simulation of pH and dissolved Fe and Mn still proved problematic,
and a simulation of the carbonate and redox reaction kinetics was still needed.
Brown et al. (2000), again improving on Brown (1996), found that HYDROGEOCHEM
and PHREEQC simulations of the Stollenwerk (1994) column experiments provided similar
numerical results, but once again, required accounting for reaction kinetics, if the experimental results were to be matched by the numerical simulations. In the absence of adequate
knowledge of reaction rates, the observed experimental results could also have been matched
by fitting irreversible mass transfers of water and reactants in PHREEQC. A common
result of the Brown et al. (2000, 1998) studies was that the numerical simulation results were
reasonable only if the number of adsorption sites was reduced to 5% of the initial value considered by Stollenwerk (1994). Stollenwerk simulated, through the use of the geochemical
code MINTEQA2, the neutralization and later rinse-out of Pinal Creek acidic groundwater
that he observed in a column experiment. Because MINTEQA2 was not a reactive transport
code, Stollenwerk used a series of batch calculations to simulate transport, and explicitly
defined the aqueous and solid-phase chemical inputs for each cell at each time step. As mentioned in Brown et al. (2000), Although output from one time step provided much of the
input for the next time step, the aqueous and solid-phase concentrations of selected constituents were manually adjusted to fit measured concentrations in column outflow and measured
changes in column mineralogy.
To improve the characterisation of calcite, dolomite, and pyrolusite (MnO2) dissolution
rates at the Pinal Creek site, Brown and Glynn (2003) conducted in-situ field experiments that
suspended well-characterized commercially-obtained mineral samples in acidic waters at
the Pinal Creek site (wells 101 and 301) for various lengths of time (from 96 to 595 days)

BUNDSCH_Book.indb 226

11/29/2011 12:32:12 PM

Integrating field observations and inverse and forward modeling 227

starting in April 1998. In-situ dissolution rates for calcite and dolomite were determined,
and found to be about three orders of magnitude slower than rates determined by Plummer
et al. (1978) and Busenberg and Plummer (1982) in laboratory experiments. Interestingly,
Brown and Glynn (2003) found that the incorporation of the in-situ rates, or alternatively,
the faster laboratory-determined rates, in PHREEQC v.2 (Parkhurst and Appelo, 1999) reactive-transport simulations did not result in better fits of Stollenwerks column experiments.
Instead, an equilibrium model adequately simulated carbonate dissolution at the shorter and
faster (relative to the field situation) spatial and temporal scales of the column experiment.
More surprisingly, Brown and Glynn (2003) observed growth in the mass of the pyrolusite
material placed in the in-situ experiments, despite thermodynamic conditions that, in theory,
favored pyrolusite dissolution. They postulated that this result could be related to the result
of Villinski et al. (2001), who found that the dissolution of MnO2 under similar conditions
in a laboratory flow-through cell resulted in the precipitation of a mixed oxidation Mn-Fe
phase with the structure of jacobsite (MnFe2O4). Brown and Glynn (2003) then used the rate
equations developed by Postma and Appelo (2000) for the reductive dissolution of birnessite
(MnO2) by dissolved Fe(II) to better fit the distribution and evolution of dissolved Mn and
Fe concentrations in the transition zone and in the acidic part of the plume (at well 402) from
1984 through 1992. Although the rate constant used in their PHREEQC v.2 simulations
was about 3 to 4 orders of magnitude smaller than the range of values used by Postma and
Appelo (2000) in their column experiments, Brown and Glynn (2003) did obtain a better fit of
the observed Fe, and especially, the observed Mn concentrations in the transition zone. Their
simulated pH values, however, remained about 1 pH unit higher than the observed values.
The field, laboratory and numerical modeling investigations of the Pinal Creek basin provide several lessons that may be useful to other investigations of complex, highly dynamic,
systems with a relative paucity of information (and limited funding to obtain information):
1. Constructing, analyzing and interpreting numerical models, regardless of the type of
model (hydrologic vs. geochemical; inverse vs. forward), forces the modeler(s), and hopefully the user(s) of the models, to reexamine and revise their conceptual model and perceptions of the available information. The modeling process forces the modelers and users to
assemble, structure, transform, and assess a wide variety of information.
2. No model is ever final. As numerical modeling leads to greater understanding of a system,
it invariably illuminates knowledge gaps and the need for more information. Frequent
iteration is needed among field observation, field experiments, laboratory experiments, and
the periodic assembly and interpretation of the available information through numerical
modeling.
3. The combination of inverse modeling (including inverse geochemical modeling) and
forward or predictive modeling is particularly powerful in helping assess available information. Inverse modeling forces the modeler to use basic principles (such as conservation of
mass and energy) to interpret the available observations in terms of properties of the system
(e.g., geochemical mass transfers, possible reactions, other intrinsic system properties). Forward, or predictive modeling, forces the modeler to test available information by extrapolating forward through time, space or other informational dimensions by using the
modelers preconceptions of system behavior. In the case of the Pinal Creek system, inverse
modeling helped Glynn and Brown (1996) identify sets of reactions that could explain the
observed geochemical evolution of groundwaters, and helped identify field and laboratory observations/experiments that might improve characterization and quantification of
active geochemical processes. Sensitivity analyses conducted through forward reactive
transport modeling identified which minerals and reaction processes, given knowledge of
reaction thermodynamics and stoichiometries, provided reasonable rates of migration of
chemical fronts and a reasonable distribution of solution properties (such as pH and Fe
and Mn concentrations) in space and time. More importantly, the geochemical modeling
at the Pinal Creek site helped identify which reactions and processes were unreasonable.

BUNDSCH_Book.indb 227

11/29/2011 12:32:13 PM

228

Geochemical modeling of groundwater, vadose and geothermal systems

4. Determining what is unreasonable is generally dependent on limited information and

observations (through time, space and process domains) and on an interpretation of
that limited information. In complex, highly dynamic, systems with limited regularity
or steadiness of observations, such as the Pinal Creek groundwater system, the problem
of eliminating unreasonable processes is exacerbated: what might seem reasonable, or
unreasonable, given available information at some point in time and space could change.
The studies conducted at the Pinal Creek site illustrate the fact that nature always keeps
surprises in reserve for its observers and interpreters. Humility, and frequent testing of
assumptions, are needed in modeling natures systems.
5. Given our often limited knowledge of natural systems, it behooves us to model these systems by considering general system behavior before interpreting, matching, and predicting
specific system behavior. Curve and point-by-point matching of numerical simulations
to the observed chemical evolution at the Pinal Creek site, or laboratory columns, was
generally difficult, and provided limited, albeit useful, returns on expended effort. Inverse
modeling and sensitivity analyses of general system behavior, however, through relatively
unconstrained reactive-transport simulations, provided substantial insights on knowledge
gaps and information needs that could improve understanding of the complex, highly
dynamic processes at the site. These modeling efforts should have been conducted much
earlier in the life of the Pinal Creek project investigations.
6. Finally, the Pinal Creek project taught us the value of spatial and temporal observations.
Systems that have a complex suite of active processes and that exhibit dynamic changes
in observed properties require extensive (and often costly) adaptive monitoring programs
that have sufficient resolution in both time and space. Considering a greater variety of
geochemical/hydrological information than might initially be considered in setting up a
monitoring program is useful and can sometimes help compensate for a lack of spatial/
temporal monitoring: different geochemical or hydrological measurements often reflect
different degrees of spatial integration and temporal evolution.
Other information on the hydrogeochemistry of acidic metal-contaminated waters, geochemical
modeling, and on obtaining U.S. Geological Survey computer codes and the PHREEQM code
Further information on the application of geochemical information and geochemical modeling
in understanding and dating groundwater systems can be found in Glynn and Plummer (2005),
and in a soon to be published IAEA guidebook on dating of old groundwaters and many
chapters therein, including Plummer and Glynn (Radiocarbon dating in groundwater systems;
in press) and Plummer et al. (Characterization and conceptualization of groundwater-flow
systems; in press). An excellent discussion of geochemical modeling tools and philosophy is
also provided by Nordstrom (2004; updated in 2007). Finally, the reader is directed to the
work of Alpers and Nordstrom (1999) and Nordstrom (2000, 2008). These authors have
used geochemical modeling and characterization to inform regulatory decision making in the
remediation of acid mine waters.
Readers interested in geochemical and hydrological investigations of the acidic and heavy
metal contamination in the Pinal Creek basin should also consider the studies conducted
on the role of groundwater-surface water interactions. Narrowing of the basin to the north
forces groundwater discharge and maintains perennial flow for 7 kilometers along Pinal Creek.
Degassing of carbon dioxide raises the pH of the stream from about 6.6 to 7.6 (Choi et al.,
1998). The return flow of the streamwater to shallow subsurface flowpaths beneath and to
the side of the stream (hyporheic zone) brings this higher pH water with high concentrations
of dissolved Mn2+ back into direct contact with microbes and geochemical coatings in the
streambed sediment. Rapid oxidation of Mn2+ in the hyporheic zone sediments is catalyzed
by Mn oxidizing bacteria on a timescale of hours, and is accompanied by the sorption of Co,
Ni, and Zn onto the biogenic manganese oxides (Bargar et al., 2009). Enhanced oxidation of
Mn in the hyporheic zone was found to remove 20% of the dissolved load of Mn reaching
the stream, before its flow out of the basin (Harvey and Fuller, 1998). Additionally, sorption

BUNDSCH_Book.indb 228

11/29/2011 12:32:13 PM

Integrating field observations and inverse and forward modeling 229

onto the manganese oxides in the streambed sediments accounted for 52%, 27%, and 24%
removal of the dissolved Co, Ni, and Zn loads, respectively (Fuller and Harvey, 2000; Kay
et al., 2001).
The latest versions of the USGS PHREEQC and NETPATH geochemical codes, and
other related codes, can be downloaded directly from the web site: http://wwwbrr. cr.usgs.gov/
projects/GWC_coupled/. Codes available include new versions of NETPATH (NetpathXL;
Parkhurst and Charlton, 2008; NETPATH-WIN, El-Kadi et al., 2011) and the 3-dimensional
PHREEQC-based reactive transport code PHAST (Parkhurst et al., 2010). USGS codes can
also be accessed through the more general USGS National Research Program software web
site: http://water.usgs.gov/nrp/models.html. PHREEQM is now largely an obsolete code
that has been replaced by PHREEQC version 2 (Parkhurst and Appelo, 1999) and by the
Windows interactive code PHREEQCi (Charlton and Parkhurst, 1999). Another interactive
code, PHREEQC for Windows, can also be downloaded from Vincent Posts web site: http://
pfw.antipodes.nl/index.html. Additional information on the PHREEQC code can also be
obtained from Tony Appelos web site: http://www.xs4all.nl/appt/, and from the Appelo and
Postma (2005) groundwater geochemistry textbook.
ACKNOWLEDGEMENTS
The original version of this manuscript, Glynn and Brown (1996), received substantial
help and reviews from David Parkhurst, Joseph Vrabel, Carl Steefel, Eric Oelkers, Don
Thorstenson and Paul Ribbe. Additionally, the current manuscript benefitted greatly from
reviews provided by Mark Fuhrman and David Parkhurst, and from additional comments,
edits, and discussion provided by Leonard Konikow, Kirk Nordstrom and Kevin Breen. The
Senior Author retains full responsibility however for any remaining errors of fact, expression
or judgment. The work presented here was funded by the Toxics Substance Hydrology and
the Hydrologic Research and Development programs of the US Geological Survey. Finally,
any use of trade, product, or firm names is for descriptive purposes only and does not imply
endorsement by the U.S. Government.
REFERENCES
Alpers, C.N. & Nordstrom, D.K.: Geochemical modelling of water-rock interactions in mining
environments. In: G.S. Plumlee & M.J. Logsdon (eds): Reviews in Economic Geology, vol. 6A, The
environmental geochemistry of mineral deposits: Part A. Processes, methods and health issues, Society
of Economic Geologists, Littleton, CO, 1999, pp. 289324.
Appelo, C.A.J.: Multicomponent ion exchange and chromatography in natural systems. In: C.I. Steefel,
P. Lichtner & E. Oelkers (eds): Reactive transport in porous media: general principles and application
to geochemical processes. Reviews in Mineralogy 34, 1996, pp. 193227.
Appelo, C.A.J. & Postma, D.: Geochemistry, groundwater and pollution. A.A. Balkema, Rotterdam, The
Netherlands, 1993.
Appelo, C.A.J. & Postma, D.: Geochemistry, groundwater and pollution. 2nd ed., A.A. Balkema, Leiden,
The Netherlands, 2005.
Appelo, C.A.J. & Willemsen, A.: Geochemical calculations and observations on salt water intrusions,
1, a combined geochemical/mixing cell model. J. Hydrology 94 (1987), pp. 313330.
Bargar, J.R., Fuller, C.C., Marcus, M.A., Brearley, A.J., Perez De la Rosa, M., Webb, S.M. & Caldwell,
W.A.: Structural characterization of terrestrial bacteriogenic Mn oxides from Pinal Creek, AZ.
Geochim. Cosmochim. Acta 73 (2009), pp. 889910.
Bigham, J.M. & Nordstrom, D.K.: Iron and aluminum hydroxysulfate minerals. In: C.N. Alpers, J.L.
Jambor & D.K. Nordstrom (eds): Sulfate mineralscrystallography, geochemistry and environmental
significance, Reviews in Mineralogy and Geochemistry 40, 2000, pp. 351403.
Breeuwsma, A., Wsten, J.H.M., Vleeshouwer, J.J., Van Slobbe, A.M. & Bouma, J.: Derivation of land qualities to assess environmental problems from soil surveys. Soil Sci. Soc. Amer. J. 50 (1986), pp. 186190.

BUNDSCH_Book.indb 229

11/29/2011 12:32:13 PM

230

Geochemical modeling of groundwater, vadose and geothermal systems

Brown, J.G.: Movement of metal contaminants in ground water in the Pinal Creek Basin, Arizona:
Model assessment and simulation of reactive transport. MSc Thesis, University of Arizona, Tucson,
AZ, USA, 1996.
Brown, J.G. & Glynn, P.D.: Kinetic dissolution of carbonates and Mn oxides in acidic water: measurement of in situ rates and reactive transport modeling. Appl. Geochem. 18:8 (2003), pp. 12251239.
Brown, J.G. & Harvey, J.W.: Hydrologic and geochemical factors affecting metal-contaminant
transport in Pinal Creek Basin near Globe, Arizona. In: D.W. Morganwalp & D.A. Aronson (eds):
U.S. Geological Survey Toxic Substances Hydrology ProgramProceedings of the Technical Meeting,
Colorado Springs, Colorado, 2024 September, 1993: U.S. Geological Survey Water-Resources
Investigations Report 944015, 1994, pp. 10351042.
Brown, J.G., Bassett, R.L., Glynn, P.D. & Parkhurst, D.L.: Reactive transport of metal contaminants in
ground water in Pinal Creek Basin, Arizona. In: A.M. Geddis (ed.): Water Use in ArizonaCooperation
or Conflict. Proceedings of the 8th Annual Symposium of the Arizona Hydrological Society, Tucson,
Arizona, 1415 September, 1995, pp. 104105.
Brown, J.G., Bassett, R. & Glynn, P.D.: Analysis and simulation of reactive transport of metal
contaminants in ground water in Pinal Creek Basin, Arizona. J. Hydrology 209 (1998), pp. 225250.
Brown, J.G., Bassett, R. & Glynn, P.D.: Reactive transport of metal contaminants in alluviummodel
comparison and column simulation. Appl. Geochem. 15 (2000), pp. 3549.
Busenberg, E. & Plummer, L.N.: The kinetics of dissolution of dolomite in CO2-water systems at 1.5 to
65C and 0.0 to 1.0 atmosphere pCO2. Amer. J. Science 282 (1982), pp. 4578.
Charlton, S.R., Macklin, C.L. & Parkhurst, D.L.: PhreeqcIA graphical user interface for the
geochemical computer program PHREEQC. U.S. Geological Survey Water-Resources Investigations
Report 974222, 1997.
Charlton, S.R. & Parkhurst, D.L.: PhreeqcIA graphical user interface to the geochemical model
PHREEQC. U.S. Geological Survey Fact Sheet FS-03102, 2002.
Choi, J., Hulseapple, S.M., Conklin, M.H. & Harvey, J.W.: Modeling CO2 degassing and pH in a streamaquifer system. J. Hydrology 209 (1998), pp. 297310.
Davis, J.A. & Kent, D.B.: Surface complexation modeling in aqueous geochemistry. In: M.F. Hochella
Jr & A.F. White (eds): Mineral-Water Interface Geochemistry. Reviews in Mineralogy 23, 1990, pp.
177260.
Dria, M.A., Bryant, S.L., Schechter, R.S. & Lake, L.W.: Interacting precipitation dissolution waves: the
movement of inorganic contaminants in groundwater. Water Resour. Res. 23 (1987), pp. 20762090.
Dzombak, D.A. & Morel, F.M.M.: Surface complexation modeling: Hydrous ferric oxide. Wiley & Sons.
New York, 1990.
El-Kadi, A.I., Plummer, L.N. & Aggarwal, P.: NETPATH-WIN: An interactive user version of the
mass-balance model, NETPATH. Ground Water 49:4 (2011), pp. 593599.
Engesgaard, P. & Kipp, K.L.: A geochemical transport model for redox-controlled movement of mineral
fronts in groundwater flow systems: a case of nitrate removal by oxidation of pyrite. Water Resour.
Res. 28 (1992), pp. 28292843.
Eychaner, J.H.: Movement of inorganic contaminants in acidic water near Globe, Arizona. In:
G.E. Mallard & S.E. Ragone (eds):U.S. Geological Survey Toxic Substances Hydrology ProgramProceedings of the Technical Meeting, Phoenix, Arizona, 2630 September, 1989: U.S. Geological Survey Water-Resources Investigations Report 894220, 1989, pp. 567575.
Eychaner, J.H.: The Globe, Arizona, research siteContaminants related to copper mining in a hydrologically integrated environment. In: G.E. Mallard & D.A. Aronson (eds): U.S. Geological Survey
Toxic Substances Hydrology ProgramProceedings of the Technical Meeting, Monterey, California,
1115 March, 1991: U.S. Geological Survey Water-Resources Investigations Report 914034, 1991,
pp. 475480.
Eychaner, J.H. & Stollenwerk, K.G.: Neutralization of acidic ground water near Globe, Arizona.
In:K.D. Schmidt (ed):Groundwater contamination and reclamation. Proceedings of a Symposium,
Tucson, Arizona, 1415 August, 1985, Bethesda, Maryland. American Water Resources Association,
1985, pp. 141148.
Ficklin, W.H., Love, A.S. & Papp, C.S.E.: Solid-phase variations in an aquifer as the aqueous solution
changes, Globe, Arizona.In: G.E. Mallard & D.A. Aronson (eds):U.S. Geological Survey Toxic
Substances Hydrology ProgramProceedings of the Technical Meeting, Monterey, California,
1115 March, 1991: U.S. Geological Survey Water-Resources Investigations Report 914034, 1991,
pp. 475480.

BUNDSCH_Book.indb 230

11/29/2011 12:32:13 PM

Integrating field observations and inverse and forward modeling 231

Fuller, C.C. & Harvey, J.W.: Reactive uptake of trace metals in the hyporheic zone of a miningcontaminated stream, Pinal Creek, Arizona.Environ. Sci. Technol. 34 (2000), pp. 11501156.
Glynn, P.D.: MBSSAS: A code for the computation of Margules parameters and equilibrium relations
in binary solid-solution aqueous-solution systems.Comput.Geosci.17 (1991a), pp. 907966.
Glynn, P.D.: Effect of impurities in gypsum on contaminant transport at Pinal Creek, Arizona.
In: G.E. Mallard & D.A. Aronson (eds):U.S. Geological Survey Toxic Substances Hydrology
ProgramProceedings of the Technical Meeting, Monterey, California, 1115 March, 1991: U.S.
Geological Survey Water-Resources Investigations Report 914034, 1991b, pp. 466474.
Glynn, P.D.: Solid-solution solubilities and thermodynamics: sulfates, carbonate and halides.In: C.N.
Alpers, J.L. Jambor & D.K. Nordstrom (eds): Sulfate minerals-crystallography, geochemistry and
environmental significance. Reviews in Mineralogy and Geochemistry 40, 2000, pp. 481511.
Glynn, P.D. & Brown, J.G.: Reactive transport modeling of acidic metal-contaminated ground water
at a site with sparse spatial information. In: C.I. Steefel, P. Lichtner & E. Oelkers (eds): Reactive
transport in porous media: general principles and application to geochemical processes. Reviews in
Mineralogy 34, 1996, pp. 377438.
Glynn, P.D. & Busenberg, E.: Dissolved gas and chlorofluorocarbon content of ground waters in the
Pinal Creek Basin, Arizona.In: D.W. Morganwalp & D.A. Aronson (eds): U.S. Geological Survey
Toxic Substances Hydrology ProgramProceedings of the Technical Meeting. Colorado Springs,
CO, 2024 September, 1993: U.S. Geological Survey Water-Resources Investigations Report 944015,
l994a, pp. 10431054.
Glynn, P.D. & Busenberg, E.: Unsaturated zone diffusion of carbon dioxide and oxygen in the Pinal
Creek Basin, Arizona. In: D.W. Morganwalp & D.A. Aronson (eds): U.S. Geological Survey Toxic
Substances Hydrology ProgramProceedings of the Technical Meeting. Colorado Springs, CO,
2024 September, 1993: U.S. Geological Survey Water-Resources Investigations Report 944015,
l994b, pp. 10551064.
Glynn, P.D. & Plummer, L.N.: Geochemistry and the understanding of ground-water systems.
Hydrogeology J. 13 (2005), pp. 263287.
Glynn, P.D., Engesgaard, P. & Kipp, K.L.: Use and limitations of two computer codes for simulating
geochemical mass transport at the Pinal Creek toxic-waste site. In: G.E. Mallard & D.A. Aronson
(eds).U.S. Geological Survey Toxic Substances Hydrology ProgramProceedings of the Technical Meeting, Monterey, California, 1115 March, 1991: U.S. Geological Survey Water-Resources Investigations
Report 914034, 1991, pp. 454460.
Glynn, P.D., Busenberg, E. & Brown, J.G.: Use of chlorofluorocarbons, dissolved gases and water
isotopes to characterize ground-water recharge in an aquifer contaminated by acidic, metal-laden
wastewater. U.S. Geological Survey Toxic Substances Hydrology ProgramProceedings of the
Technical Meeting, Charleston, South Carolina, 812 March, 1999Volume 1 of 3Contamination
from Hardrock Mining: U.S. Geological Survey Water-Resources Investigations Report 994018A,
1999, pp. 155162.
Harvey, J.W. & Fuller, C.C.: Effect of enhanced manganese oxidation in the hyporheic zone on basinscale geochemical mass balance. Water Resour. Res. 34:4 (1998), pp. 623636.
Herzer, J. & Kinzelbach, W.: Coupling of transport and chemical processes in numerical transport
models. Geoderma 44 (1989), pp. 115127.
Hydro Geo Chem Inc.: Investigation of acid water contamination along Miami Wash and Pinal Creek,
Gila County, Arizona. Cyprus Miami Mining Corporation, Claypool, AZ, 1989.
Kay, J.T., Conklin, M.H., Fuller, C.C. & ODay, P.E.: Processes of nickel and cobalt uptake by Mn-oxide
forming sediment in Pinal Creek.Environ. Sci. Technol. 35: 24 (2001), pp. 47194725.
LeBlanc, D.R.: Sewage plume in a sand and gravel aquifer, Cape Cod, Massachusetts. U.S. Geological
Survey Water-Supply Paper 2218, 1984.
Lichtner, P.C.: Continuum model for simultaneous chemical reactions and mass transport in
hydrothermal systems. Geochim.Cosmochim.Acta 49 (1985), pp. 779800.
Lichtner, P.C.: The quasi-stationary state approximation to coupled mass transport and fluid-rock
interaction in a porous medium. Geochim.Cosmochim.Acta 52 (1988), pp. 143165.
Lind, C.J. & Stollenwerk, K.G.: Alteration of alluvium of Pinal Creek, Arizona by acidic ground water
resulting from copper mining. In: D.W. Morganwalp & D.A. Aronson (eds): U.S. Geological Survey
Toxic Substances Hydrology ProgramProceedings of the Technical Meeting. Colorado Springs, CO,
2024 September, 1993: U.S. Geological Survey Water-Resources Investigations Report 94-4015, l994,
pp. 10891094.

BUNDSCH_Book.indb 231

11/29/2011 12:32:13 PM

232

Geochemical modeling of groundwater, vadose and geothermal systems

Mackay, O.M., Freyberg D.L., Roberts, P.V. & Cheny, J.A.: A natural gradient experiment on solute
transport in a sand aquifer I: Approach and overview of plume movement.Water Resour. Res. 22
(1986), pp. 20172029.
Neaville, C.C. & Brown, J.G.: Hydrogeology and hydrologic system of Pinal Creek Basin, Gila County,
Arizona. U.S. Geological Survey Water-Resources Investigations Report 934212, 1993.
Nordstrom, D.K.: Advances in the hydrogeochemistry and microbiology of acid mine waters. Int.
Geology Review 42 (2000), pp. 499515.
Nordstrom, D.K.: Modeling low-temperature geochemical processes. In: J.I. Drever (ed): Treatise
on geochemistry, v. 5, Surface and ground water, weathering, and soils. Amsterdam, Elsevier, 2004
(updated in 2007), pp. 3772.
Nordstrom, D.K.: Questa baseline and pre-mining ground-water quality investigation. 25. Summary of
results and baseline and pre-mining ground-water geochemistry, Red River Valley, Taos County, New
Mexico, 20012005. U.S. Geological Survey Professional Paper 1728, 2008.
Nordstrom, D.K., Plummer, L.N., Langmuir, D., Busenberg, E., May, H.M., Jones, B.F. & Parkhurst,
D.L.: Revised chemical equilibrium data for major water-mineral reactions and their limitations. In:
D.C. Melchior & R.L. Bassett (eds): Chemical modeling in aqueous system II, American Chemical
Society Symposium Series 416, 1990, pp. 398413.
Ortoleva, P., Auchmuty, G., Chadam, J., Hettmer, J., Merino, E., Moore, C.H. & Ripley, E.: Redox
front propagation and banding molalities. Physica 19D (1986), pp. 334354.
Parkhurst, D.L.: Users guide to PHREEQC--A computer program for speciation, reaction-path,
advective-transport, and inverse geochemical calculations. U.S. Geological Survey Water-Resources
Investigations Report 954227, 1995.
Parkhurst, D.L.: Geochemical mole-balance modeling with uncertain data. Water Resour. Res. 33:8
(1997), pp. 19571970.
Parkhurst, D.L. & Appelo, C.A.J.: Users guide to PHREEQC (Version2)A computer program
forspeciation, batch-reaction, one-dimensional transport, and inverse geochemical calculations. U.S.
Geological Survey Water-Resources Investigations Report 99-4259, 1999.
Parkhurst, D.L. & Charlton, S.R.: NetpathXL-An Excel interface to the program NETPATH. U.S.
Geological Survey Techniques and Methods 6-A26, 2008.
Parkhurst, D.L., Plummer, L.N. & Thorstenson, D.C.: PHREEQEA computer program for
geochemical calculations. U.S. Geological Survey Water-Resources Investigations Report 8096, 1980.
Parkhurst, D.L., Plummer, L.N. & Thorstenson, D.C.: BALANCE-A computer program for
geochemical calculations. U.S. Geological Survey Water-Resources Investigations Report 82-14, 1982.
Parkhurst, D.L., Kipp, K.L. & Charlton, S.R.: PHAST Version 2A program for simulating
groundwater flow, solute transport, and multicomponent geochemical reactions. U.S. Geological
Survey Techniques and Methods 6A35, 2010.
Petersen, N.P.: Geology and ore deposits of the Globe-Miami district, Arizona. U.S. Geological Survey
Professional Paper 342, 1962.
Plummer, L.N. & Glynn, P.D.: Radiocarbon dating in groundwater systems. Guidebook on dating of old
groundwaters, Chapter 4. International Atomic Energy Agency, in press.
Plummer, L.N., Wigley, T.M.L. & Parkhurst, D.L.: The kinetics of calcite dissolution in CO2-water
systems at 5 to 60C and 0.0 to 1.0 atm CO2. Amer. J. Sci. 278 (1978), pp. 179216.
Plummer, L.N., Parkhurst, D.L. & Thorstenson, D.C.: Development of reaction models for groundwater systems. Geochim.Cosmochim. Acta 47 (1983), pp. 665686.
Plummer, L.N., Prestemon, E.C. & Parkhurst, D.L.: An interactive code (NETPATH) for modeling
net geochemical reactions along a flow path. U.S. Geological Survey Water-Resources Investigations
Report914078, 1991.
Plummer, L.N., Prestemon, E.C. & Parkhurst, D.L.: An interactive code (NETPATH) for modeling
NET geochemical reactions along a flow PATH. Version 2.0.U.S. Geological Survey Water-Resources
Investigations Report 944169, 1994.
Plummer, L.N., Sanford, W.E. & Glynn, P.D.: Characterization and conceptualization of groundwaterflow systems. Guidebook on dating of old groundwaters, Chapter 2. International Atomic Energy
Agency, in press.
Postma, D. & Appelo, C.A.J.: Reduction of Mn oxides by ferrous iron in a flow system: column
experiment and reactive transport modeling. Geochim.Cosmochim.Acta 64 (2000), pp. 12371247.
Revesz, K., Coplen, T.B., Baedecker, M.J., Glynn, P.D. & Hult, M.: Methane production and
consumption monitored by stable H and C isotope ratios at a crude oil spill site, Bemidji, Minnesota.
Appl. Geochem. 10 (1995), pp. 505516.

BUNDSCH_Book.indb 232

11/29/2011 12:32:13 PM

Integrating field observations and inverse and forward modeling 233

Robertson, F.N.: Geochemistry of ground water in alluvial basins of Arizona and adjacent parts of
Nevada, New Mexico, and California. Regional Aquifer-System AnalysisSouthwest Alluvial
Basins, Arizona and Adjacent States. U.S. Geological Survey Professional Paper l406-C, 1991.
Steefel, C.I. & MacQuarrie, K.T.B.: Approaches to modeling of reactive transport in porous media. In:
C.I. Steefel, P. Lichtner & E. Oelkers (eds): Reactive transport in porous media: General principles
and application to geochemical processes. Reviews in Mineralogy 34, 1996, pp. 85129.
Stollenwerk, K.G.: Geochemical interactions between constituents in acidic groundwater and alluvium
in an aquifer near Globe, Arizona. Appl. Geochem. 9 (1994), pp. 353369.
Stollenwerk, K.G. & Eychaner, J.H.: Acidic ground water contamination from copper mining near
Globe, Arizona. In: B.J. Franks (ed): U.S. Geological Survey Toxic Substances Hydrology Program
Proceedings of the Technical Meeting, Pensacola, Florida, 2327 March, 1987: U.S. Geological Survey
Open-File Report 87109, 1987, pp. D19D24.
Tolle, S. & Arthur, G.V.: Aquifer restoration under the clean water act. In: G.E. Mallard & D.A. Aronson (eds): U.S. Geological Survey Toxic Substances Hydrology ProgramProceedings of the Technical
Meeting, Monterey, California, 1115 March, 1991: U.S. Geological Survey Water-Resources Investigations Report 914034, 1991, pp. 520523.
Van Cappellen, P. & Gaillard, J.F.: Biogeochemical dynamics in aquatic sediments.In: C.I. Steefel, P.
Lichtner & E. Oelkers (eds): Reactive transport in porous media: General principles and application
to geochemical processes. Reviews in Mineralogy 34, 1996, pp. 335376.
Villinski, J.E., ODay, P.A., Corley, T.L. & Conklin, M.H.: In situ spectroscopic and solution analyses of
the reductive dissolution of MnO2 by Fe(II). Environ. Sci. Technol. 35 (2001), pp. 11571163.
Winograd, I.J. & Robertson, F.N.: Deep oxygenated ground water: anomaly or common occurrence?
Science 216 (1982), pp. 12271230.

BUNDSCH_Book.indb 233

11/29/2011 12:32:13 PM

BUNDSCH_Book.indb 234

11/29/2011 12:32:13 PM