tracer-based characterization of a stimulation … · 2021. 2. 22. · examining the tracer...

DISS. ETH NO. 26248

TRACER-BASED CHARACTERIZATIONOF A STIMULATION-ENHANCED ROCK VOLUME

AND APPLICATION OF NOVEL DNA NANOTRACERSIN FRACTURED CRYSTALLINE ROCK

A thesis submitted to attain the degree of

DOCTOR OF SCIENCES of ETH ZURICH

(Dr. sc. ETH Zurich)

presented by

ANNIINA MARIA KAROLIINA KITTILÄ

Master of Science, University of Helsinki, Finland

Born on 31.12.1990

Citizen of Finland

accepted on the recommendation of

Prof. Dr. Martin O. Saar

Dr. Xiang-Zhao Kong

Prof. Dr. Martin Sauter

2020

“Kindness is the language which the deaf can hear and the blind can see.”

Mark Twain.

3

Abstract

Geothermal energy is one of the renewable energy sources receiving growing interest as

a consequence of increasing fossil fuel prices, low-carbon imperatives, and environmental

awareness. As traditional hydrothermal systems are scarce, gaining access to a broader

resource is fundamental in increasing the share of electricity generated from geothermal

systems. Drilling down to greater depths of about 5 kilometers gives us access to ubiqui-

tous hot rocks, where, however, permeability is low and there is scarcity of water in-place.

Thus, to engineer a commercially viable heat exchanger in low- to zero-permeability rock,

known as Enhanced Geothermal Systems (EGS), hydraulic stimulation technologies, such

as hydraulic fracturing or shearing, must be applied. Regarding such geothermal reser-

voir creation and its subsequent characterization, there exists a knowledge gap at the

intermediate deca-meter scale to understand i) the processes relevant for permeability en-

hancement, and ii) the properties of the subsurface heat exchanger and of the hydraulic

connections. This thesis contributes to improving our understanding of the hydrody-

namic changes in the fractured crystalline rock mass induced by hydraulic stimulation

experiments and the injection of hot water.

Firstly, in concert with solute tracers, I applied novel DNA-labeled silica nanoparticles

to investigate their transport properties in fractured crystalline rock. These nanoparticles,

with an approximate diameter of 166 nm, are labeled with unique DNA signatures and

encapsulated into silica spheres. The resulting nanotracers are identified based on their

DNA signature, but their transport properties can be equated with that of natural sand

particles. I observed that the stability of the recovered tracer response curves, i.e., whether

there are fluctuations between consecutive samples, is directly correlated with the injected

tracer mass. It is also evident that size exclusion, and potentially density effects, attenuate

the DNA nanotracer signal. These effects are manifested as a reduction in the following

parameters, in comparison to solute dye tracers: tracer recoveries, swept volumes, mean

residence times, and dispersion. However, lower detection limits and no susceptibility to

background concentrations promote the use of DNA nanotracers in tracer tomography

and in tracing particulate-bound contaminant transport.

By applying solute dye tracers before and after the hydraulic stimulations and the start

of hot water injection, I was able to place constraints on the evolution of preferential flow

paths and determine the changes in the tracer-swept volumes resulting from the thermo-

hydro-mechanical responses of the rock mass. Examining the tracer response curves

showed that spatial heterogeneities in the fracture network result in fluid flow channeling

and a wide distribution of residence times. As a consequence of the hydraulic stimulation

programs, tracer swept volumes increased considerably, i.e., between 43% and 316%. The

4

lack of a corresponding general trend in the other obtained moment analysis results,

that is, recovery, mean residence time, Gini coefficient, and second moment, is indicative

of spatial heterogeneities in the fractures that dominate fluid flow. The tomograms of

hydraulic conductivity, K, derived from tracer peak concentration arrival times before and

after the hydraulic shearing stimulation, showed that after the stimulation, fluid flow was

accessing pathways with higher K values. Likely due to new hydraulic connections, the

geometric mean of the computed K values increased. As the spatial distribution of flow

properties is not obtainable from temporal moments, using a tomographic approach to

complement the reservoir characterization can be decisive in estimating the performance

of a geothermal reservoir.

Finally, I observed that the thermo-mechanical response, induced by hot water injec-

tion, redistributed the fluid flow at the Grimsel Test Site (GTS) in Switzerland. This ob-

servation is supported by comparing solute dye tracer response curves and their temporal

moments from before and two weeks after the start of hot water injection. It is important

to note that, the total recovery of the tracers decreased significantly due to fluid losses to

the far field. The first fractures to be affected by reservoir stimulation and operational

activities are likely those carrying high flow rates and large fractions of injected fluid, so

that, when the permeability of these key flow paths increases (e.g., due to stimulation)

or decreases (e.g., via heat build-up), fluid flow is strongly redistributed. Understanding

the evolution of the preferential flow paths is crucial for the sustainable management of

EGS and other subsurface reservoirs. For that purpose, as it is shown in this thesis, the

analysis of tracer tests, by estimating the temporal moments of tracer response curves,

provides essential information on the hydrodynamic properties of geothermal reservoirs.

5

Zusammenfassung

Geothermie gehört zu den erneuerbaren Energien, welche derzeit aufgrund der steigenden

Preise für fossile Brennstoffe, der notwendigen Reduktion atmosphärischer Kohlendioxid-

Konzentrationen und des erhöhten Umweltbewusstseins der Bevölkerung wachsendes In-

teresse geniessen. Klassische hydrothermale Systeme sind selten, sodass man den Anteil

an Geothermie in der Stromproduktion nur dann signifikant erhöhen kann, wenn man

häufiger vorkommende geothermische Ressourcen in Anspruch nimmt. In Tiefen von ca.

fünf Kilometern im Untergrund ist ausreichend heisses Gestein vorhanden. Leider sind

die Permeabilitäten von solch tiefem Gestein meist sehr gering und es gibt sehr wenig Po-

renwasser. Möchte man nichtsdestotrotz das Gestein in diesen Tiefen als Wärmetauscher

benutzen, so muss man Fluid-Wegigkeiten im Gestein künstlich erzeugen. Dies geschieht

zumeist mit hydraulischer Stimulation des Gesteins, d. h. Klufterzeugung, und nennt sich

Enhanced Geothermal Systems (EGS). Allerdings gibt es auf der mittleren räumlichen

Dekameter Skala einige Wissenslücken bezüglich i) der Prozesse, die zur Permeabilitäts-

erhöhung führen und ii) der Eigenschaften des Wärmetauschers im Gestein und der hy-

draulischen Verbindungen, die durch die hydraulische Stimulation erzeugt werden. Diese

Dissertation trägt dazu bei, das Verständnis dafür, wie sich hydrodynamische Prozesse im

geklüfteten kristallinen Gestein aufgrund der hydraulischen Stimulation und der Injektion

von heissem Wasser verändern, zu verbessern.

Zusätzlich zu etablierten gelösten Markierstoffen (Farbstoffen) habe ich eine neuar-

tige Methode des Tracerversuchs zur Charakterisierung der Transporteigenschaften von

geklüftetem kristallinem Gestein angewandt. Diese Methode verwendet Nanopartikel mit

einem Durchmesser von ca. 166 nm, die mit einzigartigen DNA-Signaturen markiert und

in Siliziumdioxidkugeln verkapselt sind. Die resultierenden Nanotracer können mithilfe ih-

rer DNA-Kennzeichnung identifiziert werden, und ihr Transportverhalten kann mit dem

von natürlichen Sandpartikeln gleichgesetzt werden. Ich habe beobachtet, dass je weni-

ger Tracer-Masse injiziert wird, desto grösser sind die Fluktuationen (d. h. Unsicherhei-

ten) in der Tracer-Durchbruchskurve. Es ist sehr wahrscheinlich, dass Grössenausschluss-

und Dichteeffekte das DNA-Nanotracersignal abgeschwächt haben. Im Vergleich zu ge-

lösten Farbstofftracern waren die Rückgewinnungsraten, das vom Tracer durchflossenen

Gesteinsvolumen, die mittleren Verweilzeiten und die Dispersion der DNA-Nanopartikel

kleiner. Allerdings liegen die Vorteile der DNA-Nanotracer für den Einsatz in der Tracer-

Tomographie und in der Untersuchung von teilchengebundenem Schadstofftransport dar-

in, dass sie geringere Nachweisgrenzen und kein natürliches Rauschsignal haben.

Die Durchführung von Tracerversuchen vor und nach den hydraulischen Stimulationen

und vor der Injektion warmen Wassers, hat es mir ermöglicht, ein besseres Verständnis

6

dafür zu erlangen, wie sich die thermo-hydro-mechanische Reaktion des Gesteins auf die

Entwicklung der präferentiellen Fliesspfade und auf die Veränderung des vom Tracer

durchflossenen Gesteinsvolumens auswirkt. Die Auswertungen der Durchbruchskurven

zeigen, dass räumliche Heterogenitäten im Kluftnetzwerk zu präferentiellen Fliesspfaden

und zu einer grossen Bandbreite an Verweilzeiten führen. Die hydraulischen Stimulationen

führten zu deutlichen Zunahmen des durchflossenen Volumens, zwischen 43% und 316%.

Das Fehlen eines entsprechenden allgemeinen Trends in den anderen Ergebnissen mei-

ner Momentenanalyse, d. h. Rückgewinnungsrate, mittlere Verweilzeit, Gini-Koeffizient

und das zweite zentrale Moment, ist ein Hinweis darauf, dass räumliche Heterogenitäten

im Kluftnetzwerk die Fluidströmung dominieren. Die Tomogramme der hydraulischen

Durchlässigkeit, K, die aus den Ankunftszeiten der Konzentrationsspitze vor und nach

der Hydraulische Sher-Stimulation abgeleitet wurden, zeigten, dass nach der Stimulation

das Strömungsfeld sich auf Fliesspfade mit höheren K-Werten konzentriert hat. Das geo-

metrische Mittel der berechneten K-Werte stieg an, wahrscheinlich aufgrund neu entstan-

dener hydraulischer Verbindungen. Da die räumliche Verteilung der Strömungseigenschaf-

ten nicht aus den Momenten erzielbar ist, kann die Verwendung eines tomographischen

Ansatzes zur Charakterisierung eines geothermischen Reservoirs sehr wichtig sein.

Meine Beobachtungen lassen darauf schliessen, dass die durch die Heisswasserinjekti-

on induzierte thermomechanische Reaktion des Gesteins beim Grimsel-Testgelände in der

Schweiz das Strömungsfeld umverteilt hat. Diese Schlussfolgerung stützt sich auf den Ver-

gleich der Durchbruchskurven der gelösten Farbstoff-Tracer und ihrer Momente vor und

zwei Wochen nach der Heisswasserinjektion. Die Rückgewinnungsraten der Tracer sanken

dabei beträchtlich. Zu der Neuverteilung des Strömungsfelds tragen in erster Linie jene

grossen Klüfte, die die höchsten Durchflüsse aufweisen, bei. Diese Klüfte erweitern sich

während der Stimulation noch mehr bzw. verengen sich, wenn sie sich aufwärmen. Somit

verteilen sich die Stromlinien neu. Es ist wichtig für eine nachhaltige Energiegewinnung

durch EGS, diese Veränderungen der Strömung zu verstehen. Wie in dieser Dissertation

dargelegt, bekommt man durch Analyse der Tracerversuche, indem man die Momente der

Tracer-Durchbruchskurven bestimmt, essentielle Information über die hydrodynamischen

Eigenschaften des geothermischen Reservoirs.

Aus dem Englischen von Dr. Anozie Ebigbo

Acknowledgements

I sincerely thank Prof. Martin Saar for giving me the opportunity to pursue my PhD

in the Geothermal Energy and Geofluids group. Being the first PhD in the GEG group,

I was able to witness the joys, and pains, of the group’s growth, but looking back into

those four years, I am happy to have had this great opportunity to also grow, as a person

and as a scientist.

During the last year of my PhD, I had an extraordinary pleasure to have Dr. Xiang-

Zhao Kong becoming my direct supervisor. It is thanks to you that I was able to go

through writing the papers with such clearness and speed, when that was exactly what

was needed. I know you felt hesitant of starting to supervise me so close to the end of my

project, but I wish you to know that I am deeply grateful for the knowledge you shared

with me, the motivation and support you gave, and for always being there for all the

scientific discussions.

Although I’m finishing my PhD with Kong and Martin as my official supervisors, there

were several advisors and "unofficial supervisors" whose help has been pivotal: Dr. Anozie

Ebigbo, although for only a short duration of time that you were officially supervising me,

the advise and feedback you gave me were valuable in finding the right direction with the

analyses and doing some important decisions; Dr. Reza Jalali, big thanks for making the

field works at Grimsel possible and for the help you gave; Dr. Matthias Willmann, despite

not being able to complete the MRMT modelling that we were working on together, I

am very grateful for you being the source of motivation and support for me during a very

stressful period of my PhD; and finally, I am deeply honored for having had the chance

to work with Dr. Keith Evans. In particular, when it comes to the level of detail and

commitment towards scientific communication, you are my role model.

Special thanks go to Nils Knornschild for the invaluable technical support you gave

with the Grimsel tracer tests, and Dominique Ballarin for support in so many ways.

I also want to thank the Grimsel DUG-Lab team and particularly all of those kind

people I had a pleasure working with in Grimsel.

Further thanks go to Claudia, whose Master’s thesis topic on the DNA nanotracers

allowed me to share part of the PhD journey with her. I also thank those who helped

7

8

with all aspects of field work at Widen and Grimsel: Santos, Linwei, Fanny, Márk, Reto,

Gediminas, and Michela.

Very warm thanks go to Jin, Marina and Hoda, the original Pink GEG, for the endless

support and motivation. It makes me extremely happy to having being able to share the

PhD journey with you: It is not only your positivity and support that has been crucial

when there were challenges with the PhD, but it has also been a pleasure sharing stories

of motherhood, womanhood, and basically of anything ‘between heaven and the Earth’,

as a Finnish saying goes. Sooner than you notice, you will also be in the end of your

PhDs. I wish we find each other again, once our ways have departed to our new journeys.

I would like to thank my family and friends in Finland and the United States for their

support and encouragement.

Finally, most importantly, my lovely husband, without you this journey would have

been a thousand times more difficult. There are not enough words to thank you for the

unconditional love and support you have given, and continue to give to me. I am also

infinitely grateful for becoming a mother to our precious, sweet and lovely Sofia. You two

have given me the greatest moments and the most important things in life!

Hyvää yritetähän mutta priimaa pakkaa tulemahan!

Contents

Abstract 3

Zusammenfassung 5

Acknowledgements 7

1 Introduction 13

1.1 Fluid flow and mass transport in fractures . . . . . . . . . . . . . . . . . . 13

1.2 Geothermal resources . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15

1.3 Artificial tracer tests . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19

1.4 Motivation for the research . . . . . . . . . . . . . . . . . . . . . . . . . . 23

1.5 Thesis outline . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25

2 Field comparison of DNA-labeled nanoparticle and solute tracer trans-

port in a fractured crystalline rock 27

2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30

2.2 Materials and methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33

2.2.1 Tracers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33

2.2.2 Study site . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34

2.2.3 Experimental design . . . . . . . . . . . . . . . . . . . . . . . . . . 36

2.2.4 Moment analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39

2.3 Results and discussions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43

2.3.1 Effect of DNA nanotracer injection concentration . . . . . . . . . . 43

2.3.2 Tracer breakthrough curves . . . . . . . . . . . . . . . . . . . . . . 44

2.3.3 Attenuation and absence of DNA nanotracers . . . . . . . . . . . . 46

2.3.4 Residence time and the first arrival . . . . . . . . . . . . . . . . . . 47

2.3.5 Recovery and mean residence time . . . . . . . . . . . . . . . . . . 47

2.3.6 Transport processes . . . . . . . . . . . . . . . . . . . . . . . . . . 50

2.3.7 Swept pore volume . . . . . . . . . . . . . . . . . . . . . . . . . . . 52

2.3.8 Flow geometry and hydraulic connectivity . . . . . . . . . . . . . . 53

9

10 CONTENTS

2.4 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56

3 Characterization of the effects of hydraulic stimulation with tracer-

based temporal moment analysis and tomographic inversion 59

3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62

3.2 Site description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65

3.3 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67

3.3.1 Experimental design . . . . . . . . . . . . . . . . . . . . . . . . . . 67

3.3.2 Moment analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69

3.3.3 Tomographic inversion . . . . . . . . . . . . . . . . . . . . . . . . . 72

3.4 Results and discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75

3.4.1 Comparison of pre- and post-stimulation results . . . . . . . . . . . 75

3.4.2 Post-stimulation characterization . . . . . . . . . . . . . . . . . . . 80

3.4.3 Tomographic inversion . . . . . . . . . . . . . . . . . . . . . . . . . 84

3.5 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89

4 Solute tracer test quantification of the effects of hot water injection into

hydraulically stimulated crystalline rock 93

4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99

4.2 Test site . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102

4.3 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105

4.3.1 Tracer experiments . . . . . . . . . . . . . . . . . . . . . . . . . . . 105

4.3.2 Moment analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106

4.3.3 Temperature perturbations in a fracture . . . . . . . . . . . . . . . 108

4.4 Results and discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110

4.4.1 Residence time distributions . . . . . . . . . . . . . . . . . . . . . . 110

4.4.2 Redistribution of fluid flow . . . . . . . . . . . . . . . . . . . . . . 112

4.4.3 Estimation of fracture surface area . . . . . . . . . . . . . . . . . . 116

4.4.4 Data uncertainties . . . . . . . . . . . . . . . . . . . . . . . . . . . 119

4.5 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120

5 Summary and perspectives 123

A Appendix 127

A.1 Contribution to the risk report of the ISC experiment at the GTS . . . . . 127

A.1.1 Effects of tracer injection . . . . . . . . . . . . . . . . . . . . . . . 127

A.2 Supporting Information for Chapter 2 . . . . . . . . . . . . . . . . . . . . 136

A.3 Multirate Mass Transfer Model . . . . . . . . . . . . . . . . . . . . . . . . 140

A.4 Supporting Information for Chapter 3 . . . . . . . . . . . . . . . . . . . . 144

CONTENTS 11

A.5 Authored and co-authored publications and posters . . . . . . . . . . . . . 147

List of Figures 149

List of Tables 155

Bibliography 157

1Introduction

This thesis aims at contributing to the understanding of hydrodynamic changes in a

fractured crystalline rock volume induced by i) rock mass hydraulic stimulation and ii)

injection of hot water. I use conventional solute dye tracers and novel colloidal DNA

nanotracers to examine these changes in the connected pore volume of the rock mass

through cross-hole tracer testing, while also evaluating and validating the field application

of the DNA nanotracers in fractured crystalline rock. The field site is at the Grimsel Test

Site (GTS) in the Swiss Alps, where the tracer tests were conducted as part of a pre- and

post-stimulation characterization of hydraulic stimulation of the rock mass. Temporal

moments of the recorded tracer breakthrough curves are fundamental in evaluating the

hydrodynamic changes that took place in the stimulated rock mass at the GTS. The study

volume was a decameter-scale analogue to enhanced geothermal systems (EGS), albeit

with ambient temperatures of 13 °C, which is lower than in typical EGS reservoirs.

1.1 Fluid flow and mass transport in fractures

Fractured geological formations are ubiquitous, and fractures occur at all scales from

microscopic to continental (Streltsova, 1976; Berkowitz, 2002). They provide flow channels

for fluid flow and mass transport throughout different scales, and for this reason they are

of high importance in geo-engineering and hydrogeological applications. However, due to

heterogeneous aperture distributions, fluid flow usually does not take place over the entire

13

14 1 Introduction

cross-section of a fracture (Durham and Bonner, 1994; Watanabe et al., 2009; Becker and

Tsoflias, 2010; Guo et al., 2016). Neretnieks et al. (1982) showed that fluid dispersion

due to channeling occurs in fractures, as fluid flows in channels at different velocities.

More recently, Guihéneuf et al. (2017) demonstrated significant flow channeling effects

on mass transport at the fracture scale and for large distances at the network scale, and

Dou et al. (2018) showed that variable aperture alone leads to large fluid residence times,

which is, however, enhanced by the presence of flow eddies. On a larger scale, it has

been found that fracture density does not necessarily correlate with local transmissivity

(Le Borgne et al., 2007; Brixel et al., 2020), and that only a small portion of fractures

may contribute to fluid flow. These observations raise a question concerning hydraulic

connectivity in a fracture network, as highly fractured domains may not actually be well-

connected (Long and Billaux, 1987; Berkowitz et al., 2006; Lei et al., 2017). Because of

flow channeling, hydraulic connections may thus be unpredictable. Furthermore, some

portions of the fracture surface area are not efficiently swept by the fluid, which is an

issue for geothermal systems (Fu et al., 2016). Additionally, in other applications, flow

channeling may introduce preferential flow paths through which contaminants, such as

radionuclides, can travel (Rasmuson and Neretnieks, 1986; Kurosawa et al., 2006).

Considering fully saturated, single phase water flow in fractures, mass (solutes and

particulates) transported through such fractures experiences hydrodynamic dispersion

(Neretnieks et al., 1982; Roux et al., 1998), which consists of i) Taylor dispersion, due to

a parabolic velocity profile, ii) macrodispersion, due to spatial velocity contrasts, resulting

in channeling (Salamon et al., 2006a), and iii) molecular diffusion due to diffusive exchange

across a concentration gradient. At the reservoir scale, the effects of Taylor dispersion

and molecular diffusion are diminished, because the length scale of these mechanisms

is on the same order of magnitude as a fracture aperture (Robinson and Tester, 1984).

In strongly heterogeneous media, like fractured aquifers, fluid flow velocity distributions

are asymmetric, with a bias towards low velocities (Willmann et al., 2008). This is a

consequence of aperture heterogeneities that give rise to unevenly distributed fluid flow in

fractures, which in turn results in preferential flow paths (Moreno et al., 1988; Stoll et al.,

2019). Such a distribution of fluid flow also leads to anomalous mass transport behavior,

observed as a pronounced tailing of tracer breakthrough curves. This anomalous behavior

has been represented by a matrix diffusion model, which was used to account for sorption

kinetics (Neretnieks et al., 1982; Neretnieks and Rasmuson, 1984), and was widely used

in fractured aquifers (Hadermann and Heer, 1996). However, Becker and Shapiro (2000)

showed that in fractured aquifers, the tailing may not be caused by diffusion processes

alone. They argued that transport, occurring in primary fractures or channels with rapid

advection and in secondary fractures with slow advection, can lead to a significantly

1.2 Geothermal resources 15

heterogeneous flow field, which can result in heavy tailing that may falsely be interpreted

as being caused by matrix diffusion. The underlying process was coined mass transfer

(into low-permeable zones) or slow advection (Zinn and Harvey, 2003; Willmann et al.,

2008; Fiori and Becker, 2015; Henri and Fernàndez-Garcia, 2015; Tuykhova andWillmann,

2016).

The ubiquitous heterogeneities of geometrical and biogeochemical properties, occur-

ring in geological formations, can have a significant effect on mass transport. This implies

that resolving the pore space in detail is not possible, and that heterogeneities at all scales

can affect mass transport behavior (Berkowitz et al., 2006). As dispersion depends on

the scale of the observation (Gelhar et al., 1992), constant center of mass velocity and

dispersion coefficients are not sufficient to quantify the temporal evolution of transported

mass (Berkowitz, 2002). Such scale-dependent behavior is called anomalous transport,

or non-Fickian transport (Haggerty et al., 2000; Berkowitz et al., 2006; Willmann et al.,

2008), and it is displayed as early time arrivals and slowly decreasing concentrations of

mass (tracer) recovered at an observation point. Regarding the temporal concentration

distribution of an initial tracer pulse, the characteristic shape of the tracer breakthrough

curves (BTC) is thus non-Gaussian for non-Fickian transport (Becker and Shapiro, 2000;

Kosakowski and Smith, 2004; Cherubini et al., 2014). In addition, long tailing and multi-

ple breakthrough curve peaks are observed when aperture heterogeneities exist (Moreno

et al., 1988; Siirila-Woodburn et al., 2015), as discussed above. Other factors influencing

recovery and the shape of the tracer BTCs include fracture orientation with respect to

flow direction (Johnston et al., 2009; Edery et al., 2016) and reactive mass transport

(Tompson, 1993; Berkowitz, 2002; Hawkins et al., 2018).

1.2 Geothermal resources

Geothermal energy is heat stored within the Earth, and the primary sources of this heat

originate from the planet’s formation and the decay of the long-lived, naturally radioactive

isotopes of uranium, thorium, and potassium. At the Earth’s surface, an average heat flow

of 101mW/m2 (at the oceanic crust) and 65mW/m2 (at the continental crust), dissipates

into space (Pollack et al., 1993). Furthermore, the temperature of the Earth increases with

depth by about 25-30 °C/km on average. However, at some locations, higher than average

temperatures, based on the normal geothermal gradient, occur at shallow depths. The

source of this may be due to i) topography, ii) high-temperature magmatic intrusions, or

iii) thermal blanketing of deep rocks by a thick formation of rocks that have low thermal

conductivity (e.g., clays). Heat is transferred predominantly by conduction in the Earth’s

crust, but locally, permeable fault zones can enable groundwater to descend to depths of

16 1 Introduction

several kilometers, where groundwater is heated, and via advection (convection) brought

back up (Mock et al., 1997; Dickson and Fanelli, 2002; Saar, 2011; WEC, 2013). An

exploitable geothermal system thus requires a heat source, a reservoir, and a circulating

fluid. The heat source is discussed above, and the reservoir is a volume of hot and

permeable rocks, from which the heat is extracted by a circulating fluid (Dickson and

Fanelli, 2002; Evans, 2015).

The need to reduce carbon dioxide (CO2) emissions and to protect the environment

have led to the growing interest towards renewable energy sources. From renewable

sources, wind and solar are weather dependent, although growing rapidly, and biomass

fuels are subject to feedstock limitations. However, geothermal resources offer constant

energy which is available on demand (Younger, 2014). Furthermore, geothermal energy

utilization and the development of geothermal energy-producing technologies benefit from

increasing fossil fuel prices and low-carbon imperative (WEC, 2013; Zheng et al., 2015).

Additionally, as a consequence of the overwhelmingly negative public opinion regarding

nuclear power, particularly encouraged by the aftermath of large nuclear incidents such

as the Chernobyl and Fukushima nuclear accidents, geothermal development has received

more attention and further support (Pioro et al., 2019; Yasukawa, 2019).

Geothermal resources are commonly categorized into hydrothermal and petrother-

mal, or engineered/enhanced geothermal systems (Mock et al., 1997; Dickson and Fanelli,

2002; WEC, 2013). Hydrothermal resources are characterized by convective circulation

of naturally occurring liquid water or steam in a permeable medium, at depths of ap-

proximately 1-4 km and up to 350 °C. Both vapour- and liquid-dominated systems are

being used to generate electricity. While liquid-dominated systems are the most widely

distributed in the world, as vapour-dominated systems are somewhat rare, well-known

examples of vapour-dominated systems are the Larderello field in Italy, the Geysers field

in California, and the Matsukawa field in Japan. Petrothermal systems, however, do not

have natural permeability to enable fluid circulation.

The heat source is the only element that needs to be naturally present, thus the other

elements, i.e., a permeable reservoir and the heat-extracting fluid, can be artificially in-

troduced. This is the concept of Enhanced Geothermal Systems (EGS), or originally, Hot

Dry Rock (HDR) systems (Tester et al., 2006; Olasolo et al., 2016). To extract heat at

economically viable rates, these geothermal systems require stimulation to increase perme-

ability. Two different mechanisms of such hydraulic stimulation techniques are hydraulic

shearing and hydraulic fracturing, which aim to significantly increase the permeability

of pre-existing natural fractures or intact low permeability rocks, respectively (Tester

et al., 2006; Amann et al., 2018). However, artificially creating permeability and thus an

underground (geothermal) reservoir is still associated with high levels of uncertainty re-

1.2 Geothermal resources 17

garding reservoir production (Evans et al., 2005; Tester et al., 2006), including i) effective

generation of an artificial reservoir that exhibits sufficiently high permeabilities (i.e., low

reservoir impedances), ii) maintaining low reservoir impedances over decades despite cir-

culating hot, mineral-laden fluids through the reservoir, iii) optimal perforation locations

of injection and production wells to ensure sufficient well-reservoir connectivities (Luo

et al., 2013; Evans, 2015), or iv) inducing seismicity (Giardini, 2009; Evans et al., 2012).

Establishing a geothermal reservoir that eventually enables cost-effective geothermal

energy extraction, has proven to be a major challenge for EGS projects worldwide (Evans,

2015). From a technical point of view, stimulations should result in a sufficiently well-

connected fracture network that has sufficient surface area for conductive heat transfer

between the hot geothermal formation and the cooler circulating fluid, without inducing

earthquakes that could be felt by the local population. Note that acoustic emissions,

or microseismicity, is inevitable during the stimulation process (Majer and Doe, 1986;

Ishida et al., 2019). Ultimately, the success of an EGS is determined by its economic

viability, i.e., its ability to produce electric power at an acceptable Levelized Cost of

Electricity (LCOE), which currently is below 0.10 $/kWh (Clauser and Ewert, 2018).

Therefore, economic feasibility of an EGS requires sufficiently high production flow rates

(Tester et al., 2006; Lee et al., 2019). Also, with larger separations between the injection

and production wells, higher surface areas for heat exchange can be achieved, preferably

exploiting a large reservoir volume. However, with only little fluid in-place in an EGS,

it is important that the flow and pressure fields of the injection and production wells

interact (Evans, 2015). Thus, the production flow rate is determined by the injection

flow rate and by the resistance of the flow paths to flow. The latter is described by the

reservoir impedance, which is defined as the pressure difference between the injection and

the production wells needed to circulate a certain fluid volume (Kolditz and Clauser, 1998;

Evans, 2015). At high impedances, power consumed by pumps to maintain the target

production flow rate becomes too high, reducing the net power generated and increasing

LCOEs.

In the following, three key challenges in creating and operating a geothermal reservoir

are further discussed. The first challenge is a highly public one, as geothermal energy

projects are often planned near urban areas. Here, the most concerning aspect is induced

seismicity, which is inherently connected to creating an enhanced geothermal system. The

seismic risk is high near urban areas due to the combined effect of several factors; the

seismic hazard, population exposure, and vulnerability of the infrastructure. To manage

and mitigate the seismic hazard and risk, employing so-called traffic light systems is part

of the best practises in EGS projects (Ellsworth, 2013; Wiemer et al., 2015). However,

18 1 Introduction

the more conservative a traffic light system is, the lower is the likelihood of achieving an

economically viable heat exchanger through rock mass stimulation (Wiemer et al., 2015).

The second issue concerns fracture-flow based geothermal heat extraction. In a frac-

tured medium, heat is transferred through conduction and advection. However, heat

can only be economically “mined” by advection through fractures, which results in ther-

mal drawdown propagating faster in the fractures than in the intact rock (Kolditz, 1995;

Reimus et al., 2020). Consequently, geothermal systems must rely on conductive heat

transfer between the surrounding hot rock and the fluid circulating in fractures, in order

to enable commercially adequate production temperatures over reservoir lifetimes. This

is mainly influenced by the surface area available for heat transfer, but also by fracture

spacing and the volume of the rock available for heat extraction (Wu et al., 2008; Reimus

et al., 2020). If the fluid travel times between the injection and the production wells are

short, the production temperatures may rapidly decrease, together with the geothermal

power plant performance. Such rapid thermal drawdown in the production well reduces

power generating efficiency and increases the LCOE (Sun et al., 2018; Lee et al., 2019;

Reimus et al., 2020).

The third issue to be highlighted is the potential change in permeability and pore/fracture

connectivity during the long-term operation of the stimulated reservoir, which can occur

through mineral dissolution and precipitation (Tester et al., 2006; Taron and Elsworth,

2009; Yasuhara et al., 2011; Grimm Lima et al., 2019). The injection of cold water into

the reservoir at geochemical disequilibrium results in the redistribution of mineral mass

through mineral dissolution and precipitation. Consequently, the fracture aperture and

the reservoir permeability can change. Pressure dissolution of minerals can happen at

contacting asperities and fracture void surfaces. Mineral precipitation can occur on void

surfaces in the fractures or at recovery and injection wells (scaling) (Tester et al., 2006;

Taron and Elsworth, 2009; Yasuhara et al., 2011). The evolution of fracture apertures,

due to these processes, may hinder advective heat transfer, thereby potentially reducing

the reservoir’s lifetime. An option to improve well injectivity or productivity, after a per-

meability decrease due to mineral precipitation, is to use chemical stimulation, where acid

is injected into the formation to dissolve soluble minerals. However, this is only effective

in the near-wellbore region (Nami et al., 2008).

Regardless of such a long list of known challenges in creating and operating an en-

hanced geothermal system, and considering that there has been almost four decades since

the world’s first EGS test facility (Fenton Hill HDR test site, New Mexico, USA (Evans,

2015; Kelkar et al., 2016)) was opened, there still exist no economically viable EGS op-

erations anywhere in the world (Evans, 2015). The European flagship project is the

Soultz-sous-Forêts EGS in France, where the technical feasibility of EGS in fractured

1.3 Artificial tracer tests 19

crystalline rock was demonstrated (Schill et al., 2017). However, although power is gen-

erated at Soultz, the production flow rates are insufficient, the flow impedances higher

than desired, and the LCOE exceedingly high (Evans, 2015; Koelbel and Genter, 2017).

The performance of a geothermal reservoir is largely affected by how well the rock

volume is exploited by the circulating fluid. Reinjection of the produced, cooled fluid, to

maintain reasonable pore-fluid pressures, is a central component of EGS, and it is also

of high importance when aiming for the sustainable management of geothermal resources

(Axelsson, 2013). The reinjection of colder fluid, however, is often associated with certain

problems (Horne, 1985; Tester et al., 2006), the most significant of them being the highly

likely precipitation of dissolved minerals due to unfavourable fluid chemistry, which re-

duces the permeability and fluid injectivity of the system (Horne, 1985; André et al.,

2006; Taron and Elsworth, 2009; Yasuhara et al., 2011). Another concern, regarding rein-

jection, is the loss of fluid, resulting in long-term net injection and potential difficulties

in availability of make-up water (Horne, 1985; Evans, 2015). The reinjection of cold fluid

also affects the evolution of fluid flow channeling and the overall flow pattern. This is

linked to thermal drawdown in the geothermal reservoirs (Taron and Elsworth, 2009; Guo

et al., 2016; Pandey et al., 2017). For example, Fu et al. (2016) found that concentration

of fluid flow in cooled zones, a form of flow channeling at the reservoir scale, is inevitable.

Such phenomena are caused by thermal contraction of the rock, which increases the per-

meability of the cooled fractures (Taron and Elsworth, 2009; Fu et al., 2016; Pandey et al.,

2017).

The performance of a geothermal system can be investigated with tracer tests (San-

juan et al., 2006; Axelsson, 2013; Buscarlet et al., 2015; Ayling et al., 2016). This includes

the estimation of swept volume, which is an important parameter describing the physi-

cal performance of an EGS (Evans, 2015; Grant, 2016). This is the main topic of this

dissertation.

1.3 Artificial tracer tests

Tracer tests are a well-established method to assess hydraulic connectivities and fluid

flow processes (Leibundgut et al., 2009). In porous and fractured media, tracer tests

enhance our understanding of the hydrodynamic processes in the subsurface, and of the

changes in fluid movement in response to various activities, such as the stimulation of

the rock mass or the circulation of cooler water through geothermal reservoirs (Winick

et al., 2015). Tracer test configurations range from single-well (Haggerty et al., 2001;

Ghergut et al., 2014) to multiple-well (Marschall and Lunati, 2006; Ayling et al., 2016),

applying dissolved or particulate (solid) substances or heat as tracers (Leibundgut et al.,

20 1 Introduction

2009). The duration of a tracer test is largely dependent on the scale of the investigated

system, and so the inter-well tracer tests are typically conducted under forced hydraulic

head gradient conditions to benefit from earlier tracer breakthrough than under natural-

gradient conditions. This configuration is also applied, because a well-defined flow field

can be established using forced-gradient tests, particularly if the fluid flow converges to

a well. This form of tracer tests also benefit from a reduction in tracer mass losses in the

studied system (Pedretti et al., 2013; Guihéneuf et al., 2017).

Leibundgut et al. (2009) distinguished six main groups of artificial tracers (i.e., tracers

that are actively introduced into the system), namely fluorescent, salt, radioactive, acti-

vatable, and advanced solute tracers, and drifting solid particles (temperature being part

of the advanced tracer group). An ideal tracer, suitable for physical aquifer characteriza-

tion, should be conservative and only represent the water flow, whereas non-ideal tracers,

such as solids, can be useful in special applications. In the following, certain tracers or

tracer types are discussed in more detail:

• Of solute tracers, fluorescent tracers are the most popular, due to their low cost,

easy handling, and often non-toxic nature. Among numerous fluorescent dye tracers,

uranine, eosine, and sulforhodamine B are well-known. The detection of dye tracers

is based on the emission of light impulses, which occur when an energy source

excites an electron from a lower energy state to a higher state. The electron then

releases the excess energy in the form of light, i.e., fluorescence, and reverts to the

lower energy state. Each fluorescent dye tracer has specific excitation and emission

spectra that peak at a characteristic wavelength (Leibundgut et al., 2009).

• Temperature, or heat, is also a readily available and cost-effective tracer, which can

easily yield continuous time series. However, conductive heat transfer into the rock

matrix can be significant, and inaccurate results for physical aquifer characterization

may be obtained in low groundwater velocity systems, dominated by conduction

(Saar, 2011; Giambastiani et al., 2013; Somogyvári et al., 2017). While the average

fracture aperture has a significant effect on conservative tracer transport, for heat

transport the effect is negligible. For accurate heat transport prediction, the heat

transfer area needs to be adequately constrained (Hawkins et al., 2018).

• Colloids, such as the novel DNA nanotracers used here, are solid tracers with a

diameter ranging from 10−3 to 10µm. Due to their finite size, density, and surface

charge effects, colloids are not ideal tracers, but they can facilitate, for example,

particulate-enhanced transport of contaminants. Colloid attachment processes and

retention mechanisms, due to fracture wall effects, are unpredictable, which high-

lights the importance of understanding colloid transport in fractures. Also, colloid

1.3 Artificial tracer tests 21

transport is sensitive to the flow path geometry and the flow direction (Vilks and

Bachinski, 1996; Albarran et al., 2013; James et al., 2018).

• The application of smart tracers, i.e., tracers whose reactions under various physic-

ochemical and geothermal conditions are known, is a relatively new concept. How-

ever, these new tracer tools may help determine the thermal history of a reservoir,

or predicting two-phase mass and heat transport (Ghergut et al., 2007; Wu et al.,

2008; Redden et al., 2010; Nottebohm et al., 2012).

Despite the wide range of different types of tracers available, their utilization is lim-

ited due to physicochemical conditions (i.e., temperature, pH, UV light) and the ability to

distinguish tracers from each other, i.e., finding mutually compatible tracers (Leibundgut

et al., 2009). For example, if the interference of the excitation and emission spectra be-

tween different fluorescence dye tracers is not taken into account, false tracer signals can

be anticipated. Figure 1.1a shows fluorescence dye tracer sample concentrations (ppb)

of uranine and sulforhodamine B, obtained from pH-adjusted laboratory measurements

in my research. An increase in sulforhodamine B concentration is visible from sample

number 4 onwards, whereas uranine concentration begins to clearly increase from sample

number 14. Figure 1.1b shows the corresponding in-situ fluorometer signals in milli-

volts (mV) for sample numbers 1-11 for lamp 1, namely L1, measuring uranine, and L2,

measuring sulforhodamine B. The uncalibrated in-situ fluorometer measurements show

interference in L1 measurements from the presence of sulforhodamine B, when actually

no uranine was yet present.

The number of available traditional tracers (e.g., dye tracers) in the case of repeat

or multi-tracer tests is limited because of the restrictions described in the previous para-

graph. However, novel DNA nanotracers can overcome this problem, as virtually an

unlimited number of DNA tracer codes can be produced, due to the information stor-

age capacity of DNA, where DNA nanotracers can be produced with virtually unlimited

distinct signatures. The DNA is encapsulated in spherical silica particles (Figure 1.2)

(Paunescu et al., 2013), which implies that regardless of the DNA signature, all DNA

nanotracers exhibit largely identical surface and transport properties. The diameter of

the DNA nanoparticles is approximately 150-180 nm, resembling natural silica particles

regarding their surface properties. Furthermore, the concentration of the DNA can be

measured very accurately with ultralow detection limits by quantitative Polymerase Chain

Reaction (qPCR) (Mikutis et al., 2018). However, the DNA nanotracers are, as mentioned

above, of a finite size, compared to conventional solute tracers. This prevents them from

entering small pores. As a result, DNA nanotracers tend to stay in preferential fluid flow

pathways, resulting in a faster concentration peak arrival time than conventional solute

tracers (Section 2).

22 1 Introduction

0 5 10 15 20 25Sample number

0

20

40

60

Sam

ple

conc

entr

atio

n [p

pb]

Uranine, L1Sulforhodamine B, L2

0 5 10 15 20 25 30 35 40L2 [mV]

0

20

40

60

L1 [m

V]

b)

a)

Figure 1.1 – (a) Concentrations of uranine (blue) and sulforhodamine B (orange) insamples. The grey area indicates samples where uranine concentrations are below thebackground level. (b) Plot of fluorometer measurements from lamp 1, L1 (uranine),and L2 (sulforhodamine B) in mV, corresponding to the samples (n=11) outlined inthe grey area in a).

Figure 1.2 – STEM micrograph (left) and structural illustration (right) of the DNAnanotracer.

1.4 Motivation for the research 23

1.4 Motivation for the research

The Swiss Energy Strategy 2050 (Swiss Federal Office of Energy, 2018) aims for the reduc-

tion of the national energy consumption, while increasing energy efficiency. Furthermore,

the strategy states that the potential of new renewable energies, such as wind, solar,

biomass, and geothermal, requires further investigation. Additionally, according to the

same Energy Strategy, the construction of new nuclear power plants will be banned. To

meet the energy demand, as the electricity generated from nuclear energy is phased out,

supply from renewable sources, including geothermal, needs to be increased. Thus, with

an ambitious goal of increasing the supply of electricity from deep geothermal energy

from 0 to 4.4 TWh by 2050, not only legal, but financial and research efforts are required

(Geothermie Schweiz, 2019). The research aspect of this task requires that geothermal

resources are found, characterized, and developed, assuming that deep geothermal re-

sources, the target for future geothermal exploitation, are ubiquitous in, for example,

crystalline rock at about 4-5 km depth (Evans et al., 2014), and also assuming that heat

can be extracted at competitive LCOEs (Section 1.2).

The Swiss roadmap for exploiting deep geothermal resources (Evans et al., 2014)

states that in order to make progress with the creation of underground heat exchangers

(e.g., through the creation of EGS reservoirs), scaled analogue experiments need to be

conducted at a depth of ≤1 km in Deep Underground Laboratories (DUG-Lab). The

Grimsel DUG-Lab (near Grimsel Pass, Switzerland), which is located in granitic rock at

a depth of about 450m, was such an experimental facility to host such a well-controlled

and densely monitored experiment. Between 2015 and 2017, a comprehensive in-situ

Stimulation and Circulation (ISC) experiment was conducted at the Grimsel DUG-Lab,

which addressed various issues and questions related to reservoir creation in EGS sys-

tems (Amann et al., 2018). The decameter scale ISC experiment was divided into three

phases (Figure 1.3); pre-stimulation characterization, stimulation, and post-stimulation

characterization. Tracer tests were part of the pre- and post-stimulation characterization

phases, and they are highlighted in Figure 1.3. These tracer tests, using solute dye tracers

and DNA nanotracers, are the focus of this thesis. The aim of those tracer tests was to

understand the hydrodynamic changes in the fractured crystalline rock mass, induced by

hydraulic stimulation and the injection of hot water.

It is important to highlight that the Grimsel ISC experiment was a joint effort of

a number of researchers, as can be inferred from the long list of research investigations

conducted at the Grimsel Test Site (GTS) (Figure 1.3). My responsibility during the

Grimsel ISC experiment was the design, implementation and interpretation of tracer

tests that used solute dye tracers as well as the novel DNA nanotracers to characterize

24 1 Introduction

Drilling

Stress measurements

Characterization• Tunnel and core mapping• Geophysical borehole logs

(OPTV, ATV, electric resistivity, spectral gamma, full-wave sonic logs)

• Hydraulic tests (i.e., single-and cross-hole)

• Geophysical characterization (i.e., GPR, active seismics, single- and cross-hole and cross-tunnel)

• Tracer tests (dye, thermal tracer, and DNA nanotracer)

Monitoring• Strain and tilt• Pore pressure• Temperature• Microseismics

Stimulation• Stimulation of existing

fractures and fault zone• Hydraulic fracturing in

massive rock

Pre-stimulation phase Stimulation phase Post-stimulation phase

Monitoring• Pressure and flow rates in

active injection borehole• Pressure in passive injection

borehole• Microseismicity in tunnels and

boreholes• Pressure in boreholes and

tunnel surface• Strain in boreholes and tunnel

surface• Tilt at the tunnel surface• Dislocations in active injection

borehole using an acoustic televiewer

Characterization• Geophysical borehole logs in

the injection boreholes (electrical resistivity, spectral gamma, full-wave sonic logs)

• Hydraulic tests (i.e., single-and cross-hole)

• Tracer tests (dye, thermal tracer, and DNA nanotracer)

• Geophysical characterization (i.e., GPR, active seismics, single- and cross-hole and cross-tunnel)

Figure 1.3 – An overview of the ISC experiment test phases at the Grimsel Test Site(GTS), modified from Amann et al. (2018).

the stimulation-enhanced rock volume. Furthermore, several of the salt tracer tests, as

part of the Ground Penetrating Radar (GPR) surveys, were conducted in concert with the

dye tracer tests. Additionally, apart from the Grimsel ISC experiments, I co-supervised

a Master’s thesis project (of Claudia Deuber), which consisted of field and laboratory

work and was focused on investigating the transport properties of the DNA nanotracers

in porous media.

In this thesis, I approach the characterization of the stimulation-enhanced rock volume

by primarily applying a method of moments to the recovered tracer breakthrough curves

(BTCs). Calculating the temporal moments and other parameters associated with the

tracer BTCs is critical to quantify the hydraulic properties of the tracer-swept connected

flow paths and the overall performance of the formation. Of major interest when creating

and operating a reservoir, are the mean residence time, the tracer-swept volume, the

fraction of fluid lost, and fluid flow channeling, which can all be determined with tracer

tests. These parameters are necessary when evaluating and predicting the performance

of a geothermal reservoir.

1.5 Thesis outline 25

1.5 Thesis outline

The thesis is composed of three main chapters. Tracer tests play a key role throughout

the thesis. In Chapter 2, I start by investigating the transport properties of the novel

colloidal DNA nanotracers in fractured crystalline rock. I do so by comparing the tracer

breakthrough curves and the statistical quantities of the derived residence time distri-

bution curves of the DNA nanotracers and of the conventional solute dye tracers. This

study also provides encouraging results for advancing the use of DNA nanotracers in hy-

drogeological applications. In Chapter 3, I use solute dye tracers to delineate the changes

in the hydrodynamic properties of the rock mass, induced by stimulation at the Grimsel

Test Site (GTS). To meet this objective, the tracer tests were conducted before and after

the hydraulic stimulation experiments. I again examine the temporal moments of the

tracer residence time distribution curves and apply tomographic inversions to visualize

the spatial distribution of hydraulic conductivities. In Chapter 4, the solute dye tracer

tests are conducted before and two weeks after the start of hot water injection at the GTS.

I obtain evidence for fluid flow redistribution due to the thermo-mechanical response of

the rock mass to heat build-up, and I also observe a decrease in injection flow rate with an

increase in fluid injection pressure, both of which can be attributed to thermo-mechanical

expansion of the rock. I further attempt to constrain the fracture geometries by estimat-

ing fracture surface areas from tracer-swept volumes and from temperature perturbations

at monitoring locations. Finally, Chapter 5 concludes this thesis, summarizing the main

findings of Chapters 2 to 4 and discussing the importance of the results for geothermal

EGS reservoir creation and operation. Additionally, I conclude the thesis by suggesting

topics of interest for future research regarding the application and interpretation of tracer

tests in geothermal studies.

2Field comparison of DNA-labeled nanoparticle

and solute tracer transport in a fracturedcrystalline rock

Published as:

A. Kittilä, M.R. Jalali, K.F. Evans, M. Willmann, M.O. Saar, and X.-Z. Kong (2019),

Field comparison of DNA-labeled nanoparticle and solute tracer transport in a fractured

crystalline rock, Water Resources Research, doi: 10.1029/2019WR025021.

27

https://agupubs.onlinelibrary.wiley.com/doi/10.1029/2019WR025021

Abstract

Field tracer experiments were conducted to examine tracer transport properties in a

fracture-dominated crystalline rock mass at the Grimsel Test Site, Switzerland. In the

experiments reported here, both the DNA nanotracers and solute dye tracers were simul-

taneously injected. We compare the transport of DNA nanotracers to solute dye tracers

by performing temporal moment analysis on the recorded tracer breakthrough curves

(BTCs) and estimate the swept volumes and flow geometries. The DNA nanotracers, ap-

proximately 166 nm in diameter, are observed to travel at a higher average velocity than

the solutes, but with lower mass recoveries, lower swept volumes, and less dispersion.

Moreover, size exclusion and potentially, particle density effects, are observed during the

transport of the DNA nanotracers. Compared to solute tracers, the greatest strength of

DNA nanotracers is the demonstrated zero signal interference of background noise during

repeat or multi-tracer tests. This work provides encouraging results in advancing the use

of DNA nanotracers in hydrogeological applications, for example, during contaminant

transport investigations or geothermal reservoir characterization.

29

30 2 Field comparison of DNA-labeled nanoparticle and solute tracertransport in a fractured crystalline rock

2.1 Introduction

It is often necessary in environmental and earth sciences to determine and parameterize

hydraulic connections in the subsurface. However, formulating conceptual models of flow

and transport through fractured rock is particularly difficult due to the heterogeneous

structure of such systems, and the challenge to directly observe the connections within

the fracture network. Tracer tests are a useful tool for this purpose as they can provide

vital insights into the flow and transport properties of the hydraulic conduits (Davis et al.,

1980; Leibundgut et al., 2009; Lange et al., 2011; Saar, 2011). Particularly important in

fractured rocks, tracer tests allow the location of connecting in- and outflow points to be

identified, as well as providing an estimate of their hydraulic and transport properties,

and relative significance (Becker and Shapiro, 2000; Kowalsky et al., 2012; Vogt et al.,

2012; Pedretti et al., 2013; Ayling et al., 2016; Shook and Suzuki, 2017). However, due to

the heterogeneous nature of the fractured media, there are often several connections that

one may want to study separately. Finding suitable tracers to conduct such multiple-

inlet tests without the signals interfering is difficult. Additionally, the dissolved species

of conventional tracers may persist in detectable quantities within the flow system for a

long time (Mueller et al., 2014; Bero et al., 2016; Liao et al., 2018), thereby impeding

repeat tracer tests.

DNA nanotracers do not suffer from the aforementioned problems and limitations,

as they can be designed with a unique signature (Mahler et al., 1998; Sharma et al.,

2012; Aquilanti et al., 2013; Mora et al., 2015; Pang et al., 2017; Liao et al., 2018).

Consequently, repeat tracer tests and tracer tomography are possible without problems

from tracer interference. Moreover, the interpretation of the tracer breakthrough curves

(BTCs) is easier, as all unique DNA signatures (i.e., tracers) have the same transport

properties, corresponding to those of spherical silica particles. This type of novel artificial

tracer is the focus of our investigation, which presents, to our knowledge, the first field

application of DNA nanotracers in fractured crystalline rock for tracing purposes.

Previous field experiments examined the transport characteristics of DNA nanotracers

in porous media as compared to those of solute tracers (Kong et al., 2018). In those

experiments, the DNA nanotracer technology was implemented for the first time in an

unconsolidated aquifer. The results demonstrated their utility for imaging a subsurface

reservoir by travel-time based tomography (Somogyvári et al., 2016), using multiple DNA-

labeled nanoparticles to determine hydraulic parameter fields in porous media.

Flow and transport in fractured media are, however, fundamentally different from that

in porous media. It is unusual for a breakthrough curve from a tracer test in a fracture

network to exhibit the Gaussian shape that is characteristically obtained for homogeneous

2.1 Introduction 31

porous media which can be adequately predicted with the classical advection-dispersion

equation (ADE). The distribution of hydraulic properties within fractures, which is mainly

influenced by spatial aperture heterogeneities, leads to channeling and preferential flow

paths (Guo et al., 2016). Tracer tests in such fractured media typically result in break-

through curves with fast arrival times, sudden increases in concentration, and/or long

concentration tails (Becker and Shapiro, 2000; McKay et al., 2000; Kosakowski, 2004).

Additionally, depending upon the degree of fracture connectivity, a tracer can arrive

significantly sooner or later than expected, or even not appear at all. Rasmuson and

Neretnieks (1986), Watanabe et al. (2009), and Guo et al. (2016) found that most of the

fluid flow typically occurs over only a small fraction of a single fracture’s nominal area,

with concomitant tortuosity of the flow paths.

DNA nanotracers are considered colloidal tracers (Stumm, 1993; Mikutis et al., 2018).

Studies in fractured media (Vilks et al., 1997; Becker et al., 1999; McKay et al., 2000;

Albarran et al., 2013) have shown that colloids are transported with higher average veloc-

ity than water molecules and solute tracers. Because of their size, colloids cannot reach

the walls of the fractures where fluid flow velocities are lowest, and thus their average

transport velocity is larger than that of solutes. The laboratory study of flow through

a fracture in a granodiorite core by Albarran et al. (2013) showed that colloids are in-

creasingly retained in the fracture as water flow rate decreases, and that the larger the

size of the colloid, the faster it is transported and the lower is its recovery. Numerous

studies suggest that these transport properties appear to be universally true for colloids,

regardless of the transport medium (Knapp et al., 2000; Zheng et al., 2009; James and

Chrysikopoulos, 2011; Chrysikopoulos and Katzourakis, 2015). Since DNA nanoparticles

essentially constitute colloids, we might anticipate similar transport behavior for the DNA

nanotracers used in this study.

In this paper, we present the first field-scale experiments that use DNA-labeled sil-

ica nanoparticles as nanotracers in a fractured crystalline rock mass. We investigate

the transport behavior and characteristics of the DNA nanotracers by conducting tracer

experiments in a decameter-scale field site, located at the Grimsel Test Site (GTS) in

the Swiss Alps. This study is thus an extension of the porous media investigations of

Kong et al. (2018) and Mikutis et al. (2018) to fractured media. Our investigation of

the DNA nanotracers is based on a comparison with classical solute dye tracers, uranine

and sulforhodamine B. Our approach is to analyze lower-order temporal moments of the

experimental data. We use only the lowest-order temporal moments because they are the

most informative (Leube et al., 2012), allowing determination of tracer recovery, mean

residence time, and the degree of spreading from the center of mass. In addition, the


swept volume and the fluid flow geometry can also be estimated by the moment analysis

(Shook and Forsmann, 2005).

2.2 Materials and methods 33

2.2 Materials and methods

2.2.1 Tracers

The DNA nanotracers used in this study were developed by the Functional Materials

Laboratory in the Institute of Chemical and Bioengineering, ETH Zurich in Switzerland,

and were purchased from the company Haelixa GmbH. The particle synthesis and their

stability and method of detection have been described in detail in previous studies (Kong

et al., 2018; Mikutis et al., 2018). In short, the DNA nanotracers were synthesized by

encapsulating double-stranded DNA molecules into amorphous silica spheres, thereby iso-

lating the DNA from the environment and providing improved tracer stability (Paunescu

et al., 2013; Mikutis et al., 2018). The sphere-shaped DNA nanotracer particles used

in this study had an average diameter of 166 nm, a negative surface charge (ζ-potential

between -14.4 and −29.9 mV) (Figure A.7 and Table A.2 in the Appendix), and a particle

density of about 2.1 g cm−3. A total of seven nanotracers, each encoded with a different

DNA sequence (Table A.3), were used in this study.

In order to identify the DNA code and to subsequently determine the tracer concen-

tration in a water sample, the DNA molecules encoded in the silica nanoparticles need

to be extracted. The non-destructive release of the DNA molecules is done by dissolving

the silica particles with a fluoride-containing etching solution (Paunescu et al., 2013).

Subsequently, the released DNA is analyzed with quantitative real-time polymerase chain

reaction (qPCR), which is based on an exponential signal amplification (Paunescu et al.,

2013; Mikutis et al., 2018). This amplification yields a large detection range for DNA

concentrations from about 0.1 ppt to 1000 ppm, and this permits ultra-sensitive DNA

detection, theoretically to a single DNA molecule (Pang et al., 2017; Liao et al., 2018).

To quantify the number of tracer particles, the DNA threshold cycle Cq (qPCR output

signal) has to be correlated to the particle concentration. Dilution curves of known con-

centrations of each of the tracers are prepared in water collected at the experimental site

to ensure that the qPCR efficiency is identical between samples and dilution curves (Fig-

ure A.8). The measured threshold cycle values of individual samples are then converted

to the tracer concentration using the previously obtained dilution curves, this way linking

the measured tracer content to a standard curve of a specific nanotracer.

The principal focus of this study is to compare the transport of the DNA nanotracers

with traditional and well-established dye tracers, uranine and sulforhodamine B. During

the field tests, the DNA nanotracers were mixed with the dye tracers and injected si-

multaneously into the borehole intervals. For this reason, prior to the field experiments,

possible interferences between the fluorescent dyes and the qPCR quantification of the

DNA nanotracer were evaluated. It was observed that: i) no binding of uranine and


sulforhodamine B to the DNA nanotracer particles occurs and thus there is no loss of

these dyes upon removal of the DNA nanotracer from solution; and ii) the fluorescence of

the dyes is not inhibited by the fluorescence-based qPCR quantification of the DNA when

the concentrations of uranine and sulforhodamine B are less than 0.5 ppm and 50 ppm

respectively (Mikutis et al., 2018).

2.2.2 Study site

The tracer experiments reported in this paper were conducted as part of a pre- and

post-stimulation hydraulic characterization of the In-Situ Stimulation and Circulation

(ISC) experiment (Amann et al., 2018) which is part of the Deep Underground Geother-

mal Laboratory (DUG-Lab) at the GTS in Switzerland. The core purpose of the ISC

experiment was to perform well-controlled and well-monitored hydraulic stimulation ex-

periments at the decameter scale, and to gain insights into the key processes relevant for

the development of Enhanced Geothermal Systems (EGS). To meet this goal, extensive

hydrogeological, geological, and geophysical characterizations of the experimental vol-

ume have been carried out both before and after the stimulation experiments (Giertzuch

et al., 2018; Jalali et al., 2018b; Krietsch et al., 2018). In total, nine tracer tests using dye

tracers, DNA nanotracers, and/or salt tracer were conducted during the characterization

phases (Figure 2.1b). In Tests 1 and 4, the DNA nanotracers were used together with the

conventional solute dye tracers uranine and sulforhodamine B. These two tests are the

focus of this study. In the other tests, only dyes, or dyes with salt were used. Possible

interferences from previous tests (i.e., Tests 2 and 3) were considered minor, because i)

there were three months between Tests 2 and 4 that both used uranine, and ii) eosine

and Tinopal CBS-X used in Test 3 were injected in a configuration to avoid potential

influence of eosine on the use of uranine in Test 4.

The GTS, operated by the Swiss National Cooperative for the Disposal of Radioactive

Waste (Nagra), is located in the Swiss Alps at 1733m a.s.l. with 400-500m of overburden.

The host rock of the ISC experiment, known as Grimsel granodiorite, is intersected by

shear zones, brittle fractures, and lamprophyre dykes (Keusen et al., 1989; Amann et al.,

2018). Four of the shear zones are recognized as S1-type (ductile, NEN-SWS striking,

SE-dipping) and two as a younger S3-type (brittle-ductile, E-W striking, S-dipping) (Fig-

ures A.6 and 2.1a). The lamprophyres are mostly parallel to the S3 shear zones. The host

rock typically has a very low fracture density (zero to three fractures per meter). How-

ever, a highly fractured zone with about 20 brittle fractures per meter extends between

the two S3 shear zones (Jalali et al., 2018a,b).

The GTS lies in a saturated rock without significant water discharge into the tunnels.

Water mainly circulates in the shear zones and along lamprophyre contacts. Consequently,


2016 2017

Test 1 Test 2

Test 3Test 4

Test 5Test 6

Test 7Test 8

Test 9

Shearing

Fracturing

May Sep May Sep

a

b)

Figure 2.1 – a) Projection of the boreholes and intervals in the DUG-Lab (Krietschet al., 2018). b) Timeline of different tracer tests as well as the hydraulic stimulationphases. In Test 1, tracers were injected into INJ2-int3 and INJ2-int4, and monitored inINJ1 and the AU Tunnel. In Test 4, tracers were injected into INJ1-int2 and INJ1-int4,and monitored in PRP1, PRP2, INJ2-int4 and the AU Tunnel. One of the main shearzone planes (S3.2) is shown, and the intersections of S3 in the AU Tunnel are visualizedwith dark green disks. All four S1-type structures, and the two S3-type structures areshown in Figure A.6.


Table 2.1 – Interval information for Tests 1 and 4 (see also Figure 2.1).

Interval starta enda length Trans. fracture outflowd pressuree[m] [m] [m] [m2 s−1] depthb [m] IDb,c [Lmin−1] [kPa]

INJ1-int4 27.67 28.19 0.52 3.7E-07f 27.78 S3 0.061i 644jINJ1-int3 29.09 45.00 15.91 N/A several S1, S3 0.037iINJ1-int2 38.51 39.03 0.52 2.0E-07g 38.69 S1 635jINJ2-int4 22.89 23.41 0.52 4.0E-06f 23.14 S3 0.093j 107iINJ2-int3 24.31 24.83 0.52 1.8E-08f 24.96 S3 773iPRP1-int3 23.20 25.20 2.00 9.4E-07h several S3 0.048jPRP1-int2 28.90 32.00 3.10 3.4E-07h several S3 0.043jPRP2-int2 21.40 27.00 5.60 1.5E-07h several S3 0.022jPRP2-int1 40.00 44.98 4.98 N/A several S1, S3 0.083jAU Tunnel – – – N/A – S3 1.0i , 0.980jaFrom the top of the borehole. b(Krietsch et al., 2018). cThe shear structure the fracturebelongs to. dOutflow of the producing monitoring locations, which were always open tothe atmosphere. eThe mean injection pressure (absolute) at the top of the borehole. Theatmospheric pressure was not measured during Test 1, but based on the authors’experience, it typically had a value of approximately 82 kPa. However, during Test 4 theatmospheric pressure was on average 30 kPa. fBefore hydraulic shearing stimulation,constant head injection test (Jalali et al., 2018b). gFrom injectivity after hydraulicshearing stimulation, stimulation interval HS2 (38.00-40.00m) (Doetsch et al., 2018a).hBefore hydraulic shearing stimulation, pulse tests (Brixel et al., 2018), post-stimulationvalues not available. iTest 1. jTest 4.

the average water discharge into the GTS is low (Keusen et al., 1989). There is no

large discharge into the AU Tunnel from the S1 and S3 shear zones, although an average

outflow of approximately 100 ml min−1 enters the gallery through the brittle fracture zone

between the two S3 shear zones (Jalali et al., 2018b). This observation is in accord with

the conclusion of Wenning et al. (2018) that the brittle fractures between the S3 shear

zones dominate fluid flow. The transmissivity in the brittle fracture zone is about 10−12

to 10−6 m2 s−1 (Keusen et al., 1989), whereas the average transmissivity of the intact rock

matrix is less than 10−13 m2 s−1 (Jalali et al., 2017).

2.2.3 Experimental design

Two separate tracer experiments, namely Tests 1 and 4, were completed. All of the

tracer experiments (Figure 2.1b) were performed under steady-state conditions, which

were reached by injecting tracer-free water for approximately 17.5 hours before Test 1, or

for approximately 56 hours before Test 4. The tracers were injected into the rock mass

through separate 4mm ID steel lines, which were connected to isolated intervals in the

injection boreholes (Table 2.1). The intervals were isolated with hydraulic packers that

were 0.9m in length when inflated. We used both manually-operated and drill-powered

syringes to inject the tracers into the fractures through injection lines.

Test 1 was conducted in May 2016 and was intended to characterize the rock mass

prior to hydraulic stimulation. Two different tracer mixtures were prepared for injection,


each containing a solute dye tracer and one or multiple DNA nanotracers. The details of

the mixtures are as follows, where the number in parentheses indicates the mass of each

tracer in milligrams (mg) in the mixture:

• The first mixture (Test 1A) contained uranine (10 mg) and four differently-encoded

DNA nanotracers, namely PT-2 (400 mg), DAP-3 (100 mg), JS-1 (20 mg) and

AM-1 (5 mg), with a total volume of 1.0 L. This mixture was injected into the rock

mass at borehole INJ2 interval 4 (INJ2-int4) with an average injection rate of about

1.1 L min−1. INJ2-int4 (Table 2.1) contained a highly transmissive fracture with an

S3 orientation (designated S3.2) (Krietsch et al., 2018) which was known to have a

good connection to an outlet in the AU Tunnel.

• The second mixture (Test 1B) contained sulforhodamine B (1.15 mg) and DNA

nanotracer GM-2 (400 mg) with a total volume of 0.120L. This mixture was injected

into the rock mass at INJ2 interval 3 (INJ2-int3, Table 2.1) with an average injection

rate of about 0.1 L min−1.

The four different DNA nanotracers in Test 1A were used to investigate the effect of

injected tracer mass on the transport and recovery of the particles. Tracer-free water was

injected into both INJ2 intervals for approximately 25 hours after the tracer injections.

The sampling locations of Test 1 are given in Table 2.1 and were: INJ1-int4 that contains

a lamprophyre dyke that was moderately transmissive, INJ1-int3, which was a large

interval of approximately 16m in length that extended from 29.09m to the bottom of

the borehole (not shown in Figure 2.1a), and an outflow point in the wall of the AU

Tunnel (Figure 2.1a). Information on the intervals used in both Tests 1 and 4 is given in

Table 2.1.

Test 4 was conducted in April 2017, after the hydraulic shearing stimulation injections

but before the hydraulic fracturing stimulation tests had been performed. Two different

tracer mixtures were used:

• The first mixture, (Test 4A), contained sulforhodamine B (9.75 mg) and DNA

nanotracer GR-3 (200 mg), with a total volume of 0.975 L. This mixture was injected

into the reservoir at borehole INJ1 interval 4 (INJ1-int4) with an average injection

rate of 0.599 L min−1. INJ1-int4 includes a meta-basic dyke with an S3 orientation

(designated S3.1) (Krietsch et al., 2018). It corresponds closely to interval HS4 of

the shearing stimulation program (27.2-28.2m) (Doetsch et al., 2018a).

• The second mixture, (Test 4B), contained uranine (9.75 mg) and DNA nanotracer

GR-1 (200 mg), with a total volume of 0.975 L. This mixture was injected into the

reservoir at INJ1 interval 2 (INJ1-int2) with an average injection rate of 0.634 L min−1.


INJ1-int2 includes a fracture at 38.69m which was identified as part of an S1 struc-

ture (designated S1.2) (Krietsch et al., 2018). It was included in interval HS2 of the

shearing stimulation program (38.00-40.00m) (Doetsch et al., 2018a).

The injection of tracer-free water was continued for 50 hours after the tracer injections

were completed. The sampling locations of Test 4 were: PRP1 intervals 2 and 3, PRP2

intervals 1 and 2, INJ2 interval 4, and the AU Tunnel outflow point (Figure 2.1a).

During the tests, water samples were regularly collected at the monitoring locations

using 60ml non-transparent plastic (PE-HD) bottles. At every monitoring location, water

was allowed to flow out into the atmosphere from flow lines connected to the intervals. The

outflows were measured regularly, and the harmonic mean of those measurements is shown

in Table 2.1. Further, downhole pressures in the injection intervals were measured through

closed, saturated lines that extended from the intervals to the top of the borehole (note the

average pressures listed in Table 2.1 are measured in the gallery rather than downhole).

During Test 1, an automatic sampler was installed at the AU Tunnel monitoring location

to facilitate overnight collection of water samples. Unfortunately, we were unable to

collect water samples from the borehole INJ1 monitoring intervals during the night, which

impacted the analysis of the DNA nanotracers arriving at borehole INJ1.

The collected water samples were first analyzed by the company Haelixa GmbH with

support from the Functional Materials Laboratory at ETH Zurich, using qPCR (Roche

LightCycler 96 Instrument) to quantify the amount of the DNA nanotracers. Prior to

applying the qPCR, the DNA was recovered from 1 mL aliquot of each vortexed sam-

ple by dissolving the silica particles with buffered fluoride solution (0.23 g NH4FHF

(pure, Merck) plus 0.19 g NH4F (puriss, Sigma-Aldrich) in 10mL Milli-Q water yield-

ing ∼25’000 ppm F-ions), which was diluted to a 1% concentration and added to each

water sample. The recovered DNA, without any purification in 5µL solution, was then

mixed with qPCR reagents (10µL master mix, 3µL PCR grade water, and 2µL of a

0.5µM primer stock) for the qPCR analysis. With qPCR, the DNA threshold cycle, Cq,

was measured and correlated to particle concentration with dilution curves prepared for

each individual tracer in water collected at the GTS (Figure A.8). Furthermore, samples

of the injected tracer mixtures were analyzed to exactly determine the injected concen-

trations. After the qPCR measurements, the dye tracer concentrations of these water

samples were measured with a Luminescence Spectrometer (Perkin Elmer, LS 50 B). In

addition to the laboratory measurements, the dye tracer concentrations were measured in

situ with flow-through fluorometers (GGUN-FL30) to obtain a continuous concentration

history. The fluorometers measure the intensity of the emitted fluorescent light in pre-

defined wavelengths in millivolts (mV). The light intensity was then directly calibrated


and converted to the dye concentration in ppb, based on the dye concentration in samples

measured in the laboratory.

2.2.4 Moment analysis

Following the approach of Shook and Forsmann (2005) and Shook and Suzuki (2017), the

time-series tracer concentrations, c (t), at a monitoring location were normalized as age

distribution functions E (t) [T-1],

E (t) =c (t) ρqout

Minj, (2.1)

where t [T] is time, ρqout [MT-1] is the mass production rate at the monitoring location, ρ

[ML-3] is the density of the water sample, qout [L3T-1] is the volumetric outflow flux, and

Minj [M] is the mass of the injected tracer. For unit conversion, the tracer concentration,

c (t), in parts per billion (ppb), needs to be given as kg/109kg in Equation (2.1). This

normalization thus accounts for the different tracer injection masses and mass production

rates. Plotting E (t) over the recorded time yields the residence time distribution (RTD)

curve. Statistical parameters such as the temporal moments of the RTD curves can then

be calculated to describe the transport properties of the tracer in a specific reservoir.

Because of time limitations, the tests were terminated before the tracer concentrations

returned to the pre-test values. Since the temporal moments obtained from the RTD

curves are time-dependent parameters, it is important that the concentration-distribution

profiles are analyzed in a consistent manner. This consistency can be achieved by the

extrapolation of early-ended concentration profiles (Marschall and Lunati, 2006; Ayling

et al., 2016; Shook et al., 2017). Consequently, where the recorded RTD curves suggested

an exponential decline, they were extrapolated using an exponential function fitted to the

tail of the curves prior to the moment analysis (Shook et al., 2017) (Table A.1).

In this study, temporal moments derived from the RTD curves were used to determine

the mean residence time of the tracer and the variance of the travel times in the system,

the pore volume swept by the tracers, and the overall geometry of the flow paths (Robinson

and Tester, 1984; Luo et al., 2008; Shook and Suzuki, 2017). The nth temporal moment

m∗n at location x of an RTD curve is defined by

m∗n =

∞∫0

tnE (x, t) dt . (2.2)

Given this definition, the following information can be obtained from the analysis of the

moments:


• The recovered mass m [M] of the tracer is provided by:

m = Minj

∫ ∞0E (t) dt. (2.3)

• The recovery, R [%], of a tracer is obtained from the zeroth temporal moment (if

m∗0 equals unity, 100% of tracer is recovered).

• The normalization of the first temporal moment by the zeroth temporal moment

gives the mean residence time t∗ [T]:

t∗ = m∗1/m∗0. (2.4)

• The normalized and centralized second temporal moment, m2,c [T2], defines the

response of the tracer signal or the degree of spreading about the center of mass:

m2,c = m∗2/m∗0 − (t∗)2. (2.5)

• Similar to the approach of Shook and Forsmann (2005), the pore volume, Vp [L3],

swept by the tracer, can be calculated as:

Vp = Rt∗qinj, (2.6)

where qinj is the volumetric injection rate.

Note that higher-order moments (n ≥ 3) place a significant weight on the slowly

decreasing, low concentrations at the tail of the curve (Harvey and Gorelick, 1995). In

addition, the higher-order temporal moments are prone to large errors if derived from

noisy concentrations measured in the field (Leube et al., 2012). As we were not able to

record these low-concentration tails of the BTCs, and must rely on extrapolation, the

higher-order moments are thus omitted from the moment analysis in the current study.

The heterogeneity of the flow paths can be inferred by interpreting the Lorenz curve

and calculating the associated Gini coefficient, which is derived from the age distribution

function. Originally developed to measure income inequality (Lorenz, 1905), in this study

the Lorenz curves are determined from the relationships between the flow capacity, F ,

and the storage capacity, Φ, where

F (t) =

t∫0

E (τ) dτ

∞∫0

E (t) dt

(2.7)


and

Φ (t) =

t∫0

E (τ) τdτ

∞∫0

E (t) tdt

. (2.8)

In a Lorenz curve, the cumulative percent of a quantity (population in the original

study of Lorenz, but here, storage capacity, which is the time-weighted reservoir volume

seen by the tracer at time t) is typically plotted against the cumulative proportion of

observations (originally wealth held by the percentages of the population, but here, the

fraction of the tracer recovered in the production well through that volume, i.e., flow

capacity) (Lorenz, 1905; Shook, 2003; Shook and Forsmann, 2005; Hao et al., 2012; Ayling

et al., 2016; Shook and Suzuki, 2017). The primary assumptions for the analyses of flow

and storage capacities are: i) fluid flow is at steady-state and thus the swept pore volume

and flow geometry do not vary with time and; ii) the tracers behave conservatively (Shook

and Forsmann, 2005). Although the particulate tracers (here the DNA nanotracers)

might deviate from the ideal conservative tracers, depending on the groundwater velocity

field and the density difference between tracer and groundwater, the analyses of flow and

storage capacities could capture the different behaviors between the DNA nanotracers and

solute tracer, apart from the insights delivered by the comparison of their breakthrough

curves.

The F −Φ curve typically has a convex shape, where the deviation of the curve from

a diagonal line, termed the line of equality, is a measure of heterogeneity. The F − Φ

diagram facilitates the determination of what fraction of the pore volume contributes to

what fraction of the fluid flow. Therefore, the F −Φ diagram provides an estimate of the

volume-averaged heterogeneity or channeling of the flow paths. This approach has been

applied to the interpretation of tracer tests in geothermal reservoir studies (Buscarlet

et al., 2015; Winick et al., 2015; Ayling et al., 2016). In this study, the differences in the

F−Φ diagrams obtained from the same injector–producer pairs using different tracers, i.e.,

solute dyes and colloidal DNA nanotracers, provide insights into the different transport

mechanisms of the employed tracers, as discussed above.

The Gini coefficient, G, associated with the F − Φ curve, is expressed as (Shook and

Forsmann, 2005)

G = 2

1∫0

F dΦ− 1

2

. (2.9)

G characterizes the shape of the F − Φ curve and varies between 0 for a homogeneous

flow field (i.e., flow is equally distributed over the swept volume and is plotted along the

line of equality) and 1 for a heterogeneous flow field (i.e., only a small fraction of swept


volume contributes to the overall fluid flow). G is based on a ratio analysis, so that neither

absolute values nor the cause of the heterogeneity can be inferred from it. However, G

can be calculated independently for different tracers and for different tracer tests, and so

can be used to evaluate how flow path heterogeneities differ among them.

2.3 Results and discussions 43

2.3 Results and discussions

Although the presented tracer Tests 1 and 4 were conducted before and after the hydraulic

shearing (HS) stimulation experiment, respectively (Figure 2.1b), here we do not attempt

to interpret changes in the hydraulic properties of the study rock volume resulting from

the stimulations using the tracer test results, because between the two tests i) the PRP

boreholes were drilled and completed, ii) a thermal tracer experiment was conducted

in INJ2 (Doetsch et al., 2018a), and iii) tracers were injected into different boreholes.

As mentioned in the Introduction, the focus of this work is on comparing the transport

behavior and characteristics of the DNA nanotracers with those of the solute dye tracers

within individual test.

2.3.1 Effect of DNA nanotracer injection concentration

As mentioned in Section 2.2.3, during Test 1A, four DNA nanotracers (PT-2, DAP-3,

JS-1, and AM-1) were injected at INJ2-int4. Aside from DNA-signature (Table A.3),

they differ in the mass of tracer injected, PT-2 being the greatest. The masses are given

in Figure 2.2, together with the mass-normalized breakthrough curves (BTCs) for all

four nanotracers measured in the outflow in the AU Tunnel. The BTCs of all four show

similar characteristics, although there are differences in variability, the curves for tracers

with larger injected mass appearing to be more stable. The degree of fluctuation, i.e.,

the stability of a BTC, can be quantitatively described using the BTC autocorrelation

with a lag distance of one sample point (inset of Figure 2.2): the closer the lag-one

autocorrelation is to one, the smaller the fluctuations in the BTC. It is evident that

larger injected mass leads to greater stability of the curve.

Overall, Figure 2.2 shows that the different DNA nanotracer BTCs are comparable,

but that the BTC quality depends on the injected mass of the DNA nanotracer. Measure-

ment variations between samples and replicates can occur for low particle concentrations

due to the redistribution of the small number of DNA nanotracer particles. For example, a

20% difference in particle concentration between two samples can result if one contains 36

particles and the other contains 44 particles. The discrete nature and insufficient number

of nanotracer particles in the samples may have caused a significant sample-to-sample

variation when subsamples and replicates were prepared for the qPCR quantification.

However, other factors may also have influenced the DNA nanotracer transport (Wang

et al., 2019), yielding the erratic behavior of the BTCs. These factors include 1) clogging

of pores, where the simultaneously injected particles compete for filtration/retention sites

or open pathways and 2) size distribution of the differently labeled DNA nanotracers

(Figure A.7). However, we argue that the effect of the size distribution is minor, as the


0 500 1000 1500Time [min]

10-4

10-3

10-2

10-1

100

101

Nor

mal

ized

con

cent

ratio

n[p

pb /

Inje

cted

mas

s]

PT-2 (400mg)DAP-3 (100mg)JS-1 (20mg)AM-1 (5mg)

5 20 100 400Injected mass (mg)

0

0.5

1A

utoc

orre

latio

n w

itha

lag

dist

ance

of 1

0.9062

0.8819

0.52680.5302

Figure 2.2 – Comparison of the BTCs of DNA nanotracers PT-2, DAP-3, JS-1, andAM-1, injected together in INJ2-int4 and monitored in the AU Tunnel in Test 1A. Theconcentrations are normalized to the injected mass of the respective DNA nanotracer.Each sample was analyzed in triplicate, and the error bars present standard deviationsof the replicates. Inset: The concentration fluctuation is described using the autocor-relation of a BTC with a lag distance of one. The autocorrelation was performed onlywith the data points shared by all four BTCs, consisting of 23 data points.

size distributions of PT-2 and AM-1 are almost identical. From the four DNA nanotracers

injected at INJ2-int4, only the BTC of PT-2 was taken for further analysis, because the

fluctuations between consecutive sample points in the BTCs of DAP-3, JS-1, and AM-1

imposes high uncertainties for the extrapolation.

2.3.2 Tracer breakthrough curves

Figure 2.3 shows the normalized breakthrough curves (i.e., tracer concentrations in ppb

over the injection concentration, C0 =Minj

Vinj, where Vinj is the volume of the injected tracer

mixture) from Tests 1A, 4A, and 4B. The concentrations are plotted as a function of time

in minutes since the injection of the tracers. Corrections have been made to account for

the time the tracers spent travelling through the injection and sampling tubes so that

only the travel-time in the fracture system is considered. Furthermore, the background

concentration has been subtracted for the dye tracer concentration.

For Test 1A, in which a mixture of uranine and DNA nanotracer PT-2 were injected

into INJ2-int4, only the BTCs observed at the AU Tunnel outflow are displayed. No

BTCs could be obtained from the two monitoring intervals in borehole INJ1 as the tracer

signal strengths were poor. This is in part due to the low flow rate of 0.037 L min−1 and

0.061 L min−1 produced by INJ1-int3 and INJ1-int4, respectively, and the 16m length of

the latter interval which would have severely diluted any inflowing tracer. In contrast,


Figure 2.3 – Normalized breakthrough curves of dye tracers and DNA nanotracersfrom Tests 1A, 4A, and 4B. Only measured concentrations (i.e., no extrapolated data)are plotted. Line colors indicate the monitoring locations of breakthrough curves (forreference, see Figure 2.1).

the outflow at the AU Tunnel was 1.0 L min−1. No plausible BTCs were detected from

simultaneously conducted Test 1B, where sulforhodamine B and the DNA nanotracer

GM-2 were injected into interval INJ2-int3.

For Test 4A, where sulforhodamine B and DNA nanotracer GR-3 were injected at

INJ1-int4, breakthrough curves were captured at the monitoring locations of PRP1-int3,

PRP2-int2, PRP1-int2, INJ2-int4, and the AU Tunnel. For the simultaneously-conducted

Test 4B, where uranine and the DNA nanotracer GR-1 were injected at INJ1-int2, plausi-

ble BTCs of uranine were observed at the monitoring locations of PRP1-int3, PRP2-int2,

PRP1-int2, INJ2-int4, and the AU Tunnel. However, plausible BTCs of DNA nanotracer

GR-1 were observed only at the monitoring locations of INJ2-int4 and PRP1-int2.

In all tests, the recording of the tracers was terminated before the concentrations

returned back to the pre-test values. However, in Tests 1A and 4A, the tails of the

BTCs, except at PRP1-int2, were sampled for a sufficiently long time to enable their

extrapolation, i.e., a quasi-exponential decline of the tracer curves was observed (Shook


et al., 2017). In contrast, Test 4B was terminated before the arrival of the peaks of the

BTCs and thus extrapolation was not possible. Consequently, temporal moment analyses

were not applied to the BTCs of Test 4B.

2.3.3 Attenuation and absence of DNA nanotracers

The majority of the recorded and normalized DNA nanotracer signals (Figure 2.3) are

weaker than the corresponding dye tracer signals, the exception being Test 1A where

no significant difference between the signal strengths was observed. Tracer injection

intervals in Tests 1A and 4A, which cut different S3 shear zone structures (S3.2 and

S3.1, respectively), yielded BTCs characterized by relatively early first arrival times in

comparison to Test 4B, and long tails. Moreover, the BTCs of the DNA nanotracer GR-

3, and to a slightly lesser extent sulforhodamine B for Test 4A, show particularly sharp

peaks at PRP2-int2 followed by a slowly decreasing tail.

In Test 4B, tracers were injected into the interval INJ1-int2 penetrating the S1.2 shear

zone structure and were collected at monitoring locations that penetrated S3 shear zone

structures (except PRP2-int1). It is evident in Figure 2.3 that the tracer breakthrough

times at the various monitoring locations are consistently longer and the peak concentra-

tions somewhat lower than for Tests 1A and 4A. Indeed, the DNA nanotracer GR-1 was

only observed in two out of the five monitoring locations where the BTCs for uranine were

obtained. The ultra-low detection limit of the DNA nanotracers should have allowed the

detection of GR-1 at the same locations as uranine, had any been present (Sabir et al.,

1999). It is also unlikely that the DNA nanotracer measurement was faulty, as qPCR

measurements were carried out with three replicates. Rather, our favored interpretation

is that the DNA nanotracer particles were completely excluded from the flow paths that

connect the injector (INJ1-int2) to the three monitoring locations of PRP1-int3, PRP2-

int2, and the AU Tunnel.

Theoretically, due to the size exclusion effect, pore throats must be at least 1.5 times

larger than the particle diameter to permit the particles to travel through the fracture

pore throats (Sirivithayapakorn and Keller, 2003). If a section of a fracture that hosts a

key flow path between injection and production points has minimum pore throats that

are too small to allow DNA nanotracers to pass, then the result would be an attenuation

or complete absence of the particulate tracer signal at the monitoring point. Another

factor that may contribute to the low DNA tracer recovery in Test 4B is the relatively

large vertical distance between the injection and monitoring points (Figure 2.1). This

configuration could introduce other factors that may explain the low concentrations and

even the complete absence of DNA nanotracer at the monitoring locations during Test 4B:

i) density effects of the DNA nanotracers and ii) dilution of the tracer due to longer flow


paths. James and Chrysikopoulos (2011) showed that colloid density has a significant

effect on the transport of the colloid as gravitational force induces their settling. Since

the DNA nanotracer particles have approximately twice the density of water, it is rea-

sonable to assume that their recovery is suppressed by the gravitational force when they

travel upward to the monitoring locations. The effect of the longer flow paths on DNA

nanotracer recovery is less clear as relatively good recovery was observed for relatively

long flow paths leading to the AU Tunnel outflow during Tests 1A and 4A (Table 2.2).

2.3.4 Residence time and the first arrival

Figure 2.4 shows the age distribution functions, E (t), for Tests 1A, 4A, and 4B. The

values of E (t) are comparable to the probabilities of tracers arriving with respective

residence times at a monitoring location. In other words, E (t) dt denotes the percentage

of the injected tracer arriving at the monitoring location in a time interval dt that has

a residence time of t. The magnitudes of the E (t) values of the DNA nanotracers are

lower than those of the dye tracers (Figure 2.4). This suggests that, compared to the

dye tracers, the DNA nanotracers had a lower arrival probability for the same respective

residence time. More specifically, overall, the DNA nanotracers spend less travel time in

the reservoir than the dye tracers.

The first and peak times (in minutes) of the dye tracers and the DNA nanotracers

are marked on the RTD curves in Figure 2.4. The first arrival time is here defined as 1%

of the Cpeak/C0 value (Table 2.2). These marked values show that the DNA nanotracers

consistently arrived earlier than the dye tracers. Because there is no background noise

for the uniquely-encoded DNA nanotracers, ultra-low concentrations can be detected

and distinguished as a real signal, thereby allowing high-precision detection of the first

arrival of DNA nanotracer particles at a monitoring location. In contrast, confident

determination of the first arrival is not always possible with the conventional dye tracers

due to their non-negligible background concentrations that can smear the true signal

or produce a stray signal (Bailly-Comte et al., 2018). For example, the approximate

average background concentration of uranine measured in the DUG-Lab before Test 1 was

0.139 ppb in samples taken from the AU Tunnel and 0.015 ppb from the INJ1 borehole.

Therefore, background concentrations and their fluctuations impede the determination of

the actual first arrival of the conventional dye tracers.

2.3.5 Recovery and mean residence time

The results of the moment analyses of the RTD curves in Figure 2.4 are listed in Table 2.2.

During Test 1A, the dye tracer uranine and the DNA nanotracer PT-2 had almost the

same recoveries, R, of 44% and 42%, respectively. Conversely, the recoveries for the DNA


Figure 2.4 – Residence time distribution (RTD) curves of dye tracers and DNA nano-tracers from Tests 1A, 4A, and 4B. RTD values calculated from the measured concen-trations are illustrated with lines with circle markers. RTD values of the extrapolatedexponential decay are shown with solid lines. Line colors indicate the monitoring loca-tions.


Table 2.2 – Moment analysis results from Tests 1A, 4A, and 4B.

Monitoringlocation

Distancea R t∗ VP G m2,c Cpeak/C0b

[m] [%] [min] [m3] [–] [min2] [–]

Uranine 44

Tes

t1AAU Tunnelc 20.7 44 950 0.47 0.32 2.91E+5 4.79E-4

DNA nt PT-2 42AU Tunnelc 20.7 42 748 0.35 0.32 1.96E+5 7.65E-4

Sulforhodamine B 70

Tes

t4A

INJ2-int4c 10.9 5.4 2770 0.089 0.33 3.91E+6 1.95E-4PRP1-int2d 16.6 0.20 2500 0.0029 – – –PRP1-int3c 7.1 3.1 2070 0.038 0.47 4.46E+6 4.30E-4PRP2-int2c 6.5 2.7 1380 0.023 0.42 1.85E+6 1.70E-3AU Tunnelc 28.7 59 2590 0.92 0.36 3.99E+6 2.32E-4

DNA nt GR-3 6.6INJ2-int4c 10.9 0.17 1940 0.0020 0.30 1.55E+6 1.04E-5PRP1-int2d 16.6 0.047 2460 6.7E-4 – – –PRP1-int3c 7.1 0.26 988 0.0015 0.52 2.45E+6 1.51E-4PRP2-int2c 6.5 0.37 1030 0.0023 0.38 1.56E+6 4.39E-4AU Tunnelc 28.7 5.5 1570 0.051 0.34 1.30E+6 4.50E-5

Uranine 4.9

Tes

t4B

INJ2-int4d 19.0 1.7 2020 0.022 – – 1.19E-4PRP1-int2d 15.2 0.48 2480 0.0081 – – –PRP1-int3d 16.0 0.48 2130 0.0065 – – –PRP2-int2d 15.6 0.041 2500 6.5E-4 – – –AU Tunneld 37.5 2.2 2500 0.034 – – –

DNA nt GR-1 0.0015INJ2-int4d 19.0 5.2E-4 1820 5.9E-6 – – 3.88E-8PRP1-int2d 15.2 9.6E-4 2630 1.5E-5 – – –PRP1-int3 16.0 0 – – – – –PRP2-int2 15.6 0 – – – – –AU Tunnel 37.5 0 – – – – –

Note. The parameters R, t∗, VP, G, and m2,c were calculated from exponentiallyextrapolated tracer curves, i.e., when a sufficient amount of data were available to performfitting through late-time concentrations. The extrapolation was performed according toShook and Forsmann (2005).aEuclidean from the injection interval to the monitoring location. bThe peak concentrationdivided by the tracer injection concentration. cAll parameters were estimated withexponentially extrapolated late-time concentrations. dParameters R, t∗, and VP werecalculated with the monitored concentrations only, because the tests were terminated tooearly and no extrapolation of the late-time concentrations could be implemented.


nanotracers during Tests 4A and 4B were at most 10% of the recovery of the dye tracers.

The recoveries at the AU Tunnel outflow were generally higher than at other monitoring

locations, even though the Euclidean distances between the AU Tunnel outflow point and

the injection intervals were the longest ones. In general, no clear relationship between the

Euclidean distance and the recovery could be identified (Figure A.9).

The mean residence times, t∗, of the DNA nanotracers are generally smaller than those

of the solute dye tracers (Table 2.2). However, no clear trend is observed between the R

and the t∗ values at all monitoring locations (Figure A.10). The smaller t∗ values of the

DNA nanotracers suggests that the particles travel at a greater average velocity than the

solute dye tracers. In fact, using the distance between the injection points and the AU

Tunnel outflow point, and the corresponding t∗ estimates, the mean water velocities were

0.022 m min−1 in Test 1A and 0.011 m min−1 in Test 4A for the solutes, and 0.028 m min−1

in Test 1A and 0.018 m min−1 in Test 4A for the DNA nanotracers. The greater travel

velocity, as well as the lower R values of the DNA nanotracers, agree well with other

studies, employing colloidal tracers (Vilks and Bachinski, 1996; Zheng et al., 2009; Zhang

et al., 2015; Pang et al., 2017).

2.3.6 Transport processes

The calculated second temporal moments, m2,c, listed in Table 2.2, imply that the spread-

ing of tracers is stronger in Test 4A than in Test 1A, and that the DNA nanotracers are

less dispersive than the solute dye tracers. Previous studies have reported that tracer

dispersion (indicated by m2,c) and its non-Fickian transport (indicated by advanced peak

arrival time and tailing) can be strengthened by heterogeneity of the flow field (Cirpka

and Kitanidis, 2000) and different mass transfer processes, such as slow advection (Zinn

and Harvey, 2003; Willmann et al., 2008; Fiori and Becker, 2015; Henri and Fernàndez-

Garcia, 2015; Tuykhova and Willmann, 2016), immobile zones (Haggerty and Gorelick,

1995; Zinn and Harvey, 2003; Dou et al., 2018; Phirani et al., 2018), and matrix diffu-

sion (Neretnieks et al., 1982; Maloszewski and Zuber, 1993; Hadermann and Heer, 1996;

Zhou et al., 2007; Mosthaf et al., 2018), although the latter is not relevant for the DNA

nanotracers. While these processes have likely played a role in the transport of the solute

tracers and the DNA nanotracers, to date we have neither been able to distinguish be-

tween these processes, nor obtained satisfactory fits to the available RTDs with analytic

solutions. Our attempt on numerical modeling of tracer transport with simple geometries

was also unable to capture the complexity of the heterogeneous system, i.e., the lack of

correlation between the shortest Euclidean distance and the mean residence time. Hence,

the key aspects remaining to be solved in order to model the tracer transport in the Grim-


sel DUG-Lab are transmissivity and flow velocity distributions, which are the subject of

on-going research.

The concurrent use of multiple tracers with different transport properties enables the

investigation of the properties of the medium that influence differential transport phe-

nomena (Geyer et al., 2007; Dai et al., 2012). Colloidal tracers employed in different sizes,

when combined with solute tracers, can provide additional information on the subsurface

hydrodynamics, such as tracer dispersivity and fracture aperture width (Grindrod et al.,

1996; Vilks et al., 1997; James and Chrysikopoulos, 2003; Zheng et al., 2009; Albarran

et al., 2013). As suggested by our previous studies (Mikutis et al., 2018) and other column

tests (Albarran et al., 2013), the larger the (DNA nanotracer) particle size, the lower the

recoveries and peak concentrations they exhibit. This is likely due to size exclusion of

larger particles from smaller pores, which causes their lower R and Cpeak values. Vilks

et al. (1997) noted that the lower Cpeak/C0 values of colloids, when compared to solutes,

were an indication of particle loss during their transport through pores and fractures.

We observe that the differences in the normalized peak concentrations, Cpeak/C0, of

the solute dye tracers and the DNA nanotracers are not consistent at the different moni-

toring locations (Figure 2.3 and Table 2.2). For example, during Test 4A, the Cpeak/C0 of

GR-3 observed at PRP1-int3 is almost three times lower than that at PRP2-int2, whereas

the Cpeak/C0 of sulforhodamine B observed at PRP1-int3 was almost four times lower

than that at PRP2-int2. Similarly, Cpeak/C0 of sulforhodamine B observed at INJ2-

int4 and the AU Tunnel outlet are almost identical, whereas the DNA nanotracer GR-3

Cpeak/C0 observed at the AU Tunnel is almost 4.5 times higher than that at INJ2-int4.

The observed variations in the differences of Cpeak/C0 of the DNA nanotracers and the

dye tracers at different monitoring locations may be attributed to fracture aperture vari-

ations along the tracer flow paths that linked the injection and monitoring locations.

This interpretation is also supported by the observations of Vilks and Bachinski (1996),

Zheng et al. (2009), and Albarran et al. (2013), who concluded that a decrease in the flow

velocity enhances the sedimentation of the suspended particles, inasmuch as a reduction

of fracture aperture and/or connectivity enhances particle retention and filtration. The

Stokes settling velocity of a spherical particle is defined as vset = 2r2pg(ρp − ρf )(9µf )−1,

where rp is the particle radius, g is acceleration due to gravity, ρp and ρf are particle and

fluid densities, respectively, and µf is the dynamic viscosity of the fluid. This calculation

results in vset = 8.3 × 10−7 m min−1, which is clearly below the mean water velocities

calculated in the previous section. These above-mentioned factors may explain, or at

least contribute to, the lower colloid Cpeak/C0 and recovery values observed in Test 4A,

compared to Test 1A, although the settling due to fluid velocity decrease may be minor.

Assuming a constant hydraulic head gradient, a decrease in the mean fluid flow velocity in


fractures reflects a decrease in fracture transmissivity, which could be caused by a reduc-

tion of fracture aperture and/or connectivity. It is also worth noting that, occasionally,

large cavities are present in the shear zones. One such cavity was intersected by the INJ1

borehole at a depth of 10m (Krietsch et al., 2018). Fluids traversing such cavities would

yield very low velocities and would thus be prone to losing colloids through sedimentation.

Such cavities may also have contributed to the lack of correlation between the Euclidean

distances and t∗ (Section 2.3.5).

2.3.7 Swept pore volume

The swept volumes, Vp, of the different tracers, follow the same trend as the tracer

recoveries. That is, the volume swept by the DNA nanotracer is about 10% of the volume

swept by the solute tracer in Test 4A. In contrast, during Test 1A, the volume swept by

the DNA nanotracer PT-2 is about 75% of the volume swept by uranine. However, similar

trends between swept volumes and tracer recoveries are expected, based on the definition

of Vp (Equation 2.6). In both Tests 1A and 4A, the largest swept volumes are observed

between the injection intervals and the AU Tunnel outflow point. However, when the

injection interval is changed from Borehole INJ2 to INJ1 (i.e., from Test 1 to Test 4), a

significant change in the hydraulic connection between the injection interval and the AU

Tunnel outflow point is suggested by the lower R and the smaller Vp values exhibited by

the DNA nanotracer GR-3 in Test 4 (Table 2.2). Apparently, the change of the hydraulic

connection predominantly influences the transport of the DNA nanotracer.

The unpredictable nature of the differences between the Vp estimates for solute and

DNA nanotracer diminishes the suitability of the DNA nanotracers to correctly estimate

pore volumes in the subsurface. Sirivithayapakorn and Keller (2003) observed that pref-

erential flow paths become less evident the smaller the colloids, and that colloids have

a very low probability of entering dead-end pores. The former may explain the similar

swept volumes between the tracers in Test 1A. However, as only one particle size was used,

i.e., approximately 166 nm, this observation may be attributed to larger ratios of fracture

openings to colloid diameters than in Test 4A. The latter, regarding the probability of

colloids entering dead-end pores, would result in smaller Vp estimates for colloids than

for solutes. Additionally, the ζ-potential of −14.4 mV for GR-3 may have enabled parti-

cle attachment, resulting in lower tracer recovery R, and subsequently, lower Vp. These

effects are probably a contributing factor, as our results show that the DNA nanotracer

colloids invariably yield smaller Vp. However, our results do not allow a distinction among

particle exclusion from dead-end pores, entrapment due to size exclusion or attachment

processes, and flow path channeling. In any case, the unpredictable interplay of these


mechanisms results in smaller estimates of Vp from DNA nanotracers than from solute

dye tracers.

2.3.8 Flow geometry and hydraulic connectivity

The F−Φ curves and the corresponding G values mainly refer to the spatial heterogeneity

of the volume swept by the individual tracers and can be used to confirm the occurrence

of channeling. However, the shape of the F − Φ curves and, specifically, their deviation

from the line of equality, i.e., the Gini coefficient, do not directly indicate whether one

tracer experiences more channeling than the other. In fact, preferential flow paths and

channeling of the particulate DNA nanotracers may result from size exclusion, flow field

heterogeneity, and particle density effects. To assess the relative importance of these

aforementioned mechanisms requires consideration of the mass recovery, the mean resi-

dence time, and the swept volume (Table 2.2). The calculated G values in Table 2.2 all lie

in the range of 0.30-0.52, which indicates that fluid flow is not equally distributed within

the tracer swept volumes. Although the DNA nanotracers are expected to travel through

preferential flow paths, they do not yield systematically larger G values than the solute

tracers.

Figure 2.5 shows the F−Φ diagrams of the extrapolated tracer RTD curves of Tests 1A

and 4A.It is worth emphasizing that the ability of the F−Φ curves to depict the actual flow

geometries depends on the accuracy of the extrapolation.Nevertheless, the F − Φ curves

of all tracers deviate from the line of equality, which represents a homogeneous fracture

system. For Test 1A, essentially identical F − Φ curves and G values were obtained

for the solute and the DNA nanotracer at the AU Tunnel (Table 2.2 and Figure 2.5).

Moreover, these were broadly similar to the G values and the F − Φ curves obtained at

the AU Tunnel outflow for both solute and DNA nanotracer in Test 4A. Furthermore,

the F −Φ curves obtained from the DNA nanotracers measured at INJ2-int4, PRP2-int2,

and to a lesser extent, the AU Tunnel outflow point during Test 4A, deviate less from

the equilibrium line than the corresponding curves from the dye tracers. In contrast, the

F − Φ curve from the DNA nanotracer at PRP1-int3 deviates significantly more from

the line of equality than that of sulforhodamine B. A close inspection of the F −Φ curve

from the DNA nanotracer at PRP1-int3 suggests that a relatively large portion of tracer

was transported through a small portion of swept volume, i.e., 30% of the swept volume

transmits 70% of the DNA nanotracer.

In general, the flow field geometry of a given injection-monitoring pair is defined by

the pore space/fracture geometry and the hydraulic boundary conditions. Given identical

F − Φ curves and similar tracer R, t∗, Vp, and m2,c, as in Test 1A, it is reasonable to

assume that the tracers largely shared the same flow geometry. Likewise, dissimilarities


Figure 2.5 – F − Φ curves derived from the RTDs obtained in Tests 1A and 4A.Here, only the extrapolatable RTDs are reported. Line colors indicate the monitoringlocations of RTDs.


between the F −Φ curves and substantial differences in the moment analysis parameters,

as in Test 4A, suggest that large portions of flow geometry were excluded during the

transport of the DNA nanotracers.

The excluded geometries between the DNA nanotracers and the solutes, particularly

the DNA nanotracer GR-3 and sulforhodamine B during Test 4A, were likely zones of slow

advection, including the occasionally occurring large cavities, which led to the deposition

of the DNA nanotracer GR-3 (Vilks and Bachinski, 1996; Zheng et al., 2009; Albarran

et al., 2013) and subsequently resulted in decreased values of R, Vp, and m2,c derived

from the GR-3 BTCs. As discussed earlier, the exclusion from dead-end pores, the size

exclusion effect and attachment processes further reduced the Vp and t∗ values of the

GR-3 DNA nanotracer.


2.4 Conclusions

Two field campaigns were conducted to evaluate and advance the use and interpretation

of DNA-labeled silica nanoparticle tracers in a fracture-dominated crystalline rock. The

DNA nanotracers were evaluated by analyzing the temporal moments of their residence

time distribution (RTD) curves, and comparing the results to those obtained from solute

dye tracers. The results showed that, compared to solute dye tracers, the particulate DNA

nanotracers yield lower detection limits, no susceptibility to background noise, smaller

mean residence times, lower mass recoveries, less dispersion, and smaller swept volumes.

The current study shows that the solute tracers are better suited than DNA nano-

tracers for the estimation of mass recovery, swept volume, and the Gini coefficient. This

was expected as the theory of the tools applied here was developed for conventional trac-

ers. However, the large difference in the results observed shows the potential of using

DNA nanotracers together with solute tracers to define the connected structures within

the fractured media and finally its connectivity. Further research, both experimental

and numerical studies, are still needed to better understand the differences. Also, the

shorter mean residence time, smaller dispersion, and ability to prepare infinite varieties

with unique signatures, even in different sizes, gives the particulate DNA nanotracers

an advantage for use in tracer tomography (Kong et al., 2018), or in tracing particulate

or particulate-bound contaminant transport, such as sediment or radionuclides. Impor-

tantly, the combined use of DNA nanotracers and solute tracers yields more information

than single usage of one type of tracer. In this regard, we showed that the DNA nano-

tracers can help in determining additional flow path connectivities, mainly due to their

low detection limit and smaller swept volumes when compared to solute tracers. Indeed,

we anticipate that the results of this study will facilitate the subsequent characterization

of the hydraulic stimulation at the GTS. However, an important issue to be resolved,

regarding the transport of the DNA-labeled nanoparticles, is their tendency to settle,

particularly at low fluid flow velocities, which is a consequence of their density being

approximately twice that of water.

We have identified different transport properties of the DNA nanotracers in compar-

ison to solute tracers. The transport of the particulate DNA nanotracers in fractured

rock is strongly influenced by the heterogeneity of the flow field, and yields significantly

different tracer BTCs than observed in porous media (Kong et al., 2018; Mikutis et al.,

2018). The results presented in this paper suggest that DNA nanotracers could be well

suited to the characterization of liquid-bearing fracture or karst systems, particularly in

case of characterizing particulate transport, such as in some hydrogeological, petroleum

engineering, and geothermal energy applications.

Acknowledgements

The ISC is a project of the Deep Underground Laboratory at ETH Zurich, established by

the Swiss Competence Center for Energy Research - Supply of Electricity (SCCER-SoE)

with the support of the Swiss Commission for Technology and Innovation (CTI). Funding

for the ISC project was provided by the ETH Foundation with grants from Shell and

EWZ and by the Swiss Federal Office of Energy through a P&D grant. The Grimsel

Test Site is operated by Nagra, the National Cooperative for the Disposal of Radioactive

Waste. We are indebted to Nagra for hosting the ISC experiment in their GTS facility

and to the Nagra technical staff for onsite support. The authors are grateful for the

contributions of N. Knornschild, F. Leuenberger, G. Mikutis and M. Puddu. The three

anonymous reviewers and the Associate Editor D. O’Carroll are further thanked for their

constructive reviews that helped to improve the paper. Editor J. Bahr is thanked for her

handling of the manuscript. The Werner Siemens Foundation (Werner Siemens-Stiftung)

is further thanked by M.O. Saar for its support of the Geothermal Energy and Geofluids

Group at ETH Zurich. The data used in this study is available at https://www.research-

collection.ethz.ch/handle/20.500.11850/318320.

57

3Characterization of the effects of hydraulic

stimulation with tracer-based temporal momentanalysis and tomographic inversion

Accepted to be published as:

A. Kittilä, M.R. Jalali, M. Somogyvári, K.F. Evans, M.O. Saar, and X.-Z. Kong (2020),

Characterization of the effects of hydraulic stimulation with tracer-based temporal mo-

ment analysis and tomographic inversion, Geothermics.

59

Abstract

Tracer tests were conducted as part of decameter-scale in-situ hydraulic stimulation ex-

periments at the Grimsel Test Site to investigate the hydraulic properties of a stimulated

crystalline rock volume and to study the stimulation-induced hydrodynamic changes.

Temporal moment analysis yielded an increase in tracer swept pore volume with promi-

nent flow channeling. Post-stimulation tomographic inversion of the hydraulic conductiv-

ity, K, distribution indicated an increase in the geometric mean of logK and a decrease in

the Dykstra-Parsons heterogeneity index. These results indicate that new flow path con-

nections were created by the stimulation programs, enabling the tracers to sweep larger

volumes, while accessing flow paths with larger hydraulic conductivities.

61

62 3 Characterization of the effects of hydraulic stimulation withtracer-based temporal moment analysis and tomographic inversion

3.1 Introduction

As global energy production and opinions regarding its future are shifting towards the use

of sustainable energy, significant interest has been focused on renewable energy sources

which are expected to play a larger role in the energy sector. Geothermal energy has

enormous potential (Kirli and Fahrioğlu, 2019; Lee et al., 2019) and is baseload capa-

ble (Tester et al., 2006). The two most commonly discussed forms of geothermal energy

utilization are standard hydrothermal and petrothermal, the latter being also referred

to as Enhanced or Engineered Geothermal Systems (EGS) or originally, Hot-Dry Rock

(HDR) systems. Traditional hydrothermal systems require high reservoir temperatures

of at least ∼ 170°C at relatively shallow depths of approximately 2 to 3 kilometers to

enable electricity generation at a reasonable Levelized Cost of Electricity (LCOE). Such

resources are scarce and globally unevenly distributed (Mock et al., 1997; Tester et al.,

2006; WEC, 2013). One promising but challenging idea to alleviate this scarcity issue is

to drill down to greater depths of 5 to 7 kilometers, where crystalline basement rocks are

typically found. Rock temperatures at such depths commonly reach or exceed ∼ 170°C,

even in areas with a standard geothermal gradient for the continental-crust of ∼ 30°C/km.

The drawback of this approach, however, is that the permeability of crystalline basement

rocks is usually too low to permit the flow rates needed to yield commercially viable

LCOEs. As a result, such petrothermal systems require hydraulic stimulation to enhance

the overall system permeability using techniques such as hydraulic fracturing (as originally

contemplated) or, more likely, ‘hydroshearing’, whereby shearing and attendant dilation

of critically stressed but hydraulically tight natural fractures is induced by increasing the

pore-fluid pressure.

Experience has shown that hydraulic stimulation operations can occasionally induce

earthquakes large enough to be felt by local populations (Häring et al., 2008; Edwards

et al., 2015). A promising strategy to reduce this risk, without compromising the volume

of rock that is stimulated, is to adapt the techniques developed for shale gas formations

to a geothermal context and perform a series of smaller-volume fluid injections along iso-

lated borehole intervals within the formation, so that each stimulation event only induces

micro-seismicity. In this scenario, the intervals would be close enough to promote linkage

of the individually stimulated volumes, thereby forming an extensive stimulated fracture

network. It remains to be seen whether a network can be generated that, indeed, has the

requisite properties of a large swept area and low impedance needed to support efficient

long-term advective heat extraction and economical electricity generation. Aside from the

need for fundamental research into the stimulation processes underpinning permeability

enhancement and fracture linkage, the development of stimulation technologies also re-

3.1 Introduction 63

quires improvements of the tools and techniques needed for characterizing the geologic

formation and assessing its potential for stimulation so as to create an artificial reservoir,

albeit small-scale. This contribution provides one such technique that we believe is par-

ticularly well suited for characterizing the pre- and post-stimulation hydraulic properties

of formations.

Once stimulation has resulted in a formation that may be called a reservoir, i.e., a

formation that can store and/or transmit a significant amount of fluid, its performance

as an efficient heat exchanger must be evaluated. Grant (2016) provides three physical

parameters to evaluate the performance of an EGS: reservoir impedance, recovery factor,

and tracer swept volume. In other words, the success of an EGS depends on the ability of

the circulating working fluid to access a substantial volume of the hot reservoir rock and

extract heat from it (Olasolo et al., 2016). Clearly the characterization of the amount and

significance of undesirable preferential flow paths is key in this regard. Whilst borehole

logs can identify the inlet and outlet points along the injection and production wells,

respectively, tracer tests are the standard tool for characterizing the flow paths themselves

(Chrysikopoulos, 1993; Rose et al., 2006; Vogt et al., 2012; Axelsson, 2013; Buscarlet et al.,

2015; Winick et al., 2015; Ayling et al., 2016; Shook and Suzuki, 2017).

Most previous EGS field projects that aimed at improving our understanding of the

processes relevant for permeability creation and/or enhancement sought to develop reser-

voirs at scales (i.e., well separations) of 100m or more. At the other extreme, there are

numerous laboratory studies of EGS-relevant processes at scales of a metre or less. How-

ever, there have been relatively few in-situ investigations conducted that explore aspects

of reservoir creation at intermediate scales, notable exceptions being the projects at Rose-

manowes (Phase 1) (Batchelor, 1982), Falkenberg (Rummel and Kappelmeyer, 1983), Le

Mayet (Cornet, 1987), and the Gamma project (Niitsuma, 1989). Thus, a knowledge

gap still exists at the deca- to hectometer scale (Evans, 2015). The in-situ Stimulation

and Circulation (ISC) experiments conducted at the Grimsel Test Site, which is operated

by NAGRA (Swiss National Cooperative for the Disposal of Radioactive Waste), were

intended as a contribution towards bridging this gap by performing stimulations at the

decameter scale in an underground environment.

The ISC experiment consisted of well-controlled and densely-monitored hydraulic

stimulation experiments designed to: i) study the processes relevant for permeability en-

hancement and creation, ii) investigate mitigation measures to reduce the risk of inducing

significant seismicity during hydraulic stimulation, and iii) improve our understanding of

the heat exchange efficiency and the hydraulic properties of enhanced geothermal sys-

tems (Amann et al., 2018; Gischig et al., 2019). To meet the latter objective, tracer tests

were conducted in the study volume before and after the hydraulic stimulation exper-


iments. The purpose of this paper is to report on the results of applying a temporal

moment analysis and a tomographic inversion to these tracer test data. Moment anal-

yses of tracer breakthrough curves (BTC) yield BTC-intrinsic statistical properties of

the subsurface, including the tracer swept volume, the fluid residence time distribution,

and potential short-circuit flow paths (Shook and Forsmann, 2005; Sanjuan et al., 2006;

Tester et al., 2006). Moreover, to resolve the spatial hydraulic properties of the pre- and

post-stimulation rock volume, a tomographic procedure that is similar to those in geo-

physics (Mandal et al., 2019; Nieto et al., 2019) was performed, using tracer peak travel

times (Somogyvári et al., 2016; Kong et al., 2018). Applying the tomographic proce-

dure separately to the pre- and post-stimulation tracer test data enabled us to define the

stimulation-enhanced changes in the hydraulic properties of the rock mass. Our results

help bridge the gap between laboratory- and EGS-scale experiments and, importantly,

promote the understanding and interpretation of tracer tests that may be conducted at

future full-scale EGS sites (Evans, 2015).

3.2 Site description 65

3.2 Site description

The Grimsel Test Site (GTS) is located in the Swiss Alps at 1733m a.s.l. with 400-

500m of overburden. The host rocks of the GTS are the crystalline rocks of the Central

Aar Granite (CAGr) and Grimsel Granodiorite (GrGr) (Keusen et al., 1989), and are

considered a suitable analogue for EGS reservoirs that are located in deep crystalline

basements (Amann et al., 2018; Krietsch et al., 2018; Gischig et al., 2019). The test volume

of the in-situ Stimulation and Circulation (ISC) experiment is situated at a location in the

GTS where the rock is strongly-foliated granodiorite. The experiments were conducted

between 2015 and 2017 (Amann et al., 2018).

The test volume is intersected by two types of shear zones: i) ductile structures, desig-

nated S1 (052 °/77 °), and ii) younger brittle-ductile structures, designated S3 (093 °/65 °)

(Fig. 3.1) (Keusen et al., 1989). There are four S1 and two S3 shear zone structures

within the test volume. The two S3 zones coincide with meta-basic dykes, which locally

bound a brittle fracture zone with a high fracture density of about 20m−1 (Jalali et al.,

2017; Krietsch et al., 2018) that contains cataclasites, breccias, and fault gouge (Ziegler

et al., 2013). Aside from this zone, the fracture density in the rock mass is exceptionally

low (0-3m−1) (Gischig et al., 2018). The ambient temperature at the GTS is 13 °C, and

under natural flow conditions, there is no significant water discharge into the tunnels at

the GTS (Keusen et al., 1989). Measurements indicate that the brittle faults, bounded

by the two S3 shear zones, dominate the fluid flow (Jalali et al., 2017; Krietsch et al.,

2018), producing an approximate outflow of 100mlmin−1 at the AU Tunnel (Jalali et al.,

2018b). During the tracer tests, the principal outflow point in the AU Tunnel served

as one of the monitoring locations and was named the AU Tunnel outflow point. The

transmissivity of the shear zones is approximately 10−12 to 10−6 m2 s−1, and less than

10−13 m2 s−1 in the intact rock (Keusen et al., 1989; Jalali et al., 2017).


a)

c)

b)a)

Figure 3.1 – Overview of the experiment setup at the Grimsel Test Site (GTS). a)The shear zones S1.3 and S3.2 are shown with red and green planes, respectively. Thecontours on the structures are drawn to give a sense of curvature and are not relatedto the tomographic inversion grid. A second S3 structure (S3.1) lies close to S3.2 butis not shown in the figure to avoid complication. However, the intersections of the S3shear zones with the AU Tunnel are indicated as green disks in the AU Tunnel (Krietschet al., 2018). The monitoring intervals (red cylinders) in the INJ and PRP boreholes(blue and green lines, respectively) and the AU Tunnel outflow point (red circle) are alsomarked. The dashed blue lines define the boundaries of the 5m thick section throughthe 3D tomographic inversion grid. b) Top-view of (a), showing the positions of theinlet and outlet points with respect to the tomographic section, which approximatesthe S3 structures. Note that the grid on the structures does not represent the grid usedfor the tomographic inversions. c) The red boxes in the core images (Krietsch et al.,2018) mark the 0.52m long intervals (except Test 7, which had a length of 1.0m) inthe INJ1 and INJ2 boreholes, which were isolated by hydraulic packer systems. Theseintervals are indicated by the bold font in (a).

3.3 Methods 67

3.3 Methods

3.3.1 Experimental design

A total of nine solute tracer tests were conducted during the ISC experiments. Tests 1 to 8

were pre- and post-stimulation reservoir characterization tests, and Tests 8 and 9 were

part of the circulation experiment (Table 3.1). In this paper, we report on the results

from the eight pre- and post-stimulation tracer tests in which solute dye tracers were

used (Column 7 in Table 3.1). The solute dye tracers were accompanied by colloidal DNA

nanotracers in Tests 1 and 4 (Kittilä et al., 2019), and by salt injections in Tests 2, 5, 6, 8,

and 9 (Doetsch et al., 2018a).

Tracer sampling points within the test volume were located along the three PRP

boreholes and whichever INJ borehole was being used for production during the test in

question. All boreholes were drilled from the AU Cavern (Fig. 3.1). The PRP boreholes

were drilled between Tests 1 and 2, and were hard-completed with grout, leaving open

intervals of lengths 2.00 to 6.11m. The INJ boreholes were left entirely open so that

hydraulic packers could be used to isolate 0.52 to 1.00m zones containing fractures that

were of interest for tracer injection or monitoring (Table 3.2). The grouted intervals in

the PRP boreholes and the packer-isolated intervals in the INJ boreholes were each ac-

cessed by two separate 4mm ID lines: one for fluid pressure monitoring and one for fluid

injection or recovery. All pressure-monitoring lines were fully saturated before connecting

to the pressure sensors, located in the AU gallery (Fig. 3.1 and Table 3.1). Tap water or

tracer for the test was injected into the flow line leading to the injection interval. The

flow lines leading from the production intervals were left open in the AU gallery for the

collection of fluid. Since the pressure at the outlet is atmospheric, we do not report the

interval fluid pressures at the monitoring locations in Table 3.1. Note that the AU Tunnel

outflow point at the tunnel wall is in direct contact with the atmosphere. Table 3.1 lists

the injection and monitoring information for the tracer tests. All tests were conducted un-

der steady-state flow conditions. Tracers were added as short-pulse inputs to the interval

injection streams using 1 liter syringes. The injection concentrations of the tracers were

10 ppm in all tests except Tests 3, 8 and 9, where concentrations of 20 ppm were used.

The outflows (Table 3.2) at the monitoring locations were passed through fluorometers

(GGUN-FL30) to obtain continuous profiles of solute dye tracer concentrations. Effluent

samples were also collected from the produced outflow for laboratory tracer concentration

analyses using a luminescence spectrometer (Perkin Elmer, LS 50 B). These laboratory

measurements were then used for calibrations to convert the continuous fluorometer sig-

nals in millivolts (mV) to concentrations in parts per billion (ppb).


Table

3.1–Sum

mary

ofthesolute

dyetracer

testsconducted

atthe

Grim

selTestSite

(GTS)

duringthe

ISCexperim

ents.

Test

StartEnd

InjectionQ

binj

Pcin

jTracer

Monitoring

datedate

interval[L

min

−1]

[kPa]

Type

dMass[m

g]locations

e

Test

1A03.05.2016

04.05.2016IN

J2-i41.12

107Urf

10.0

AU

(INJ1-i4,IN

J1-i3)Test

1B03.05.2016

04.05.2016IN

J2-i30.095

773SB

f1.2

(AU,IN

J1-i4,INJ1-i3)

Test

2a

26.01.201730.01.2017

INJ2-i4

1.02535

Ur

9.5

PRP1-i3,

PRP2-i2,

AU

(INJ1-i4,P

RP1-i2)

Hyd

raulic

shearin

gstim

ulation

,06.-15.02.2017

Test

3A20.04.2017

21.04.2017IN

J2-i40.360

658Tin

19.2

PRP1-i3

(AU,

INJ1-i4,

PRP2-i2,P

RP2-i1)

Test

3B20.04.2017

21.04.2017IN

J2-i20.340

495Eo

19.2

PRP2-i1

(PRP2-i2,

PRP1-i3,

AU,IN

J1-i4)Test

4A24.04.2017

26.04.2017IN

J1-i40.599

644SB

f9.5

PRP1-i3,

PRP2-i2,

AU,

INJ2-i4

(PRP1-i2)

Test

4B24.04.2017

26.04.2017IN

J1-i20.634

635Urf

9.5

(PRP1-i3,

PRP1-i2,

PRP2-

i2,AU,IN

J2-i4)Test

5a

28.04.201730.04.2017

INJ2-i2

0.363663

Eo

9.0

PRP2-i1

(PRP1-i1)

Test

6Aa

03.05.201704.05.2017

INJ1-i4

0.491677

SB9.5

PRP1-i3

(PRP1-i1)

Test

6B03.05.2017

04.05.2017IN

J1-i20.609

666Ur

9.3

PRP1-i3

(PRP1-i1)

Hyd

raulic

fracturin

gstim

ulation

,15.-18.05.2017

Test

716.10.2017

18.10.2017IN

J2-i20.680

340SB

9.4

PRP2-i1

Test

8a

22.11.201729.11.2017

INJ2-i4

2.10374

Eo

19.5

AU,

PRP1-i3,

PRP2-i2

(PRP1-i2)

Circu

lationphase

with

hot

water

injection

,29.11.2017-10.01.2018

g

Test

9a

13.12.201720.12.2017

INJ2-i4

1.79450

Eo

19.0

AU,

PRP1-i3,

PRP2-i2

(PRP1-i2)

aSaltw

aterinjection

inconjunction

with

GP

R(L

eresche,2018).bH

armonic

mean.

cAbsolute

pressureat

thew

ellhead.A

tmospheric

pressurew

as82.5

kPa

onaverage

duringthe

tracertests,except

duringTest

1,when

itw

asnot

measured

andduring

Test

4,when

itw

asabout

30kP

a.dU

r=U

ranine,Eo=

Eosine,T

in=T

inopalCB

S-X,SB

=Sulforhodam

ineB

.eL

ocationsin

bracketsindicate

thatincom

pleteor

notracer

BT

Cs

were

observed.fD

NA

nanotracersw

erealso

injected,seeK

ittiläet

al.(2019).gD

oetschet

al.(2018a).

3.3 Methods 69

The following presents a short chronological description of the tracer tests and high-

lights their relationships to other activities, such as hydraulic stimulations (Table 3.1).

Tests 1 and 2 were conducted before the hydraulic shearing stimulation experiments to

aid in formation characterization. Note that during Test 1, DNA nanotracers (Kittilä

et al., 2019) were simultaneously injected with the solute dye tracers and during Test 2,

the injection of the dye tracer was immediately followed by salt water injection to enable

cross-hole ground penetrating radar (GPR) imaging at the site (Giertzuch et al., 2018).

After Test 2, hydraulic shearing stimulation experiments were performed (Doetsch et al.,

2018a). This was followed by a thermal tracer test (Doetsch et al., 2018a) that was con-

ducted shortly before Test 3. During Test 4, the injection of the dye tracer was again

accompanied by an injection of the DNA nanotracers. During Tests 5 and 6, GPR surveys

were performed using salt water injection in conjunction with the dye tracer tests. Af-

ter Test 6, hydraulic fracturing stimulation experiments were conducted (Doetsch et al.,

2018a) followed by dye tracer Tests 7 and 8. Hot water was then injected through the

system and Test 9 conducted during that injection (Table 3.1). Of the injection and

monitoring locations used in the tracer tests reported here, intervals INJ1-i4, INJ1-i2 and

both INJ2-i2 intervals (Table 3.2) were subjected to stimulation injections in either the

hydraulic shearing (HS) or hydraulic fracturing (HF) experiments HS4, HS2, HF6, and

HS1 (Table 3.3) (Doetsch et al., 2018a).


Tracer breakthrough curves (BTCs) can be viewed as the probability distribution func-

tions (PDFs) of the injected tracers recorded at the monitoring locations. Thus, statistical

properties of the BTCs can be calculated to quantify aspects of tracer transport prop-

erties in the formation. This is often referred to as moment analysis. To evaluate the

hydraulic and mass transport properties at the GTS before and after hydraulic stimu-

lation, the calibrated tracer BTCs were first normalized to derive the age distribution

functions, E (t) (Shook and Forsmann, 2005; Shook and Suzuki, 2017),

E (t) =c (t) ρqout

Minj, (3.1)

where c (t) is the tracer concentration at time t, ρ is the density of the effluent fluid sample,

qout is the volumetric outflow rate, and Minj is the mass of the injected tracer. Plotting

E (t) over the recorded time yields the so-called residence time distribution (RTD) curve.

The tracer concentration, specified in parts per billion (ppb), must be given as kg/109kg

in Eq. (3.1). Attributes of the tracer response, such as signal strength, mean and variance,

can then be determined from E (t) by calculating the temporal moments (Leube et al.,


Table

3.2–Borehole

intervalinformation.

IntervalStart

aEnd

aLength

Trans.

FractureOutflow

d

[m]

[m]

[m]

[m2s −

1]depth

b[m

]ID

b,c

[Lmin

−1]

INJ1-i4

27.6728.19

0.523.7E

-07g

27.78S3

0.061(T

1),0.075(T

3),0.001(T

8)IN

J1-i329.09

45.0015.91

N/A

severalS1,S3

0.037(T

1)IN

J1-i238.51

39.030.52

2.0E-07

h38.69

S10.084

(T2)

INJ2-i4

22.8923.41

f0.52

4.0E-06

g23.14

S30.093

(T4),0.059

(T6)

INJ2-i3

24.3124.83

0.521.8E

-08g

24.96S3

–IN

J2-i2e

38.4039.40

1.003.8E

-07h

38.68S1

–IN

J2-i239.73

40.250.52

1.5E-07

h40.09

S1–

PRP1-i3

23.2025.20

2.009.0E

-07i

severalS3

0.056(T

2),0.009

(T3),

0.048(T

4),0.049

(T6),0.090

(T8),0.051

(T9)

PRP1-i2

28.9032.00

3.104.0E

-07i

severalS3

0.001(T

2),0.043

(T4),

0.037(T

6),0.122

(T8),0.124

(T9)

PRP1-i1

41.8047.91

6.11N/A

severalS1

0.020(T

5),0.030(T

6)PRP2-i2

21.4027.00

5.603.0E

-07i

severalS3

0.013(T

2),0.06

(T3),

0.022(T

4),0.018

(T6),0.069

(T8),0.043

(T9)

PRP2-i1

40.0044.98

4.98N/A

severalS1,S3

0.505(T

3),0.083

(T4),

0.429(T

5),0.369

(T6),0.622

(T7)

AU

Tunnel

––

–N/A

–S3

1.0(T

1),0.923

(T2),

0.367(T

3),0.980

(T4),1.0

(T8)

j,0.85(T

9)j

aFromthe

topof

theborehole.

bFromKrietsch

etal.(2018).

cThe

shearstructure

thefracture

belongsto.

dOutflow

ofthe

producingmonitoring

locations.eInform

ationof

Test

7.f23.89

duringTests

8and

9.gB

eforehydraulic

shearingstim

ulation,constanthead

injectiontest

(Jalalietal.,2018b).

hFrominjectivity

afterhydraulic

stimulation

(Doetsch

etal.,2018a).

iBefore

hydraulicshearing

stimulation,pulse

tests(B

rixeletal.,2020),post-stim

ulationvalues

notavailable.

jNodata

available,outflowestim

ated.

3.3 Methods 71

Table 3.3 – Hydraulic shearing (HS) and hydraulic fracturing (HF) borehole intervalinformation (Doetsch et al., 2018a), relevant for the tracer tests presented here.

Interval Borehole Start End Length[m] [m] [m]

HS1 INJ2 39.75 40.75 1HS2 INJ1 38.00 40.00 2HS4 INJ1 27.20 28.20 1HF6 INJ2 38.40 39.40 1HF8 INJ2 15.20 16.20 1

2012),

m∗n =

∞∫0

tnE (x, t) dt , (3.2)

where m∗n yields the n-th temporal moment at location x.

The zeroth temporal moment, m∗0, yields the strength of the response, which in the

present context is the fraction of injected tracer that is recovered, R. The values of m∗0range from zero, for no recovery, to unity, for full recovery. The tracer mean residence

time, t∗, can be determined by normalizing the first temporal moment, m∗1, by the zeroth

temporal moment, m∗0,

t∗ =m∗1m∗0

. (3.3)

The dispersion of the injected tracer can be evaluated by calculating the variance of the

tracer response, i.e., the normalized and centralized second temporal moment, m2,c, the

value of which is elevated by heterogeneity (Cirpka and Kitanidis, 2000),

m2,c =m∗2m∗0−(m∗1m∗0

)2

. (3.4)

Following the approach of Shook and Suzuki (2017), the total pore volume swept by

the tracer, Vp, can be calculated from


where qinj is the volumetric injection rate. Moreover, the E (t) of the tracer BTCs can also

be used to infer aspects of the fluid flow geometry, such as the degree of flow channeling.

The storage capacity, Φ, which represents the time-weighted reservoir volume seen by

the tracer, and the flow capacity, F , which describes the fraction of the tracer recovered

at the monitoring location through that volume, can both be calculated as functions of


time, t (Shook, 2003; Shook and Forsmann, 2005),

Φ (t) =

t∫0

E (τ) τdτ

∞∫0

E (t) tdt

, (3.6)

and

F (t) =

t∫0

E (τ) dτ

∞∫0

E (t) dt

. (3.7)

The heterogeneity, or channeling, of the flow paths can then be quantified by the Gini

coefficient, G, which yields 0 for a homogeneous and 1 for a heterogeneous case (Shook

and Forsmann, 2005),

G = 2

1∫0

F dΦ− 1

2

. (3.8)

3.3.3 Tomographic inversion

We evaluated the hydraulic conductivity, K, profiles of the study site before and after

the hydraulic shearing stimulation experiment by employing the travel-time-based tomo-

graphic inversion approach of Somogyvári et al. (2016) and Kong et al. (2018), which is

valid when solute transport is advection-dominated. Thus, the apparent Peclet number

(Cirpka and Kitanidis, 2000),

Pea =2 (m∗1)

2

m2,c (m∗0)2 , (3.9)

was calculated, to assess whether tracer transport was dominated by advection (i.e.,

Pea �1) or dispersion (i.e., Pea �1). Note that the moments in Eq. (3.9) are defined

using the age distribution, E(t), curves, not the concentration, c(t), curves, as in Cirpka

and Kitanidis (2000). The calculated Pea values range between 0.49 and 6.20, out of

which only two Pea values are below one, resulting in an average Pea of 2.38 for all tracer

breakthrough curves reported in this study. Although the calculated Pea values suggest

a somewhat dispersive character of the tracer breakthrough curves, we believe it is still

meaningful to assume that the transport of tracers is advection-dominated and that, as

a result, the tracer travel times, tT (i.e., the peak arrival times), can be used to derive

hydraulic conductivity profiles for the test volume at the GTS.

When the transport equation is transformed into an eikonal equation, reported by

Vasco and Datta-Gupta (1999), the tracer travel times can be related to the inverse of

3.3 Methods 73

the mean velocity, u, through a line integral along a tracer trajectory, s,

tT (xr) =

∫ xr

xs

ds

u, (3.10)

where xs and xr are the injection (source) and monitoring (receiver) locations, respec-

tively. This line integral formulation shows that tracer travel times only depend on the

mean tracer velocity distribution along the transport trajectory. Therefore, if enough

tracer travel times are available, the mean tracer velocity distribution can be resolved.

A relation between mean tracer velocity, u, and hydraulic conductivity, K, is given by

Darcy’s law:

u =q

φ=K∇hφ

, (3.11)

where q is the Darcy velocity or specific discharge, φ is porosity, and ∇h is the hydraulic

head gradient.

As the values of K typically span orders of magnitude, and the variations of porosity

are generally small, porosity in Eq. 3.11 can be approximated by a constant value (Somo-

gyvári et al., 2016). Similarly, variations in hydraulic heads are much smaller than those

of K. Note that φ and ∇h in Eq. 3.11 are only scaling factors and, as such, do not affect

the spatial distribution of the calculated K values. In this study, we assume that the

tomographic inversion yields an equivalent hydraulic conductivity field of the fractured

rock mass. For this equivalency calculation, we assume that the porosity is adequately

estimated by a value of φ=0.25 (Bossart and Mazurek, 1991; Marschall and Lunati, 2006).

Also, for simplicity, a mean hydraulic head gradient of ∇h=5m m−1 (Supplementary Ta-

ble A.4) is used to determine K. In summary, applying Eq. 3.10 to the recorded tracer

travel times, each representing one transport trajectory, results in an inverse problem that

can be used to infer the distribution of K values between source and receiver coordinates.

For the pre-stimulation inversion we had fewer source and receiver locations (2 and 4,

respectively) than for the post-stimulation inversion (4 and 6 points, respectively) (Supple-

mentary Table A.4). The additional source and receiver locations in the post-stimulation

inversion were the lower intervals in the INJ and PRP boreholes. These are associated

with the S1 shear zone (Table 3.2) and did not yield sufficient flow rates before the stim-

ulation to be included in the tracer tests. To improve the quality of the pre-stimulation

inversion, virtual travel times were introduced into the inversion, following the recom-

mendations of Somogyvári and Bayer (2017). These virtual travel times account for the

information from the source-receiver pairs where no tracer breakthrough was observed,

thereby improving the identification of low-K regions at the expense of inferring their K

values with higher uncertainties.


The tomographic inversions were performed with the SIRT (simultaneous iterative

reconstruction technique) algorithm, implemented in the GeoTOM3D software (Jack-

son and Tweeton, 1996). The 3D rectilinear inversion grid was oriented such that the

x-, y-, and z-directions corresponded to east, north and vertical, respectively. With

this orientation, the principal structures of interest lay approximately in the x-z plane.

Source-receiver locations and tracer trajectories were defined in three dimensions (3D)

(Somogyvári et al., 2016). Due to the low spatial resolution of the source-receiver loca-

tions (Fig. 3.1), a coarse initial grid with a size of 5m in all directions was used. This

enabled us to perform a low-resolution tomographic inversion to estimate the mean tracer

velocity distribution between the source and receiver locations. The model discretization

was refined using the staggered grid method (Brauchler et al., 2013; Jiménez et al., 2013;

Somogyvári et al., 2016) with the x- and z-directions each staggered three times to reach

a final grid size of 1.25m×1.25m. The grids were initialized with a homogeneous fluid

flow velocity distribution. The SIRT algorithm then computed the tracer transport tra-

jectories iteratively over a discretized grid of mean velocities until the reconstructed and

observed tracer travel times matched closely. The inversions in this study were carried

out with 20 curved ray iterations, where tracer velocities between 0 and 200m day−1 were

permitted. The 3D mean tracer velocity distribution obtained from the iterations was

then converted to the hydraulic conductivity field, using Eq. 3.11, yielding K=uφ/∇h(Jiménez et al., 2013; Somogyvári et al., 2016; Kong et al., 2018). The results are pre-

sented as a 5 m wide section through the 3D tomograms within the x-z plane, whose

location approximately includes the planes of the S3.1 and S3.2 shear zone structures.

The 5 m width of the section is indicated by the blue dashed lines in Figs. 3.1a and 3.1b,

and corresponds to one grid element in the y-direction. The K distributions on each of

the two surfaces bounding the 5 m wide section, were found to be essentially identical and

so the hydraulic conductivity distribution is taken as uniform throughout the section.

We later quantify the heterogeneity of the resulting hydraulic conductivity fields with

the Dykstra-Parsons coefficient (Sahimi, 2011; Kong and Saar, 2013), VDP , which can be

estimated by:

VDP = 1− e−σK , (3.12)

where σK is the standard deviation of logK. The range of this coefficient is 0<VDP<1,

where 0 corresponds to an ideal, homogeneous reservoir and 1 is a perfectly heterogeneous

reservoir (Tiab and Donaldson, 2015).

3.4 Results and discussion 75

3.4 Results and discussion

In this section, we first compare the results of tracer injection into INJ2-i4 before and

after the hydraulic stimulation experiments (Tests 1, 2, 3A and 8). INJ2-i4 was the only

injection location used during the pre-stimulation tests. We then examine the results

of the remaining tracer tests (Tests 3B to 7) to characterize the hydraulic properties of

the post-stimulation rock mass. Finally, we present the results of a tomographic tracer

test inversion to delineate the changes to the hydraulic conductivity field due to the

hydroshearing stimulation experiments. The peak arrival times from Tests 1 to 6 were

used in the tomographic inversion.

It is important to note that Tests 1 to 6 had to be terminated before all of the observed

tracer signals had returned to pre-test values. Consequently, these tracer concentration

profiles were extrapolated using exponential decline curves (Shook and Forsmann, 2005;

Kittilä et al., 2019) prior to performing the moment analysis. This extrapolation ensures

consistency in deriving the time-dependent temporal moments from the residence time

distribution (RTD) curves. Additionally, the time series of tracer concentrations were

corrected to account for the time it took the tracer to descend the injection well and to

ascend the production well, so that only the time spent in the fractured rock volume is

considered in the moment analysis.

3.4.1 Comparison of pre- and post-stimulation results

The modification of the test rock volume due to the stimulations can be assessed by com-

paring the results of the analysis of the RTD curves from pre-stimulation Tests 1 and 2

with those from post hydroshearing stimulation Test 3A and the later post-hydrofracturing

Test 8, all of which featured tracers injected into interval INJ2-i4. The RTD curves for

these tests are shown in Fig. 3.2. The curves are color-coded to denote the monitoring

locations. The peak concentrations of the RTD curves for locations PRP1-i3, PRP2-i2

and the AU Tunnel outflow are marked by open squares for the pre-stimulation tests and

closed squares for Test 8, which followed the hydrofracturing stimulations. Evidently, the

cumulative effect of both stimulation programs led to earlier arrival times of the peaks, the

advancements at PRP1-i3, PRP2-i2, and the AU Tunnel being 56, 354 and 405 minutes,

respectively.

The changes to the flow paths in the rock mass underpinning the early tracer arrival

times in Test 8 appear to have occurred during the hydrofracturing program. The RTD

curves from Test 3A, which was conducted one month before the hydrofracturing stimu-

lations and two months after the hydroshearing stimulations, show much later first and

peak arrival times than seen during the pre-stimulation tests (Fig. 3.2). The reason for


this increase in residence times is unclear, but is likely related to changes of the fracture

network induced by a thermal tracer test that was conducted on the same interval immedi-

ately before Test 3 (Brixel et al., 2019). In that test, hot water at 40 °C (∆T ≈ 27 °C) was

injected into both INJ2-i4 and INJ2-i2 for almost 5 days and then cold water was injected

for a further 7 days (Doetsch et al., 2018a). During the 5 days of hot water injection, the

injection rate gradually declined to slightly less than half the initial injection rate, despite

a modest increase in fluid pressure (Brixel et al., 2019). Injection of hot water promotes

the closure of fractures through thermal expansion of the surrounding rock volume. It is

also possible that the long tracer residence times reflect changes to the system resulting

from the hydroshearing stimulation program. However, this is considered less likely, as

the only interval in INJ2 that served as a tracer injection location during that program

was at a depth of 39.75-40.75 m (Interval HS1, Table 3.3), some 17 m below INJ2-i4.

In contrast, during the hydrofracturing program, Interval HF8 (15.2-16.2 m) was located

only 7 m above INJ2-i4 and Interval HF6 (38.4-39.4 m) was located 16 m below INJ2-i4.

This relative proximity to INJ2-i4 might explain the large differences in the RTD curves

obtained in Tests 3A and 8.

Tests 1 and 2 were both conducted before any stimulations were performed and hence

it is of interest to compare their results. The RTD curves at the AU Tunnel outflow

point in Fig. 3.2 are similar for Tests 1 and 2, although the moment analysis results

in Fig. 3.3 show that Test 2 exhibits a slightly larger mean residence time, t∗, tracer

recovery, R, swept volume, Vp, and Gini coefficient, G, compared to Test 1. However, the

second moment, m2,c, of Test 2 is an order of magnitude larger than that inferred from

Test 1. The principal difference in the conditions between these two tests is that the PRP

boreholes were drilled after Test 1 had been conducted and were producing fluid during

Test 2. We thus infer that such a change in the hydraulic boundary conditions between

Tests 1 and 2 can lead to more dispersed fluid flow.

Comparing the pre-stimulation RTDs at the AU Tunnel monitoring point with those

following the completion of the stimulations (i.e., Test 8), the first and peak arrivals

occurred earlier following the stimulations (Fig. 3.2), and the mean residence time in-

creased by almost a factor of two, while tracer recovery increased slightly (Fig. 3.3).

Furthermore, the tracer swept a significantly larger volume after the stimulation program

was completed. We thus infer that the hydraulic stimulations enhanced the flow path

connectivity, yielding earlier tracer arrivals at the AU Tunnel, but also enhanced the

connectivity of flow paths characterized by longer residence times. Such flow paths may

be located either far beyond the main flow paths or merely between the main fractures

(Robinson and Tester, 1984).


10 100 1000

Time [min]

10-7

10-6

10-5

10-4

10-3

E [

1/m

in]

Test 1, AU TunnelTest 2, PRP1-i3Test 2, PRP2-i2Test 2, AU TunnelTest 3A, PRP1-i3Test 3A, INJ1-i4Test 3A, AU TunnelTest 8, AU TunnelTest 8, PRP1-i3Test 8, PRP2-i2

Test 3A

Figure 3.2 – Pre- and post-stimulation residence time distribution (RTD) curves of thetracers injected into the INJ2-i4 interval (dotted and solid lines – before stimulation,dashed lines – after stimulation). E (t) is the age distribution function. The shifts of thepeaks, a result of the hydraulic stimulation experiments, are shown with square symbols(open – before stimulation, filled – after stimulation): at the AU Tunnel, the shift wasfrom 638 to 233min, at PRP1-i3 from 87 to 31min and at PRP2-i2 from 477 to123min. The RTDs from Test 3A all fall in the lower-right corner of the graph and aremost likely affected by the thermal tracer test, conducted at the test site immediatelybefore Test 3, as discussed in the main text.

At PRP1-i3, multiple peaks are seen in both the pre-stimulation RTD curve and

that of Test 8, which followed the hydrofracturing stimulations (Fig. 3.2). The first two

peaks seen in the pre-stimulation curve appear to be partially merged and are assumed

to correlate with the two prominent peaks in the post-hydrofracturing curve. As noted

earlier, the main post-hydrofracturing stimulation peak arrives 56 min earlier than the

main pre-stimulation peak, even though the first arrival times were similar. The ratio

between the larger of the two peaks and the injected concentrations, Cpeak/Cinj , decreased

significantly after the hydrofracturing stimulations, so that tracer recovery was lower as

well, whereas the mean residence time, swept volume, Gini coefficient, and the second

moment increased (Fig. 3.3).

At PRP2-i2, the RTDs indicate that the tracer arrived significantly earlier after the hy-

drofracturing stimulations. The form of the RTD obtained after the stimulation programs

is similar to the corresponding RTD for PRP1-i3. Specifically, both post-stimulation

curves display multiple, relatively sharply-defined peaks that are markedly similar in


Coord

inate

INJ1-i4

INJ1-i2

PR

P1-i2

PR

P1-i3

PR

P2-i1

PR

P2-i2

AU

Tunnel

INJ2-i4

INJ2-i2

Pre-s

timu

latio

n

Te

st 1

AT

est 2

Te

st 2

Te

st 2

0

10

20

30

40

50

60

65

0

10

20

30

40

10

50

100

150

200

250

300

330

10

500

1000

1500

2000

2500

2770

0 2 4 6 8

10

1.4

3 5

10

15

17

×10

-4

0.0

01

0.0

1

0.0

2

0.0

4

0.0

6

0.0

8

0.1

0.2

0.3

0.4

0.5

0.6

0.0

7 1 3 5 7×

10

6

INJ1-i4

INJ1-i2

PR

P1-i2

PR

P1-i3

PR

P2-i1

PR

P2-i2

AU

Tunnel

INJ2-i4

INJ2-i2

INJ1-i4

INJ1-i2

PR

P1-i2

PR

P1-i3

PR

P2-i1

PR

P2-i2

AU

Tunnel

INJ2-i4

INJ2-i2

Po

st-s

timu

latio

n

Te

st 3

AT

est 8

Te

st 8

Te

st 8

0

10

20

30

40

50

60

65

0

10

20

30

40

10

50

100

150

200

250

300

330

10

500

1000

1500

2000

2500

2770

0 2 4 6 8

10

1.4

3 5

10

15

17

×10

-4

0.0

01

0.0

1

0.0

2

0.0

4

0.0

6

0.0

8

0.1

0.2

0.3

0.4

0.5

0.6

0.0

7 1 3 5 7×

10

6

INJ1-i4

INJ1-i2

PR

P1-i2

PR

P1-i3

PR

P2-i1

PR

P2-i2

AU

Tunnel

INJ2-i4

INJ2-i2

574

45.2

7E

-03

0.7

3

0.4

7

65

3.1

0.6

4

∆h

[m]

Monito

ring

locatio

nD

ista

nce

[m]

Firs

t arriv

al

[min

]M

ean re

sid

ence tim

et* [m

in]

Recovery

R [%

]C

peak /C

inj

[-]

Inje

ctio

nlo

catio

nG

ini c

oeffic

ient

G [-]

2nd m

om

ent

m2,c [m

in2]

Sw

ept v

olu

me

Vp [m

3]

Figure3.3

–Parallelcoordinates

plotof

therecorded

tracerbreakthrough

curvespresented

inFig.

3.2and

theirtem

poralmom

entsfrom

boththe

pre-and

post-stimulation

tracertests.

Note

thatsom

eof

thecurves

exceedthe

plottedcoordinate

values,such

asthe

Rand

Vpvalues

fromthe

AU

Tunnel.These

valuesare

indicatedwith

arrowsnext

tothe

correspondingcoordinate

axes.The

hydraulichead

difference,∆h,

was

calculatedby

subtractingthe

atmospheric

pressurefrom

theabsolute

fluidinjection

pressureat

thewellhead

(Table3.1)

andconverting

thepressure

valueto

equivalenthydraulic

head,usingawater

densityof

1000kg/m

3for

simplicity.

Ingeneral,the

elevationsof

theinjection

andproduction

pointswere

essentiallythe

same,

theonly

exceptionbeing

theAU

Tunnelmonitoring

point,which

was

approximately

0.5m

lower

(Supplementary

TableA.4).

Acorrection

was

made

forthis

exceptionby

adding0.5

mto

thehydraulic

headdifference.

The

datain

thisfigure

canalso

befound

inthe

Supplementary

TableA.5.


terms of relative timing and amplitude. This similarity suggests that the flow paths from

the injection location, INJ2-i4, to PRP2-i2 and PRP1-i3 become shared to a higher degree

following the hydrofracturing stimulation, and that a few preferential flow paths receive

the majority of the injected fluid (Moreno and Tsang, 1991). Among the three monitoring

locations discussed here, PRP2-i2 is the only location where the mean residence time was

smaller after the hyfrofracture stimulation. The tracer recovery increased only slightly

from pre-stimulation levels, despite the significantly earlier tracer arrival time. Also, the

swept volume after the stimulation programs was higher than before, whereas the Gini

coefficient remained unchanged, suggesting that the increased volume most likely stems

from newly created flow pathways, where the tracer is distributed over the flow path

network in a similar manner as before stimulation.

The RTD obtained at the AU Tunnel outflow point after the hydroshearing but be-

fore the hydrofracturing stimulations (i.e., Test 3A) shows that the tracer arrived about

3.5 times later than before the hydroshearing stimulation, and that the tracer concen-

tration increased more slowly towards the peak. Taken at face value, these observations

suggest lower average seepage velocities, u. Similarly, the tracer first and peak arrivals

at PRP1-i3 during Test 3A were delayed from the pre-stimulation results by one order of

magnitude, while recovery and swept volume remained low (Fig. 3.3). As noted earlier,

this may be related to the changes of the fracture network induced during the thermal

tracer test that was conducted at the same interval immediately before Test 3A. These

fracture network changes likely resulted in a redistribution of the flow field in the system

(Robinson and Tester, 1984; Ghergut et al., 2016) so that no tracer signal was observed in

PRP2-i2 during Test 3A, but INJ1-i4 yielded a relatively strong signal before the tracer

test was terminated. It is also worth noting that the injection flow rate during Test 3A

was 0.360L min−1 (Table 3.1), which is less than half of the rate prior to stimulation.

This would also contribute to the late tracer arrivals and lower average seepage velocities

than during the other tests that injected tracers into INJ2-i4.

The differences between pre- and post-stimulation results from the tracer injection into

INJ2-i4 are summarized in Table 3.4. Specifically, the table lists the percentage change

for recovery, R, mean residence time, t∗, swept volume, Vp, Gini coefficient, G, and second

moment, m2,c, from before (Test 2) to after the stimulation programs had been completed

(Test 8). The swept volume increased considerably at all three monitoring locations,

namely the AU Tunnel outflow point, PRP2-i2 and PRP1-i3, suggesting that new and/or

additional flow paths were accessed by the tracers as a consequence of the stimulations.

Were this to be realized in EGS stimulations, then it would imply potentially larger heat

exchange efficiencies due to the increase in the accessible rock volume. However, there

is no corresponding general trend of increase or decrease in the values of R, t∗, G and


Table 3.4 – Summary of the pre- and post-stimulation moment analysis results, whereR is tracer recovery, t∗ is mean residence time, Vp is tracer swept volume, G is Ginicoefficient and m2,c is second temporal moment.

Monitoringlocation

Change in %R t∗ Vp G m2,c

AU Tunnel 14 78 316 21 -28PRP2-i2 69 -45 91 -2 -86PRP1-i3 -55 55 43 9 163

m2,c. In general, an increase in R and t∗ is desirable for an EGS, because they imply less

loss (i.e., higher R) and overall longer travel times (i.e., larger t∗) of the fluid circulating

through the reservoir. The latter in particular postpones the breakthrough of the cold

front (Horne, 1985; Axelsson, 2013). A decrease in m2,c, such as that observed at the

AU Tunnel and at PRP2-i2, indicates that the range of travel times decreased, whereas the

Gini coefficients either remained almost the same or increased. The post-hydrofracturing

G values at the AU Tunnel outflow point, PRP2-i2 and PRP1-i3 are 0.50, 0.54 and 0.64,

respectively (Supplementary Table A.5), and are larger than the values reported from tests

at several other locations: for example, at the Ngatamariki (0.24-0.35) (Buscarlet et al.,

2015) or Habanero (0.346-0.437) (Ayling et al., 2016) geothermal sites. The largerG values

indicate that, following the stimulation programs, the flow paths are more heterogeneous,

i.e., a larger fraction of the fluid flow was supported by fewer key flow paths.

3.4.2 Post-stimulation characterization

To further describe the effect of the hydroshearing and hydrofracturing stimulations on

the subsequent tracer tests, only the RTD curves from the post-stimulation tests are

plotted in Fig. 3.4, where tracers were injected either into INJ2-i2 (dashed lines) or INJ1-

i4 (solid lines). In the following, we group the results and discussions according to the

injection intervals, namely INJ2-i2 and INJ1-i4.

Injection into INJ2-i2

The interval INJ2-i2 cuts through part of the S1 shear zone (Fig. 3.1) and closely corre-

sponds to hydroshearing Interval HS1 (39.75-40.75m, Table 3.3) (Doetsch et al., 2018a).

Three tracer tests used INJ2-i2 as the injection interval: they are Tests 3B, 5 and 7. Dur-

ing Tests 3B and 5, the injection interval spanned 39.73-40.25m, whereas during Test 7,

it was changed slightly to 38.40-39.40m (Table 3.2) to include fractures that were con-

tained in hydrofracturing Interval HF6 (Table 3.3) (Doetsch et al., 2018a). During all

three tests, the tracers were recovered only from PRP2-i1.

The RTD curves obtained at PRP2-i1 during Tests 3B and 5 have almost identical

tracer first arrival times (Figs. 3.4 and 3.5). Moreover, both show multiple peaks or


shoulders, which are often attributed to individual preferential flow paths (Moreno and

Tsang, 1991; Siirila-Woodburn et al., 2015; Stoll et al., 2019). Such an interpretation

suggests that Test 3B likely encountered two major flow paths, causing a peak to occur

after ∼230min and a shoulder after ∼600min, whereas Test 5 encountered three major

flow paths, producing peaks at ∼180 , ∼240 and ∼400min (Fig. 3.4). The parameter

values derived from the moment analysis of the two RTD curves are presented graphically

in Fig. 3.5. They show that the mean residence time and swept volume of the tracer at

PRP2-i1 during Test 5 is almost twice as large as during Test 3B, which may be due

to the injection of a salt-ethanol-water mixture (Shakas et al., 2017; Leresche, 2018) for

90min immediately after the dye tracer injection during Test 5 (Table 3.1). Otherwise,

the hydraulic boundary conditions between Tests 3B and 5 were largely similar. Although

the salt-ethanol-water mixture is neutrally buoyant when prepared with correct quantities

(Shakas et al., 2017), there may be other unknown effects that played a role in pushing

the dye tracer into the fracture network. The tracer recoveries at PRP2-i1 were similar

(∼9%) during both tests.

Test 7 featured a slightly different interval and, importantly, was conducted after the

hydrofracturing stimulations, with one injection (HF6) being performed at an interval

that overlapped INJ2-i2. The resultant effect on the RTD curve obtained at PRP2-i1 is

as follows: i) the tracer arrives almost 100minutes sooner than during Tests 3B and 5,

ii) the mean residence time and second moment, m2,c, show the lowest values obtained

during the tracer tests reported here and iii) the recovery increases sixfold to 59%, even

though the swept volume increases by a factor of only two to 0.094m3 (Fig. 3.5 and

Supplementary Table A.5). These three observations indicate that most of the tracer

mass is transported quickly through the flow paths to the PRP2-i1 outlet, as indicated by

low t∗ and m2,c values, and that most of the recovered tracer is provided by a relatively

small fraction of the swept volume, i.e., through one or a few preferential flow paths,

resulting in a high G value.

Injection into INJ1-i4

The interval INJ1-i4 closely corresponds to the hydroshearing stimulation Interval HS4

(27.20-28.20m, Table 3.3), which targeted Shear Zone S3.1 (Doetsch et al., 2018a). It

was first used as a tracer injection interval during Tests 4A and 6A, which followed the

hydroshearing stimulation. The results of Test 4A were presented in detail in Kittilä et al.

(2019), where the RTDs of solute dye tracers were compared with those of particulate

DNA nanotracers. Here, the results of Test 4A are analyzed further to characterize the

hydraulic properties of the test volume. During Test 4A, tracer RTD curves were obtained

at INJ2-i4, AU Tunnel, PRP1-i3 and PRP2-i2 (Table 3.1 and Fig. 3.5). During Test 6A,


30 100 1000 2000Time [min]

10-9

10-8

10-7

10-6

10-5

10-4

10-3

10-2

E [

1/m

in]

Test 3B, PRP2-i1Test 3A, PRP1-i3Test 4A, PRP1-i3Test 4A, PRP2-i2Test 4A, AU TunnelTest 4A, INJ2-i4Test 5, PRP2-i1Test 6A, PRP1-i3Test 7, PRP2-i1

First peakSecond peakThird peak

Figure 3.4 – Post-stimulation residence time distribution (RTD) curves of the tracersfrom Tests 3 to 7. E (t) is the age distribution function. The solid and dashed linesdenote injection into INJ1-i4 and INJ2-i2, respectively. The squares on the RTD curvesindicate the peaks that are associated with major flow paths (see main text), as thelogarithmic scale hides some peaks.

tracer monitoring only took place at PRP1-i1 and PRP1-i3. The following describes the

interpretation of the test results at each monitoring location.

The RTD curve obtained at INJ2-i4 during Test 4A features a remarkably late first

arrival time and large mean residence time, given that the distance from the injection

location is only 10.9 m (Figs. 3.4 and 3.5). Note that a pump was used at INJ2-i4

to enable sufficient outflow for water sampling. The tracer recovered at INJ2-i4 swept

a considerably larger volume compared to the corresponding volumes derived from the

RTD curves at PRP1-i3 and PRP2-i2, which are relatively close to the tracer injection

interval INJ1-i4. Furthermore, the low Gini coefficient of 0.33 at INJ2-i4 indicates that

fluid flow was rather homogeneously distributed within the swept volume with no evidence

for preferential flow paths.

The RTD curves at PRP1-i3 during Tests 4A and 6A (Fig. 3.4) show almost identical

first arrival times (Fig. 3.5), albeit with slight dissimilarities in curve shape and peak

timing. These dissimilarities may have been caused by the injection of salt water (31.4mS

cm−1) into INJ1-i4 for 73 minutes (Doetsch et al., 2018a) before the injection of the dye

tracer during Test 6A.


INJ1-i4

INJ1-i2

PR

P1-i2

PR

P1-i3

PR

P2-i1

PR

P2-i2

AU

Tunnel

INJ2-i4

INJ2-i2

Po

st-

sti

mu

lati

on

Te

st

3B

Te

st

3A

Te

st

4A

Te

st

4A

Te

st

4A

Te

st

4A

Te

st

5T

est

6A

Te

st

7

0

10

20

30

40

50

60

65

0

10

20

30

40

10

50

100

150

200

250

300

330

10

500

1000

1500

2000

2500

2770

02468

10

1.4

35

10

15

17

×10

-4

0.0

01

0.0

1

0.0

2

0.0

4

0.0

6

0.0

8

0.1

0.2

0.3

0.4

0.5

0.6

0.0

71357×

10

6

INJ1-i4

INJ1-i2

PR

P1-i2

PR

P1-i3

PR

P2-i1

PR

P2-i2

AU

Tunnel

INJ2-i4

INJ2-i2

Monitoring

location

∆h

[m]

Dis

tance

[m]

First arr

ival

[min

]M

ean r

esid

ence tim

et*

[m

in]

Recovery

R [%

]C

pe

ak/C

inj

[-]

Sw

ept volu

me

Vp [m

3]

Gin

i coeffic

ient

G [-]

2nd m

om

ent

m2

,c [m

in2]

Inje

ction

location

3376

6.0

4E

-03

0.9

2

59

59

7.7

1E

+06

Figu

re3.5–

Parallelcoordinatesplot

oftheparametersderiv

edfrom

theresidencetim

edistrib

ution(R

TD)curves

from

teststhat

follow

the

hydroshearingstim

ulationprogram.The

Test

3Acurvein

red(injectedinto

INJ2-i4

)is

show

nas

areferenceto

Fig.

3.3.

Tracersinjected

into

INJ2-i2

areplottedas

dashed

lines

andtracersinjected

into

INJ1-i4

areplottedas

solid

lines.The

data

inthis

graphcanalso

befoun

din

the

Supp

lementary

TableA.5.


The RTD curve obtained from PRP2-i2 during Test 4A exhibits only one peak (Fig. 3.4)

whose concentration in relation to the injection concentration is high, i.e., a ratio of

1.70×10−3 (Fig. 3.5). However, the recovery and swept volume of 2.7% and 0.023m3, re-

spectively, are small in comparison to the corresponding values from INJ2-i4 (R=5.4% and

Vp=0.089m3), and are more similar to those inferred from the RTD curves from PRP1-i3

(R=3.1% and Vp=0.038m3, Fig. 3.5). These observations, together with the earlier first

tracer arrival times and shorter mean residence times at PRP1-i3 and PRP2-i2, compared

to INJ2-i4, suggest that the tracer most likely arrived at PRP1-i3 and PRP2-i2 through

short-circuiting flow paths. Yet, those flow paths do not receive significant amounts of

the injected fluid (as indicated by the low tracer recoveries at PRP1-i3 and PRP2-i2).

The RTD curve obtained at the AU Tunnel in Test 4A has a relatively late first

arrival time and a high peak concentration, features that were seen in the RTD curves

at this location during the pre-stimulation and post-hydrofracturing tests when tracers

were injected into INJ2-i4 (Fig. 3.2). The Gini coefficient of 0.36 (Fig. 3.5) suggests

that the permeability distribution between INJ1-i4 and the AU Tunnel is rather homoge-

neous for the site. The large values of first arrival time, mean residence time and swept

volume (291min, 2595min and 0.92m3, respectively) can mainly be attributed to the

long distance of 28.7 m between the injection location and the AU Tunnel outflow point.

Furthermore, the high tracer recovery of 59%, with respect to a rather low Cpeak/Cinj

value of 2.32×10−4, can be explained by the high water outflow rate at the AU Tunnel

(Table 3.2), which is promoted by the natural hydraulic head gradient towards the AU

Tunnel (Jalali et al., 2018a; Krietsch et al., 2018).

3.4.3 Tomographic inversion

The tomograms of hydraulic conductivity, K, derived from the arrival times of the tracer

concentration peaks during Tests 1 and 2 (the period before the hydroshearing stimula-

tions) and Tests 3 to 6 (the period between the hydroshearing and hydrofracturing stimu-

lations), are shown in Figs. 3.6a and 3.6b, respectively. Because of the sparse peak arrival

time data in the y-direction (Fig. 3.1b), the spatial distribution of the reconstructed K

values was restricted to a 5 m wide, vertical E-W-striking section that included the AU

Tunnel sampling point (blue dashed lines in Fig. 3.1). Sparse data also limited the distri-

bution of resolved K values near the AU Tunnel (Fig. 3.1). Note that the injection and

monitoring locations are simply projected onto the 5 m thick section, even though they

may lie outside the section (e.g., INJ2-i2 in Fig. 3.1b). Changes in hydraulic conductivity

between the pre- and post-stimulation tomograms represent changes integrated over the

5 m length of the inversion cells in the north-direction.


The distribution of K values prior to the shearing stimulation shown in Fig. 3.6a,

indicates large variations and clustering of the K values. The sample points INJ1-i4,

PRP1-i3 and PRP2-i2 lie around the boundary of a low-K zone, whereas a high-K zone

is identified around INJ2-i3 and INJ2-i4. An intermediate-K zone is identified towards

the AU Tunnel outflow point. Strong heterogeneity of the reconstructed K values is also

shown by the bimodal histogram (inset in Fig. 3.6a), where the highest frequencies are

at logK = −7.5 to −7 and −5 to −4.5 , with K given in m s−1. The strong contrast in

K between boreholes INJ1 and INJ2 is consistent with the high contrast in ground pen-

etrating radar (GPR) velocities, imaged between these two boreholes before the shearing

stimulation (Doetsch et al., 2018a), where a zone of low GPR velocities identified around

the INJ2 borehole from approximately 20m downwards contrasts with a zone of higher

GPR velocities surrounding the INJ1 borehole. In general, lower GPR velocities indicate

higher water content, i.e., higher porosities, which are typically associated with higher hy-

draulic conductivities (Doetsch et al., 2018a). On the other hand, our results somewhat

contradict the results obtained from pre-stimulation cross-hole hydraulic tests (Brixel

et al., 2020) and from hydraulic tomography results that are based on a simplified dis-

crete fracture network (DFN) concept (Klepikova et al., 2020). Both of these analyses

showed that the INJ1 and INJ2 boreholes exhibit good hydraulic connections across the

S3 shear zone. However, the hydraulic tests conducted during those two studies employed

2m-long borehole intervals and no water production from monitoring locations, i.e., the

hydraulic boundary conditions and the experimental set-up were different from the tracer

tests presented here. Whether these differences can account for the contradiction in the

results is uncertain.

It appears that the shearing stimulation not only enhanced the equivalent hydraulic

conductivity field of the test rock volume (Fig. 3.6b), but also narrowed the span of

inferred K values (inset of Fig. 3.6b). For example, the shearing stimulation converted a

highly heterogeneous (with respect to the K values) region between the INJ1 and INJ2

boreholes into a rather homogeneous region of intermediate K values. Doetsch et al.

(2018b) show that most of the seismic events during HS4 were concentrated within this

zone to the east of INJ1-i4.

Towards the AU Tunnel from the INJ2 borehole extends a region with relatively

high pre-stimulation K values, which were further enhanced following the stimulation

(Figs. 3.6a and 3.6b). This result is in good agreement with the seismic tomography

results at the study site, which suggest that a low-velocity zone is present between the

two S3 shear zones (Krietsch et al., 2018). This low water flow velocity zone is attributed

to a highly fractured zone which provides an outflow into the AU Tunnel.


a) b)

0 10 20 30 40 50

Observed [days]

0

10

20

30

40

50

Reconstr

ucte

d [days]

c)

0 0.2 0.4 0.6 0.8 1 1.2

Observed [days]

0

0.2

0.4

0.6

0.8

1

1.2

1.4

Reconstr

ucte

d [days]

d)

Figure 3.6 – Sections through tomograms of the hydraulic conductivity, K, distri-butions reconstructed using the peak arrival times of the tracer breakthrough curves(BTCs) before (a) and after (b) the hydraulic shearing stimulations. The boundaries ofthe 5 m wide sections are denoted by the pair of dashed blue lines in Fig. 3.1. The dis-tribution of K values was the same throughout the thickness of the sections. The insetsin (a) and (b) provide the histograms of the hydraulic conductivity, K, distributions.The tracer injection/monitoring intervals are also shown in the respective sections. Theobserved travel times are plotted against the tomographically reconstructed ones for thepre- (c) and post-stimulation (d) calculations. The error bars in (c) and (d) representstandard deviations of the travel times, obtained by different positions of the staggeredgrids.


The hydraulic conductivity field surrounding the injection interval INJ2-i2, associated

with the S1 shear zone (Fig. 3.1), could only be reconstructed after the hydroshearing

stimulation. The reconstructed tomogram indicates that INJ2-i2 is located in a region

with high K values. Interval PRP2-i1, which is also associated with the S1 shear zone

and which was the only monitoring location that produced a signal when tracers were

injected into INJ2-i2, is located immediately outside this high-K region. INJ2-i2 and

PRP2-i1 appear to be connected with the other injection and monitoring locations in the

post-stimulation tomogram. However, it should be emphasized that the lower part of the

tomogram is essentially a reconstruction of the hydraulic connectivity in the S1 shear

zone, which is offset towards the north of the tomogram cross section (see Fig. 3.1b).

The apparent connection is due to the combined effects of the coarse inversion grid in

the y-direction, and a reconstruction that is based on only two data points that are also

located near the border of the inversion domain. Thus, to improve the reconstruction of

the hydraulic conductivity field across the S1 and S3 shear zone structures, the density

of peak arrival time data in the y-direction would need to be increased.

A comparison of the observed and reconstructed tracer travel times for the pre- and

post-stimulation tracer tomography inversions is provided in Figs. 3.6c and 3.6d, respec-

tively. The different staggered fields yield slightly different residuals of the observed tracer

travel times, which increases the standard deviation of the reconstructed travel times. In

the pre-stimulation tomographic inversion, following the approach of Somogyvári and

Bayer (2017), a virtual travel time was assigned from the INJ2-i3 to the INJ1-i4 source-

receiver pair in order to improve the tomographic reconstruction at locations where no

tracer breakthrough time could be determined. In this study, a value of K=1×10−6m s−1

is selected for the calculation of the virtual travel time. The observed and reconstructed

travel times align well with the identity line (1:1), indicating a satisfactory reconstruction

of the hydraulic conductivity fields in the test volume. The data points from the pre-

stimulation experiment show larger variance, mainly due to fewer observations (note that

the scales of the plots are different). The travel-time-based tomographic inversion was

designed to reconstruct smooth parameter transitions and not to capture local disconti-

nuities. Considering this limitation, the match between the observed and reconstructed

travel times is considered good, particularly given the geological conditions at the test

site, such as the water-conductive and connected fractures embedded in the tight host

rock.

Table 3.5 documents the statistical parameters of the logK distribution (minimum,

maximum, geometric mean, standard deviation and Dykstra-Parsons heterogeneity index,

VDP ) from the tomographic inversion profiles of the pre- and post-stimulation tracer tests.

The maximum reconstructed value of K is 1.12×10−4m s−1 (logK = −3.95), which is


Table 3.5 – Statistical parameters for the tomographic inversion profiles of logK.

Tomogram min max geometric mean σlogK VDP [–]pre-HS -7.20 -3.98 -5.03 0.95 0.89post-HS -5.43 -3.95 -4.63 0.39 0.59

approximately an order of magnitude larger than the hydaulic conductivity estimates re-

ported by Jalali et al. (2018a). This overestimation may be caused by a number of factors.

Firstly, the mean tracer velocities derived from the inversion are generally larger than the

actual seepage velocities, due to the influence of dispersion on the tracer travel times. Sec-

ondly, the mean porosity of the rock in the test volume is taken to be 0.25, which may be

too high, although it is worth noting that large cavities are present within the connected

fracture porosity of the rock mass. Thirdly, the transmissivity values reported by Jalali

et al. (2018a) sample a volume which is likely to be smaller than the volume used for the

estimation of hydraulic conductivities from tracer tests. Often, the hydraulic conductivity

tends to increase with the scale over which it is measured (Clauser, 1992; Sanchez-Vila

et al., 2006; Saar, 2011). Thus, we conclude that the presented tracer tomography inver-

sion yielded satisfactory estimates of the changes in the hydraulic conductivity, K, field

for the studied rock volume due to hydraulic stimulation.

To conclude, the comparison of the pre- and post-stimulation tomograms indicates

that fluid flow was accessing pathways with higher hydraulic conductivities, K, after

the hydroshearing stimulation. Compared to the pre-stimulation tomogram, the post-

stimulation tomogram shows an increase in the geometric mean of K and a decrease

in both σlogK and VDP (Table 3.5). These changes indicate that the permeability (hy-

draulic conductivity) has been enhanced by the hydraulic shearing stimulation experi-

ments (at least temporarily) and that new hydraulic connections have been created (at

least temporarily). Indeed, before the hydraulic shearing stimulation, regions with high

K values appeared to be more isolated and, hence, high σlogK and VDP values were ob-

tained. In fact, the VDP = 0.89 value from the pre-stimulation K field indicates that the

tracer swept volume was ‘extremely heterogeneous’, as stated also by Tiab and Donald-

son (2015). Therefore, one may state that the hydraulic shearing stimulation converted

the ‘extremely heterogeneous’ to a merely ‘very heterogeneous’ hydraulic conductivity

system.

3.5 Conclusions 89

3.5 Conclusions

Using temporal moments and tomograms derived from tracer tests conducted at the

Grimsel Test Site (GTS), we have quantified the effects of hydraulic stimulation experi-

ments on fluid flow and solute mass transport in a fractured crystalline rock volume. We

observe a clear increase of 43% to 316% in the tracer swept volume after the stimulation

experiments. However, the other temporal moments (i.e., tracer recovery, tracer mean

residence time and variance of the tracer response) and the Gini coefficient derived from

them yielded inconclusive results. This once more illustrates that the investigated rock

volume at the GTS is highly heterogeneous with respect to hydraulic conductivity, as we

have observed in a previous study which presented a comparison of DNA nanotracers and

dyes as tracers (Kittilä et al., 2019).

The hydraulic properties of the S1 and S3 shear zone structures in the rock volume

were mainly investigated by injecting tracers into INJ2-i2 (S1) or INJ1-i4 and INJ2-

i4 (S3) (Table 3.2). These tests demonstrated that the S1 and S3 shear zones do not

appear to communicate via those injection intervals. However, the recovery of tracers

injected into INJ1-i2 (associated with the S1 shear zone) at monitoring locations in the

S3 shear zone during Test 4B (Kittilä et al., 2019) suggests that these shear zones do

have a weak hydraulic connection, at least (immediately) following hydraulic stimulation.

The results also confirm the significance of the drainage effect of the AU Tunnel and the

highly fractured zone (Krietsch et al., 2018) on fluid flow in the studied rock volume.

The results of the tracer tomographic inversions obtained both before and after the

hydraulic stimulation experiments agree well with other geophysical (GPR and seismic)

tomograms from the study site (Doetsch et al., 2018a,b). Unlike the temporal moments,

the tracer tomographic inversion enables spatial evaluation of the hydraulic conductivity

distribution. Nonetheless, a quantitative analysis of the tracer residence time distribution

curves (RTDs), employing temporal moment analysis, permits the determination of the

swept volume and other fluid flow and solute mass transport parameters that are central

for the hydraulic characterization of a subsurface formation, such as a geothermal system

(Grant, 2016). The tracer-based hydraulic characterization of a stimulated rock volume

can help determine, for example, whether flow paths were predominantly opened/created

or constricted and, in operated geothermal systems, whether an early breakthrough of

the cold front is expected.

Acknowledgements

The Werner Siemens Foundation (Werner Siemens-Stiftung) is thanked by M.O.S. for its

support of the Geothermal Energy and Geofluids (GEG.ethz.ch) Group at ETH Zurich.

We also thank N. Knornschild for his invaluable technical contributions in the GEG

group laboratory and in the field as well as F. Leuenberger for her support with the so-

lute dye tracer analyses. This research project was financially supported in part by the

Swiss Innovation Agency Innosuisse, Switzerland, and is part of the Swiss Competence

Center for Energy Research - Supply of Electricity (SCCER– SoE). Further financial

support is gratefully acknowledged from ETH Zurich. The Grimsel Test Site (GTS)

is operated by the National Cooperative for the Disposal of Radioactive Waste (NA-

GRA). We are indebted to NAGRA for hosting the in-situ Stimulation and Circulation

(ISC) experiment at their GTS facility and to the NAGRA technical staff for onsite sup-

port. We thank the reviewers for their invaluable comments that improved the quality

of this manuscript. The data used in this study are available at https://www.research-

collection.ethz.ch/handle/20.500.11850/348073.

91

4Solute tracer test quantification of the effects ofhot water injection into hydraulically stimulated

crystalline rock

Revised and submitted as:

A. Kittilä, M.R. Jalali, M.O. Saar, and X.-Z. Kong (2020), Solute tracer test quantifi-

cation of the effects of hot water injection into hydraulically stimulated crystalline rock,

Geothermal Energy.

93

Abstract

When water is injected into a fracture-dominated reservoir that is cooler or hotter than

the injected water, the reservoir permeability is expected to be altered by the injection-

induced thermo-mechanical effects, resulting in the redistribution of fluid flow in the

reservoir. These effects are important to be taken into account when evaluating the per-

formance and lifetime particularly of Enhanced Geothermal Systems (EGS). In this paper,

we compare the results from two dye tracer tests, conducted before (at ambient tempera-

ture of 13 °C) and during the injection of 45 °C hot water into a fractured crystalline rock

at the Grimsel Test Site in Switzerland. Conducting a moment analysis on the recovered

tracer residence time distribution (RTD) curves, we observe, after hot water injection, a

significant decrease in the total tracer recovery. This recovery decrease strongly suggests

that fluid flow was redistributed in the studied rock volume and that the majority of

the injected water was lost to the far-field. Furthermore, by using temperature measure-

ments, obtained from the same locations as the tracer RTD curves, we conceptualize an

approach to estimate the fracture surface area contributing to the heat exchange between

the host rock and the circulating fluid. Our moment analysis and simplified estimation

of fracture surface area provide insights into the hydraulic properties of the hydraulically

active fracture system and the changes in fluid flow. Such insights are important to assess

the heat exchange performance of a geothermal formation during fluid circulation and to

estimate the lifetime of the geothermal formation, particularly in EGS.

95

96 4 Solute tracer test quantification of the effects of hot water injectioninto hydraulically stimulated crystalline rock

Nomenclature

φ Porosity [-]

Φ (t) Storage capacity [-]

ρ density [kg/m3]

A Surface area [m2]

b Aperture of the fracture [m]

Cp,i Heat capacity(i=water W, rock R) [J/kg/K]

E (t) Tracer residence time [1/s]

F (t) Flow capacity [-]

G Gini coefficient [-]

KR Thermal conductivity of the rock [J/s/m/K]

L Fracture length [m]

m∗0 Zeroth temporal moment [-]

m∗1 First temporal moment [s]

m2,c Second centralized and normalized temporal moment [s2]

Minj Mass of tracer injected [kg]

qinj Volumetric injection rate [m3/s]

qpro Volumetric production rate [m3/s]

R Recovery [%]

S Surface area of fracture [m2]

t Time [s]

97

98 4 Solute tracer test quantification of the effects of hot water injectioninto hydraulically stimulated crystalline rock

t∗ Mean residence time [s]

TI Initial temperature [° C]

TJ Injected temperature [° C]

TR Temperature in the rock [° C]

TW Temperature in the fracture [° C]

Vp Swpt volume [m3]

W Fracture width [m]

z Axis perpendicular to the fracture [m]

4.1 Introduction 99

4.1 Introduction

To produce energy from geothermal resources, fluid is injected and circulated through

natural or artificially created reservoirs, the latter being so-called Enhanced/Engineered

Geothermal Systems (EGS) (Tester et al., 2006; Evans, 2015). As the fluid flows through

the natural or artificial reservoir, heat is exchanged between the host rock and the cir-

culating fluid. The performance of a geothermal system, in terms of fluid circulation

and heat extraction, depends on several factors, where the reservoir impedance, the heat

recovery, and the tracer-swept volume are key factors (Tester et al., 2006; Grant, 2016;

Olasolo et al., 2016). The larger the heat exchange surface area between the fluid and the

rock, the better the geothermal resource can be exploited and the longer is the expected

lifetime of the reservoir before cold-front breakthrough occurs at the production well.

However, thermo-hydro-mechanical-chemical responses of the formation can have detri-

mental effects on the performance of a geothermal reservoir (André et al., 2006; Taron

and Elsworth, 2009; Fu et al., 2016; Pandey et al., 2017).

Whenever the temperature of the injected fluid is different to the one of the reservoir,

particularly in crystalline, fractured rock, it is likely that thermal, mechanical, and chem-

ical processes alter the hydraulic properties of the rock mass. It has been reported that

both heat extraction and heat storage in fractured reservoirs can cause rock deformation,

inducing changes in fracture aperture widths (Fu et al., 2016; Pandey et al., 2017). In

fracture-dominated systems, such as EGS, heat production, by injecting water that is

cooler than the natural rock, eventually results in flow channeling, i.e., concentration of

fluid flow in cooled zones (Fu et al., 2016). In a heating experiment reported by Rutqvist

et al. (2001), elevated fluid temperatures caused the surrounding rock to expand, yielding

vertical rock displacement. Kumari et al. (2018) conducted flow-through experiments in

a granite core under different temperature conditions and demonstrated that increasing

the temperature of the injected fluid from 20 °C to 300 °C, caused an 86% reduction in

permeability before reaching 100 °C. Recently, Grimm Lima et al. (2019) observed a 20-

75% decrease in the hydraulic apertures of naturally fractured granodiorite cores from the

Grimsel Test Site (GTS) in Switzerland (which is also the field site in this paper), when

increasing the temperature of the system from 22 °C to 140 °C. Kumari et al. (2018) and

Grimm Lima et al. (2019) identified that the reductions in permeability and in fracture

apertures as well as the closure of fractures are due to the thermal expansion of the rock.

Moreover, Grimm Lima et al. (2019) demonstrated pressure dissolution of contacting

asperities.

Although the injection of heat induces the aforementioned thermo-mechanical effects,

heat is a well-established tracer in the subsurface both as a natural (Saar, 2011) and an ar-

100 4 Solute tracer test quantification of the effects of hot waterinjection into hydraulically stimulated crystalline rock

tificially introduced tracer, with typical applications in reservoir characterization (Colom-

bani et al., 2015; Irvine et al., 2015; Ayling et al., 2016; Sarris et al., 2018) and tomog-

raphy (Linde et al., 2006; Somogyvári et al., 2016; Somogyvári and Bayer, 2017). Heat

can be a valuable tool in acquiring additional information on subsurface fluid flow and

(solute, energy) transport processes, particularly when estimating permeability (McCord

et al., 1992; Anderson, 2005). However, the solute and thermal Peclet numbers can dif-

fer by orders of magnitude for the same Darcy flow velocity (de Marsily, 1986), which

can cause problems, for example, when solute transport is of interest but only heat is

used as a tracer. This discrepancy between solute and heat transport is particularly se-

vere in fine-grained sediments, as heat transport is relatively insensitive to changes in

longitudinal dispersivity, which is a particularly relevant parameter for solute transport

modeling (Rau et al., 2012; Giambastiani et al., 2013). Notwithstanding these differences,

Marschall et al. (1995) used solutes and heat to investigate the role of diffusive transport

in a fractured rock. However, in that experiment the thermal expansion of the rock was

not considered, although an increase in injection pressure and a decrease in injection flow

rate were observed. The study of Kumari et al. (2018), on the other hand, considered the

temperature effect on permeability during water injection at the core scale. Colombani

et al. (2015) validated the solute transport model with the heat transport in a sandy

aquifer. Kocabas (2005) and Ma et al. (2012) compared solute and heat transport in

evaluating the hydraulic properties of the subsurface.

In this study, we estimate fluid flow characteristics (solute tracer recovery, mean res-

idence time, swept volume, and mean fluid velocity) of a fractured and stimulated crys-

talline rock mass at ambient temperatures (about 13 °C) and approximately two weeks

after the start of hot water (about 45 °C) injection. These estimated characteristics are

derived from two dye tracer tests, conducted before and during the hot water injection.

Our experiments were conducted at the Grimsel Test Site (GTS) in Switzerland, which is

operated by the Swiss National Cooperative for the Disposal of Radioactive Waste (Na-

gra). By comparing those results before and during hot water injection, we investigate the

effects of the hot water injection on the fluid flow characteristics of the fracture-dominated

rock mass. Moreover, at the same locations, where solute tracer breakthrough curves were

recorded, temperature measurements were collected throughout the hot water injection

period. Using these temperature measurements, we estimate the fracture surface area

that may have contributed to the heat exchange between the host rock and the circu-

lating fluid in the fractures. These estimates, based on a simple parallel plate model,

facilitate delineating the fracture geometry at the test site. Our results can be useful for

establishing a three-dimensional discrete fracture network (DFN) model of the study site.

With the decameter-scale field tracer experiments presented here, we can quantitatively

4.1 Introduction 101

address the knowledge gap regarding inter-well changes in fluid flow properties that are

associated with hot water injection into a fracture-dominated crystalline rock.


4.2 Test site

The Grimsel Test Site (GTS) is located in the Swiss Alps at 1733m a.s.l. with 400-500m

of overburden, at the boundary between the crystalline rocks of the Central Aare Granite

and the Grimsel Granodiorite (Keusen et al., 1989). Between years 2015 and 2017, the

GTS hosted the in-situ Stimulation and Circulation (ISC) experiment, which aimed at

studying the thermo-hydro-mechanical and seismic (THMS) processes relevant for per-

meability enhancement during high pressure fluid injections at the decameter scale, and

to evaluate the creation of a sustainable heat exchanger (Amann et al., 2018; Krietsch

et al., 2018). Although the GTS is cold (the ambient rock temperature is 13 °C) and shal-

low (approximately 500 m deep), in comparison to actual EGS conditions, the test site

enables detailed characterizations of the rock mass and comprehensive observations of the

permeability enhancements to be made during actual EGS developments. The rock mass

permeability was shown to be enhanced, at least temporarily, after the hydraulic stimu-

lation experiments (first hydraulic shearing and then hydraulic fracturing). Additionally,

new hydraulic connections, enabling larger tracer-swept volumes, were observed (Kittilä

et al., 2020).

At the GTS, two distinguishable shear zones are recognized, intersecting the granodi-

oritic host rock: i) the ductile NEN-SWS striking S1 and ii) the younger brittle-ductile

E-W striking S3 shear zone (Fig. 4.1 and Keusen et al. (1989)). There exists also a duc-

tile shear zone, classified as S2, which is slightly discordant to S1, however, the S1 and

S2 shear zones cannot be distinguished in the field (Keusen et al., 1989; Krietsch et al.,

2018). The dominant S3 shear zone is composed of two structures that are associated with

biotite-rich metabasic dykes that are up to 1m thick and approximately 2.5m apart (Kri-

etsch et al., 2019). The two S3 shear zone structures, varying in thickness from 38mm

to 312mm (Krietsch et al., 2018), bound a highly fractured zone, where the fracture den-

sity is about 20m−1 (Jalali et al., 2017; Krietsch et al., 2018). Most of the open fractures,

identified in the INJ1 and INJ2 boreholes (Fig. 4.1), employing optical televiewer (OPTV)

logs, are associated with this highly fractured zone (Jalali et al., 2018b; Krietsch et al.,

2018). There are also some partially open fractures (Jalali et al., 2018b; Krietsch et al.,

2018), containing cataclasites, breccias, and fault gouge (Ziegler et al., 2013). The highly

fractured zone is seen in seismic tomograms as a low-velocity zone (Krietsch et al., 2018)

and in tracer tomograms as a zone of high hydraulic conductivity (Kittilä et al., 2020).

The host rock beyond the shear zone structures is remarkably intact, with 0-3 fractures

per meter (Gischig et al., 2018).

Fluid flow at the ISC test site is dominated by the highly fractured zone and the

drainage effect of the AU Tunnel (Jalali et al., 2017; Krietsch et al., 2018; Kittilä et al.,

4.2 Test site 103

Figure 4.1 – Overview of the experiment setup at the Grimsel Test Site (GTS) (mod-ified from Krietsch et al. (2018)). The shear zone structures, designated S1 and S3,are shown as red and green planes, respectively. There is also a shear zone classifiedas S2, which is slightly discordant to S1, however, the S1 and S2 shear zones cannotbe distinguished in the field (Keusen et al., 1989; Krietsch et al., 2018). The injectionand monitoring intervals, associated with this study, in boreholes INJ1, INJ2, PRP1,and PRP2 (black cylinders) and the AU Tunnel outflow point (black circle) are marked.The orange arrows indicate the interpreted flow directions in injection interval INJ2-int4, where the arrows with solid lines are associated with the more prominent flowdirections (see Section ’Redistribution of fluid flow’ for more information).


2020), with an average natural discharge of 100mlmin−1 at the AU Tunnel (Jalali et al.,

2018b). The transmissivity of the shear zones ranges from 10−12 to 10−6 m2 s−1 (Brixel

et al., 2020), and in the intact rock, the average transmissivity is less than 10−13 m2 s−1 (Keusen

et al., 1989; Jalali et al., 2018a).

4.3 Methods 105

Table 4.1 – Summary of tracer injection and production during Tests 8 and 9.

Metric Test 8 Test 9Test start 22.11.2017 13.12.2017Hot water injection 29.11.2017 – 10.01.2018Test end 29.12.2017 20.12.2017Injection interval INJ2-int4Injection depth [m] 22.89-23.89Injection flow rate, Qinj [L min−1] 2.1 1.8Injection pressure, Pinj [kPa] 374 450Injection temperature, TJ [°C] 13 45Tracer EosineInjected tracer mass, Minj [mg] 19.5 19.0Injected tracer volume, Vinj [L] 0.975 0.950Monitoring locations AU Tunnel, PRP1-int3

PRP2-int2, PRP1-int2Production flow rate, Qpro [L min−1] 1a, 0.090 0.85a, 0.051

0.069, 0.12 0.043, 0.12Distance to monitoring location [m] 20.7, 4.6, 6.4, 8.6aThe production flow rate at the AU Tunnel is estimated for bothTest 8 and Test 9, as no measurements are available. Details are givenin the Section Results and discussion.

4.3 Methods

4.3.1 Tracer experiments

Two solute dye tracer tests were conducted during the ‘Circulation’ phase of the ISC

experiment (no actual fluid reinjection took place). The objective was to study the effects

of hot water injection on fluid flow in the stimulated fractured rock mass. The first test,

namely Test 8, was conducted immediately before the start of the hot water injection.

The second test, namely Test 9, was conducted during the hot (45 ◦C) water injection,

which, at that time, had been continued for two weeks. In both tests, the solute dye

tracers were injected into the fractured rock as a short pulse. Table 4.1 summarizes the

details of tracer injection and production during Tests 8 and 9, and Figure 4.2 shows the

injection temperature, pressure, and flow rate during the hot water injection experiment.

The tracers and the hot water were injected into Interval 4 in Borehole INJ2 (hereafter

INJ2-int4), while tracer concentrations and water temperature were monitored at borehole

intervals INJ1-int4, PRP1-int3, PRP2-int2, and PRP1-int2 as well as at the AU Tunnel

outflow point (Fig. 4.1 and Table 4.1). However, at INJ1-int4 the production flow rate

was approximately 1mL min−1 during Test 8 and during Test 9, the outflow at that

location had already ceased. Hence, no tracer data were recovered from INJ1-int4.

Injection of tap water was continued throughout the ‘Circulation’ phase, with a mean

flow rate of 2.1 L min−1 during Test 8 and 1.8 L min−1 during Test 9. While the in-

jection flow rate decreased, the injection pressure (absolute pressure at the wellhead)


21.11.2017 28.11. 5.12. 12.12 19.12 26.12. 2.1.2018 9.1.0

10

20

30

40

50

60T

empe

ratu

re [°

C]

0

1

2

3

4

5

6

Flo

w r

ate

[L/m

in]

0

100

200

300

400

500

600

Pre

ssur

e [k

Pa]

Test 8 Test 9

2-day breakdown

Figure 4.2 – Injection temperature, pressure, and flow rate during fluid injection intoINJ2-int4 (modified from Doetsch et al. (2018a)). It is worth noting that the systemwas likely not yet at steady-state during Test 8.

increased from about 374 kPa during Test 8 to about 450 kPa during Test 9 (Table 4.1

and Fig. 4.2). At the monitoring locations, tracer signals were continuously monitored

using flow-through fluorometers (GGUN-FL30). The signals (in millivolts) from these

fluorometers were converted to tracer concentrations using laboratory-analyzed discrete

samples (Luminescence Spectrometer, Perkin Elmer, LS 50 B), collected at the monitoring

locations (Table 4.1).


In this study, the effect of hot water injection on fluid flow redistribution in the fractured

crystalline rock at the GTS is characterized using tracer-determined residence time distri-

bution (RTD) curves. It is well known that RTD curves can be described statistically by

determining the mode (tracer recovery), integral mean (tracer mean residence time), and

width (dispersion) of the distribution (Robinson and Tester, 1984; Leube et al., 2012).

Further interpretation of the RTD curve moments allows the calculation of the volume

swept by the tracer, the flow geometry, and the Gini coefficient which expresses the flow

heterogeneity in the fracture system (Shook and Forsmann, 2005; Shook and Suzuki,

2017).

The concept of RTD curves was developed by Danckwerts (1953), where the distri-

bution of tracer residence times, E (t), depends on the fraction of the tracer that has a

residence time between time t and t+ dt in the system. This fraction is given by E (t) dt.

Thus, for the tracer concentration, c (t), at time t of the effluents at a monitoring location,

4.3 Methods 107

the RTD curve is calculated as

E (t) =c (t) ρqpro

Minj, (4.1)

where ρ is the effluent density, qpro is the volumetric production flow rate at the monitoring

location, and Minj is the mass of injected tracer at the injection location (Robinson and

Tester, 1984; Shook and Forsmann, 2005). The n-th temporal moment of an RTD curve

is then defined as

m∗n =

∞∫0

tnE (x, t) dt . (4.2)

The zeroth temporal moment yields the tracer recovery, R = m∗0. The first normalized

temporal moment defines the mean residence time,

t∗ =m∗1m∗0

, (4.3)

and the second centralized and normalized temporal moment provides the measure on

tracer dispersion,

m2,c =m∗2m∗0−(m∗1m∗0

)2

. (4.4)

For tracer tests, where tracers are injected as a pulse, the tracer-swept volume is

defined as (Shook and Suzuki, 2017)


where qinj is the volumetric injection flow rate. The flow geometry in a fractured medium

can be characterized by the flow capacity – storage capacity curve (F − Φ curve). The

F−Φ curve is a cumulative contribution of individual flow paths, where the flow capacity,

F , is the specific velocity, divided by the bulk velocity, and the storage capacity, Φ, is the

fraction of the pore volume associated with that flow path (Shook and Forsmann, 2005;

Shook and Suzuki, 2017). Specifically, the F − Φ curve is mathematically expressed as

F (t) =

t∫0

E (τ) dτ

∞∫0

E (t) dt

, (4.6)

and

Φ (t) =

t∫0

E (τ) τdτ

∞∫0

E (t) tdt

. (4.7)


A homogeneous system yields a diagonal line in the F − Φ plot, where F and Φ range

between 0 and 1. For a heterogeneous system, on the other hand, abrupt breaks in the

F −Φ curve slope give insights into the presence of different permeabilities (Shook, 2003;

Shook and Suzuki, 2017). Furthermore, the heterogeneity of a system can be quantified

from the F − Φ curve by determining the Gini coefficient,

G = 2

1∫0

F dΦ− 1

2

. (4.8)

The G value varies between 0 and 1, where a homogeneous system yields 0, and 1 means

that a negligibly small fraction of the total tracer-swept volume provides almost all of the

fluid (Shook and Forsmann, 2005).

4.3.3 Temperature perturbations in a fracture

To estimate the fracture surface area between the injection and monitoring locations, we

apply an analytic solution, given by Shook and Suzuki (2017), for a single fracture with

uniform aperture. This analytic solution utilizes the tracer-swept volume from the conser-

vative tracer tests and fluid temperature measurements to approximate the temperature

distribution in a single fracture. Following the governing equations in Gringarten and

Sauty (1975), which are similar to those published by Lauwerier (1955) and Carlslaw and

Jaeger (1959), the heat transport in a half of a single fracture, with heat flow across the

rock matrix-fracture interface, is given by

b

2(ρCp)F

∂TW∂t

+qpro

2(ρCp)W

∂TW∂S−KR

∂TR∂z|z=b/2= 0, (4.9)

where b is the fracture aperture, Cp is heat capacity, and T is temperature, with subscripts

F for fracture, W for water, and R for rock matrix. S is the surface area of the half

fracture, KR is the thermal conductivity of the rock matrix, and (ρCp)F = φ (ρCp)W +

(1− φ) (ρCp)R. In the current setup, the half fracture is bounded by the center of the

fracture at z = 0 and the rock matrix-fracture interface at z = b/2. The temperature

evolution of the surrounding rock matrix is governed by the heat conduction equation,

∂2TR∂z2

=(ρCp)RKR

∂TR∂t

for z ≥ b

2. (4.10)

The temperatures must also satisfy the following boundary and initial conditions:

TW (S, t) = TR (S, z, t) = TI for t ≤ φbS

qpro, (4.11)

4.3 Methods 109

TW (0, t) = TJ for t > 0, (4.12)

TW (S, t) = TR (S, b/2, t) ∀S, t, (4.13)

limz→∞

TR (S, z, t) = TI ∀S, z, t, (4.14)

where TI and TJ are the initial and the injection temperature, respectively. The analytical

solution of Eqs. (4.9) and (4.10), subject to the conditions given by Eqs. (4.11)-(4.14), is

given by Gringarten and Sauty (1975) as

TI − TW (t)

TI − TJ= erfc

[(ρCp)

2W

KR (ρCp)R

(qpro

S

)2{t−

(ρCp)T(ρCp)W

bS

qpro

}]−1/2

. (4.15)

By taking the tracer-swept volume, Vp = bWLφ = bSφ, and the total fracture surface

area, A = 2S, where W and L are the fracture width and length, respectively, and

multiplying the last term in Eq. (4.15) by φ/φ and simplifying, Shook and Suzuki (2017)

obtained

TW (L, t) = TI − (TI − TJ) erfc

1

(ρCp)W

A

2qpro

√√√√ KR (ρCp)R

t− (ρCp)F(ρCp)W

Vpφqpro

. (4.16)

Eq. (4.16) can be used to estimate the surface area of a fracture, A, through which heat

exchange between the fluid and the rock takes place. Here, A is constrained by the mea-

surable Vp from the solute tracer tests and by temperature observations in the monitoring

borehole. The variables T , qpro, KR, and Cp can be obtained from measurements or from

literature data, while porosity, φ, can be estimated, as its influence on the results is

minor (Shook and Suzuki, 2017).


4.4 Results and discussion

4.4.1 Residence time distributions

Figure 4.3 shows the residence time distributions (RTDs) and the F − Φ curves from

the four monitoring locations, obtained during Tests 8 and 9. The minor changes in

the F − Φ curves between Tests 8 and 9 suggest that the injected heat had a minor

influence on the distribution of fluid within the tracer-swept volumes. The F −Φ curves

from PRP1-int2 display exceptionally negligible deviations from the F −Φ diagonal line,

implying a rather homogeneous flow distribution. In contrast, the largest deviations of

the F −Φ curves from the diagonal line are observed at PRP1-int3, where approximately

81% of the recovered tracer is transported through only 30% of the total tracer-swept

volume. In another tracer test at the same test site and at the same observation interval

(i.e., PRP1-int3) but injecting a tracer at a different interval, namely INJ1-int4 instead

of INJ2-int4 (as reported in this study, Fig. 4.1), Kittilä et al. (2019) reported that 30%

of the tracer-swept volume contributed to 70% of the tracer recovered from PRP1-int3.

Robinson and Tester (1984) observed that the flow impedance of local fracture outlets

can mask dispersive changes within a system by dominating the distribution of fluid

flow. Therefore, the observed similarity in the distribution of flow at PRP1-int3 from two

opposite injection locations suggests that most of the flow distribution may occur near

the fracture outlet at PRP1-int3.

Contrary to the similarity of the F−Φ curves, the RTD curves show distinct differences

between Tests 8 and 9 (Fig. 4.3). During Test 9, the tracer was less likely arriving at the

AU Tunnel and PRP2-int2 than before the hot water injection, as indicated by the overall

lower age distribution, i.e., E values between the tests, recorded at these two monitoring

locations. However, tracer arriving at PRP1-int3 was more likely transported in the

main preferential flow paths than before the hot water injection, resulting in higher peak

E values in the RTD curve. At PRP1-int2, the E values were also higher during Test 9

but accompanied by later tracer arrival times than during Test 8. Furthermore, the RTD

curves from the AU Tunnel and PRP2-int2 are highly similar in shape during both tracer

tests. At PRP1-int3, in contrast, the long tailing during Test 8 is not as prominent in the

RTD curve from Test 9, but the peak E value more than doubled. At PRP1-int2, the peak

arrival time (Table 4.2) increased by more than 200%. Unfortunately, we were not able

to obtain a sufficient record of the tailing before the test was terminated. The x-symbol

in the RTD curve, obtained from PRP1-int2 during Test 9 (Fig. 4.3), marks the start

of a 2-day breakdown of the water injection system, i.e., approximately 4800minutes, or

about 80 hours, after the pulse injection of the solute dye tracer (Fig. 4.2), which likely

had an effect on the tailing of the PRP1-int2 RTD curve by changing its decay rate.


0 1000 2000 3000 4000 50000

2

4

E [1

/min

]

#10-4 AU Tunnel

Test 8Test 9

0 100 200 300 4000

1

2

3

E [1

/min

]

#10-4 PRP1-int3

0 500 1000 1500 20000

2

4

6

8

E [1

/min

]

#10-5 PRP2-int2

0 2000 4000 6000Time [min]

0

2

4

6

E [1

/min

]

#10-6 PRP1-int2

0 0.5 10

0.5

1

Flo

w c

apac

ity, F

AU Tunnel

0 0.5 10

0.5

1

Flo

w c

apac

ity, F

PRP1-int3

0 0.5 10

0.5

1

Flo

w c

apac

ity, F

PRP2-int2

0 0.5 1Storage capacity, )

0

0.5

1

Flo

w c

apac

ity, F

PRP1-int2

Figure 4.3 – Comparison of the residence time distribution (RTD) curves (left) andthe F − Φ curves (right). The RTD and the F − Φ curves are from before (Test 8)and during (Test 9) hot water injection at the four monitoring locations, namely theAU Tunnel outflow point, PRP1-int3, PRP2-int2, and PRP1-int2. The x-symbol on theTest 9 RTD curve, obtained from PRP1-int2, marks the start of a 2-day breakdown ofthe water injection system (Fig. 4.2) and the dashed diagonal lines in the F −Φ plotsrepresent a homogeneous fracture system. Note the different scales of the axes for theRTD curves.


4.4.2 Redistribution of fluid flow

Changes in flow rate and tracer recovery

Table 4.2 shows the tracer transport results from the moment analysis. These results

are calculated from the obtained RTD curves without extrapolation. The recovery, R,

becomes smaller at all of the monitoring locations during Test 9, except at PRP1-int2, in-

dicating that the injection of ∼ 45 °C hot water redistributed the fluid flow in the fracture

network of the S3 shear zone structures, resulting in a substantial overall reduction of total

tracer recovery (Table 4.2). It is worth noting that the tracer recoveries at the AU Tunnel

are calculated with an estimated production flow rate, due to an overflow at the outflow

collection point, using a “tipping-bucket” measuring device. In our previous tracer tests,

at the GTS (Kittilä et al., 2020), an outflow of about 1.0 L min−1 was documented at the

AU Tunnel. Therefore, we estimate a production flow rate of 1.0 L min−1 during Test 8,

and yield a tracer recovery of 65% at the AU Tunnel (Table 4.2). This calculated recovery

is a typical value, observed during previous tracer tests at the GTS (Kittilä et al., 2020).

Consequently we also use 1.0 L min−1 in estimating the tracer-swept volume (Table 4.2).

During the tracer tests, we have noticed that the overflow at the outflow collection

point at the AU Tunnel was smaller during Test 9 than during Test 8. The reduction

of production flow rate at the AU Tunnel during Test 9 is physically rational because a

good hydraulic connection between the injection interval INJ2-int4 and the AU Tunnel

was observed during previous tests at the GTS (Jalali et al., 2018b; Brixel et al., 2020;

Kittilä et al., 2020). Here, we propose two methods to estimate the production flow rate

at the AU Tunnel during Test 9. Method I uses a direct proportion of the injection rate.

i.e., QAUpro,Test 9 = QAUpro,Test 8×Qinj,Test 9/Qinj,Test 8. As previously discussed, we estimate

that Qpro,Test 8=1.0 L min−1. Given that Qinj,Test 8=2.1 L min−1 and Qinj,Test 9=1.8 L

min−1 (Table 4.1), QAUpro,Test 9 is calculated to be 0.85 L min−1. Method II uses the

proportional fluid lost to the far-field. In this method, the fraction of fluid lost to the

far-field is defined in terms of injection vs. production flow rate, i.e., not referring to

recovered tracer mass. We assume that the fraction of fluid lost to the far-field is the

same during Tests 8 and 9. During Test 8, the fraction of fluid lost to the far-field

is calculated as flost = 1 −∑

iQipro,Test 8/Qinj,Test 8, where the summation applies to

all monitoring locations (i.e., i represents the AU Tunnel, PRP1-int3, PRP2-int2, and

PRP1-int2). Taking the injection and the production flow rates in Table 4.1, flost is

calculated to be 0.39. According to the assumption made for Method II, QAUpro,Test9 =

Qinj,Test 9 × (1 − flost) −∑

j Qjpro,Test 8, where j represents the monitoring locations,

excluding the AU Tunnel. Method II yields QAUpro,Test 9 = 0.88 L min−1. Both Methods

(I and II) yield very similar production flow rates at the AU Tunnel during Test 9. The


Table 4.2 – Summary of the tracer transport parameters.

Monitoring location Test 8 Test 9 Changea [%]Recovery, R [%]AU Tunnel ∼65 ∼23 -64PRP1-int3 2.84 2.02 -29PRP2-int2 2.47 1.46 -41PRP1-int2 0.97 2.17 123Peak arrival time, tp [min]AU Tunnel 235 375 60PRP1-int3 30 18 -41PRP2-int2 122 183 50PRP1-int2 950 3020 218Mean residence time, t∗ [min]AU Tunnel 2242 1383 -38PRP1-int3 1232 286 -77PRP2-int2 866 1511 75PRP1-int2 2126 3373 59Swept volume, Vp [m3]AU Tunnel ∼3.05 ∼0.57 -81PRP1-int3 0.074 0.010 -86PRP2-int2 0.045 0.040 -12PRP1-int2 0.043 0.13 202Dispersion, m2,c [min2]AU Tunnel 4.63E+06 1.31E+06 -72PRP1-int3 2.76E+06 1.91E+05 -93PRP2-int2 8.54E+05 1.94E+06 127PRP1-int2 1.24E+06 2.15E+06 73Gini coefficient, G [–]AU Tunnel 0.50 0.44 -13PRP1-int3 0.64 0.67 4PRP2-int2 0.54 0.50 -7PRP1-int2 0.30 0.25 -16aChange from Test 8 to Test 9.

production flow rate of 0.85 L min−1 yields a tracer recovery of 23% and a swept volume

of 0.57m3 at the AU Tunnel (Table 4.2). Even taking the same production flow rate

(i.e., 1.0 L min−1) as estimated during Test 8, we obtain a tracer recovery of 27% and a

swept volume of 0.68m3 at the AU Tunnel. It is thus evident that the decrease in tracer

recovery is significant between Tests 8 and 9, implying that the hot water injection greatly

impaired the hydraulic connection from INJ2-int4 to the AU Tunnel. Furthermore, these

results suggest that a significant portion of the injected tracer was lost to the far-field.

Nelson et al. (1981) showed that the injection of hot water promotes the closure

of fractures, particularly near the injection borehole. Such closures occur particularly

in fractures that initially carried higher fluid flow rates, thereby closing sooner. This

fracture closure upon hot fluid injection into rock fractures is caused by the thermal

expansion of the fluid-heated rock matrix (Rutqvist et al., 2001). The closure of the

fractures in turn causes the fluid to flow through new pathways. The opposite effect,


that is the focusing of fluid flow in cooled zones, resulting from thermal contraction

of the rock matrix when cold fluids are injected into a rock fracture, was observed by

Fu et al. (2016). Consequently, hydraulic connections between injection and monitoring

locations can become weakened either i) by hot fluid injection, when key flow paths are

constricted, or ii) by cold fluid injection, when certain flow paths begin receiving more

fluid, diminishing fluid flow through other flow paths, compromising previously dominant

hydraulic connections.

Visual inspections of the S3 shear zone structures show that the injection interval

INJ2-int4 is located in-between the S3.1 and S3.2 shear zones (Fig. 4.1 and Brixel et al.

(2020)). It is also observed that the AU Tunnel outflow point is located in-between these

two structures (Krietsch et al., 2018). It is thus likely that the solute tracer injected in

the current study during Test 8 first travelled via a linking damage zone, i.e., fractures

linking S3.1 and S3.2, to a wall damage zone, i.e., fractures parallel to and associated with

S3.1 and/or S3.2. We can speculate that, subsequently, the tracers entered the highly

fractured zone between S3.1 and S3.2 again, and then exited at the AU Tunnel outflow

point (Fig. 4.1). This interpretation is further supported by the Time-Lapse Difference

Reflection Ground Penetrating Radar (GPR) surveys conducted at the GTS (Giertzuch

et al., 2019). However, from INJ2-int4 towards the west, i.e., towards the other monitoring

locations, the solute tracer travelled faster through the S3.1 shear zone (RTDs from PRP1-

int3 and PRP2-int2) than through the S3.2 shear zone (RTD from PRP1-int2).

During Test 9, the injection flow rate, Qinj , was lower than during Test 8, whereas the

injection pressure, Pinj , had increased. Note that the injection pressures during Tests 8

and 9 never exceeded the minimum stress at the site, nor the pore pressure required to

initiate rock failure (8.6-9.7MPa for σ3 and 5MPa, respectively (Krietsch et al., 2019)).

Therefore, the fracture system was not supported by the injection pressure. In addition

to the changes in the injection flow rate and fluid pressure, the production flow rates,

Qpro, at PRP1-int3, PRP2-int2, and PRP1-int2 (Table 4.1) were 43%, and 38% less, and

2% more, respectively, than during Test 8. Such changes in the injection and production

flow rates and fluid injection pressures were also observed in a hot water thermal tracer

test in another fracture zone, located only 300m north from our test volume, at the

GTS (Marschall et al., 1995).

In terms of negative or positive changes, regarding production flow rates and tracer

recoveries during Test 8 and Test 9, the two parameters behave similarly (Table 4.2). In

terms of quantity, however, the changes between these two parameters do not match. This

may be due to different contributions of fluid flow from the far-field into the monitoring

locations, due to the redistribution of flow (and changes in the fluid pressure gradient) in

the fracture network.


Tracer-swept volumes and residence times

Similar to tracer recoveries, the tracer-swept volumes, Vp, decreased at all monitoring

locations, except at PRP1-int2 (Table 4.2). Although the full RTD curve could not be

recorded at PRP1-int2 during Test 9, this interval indicated a strong increase in tracer-

swept volume from 0.043m3 to 0.13m3. The decrease in tracer-swept volume at the other

three monitoring locations likely reflects the diminished quantities of the traced water,

travelling to the monitoring locations. It is important to note that the tracer-swept volume

yields an estimate of the total volume of all fractures that contribute to fluid flow and

that produce fluid at a monitoring location, regardless of the fractures’ flow impedances.

Typically, the majority of the injected fluid flows in low-impedance fractures, yielding an

F −Φ curve with a deviation from the F −Φ diagonal line. However, extremely long fluid

residence times usually indicate that low permeabilities and/or long-distance flow paths

are present.

As discussed earlier, thermally driven fracture closure can significantly affect the dis-

tribution of fluid flow, as it diminishes fluid flow in the heated zones. The mean residence

times, t∗, were shorter during Test 9 at the AU Tunnel and PRP1-int3, and longer at

PRP2-int2 and PRP1-int2, than during Test 8. Furthermore, the change in the mean res-

idence time appear to be directly proportional to the change in dispersion, m2,c, during

the tracer tests (Table 4.2). The dispersion values describe the spread of tracer residence

times. If m2,c yielded zero, tracer introduced to the system by a Dirac pulse, and ob-

served at a monitoring location, would have experienced zero dilution. However, natural

systems invariably yield nonzero m2,c, as the tracer is always transported with at least

some variation in the fluid flow velocity field in the systems (Cirpka and Kitanidis, 2000).

These results suggest that, at the AU Tunnel and at PRP1-int3, not only was the bulk

of the tracer transported faster during Test 9, but also that the occurrence of the tracer

mass recovered from the flow paths with long residence times decreased. Consequently,

the opposite describes the RTD curves obtained from PRP2-int2 and PRP1-int2. That

is, the slower-travelling tracers were more likely diffusing into stagnant zones.

Gini coefficients

The Gini coefficient, G, indicates how channelized the fluid flow is. Only the RTD curve,

obtained from PRP1-int3, yielded a increased Gini coefficient after the start of hot water

injection. This likely reflects the domination of the preferential flow paths over the total

tracer transport towards PRP1-int3 during Test 9, which is exhibited as a higher peak

of E values in the RTD curve (Fig. 4.3). In contrast, the changes in values of dispersion

indicator (m2,c) and Gini coefficients at the AU Tunnel indicate that both the spread of

the residence times had decreased and the fluid flow was less channelized during Test 9


Table 4.3 – Summary of thermal and petrophysical properties used.

Model parametersρR 2706aCp,R 768bKR 3.25bρW (13 °C) 999Cp,W (13 °C) 4180TI 13cTJ 45cφ 0.8t (days) 15TW (AU Tunnel) N/Ac

TW (PRP1-int3) 20.2cTW (PRP2-int2) 13.5cTW (PRP1-int2) 13.0caWenning et al. (2018), bKant et al. (2017),cBrixel et al. (2019).

than during Test 8. However, despite the decrease, or only small increase, in all of the

Gini coefficient values during Test 9, the fracture network, associated with the S3 shear

zone, still exhibited a rather heterogeneous distribution of fluid flow (i.e., Gini coefficients

ranging from 0.44 to 0.67).

4.4.3 Estimation of fracture surface area

The initial fluid temperature at the GTS was approximately 13 °C (Table 4.3). Combined

data sets of solute dye tracer RTD curves and temperature measurements were recorded

at the AU Tunnel outflow point, PRP1-int3, PRP2-int2, and PRP1-int2. The tempera-

ture data used in this paper, recorded at these monitoring locations, were obtained with

PT1000 temperature sensors, installed in the open intervals (Doetsch et al., 2018a). At

the AU Tunnel outflow point, the temperature measurements were strongly influenced

by the direct contact of the outflow, at the tunnel wall, with the atmosphere (Brixel

et al., 2019), and are thus not analyzed in this study. During Test 9, the temperature

at PRP1-int3 was fluctuating, with a harmonic mean of 20.2 °C. At PRP2-int2, the tem-

perature was still steadily increasing, by approximately 0.05 °C per day, with a harmonic

mean during Test 9 of 13.5 °C. At PRP1-int2, no increase in temperature was observed

(Table 4.3 and Fig. 4.4).

When injecting water that is cooler or hotter than the in-situ fracture temperature,

the induced temperature perturbations at a monitoring location at time t are related to

the surface area of the fractures contributing to fluid flow (Kolditz, 1995; Finsterle et al.,

2013; Guo et al., 2016). Thus, to provide constraints on fracture geometry, Eq. (4.16)

is solved for the three monitoring locations during Test 9 (Table 4.4), using a fracture

porosity of φ = 0.80 (Marschall et al., 1995) (Table 4.3). However, it is worth noting


22.11. 28.11. 4.12. 10.12. 16.12. 22.12.2017Date

12

14

16

18

20

22

Tem

pera

ture

[°C

] PRP1-int3PRP2-int2PRP1-int2

Figure 4.4 – Temperature at monitoring locations PRP1-int3, PRP2-int2, and PRP1-int2 during Tests 8 and 9 (Brixel et al., 2019). Temperatures between Tests 8 and 9are not shown in this figure due to data copyright.

that only at PRP1-int3 the temperature perturbation signal is significant, in addition to

apparently having reached steady-state. As the temperature perturbations in Table 4.3

and Fig. 4.4 show, heat takes a long time to break through when the scale of a system

is increased. For typical reservoir scales of hundreds of meters, such long breakthrough

times would often be unfeasible to estimate the fracture surface area from the comparison

of heat and conservative tracer breakthrough data. Consequently, thermally degrading or

reactive tracers have been suggested to be compared with conservative and non-reactive

tracers to estimate fracture surface areas (Hawkins et al., 2017b, 2018).

As pointed out by Shook and Suzuki (2017), Eq. (4.16), i.e., the estimation of the

fracture surface area, A, is not particularly sensitive to the estimation of φ. In fact,

the aperture of a single fracture does not influence the estimate of its surface area. In

contrast, for highly fractured reservoirs, the fracture porosity (taken as 2b/L, where b

and L are the fracture aperture and spacing, respectively) strongly influences surface

area. However, the breakthrough time of a conservative tracer is strongly influenced

by the reservoir volume, i.e., the mean fracture aperture, whereas the heat exchange in

fractured reservoirs is dominated by the surface area available for heat exchange (Hawkins

et al., 2017a,b). Therefore, there is an uncertainty in estimating thermal breakthrough

times and fracture surface areas using conservative tracer tests. In summary, the non-

unique relationship between solute and heat transport limits the accuracy of determining

fracture surface areas, employing the simple model proposed in this study.

Nonetheless, we set t at 15 days in Eq. 4.16, i.e., 15 days after the start of hot water

injection, to estimate the fracture surface area (Table 4.4), A, with the temperature

values measured at the monitoring locations (Table 4.3). At PRP1-int3, a surface area of

2.65m2 is calculated, which enabled the highest temperature change of ∆T ≈7 °C to be


Table 4.4 – Results of solving fracture surface area using Eq. (4.16).

Monitoring Test 9location A [m2] b [mm]PRP1-int3 2.65 9.4PRP2-int2 4.40 22.7PRP1-int2 >19.0 <17.1

measured during hot water injection (Fig. 4.4 and Doetsch et al. (2018a)). At PRP2-int2,

where only a slight and not yet stabilized temperature perturbation was observed, the

production flow rate was lower than at PRP1-int3, and the tracer-swept volume was four

times higher. In comparison to PRP1-int3, PRP2-int2 yielded a larger fracture surface

area of A = 4.40 m2. As no temperature perturbation was observed at PRP1-int2 during

Test 9, the fracture surface area is estimated to be 19.0m2 (when ∆T=0 °C) or larger

(Table 4.4).

The fracture aperture, b, values, shown in the last column of Table 4.4, and calculated

using the tracer-swept volume, are directly scaled by porosity, φ, which is estimated in

this study. Robinson and Tester (1984) point out that placing the flow, provided by the

tracer-swept volume, Vp, within a bulk volume of rock requires assuming a porosity, which

is difficult to determine. They explain that with a porosity of 1-10%, fluid flow is localized

to a small rock volume, such as that formed by interconnected flow paths between main

fractures. In contrast, porosities below 1% imply that a larger fraction of the fluid is

sweeping through the rock beyond the main fractures. Furthermore, the aperture, b,

values are non-unique and are thus only meant to illustrate the potential magnitude of b.

Through a numerical model of fluid flow through a nearby fracture system at the GTS,

Marschall et al. (1995) found that the best fit for solute breakthrough curves was obtained

when fracture apertures were estimated to be approximately 10-80mm. Thus, our results

for b are on the same order of magnitude with the results obtained by Marschall et al.

(1995).

Robinson and Tester (1984), Luhmann et al. (2015), and de La Bernardie et al. (2018)

point out that heat transfer in fractured media cannot be modeled precisely with a single

structure, having constant hydraulic and transport properties. As our results are based

on a highly simplified model, they should only serve as scoping calculations. As Fu et al.

(2016) and Guo et al. (2016) showed, complex structures can cause thermal perturbations

in a rock and a monitoring borehole that deviate from simple models, for example due to

flow channeling or as a result of thermally induced fracture aperture variations (Pandey

et al., 2017). As the examination of the RTD curves suggests (Fig. 4.3 and Table 4.2),

it is reasonable to assume that both processes (flow channeling and thermo-mechanical

effects) likely played a role during the hot water injection experiment at the GTS. We be-


lieve that the main cause contributing to the redistribution of fluid flow, observed in this

study, is most likely the thermo-mechanical expansion of the rock and not, for example,

the result of geochemical dissolution of asperities (Grimm Lima et al., 2019), as chem-

ical processes typically involve longer timescales (Yasuhara et al., 2011). Despite these

complications when estimating heat transfer, employing simple models of solute tracer

and temperature data, our results show that this approach still helps placing constraints

on fracture geometry. This in turn can improve predictions of the geothermal energy

transport performance of a formation.

4.4.4 Data uncertainties

In addition to uncertainties regarding the simplification of the thermal perturbation

model, it is worth noting that there are other potential sources of uncertainty, including:

• Sorption processes and decay of the tracer eosine may have affected the recovered

RTD curves. Irreversible sorption would reduce the recovered tracer mass, while re-

versible sorption would increase the fraction of tracer mass exhibiting long residence

times, resulting in long tailing.

• The Test 9 fluorometer data were converted to concentrations (ppb) using the cali-

bration curves from Test 8 and, although there were only two weeks between these

tests, it remains unknown how well the calibration curve from Test 8 matches the

concentrations during Test 9.

• It is possible that during Test 9, the flow field in the rock volume was still changing,

due to the thermal perturbations, i.e., the system was not at steady-state. Con-

sequently, the constant production flow rates, used in Eq. (4.16) (a mean of the

flow rates of Tests 8 and 9), may not have been properly estimated. Also, other

fractures than the one(s) producing the injected tracer may have contributed to the

total production flow rates. Furthermore, since the start of the hot water injection,

the tracer-swept volumes changed. Using the value obtained during Test 9 may

thus have underestimated the fracture surface area through which the fluid and

rock exchanged heat.

• It may be possible that PRP2-int2 acts as a short-circuiting pathway between the

S3.1 and the S3.2 shear zone structures (Fig. 4.1), because it is intersected by

fractures associated with both of these structures (Krietsch et al., 2018; Brixel et al.,

2020). The use of tracers in subsurface reservoirs requires installing boreholes, which

are likely to change the hydraulic conditions in the subsurface.


4.5 Conclusions

We investigated the effects of hot water injection on the hydrodynamic properties of a

hydraulically stimulated crystalline rock, using solute dye tracers. The hydraulic stim-

ulations and the tracer tests were part of scaled Enhanced geothermal System (EGS)

analogue experiments, which were conducted to gain insights into the processes under-

pinning permeability enhancement and creation of an efficient subsurface heat exchanger,

relevant for the development of actual EGS reservoirs. In addition to comparing the

temporal moments and the associated tracer-swept volumes and flow geometries, we put

constraints on fracture geometries by estimating the fracture surface areas contributing to

heat exchange between the host rock and the fluid flowing in the fractures. Importantly,

we observe redistribution of fluid flow and loss of injected fluid to the far-field after two

weeks of hot water injection, while the key flow paths between the injection and monitor-

ing locations did not change. The obtained results can be interpreted as a manifestation

of the thermo-mechanical response of the fractured rock to the hot water injection.

The moment analysis results of the effects of hot water injection on fluid flow properties

can also facilitate numerically quantifying the thermo-mechanical behavior of the test

rock volume and the subsequent changes in permeability of the fracture network. We also

showed a case of combining solute tracer concentration measurements with temperature

perturbations to delineate fracture geometries, based on a simple single parallel plate

model. Despite the simplicity of our model, further numerical studies, attempting to

model the thermo-hydro-mechanical behavior of the study site, will likely benefit from

the scoping calculations of the fracture geometries presented here. Finally, by conducting

field-scale tracer experiments in hydraulically stimulated fractured crystalline rock, we

provided insights into the evolution of fluid flow distribution and hydraulic connections

during fluid circulation. Such processes may be critical during the operation of an actual

EGS, as it is necessary to understand the induced changes in fluid flow geometry and

rate, due to the injection of a fluid with a temperature that is different from the in-situ

rock temperature. With such understanding, one can better estimate the lifetime of a

geothermal formation from which geothermal energy is being produced.

Acknowledgements

The ISC is a project of the Deep Underground Laboratory at ETH Zurich, established

by the Swiss Competence Center for Energy Research - Supply of Electricity (SCCER-

SoE) with the support of the Swiss Commission for Technology and Innovation (CTI).

The Grimsel Test Site is operated by Nagra, the National Cooperative for the Disposal of

Radioactive Waste. We are indebted to Nagra for hosting the ISC experiment in their GTS

facility and to the Nagra technical staff for onsite support. The authors are grateful for

the invaluable technical contributions of N. Knornschild in the GEG group at ETH Zurich

and for the support of F. Leuenberger with the solute dye tracers. The Werner Siemens

Foundation (Werner Siemens-Stiftung) is further thanked by M.O. Saar for its support of

the Geothermal Energy and Geofluids (GEG.ethz.ch) Group at ETH Zurich. We thank

the two anonymous reviewers for their suggestions and comments that greatly improved

an earlier version of this manuscript and the editor, Thomas Kohl, for his handling of the

manuscript.

121

5Summary and perspectives

The focus of this thesis is to quantify the effects of hydraulic stimulation and hot water

injection on the fluid flow and solute mass transport characteristics of a fractured crys-

talline rock at the decameter scale, by performing solute tracer tests before and after the

stimulation and the start of the hot water injection.

Tracer tests are controlled field experiments that improve our understanding of fluid

flow as well as mass and heat transport through porous and/or fractured media. In

subsurface reservoir studies, tracer tests are particularly powerful in assessing inter-well

hydraulic properties that are otherwise beyond the scope of other hydraulic, geophysical,

and geological characterization methods. Tracer tests can yield information on preferen-

tial flow paths, fluid flow channeling, and tracer-accessed reservoir volume. Furthermore,

applied tracers can serve as harmless surrogates of contaminants or as colloid particles

mediating contaminant (e.g., pathogens, heavy metals, radionuclides) transport. The

latter is particularly important in contaminant migration and site vulnerability studies.

In Chapter 2, I evaluated the transport properties of novel colloidal DNA-labeled silica

nanoparticles in fractured crystalline rock. With an approximate diameter of 166 nm, the

DNA nanotracers tend to stay in the center of flow channels, thus exhibiting higher

average velocities than solute tracers. This behavior is due to the size exclusion of the

colloidal DNA nanoparticles, whereby the particles are excluded from certain flow paths

due to their finite size. In addition to this, and potentially further contributed by density

123

124 5 Summary and perspectives

effects, the DNA nanotracers yielded lower mass recoveries and swept volumes, and less

dispersion with reference to solute dye tracers. Moreover, I observed that the recovery and

the response curves of the DNA nanotracers are strongly influenced by the size exclusion,

and by complex flow fields, induced by spatial heterogeneities in the fracture network.

A specific application of these particulate tracers is thus the mimicking of particulate-

bound contaminant transport. This study presented, to the best of my knowledge, the first

field campaigns of the novel DNA nanotracer in fractured crystalline rock. Our previous

studies demonstrated i) their utility to image subsurface reservoirs by travel-time-based

tomography in an unconsolidated aquifer (Kong et al., 2018), and ii) the impact of the

DNA nanotracer particle size and surface charge on their transport in a sand column

(Mikutis et al., 2018). These experiments were designed to validate the use of the DNA

nanoparticles in hydrogeological applications and in tracer tomography.

The assessment of geothermal reservoir performance is crucial for the reservoir lifetime

and economic evaluation of the reservoir. For enhanced geothermal systems (EGSs), a

fundamental uncertainty, regarding the future performance of the EGS, is the ability to

create and maintain an efficient heat exchanger via rock mass stimulation (Evans, 2015).

In Chapter 3, the use of conventional solute dye tracers allowed me to investigate the

stimulation-induced changes in the hydraulic properties of a scaled EGS analogue rock

mass at the Grimsel Test Site (GTS) in Switzerland. I accomplished this by using moment

analysis and tomographic inversion approaches on the recovered tracer response curves.

The decameter scale tracer experiments were conducted before and after the rock mass

was stimulated via hydraulic shearing and hydraulic fracturing. The tracer experiments

yielded larger tracer-swept volumes after the stimulation programs. The increment in

swept volumes can be attributed to newly created flow path connections. However, the

analysis of the Gini coefficients, a parameter that quantifies flow path channeling, showed

that a larger fraction of the fluid flow was supported by fewer key flow paths after the

stimulation programs. I also identified the occurrence of multiple peaks in several tracer

response curves, which are indicative of multiple separate preferential flow channels. The

creation of new hydraulic connections, as a result of the hydraulic shearing stimulation,

was also observed in the hydraulic conductivity tomograms, which indicated that fluid

flow was accessing pathways with higher hydraulic conductivity values after the hydraulic

shearing stimulation. My results highlighted the importance of understanding the evolu-

tion of preferential flow paths in a stimulated rock mass. The improved knowledge of the

distribution of flow within an EGS can be used to improve the configuration of injection

and production wells, and thus to improve the reservoir’s performance.

Finally, in Chapter 4, I examined the effects of hot water injection on the hydrody-

namic characteristics of the stimulated rock mass at the GTS. The estimated flow geome-

125

tries and the associated Gini coefficients suggested little change took place in the distri-

bution of flow within the tracer-swept volumes between each of the injection-monitoring

well pairs. However, with a significant decrease in the total recovered tracer mass two

weeks after the start of hot water injection, it is clear that the fluid flow was redistributed

within the fracture network. As a result, the majority of the injected water was lost to

the far-field. A potential cause for the redistribution of flow may have been the thermo-

mechanical response of the rock mass to the heat buildup. An intuitive process for such

a heat buildup-induced redistribution of flow is the following: at first, the fractures with

higher flow rates are likely to warm up faster, whereby the surrounding rock expands,

reducing the fracture’s permeability. Consequently, the injected fluid will seek new flow

paths. The heat buildup-induced constriction of fractures thus results in a higher flow

impedance of the system, which may have resulted in the observed increase in fluid in-

jection pressure and a decrease in injection flow rate during the course of the hot water

injection experiment. Furthermore, constraining the fracture geometries, specifically the

fracture surface area, employing a simple parallel-plate model yielded fracture apertures

in the range of 10-30mm. The magnitude of our results is in good agreement with the

results of Marschall et al. (1995).

Changes in the tracer-swept volume and in the distribution of flow in an EGS reservoir

may potentially have detrimental effects on the economic viability of the EGS. Let us,

however, take a closer look at the implications of different scenarios of tracer-swept volume

and fluid flow redistributions on EGS performance:

• As shown by Robinson and Tester (1984), there is a correlation between the effective

heat transfer area and reservoir volume, which directly implies that a decrease in

tracer-swept volume results in an undesired decrease in how well the heat in the

rock volume is extracted by the circulating fluid.

• The number of accessible flow paths can increase as a result of changes in hydraulic

connectivity. If the new flow paths are located in-between the main fractures, heat

can be extracted more efficiently but from a limited volume of rock.

• If the new flow paths reach regions far beyond the main fractures, heat can be ex-

tracted from a larger extent of the reservoir, and reservoir lifetime may be improved.

The results presented in this thesis are promising with regards to the understanding

and advancement of tracer-based characterizations of stimulated, fractured rock, and in

pinpointing the stimulation-induced or operational changes in the EGS. The next logical

step may be to increase the scale of the experiments. With an inauguration in May 2019,

ETH’s Bedretto Underground Laboratory for Geoenergies (BULG) is a hectometer-scale

follow-up underground experiment of the Grimsel DUG-Lab experiments (Hertrich et al.,

126 5 Summary and perspectives

2019). The larger scale enables more realistic experiments to investigate processes related

to EGS reservoir creation and characterization.

At the Grimsel DUG-Lab, the vicinity of the tunnels had a strong influence on fluid

flow (Chapter 3 and Krietsch et al., 2018). Particularly the outflow at the AU Tunnel

dominated the hydraulic experiments, potentially masking some of the effects induced by

the stimulation experiments or by the hot water injection. In addition, due to the small

scale and the heterogeneous distribution of permeability, minor changes in, for example,

the positioning of the packer systems may have induced errors in the repeated experi-

ments. Therefore, thanks to the larger scale at the BULG, these sources of errors can

potentially be reduced. The Grimsel DUG-Lab and the BULG provide unique opportu-

nities to test EGS-related technologies and to fill the knowledge gap regarding reservoir

creation at intermediate spatial scales, with the final goal of establishing a full-scale EGS

(Evans, 2015).

Interpretation of tracer tests conducted at full-scale EGS sites provides valuable infor-

mation of the reservoir’s hydrodynamic characteristics. Importantly, fluid loss fraction,

residence time, and tracer-swept volume can be derived. An interesting follow-up, regard-

ing larger-scale experiments and the associated tracer test analyses presented here, is the

comparison of tracer-swept volume and the volume of stimulated rock, where the latter

may be based on acoustic emission (AE) techniques. Grant (2016) presented tracer-swept

volume as one of three physical parameters to describe how well the stimulated rock is

exploited by a circulating fluid. Grant (2016) pointed out that the tracer-swept volume

is typically 0.001% of the AE volume, but further studies are needed to evaluate this re-

lationship in more detail. Furthermore, not only comparing the respective volumes from

tracer tests and acoustic emissions, I also recommend constraining the spatial extent of

the tracer-swept volume by geophysical imaging methods (e.g., using acoustic emissions)

and tomographic approaches. This joint consideration can enable a more comprehensive

utilization of the methods used in reservoir characterization, giving insights into reservoir

performance and improving predictions of reservoir lifetime.

AAppendix

A.1 Contribution to the risk report of the ISC experiment at

the GTS

On April 2016, a risk report, prepared by the SCCER-SoE team, regarding the impact

of the ISC experiment at the Grimsel Test Site was given to NAGRA (the National

Cooperative for the Disposal of Radioactive Waste). The title of this report was ‘Impact

of the ISC experiment at the Grimsel Test Site - Seismic hazard and disturbances to

nearby experiments and KWO infrastructure´, and the contribution that I provided for

this risk report, concerning the effect of tracer injection, is given below.

A.1.1 Effects of tracer injection

Pre-stimulation tracer tests will be conducted in May 2016 (2.-4.5.2016) between bore-

holes INJ15.001 (INJ1) and INJ15.002 (INJ2) (Figure A.1) in packered-off intervals. The

intervals of interest (Figure A.2) have been chosen based on preliminary investigations of

cross-hole pressure tests and electrical conductivity fluid logging. INJ2 is the injection

well, where tracers will be added, and INJ1 is the production/sampling well for this pur-

pose. Samples will also be collected at the intersection point of the conductive shear zone

in the AU tunnel.

127

128 A Appendix

Figure A.1 – Boreholes at the GTS.

Figure A.2 – Injection and production intervals in the boreholes. The estimated flowrates in INJ2 have been calculated based on previous pulse tests.

A.1 Contribution to the risk report of the ISC experiment at the GTS 129

After an introduction to the planned tracer tests and a short review of previous tracer

tests nearby, this risk analysis concerns the effects of pressure and tracer propagation.

Finally, a summary of this study and recommendations are provided.

Tracer tests

The main tracer to be used consists of silica particles encapsulating short fragments

of DNA. These novel so-called DNA nanotracers are environmentally friendly particles

whose surface properties resemble any natural silica particle, such as sand grains, and

their diameter is approximately 150 nm. The zeta potential of these particles is between

-30 and -45mV, and the DNA comprises 0.1% weight of the particles. The benefit of

using DNA in tracers is that it enables the development of a virtually unlimited number

of distinct tracers. As a consequence, the limitation of noise from tracer background

levels can be excluded. In the tracer experiments in the DUG-Lab, amounts up to 200mg

of the DNA nanotracer are used in each of the injection intervals. Additionally, along

the DNA nanotracers, other well-established solute dye tracers are used. The dye tracers

serve as a reference for the DNA nanotracers due to the novelty of the DNA nanotracers.

This combination does not only strengthen the study evaluating the usage of the DNA

nanotracers in a hydrogeological setting, but also provides further information on the

transport properties of the fractured rock system, because it is expected that the dyes, as a

solute, are transported slightly differently in comparison to the suspended particles of the

DNA nanotracer. As flow in fractures takes place in complex structures of channels with

variable size and length, the use of tracers with different transportation properties allows

more detailed investigation of the flow paths within the studied rock mass. The dyes used

in this study are Uranine and Sulforhodamine B, for which the injection concentrations of

10 ppm are planned. By using two different dye tracers, each can be injected in separate

intervals together with the DNA nanotracers without causing interferences between the

two tracer tests.

The goals of the tracer tests are to obtain a benchmark case for the novel DNA

nanoparticle tracers, and to estimate their properties as tracers in hydrogeological appli-

cations. Another important aspect of the tracer tests is to characterize the groundwater

flow properties and flow pathways in the fracture-dominated system as pre-stimulation

conditions. The tracer tests are expected to last for approximately a half day, although

prior to the injection of the tracers, we will inject water at a surface overpressure of

≤ 6 bar for several hours to reach steady-state conditions of equal water inflow to outflow

rates.

130 A Appendix

Short review of previous tracer tests nearby

Uranine has often been used in the Grimsel test site in hydrogeological investigations dur-

ing the last three decades. For example, in the Migration Experiment (MI) site according

to Frick et al. (1992), and Hadermann and Heer (1996), it has been injected during several

tracer tests between 1985 and the early 1990s. Pfingsten and Soler (2003) have summa-

rized the use of different tracers in the HPF shear zone between 1999 and 2000, where

Uranine has also been frequently used, as well as other fluorescent dyes. In the CRR

experiment site (former MI site) the breakthrough curves of Uranine, bentonite colloids

(50-500 nm), and different radionuclides were measured in the context of studying the

transport of colloids and radionuclides to understand the long-term behavior of radioac-

tive waste repositories (Möri et al., 2003; Kosakowski, 2004). Additionally, Marschall and

Lunati (2006) report the use of three non-sorbing solute tracers (Uranine, Napthionate

and Sulphurhodamine), fluorescent latex microspheres with 1µm diameter, nanospheres

of 1 nm diameter, and biocolloids with sizes between 20 and 350 nm in the GAM project.

The use of different fluorescent tracers was chosen in order to avoid residual concentrations

“contaminating” the study volume for subsequent tracer tests.

Modelling results of pressure propagation

The modeling of pressure and tracer propagation during the tracer tests was done with

PetraSim, a pre- and post-processor for the TOUGH2 simulator. The 2D model was

structured as shown in the first panel of Figure A.3, with a constant water injection rate

of 0.5 l min−1. The model assumes a shear-zone permeability of 1.00×10−13m2 and a

rock matrix permeability of 1.66×10−18m2. Additionally, the porosity of the shear-zone

was set to 0.05 and of the matrix to 0.01. The initial system pressure condition in the

model was set to 1 bar (105Pa), with fixed pressure head in the tunnels.

The pressure propagation is seen in Figure A.3 after 10minutes (a), 30minutes (b),

1 hour (c), 6 hours (d), and 12 hours (e) from the start of injection. The pressure increase

in the system is highest during the first hours. At the beginning, the pressure propaga-

tion clearly follows the structures of the shear zones, but later the pressure perturbation

also diffuses into the (relatively low-permeability) rock matrix. This occurs because the

hydraulic diffusivity of pressure is larger in the higher-permeability fractures than in the

rock matrix and, thus, the pressure perturbations travel farther within the fractures dur-

ing a given amount of time than within the matrix. After approximately 6 hours, the

pressure stabilizes and although the pressure increase is relatively high in the immedi-

ate vicinity of the injection point, farther within the structures, the close-to steady-state

pressure is elevated by only approximately 1.1 to 1.7 bar (i.e., 1.1×105 to 1.7×105 Pa,Figure A.4). The tunnels play a major role in confining the pressure perturbations, as


seen in Figure A.3 (particularly d and e). This has also been noticed during hydraulic

tests, where perturbations in boreholes, intersecting the shear zones, cause an increase in

water seeping to the tunnels through the shear zone.

Numerical modelling results of tracer propagation

Injection of a tracer solution with a concentration of 1 ppm was modeled. The tracer

injection took place once steady-state, with a constant pure-water injection rate of 0.5 l

min−1 was reached, i.e., after 6 hours from the start of the constant water injection. The

tracer was injected as a pulse over 60 seconds with a total volume of 0.5 liters. The tracer

was modeled as being perfectly conservative so that only the injected water, representing

the tracer, was “tagged” during the simulation.

The simulation result depicts the tracer transport as a mass fraction of the tracer

solution from the initial water, in this case, the water injected at constant rate. It does not

distinguish between different types of tracers, but considering the use of an ideal tracer,

i.e., salt or dye, the transport properties do not differ from those of tracer-free groundwater

in the formation. The results after 1, 4, and 18 hours of tracer injection are shown in

Figure A.5, which suggests that the tracer would not be transported over long distances

over short periods of time. However, due to the simple fracture system structure created

for the model, it might underestimate the water velocity and thus also the tracer transport

distance. The direct path east from the injection point, seen in Figure A.5c, for example,

has a peak concentration of approximately 28.5 ppb (concentration of tracer tagged water

in the initial water). Based on the tracer transport simulation and observation of the

influence of the tunnels on perturbations in the system, the risk of the injected tracer

influencing the water chemistry outside the investigated shear zone, travelling through

preferential flow paths, appears very low. On the other hand, the probability of the tracer

seeping into the AU tunnel, where the shear zone intersects the tunnel, and subsequently

being spread around in shoes or clothing by those using the tunnels, is likely higher.

However, even in this case, the tracer concentrations would be diluted and, thus, their

influence on water chemistry further reduced. Collection of the water seeping to the AU

tunnel during the peak of the tracer arrival with separate pumps to a tank would reduce

the risk of these higher tracer concentrations being spread in the tunnels.

Summary and recommendations

Section A.1 considers the risk of fluid pressure and tracer propagation during the tracer

injection experiment. Fluid pressure perturbations are low and do not reach far and

thus likely do not require to be mitigated. The risk of influencing the water chemistry

of surrounding experiments during tracer tests using DNA nanoparticle tracers and dye

132 A Appendix

Figure A.3 – In the first panel, top left, the layout of the tunnels (blue), the shearzones (red lines), and the modelled water injection point (green dot) at the GTS areshown. Pressure propagation with 0.5 l min−1 injection rate after a) 10minutes, b)30minutes, c) 1 hour, d) 6 hours, and e) 12 hours from the start of injection.


Figure A.4 – Pressure changes from chosen cells near the intersection of the shearzones and the tunnels.

134 A Appendix

Figure A.5 – Tracer transportation 1 h (a), 4 h (b), and 18 h (c) after the injection ofthe tracer, with simultaneous constant 0.5 l min−1 injection of tracer free water. Thetracer concentrations are shown as mass fraction from the initial water.


tracers (Uranine, Sulforhodamine B) is negligible to low. Only at the intersection of

the S3 faults with the AU tunnel, tracers may seep into the tunnel where they could

be spread by people, although at very small quantities. A mitigation strategy could be

here to collect the seeping water with the tracer over the expected 1-3 hours of elevated

tracer concentrations during tracer breakthrough times (there would have to be some

elevated tracer cut-off concentration as (low but non-zero) concentration tails will persist

for several more hours. The, environmentally harmless, water (with tracer) collected in

the AU tunnel (about 100 to 200 liters) would be removed from the GTS site. If desired,

a similar strategy could be followed at the borehole (INJ1) sample collection location.

Injected salt tracer, although not used in this particular tracer experiment, could be

expected to propagate in a very similar manner to the tracer propagation results presented

here (see also Figure A.5). Analogous to dyes, salt is defined as a (quasi-)ideal tracer,

i.e., the salt travels at the velocity of the injected water because of its high solubility, and

it is nonreactive with the solid material.

136 A Appendix

Figure A.6 – An illustration of the shear zones S3 (green) and S1 (red) (Krietschet al., 2018), intersected by the boreholes INJ1, INJ2, PRP1, and PRP2, and the AUTunnel monitoring location (red circle).

A.2 Supporting Information for Chapter 2

This supporting information provides an illustration of the S1 and S3 shear zone struc-

tures at the study site (Figure A.6), additional detail regarding the DNA nanotracer size

distributions (Figure A.7) and their qPCR dilution curves (Figure A.8), and comparisons

of the obtained moment analysis parameters as a function of distance from the injection to

the monitoring locations (Figure A.9) and as a function of tracer recovery (Figure A.10).

We also list the fitted functions of the exponential extrapolation (Table A.1), provide

characteristics of the DNA nanotracer particles (Table A.2), and give the complete DNA

sequences and primers of the seven DNA nanotracers (Table A.3).

A.2 Supporting Information for Chapter 2 137

50 100 150 200 250 300 350 400 450Particle size [nm]

0

10

20

30

40

50

60

70

80

90

100

Cum

ulat

ive

volu

me

perc

enta

ge

PT-2DAP-3JS-1AM-1GM-2GR-3GR-1

0 100 200 300 400Particle size [nm]

0

0.5

1

1.5

2

2.5

Vol

ume

perc

enta

ge

Figure A.7 – Particle size distributions of the DNA nanotracers.

0 1 2 3 4 5 6 7 8Log Quantity

0

5

10

15

20

25

30

35

40

45

qPC

R th

resh

old

cycl

e (C

q)

PT-2Fit for PT-2 (-3.526 31.675 0.9976)DAP-3Fit for DAP-3 (-3.322 28.417 0.9985)JS-1Fit for JS-1 (-3.629 31.335 0.9993)AM-1Fit for AM-1 (-3.177 26.987 0.9994)GM-2Fit for GM-2 (-3.397 31.148 0.9961)GR-3Fit for GR-3 (-4.587 39.845 0.9903)GR-1Fit for GR-1 (-3.733 30.430 0.9968)

Figure A.8 – qPCR dilution curves, where values in parenthesis of the linear fittings aregiven as (slope Y-intercept R2). Since both the standard dilution curve and the samplesare prepared in the same water, using the same qPCR reagents, and the standard tracerdilution in all cases is linear (R2 ≥0.99), the tracer quantification is expected to beaccurate.

138 A Appendix

0 10 20 30Distance [m]

0

20

40

60

R [%

]


0

1000

2000

3000

t* [min

]


0

0.5

1

Vp [m

3 ]


0.3

0.4

0.5

0.6

G [-

]


0

2

4

6

m2,

c [m

in2 ]

106

AU Tunnel

INJ2-int4

PRP1-int3

PRP2-int2

AU Tunnel

Figure A.9 – Comparison of the tracer recovery, R, the mean residence time, t∗, theswept volume, Vp, the Gini coefficient, G, and the second normalized and centralizedtemporal moment, m2,c, as a function of distance from different locations (indicatedby colors, see legend) for the different tracers, shown with squares (dye tracer) anddiamonds (DNA nanotracer). Results from Test 1A are indicated by filled symbols andfrom Test 4A by empty symbols.


0 20 40 60R [%]

500

1000

1500

2000

2500

3000t* [m

in]

0 20 40 60R [%]

0

0.2

0.4

0.6

0.8

1

Vp [m

3 ]

0 20 40 60R [%]

0.3

0.35

0.4

0.45

0.5

0.55

G [-

]

AU TunnelINJ2-int4PRP1-int3PRP2-int2AU Tunnel

0 20 40 60R [%]

0

1

2

3

4

5

m2,

c [m

in2 ]

106

Figure A.10 – Comparison of the mean residence time, t∗, the swept volume, Vp, theGini coefficient, G, and the second normalized and centralized temporal moment, m2,c,as a function of tracer recovery, R, from different locations (indicated by colors, seelegend) for the different tracers, shown with squares (dye tracer) and diamonds (DNAnanotracer). Results from Test 1A are indicated by filled symbols and from Test 4A byempty symbols.

Table A.1 – Exponential extrapolation parameters in Tests 1A and 4A, where t is timein minutes and tb is the time at which the exponential extrapolation is started.

Tracer Monitoring Fitted function for t > tb tb R2

location [min] [–]Uranine AU Tunnel f = (1.31E−3) exp((−1.67E−3)t) 711 0.998PT-2 AU Tunnel f = (1.55E−3) exp ((−2.08E−3)t) 456 0.943Sulforhodamine B INJ2-int4 f = (4.38E−5) exp ((−5.46E−4)t) 1830 0.999GR-3 INJ2-int4 f = (2.63E−6) exp ((−8.22E−4)t) 1396 0.980Sulforhodamine B PRP1-int3 f = (1.36E−5) exp ((−4.74E−4)t) 1205 0.985GR-3 PRP1-int3 f = (5.28E−7) exp ((−4.93E−4)t) 974 0.940Sulforhodamine B PRP2-int2 f = (1.18E−5) exp ((−6.25E−4)t) 2188 0.971GR-3 PRP2-int2 f = (1.12E−6) exp ((−6.22E−4)t) 769 0.920Sulforhodamine B AU Tunnel f = (4.07E−4) exp ((−5.28E−4)t) 1291 0.997GR-3 AU Tunnel f = (6.91E−5) exp ((−8.75E−4)t) 920 0.974

140 A Appendix

Table A.2 – DNA-based particle tracer characterization.

DNA code DNA length Sizea ζ-potentialb

[base pairs] [nm] [mV]PT-2 120 164±38 -28.8±0.8DAP-3 108 173±44 -29.6±0.2JS-1 105 160±44 -29.5±0.4AM-1 94 166±42 -26.2±0.4GM-2 76 150±25 -29.6±1.1GR-3 67 181±31 -14.4±0.4GR-1 65 167±23 -29.9±0.8aParticle diameter, measured by Luminizer analyticalphotocentrifuge. bMeasured by Malvern Zetasizer Nano.

A.3 Multirate Mass Transfer Model

An attempt was made to compare the transport properties of the DNA nanotracers and

solute dye tracers using a multirate mass transfer model combined with a random walk

particle-tracking algorithm (Salamon et al., 2006b), which is able to account for distinctive

mass transfer processes.

In order to separate different sources of transport heterogeneity in a strongly hetero-

geneous media like fractured aquifers, the general flow configuration has to be resolved.

Otherwise, artifacts from the flow configuration will be interpreted as heterogeneity. The

model geometry and the flow field of the tracer transport simulations were attempted

to be generated using MODFLOW (Harbaugh et al., 2000). However, generating a flow

field that would have resulted in the observed tracer peak arrival times at the different

monitoring locations, i.e., the lack of correlation between the Euclidean distance and the

mean residence time shown in Figure A.9, proved not to be possible with the data cur-

rently available. The following is a documentation of what was achieved despite the lack

of adequate flow field generation.

A model that can account for mass transfer processes is the multirate mass transfer

(MRMT) model (Haggerty and Gorelick, 1995). The model consists of a mobile zone

which accounts for the fast pathways and a series of immobile zones which exchange

mass with the mobile zone via linear mass transfer. The physical representations of the

immobile zones can be imagined either as stagnant zones (Haggerty and Gorelick, 1995)

or zones of slow advection (Fiori and Becker, 2015; Tuykhova and Willmann, 2016) where

the particles are delayed. However, distinguishing the influence of these zones in field-scale

is often very difficult. Nevertheless, the immobile zones account for the slower pathways

by buffering mass transport.

A.3 Multirate Mass Transfer Model 141

Table

A.3

–Com

pleteDNA

sequ

encesandprim

ersof

thesevenDNAnano

tracersused

inthisstud

y.

Test

1Fo

rwar

dTC

CCCT

TCCT

TTGA

TTCC

TTTT

GTGA

TTCT

TTAA

TAAG

AGAA

CAAG

AAAA

ACTC

TTAC

ACCT

TAGT

CTTC

TTAA

TCTT

GGAA

ACTC

GTCT

AAGA

AAGC

CTTA

ACTG

CCCA

ACA

ACAA

CCA

Reve

rse

TGGT

TGTT

GTTG

GGCA

GTTA

AGGC

TTTC

TTAG

ACGA

GTTT

CCAA

GATT

AAGA

AGAC

TAAG

GTGT

AAGA

GTTT

TTCT

TGTT

CTCT

TATT

AAAG

AATC

ACAA

AAGG

AATC

AAA

GGAA

GGGG

APr

imer

1CC

TTCC

TTTG

ATTC

CTTT

TGTG

ATTC

Prim

er 2

TGGT

TGTT

GTTG

GGCA

GTTA

AGFo

rwar

dTA

CCGA

TGCT

GAAC

AAGT

CGAT

GCAG

GCTC

CCGT

CTTT

GAAA

AGGG

GTAA

ACAT

ACAA

GTGG

ATAG

ATGA

TGGG

TAGG

GGCC

TCCA

ATAC

ATCC

AACA

CTCT

ACGC

CCRe

vers

eGG

GCGT

AGAG

TGTT

GGAT

GTAT

TGGA

GGCC

CCTA

CCCA

TCAT

CTAT

CCAC

TTGT

ATGT

TTAC

CCCT

TTTC

AAAG

ACGG

GAGC

CTGC

ATCG

ACTT

GTTC

AGCA

TCGG

TAPr

imer

1TA

CCGA

TGCT

GAAC

AAGT

CGPr

imer

2GG

GCGT

AGAG

TGTT

GGAT

GTFo

rwar

dGA

TTAG

CTTG

ACCC

GCTC

TGTA

GGGT

CGCG

ACTA

CGTG

AGCT

AGGG

CTCC

GGAC

TGGG

CTGT

ATAG

TCGA

GTCT

GATC

TCGC

CCCG

ACAA

CTGC

AAAC

CCCA

ACT

Reve

rse

AGTT

GGGG

TTTG

CAGT

TGTC

GGGG

CGAG

ATCA

GACT

CGAC

TATA

CAGC

CCAG

TCCG

GAGC

CCTA

GCTC

ACGT

AGTC

GCGA

CCCT

ACAG

AGCG

GGTC

AAGC

TAAT

CPr

imer

1GA

TTAG

CTTG

ACCC

GCTC

TGPr

imer

2AG

TTGG

GGTT

TGCA

GTTG

TCFo

rwar

dGC

TTGG

TCTC

TCGT

ACTT

CTCC

TGGA

GATC

AAGG

AAAT

GTTT

CTTG

TCCA

AGCG

GACA

GCGG

TTCT

ACGG

AATG

GATC

TACG

TTAC

TGCC

TGCA

Reve

rse

TGCA

GGCA

GTAA

CGTA

GATC

CATT

CCGT

AGAA

CCGC

TGTC

CGCT

TGGA

CAAG

AAAC

ATTT

CCTT

GATC

TCCA

GGAG

AAGT

ACGA

GAGA

CCAA

GCPr

imer

1GC

TTGG

TCTC

TCGT

ACTT

CTC

Prim

er 2

TGCA

GGCA

GTAA

CGTA

GATC

Forw

ard

ATTG

CACC

CTTA

CCAC

GAAG

ACAG

GTTT

GTCC

AATC

CCAT

CGTT

GCTG

AAGG

CTCA

GGCT

TGGA

CCAG

CTTT

AGTC

Reve

rse

GACT

AAAG

CTGG

TCCA

AGCC

TGAG

CCTT

CAGC

AACG

ATGG

GATT

GGAC

AAAC

CTGT

CTTC

GTGG

TAAG

GGTG

CAAT

Prim

er 1

ATTG

CACC

CTTA

CCAC

GAA

Prim

er 2

GACT

AAAG

CTGG

TCCA

AGC

Test

4Fo

rwar

dTT

CGGA

CAAT

CCTT

TCCA

TATT

ACGC

TCTG

AAGG

CTAC

TACT

CCTT

CTTA

TTAA

CTGG

GTCT

CGTT

TRe

vers

eAA

ACGA

GACC

CAGT

TAAT

AAGA

AGGA

GTAG

TAGC

CTTC

AGAG

CGTA

ATAT

GGAA

AGGA

TTGT

CCGA

APr

imer

1CG

GACA

ATCC

TTTC

CATA

Prim

er 2

ACGA

GACC

CAGT

TAAT

AAG

Forw

ard

GCGA

GATA

CACT

GCCA

AAAA

TCCG

CGTG

ATTA

CGAG

TCGT

GGCA

AATT

TGGT

CTGG

CTGT

GGTC

TRe

vers

eAG

ACCA

CAGC

CAGA

CCAA

ATTT

GCCA

CGAC

TCGT

AATC

ACGC

GGAT

TTTT

GGCA

GTGT

ATCT

CGC

Prim

er 1

GCGA

GATA

CACT

GCCA

AAAA

Prim

er 2

AGAC

CACA

GCCA

GACC

AAAT

GR-1

PT-2

DAP-

3

JS-1

AM-1

GM-2

GR-3

142 A Appendix

The governing equation of the MRMTmodel has advective, dispersive, and source/sink

components, and it can be written for concentration in the mobile zone, cm, as follows

(Carrera et al., 1998; Willmann et al., 2010):

φm∂cm∂t

= ∇ · (D∇cm)− q · ∇cm − Γ (A.1)

where φm is the porosity of the mobile zone, D is the dispersion coefficient, and Γ is the

source/sink term controlling mass transfer between mobile and immobile zones, and can

be written as (Carrera et al., 1998; Haggerty et al., 2000):

Γ = φi,tot

[g ∗ ∂cm

∂t+ gcm0

](A.2)

where cm0 is the initial mobile concentration, ∗ represents convolution product in time,

and φitot is the total immobile porosity. The general form of the memory function, g, is

(Willmann et al., 2010):

g (t) =N∑i=1

αibie−αit (A.3)

where αi is the rate coefficient of the ith immobile zone, and bi is the corresponding

fraction of the immobile porosity. In this form, the memory function implies that N

immobile zones exchange mass with rate αi (Haggerty and Gorelick, 1995; Willmann

et al., 2010):∂cim,i∂t

= αi (cm − cim,i) (A.4)

where cim,i is the concentration in the ith immobile zone. Thus, combining Equa-

tions (A.3) and (A.4), Equation (A.2) can be rewritten as (Willmann et al., 2010):

Γ =

N∑i=1

αiφi (cm − cim,i) (A.5)

where φi = biφi,tot is the porosity of the immobile zone i with the corresponding exchange

rate αi. We use the MRMT model implemented in the particle-tracking code RW3D

(Fernàndez-Garcia et al., 2005; Salamon et al., 2006b,a).

Had the generation of flow field in MODFLOW succeeded, the resulting cell interface

velocities during injection, calculated with MODFLOW, would have then been imported

to the RW3D code.

Due to the finite size of the DNA nanotracers in comparison to conventional solute

tracers, they are prevented from entering small pores. As a result, DNA nanotracers tend

to stay in preferential pathways and show a faster peak arrival time than conventional

solute tracers. Such behavior can be cast naturally in a multirate mass transfer framework.

A.3 Multirate Mass Transfer Model 143

The modeling concept was as follows: The pathways of the DNA nanotracer take place

in the most permeable part of the fracture zone. This part is modeled as the mobile zone

with a fast advective velocity, which is imported from the MODFLOW simulation. A part

of the DNA nanotracers will be delayed, due to a lower permeability of the corresponding

pathways. This will be zone of heterogeneity 1 and will be implemented as the first block

of immobile porosities. The DNA nanotracer will be modeled by these two parts. The

fluorescent tracer has additionally two more parts. First, there is the more permeable part

where the solute fluorescent tracer enters but DNA nanotracer not. Still, the resulting

velocities are so large that they effectively become part of the mobile zone. This is zone of

heterogeneity 2, with large exchange coefficient which will make them part of the mobile

porosity and decrease the mobile advective velocity. And finally, zone 3 of heterogeneity

is the low permeable part which is only accessed by the fluorescent tracer and contributes

to tailing.

Figure A.11 illustrates tracer breakthrough curves of the DNA nanotracer and a solute

dye tracer, marking the main transport processes that were considered in the MRMT

model. Advection yielding the peak of the DNA nanotracer breakthrough curve (marked

with number 1 in Figure A.11) is matched for the mobile zone (marked with letter a).

Dispersion (2) is responsible for the spread of the breakthrough curve. To match the

solute dye tracer peak arrival time, the properties of the zone of heterogeneity 2 are

adjusted (b). Finally, the long tailing (3) is modelled with immobile zones, corresponding

to zones of heterogeneity 1 and 3 (c).

144 A Appendix

Figure A.11 – An illustration of the transport processes and the relevant steps con-sidered in the MRMT model.

A.4 Supporting Information for Chapter 3

Table A.4 gives input parameters for the tomographic inversions in Chapter 3, that is,

the source and receiver coordinated, the tracer peak concentration arrival times, and the

hydraulic head gradients measured in all tests that yielded sufficient tracer breakthrough

curves to determine the tracer peak arrivals. Table A.5 shows all the moments analysis

results for Chapter 3 in tabular format.


Table

A.4

–So

urce

andreceiver

coordinatesof

injectionandmon

itorin

glocatio

ns,respectiv

ely,

used

inthepre-

andpo

st-stim

ulation

tomograph

icinversions,w

ithinform

ationof

tracer

peak

concentrationtravel

times,hydraulic

head

gradients,andEu

clideandistances.

trave

l tim

etra

vel t

ime

delta

hdi

stan

cegr

adie

ntx

yz

xy

zm

inda

ysm

mm

/m59

105.

217

.172

107.

533

476

0.33

3Te

st 1

APr

e-HS

20.7

0.15

58.4

106.

216

.272

107.

533

1370

0.95

69Te

st 1

BPr

e-HS

21.7

3.19

5910

5.2

17.1

7210

7.5

3363

80.

4446

.5Te

st 2

Pre-

HS20

.72.

2559

105.

217

.153

.610

5.5

20.5

477

0.33

46Te

st 2

Pre-

HS6.

47.

2059

105.

217

.155

.210

2.8

16.1

870.

0646

Test

2Pr

e-HS

4.6

9.99

53.2

115.

75.

344

.311

7.5

1123

20.

1642

Test

3B

Post

-HS

10.7

3.92

5910

5.2

17.1

55.2

102.

816

.155

00.

3858

.5Te

st 3

APo

st-H

S4.

612

.71

5910

5.2

17.1

48.2

510

3.4

17.5

459

0.32

58.5

Test

3A

Post

-HS

10.9

5.36

48.2

510

3.4

17.5

55.2

102.

816

.125

50.

1862

.5Te

st 4

APo

st-H

S7.

18.

7848

.25

103.

417

.553

.610

5.5

20.5

475

0.33

62.5

Test

4A

Post

-HS

6.5

9.64

48.2

510

3.4

17.5

7210

7.5

3394

50.

6663

Test

4A

Post

-HS

28.7

2.20

48.2

510

3.4

17.5

5910

5.2

17.1

1390

0.97

62.5

Test

4A

Post

-HS

10.9

5.73

41.3

109.

211

.659

105.

217

.116

801.

1762

.5Te

st 4

BPo

st-H

S19

.03.

3053

.211

5.7

5.3

44.3

117.

511

240

0.17

59Te

st 5

Post

-HS

10.7

5.50

48.2

510

3.4

17.5

55.2

102.

816

.144

50.

3160

.5Te

st 6

APo

st-H

S7.

18.

50

5.89

sour

ce co

ordi

nate

sre

ceive

r coo

rdin

ates

note

s

Aver

age

grad

ient

146 A Appendix

Table

A.5

–Mom

entanalysis

resultsvisualized

inFigures

3.3and

3.5.

Te

st

Tra

ce

rd

hD

ista

nce

Firs

t arriv

al

Me

an

resid

en

ce

time

Re

co

ve

ryC

(pe

ak)/

C(in

j)S

we

pt v

olu

me

Gin

i co

effic

ien

t2

nd

mo

me

nt

[m]

[m]

[min

]t*

[min

]R

[%]

[-]V

(p) [m

3]

G [-]

m(2

,c) [m

in2

]

Te

st 1

AU

ran

ine

AU

Tu

nn

el

32

0.7

15

79

50

44

4.7

9E

-04

0.4

70

.32

2.9

1E

+0

5IN

J2-in

t4

Te

st 1

AU

ran

ine

INJ1

-int4

2.5

10

.96

06

INJ2

-int4

Te

st 2

Ura

nin

eP

RP

1-in

t34

64

.61

07

96

6.3

5.2

7E

-03

0.0

51

0.5

91

.05

E+

06

INJ2

-int4

Te

st 2

Ura

nin

eP

RP

2-in

t24

66

.41

20

15

71

1.5

1.3

1E

-03

0.0

23

0.5

56

.23

E+

06

INJ2

-int4

Te

st 2

Ura

nin

eA

U T

un

ne

l4

6.5

20

.71

65

12

62

57

4.1

4E

-04

0.7

30

.42

6.4

4E

+0

6IN

J2-in

t4

Te

st 2

Ura

nin

eIN

J1-in

t44

61

0.9

10

30

INJ2

-int4

Te

st 2

Ura

nin

eP

RP

1-in

t24

68

.6~

21

60

-57

60

INJ2

-int4

Te

st 3

AT

ino

pa

l CB

S-X

PR

P1

-int3

58

.54

.61

24

94

20

.98

1.1

2E

-03

0.0

03

30

.36

4.8

1E

+0

5IN

J2-in

t4

Te

st 3

AT

ino

pa

l CB

S-X

INJ1

-int4

58

.51

0.9

34

43

.18

E-0

4IN

J2-in

t4

Te

st 3

AT

ino

pa

l CB

S-X

AU

Tu

nn

el

59

20

.76

12

INJ2

-int4

Te

st 3

BE

osin

eP

RP

2-in

t14

21

0.7

12

67

81

9.6

3.4

9E

-04

0.0

25

0.4

04

.38

E+

05

INJ2

-int2

Te

st 4

AS

ulfo

rho

da

min

e B

PR

P1

-int3

62

.57

.17

32

06

93

.14

.30

E-0

40

.03

80

.47

4.4

6E

+0

6IN

J1-in

t4

Te

st 4

AS

ulfo

rho

da

min

e B

PR

P2

-int2

62

.56

.52

64

13

83

2.7

1.7

0E

-03

0.0

23

0.4

21

.85

E+

06

INJ1

-int4

Te

st 4

AS

ulfo

rho

da

min

e B

AU

Tu

nn

el

63

28

.72

91

25

94

59

2.3

2E

-04

0.9

20

.36

3.9

9E

+0

6IN

J1-in

t4

Te

st 4

AS

ulfo

rho

da

min

e B

INJ2

-int4

62

.51

0.9

32

82

77

25

.41

.95

E-0

40

.08

90

.33

3.9

1E

+0

6IN

J1-in

t4

Te

st 4

AS

ulfo

rho

da

min

e B

PR

P1

-int2

62

.51

6.6

17

30

INJ1

-int4

Te

st 4

BU

ran

ine

INJ2

-int4

62

19

69

31

.19

E-0

4IN

J1-in

t2

Te

st 4

BU

ran

ine

PR

P1

-int3

62

16

85

8IN

J1-in

t2

Te

st 4

BU

ran

ine

AU

Tu

nn

el

62

.53

7.5

13

80

INJ1

-int2

Te

st 4

BU

ran

ine

PR

P1

-int2

62

15

.21

45

0IN

J1-in

t2

Te

st 4

BU

ran

ine

PR

P2

-int2

62

15

.61

35

0IN

J1-in

t2

Te

st 5

Eo

sin

eP

RP

2-in

t15

91

0.7

13

31

29

88

.82

.28

E-0

40

.04

20

.43

1.4

0E

+0

6IN

J2-in

t2

Te

st 6

AS

ulfo

rho

da

min

e B

PR

P1

-int3

60

.57

.18

43

37

61

.62

1.4

3E

-04

0.0

27

0.4

57

.71

E+

06

INJ1

-int4

Te

st 6

BU

ran

ine

PR

P1

-int3

59

.51

69

10

INJ1

-int2

Te

st 7

Su

lforh

od

am

ine

BP

RP

2-in

t12

61

0.7

39

23

65

96

.04

E-0

30

.09

40

.46

8.8

1E

+0

4IN

J2-in

t2

Te

st 8

Eo

sin

eA

U T

un

ne

l3

0.5

20

.77

02

24

26

54

.42

E-0

43

.10

.50

4.6

3E

+0

6IN

J2-in

t4

Te

st 8

Eo

sin

eP

RP

1-in

t33

04

.61

41

23

22

.81

.37

E-0

30

.07

40

.64

2.7

6E

+0

6IN

J2-in

t4

Te

st 8

Eo

sin

eP

RP

2-in

t23

06

.45

78

66

2.5

1.0

4E

-03

0.0

45

0.5

48

.54

E+

05

INJ2

-int4

Mo

nito

ring

loca

tion

Inje

ctio

n

loca

tion

A.5 Authored and co-authored publications and posters 147

A.5 Authored and co-authored publications and posters

Publications in peer-reviewed journals

• Kittilä, A., Jalali, M.R., Somogyvári, M., Evans, K.F., Saar, M.O. and Kong, X.-Z.

(2020). Characterization of the effects of hydraulic stimulation with tracer-based

temporal moment analysis and tomographic inversion, Geothermics.

• Kittilä, A., Jalali, M.R., Evans, K.F., Willmann, M., Saar, M.O. and Kong, X.-Z.

(2019). Field comparison of DNA-labeled nanoparticle and solute tracer transport

in a fractured crystalline rock. Water Resources Research, 55 (8): 6577-6595.

• Gischig, V.S., Giardini, D., Amann, F., Hertrich, M., Krietsch, H., Loew, S., Mau-

rer, H., Villiger, L., Wiemer, S., Bethmann, F., Brixel, B., Doetsch, J., Doonechaly,

N.G., Driesner, T., Dutler, N., Evans, K.F., Jalali, M., Jordan, D., Kittilä, A.,

Ma, X., Meier, P., Nejati, M., Obermann, A., Plenkers, K., Saar, M.O., Shakas,

A. and Valley, B. (2020). Hydraulic stimulation and fluid circulation experiments

in underground laboratories: Stepping up the scale towards engineered geothermal

systems. Geomechanics for Energy and the Environment, 100175.

• Amann, F., Gischig, V., Evans, K., Doetsch, J., Jalali, R., Valley, B., Krietsch,

H., Dutler, N., Villiger, L., Brixel, B., Klepikova, M., Kittilä, A., Madonna, C.,

Wiemer, S., Saar, M.O., Loew, S., Driesner, T., Maurer, H. and Giardini, D. (2018).

The seismo-hydro-mechanical behaviour during deep geothermal reservoir stimula-

tions: open questions tackled in a decameter-scale in-situ stimulation experiment.

Solid Earth, 9: 115-137.

• Kong, X.-Z., Deuber, C., Kittilä, A., Somogyvari, M., Mikutis, G., Bayer, P.,

Stark, W.J. and Saar, M.O. (2018). Tomographic reservoir imaging with DNA-

labeled silica nanotracers: The first field validation. Environmental Science &

Technology, 52 (23): 13681-13689.

• Mikutis, G., Deuber, C.A., Schmid, L., Kittilä, A., Lobsiger, N., Puddu, M., As-

geirsson, D.O., Grass, R.N., Saar, M.O. and Stark, W.J. (2018). Silica-encapsulated

DNA-based tracers for aquifer characterization. Environmental Science & Technol-

ogy, 52: 12142-12152.

Other publications

• Doetsch, J., Gischig, V., Krietsch, H., Villiger, L., Amann, F., Dutler, N., Jalali,

M., Brixel, B., Roques, C., Giertzuch, P., Kittilä, A. and Hochreutener, R. (2018).

Grimsel ISC experiment description. Technical report, SCCER-SoE, Zurich.

148 A Appendix

• Kittilä, A., Deuber, C., Mikutis, G., Evans, K., Puddu, M., Grass, R.N., Stark,

W.J. and Saar, M.O. (2016). Comparison of novel synthetic DNA nano-colloid

tracer and classic solute tracer behavior. Conference proceedings, European Geother-

mal Congress 2016.

Posters

• Kittilä, A., Jalali, M., Evans, K., Kong, X.-Z. and Saar, M.O. (2018). Comparison

between DNA nanotracer and solute tracer tests in a fractured crystalline rock –

GTS case study. SCCER-SoE Annual Conference 2018.

• Kittilä, A., Evans, K., Jalali, M., Willmann, M. and Saar, M.O. (2017). Tracer

based characterization of the connected fracture volume in the DUG Lab at the

Grimsel Test Site. SCCER-SoE Annual Conference 2017.

• Kittilä, A., Evans, K., Jalali, M., Willmann, M. and Saar, M.O. (2017). DNA

nanotracers in characterization of stimulation enhanced pore space in fractured

rock. 44th IAH Congress.

• Kittilä, A., Evans, K., Deuber, C. and Saar, M.O. (2016). Flow path characteriza-

tion using DNA-based smart tracers in the Grimsel Deep Underground Geothermal

Laboratory (DUG-Lab). SCCER-SoE Annual Conference 2016.

• Kittilä, A., Evans, K., Puddu, M., Mikutis, G., Grass, R.N., Deuber, C. and

Saar, M.O. (2016). The use of novel DNA nanotracers to determine groundwater

flow paths – a test study at the Grimsel Deep Underground Geothermal (DUG)

Laboratory in Switzerland. EGU General Assembly 2016.

List of Figures

1.1 (a) Concentrations of uranine (blue) and sulforhodamine B (orange) in

samples. The grey area indicates samples where uranine concentrations are

below the background level. (b) Plot of fluorometer measurements from

lamp 1, L1 (uranine), and L2 (sulforhodamine B) in mV, corresponding to

the samples (n=11) outlined in the grey area in a). . . . . . . . . . . . . . 22

1.2 STEM micrograph (left) and structural illustration (right) of the DNA

nanotracer. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22

1.3 An overview of the ISC experiment test phases at the Grimsel Test Site

(GTS), modified from Amann et al. (2018). . . . . . . . . . . . . . . . . . 24

2.1 a) Projection of the boreholes and intervals in the DUG-Lab (Krietsch

et al., 2018). b) Timeline of different tracer tests as well as the hydraulic

stimulation phases. In Test 1, tracers were injected into INJ2-int3 and

INJ2-int4, and monitored in INJ1 and the AU Tunnel. In Test 4, trac-

ers were injected into INJ1-int2 and INJ1-int4, and monitored in PRP1,

PRP2, INJ2-int4 and the AU Tunnel. One of the main shear zone planes

(S3.2) is shown, and the intersections of S3 in the AU Tunnel are visualized

with dark green disks. All four S1-type structures, and the two S3-type

structures are shown in Figure A.6. . . . . . . . . . . . . . . . . . . . . . . 35

2.2 Comparison of the BTCs of DNA nanotracers PT-2, DAP-3, JS-1, and

AM-1, injected together in INJ2-int4 and monitored in the AU Tunnel in

Test 1A. The concentrations are normalized to the injected mass of the

respective DNA nanotracer. Each sample was analyzed in triplicate, and

the error bars present standard deviations of the replicates. Inset: The

concentration fluctuation is described using the autocorrelation of a BTC

with a lag distance of one. The autocorrelation was performed only with

the data points shared by all four BTCs, consisting of 23 data points. . . 44

149

150 LIST OF FIGURES

2.3 Normalized breakthrough curves of dye tracers and DNA nanotracers from

Tests 1A, 4A, and 4B. Only measured concentrations (i.e., no extrapo-

lated data) are plotted. Line colors indicate the monitoring locations of

breakthrough curves (for reference, see Figure 2.1). . . . . . . . . . . . . . 45

2.4 Residence time distribution (RTD) curves of dye tracers and DNA nano-

tracers from Tests 1A, 4A, and 4B. RTD values calculated from the mea-

sured concentrations are illustrated with lines with circle markers. RTD

values of the extrapolated exponential decay are shown with solid lines.

Line colors indicate the monitoring locations. . . . . . . . . . . . . . . . . 48

2.5 F − Φ curves derived from the RTDs obtained in Tests 1A and 4A. Here,

only the extrapolatable RTDs are reported. Line colors indicate the mon-

itoring locations of RTDs. . . . . . . . . . . . . . . . . . . . . . . . . . . 54

3.1 Overview of the experiment setup at the Grimsel Test Site (GTS). a) The

shear zones S1.3 and S3.2 are shown with red and green planes, respectively.

The contours on the structures are drawn to give a sense of curvature and

are not related to the tomographic inversion grid. A second S3 structure

(S3.1) lies close to S3.2 but is not shown in the figure to avoid complication.

However, the intersections of the S3 shear zones with the AU Tunnel are

indicated as green disks in the AU Tunnel (Krietsch et al., 2018). The

monitoring intervals (red cylinders) in the INJ and PRP boreholes (blue

and green lines, respectively) and the AU Tunnel outflow point (red circle)

are also marked. The dashed blue lines define the boundaries of the 5m

thick section through the 3D tomographic inversion grid. b) Top-view of

(a), showing the positions of the inlet and outlet points with respect to the

tomographic section, which approximates the S3 structures. Note that the

grid on the structures does not represent the grid used for the tomographic

inversions. c) The red boxes in the core images (Krietsch et al., 2018) mark

the 0.52m long intervals (except Test 7, which had a length of 1.0m) in the

INJ1 and INJ2 boreholes, which were isolated by hydraulic packer systems.

These intervals are indicated by the bold font in (a). . . . . . . . . . . . . 66

LIST OF FIGURES 151

3.2 Pre- and post-stimulation residence time distribution (RTD) curves of the

tracers injected into the INJ2-i4 interval (dotted and solid lines – before

stimulation, dashed lines – after stimulation). E (t) is the age distribution

function. The shifts of the peaks, a result of the hydraulic stimulation ex-

periments, are shown with square symbols (open – before stimulation, filled

– after stimulation): at the AU Tunnel, the shift was from 638 to 233min,

at PRP1-i3 from 87 to 31min and at PRP2-i2 from 477 to 123min. The

RTDs from Test 3A all fall in the lower-right corner of the graph and are

most likely affected by the thermal tracer test, conducted at the test site

immediately before Test 3, as discussed in the main text. . . . . . . . . . . 77

3.3 Parallel coordinates plot of the recorded tracer breakthrough curves pre-

sented in Fig. 3.2 and their temporal moments from both the pre- and post-

stimulation tracer tests. Note that some of the curves exceed the plotted

coordinate values, such as the R and Vp values from the AU Tunnel. These

values are indicated with arrows next to the corresponding coordinate axes.

The hydraulic head difference, ∆h, was calculated by subtracting the atmo-

spheric pressure from the absolute fluid injection pressure at the wellhead

(Table 3.1) and converting the pressure value to equivalent hydraulic head,

using a water density of 1000 kg/m3 for simplicity. In general, the eleva-

tions of the injection and production points were essentially the same, the

only exception being the AU Tunnel monitoring point, which was approxi-

mately 0.5m lower (Supplementary Table A.4). A correction was made for

this exception by adding 0.5m to the hydraulic head difference. The data

in this figure can also be found in the Supplementary Table A.5. . . . . . 78

3.4 Post-stimulation residence time distribution (RTD) curves of the tracers

from Tests 3 to 7. E (t) is the age distribution function. The solid and

dashed lines denote injection into INJ1-i4 and INJ2-i2, respectively. The

squares on the RTD curves indicate the peaks that are associated with

major flow paths (see main text), as the logarithmic scale hides some peaks. 82

3.5 Parallel coordinates plot of the parameters derived from the residence time

distribution (RTD) curves from tests that follow the hydroshearing stimu-

lation program. The Test 3A curve in red (injected into INJ2-i4) is shown

as a reference to Fig. 3.3. Tracers injected into INJ2-i2 are plotted as

dashed lines and tracers injected into INJ1-i4 are plotted as solid lines.

The data in this graph can also be found in the Supplementary Table A.5. 83

152 LIST OF FIGURES

3.6 Sections through tomograms of the hydraulic conductivity, K, distribu-

tions reconstructed using the peak arrival times of the tracer breakthrough

curves (BTCs) before (a) and after (b) the hydraulic shearing stimula-

tions. The boundaries of the 5 m wide sections are denoted by the pair of

dashed blue lines in Fig. 3.1. The distribution of K values was the same

throughout the thickness of the sections. The insets in (a) and (b) pro-

vide the histograms of the hydraulic conductivity, K, distributions. The

tracer injection/monitoring intervals are also shown in the respective sec-

tions. The observed travel times are plotted against the tomographically

reconstructed ones for the pre- (c) and post-stimulation (d) calculations.

The error bars in (c) and (d) represent standard deviations of the travel

times, obtained by different positions of the staggered grids. . . . . . . . . 86

4.1 Overview of the experiment setup at the Grimsel Test Site (GTS) (modi-

fied from Krietsch et al. (2018)). The shear zone structures, designated S1

and S3, are shown as red and green planes, respectively. There is also a

shear zone classified as S2, which is slightly discordant to S1, however, the

S1 and S2 shear zones cannot be distinguished in the field (Keusen et al.,

1989; Krietsch et al., 2018). The injection and monitoring intervals, asso-

ciated with this study, in boreholes INJ1, INJ2, PRP1, and PRP2 (black

cylinders) and the AU Tunnel outflow point (black circle) are marked. The

orange arrows indicate the interpreted flow directions in injection inter-

val INJ2-int4, where the arrows with solid lines are associated with the

more prominent flow directions (see Section ’Redistribution of fluid flow’

for more information). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103

4.2 Injection temperature, pressure, and flow rate during fluid injection into

INJ2-int4 (modified from Doetsch et al. (2018a)). It is worth noting that

the system was likely not yet at steady-state during Test 8. . . . . . . . . 106

4.3 Comparison of the residence time distribution (RTD) curves (left) and the

F − Φ curves (right). The RTD and the F − Φ curves are from before

(Test 8) and during (Test 9) hot water injection at the four monitoring

locations, namely the AU Tunnel outflow point, PRP1-int3, PRP2-int2,

and PRP1-int2. The x-symbol on the Test 9 RTD curve, obtained from

PRP1-int2, marks the start of a 2-day breakdown of the water injection

system (Fig. 4.2) and the dashed diagonal lines in the F−Φ plots represent

a homogeneous fracture system. Note the different scales of the axes for

the RTD curves. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111

LIST OF FIGURES 153

4.4 Temperature at monitoring locations PRP1-int3, PRP2-int2, and PRP1-

int2 during Tests 8 and 9 (Brixel et al., 2019). Temperatures between

Tests 8 and 9 are not shown in this figure due to data copyright. . . . . . 117

A.1 Boreholes at the GTS. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128

A.2 Injection and production intervals in the boreholes. The estimated flow

rates in INJ2 have been calculated based on previous pulse tests. . . . . . 128

A.3 In the first panel, top left, the layout of the tunnels (blue), the shear zones

(red lines), and the modelled water injection point (green dot) at the GTS

are shown. Pressure propagation with 0.5 l min−1 injection rate after a)

10minutes, b) 30minutes, c) 1 hour, d) 6 hours, and e) 12 hours from the

start of injection. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132

A.4 Pressure changes from chosen cells near the intersection of the shear zones

and the tunnels. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133

A.5 Tracer transportation 1 h (a), 4 h (b), and 18 h (c) after the injection of

the tracer, with simultaneous constant 0.5 l min−1 injection of tracer free

water. The tracer concentrations are shown as mass fraction from the

initial water. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134

A.6 An illustration of the shear zones S3 (green) and S1 (red) (Krietsch et al.,

2018), intersected by the boreholes INJ1, INJ2, PRP1, and PRP2, and the

AU Tunnel monitoring location (red circle). . . . . . . . . . . . . . . . . . 136

A.7 Particle size distributions of the DNA nanotracers. . . . . . . . . . . . . . 137

A.8 qPCR dilution curves, where values in parenthesis of the linear fittings are

given as (slope Y-intercept R2). Since both the standard dilution curve

and the samples are prepared in the same water, using the same qPCR

reagents, and the standard tracer dilution in all cases is linear (R2 ≥0.99),the tracer quantification is expected to be accurate. . . . . . . . . . . . . . 137

A.9 Comparison of the tracer recovery, R, the mean residence time, t∗, the

swept volume, Vp, the Gini coefficient, G, and the second normalized and

centralized temporal moment, m2,c, as a function of distance from different

locations (indicated by colors, see legend) for the different tracers, shown

with squares (dye tracer) and diamonds (DNA nanotracer). Results from

Test 1A are indicated by filled symbols and from Test 4A by empty symbols.138

154 LIST OF FIGURES

A.10 Comparison of the mean residence time, t∗, the swept volume, Vp, the Gini

coefficient, G, and the second normalized and centralized temporal mo-

ment, m2,c, as a function of tracer recovery, R, from different locations (in-

dicated by colors, see legend) for the different tracers, shown with squares

(dye tracer) and diamonds (DNA nanotracer). Results from Test 1A are

indicated by filled symbols and from Test 4A by empty symbols. . . . . . 139

A.11 An illustration of the transport processes and the relevant steps considered

in the MRMT model. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 144

List of Tables

2.1 Interval information for Tests 1 and 4 (see also Figure 2.1). . . . . . . . . 36

2.2 Moment analysis results from Tests 1A, 4A, and 4B. . . . . . . . . . . . . 49

3.1 Summary of the solute dye tracer tests conducted at the Grimsel Test Site

(GTS) during the ISC experiments. . . . . . . . . . . . . . . . . . . . . . . 68

3.2 Borehole interval information. . . . . . . . . . . . . . . . . . . . . . . . . . 70

3.3 Hydraulic shearing (HS) and hydraulic fracturing (HF) borehole interval

information (Doetsch et al., 2018a), relevant for the tracer tests presented

here. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71

3.4 Summary of the pre- and post-stimulation moment analysis results, where

R is tracer recovery, t∗ is mean residence time, Vp is tracer swept volume,

G is Gini coefficient and m2,c is second temporal moment. . . . . . . . . . 80

3.5 Statistical parameters for the tomographic inversion profiles of logK. . . . 88

4.1 Summary of tracer injection and production during Tests 8 and 9. . . . . . 105

4.2 Summary of the tracer transport parameters. . . . . . . . . . . . . . . . . 113

4.3 Summary of thermal and petrophysical properties used. . . . . . . . . . . 116

4.4 Results of solving fracture surface area using Eq. (4.16). . . . . . . . . . . 118

A.1 Exponential extrapolation parameters in Tests 1A and 4A, where t is time

in minutes and tb is the time at which the exponential extrapolation is

started. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139

A.2 DNA-based particle tracer characterization. . . . . . . . . . . . . . . . . . 140

A.3 Complete DNA sequences and primers of the seven DNA nanotracers used

in this study. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141

A.4 Source and receiver coordinates of injection and monitoring locations, re-

spectively, used in the pre- and post-stimulation tomographic inversions,

with information of tracer peak concentration travel times, hydraulic head

gradients, and Euclidean distances. . . . . . . . . . . . . . . . . . . . . . . 145

A.5 Moment analysis results visualized in Figures 3.3 and 3.5. . . . . . . . . . 146

155

Bibliography

Albarran, N., Missana, T., Alonso, U., García-Gutiérrez, M., and López, T. (2013).

Analysis of latex, gold and smectite colloid transport and retention in artificial fractures

in crystalline rock. Colloids and Surfaces A: Physicochemical and Engineering Aspects,

435:115–126. 21, 31, 51, 55

Amann, F., Gischig, V., Evans, K., Doetsch, J., Jalali, R., Valley, B., Krietsch, H.,

Dutler, N., Villiger, L., Brixel, B., Klepikova, M., Kittilä, A., Madonna, C., Wiemer,

S., Saar, M. O., Loew, S., Driesner, T., Maurer, H., and Giardini, D. (2018). The

seismo-hydromechanical behavior during deep geothermal reservoir stimulations: Open

questions tackled in a decameter-scale in situ stimulation experiment. Solid Earth,

9(1):115–137. 16, 23, 24, 34, 63, 65, 102, 149

Anderson, M. P. (2005). Heat as a Ground Water Tracer. Ground Water, 43(6):951–968.

100

André, L., Rabemanana, V., and Vuataz, F.-D. (2006). Influence of water-rock inter-

actions on fracture permeability of the deep reservoir at Soultz-sous-Forêts, France.

Geothermics, 35(5-6):507–531. 19, 99

Aquilanti, L., Clementi, F., Landolfo, S., Nanni, T., Palpacelli, S., and Tazioli, A. (2013).

A DNA tracer used in column tests for hydrogeology applications. Environmental Earth

Sciences, 70(7):3143–3154. 30

Axelsson, G. (2013). Tracer tests in geothermal resource management. EPJ Web of

Conferences, 50:02001. 19, 63, 80

Ayling, B. F., Hogarth, R. A., and Rose, P. E. (2016). Tracer testing at the Habanero

EGS site, central Australia. Geothermics, 63:15–26. 19, 30, 39, 41, 63, 80, 100

Bailly-Comte, V., Durepaire, . X., Batiot-Guilhe, . C., and Schnegg, P.-A. (2018). In situ

monitoring of tracer tests: how to distinguish tracer recovery from natural background.

Hydrogeology Journal, 26:2057–2069. 47

Batchelor, A. S. (1982). The creation of Hot Dry Rock systems by combined explosive

and hydraulic fracturing. In International conference on geothermal energy, volume 2,

pages 321–342, Florence, Italy. BHRA Fluid Engineering, Cranfield, Beds. 63

157

158 BIBLIOGRAPHY

Becker, M. W., Reimus, P., and Vilks, P. (1999). Transport and Attenuation of

Carboxylate-Modified Latex Microspheres in Fractured Rock Laboratory and Field

Tracer Tests. Ground Water, 37(3):387–395. 31

Becker, M. W. and Shapiro, A. M. (2000). Tracer transport in fractured crystalline rock:

Evidence of nondiffusive breakthrough tailing. Water Resources Research, 36(7):1677.

14, 15, 30, 31

Becker, M. W. and Tsoflias, G. P. (2010). Comparing flux-averaged and resident con-

centration in a fractured bedrock using ground penetrating radar. Water Resources

Research, 46:W09518. 14

Berkowitz, B. (2002). Characterizing flow and transport in fractured geological media: A

review. Advances in Water Resources, 25(8-12):861–884. 13, 15

Berkowitz, B., Cortis, A., Dentz, M., and Scher, H. (2006). Modeling non-Fickian trans-

port in geological formations as a continuous time random walk. Reviews of Geophysics,

44(2):49. 14, 15

Bero, N. J., Ruark, M. D., and Lowery, B. (2016). Bromide and chloride tracer application

to determine sufficiency of plot size and well depth placement to capture preferential

flow and solute leaching. Geoderma, 262:94–100. 30

Bossart, P. and Mazurek, M. (1991). Structural geology and water flow-paths in the

migration shear-zone. Technical report, Nagra, Wettingen, Switzerland. 73

Brauchler, R., Hu, R., Hu, L., Jiménez, S., Bayer, P., Dietrich, P., and Ptak, T. (2013).

Rapid field application of hydraulic tomography for resolving aquifer heterogeneity in

unconsolidated sediments. Water Resources Research, 49(4):2013–2024. 74

Brixel, B., Jalali, M., Klepikova, M., Roques, C., and Löw, S. (2018). In-situ characteri-

zation of fluid flow in an EGS-analog reservoir. 36

Brixel, B., Jalali, M., and Klepikova, M. V. (2019). Single- and cross-hole hydraulic char-

acterization of the ISC support volume at the Grimsel Test Site. https://www.research-

collection.ethz.ch/handle/20.500.11850/306205. 76, 116, 117, 153

Brixel, B., Klepikova, M., Jalali, M., Roques, C., Lei, Q., Krietsch, H., and Loew, S.

(2020). Tracking fluid flow in shallow crustal fault zones: 1. Insights from single-hole

permeability measurements. Submitted to JGR: Solid Earth. 14, 70, 85, 104, 112, 114,

119

Buscarlet, E., Moon, H., Wallis, I., and Quinao, J. (2015). Reservoir Tracer Test at

the Ngatamariki Geothermal Field , New Zealand. In 37th New Zealand Geothermal

Workshop, page 8, Taupo. 19, 41, 63, 80

BIBLIOGRAPHY 159

Carlslaw, H. S. and Jaeger, J. C. (1959). Conduction of heat in solids. Clarendon Press,

Oxford, 2nd edition. 108

Carrera, J., Sánchez-Vila, X., Benet, I., Medina, A., Galarza, G., and Guimerà, J. (1998).

On matrix diffusion: formulations, solution methods and qualitative effects. Hydroge-

ology Journal, 6(1):178–190. 142

Cherubini, C., Giasi, C. I., and Pastore, N. (2014). On the reliability of analytical models

to predict solute transport in a fracture network. Hydrology and Earth System Sciences,

18(6):2359–2374. 15

Chrysikopoulos, C. V. (1993). Artificial tracers for geothermal reservoir studies. Envi-

ronmental Geology, 22(1):60–70. 63

Chrysikopoulos, C. V. and Katzourakis, V. E. (2015). Colloid particle size-dependent

dispersivity. Water Resources Research, 51(6):4668–4683. 31

Cirpka, O. A. and Kitanidis, P. K. (2000). Characterization of mixing and dilution in

heterogeneous aquifers by means of local temporal moments. Water Resources Research,

36(5):1221–1236. 50, 71, 72, 115

Clauser, C. (1992). Permeability of Crystalline Rocks. EOS Transactions AGU,

73(21):237–238. 88

Clauser, C. and Ewert, M. (2018). The renewables cost challenge: Levelized cost of

geothermal electric energy compared to other sources of primary energy – Review and

case study. Renewable and Sustainable Energy Reviews, 82(3):3683–3693. 17

Colombani, N., Giambastiani, B., and Mastrocicco, M. (2015). Combined use of heat

and saline tracer to estimate aquifer properties in a forced gradient test. Journal of

Hydrology, 525:650–657. 100

Cornet, F. (1987). Results from Le Mayet de Montagne project. Geothermics, 16(4):355–

374. 63

Dai, Z., Wolfsberg, A., Reimus, P., Deng, H., Kwicklis, E., Ding, M., Ware, D., and Ye, M.

(2012). Identification of sorption processes and parameters for radionuclide transport

in fractured rock. Journal of Hydrology, 414-415:516–526. 51

Danckwerts, P. (1953). Continuous flow systems. Chemical Engineering Science, 2(1):1–

13. 106

Davis, S. N., Thompson, G. M., Bentley, H. W., and Stiles, G. (1980). Ground-Water

Tracers - A Short Review. Ground Water, 18(1):14–23. 30

de La Bernardie, J., Bour, O., Le Borgne, T., Guihéneuf, N., Chatton, E., Labasque,

T., Le Lay, H., and Gerard, M.-F. (2018). Thermal Attenuation and Lag Time in

Fractured Rock: Theory and Field Measurements From Joint Heat and Solute Tracer

160 BIBLIOGRAPHY

Tests. Water Resources Research, 54(12):053–10. 118

de Marsily, G. (1986). Quantitative Hydrogeology: Groundwater Hydrology for Engineers.

Academic, Orlando, Fla. 100

Dickson, M. H. and Fanelli, M. (2002). What is geothermal energy?

http://users.metu.edu.tr/mahmut/pete450/Dickson.pdf. 16

Doetsch, J., Gischig, V., Krietsch, H., Villiger, L., Amann, F., Dutler, N., Jalali, M.,

Brixel, B., Roques, C., Giertzuch, P.-L., Kittilä, A., and Hochreutener, R. (2018a).

Grimsel ISC Experiment Description. Technical report, SCCER-SoE, Zurich. 36, 37,

38, 43, 67, 68, 69, 70, 71, 76, 80, 81, 82, 85, 89, 106, 116, 118, 152, 155

Doetsch, J., Gischig, V. S., Villiger, L., Krietsch, H., Nejati, M., Amann, F., Jalali,

M., Madonna, C., Maurer, H., Wiemer, S., Driesner, T., and Giardini, D. (2018b).

Subsurface Fluid Pressure and Rock Deformation Monitoring Using Seismic Velocity

Observations. Geophysical Research Letters, 45(19):389–10. 85, 89

Dou, Z., Chen, Z., Zhou, Z., Wang, J., and Huang, Y. (2018). Influence of eddies on

conservative solute transport through a 2D single self-affine fracture. International

Journal of Heat and Mass Transfer, 121:597–606. 14, 50

Durham, W. B. and Bonner, B. P. (1994). Self-propping and fluid flow in slightly off-

set joints at high effective pressures. Journal of Geophysical Research: Solid Earth,

99(B5):9391–9399. 14

Edery, Y., Geiger, S., and Berkowitz, B. (2016). Structural controls on anomalous trans-

port in fractured porous rock. Water Resources Research, 52(7):5634–5643. 15

Edwards, B., Kraft, T., Cauzzi, C., Kästli, P., and Wiemer, S. (2015). Seismic moni-

toring and analysis of deep geothermal projects in St Gallen and Basel, Switzerland.

Geophysical Journal International, 201:1022–1039. 62

Ellsworth, W. L. (2013). Injection-Induced Earthquakes. Science, 341(6142):142–149. 17

Evans, K., Wieland, U., Wiemer, S., and Giardini, D. (2014). Deep Geothermal Energy,

R&D Roadmap for Switzerland, 2014. Technical report, Swiss Competence Center for

Energy Research Supply of Electricity, SCCER-SoE, Zurich. 23

Evans, K. F. (2015). Reservoir creation. In Hirschberg, S., Wiemer, S., and Burgherr,

P., editors, Energy from the Earth, chapter 3.2, pages 82–118. vdf Hochschulverlag AG,

Zürich. 16, 17, 18, 19, 63, 64, 99, 124, 126

Evans, K. F., Moriya, H., Niitsuma, H., Jones, R. H., Phillips, W. S., Genter, A., Sausse,

J., Jung, R., and Baria, R. (2005). Microseismicity and permeability enhancement of

hydrogeologic structures during massive fluid injections into granite at 3 km depth at

the Soultz HDR site. Geophysical Journal Internation, 160(1):388–412. 17

BIBLIOGRAPHY 161

Evans, K. F., Zappone, A., Kraft, T., Deichmann, N., and Moia, F. (2012). A survey of

the induced seismic responses to fluid injection in geothermal and CO2 reservoirs in

Europe. Geothermics, 41:30–54. 17

Fernàndez-Garcia, D., Illangasekare, T. H., and Rajaram, H. (2005). Differences in

the scale-dependence of dispersivity estimated from temporal and spatial moments in

chemically and physically heterogeneous porous media. Advances in Water Resources,

28(7):745–759. 142

Finsterle, S., Zhang, Y., Pan, L., Dobson, P., and Oglesby, K. (2013). Microhole arrays

for improved heat mining from enhanced geothermal systems. Geothermics, 47(July

2013):104–115. 116

Fiori, A. and Becker, M. W. (2015). Power law breakthrough curve tailing in a fracture:

The role of advection. Journal of Hydrology, 525:706–710. 15, 50, 140

Frick, U., Alexander, W., Baeyens, B., Bossart, P., Bradbury, M., Bühler, C., Eikenberg,

J., Fierz, T., Heer, W., Hoehn, E., McKinley, I., and Smith, A. (1992). Grimsel test

site - The radionuclide migration experiment: overview of investigations 1985-1990.

Technical report, Nagra. 130

Fu, P., Hao, Y., Walsh, S. D., and Carrigan, C. R. (2016). Thermal Drawdown-Induced

Flow Channeling in Fractured Geothermal Reservoirs. Rock Mechanics and Rock En-

gineering, 49(3):1001–1024. 14, 19, 99, 114, 118

Gelhar, L. W., Welty, C., and Rehfeldt, K. R. (1992). A critical review of data on

field-scale dispersion in aquifers. Water Resources Research, 28(7):1955–1974. 15

Geothermie Schweiz (2019). Unterstützung für tiefe Geothermie. https://geothermie-

schweiz.ch/politik/. 23

Geyer, T., Birk, S., Licha, T., Liedl, R., and Sauter, M. (2007). Multitracer Test Approach

to Characterize Reactive Transport in Karst Aquifers. Ground Water, 45(1):36–45. 51

Ghergut, I., Behrens, H., and Sauter, M. (2014). Tracer-based Quantification of Indi-

vidual Frac Discharge in Single-well Multiple-frac Backflow: Sensitivity Study. Energy

Procedia, 59:235–242. 19

Ghergut, I., Sauter, M., Behrens, H., Licha, T., Mcdermott, C. I., Herfort, M., Rose,

P., Zimmermann, G., Orzol, J., Jung, R., Huenges, E., Kolditz, O., Lodemann, M.,

Fischer, S., Wittig, U., Güthoff, F., and Kühr, M. (2007). Tracer tests evaluating

hydraulic stimulation at deep geothermal reservoirs in Germany. In Proceedings, 31st

Workshop on Geothermal Reservoir Engineering, page 8, Stanford, California, USA.

Stanford University. 21

Ghergut, J., Behrens, H., and Sauter, M. (2016). Petrothermal and aquifer-based EGS in

the Northern-German Sedimentary Basin, investigated by conservative tracers during

162 BIBLIOGRAPHY

single-well injection-flowback and production tests. Geothermics, 63:225–241. 79

Giambastiani, B. M. S., Colombani, N., and Mastrocicco, M. (2013). Limitation of using

heat as a groundwater tracer to define aquifer properties: experiment in a large tank

model. Environmental Earth Sciences, 70(2):719–728. 20, 100

Giardini, D. (2009). Geothermal quake risks must be faced. Nature, 462:848–849. 17

Giertzuch, P.-L., Doetsch, J., Jalali, M., Shakas, A., Schmelzbach, C., and Maurer, H.

(2019). High-Resolution Time-Lapse GPR Monitoring of Saline Tracer Flow in Frac-

tured Rock. In Geophysical Research Abstracts, volume 21, page 13983, Vienna. EGU

General Assembly 2019. 114

Giertzuch, P.-L., Doetsch, J., Kittilä, A., Jalali, M., Schmelzbach, C., Maurer, H., and

Shakas, A. (2018). Monitoring salt tracer transport in granite rock using ground pen-

etrating radar reflection imaging. In 2018 17th International Conference on Ground

Penetrating Radar (GPR), pages 1–6, Rapperswil. IEEE. 34, 69

Gischig, V. S., Doetsch, J., Maurer, H., Krietsch, H., Amann, F., Evans, K. F., Nejati,

M., Jalali, M., Valley, B., Obermann, A. C., Wiemer, S., and Giardini, D. (2018). On

the link between stress field and small-scale hydraulic fracture growth in anisotropic

rock derived from microseismicity. Solid Earth, 9(1):39–61. 65, 102

Gischig, V. S., Giardini, D., Amann, F., Hertrich, M., Krietsch, H., Loew, S., Maurer,

H., Villiger, L., Wiemer, S., Bethmann, F., Brixel, B., Doetsch, J., Doonechaly, N. G.,

Driesner, T., Dutler, N., Evans, K. F., Jalali, M., Jordan, D., Kittilä, A., Ma, X., Meier,

P., Nejati, M., Obermann, A., Plenkers, K., Saar, M. O., Shakas, A., and Valley, B.

(2019). Hydraulic stimulation and fluid circulation experiments in underground labo-

ratories: Stepping up the scale towards engineered geothermal systems. Geomechanics

for Energy and the Environment. 63, 65

Grant, M. A. (2016). Physical performance indicators for HDR/EGS projects. Geother-

mics, 63:2–4. 19, 63, 89, 99, 126

Grimm Lima, M., Vogler, D., Querci, L., Madonna, C., Hattendorf, B., Saar, M. O., and

Kong, X.-Z. (2019). Thermally driven fracture aperture variation in naturally fractured

granites. Geothermal Energy, 7(1):28. 18, 99, 119

Grindrod, P., Edwards, M. S., Higgo, J. J. W., and Williams, G. M. (1996). Analysis

of colloid and tracer breakthrough curves. Journal of Contaminant Hydrology, 21(1-

4):243–253. 51

Gringarten, A. C. and Sauty, J. P. (1975). A theoretical study of heat extraction from

aquifers with uniform regional flow. Journal of Geophysical Research, 80(35):4956–4962.

108, 109

BIBLIOGRAPHY 163

Guihéneuf, N., Bour, O., Boisson, A., Le Borgne, T., Becker, M., Nigon, B., Wajiduddin,

M., Ahmed, S., and Maréchal, J.-C. (2017). Insights about transport mechanisms and

fracture flow channeling from multi-scale observations of tracer dispersion in shallow

fractured crystalline rock. Journal of Contaminant Hydrology, 206:18–33. 14, 20

Guo, B., Fu, P., Hao, Y., Peters, C. A., and Carrigan, C. R. (2016). Thermal drawdown-

induced flow channeling in a single fracture in EGS. Geothermics, 61:46–62. 14, 19, 31,

116, 118

Hadermann, J. and Heer, W. (1996). The Grimsel (Switzerland) migration experiment:

Integrating field experiments, laboratory investigations and modelling. Journal of Con-

taminant Hydrology, 21(1-4):87–100. 14, 50, 130

Haggerty, R., Fleming, S. W., Meigs, L. C., and McKenna, S. A. (2001). Tracer tests in

a fractured dolomite: 2. Analysis of mass transfer in single-well injection-withdrawal

tests. Water Resources Research, 37(5):1129–1142. 19

Haggerty, R. and Gorelick, S. M. (1995). Multiple-rate mass transfer for modeling diffusion

and surface reactions in media with pore-scale heterogeneity. Water Resources Research,

31(10):2383–2400. 50, 140, 142

Haggerty, R., McKenna, S. A., and Meigs, L. C. (2000). On the late-time behavior of

tracer test breakthrough curves. Water Resources Research, 36(12):3467–3479. 15, 142

Hao, C., Shook, G. M., Taimur, M., Dwarakanath, V., Smith, B. R., Muhammad, S.,

Masduki, A., Mario, C. C., and Putra, K. A. D. (2012). Interwell tracer tests to optimize

operating conditions for a surfactant field trial: Design, evaluation and implications.

Society of Petroleum Engineers, 2(1):1185–1202. 41

Harbaugh, A. W., Banta, E. R., Hill, M. C., and Mcdonald, M. G. (2000). Modflow-2000,

the U.S. Geological Survey modular ground-water model: User guide to modularization

concepts and the ground-water flow process. Technical report, U.S. Geological Survey.

140

Häring, M. O., Schanz, U., Ladner, F., and Dyer, B. C. (2008). Characterisation of the

Basel 1 enhanced geothermal system. Geothermics, 37(5):469–495. 62

Harvey, C. F. and Gorelick, S. M. (1995). Temporal Moment-Generating Equations:

Modeling Transport and Mass Transfer in Heterogeneous Aquifers. Water Resources

Research, 31(8):1895–1911. 40

Hawkins, A. J., Becker, M. W., and Tester, J. W. (2018). Inert and Adsorptive Tracer

Tests for Field Measurement of Flow-Wetted Surface Area. Water Resources Research,

54(8):5341–5358. 15, 20, 117

Hawkins, A. J., Becker, M. W., and Tsoflias, G. P. (2017a). Evaluation of inert tracers in

a bedrock fracture using ground penetrating radar and thermal sensors. Geothermics,

164 BIBLIOGRAPHY

67:86–94. 117

Hawkins, A. J., Fox, D. B., Becker, M. W., and Tester, J. W. (2017b). Measurement and

simulation of heat exchange in fractured bedrock using inert and thermally degrading

tracers. Water Resources Research, 53(2):1210–1230. 117

Henri, C. V. and Fernàndez-Garcia, D. (2015). A random walk solution for modeling

solute transport with network reactions and multi-rate mass transfer in heterogeneous

systems: Impact of biofilms. Advances in Water Resources, 86(Part A):119–132. 15, 50

Hertrich, M., Villiger, L., Doetsch, J., Obermann, A., Ma, X., and

Gholizadeh Doonechaly, N. (2019). Design of the seismic monitoring network

for the stimulation experiments in the Bedretto Deep Underground Rock Laboratory.

In 3rd Schatzalp Workshop on Induced Seismicity, Davos, Switzerland. ETH Zürich.

125

Horne, R. (1985). Reservoir engineering aspects of reinjection. Geothermics, 14(2-3):449–

457. 19, 80

Irvine, D. J., Simmons, C. T., Werner, A. D., and Graf, T. (2015). Heat and Solute

Tracers: How Do They Compare in Heterogeneous Aquifers? Groundwater, 53(S1):10–

20. 100

Ishida, T., Fujito, W., Yamashita, H., Naoi, M., Fuji, H., Suzuki, K., and Matsui, H.

(2019). Crack Expansion and Fracturing Mode of Hydraulic Refracturing from Acoustic

Emission Monitoring in a Small-Scale Field Experiment. Rock Mechanics and Rock

Engineering, 52(2):543–553. 17

Jackson, M. J. and Tweeton, D. R. (1996). 3DTOM, Three-dimensional Geophysical

Tomography in Rep. of Invest. 9617. Technical report, US Department of the Interior,

Bureau of Mines, Minneapolis, MN. 74

Jalali, M., Gischig, V., Doetsch, J., Krietsch, H., Amann, F., and Klepikova, M. (2017).

Mechanical, Hydraulic and Seismological Behavior of Crystalline Rock as a Response

to Hydraulic Fracturing at the Grimsel Test Site. In Proceedings of 51st US Rock

Mechanics/Geomechanics Symposium, pages 17–0962, San Francisco, USA. 36, 65, 102

Jalali, M., Gischig, V., Doetsch, J., Näf, R., Krietsch, H., Klepikova, M., Amann, F., and

Giardini, D. (2018a). Transmissivity Changes and Microseismicity Induced by Small-

Scale Hydraulic Fracturing Tests in Crystalline Rock. Geophysical Research Letters,

45(5):2265–2273. 34, 84, 88, 104

Jalali, M., Klepikova, M., Doetsch, J., Krietsch, H., Brixel, B., Dutler, N., Gischig, V., and

Amann, F. (2018b). A Multi-Scale Approach to Identify and Characterize Preferential

Flow Paths in a Fractured Crystalline Rock. In American Rock Mechanics Association

(ARMA), page 9, Seattle, USA. 34, 36, 65, 70, 102, 104, 112

BIBLIOGRAPHY 165

James, S. C. and Chrysikopoulos, C. V. (2003). Analytical solutions for monodisperse and

polydisperse colloid transport in uniform fractures. Colloids and Surfaces A: Physico-

chemical and Engineering Aspects, 226(1-3):101–118. 51

James, S. C. and Chrysikopoulos, C. V. (2011). Monodisperse and polydisperse colloid

transport in water-saturated fractures with various orientations: Gravity effects. Ad-

vances in Water Resources, 34(10):1249–1255. 31, 47

James, S. C., Wang, L., and Chrysikopoulos, C. V. (2018). Modeling colloid transport in

fractures with spatially variable aperture and surface attachment. Journal of Hydrology,

566:735–742. 21

Jiménez, S., Brauchler, R., and Bayer, P. (2013). A new sequential procedure for hydraulic

tomographic inversion. Advances in Water Resources, 62(PA):59–70. 74

Johnston, P. B., Atkinson, T., Barker, J., and Odling, N. (2009). Constraining the

uncertainty in fracture geometry using tracer tests. Hydrogeology Journal, 17(3):527–

539. 15

Kant, M. A., Ammann, J., Rossi, E., Madonna, C., Höser, D., and Rudolf von Rohr,

P. (2017). Thermal properties of Central Aare granite for temperatures up to 500◦C:

Irreversible changes due to thermal crack formation. Geophysical Research Letters,

44:771–776. 116

Kelkar, S., WoldeGabriel, G., and Rehfeldt, K. (2016). Lessons learned from the pioneer-

ing hot dry rock project at Fenton Hill, USA. Geothermics, 63:5–14. 18

Keusen, H., Ganguin, J., Schuler, P., and Buletti, M. (1989). Grimsel Test Site - Geology.

Technical report, NAGRA, Baden. 34, 36, 65, 102, 103, 104, 152

Kirli, M. S. and Fahrioğlu, M. (2019). Sustainable development of Turkey: Deployment

of geothermal resources for carbon capture, utilization, and storage. Energy Sources,

Part A: Recovery, Utilization, and Environmental Effects, 41(14):1739–1751. 62

Kittilä, A., Jalali, M., Somogyvári, M., Evans, K. F., Saar, M. O., and Kong, X.-Z. (2020).

Characterization of the effects of hydraulic stimulation with tracer-based temporal mo-

ment analysis and tomographic inversion. Geothermics. 102, 112

Kittilä, A., Jalali, M. R., Evans, K. F., Willmann, M., Saar, M. O., and Kong, X.-Z.

(2019). Field comparison of DNA-labeled nanoparticle and solute tracer transport in a

fractured crystalline rock. Water Resources Research, 55(8):6577–6595. 67, 68, 69, 75,

81, 89, 110

Klepikova, M., Brixel, B., and Jalali, M. (2020). Transient hydraulic tomography approach

to characterize main flowpaths and their connectivity in fractured media. Advances in

Water Resources, 136:103500. 85

166 BIBLIOGRAPHY

Knapp, R., Chiarappa, M., and Durham, W. (2000). An experimental exploration of the

transport and capture of abiotic colloids in a single fracture. Water Resources Research,

36(11):3139–3149. 31

Kocabas, I. (2005). Geothermal reservoir characterization via thermal injection backflow

and interwell tracer testing. Geothermics, 34(1):27–46. 100

Koelbel, T. and Genter, A. (2017). Enhanced Geothermal Systems: The Soultz-sous-

Forêts Project. In Uyar, T. S., editor, Towards 100% Renewable Energy, pages 243–248.

Springer. 19

Kolditz, O. (1995). Modeling flow and heat-transfer in fractured rocks - Conceptual model

of a 3-D deterministic fracture network. Geothermics, 24(3):451–470. 18, 116

Kolditz, O. and Clauser, C. (1998). Numerical simulation of flow and heat transfer in

fractured crystalline rocks: application to the hot dry rock site in Rosemanowes (U.K.).

Geothermics, 27(1):1–23. 17

Kong, X.-Z., Deuber, C. A., Kittilä, A., Somogyvári, M., Mikutis, G., Bayer, P., Stark,

W. J., and Saar, M. O. (2018). Tomographic Reservoir Imaging with DNA-Labeled

Silica Nanotracers: The First Field Validation. Environmental Science & Technology,

52(23):13681–13689. 30, 31, 33, 56, 64, 72, 74, 124

Kong, X.-Z. and Saar, M. O. (2013). Numerical study of the effects of permeability

heterogeneity on density-driven convective mixing during CO2 dissolution storage. In-

ternational Journal of Greenhouse Gas Control, 19:160–173. 74

Kosakowski, G. (2004). Anomalous transport of colloids and solutes in a shear zone.

Journal of Contaminant Hydrology, 72(1-4):23–46. 31, 130

Kosakowski, G. and Smith, P. (2004). Modelling the transport of Solutes and Colloids

in a Water Conducting Shear Zone at the Grimsel Test Site. Technical Report 04-01,

National Cooperative for the Disposal of Radioactive Waste, NAGRA. 15

Kowalsky, M. B., Finsterle, S., Williams, K. H., Murray, C., Commer, M., Newcomer,

D., Englert, A., Steefel, C. I., and Hubbard, S. S. (2012). On parameterization of the

inverse problem for estimating aquifer properties using tracer data. Water Resources

Research, 48(6). 30

Krietsch, H., Doetsch, J., Dutler, N., Jalali, M., Gischig, V., Loew, S., and Amann,

F. (2018). Comprehensive geological dataset describing a crystalline rock mass for

hydraulic stimulation experiments. Scientific Data, 5:12. 34, 35, 36, 37, 38, 52, 65, 66,

70, 84, 85, 89, 102, 103, 114, 119, 126, 136, 149, 150, 152, 153

Krietsch, H., Gischig, V., Evans, K., Doetsch, J., Dutler, N. O., Valley, B., and Amann,

F. (2019). Stress Measurements for an In Situ Stimulation Experiment in Crystalline

Rock: Integration of Induced Seismicity, Stress Relief and Hydraulic Methods. Rock

BIBLIOGRAPHY 167

Mechanics and Rock Engineering, 52(2):517–542. 102, 114

Kumari, W., Ranjith, P., Perera, M., and Chen, B. (2018). Experimental investigation

of quenching effect on mechanical, microstructural and flow characteristics of reser-

voir rocks: Thermal stimulation method for geothermal energy extraction. Journal of

Petroleum Science and Engineering, 162:419–433. 99, 100

Kurosawa, S., James, S. C., Yui, M., and Ibaraki, M. (2006). Model analysis of the colloid

and radionuclide retardation experiment at the Grimsel Test Site. Journal of Colloid

and Interface Science, 298(1):467–475. 14

Lange, J., Schuetz, T., Gregoire, C., Elsässer, D., Schulz, R., Passeport, E., and

Tournebize, J. (2011). Multi-tracer experiments to characterise contaminant mitigation

capacities for different types of artificial wetlands. International Journal of Environ-

mental and Analytical Chemistry, 91(7-8):768–785. 30

Lauwerier, H. A. (1955). The transport of heat in an oil layer caused by the injection of

hot fluid. Applied Scientific Research, 5(2-3):145–150. 108

Le Borgne, T., Bour, O., Riley, M. S., Gouze, P., Pezard, P. A., Belghoul, A., Lods,

G., Le Provost, R., Greswell, R. B., Ellis, P. A., Isakov, E., and Last, B. J. (2007).

Comparison of alternative methodologies for identifying and characterizing preferential

flow paths in heterogeneous aquifers. Journal of Hydrology, 345(3-4):134–148. 14

Lee, I., Tester, J. W., and You, F. (2019). Systems analysis, design, and optimization of

geothermal energy systems for power production and polygeneration: State-of-the-art

and future challenges. Renewable and Sustainable Energy Reviews, 109:551–577. 17,

18, 62

Lei, Q., Latham, J.-P., and Tsang, C.-F. (2017). The use of discrete fracture networks

for modelling coupled geomechanical and hydrological behaviour of fractured rocks.

Computers and Geotechnics, 85:151–176. 14

Leibundgut, C., Maloszewski, P., and Külls, C. (2009). Tracers in hydrology. John Wiley

& Sons, Ltd, Chichester, UK. 19, 20, 21, 30

Leresche, S. (2018). Characterization of Heterogeneous Fractured Reservoirs Using Solute

Tracer Tests: Application to the Underground Laboratory of Grimsel (MSc Thesis).

ETH Zurich. 68, 81

Leube, P. C., Nowak, W., and Schneider, G. (2012). Temporal moments revisited: Why

there is no better way for physically based model reduction in time. Water Resources

Research, 48(11):W11527. 31, 40, 69, 106

Liao, R., Yang, P., Wu, W., Luo, D., and Yang, D. (2018). A DNA Tracer System for

Hydrological Environment Investigations. Environ. Sci. Technol, 52:3. 30, 33

168 BIBLIOGRAPHY

Linde, N., Finsterle, S., and Hubbard, S. (2006). Inversion of tracer test data using

tomographic constraints. Water Resources Research, 42(4):1–15. 100

Long, J. C. S. and Billaux, D. M. (1987). From field data to fracture network modeling:

An example incorporating spatial structure. Water Resources Research, 23(7):1201–

1216. 14

Lorenz, M. O. (1905). Methods of Measuring the Concentration of Wealth. Publications

of the American Statistical Association, 9(70):209–219. 40, 41

Luhmann, A. J., Covington, M. D., Myre, J. M., Perne, M., Jones, S. W., Alexander Jr,

E. C., and Saar, M. O. (2015). Thermal damping and retardation in karst conduits.

Hydrol. Earth Syst. Sci, 19:137–157. 118

Luo, F., Xu, R.-N., and Jiang, P.-X. (2013). Numerical Study of the Influence of In-

jection/Production Well Perforation Location On Co2-EGS System. Energy Procedia,

37:6636–6643. 17

Luo, J., Cirpka, O. A., Dentz, M., and Carrera, J. (2008). Temporal moments for transport

with mass transfer described by an arbitrary memory function in heterogeneous media.

Water Resources Research, 44(1):W01502. 39

Ma, R., Zheng, C., Zachara, J. M., and Tonkin, M. (2012). Utility of bromide and heat

tracers for aquifer characterization affected by highly transient flow conditions. Water

Resources Research, 48(8). 100

Mahler, B. J., Winkler, M., Bennett, P., and Hillis, D. M. (1998). DNA-labeled clay: A

sensitive new method for tracing particle transport. Geology, 26(9):831–834. 30

Majer, E. L. and Doe, T. W. (1986). Studying hydrofractures by high frequency seismic

monitoring. International Journal of Rock Mechanics and Mining Sciences & Geome-

chanics Abstracts, 23(3):185–199. 17

Maloszewski, P. and Zuber, A. (1993). Tracer experiments in fractured rocks: Matrix

diffusion and the validity of models. Water Resources Research, 29(8):2723–2735. 50

Mandal, A., Basantaray, A. K., Chandroth, A., and Mishra, U. (2019). Integrated geo-

physical investigation to map shallow surface alteration/fracture zones of Atri and

Tarabalo hot springs, Odisha, India. Geothermics, 77:24–33. 64

Marschall, P. and Lunati, I. (2006). GAM - Gas Migration Experiments in a Hetero-

geneous Shear Zone of the Grimsel Test Site. Technical Report NTB 03-11, Nagra,

Wettingen. 19, 39, 73, 130

Marschall, P., Vomvoris, S., Bossart, P., Correa, N., Guyonnet, D., Heinemann-Glutsch,

B., Hurtig, E., Johns, R., Kohl, T., Kuhlmann, U., Müri, F., and Rybach, L. (1995).

Developments in Hydrotesting, Fluid Logging and Combined Salt/Heat Tracer Exper-

BIBLIOGRAPHY 169

iments in the BK Site (Phase III). Technical Report NTB 93-47, Nagra, Wettingen.

100, 114, 116, 118, 125

McCord, J., Reiter, M., and Phillips, F. (1992). Heat-flow data suggest large ground-

water fluxes through Fruitland coals of the northern San Juan basin, Colorado-New

Mexico. Geology, 20(5):419–422. 100

McKay, L., Sanford, W., and Strong, J. (2000). Field-Scale Migration of Colloidal Tracers

in a Fractured Shale Saprolite. Ground Water, 38(1):139–147. 31

Mikutis, G., Deuber, C. A., Schmid, L., Kittilä, A., Lobsiger, N., Puddu, M., Asgeirs-

son, D. O., Grass, R. N., Saar, M. O., and Stark, W. J. (2018). Silica-Encapsulated

DNA-Based Tracers for Aquifer Characterization. Environmental Science & Technol-

ogy, 52(21):12142–12152. 21, 31, 33, 34, 51, 56, 124

Mock, J. E., Tester, J. W., and Wright, P. M. (1997). Geothermal Energy from the Earth:

Its Potential Impact as an Environmentally Sustainable Resource. Annual Review of

Energy and the Environment, 22:305–356. 16, 62

Mora, C. A., Paunescu, D., Grass, R. N., and Stark, W. J. (2015). Silica particles with

encapsulated DNA as trophic tracers. Molecular Ecology Resources, 15(2):231–241. 30

Moreno, L. and Tsang, C. F. (1991). Multiple-Peak Response to Tracer Injection Tests

in Single Fractures: A Numerical Study. Water Resources Research, 27(8):2143–2150.

79, 81

Moreno, L., Tsang, Y. W., Tsang, C. F., Hale, F. V., and Neretnieks, I. (1988). Flow and

tracer transport in a single fracture: A stochastic model and its relation to some field

observations. Water Resources Research, 24(12):2033–2048. 14, 15

Möri, A., Alexander, W. R., Geckeis, H., Hauser, W., Schäfer, T., Eikenberg, J., Fierz,

T., Degueldre, C., and Missana, T. (2003). The colloid and radionuclide retardation

experiment at the Grimsel Test Site: Influence of bentonite colloids on radionuclide mi-

gration in a fractured rock. Colloids and Surfaces A: Physicochemical and Engineering

Aspects, 217(1-3):33–47. 130

Mosthaf, K., Brauns, B., Fjordbøge, A. S., Rohde, M. M., Kerrn-Jespersen, H., Bjerg,

P. L., Binning, P. J., and Broholm, M. M. (2018). Conceptualization of flow and trans-

port in a limestone aquifer by multiple dedicated hydraulic and tracer tests. Journal

of Hydrology, 561:532–546. 50

Mueller, M. H., Alaoui, A., Kuells, C., Leistert, H., Meusburger, K., Stumpp, C., Weiler,

M., and Alewell, C. (2014). Tracking water pathways in steep hillslopes by δ18O depth

profiles of soil water. Journal of Hydrology, 519:340–352. 30

Nami, P., Schellschmidt, R., Schindler, M., and Tischner, T. (2008). Chemical stimu-

lation operations for reservoir development of the deep crystalline HDR/EGS system

170 BIBLIOGRAPHY

at Soultz-Sous-Forêts (France). In Proceedings of the 32nd workshop on geothermal

reservoir engineering, pages SGP–TR–185, Stanford. 18

Nelson, P., Rachiele, R., Remer, J., and Carlsson, H. (1981). Water inflow into boreholes

during the Stripa heater experiments. Technical Report LBL-12574, Lawrence Berkeley

Laboratory, California. 113

Neretnieks, I., Eriksen, T., and Tähtinen, P. (1982). Tracer movement in a single fissure

in granitic rock: Some experimental results and their interpretation. Water Resources

Research, 18(4):849–858. 14, 50

Neretnieks, I. and Rasmuson, A. (1984). An approach to modelling radionuclide migration

in a medium with strongly varying velocity and block sizes along the flow path. Water

Resources Research, 20(12):1823–1836. 14

Nieto, I. M., Martín, A. F., Blázquez, C. S., Aguilera, D. G., García, P. C., Vasco, E. F.,

and García, J. C. (2019). Use of 3D electrical resistivity tomography to improve the

design of low enthalpy geothermal systems. Geothermics, 79:1–13. 64

Niitsuma, H. (1989). Fracture mechanics design and development of HDR reservoirs -

Concept and results of the Γ-project, Tohoku University, Japan. International Journal

of Rock Mechanics and Mining Sciences & Geomechanics Abstracts, 26(3-4):169–175.

63

Nottebohm, M., Licha, T., and Sauter, M. (2012). Tracer design for tracking thermal

fronts in geothermal reservoirs. Geothermics, 43:37–44. 21

Olasolo, P., Juárez, M. C., Morales, M. P., Damico, S., and Liarte, I. A. (2016). Enhanced

geothermal systems (EGS): A review. Renewable and Sustainable Energy Reviews,

56:133–144. 16, 63, 99

Pandey, S., Chaudhuri, A., and Kelkar, S. (2017). A coupled thermo-hydro-mechanical

modeling of fracture aperture alteration and reservoir deformation during heat extrac-

tion from a geothermal reservoir. Geothermics, 65:17–31. 19, 99, 118

Pang, L., Robson, B., Farkas, K., Mcgill, E., Varsani, A., Gillot, L., Li, J., and Abraham,

P. (2017). Tracking effluent discharges in undisturbed stony soil and alluvial gravel

aquifer using synthetic DNA tracers. Science of the Total Environment, 592:144–152.

30, 33, 50

Paunescu, D., Puddu, M., Soellner, J. O. B., Stoessel, P. R., and Grass, R. N. (2013).

Reversible DNA encapsulation in silica to produce ROS-resistant and heat-resistant

synthetic DNA ’fossils’. Nature Protocols, 8(12):2440–2448. 21, 33

Pedretti, D., Fernàndez-Garcia, D., Bolster, D., and Sanchez-Vila, X. (2013). On the

formation of breakthrough curves tailing during convergent flow tracer tests in three-

dimensional heterogeneous aquifers. Water Resources Research, 49(7):4157–4173. 20,

BIBLIOGRAPHY 171

30

Pfingsten, W. and Soler, J. M. (2003). Modelling of nonreactive tracer dipole tests in a

shear zone at the Grimsel test site. Journal of Contaminant Hydrology, 61:387–403.

130

Phirani, J., Roy, S., and Pant, H. (2018). Predicting stagnant pore volume in porous

media using temporal moments of tracer breakthrough curves. Journal of Petroleum

Science and Engineering, 165:640–646. 50

Pioro, I., Duffey, R. B., Kirillov, P. L., Pioro, R., Zvorykin, A., and Machrafi, R. (2019).

Current Status and Future Developments in Nuclear-Power Industry of the World.

Journal of Nuclear Engineering and Radiation Science, 5(2):UNSP 024001. 16

Pollack, H. N., Hurter, S. J., and Johnson, J. R. (1993). Heat flow from the Earth’s

interior: Analysis of the global data set. Reviews of Geophysics, 31(3):267. 15

Rasmuson, A. and Neretnieks, I. (1986). Radionuclide Transport in Fast Channels in

Crystalline Rock. Water Resources Research, 22(8):1247–1256. 14, 31

Rau, G. C., Andersen, M. S., and Acworth, R. I. (2012). Experimental investigation of

the thermal dispersivity term and its significance in the heat transport equation for

flow in sediments. Water Resources Research, 48:W03511. 100

Redden, G., Stone, M., Wright, K. E., Mattson, E., Palmer, C. D., Rollins, H., Harrup,

M., and Hull, L. C. (2010). Tracers for characterizing enhanced geothermal systems. In

Proceedings, 35th Workshop on Geothermal Reservoir Engineering, page 7, Stanford,

California, USA. Stanford University. 21

Reimus, P., Caporuscio, F., Marina, O., and Janney, D. (2020). Field demonstration

of the combined use of thermally-degrading and cation-exchanging tracers to predict

thermal drawdown in a geothermal reservoir. Geothermics, 83. 18

Robinson, B. A. and Tester, J. W. (1984). Dispersed fluid flow in fractured reservoirs:

An analysis of tracer-determined residence time distributions. Journal of Geophysical

Research, 89(B12):10374–10384. 14, 39, 76, 79, 106, 107, 110, 118, 125

Rose, P., Mella, M., and McCullough, J. (2006). A comparison of hydraulic stimulation

experiments at the soultz, france and coso, california engineered geothermal systemS. In

31st Workshop on Geothermal Reservoir Engineering, pages 60–64, Stanford, California.

63

Roux, S., Plouraboué, F., and Hulin, J. (1998). Tracer Dispersion in Rough Open Cracks.

Transport in Porous Media, 32(1):97–116. 14

Rummel, F. and Kappelmeyer, O. (1983). The Falkenberg Geothermal Frac-Project:

Concepts and Experimental Results. In Nemat-Nasser, S., Abé, H., and Hirakawa, S.,

172 BIBLIOGRAPHY

editors, Hydraulic fracturing and geothermal energy, pages 59–74, Dordrecht, Nether-

lands. Springer. 63

Rutqvist, J., Börgesson, L., Chijimatsu, M., Nguyen, T., Jing, L., Noorishad, J., and

Tsang, C.-F. (2001). Coupled thermo-hydro-mechanical analysis of a heater test in

fractured rock and bentonite at Kamaishi Mine - comparison of field results to predic-

tions of four finite element codes. International Journal of Rock Mechanics & Mining

Sciences, 38(1):129–142. 99, 113

Saar, M. O. (2011). Review: Geothermal heat as a tracer of large-scale groundwater flow

and as a means to determine permeability fields. Hydrogeology Journal, 19(1):31–52.

16, 20, 30, 88, 99

Sabir, I. H., Haldorsen, S., Torgersen, J., Aleström, P., Sabir, I. H., and Haldorsen, S.

(1999). DNA tracers with information capacity and high detection sensitivity tested in

groundwater studies. Hydrogeology Journal, 7:264–272. 46

Sahimi, M. (2011). Flow and transport in porous media and fractured rock : From classical

methods to modern approaches. Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim,

Germany, 2nd, rev. edition. 74

Salamon, P., Fernàndez-Garcia, D., and Gómez-Hernández, J. J. (2006a). A review and

numerical assessment of the random walk particle tracking method. Journal of Con-

taminant Hydrology, 87(3-4):277–305. 14, 142

Salamon, P., Fernàndez-Garcia, D., and Gómez-Hernández, J. J. (2006b). Modeling mass

transfer processes using random walk particle tracking. Water Resources Research,

42(11):W11417. 140, 142

Sanchez-Vila, X., Guadagnini, A., and Carrera, J. (2006). Representative hydraulic con-

ductivities in saturated groundwater flow. Reviews of Geophysics, 44(3):RG3002. 88

Sanjuan, B., Pinault, J.-L., Brach, M., Braibant, G., Crouzet, C., Foucher, J.-C., Gautier,

A., and Touzelet, S. (2006). Tracer testing of the geothermal heat exchanger at Soultz-

sous-Forêts (France) between 2000 and 2005. Geothermics, 35(5-6):622–653. 19, 64

Sarris, T. S., Close, M., and Abraham, P. (2018). Using solute and heat tracers for aquifer

characterization in a strongly heterogeneous alluvial aquifer. Journal of Hydrology,

558:55–71. 100

Schill, E., Genter, A., Cuenot, N., and Kohl, T. (2017). Hydraulic performance history at

the Soultz EGS reservoirs from stimulation and long-term circulation tests. Geother-

mics, 70:110–124. 19

Shakas, A., Linde, N., Baron, L., Selker, J., Gerard, M.-F., Lavenant, N., Bour, O., and

Le Borgne, T. (2017). Neutrally buoyant tracers in hydrogeophysics: Field demonstra-

tion in fractured rock. Geophysical Research Letters, 44(8):3663–3671. 81

BIBLIOGRAPHY 173

Sharma, A. N., Luo, D., and Todd Walter, M. (2012). Hydrological Tracers Us-

ing Nanobiotechnology: Proof of Concept. Environmental Science & Technology,

46(16):8928–8936. 30

Shook, G. M. (2003). A Simple, Fast Method of Estimating Fractured Reservoir Geometry

from Tracer Tests. Geothermal Resources Council Transactions, 27:407–411. 41, 72,

108

Shook, G. M. and Forsmann, J. H. (2005). Tracer Interpretation Using Temporal Moments

on a Spreadsheet. Technical report, Idaho National Laboratory. 32, 39, 40, 41, 49, 64,

69, 72, 75, 106, 107, 108

Shook, G. M., Sharma, A., and Pope, G. A. (2017). Early-Time Analysis of Tracers

for Use in Enhanced-Oil-Recovery Flood Optimization. SPE Reservoir Evaluation &

Engineering, 20(1):30–41. 39, 45

Shook, G. M. and Suzuki, A. (2017). Use of tracers and temperature to estimate fracture

surface area for EGS reservoirs. Geothermics, 67:40–47. 30, 39, 41, 63, 69, 71, 106, 107,

108, 109, 117

Siirila-Woodburn, E. R., Sanchez-Vila, X., and Fernàndez-Garcia, D. (2015). On the

formation of multiple local peaks in breakthrough curves. Water Resources Research,

51(4):2128–2152. 15, 81

Sirivithayapakorn, S. and Keller, A. (2003). Transport of colloids in saturated porous

media: A pore-scale observation of the size exclusion effect and colloid acceleration.

Water Resources Research, 39(4):12. 46, 52

Somogyvári, M. and Bayer, P. (2017). Field validation of thermal tracer tomography for

reconstruction of aquifer heterogeneity. Water Resources Research, 53(6):5070–5084.

73, 87, 100

Somogyvári, M., Bayer, P., and Brauchler, R. (2016). Travel-time-based thermal tracer

tomography. Hydrology and Earth System Sciences, 20(5):1885–1901. 30, 64, 72, 73,

74, 100

Somogyvári, M., Jalali, M., Jimenez Parras, S., and Bayer, P. (2017). Synthetic fracture

network characterization with transdimensional inversion. Water Resources Research,

53(6):5104–5123. 20

Stoll, M., Huber, F., Trumm, M., Enzmann, F., Meinel, D., Wenka, A., Schill, E., and

Schäfer, T. (2019). Experimental and numerical investigations on the effect of fracture

geometry and fracture aperture distribution on flow and solute transport in natural

fractures. Journal of Contaminant Hydrology, 221:82–97. 14, 81

Streltsova, T. D. (1976). Hydrodynamics of groundwater flow in a fractured formation.

Water Resources Research, 12(3):405–414. 13

174 BIBLIOGRAPHY

Stumm, W. (1993). Aquatic colloids as chemical reactants: surface structure and re-

activity. Colloids and Surfaces A: Physicochemical and Engineering Aspects, 73:1–18.

31

Sun, J., Liu, Q., and Duan, Y. (2018). Effects of evaporator pinch point temperature dif-

ference on thermo-economic performance of geothermal organic Rankine cycle systems.

Geothermics, 75:249–258. 18

Swiss Federal Office of Energy (2018). Energy Strategy 2050.

https://www.bfe.admin.ch/bfe/en/home/policy/energy-strategy-2050.html. 23

Taron, J. and Elsworth, D. (2009). Thermal-hydrologic-mechanical-chemical processes

in the evolution of engineered geothermal reservoirs. International Journal of Rock

Mechanics and Mining Sciences, 46(5):855–864. 18, 19, 99

Tester, J. W., Anderson, B. J., Batchelor, A. S., Blackwell, D. D., DiPippo, R., Drake,

E. M., Garnish, J., Livesay, B., Moore, M. C., Nichols, K., Petty, S., Toksöz, M. N.,

and Veatch, R. W. (2006). The Future of Geothermal Energy: impact of enhanced

geothermal systems (EGS) on the United States in the 21st century. Technical Report

INL/EXT-06-11746, Massachusetts Institute of Technology, Cambridge, MA, USA. 16,

17, 18, 19, 62, 64, 99

Tiab, D. and Donaldson, E. C. (2015). Petrophysics : theory and practice of measuring

reservoir rock and fluid transport properties. Gulf Professional Publishing, an imprint

of Elsevier, Amsterdam, fourth ed. edition. 74, 88

Tompson, A. F. B. (1993). Numerical simulation of chemical migration in physically and

chemically heterogeneous porous media. Water Resources Research, 29(11):3709–3726.

15

Tuykhova, A. R. and Willmann, M. (2016). Conservative transport upscaling based on

information of connectivity. Water Resources Research, 52:14. 15, 50, 140

Vasco, D. W. and Datta-Gupta, A. (1999). Asymptotic solutions for solute transport: A

formalism for tracer tomography. Water Resources Research, 35(1):1–16. 72

Vilks, P. and Bachinski, D. B. (1996). Colloid and suspended particle migration experi-

ments in granite fracture. Journal of Contaminant Hydrology, 21:269–279. 21, 50, 51,

55

Vilks, P., Frost, L. H., and Bachinski, D. B. (1997). Field-scale colloid migration exper-

iments in a granite fracture. Journal of Contaminant Hydrology, 26(1-4):203–214. 31,

51

Vogt, C., Kosack, C., and Marquart, G. (2012). Stochastic inversion of the tracer experi-

ment of the enhanced geothermal system demonstration reservoir in Soultz-sous-Forêts

- Revealing pathways and estimating permeability distribution. Geothermics, 42:1–12.

BIBLIOGRAPHY 175

30, 63

Wang, C., McNew, C. P., Lyon, S. W., Walter, M. T., Volkman, T. H., Abramson, N.,

Sengupta, A., Wang, Y., Meira Neto, A. A., Pangle, L., Troch, P. A., Kim, M., Harman,

C., and Dahlke, H. E. (2019). Particle tracer transport in a sloping soil lysimeter under

periodic, steady state conditions. Journal of Hydrology, 569:61–76. 43

Watanabe, N., Hirano, N., and Tsuchiya, N. (2009). Diversity of channeling flow in

heterogeneous aperture distribution inferred from integrated experimental-numerical

analysis on flow through shear fracture in granite. Journal of Geophysical Research,

114(B4):B04208. 14, 31

WEC (2013). World Energy Resources: 2013 Survey. Technical report, World Energy

Council, London. 16, 62

Wenning, Q., Madonna, C., de Haller, A., and Burg, J. P. (2018). Permeability and

seismic velocity anisotropy across a ductile-brittle fault zone in crystalline rock. Solid

Earth, 9:683–698. 36, 116

Wiemer, S., Kraft, T., and Landtwing, D. (2015). Seismic risk. In Hirschberg, S., Wiemer,

S., and Burgherr, P., editors, Energy from the earth, chapter 6.2, pages 263–295. vdf

Hochschulverlag AG. 17, 18

Willmann, M., Carrera, J., and Sánchez-Vila, X. (2008). Transport upscaling in heteroge-

neous aquifers: What physical parameters control memory functions? Water Resources

Research, 44(12):W12437. 14, 15, 50

Willmann, M., Carrera, J., Sanchez-Vila, X., Silva, O., and Dentz, M. (2010). Coupling

of mass transfer and reactive transport for nonlinear reactions in heterogeneous media.

Water Resources Research, 46(7):W07512. 142

Winick, J., Siega, F., Addison, S., Richardson, I., Mountain, B., and Barry, B. (2015).

Coupled Iodine-125 and 2NSA Reservoir Tracer Testing at the Rotokawa Geothermal

Field, New Zealand. In Proceedings World Geothermal Congress 2015, page 10, Mel-

bourne, Australia. 19, 41, 63

Wu, X., Pope, G. A., Shook, G. M., and Srinivasan, S. (2008). Prediction of enthalpy pro-

duction from fractured geothermal reservoirs using partitioning tracers. International

Journal of Heat and Mass Transfer, 51(5-6):1453–1466. 18, 21

Yasuhara, H., Kinoshita, N., Ohfuji, H., Lee, D. S., Nakashima, S., and Kishida, K. (2011).

Temporal alteration of fracture permeability in granite under hydrothermal conditions

and its interpretation by coupled chemo-mechanical model. Applied Geochemistry,

26(12):2074–2088. 18, 19, 119

Yasukawa, K. (2019). Issues Around Geothermal Energy and Society in Japan. In

Manzella, A., Allansdottir, A., and Pellizzone, A., editors, Geothermal Energy and

176 BIBLIOGRAPHY

Society, pages 179–191. Springer International Publishing. 16

Younger, P. L. (2014). Missing a trick in geothermal exploration. Nature Geoscience,

7:479–480. 16

Zhang, W., Tang, X. Y., Weisbrod, N., Zhao, P., and Reid, B. J. (2015). A coupled

field study of subsurface fracture flow and colloid transport. Journal of Hydrology,

524:476–488. 50

Zheng, B., Xu, J., Ni, T., and Li, M. (2015). Geothermal energy utilization trends from

a technological paradigm perspective. Renewable Energy, 77:430–441. 16

Zheng, Q., Dickson, S. E., and Guo, Y. (2009). Differential transport and dispersion of

colloids relative to solutes in single fractures. Journal of Colloid and Interface Science,

339(1):140–151. 31, 50, 51, 55

Zhou, Q., Liu, H. H., Molz, F. J., Zhang, Y., and Bodvarsson, G. S. (2007). Field-scale

effective matrix diffusion coefficient for fractured rock: Results from literature survey.

Journal of Contaminant Hydrology, 93(1-4):161–187. 50

Ziegler, M., Loew, S., and Moore, J. R. (2013). Distribution and inferred age of exfoliation

joints in the Aar Granite of the central Swiss Alps and relationship to Quaternary

landscape evolution. Geomorphology, 201:344–362. 65, 102

Zinn, B. and Harvey, C. F. (2003). When good statistical models of aquifer heterogeneity

go bad: A comparison of flow, dispersion, and mass transfer in connected and multi-

variate Gaussian hydraulic conductivity fields. Water Resources Research, 39(3). 15,

50

tracer-based characterization of a stimulation … · 2021. 2. 22. · examining the tracer...

Documents