• Tidak ada hasil yang ditemukan

Decoding the state of stress and fluid pathways along the Andean Southern Volcanic Zone

N/A
N/A
Protected

Academic year: 2025

Membagikan "Decoding the state of stress and fluid pathways along the Andean Southern Volcanic Zone"

Copied!
16
0
0

Teks penuh

(1)

Decoding the state of stress and fl uid pathways along the Andean Southern Volcanic Zone

Nicolás Pérez-Estay 1✉, Javiera Ruz-Ginouves2,3, Pamela Pérez-Flores 4, Gerd Sielfeld5,6, Tomás Roquer7,8,9 & José Cembrano1,3

Decoding means decrypting a hidden message. Here, the encrypted messages are the state of stress,fluid pathways, and volcano tectonic processes occurring in volcanoes of the Andean Southern Volcanic Zone (SVZ). To decode these messages, we use earthquake focal mechanisms, fault slip data, and a Monte Carlo simulation that predicts potential pathways for magmatic and hydrothermal fluids. From this analysis, we propose that SVZ volcanoes have three end-member stress patterns: (i) Stress-A, a strike-slip regime coupled with the regional far-field tectonic stress; (ii) Stress-B, an extensional regime that may be promoted by volcanic edifice loading and upward pressure due to magma inflation occurring within the upper brittle-crust; and (iii) Stress-C, a local and transient fluid-driven stress rotated ~90 degrees from Stress-A. Notoriously, Stress-C pattern was observed in most volcanoes with historical eruptions. We propose that volcanoes presenting Stress-B are attractive geother- mal targets, while Stress-C could be used as a predicting signal for impending eruptions.

https://doi.org/10.1038/s43247-023-01040-9 OPEN

1Andean Geothermal Center of Excellence (CEGA), Facultad de Ciencias Físicas y Matemáticas, Universidad de Chile, Santiago, Chile.2Geology Department, University of Otago, Dunedin, New Zealand.3Departamento de Ingeniería Estructural y Geotécnica, Ponticia Universidad Católica de Chile, Santiago, Chile.

4Consultoría e Investigación Geológico Ambiental Ldta., Huasco, Chile.5Cirrus SpA, Santiago, Chile.6School of Environment, University of Auckland, Auckland, New Zealand.7Departamento de Ingeniería de Minería, Ponticia Universidad Católica de Chile, Santiago, Chile.8Departamento de Ingeniería de Minas, Facultad de Ciencias Físicas y Matemáticas, Universidad de Chile, Santiago, Chile.9Advanced Mining Technology Center (AMTC), Universidad de Chile, Santiago, Chile. email:[email protected]

1

1234567890():,;

(2)

M

agmatic and hydrothermalfluidflow through the brittle crust is governed by fracture mechanics laws that account for the formation of open-fractures (i.e., extension or hybrid fractures)1–3, and which depends on the balance between the state of stress and the total fluid-pressure4. Around volcanoes, it is crucial to understand these variables to address processes that control fluid circulation, and thereby potential volcanic hazards and geothermal resources57. The state of stress around volcanoes has been studied in different tectonic settings6,8, from which oblique convergence margins represent one of the most recorded tectonic settings worldwide, including long-lived volcanic arcs such as the Sunda Arc (Indonesia), Northeast Honshu (Japan), Central America volcanic arc, Aleutian-Alaska volcanic arc, and the Andes (South America)914. In this study, we focus on understanding the state of stress and the conditions required to generate extension frac- tures in the Southern Volcanic Zone of the Andes (SVZ), where oblique convergence between the Nazca and South-American plates generates a volcanic arc that is closely linked to partitioned continental deformation14,15.

Here, we defined the regional and local stress state for sixteen volcanic complexes of the SVZ using seismological (short-term) and structural geology (long-term) data (Fig. 1). Then, we ana- lyzed dilatational tendency and fluid-pressure conditions that can trigger extension fractures using a novel Monte Carlo simulation, to reveal the preferred pathways of magmatic and hydrothermal fluids. We then discuss (i) the reliability of our

results, (ii) the similarity between short- and long-term states of stress, (iii) propose a holistic conceptual model for the volcano- tectonic processes occurring in volcanoes within the SVZ, and (iv) discuss how our results can be extended to other volcanic arcs within oblique subduction settings.

Revealing the state of stress, dilatational tendency and fluid- pressure to trigger extension fractures in a regional context and in individual volcanic complexes. Here we present a con- solidated compilation and careful selection of published focal mechanisms (short-term“instantaneous”data) and fault slip data (FSD, long-term data) from the SVZ16–36 used to unravel the regional and local state of stress around volcanoes. The double- couple component of focal mechanisms and FSD represent shear fractures6,37,38, and can be used to identify the orientation of the principal stress axes and the stress ratio (ɸ)37,38 by using the same physical and mathematical framework, as the Wallace-Bott hypothesis39,40. In this work, the term“state of stress”or“state stress”is used to refer to the orientation of the principal stress axes and the stress ratio (ɸ).

Present regional stress state was calculated from short-term/

instantaneous focal mechanism data18,19,22,23,41,42 for five regions. These were defined by considering the latitudinal volcano-tectonic segmentation of the SVZ14,15: Maipo (33°–34°S), Maule (34°–37°), Araucanía (38°–40°), Chiloé (42°–43°S) and Cisnes (44°–45°S) shown in Fig.1.

10 30 20

40 60 50

70 80 90

σ axis plunge 0

10 20

30 40

50 60

70 80

90

σ axis plunge

0

20

30

40

50

60

70

80

90

σ axis pl

unge Puyuhuapi Puyuhuapi Los Pescados

Los Pescados Fui

Tolhuaca Villarica

Laguna Maule Laguna Maule

Mentolat

Callaqui Copahue-Caviahue

Copahue-Caviahue Puyehue-Cordón Caulle San Pedro-Tatara

Planchon

Sierra Nevada

Tinguiririca

Cisnes Region Chiloé Region

Araucania Region

Maule Region

Maipo Region

Regional Stress Volcano (Short-Term) Volcano (Long-Term)

20 10 40 30 60 50 80 70 90 0 10

20 30

40 50

60 70

80 90

σ axis plunge

0 10

20 30

40 50

60

0 70 80

90 σ axi

s plunge

Normal Reverse

Strike-slip

Strike-slip Normal Normal Strike-slip

Strike-slip Reverse

Reverse Strike-slip

σ axis plunge

0 75 150 km

Liquiñe-Ofqui Fault System

Andean Transverse Faults Inferred fault

Inferred fault Dextral strike slip fault

Sinistral-reverse strike slip fault

Volcano Analyzed volcano

Legend

N

35°S

40°S

45°S

75°W 70°W

Planchón-Peteroa

Tinguiririca

Callaqui Copahue-Caviahue Tolhuaca

Villarica

Sierra Nevada

Fui V.G.

Puyuhue- Cordón Caulle

Puyuhuapi V.G.

Mentolat

Los Pescados V.G.

San Pedro - Tatara Laguna del Maule

Nevados de Chillán

Carrán Los Venados

Nevados de Chillán

Carrán Los Venados

NP

AP

SAP

74 mm/yr

73.5 mm/yr

73 mm/yr

20 mm/yr

Chiloé Region

Cisne Region Araucanía

Region Maule Region Maipo Region

Volcanic Gap

Southern Volcanic Zone

b) a)

c)

Fig. 1 Analyzed volcanoes of the SVZ. aMap of the Andean Southern Volcanic Zone showing theve regions dened in this study and the analyzed volcanoes (red triangles). NP Nazca Plate, AP Antarctic Plate, SAP South American Plate. The background layer is the ERSI Shaded Relief freely available through QuickMapService of QGis software.bKaverina plot of volcanoes and analyzed regions.cStress/strainelds as represented in the Kaverina plot.

2

(3)

For individual volcanoes, we calculated both the short- and long- term stress state from focal mechanisms and FSD16–21,24–36. The raw seismological and structural database can be found in the Supplementary Note S1, Supplementary Data 1, Supplementary Table 1, and in an online repository declared in Data availability.

Earthquake magnitudes range from Mw 1.0 to 4.1, and fault slip data include a wide range of fault sizes, from millimeters to meters.

Each component of the database employed here has already been collected, discussed, and published in previous works16–21,24–36, but here they are analyzed holistically to understand the tectono- magmatic processes occurring in the SVZ rather than in individual volcanoes only. To build our database we only considered volcanoes with existing seismic and/or field data within a conservative radius of 2 kilometers from the volcanic edifice or located in major faults spatially associated with volcanoes. The data shortage (due to inaccessibility to some volcanicflanks) and short duration of seismic temporal records generate different spatial and temporal biases, which are detailed individually for each volcanic complex in Supplementary Discussion 1, and later is discussed how these biases affect our results and conclusions.

In total, we conducted seven short-term1621and thirteen long- term24–36analyses, accounting for 23% of the volcanoes within the SVZ, that include 16 different volcanic complexes (i.e., 12 polygenetic volcanoes and 4 monogenetic volcanicfields; Figs.1, 2). The selected volcanoes are broadly representative of the SVZ tectono-magmatic context since they are distributed along the arc’s length and volcano-tectonic segmentation. They also represent a wide range of lava compositions, edifice volumes, and fault settings including the most relevant fault systems within the intra-arc region, namely the Liquiñe-Ofqui Fault System and Andean Transverse Faults (Fig.1).

The likely distribution of principal stress axes and stress ratio (ɸ) were calculated using the Multiple Inverse Method (MIM)37,38. We chose this method because it is adapted to analyze focal mechanisms as well as FSD, and allows for the identification of heterogeneous or polyphase stress states, which is crucial when analyzing long-term data24,29,31. From the MIM results, we selected the most representative stress solution for each volcano and evaluated the compatibility of the data with this selected solution (see Methods section). Compatibility was evaluated by quantifying the amount of data with a misfit angle (i.e. angle between the slip-direction of the data and the predicted-slip direction) lower than 30°31,37,43. If less than 70%

of the data were compatible with the most representative solution, we calculated a secondary stress solution, excluding the data that were compatible with the primary stress solution. The results of principal stress directions and stress ratio are presented in Fig.2.

Potential fluid-pathways in volcanoes can be inferred by identifying the planes showing the highest values of dilatation tendency, or the lowest values of total fluid-pressure required to trigger hydrofractures. Both scenarios should promote the genera- tion of extension fractures. To calculate the dilatation tendency and total fluid-pressure, we used the principal stress orientation and stress ratio previously calculated for each volcano, and the non- Andersonian equations (Methods). Unknown variables for these equations are the tensile strength (T), the magnitude of differential stress (τ¼σ1σ3), and the plunge of the principal stresses (φ123), which makes the calculation of dilatational tendency and totalfluid-pressure uncertain. Therefore, we created a Monte Carlo simulation of 1000 cases for each volcano to test the most likely values for dilatation tendency and total fluid-pressure. These parameters depend on six variables (z;φ13;ϕ;τ;T- depth, plunge of σ1 and σ3, stress ratio, differential stress, and tensile strength, respectively). All six-variables varied randomly within a reasonable range in either linear or probabilistic distributions. As a result, we

obtained 1000 simulations that can be re-organized to show statistical parameters as percentiles, median, or standard deviations (see Fig.3a). This novel Monte-Carlo simulation reveals the likely orientation of extension fractures and thus, of geofluid circulation pathways at individual volcanoes (Figs.3b,4). For readers interested in reproducing or expanding upon this analysis in other volcanic arc settings, the codes developed here can be found in Supplementary Note 2, Supplementary Folder“Code”, and the link declared in Code availability statement, where a test case is also provided.

Results and discussions

Regional and local states of stress in the Southern Volcanic Zone. Our results from the short-term analysis show that the regional stress state in the SVZ is governed by a sub-horizontalσ1

trending between N30°E and N60°E, and a sub-horizontal σ3

trending between N05°E and N40°W (Fig. 2). Apart from the Maipo Region, all regional stress state can be classified as strike- slip normal according to the Kaverina plot44,45 (Fig. 1b). The outlier compressional solution obtained for the Maipo Region (33°–34°S) is, however, in agreement with previous works doc- umenting a compressional regime at these latitudes, linked to the Chile-Argentina flat-slab (28°–33°S)14,22. In turn, the sub- horizontal ~NE-trending σ1 for remaining regions suits an intra-arc strike-slip regime associated with strain/deformation partitioning described by many authors14,23,31,35.

For individual volcanoes, a wide range of stress classifications is evident in the Kaverina plot ranging from normal to strike-slip reverse (Fig. 1b). Overall, we determined four states of stress patterns:

Type A: Strike-slip state of stress with a sub-horizontalσ1

andσ3, consistent with the far-field regional tectonic stress imposed by the oblique convergence. This pattern is commonly observed in monogenetic volcanic fields such as Puyuhuapi and Los Pescados (Fig.2).

Type B: Local extensional state of stress defined by a verticalσ1and a horizontalσ3that is not necessarily equal to the regional far-field tectonicσ3 trend. This is the case for several polygenetic volcanoes with large composite edifices such as Copahue-Caviahue (long-term), Tolhuaca, and San Pedro-Tatara.

Type C: Strike-slip state of stress whereσ1is horizontally rotated between 50–90° from Type-A, indicating a NW- trendingσ1. This pattern is observed as the primary stress state solution for the Villarrica volcano, and as secondary stress state solutions for Tolhuaca, Copahue-Caviahue (long-term), Planchón-Peteroa, and Nevados de Chillán volcanic complexes.

Hybrid Type (A+B, or B+C): A single state of stress con- taining two of the aforementioned stress patterns/types, for example, Type A and B, whereσ3is equal to the regional far- field direction, and σ1varies in a NE-oriented sub-vertical plane; observable, for example, in Copahue-Caviahue or Mentolat (short-term). Another example is the primary stress state of Puyehue-Cordón Caulle, whereσ1varies in a NW-striking sub-vertical plane containing Type B and C.

The outlier solutions with a single case are Carrán Los Venados and Callaqui.

Orientation of dikes and veins as a proxy for magma and hydrothermal fluid migration plumbing system. The results of the total fluid-pressure required to trigger extension fractures are presented in Fig.3b, where eight volcanoes were selected to show the expected dike and vein orientations for the four, previously

3

(4)

described, volcanic stress patterns. We focus on totalfluid-pressure conditions to induce extension fractures, rather than dilatational tendency, as the former provides a more accurate prediction of dike and vein orientation (discussed in the next section).

Results of the Monte Carlo simulation of Type A pattern show two main orientations where fluid circulation should be

concentrated, as depicted by the blue bands (Type A results in Fig.3b). Normally, these bands are limited to the range 50–80°

in the NE-quadrant, which is consistent with the occurrence of NE-striking dikes and veins with dip angles >40–50°. In contrast, Type B stress pattern promotefluid circulation in dikes and veins with all possible strikes (0° to 360°) and with dips >20–30°. Type

σ₁ σ₃

Regional Short-term (from focal mechanisms) Region Principal stress axes Maipo Region

33°S 70.8°W 69.9°W

34°S

Sol. 62%

n=39

Period

01-2017 to 03-2019

07-2016 to 05-2017 03-2015 to

06-2016

12-2004 to 11-2005

Short-term (from focal mechanisms) Volcano Principal stress axes

Mentolat [44.7°S, 73.1°W]

Period

12-2004 to 01-2007 07-2016 to

05-2017

07-2016 to 05-2017 03-2015 to

06-2016 04-2011 to

10-2014

Long-term (from fault slip data)

Volcano Principal stress

axes

Sol. 81%

Sol. 48%

Sol. 61%

Sol. 23%

Max age of

FSD

Sol. 68%

Puyuhuapi Sol. 52%

Sol. 46%

01-1985 to 01-2015 Maule Region

34°S 71.5°W 70.0°W

37°S n=7

[44.3°S,72.5°W]

Los Pescados [45.4°S,73.0°W]

Araucanía Region

38°S 72.3°W 70.9°W

40°S n=37

Sol. 86%

=0.36

Sol. 57%

=0.81

Sol.100%

=0.45

Sol. 65%

=0.64 Chiloé Region

42°S 73.0°W 72.0°W

43°S n=8 Cisne Region

44°S 73.5°W 72.3°W

45°S n=20

Los Pescados [45.4°S,73.0°W]

Puyuhuapi [44.3°S,72.5°W]

Villarica [39.4°S,71.9°W]

n=18 Laguna del

Maule [36.1°S,70.5°W]

Sol. 76%

=0.39 Sol. 71%

=0.13

Sol. 87%

=0.83 Sol. 73%

=0.65 Sol. 94%

=0.38

n=31 Tolhuaca [38.3°S,71.6°W]

n=16

Sierra Nevada [38.6°S,71.6°W]

n=37 Copahue Caviahue [37.8°S,71.2°W]

n=31 Nevados de

Chillán [36.9°S,71.4°W]

n=25 Laguna del

Maule [36.1°S,70.5°W]

n=12

=0.19

=0.75

=0.29

=0.15

=0.40

Sol. 30%

=0.92

Sol. 24%

=0.42

n=20

0 1

0 1

0 1

0 1

0 1

Stress ratio

Qa Qa Ple

Mio Plio- Qa Mio

Plio- Ple

Long-term (from fault slip data)

Volcano Principal stress

axes

Sol. 44%

n=43 Tinguiririca

[34.8S,70.3°W]

=0.53

n=24

Sol. 23%

=0.39 Plio-

Qa

% of data compatible with the most representative solution

n= Amount of data

Name of volcano

[Lat, Lon]

Median Age or period of

time that are represented by the availabe data Histogram of fhi

Location

Stress distribution obtained from MIM results and the Kamb method

Sol. 41%

n=34 Planchón

Peteroa [35.2°S,70.6°W]

n=20

=0.37

Sol. 24%

=0.75 Qa

Sol. 42%

Sol. 24%

n=289 San Pedro

Tatara [36.0°S,70.8°W]

n=167

=0.27 Qa

Carrán Los Venados [40.3°S,72.1°W]

Sol. 61%

=0.26

Sol. 29%

=0.09

Sol. 75%

=0.51

0 0.5 1

n=12

Plio- Qa 12-2017 to

03-2018 Copahue

Caviahue [37.8°S,71.2°W]

Sol. 79%

=0.57 Stress ratio

Stress ratio

Stress ratio

n=56

n=15

n=8

n=8

n=25

n=36

n=12

n=99

n=16

Max age of

FSD

21

17 40

22

18

15

16

17

18

18

19

23

24

27

26

28

29

30

31

33

34

35 n° reference

or area

During eruption (from focal mechanisms)

06-2011 to 01-2012 Puyehue

Cordón Caulle [40.6°S,70.1°W]

n=9

Sol. 61%

=0.32

Sol. 22%

=0.89 n=23

20

Callaqui [37.9°S,71.4°W]

Sol. 54%

=0.98 Plio-

Ple n=13

Fui [39.9°S,71.9°W]

Sol. 61%

=0.27 Jr n=13

32

25

Fig. 2 State of stress of regions and volcanoes of the SVZ.Thegure shows the density distribution ofσ1- andσ3- directions colored in red and blue contours, respectively. We include the medianɸvalue, the percentage of data compatible with the most representative stress setting (Methods), and the amount of datan(focal mechanisms or FSD). Some volcanoes present two stress states separated in two rows. The specic location of short- and long- term data for each volcano can be found in the folderraw dataexplained in Supplementary Note 1. The time window of short-term data represents the period of recorded seismicity by local seismic networks. The relative age of fault slip data was taken from literature16–21,24–36, and mostly ranges from Miocene to Quaternary. They are considered maximum faulting ages because they are dened from the rock unit(s) they crosscut. However, the internal consistency between the geometry and kinematics of faults cross-cutting Pliocene to Pleistocene rock units with those from older plutonic and volcanic units strongly suggests that the fault populations analyzed represent post-Pliocene deformation. This interpretation is further supported by the stability of the tectonic regime governing the SVZ, as well as the magnitude and direction of the convergence vector, which has remained nearly constant since the mid-Miocene57,81.

4

(5)

Fig. 3 Results of the Monte Carlo simulation. aGraphical explanation of the Monte Carlo simulation, and the percentile calculations, for the case of Copahue-Caviahue Volcanic Complex. Top left shows an example of one Monte Carlo simulation, randomly chosen. At the middle, 1000 Monte Carlo simulations are organized by the number of the simulation. Top right shows the results of the 1000 simulations reorganized statistically in percentiles. The totaluid-pressure that triggers extension fractures is presented normalized by the lithostatic pressure or the vertical stress. Thus, this variable is the fraction of the lithostatic pressure required foruid-pressure to trigger extension fractures in a specic plane of strike (x-axis) and dip (y-axis).bResults of Monte Carlo simulation separated by the four types of stress patterns identied in the SVZ. The volume of volcanic edices, lava composition, and the orientation ofσ2axes also are shown. The totaluid-pressure results are displayed in the space of strike (x-axis) and dip (y-axis), where each coordinate (x,y) represents a specic plane with strike/dip orientation. We show the 10- and 90- percentile results of the Monte Carlo simulation because they allow the observation of a representative range that includes 80% of the results of the Monte Carlo simulation. 5- 25- 75- and 95- percentiles, mean, and standard deviation can be found in Supplementary Note 3, in theSupplementary Figurefolder. Synthetic fractures are modeled by randomly selecting their orientation (strike/dip) considering the lower totaluid-pressure values. The purpose of showcasing these synthetic fractures is to illustrate the results of the Monte Carlo simulation in terms of how the orientation of extension fractures should be according with the models. Comparison between the models and actual data is shown in Fig.4and in the discussion section of the reliability of results.

5

(6)

C concentrates its most-likely conditions for extension fracture development in NW-striking dikes and veins. Lastly, some examples of Hybrid Type are consistent with Type A, but extension fractures can occur in a wider range of orientations (~100°) falling in the NE-quadrant.

The spatial and temporally biases of database and reliability of results. The reliability of the results presented here may be affected by several biases due to data limitations. Firstly, there is a

spatial bias due to inaccessibility and limited outcrops that restricts the structural database. Secondly, there is a temporal bias resulting from the limited temporal seismic networks in the region (see Supplementary Discussion 1, for spatial and temporal bias of each volcano). These biases limit the comprehensive definition of stress states related to volcanoes, which can be mostly resolved when instrumentation records several eruptions cycles. Even then, a new different stress signature could emerge as a result of tectonic variability or a stress disequilibrium. Despite these issues, here we have partially revealed the state of stress of

10-percentile

median

0.0 0.2 0.4 0.6 0.8 1.0

Dilatational Tendency [MPa/MPa]

0.0 0.3 0.6 0.9 1.2 1.5 1.8 Puyehue - Cordón Caulle

90-percentile median 10-percentile 90-percentile median 10-percentile

300

100 200

0

Strike [°]

80 60 40 20 0

Dip [°]

80 60 40 20 0

Dip [°]

80 60 40 20 0

Dip [°]

90-percentile

80 60 40 20 0

Dip [°]

10-percentile

80 60 40 20 0

Dip [°]

median

80 60 40 20 0

Dip [°]

90-percentile

80 60 40 20 0

Dip [°]

80 60 40 20 0

Dip [°]

80 60 40 20 0

Dip [°]

80 60 40 20 0

Dip [°]

80 60 40 20 0

Dip [°]

80 60 40 20 0

Dip [°]

Puyuhuapi Tolhuaca San Pedro - Tatara

Veins (Hydrothermal process) Dikes (Magmatic process)

300

100 200

0

Strike [°]

300

100 200

0

Strike [°]

300

100 200

0

Strike [°]

10-percentile

median

90-percentile

10-percentile

median

90-percentile 90-percentile

median 10-percentile 90-percentile median 10-percentile

b) a)

c)

n = 59 n = 44 n = 150 n = 18

Total fluid pressure normalized “pt/σv” [MPa/MPa]

Fig. 4 Reliability of Monte Carlo simulation. aRose diagram of veins and dikes considered in this work to test Monte Carlo simulation in four volcanoes of the SVZ: Puyuhuapi volcanic group; Tolhuaca; San Pedro-Tatara and Puyehue-Cordón Caulle. Note that these data only represent a small proportion of dikes around a volcano, because near 90% of dikes are arrested or deected at depth, considering the small percentage of feeder dikes observed worldwide7,49.bDilatational tendency and (c) totaluid pressure required to trigger extension fractures, as revealed by the Monte Carlo approach in the same cases as (a). Black dots represent vein and dike orientation measured in theeld. At Puyuhuapi and Tolhuaca volcanoes, the radius of black dots is related to veins width. Vein and dike orientations were obtained from previous works in Tolhuaca32, San Pedro-Tatara28and Caulle82volcanoes, and represent Quaternary processes. Data from Puyuhuapi were obtained directly by the authors. All Puyuhuapi veins crosscut the Holocene basaltic rocks of Puyuhuapi volcanic group.

6

(7)

volcanoes in the SVZ, which shows the existence of three end- member stress patterns repeated throughout the region (Type-A, B, and C). We acknowledge that the spatio-temporal biases in our database do not allow us to conclude that these are the only stress patterns in the region, and that patterns not reported here might emerge as more data becomes available.

The reliability of the Monte Carlo simulations is discussed and evaluated by comparing the results withfield data (strike/dip) of dikes and hydrothermal veins documented within and around four volcanoes (Fig.4). We observed that the orientations of dikes and veins (black circles in Fig.4b, c) coincide with the preferred dilatational planes and the orientations where the total fluid- pressure required to trigger extension fractures is the lowest.

Compatibility between the models andfield data was evaluated by quantifying the number of dikes and veins that are constrained between a range of one standard deviation from the highest value of dilatational tendency and the lowest totalfluid-pressure value.

Median models contain 65% and 77% offield data for dilatational tendency and total fluid-pressure models. The 10-percentile models contain 84% and 80% of data from dikes and veins for dilatational tendency and total fluid-pressure respectively, whereas the 90-percentile models contain 38% and 72%, respectively. These percentages suggest thatfluid-pressure models are better predictors of dike and vein occurrence than dilatational tendency. Moreover, there are populations of dikes and veins that are not consistent with our modeling (i.e., the remaining percentage of data) which can be explained by other local geologic and tectonic factors not considered in the present work.

These may include (1) the local rotation of principal stresses due to dike emplacement46–48; (2) local mechanical heterogeneities that can arrest or redirect propagation pathways7,49; (3) spatial and temporal variation in the local stress at the vicinity of a magma chamber, or ahead of the dike tip7,50. However, considering that most of the field data fit with the total fluid- pressure models (previous testing percentage greater than 70%), the state of stress calculated from shear fractures printed in focal mechanisms and fault-slip data coupled with the Monte Carlo method presented here seem to be reliablefirst-order predictors of fluid pathways within the upper crust in volcanic regions, at least in cases where extension fractures are associated with the documented state of stress.

Whilst some authors have argued that hydrothermal veins are commonly restricted to fault zones51,52, the vein data presented here are recorded within Holocene volcanic edifices and are not restricted to fault zones. Therefore, they are most likely dominated by a volcano-driven stress effect rather than tectonic-driven stress occurring within a fault. As shown in Fig.4, the presented Monte Carlo simulation can predict vein orienta- tions when they are related to the documented stress state in the volcanoes. In the case that veins are restricted to fault-zones, our method should be tested considering the specific stress state of the fault, which is beyond the scope of this work.

Comparison between long-term and short-term state of stress.

In the four cases where evidence of both long- and short-term stress states can be compared, three volcanic displayed striking similarities (Los Pescados, Puyuhuapi, and Laguna del Maule).

Although a subtle rotation in the orientation of the principal stress axes and some variability in the ɸ histograms were noticeable in Laguna del Maule, the orientation of the principal stress axes was nearly identical in all three cases (Fig.2).

The Copahue-Caviahue volcanic complex (Fig.2) is the only exception where long- and short-term solutions differ. A reasonable explanation of these differences could be the temporal variability of the stress state in the volcanic complex during an

eruption46,47,53, or in the subduction cycle29,54–56. This implies short-term data with earthquakes Mw<4 reveal only a part of the entire deformation process. During an eruption or a megathrust subduction earthquake, significant changes in the stress state should vary fault-kinematics producing afingerprint in the long- term that is not necessarily comparable to the short-term seismological observation, that in the Copahue-Caviahue data- base, is restricted to the subduction interseismic period. An alternative explanation would be local stress rotations47 in the walls of the long-lived fissure volcanism reported for the bulk Callaqui-Copahue-Caviahue-Mandolehue volcanic chain26.

The long-term analysis shows that seven out 13 cases present two stress state solutions, suggesting a complex stress distribution around volcanoes. The most significant solution of long-term cases (i.e, the primary stress solution that represents the most % of the data) is compatible with 41% up to 81% of the FSD, whereas the secondary stress solution is subordinate, and compatible with 23% up to 30% of the FSD. The deformation pattern during and after the 2011 Cordón Caulle eruption confirms the above-mentioned stress distribution complexity, showing that extensional and the opposite compressional state of stress coexist for about 8 months (short-term local variability), that could be attributed to a compressional environment coupled with extensional mechanisms triggered by the evacuation of magma from the chamber during the eruption21. The absence of a secondary stress solution in most of the short-term cases suggests that the datasets used to define the short-term stress state do not reveal the entire deformation process occurring around volcanoes (which can be as long as hundreds or thousands of years). This observation underscores the significance of long-term data as a valuable source of information for comprehending the volcano- tectonic processes that govern volcanic arcs.

The geological significance of the different stress patterns in the SVZ. The geological processes accounting for the previously described stress patterns are discussed below, and summarized in Fig. 5. The main finding is that, unlike Hybrid Types, stress patterns A, B, and C characterize end-member state of stress.

The strike-slip Type A stress pattern represents volcanoes that are coupled with the far-field regional tectonic stress setting of the interseismic deformation on the overriding South American plate14,54,57 (Fig. 5). This stress pattern is expected to control volcano-tectonic processes in monogenetic groups, as Puyuhuapi volcanic group; around volcano edifice flanks58, influencing the geometry and location of parasitic cones26; or within kinematically-coupled volcanoes14. In terms of magma chambers, this stress pattern indicates that any hypothetical shallow chamber does not significantly alter the regional tectonic stress, thus, there are at least three possibilities: (i) the absence of a shallow magma chamber located in the brittle crust (where data are located), (ii) the stress effect of a magma chamber is enclosed in another stress state solution (i.e., the primary or the secondary state of stress, in volcanoes that have more than one stress state solution); or (iii) the spatial and/or temporal biases of the data do not capture the potential magma chamber-driven stress state. In the case of the small volcanic groups like Puyuhuapi or Los Pescados, an absence of a shallow magma chamber in the brittle crust is plausible, given their basaltic nature and the fact they have erupted less than 2 km3of lava. In the case of Laguna del Maule, the primary Type-A pattern should reflect a local stress dominated by the far-field tectonic stress, as shown by the dextral NE-striking Troncoso fault16, whereas the secondary stress state solution in the long-term results (Fig. 2) may reflect the inflation of a sill at ~5 km depth59, fitting with the second possibility. In the case of Planchón-Peteroa, MT surveys reveal

7

(8)

the absence of a magma chamber immediately beneath the volcanic edifice60, and NNE to NE-striking normal fault and fractures, suggest the existence of extension fractures aligned with σ1of a Type-A stress pattern25. Similarly, Nevados de Chillán is spatially associated predominantly with NE-striking dextral faults and dikes aligned with σ1 of a Type-A stress pattern29. Normal faults, fractures and dikes aligned with σ1 of a Type-A stress pattern suggest that fluid circulation in the brittle crust at Nevados de Chillan and Planchón-Peteroa volcanic complexes should be controlled, at least in part, by this stress pattern.

Indeed, if fluid circulation only depends on that, it should be restricted to subvertical NE-striking dikes and veins (e.g, Puyuhuapi in Fig. 5). However, we cannot discard that the

migration of magmatic and hydrothermal fluids also occurs within preexisting high permeability zones, such as faults, which are not necessarily NE-striking. Overall, the Type-A stress pattern is attributed to the far-field tectonic stress resulting from oblique subduction, should promote the occurrence of NE-striking dikes and veins, and it is associated with scenarios such as the absence of a shallow magma chamber (e.g., Puyuhuapi or Planchón- Peteroa) or the presence of a magma chamber in the brittle crust, as indicated by other state of stress solution (e.g., Laguna del Maule).

The extensional Type-B stress pattern was observed in the largest volcanic edifices, such as San Pedro-Tatara and Puyehue- Cordón Caulle. This stress pattern can be explained by (1) a

Type-A Type-B Type-C

Oblique Subduction N

Z E

~NE-trending interseismic compression

~NE-striking Dikes and veins

N

Z E Z

H (N or E) 80

60 40 20

Dip [°]

0 10-percentile

80 60 40 20

Dip [°]

0 10-percentile

80 60 40 20

Dip [°]

0 10-percentile

300

100 200

0 Strike [°] 0 100 Strike [°]200 300 0 100 Strike [°]200 300

e.g. Primary stress state solution of Tolhuaca e.g. Primary stress state

solution of Puyuhuapi

e.g. Secondary stress state solution of Planchón-Peteroa

Hypothesis n°2:

Hypothesis n°1:

Magma passing through NE-striking dike

Local σ 90° rotated at the dike walls Representative example of the stress pattern:

Monte Carlo simulation results:

Predicted extension fracture attitude (synthetic examples):

Volcano-tectonic processes for each stress pattern:

n = 500

Laccoliths Sills

Diapir

Brittle Ductile

σ₁ σ₃

Dip [°]

Strike [°]

0 to 360°-striking Dikes and veins Dip [°]

Strike [°]

~NW-striking Dikes and veins Dip [°]

Strike [°]

σ₁

σ₁ σ₃

σ₁ σ₃

n = 500 n = 500

Fluids passing through ancient NW-striking

discontinuities Transient ~NE-trending

extension 0.0

0.6 1.2 1.8

Fig. 5 Summary of the volcano-tectonic process dominating the SVZ, as discussed in the text.Each of the three end-member stress patterns identied (Types A, B, and C) represent different volcano-tectonic processes. Note that Type-C stress can be explained by two volcano-tectonic processes, hypothesis 1 or hypothesis 2.

8

(9)

progressive increase in the magnitude of the vertical stress due to edifice growth during volcano building61, (2) increase in the magnitude of vertical stress above a magma storage zone (sills, laccoliths or chambers7,62), or alternatively (3) local stress rotation in dilatation jogs or horse-tail structures5. These three processes can coexist. The existence of this stress pattern may suggest the presence of a shallow magma chamber in the brittle crust, or a magma reservoir located in the ductile crust, either of which may be capable of modifying and controlling the stress at shallower levels where focal mechanisms and FSD occur. Several magnetotelluric (MT) surveys conducted in the SVZ have documented electrically conductive volumes in the upper crust associated with magma storage regions, including San Pedro- Tatara, Tiguiririca, Copahue-Caviahue and Tolhuaca volcanoes60,6365, all of which display a Type B stress signature.

Additionally, from surface deformation modeling using InSAR, an inflating sill at 5 km depth was documented in the Laguna del Maule volcanic complex59 and several pulses of upper crustal magma injection were interpreted for Cordón Caulle66, both presenting Type-B pattern. Lastly, Mentolat volcano has geobaro- metric evidence indicating a magma storage depth as shallow as 5 km67, which allows us to state that in all seven cases presenting the Type-B stress pattern there is evidence of magma stored in the upper brittle crust. Type-B stress pattern promotes dike and vein formation in radial directions (0 to 360°, see Fig. 5). This is suggested by Monte Carlo models and confirmed withfield data from Tolhuaca, as seen in Fig. 4. Notoriously, Type B stress pattern have the lowest total fluid-pressure values compared to the other stress patterns (Fig.3) suggesting that this stress pattern promotes the formation of extension fractures more easily. Thus, representing a positive feedback process that promotes magma injection. In other words, magma storage should increase the magnitude of vertical stress, which in turn facilitates later magma injection in the form of dikes and inclined sheets and generating more magma storage and eruptions. This incresed rate of magma injection, together with the plausible magma storage zone located in the upper crust, would likely create a heated volume in the crust near Type-B volcanoes, and thus, the vicinity of these volcanoes could represent ideal geothermal targets.

Finally, the strike-slip Type-C pattern is an enigmatic signature present in several volcanoes that have had eruptions within the past 300 years (Fig.6). In fact, considering the volcanoes analyzed here, six out of nine volcanoes with recorded historical eruptions have a Type C signature, either as a primary or secondary stress solution. The exceptions are Mentolat, Carrán Los Venados and Callaqui. Previous authors have documented a similar Type-C pattern in Nevados de Chillán and Puyehue-Cordón Caulle29,34 and have hypothesized that: (i) it is linked to the co-seismic period of megathrust earthquakes where the elastic rebound generated in the overlying plate allows for extension and/or (ii) the reactivation of NW-striking basement faults. However, the former seems to be small to generate the stress change required to trigger an eruptive cycle by itself55,68. The second hypothesis is also difficult to reconcile in all cases, since NW-striking basement faults have not been identified in many of the volcanoes with Type-C signature. Therefore, there is no unequivocal geological process that can account for this signature.

We hypothesize that at least two main processes can explain this stress pattern: (1) reactivation of preexisting NW-striking faults as parallel-to-the-fault hydrofractures promoted by the ascent of overpressurized deep magmatic or hydrothermalfluids, and (2) local ninety-degree rotation of principal stress axes during intrusion and inflation of a NE-striking dike at its walls47(Fig.5).

Ourfirst hypothesis is similar to those previously suggested, but here we highlighted the reactivation of the NW-structures as hydrofractures (extension fracture) that require an overpressure

fluid. The first hypothesis implies transient perturbations of magmatic or hydrothermal reservoirs creating a large-scale NW- striking channel of fluids capable of modifying the local stress.

The second hypothesis suggests that ongoing magmatic injections aligned with the far-field tectonic regional stress axes exert a fluid-driven force against wall rocks. When subduction mega- thrust events occur, hypothesis (1) is plausible considering the permeability enhancement in inherited faults after dynamic alteration of seismic waves69 and volumetric dilatation of the crust70. Otherwise, we propose that the long-term secondary Type-C stress pattern documented in Planchón-Peteroa, Nevados de Chillán, Copahue-Caviahue, and Tolhuaca volcanoes is coherent with the second hypothesis since, in all these cases, Type-C is rotated 90-degrees from the primary stress state solution (see Fig.2). On the contrary, at Villarrica and Puyehue- Cordón-Caulle volcanic complexes, thefirst hypothesisfits better with the geological context, considering previously defined NW- faults associated with both volcanoes14,18,34, and seismicity elongated in a NW-trending direction (see expanded discussion in Supplementary Discussion 2). Interestingly, both hypotheses used to explain the Type-C pattern implies pressure made byfluid circulation through open-fractures (i.e, extension or hybrid fracture), which leads us to consider that Type-C is a potential eruption footprint. Of the total number of eruptions that occurred in the last 300 years, 49% have been recorded in volcanoes with a Type-C signature, and this percentage increases up to 96% if we only consider the volcanoes analyzed in this work (Fig.6). This demonstrates the importance of the Type-C stress pattern for understanding and assessing volcanic hazards in the SVZ. Although the volcanoes associated with this stress pattern are the most active along the SVZ (Fig. 6) and appear to have a positive correlation with megathrust earthquakes, eruptions from these volcanoes cannot be exclusively explained by megathrust earthquakes, since most of the eruptions occurred within interseismic periods. Therefore, subduction earthquakes, along the megathrust, should be viewed as a disturbance for a system ready to erupt rather than as the trigger for eruptions in Type-C volcanoes.

Summarizing the fundamental findings of this investigation, the three pivotal end-member stress patterns proposed here are:

Type-A, the regional far-field tectonic stress; Type-B, the local, volcanic load and magmatic-dominated stress state; and Type-C, the local and probably transient stress driven by fluid passing through open-fractures. We propose a classification of volcanoes based on the stress state constrained in this work, which is presented in Fig.7. If volcanoes exhibit a Hybrid-type stress state, or have two or more distinguishable stress state solutions (two rows in Fig.2) representing different stress pattern/types, they are classified as A+B, A+C, B+C or A+B+C (details in Supplementary Table 2). There is no direct correlation between the stress patterns and the geochemical composition of volcanoes, the volume of volcanoes, nor the type of volcanoes (i.e., stratovolcanoes, calderas), as can be observed in Fig. 3 and Supplementary Table 3. However, minor eruptive centers, commonly of basaltic composition, mostly exhibit a Type-A signature. Importantly, the classification presented in Fig.7could be improved in the future, as new data emerge, reducing the spatial and temporal biases of our database. For example, petrological data71,72, crustal deformation revealed by InSAR73, long-period seismic signals74, and MT-surveys75within Villarrica volcano provide evidence of a shallow magma chamber in the upper ~5 km that should imply the Type-B pattern, but the stress inversion conducted here does not reveal this pattern in Villarica (Figs.2,7). The limited short-term seismic data restricted to ~one year of measurements, with earthquakes confined to the eastern flank of the volcano within 7 and 9 km depth, does not fully

9

(10)

reveal the full complex stress pattern of this stratovolcano.

Similarly, other long-lived stratovolcanoes and calderas with variable compositions, eruption styles, and chemical evolution might have a complex stress history that could not be completely represented in our database, or in our classification.

The contribution of this work in the context of other volcanic arcs. Our results can be compared with the state of stress docu- mented in other compressional and strike-slip volcanic arcs worldwide, related to oblique subduction (e.g., Aleutians-Alaska, Northeast Honshu, Java arcs). A remarkable similarity with the Aleutian-Alaska arc is that the most common stress state recor- ded in volcanoes is consistent with the far-field tectonic stress arising from oblique convergence and slip partitioning, named here as Type-A (i.e., 40 out of 49 in Aleutian-Alaska volcanoes12, and 10 out of 16, exposed here). The type-A pattern is also the

most recurrent in Quaternary stratovolcanoes and Mio-Pliocene calderas in the Northeast Honshu arc, and in different types of volcanic cones along the Java volcanic arc9,10. This suggests that the far-field tectonic stress exerts a first-order control on the fracturing process occurring in compressional and strike-slip volcanic arcs. Type-B pattern has been suggested in NE Honshu in volcanoes with recurrent sill emplacement6,10, and has been documented as the background state of stress in the Redoubt volcano in the Aleutians53. Additionally, local horizontal rotation of the far-field tectonic stress, similar to Type-C pattern, has been documented in volcanoes in the Aleutians12,46,53, and in Mt.

Argopuro within the Java arc9, but not clearly in NE Honshu.

Hence, the identified stress patterns likely reflect volcano-tectonic processes occurring in both strike-slip- and compressional- dominated volcanic arcs, extending beyond the SVZ.

Lastly, more than one stress state has been observed in seven out of 49 volcanoes of the Aleutians12, but none out of the 23 in

Fig. 6 Contrasting the stress patterns with historical eruptions.Plot showing the last 300 years of documented eruptions and subduction megathrust earthquakes in the SVZ. Volcanoes are organized by latitude (y-axis) and color-coded by the type of stress setting (A, B, and C). The volcanoes are subclassied as A+B; A+C; or B+C, when they have a stress state that included two end-member stress pattern (i.e., hybrid stress type) or when they have two stresselds containing two different end-member types. Further details of the stress state classication are shown in Fig.7, and Supplementary Table 2. Eruption dates were obtained from the Global Volcanism Program83. The slip segment of each megathrust earthquake was approximated from Watt et al.84. The area affected by elastic and viscoelastic deformation was approximated to be two times longer than the slip segment; the period of elastic effects was set to 3 years70, and viscoelastic effects were constrained to 35 years considering the discussion of Bonali et al.55.

10

(11)

the Java volcanic arc9. In contrast, at least nine of the 16 volcanoes analyzed in the SVZ have been found to exhibit more than one stress state (Figs.2,7). This difference can be explained by the stress inversion performed in this study, which has been shown to better describe the complex and variable tectonic setting around volcanoes, compared to traditional morphometric tools.

Traditional morphometric analyses rely on surface strain expressions at the volcano-scale, limiting interpretation to horizontal principal stress axes. The high number of volcanoes with more than one stress state (>50% in SVZ) highlight the importance of the formal inversion of the stress around volcanoes, which could include all three stress patterns in the same volcano (e.g., Copahue-Caviahue, Fig.7).

Furthermore, evidence has shown that the local state of stress can rotate during eruption cycles, as seen along the Aleutian- Alaska, the Japanese arcs and the Andean Northern Volcanic Zone46,47,53. Here we have demonstrated that the local rotation of the stress state around volcanoes is also recorded in long-term structural data, and most importantly, is present in most of the volcanoes with historical eruptions (Fig. 6). Thus, defining the background state of stress for individual volcanoes is crucial to detecting changes in the stress state that can be correlated with coming eruptions.

Conclusions

From our results, we conclude that volcanic stress state can be decoded by combining an analysis of fault slip data and earth- quake focal mechanisms. The coupling between long-term and short-term signatures allows to extension of the understanding of the stress state in volcanoes without decades of instrumental seismic data, usingfield structural geology studies. The long-term approach should be important considering that volcanoes that are likely to yield destructive eruptions during the next decades might lack prior instrumental data76.

Three main volcano-tectonic processes occur in the SVZ, and each one generates an end-member stress pattern. These pro- cesses can be summarized in: (i) regional far-field tectonic

deformation (Type A); (ii) local state of stress driven by volcanic edifice and magma storage processes (Type B), and (iii) local transient stress driven byfluid circulation through open-fractures (Type-C). These stress patterns have been observed in other volcanic arcs, suggesting that these stress patterns are not restricted to the SVZ. The most recurrent pattern in the SVZ, as well as in Aleutians-Alaska, Northeast Honshu, and Java volcanic arcs, is Type-A.

Stress analysis from earthquake focal mechanisms and fault slip data should be further combined with the Monte Carlo method proposed here to determine likely fluid pathways in individual volcanoes. This analysis could be extended to most, if not all, active volcanoes in the SVZ, to assess potential fluid pathways within the upper crust, and can be applied to other volcanic arcs.

Applying the Monte Carlo simulation to the SVZ, ourfindings indicate that volcanoes with Type-B stress state have the most favorable conditions for triggering extension fractures, that pro- mote magmatic and hydrothermalfluid circulation. Furthermore, this stress pattern may reflect the presence of a magma storage zone nested in the brittle upper crust, as all seven of the volcanoes with Type-B patterns have shown evidence of this. Therefore, these volcanoes may have a heated crustal volume in their vicinity generated by hot-fluid circulation around magma bodies, making them attractive geothermal targets.

Type-C stress patterns should be carefully examined in the future as an eruption forecast, considering that from the analyzed volcanoes, 96% of eruptions in the last 300 years have occurred in volcanoes with this stress signature, and both reliable hypotheses to explain its occurrence imply fluid migration. Therefore, it is necessary to further evaluate the causality between eruptions and Type-C stress patterns.

Methods section

Principal stress definition. The calculation of the principal stress directions and the stress ratio was done using the Multiple Inverse Method (MIM) software of Yamaji et al.37,38. The MIM reveals the stress state from shear fracture data, by minimizing

Mentolat

Caulle

Los Pescados Puyuhuapi

Fui

Tolhuaca

Sierra Nevada

Laguna del Maule

San Pedro-Tatara

Planchón- Peteroa Tinguiririca

Copahue-Caviahue

Nevados de Chillán

Villarrica

Type A

Type B

Type C

Local stress dominated by volcanic-load and/or magmatic chamber

State of stress dominated by regional far-field tectonic stress

Local and transient stress driven by fluids passing through open-fractures

Fig. 7 Volcanoes of the SVZ, as classied in this work onto a triangular plot, which is based on the amount of compatible data with each stress type (Type A, B, C).The volcanoes are subclassied as A+B; A+C; or B+C, when they have a stress state that includes two end-member stress patterns (for example, volcanoes as Mentolat, with hybrid stress type that include Type A and B) or when the volcanoes that record two state of stress containing two different end-member types (for example, Planchón-Peteroa that has a primary stress state classied as Type-A, and its secondary stress state pertaining to Type C, shown in Fig.2). The relative proportion of each stress pattern for each volcano is explained in Supplementary Table 2.

11

Referensi

Dokumen terkait