Seismic Hazard and Risk Assessment for Induced SeismicityEllen M. Rathje, PhD, PE, F.ASCE

Janet S. Cockrell Centennial Chair in EngineeringDepartment of Civil, Architectural, and Environmental Engineering, University of Texas, Austin

Prof. Patricia Clayton, Prof. Brady CoxIason Grigoratos, Farid Khosravikia, Meibai Li, Michael Yust

Department of Civil, Architectural, and Environmental Engineering, University of Texas, AustinDr. Alexandros Savvaidis

Bureau of Economic Geology, University of Texas, Austin

Seismic Risk Assessment

Risk = Hazard

x Exposure

x Vulnerability

x Consequences

Measure of ground shaking and its probability

Characterization of built environment and inhabitants

Susceptibility of the exposure to damage/undesirable consequences

$$, number of people adversely affected

Beer!Food!

Clayton, Khosravikia, Rathje, Grigoratos

Rathje, Grigoratos, Cox, Li, Yust, Savvaidis

Seismic Hazard AssessmentRisk = Hazard x Exposure x Fragility x Consequences

Seismic Source Characterization

Ground Motion Characterization

(Reiter, 1990)

Requires:• Rate of earthquakes• Magnitude (M)

distribution• Locations

Requires:• Ground shaking as a function of

M, distance (R), and soil/rock conditions (Vs30)

• Variability

For tectonic EQs: • Stationary• Use historical EQ

catalog

For induced EQs:• Time-dependent• Relate seismicity to oil/gas

operations (e.g., injection)

For induced EQs:Use recordings from events in

geologically similar regions

4

Temporal and Spatial Variations in Seismicity: Oklahoma

Seismicity

Wastewater Injection

Temporal Spatial

5

Time-Dependent Gutenberg-Richter Relationship

𝜆𝜆 = 10𝑎𝑎𝑡𝑡𝑡𝑡𝑡𝑡−𝑏𝑏�𝑚𝑚

𝜆𝜆[𝑡𝑡] = 10𝑎𝑎𝑡𝑡𝑡𝑡𝑡𝑡 + 𝑣𝑣𝑙𝑙𝑙𝑙𝑙𝑙[t] � 10𝚺𝚺 � 10−𝑏𝑏�𝑚𝑚

monthly lagged distributed

injection rate

Seismogenic Index

with 𝑡𝑡𝑙𝑙𝑎𝑎𝑙𝑙[t] = 𝜽𝜽𝑣𝑣[𝑡𝑡]

monthly distributed injection rate

Free parameters:𝜽𝜽 and Σ from monthly injection and seismicity data𝑙𝑙𝑡𝑡𝑡𝑡𝑡𝑡 and b from background seismicity (<2009)

Spatial and temporal resolution:Monthly injection/seismicity for 5-km x 5-km blocks

Semi-empirical model with parameters derived from seismicity and injection data

6

Application to OklahomaMonthly EQs and injection volumes 2000-2018

Cumulative injection volume (m3) 2006-2018 Spatially Varying Model Parameters

Gridded and Spatially Distributed Injection VolumesCumulative 2006-2018

Seismogenic Index (Σ) Hydromechanical Parameter (θ)

7

Simulated seismicity rates

Sub-Regions

Full Study Area

Calibration: Jan 2006 – Dec 2017

ObservedSimulated

8

Hindcasting: calibrate parameters through Dec 2014

Calibration thru Dec 2014

Calibration Forecast

Model performance when we calibrate parameters thru Dec 2014 and then forecast the EQs, given the injection ratesCalibration thru

Dec 2017

Observed

Simulated

9

Hindcasting: calibrate parameters through **Dec 2011**

Calibration thru Dec 2011

We need to spatially extrapolate Σ and θ

Calibration Forecast

Model performance when we calibrate parameters thru Dec 2011 and then forecast the EQs, given the injection ratesCalibration thru

Dec 2014

Ground Motion CharacterizationEmpirical Ground Motion

Model (GMM)

Peak

Gro

und

Acce

l, PG

A (g

)

Distance (km)

Rock Conditions

Characterization of Shear Wave Velocity (Vs30)

Zalachoris and Rathje (2019) EQ Spectra

10

Using data from Texas, Oklahoma, and Kansas

Ground Motion Model (GMM) Development• Events in TX, OK and KS with M > 3.0 between 2005-2017• Recordings from TX, OK and KS seismic stations (past and existing)

2.5

3

3.5

4

4.5

5

5.5

6

1 10 100

Mom

ent M

agni

tude

, MW

Hypocentral Distance, Rhyp (km)

(a)• 4,815 records• 398 Events• 223 Stations

4.5 < M < 5.0 M > 5.0

M < 3.5 3.5 < M < 4.0 4.0 < M < 4.5

Gulf Coast Plain Stations

Events Stations

Zalachoris and Rathje (2019) EQ Spectra

Assessment of Existing GMMs

VS30 = 760 m/s VS30 = 760 m/s

Hassani & Atkinson: NGA-East (2015)

Atkinson (2015): PotentiallyInduced EQs (PIEs)

• Develop empirical adjustment for Hassani and Atkinson (2015) GMM using TX-OK-KS ground motion data

Reference Empirical Approach

Underprediction

Overprediction

12

Adjustment Factors

𝐶𝐶 = 𝑡𝑡𝑐𝑐𝑐𝑐𝑐𝑐𝑡𝑡.

ln𝐹𝐹 = 𝐶𝐶 + 𝐹𝐹𝑅𝑅 + 𝐹𝐹𝑆𝑆 + 𝐹𝐹𝑀𝑀 ⟹

Overall Adjustment:

At each frequency, f:

Distance Scaling:

VS30 Scaling:

Magnitude Scaling:

Final Model for Adjustment Factors:

𝐹𝐹𝑅𝑅 = �0 𝑅𝑅 ≥ 𝑅𝑅𝑏𝑏

𝑙𝑙 ∙ ln �𝑅𝑅𝑅𝑅𝑏𝑏� 𝑅𝑅 < 𝑅𝑅𝑏𝑏

𝐹𝐹𝑆𝑆 = �0 𝑉𝑉𝑆𝑆30 > 𝑉𝑉𝐶𝐶∗

𝑡𝑡∗ ∙ ln�𝑉𝑉𝑆𝑆30

𝑉𝑉𝐶𝐶∗� 𝑉𝑉𝑆𝑆30 ≤ 𝑉𝑉𝐶𝐶∗

𝐹𝐹𝑀𝑀 = 𝑏𝑏0 + 𝑏𝑏1 ∙ 𝑀𝑀

𝑅𝑅 = �𝑅𝑅ℎ𝑦𝑦𝑦𝑦2 + ℎ𝑡𝑡𝑒𝑒𝑒𝑒2 heff: effective depth (Yenier & Atkinson, 2014)

⟹ 𝑃𝑃𝑆𝑆𝑃𝑃𝑇𝑇𝑇𝑇−𝑂𝑂𝑂𝑂−𝐾𝐾𝑐𝑐 = 𝐹𝐹 ∙ 𝑃𝑃𝑆𝑆𝑃𝑃𝐻𝐻𝑃𝑃15

Zalachoris and Rathje (2019) Ground Motion ModelM ≥ 5.0, Vs30 = 760 m/s

0

1

2

3

4

5

6

7

8

100 1000

Am

plifi

catio

n (V

S30

= 76

0 m

/s)

VS30 (m/s)

T = 1.0 sec BSSA14

Adjusted BSS14

NGA-East

VC

Adjusted Vs30 Scaling

ZR19NGA-West2

NGA-East

Shear Wave Velocity CharacterizationP-wave Seismogram Method

Incident PReflected SV

Reflected P

UZ

URx

ii

j

Ni et al. (2014), Kim et al. (2016)

VSZ = 1578 m/sVS30 = 1107 m/s

VSZ = 519 m/sVS30 = 436 m/s

Vs30 = 1107 m/s

Vs30 = 436 m/s

+±

=

Z

R

Z

R

S

UUp

UUjj

V

2

2coscos2

2

15

Shear Wave Velocity CharacterizationP-wave Seismogram Method

Incident PReflected SV

Reflected P

UZ

URx

ii

j

Ni et al. (2014), Kim et al. (2016)

+±

=

Z

R

Z

R

S

UUp

UUjj

V

2

2coscos2

2

Vs30 at Recording Stations

16

In Situ Vs30 Measurements

Field Measurements

Performed by Cox and YustZalachoris et al. (2017) EQ Spectra

Development of Comprehensive Vs30 MapGeologic Proxy for Vs30:

Age and Rock Type

17

Geologic age Rock Type Group

Npts

Quaternary-Holocene (Outside of Gulf Coast) A/C 11 484Quaternary-Pleistocene (Outside of Gulf Coast) A/B/C/D 30 526Quaternary-Undivided (Outside of Gulf Coast) A/B/C 18 588

Quaternary-Holocene (In Gulf Coast) A/C 62 211Quaternary-Pleistocene (In Gulf Coast) B/C 7 242Quaternary-Undivided (In Gulf Coast) A/B 4 213

B 11 386C/D 30 466

E 1 696F 2 838

B/C/D 42 517E 37 765D 80 747E 12 971F 3 1638

Precambrian F 2 1434

Tertiary

Mesozoic

Paleozoic

𝝁𝒍𝒏𝑽 (𝒎/𝒔)

Inputs:1. Geologic Atlas of

Texas from BEG2. P-wave seismogram

estimated Vs303. In situ measurement of

Vs30

Intermediate Result: 1. Vs30 map based on

Geologic proxy

Final Output: 1. Kriged Vs30 map

Calculate residuals

Geostatistical interpolation

Vs30 observations

Mapping Approach

Geologic Age Rock Type

Group A: Alluvial and terrace depositsGroup B: Clay, silt, and loess; not alluviumGroup C: Sand and gravel; not alluvium.Group D: Mud/clay/silt/sand stone, conglomerate, marl, and shale Group E: Limestone and chalkGroup F: Chert, basalt, granite, and rhyolite

Rock Type Groups

Zalachoris et al. (2017) EQ SpectraLi et al. (in prep) EQ Spectra

#pts

Vs(m/s)

Comparison of Vs30 Maps

USGS Global Vs30 from Topographic Proxy

18

Texas-specific Vs30 Map(This study)

Ratio = Texas/USGS

Seismic Risk Assessment

Risk = Hazard

x Exposure

x Fragility

x Consequences

Measure of ground shaking and its probability

Characterization of built environment and inhabitants

Susceptibility of the exposure to damage/undesirable consequences

$$, number of people adversely affected

Beer!Food!

19

Fragility Curves

Ground Shaking Intensity (e.g. Peak Ground Acceleration,

Peak Ground Velocity)

Prob

abili

ty o

f Dam

age

Texas Building

California Building

0.5

1.0

Less fragile

More fragile

Used to predict the likelihood of

damage for a given ground shaking

intensity

20

Clayton and Khosravikia

21

Masonry Facades: Effect of Construction Practices

M5.7 Prague, OK (Nov. 2011)

In seismically-active areas:• Brick facades and chimneys are avoided• When used, more bracing is used to

prevent collapse

Unreinforced brick:• Commonly used in Texas• Known to be vulnerable in earthquakes

Fragility curves must reflect local construction practices

22

Effect of Ground Motions

Fragility curves must reflect ground motion characteristics

0 0.5 1 1.5 2 2.5 3

Tn [s]

0

1

2

3

4

5

6

Sa

/PG

A

Mean

“West coast” design motions

0 0.5 1 1.5 2 2.5 3

Tn

[s]0

1

2

3

4

5

6

Sa

/PG

A [

g]

Mean

Texas/Oklahoma motions

Earthquake Frequency ContentHigh frequency content = More

damage to buildings with low natural

period

Low energy for long period waves = Less damage to buildings with higher natural

period

23

Masonry Façade Fragility: Comparison of intensity measures

0 0.5 1 1.5 20

0.2

0.4

0.6

0.8

1

Sa05

Prob

abili

ty o

f dam

age

Rand 4Max MwMax PGAMin Dist

PSA(0.5) (g)

Prob

abili

ty o

f dam

age

PGA (g)

Prob

abili

ty o

f dam

age

PGV (cm/s)

Prob

abili

ty o

f dam

age

Sa(0.2) (g) PSA(0.2) (g)

Prob

abili

ty o

f dam

age

PGA PGV

PSA(0.2) PSA(0.5)

0 1 2 3 4 50

0.2

0.4

0.6

0.8

1

Sa(0.2) (g)

Prob

abili

ty o

f dam

age

22ecc-DS122ecc-DS228min-DS128min-DS2

Collapse

Cracking

Fragility of Different Types of Infrastructure

Residential Masonry Facades(TexNet - CISR)

Bridges (TxDOT)

(Clayton, Kurkowski, Khosravikia) (Clayton, Cox, Rathje, Williamson, Khosravikia, Potter, Prakhov, Zalachoris)

2016 M5.8 Pawnee, OK (source: P. Clayton) 2016 M5.8 Pawnee, OK

(source: P. Clayton)

Characterize Bridge Inventory

1920 1940 1960 1980 2000 20200

200

400

600

800

Year of Construction#

Brid

ges

Pre-stressed Concrete Girders(1970s-present)

Steel Girders(1950s-60s)

~53,000 bridges & culverts in Texas

26

Characterize Bridge Vulnerability

• Used computer models to simulate:• Different geometries (height, span length, etc.)• Different construction materials & designs• Wide range of ground motions

Verti

cal

Rigid linkBearing springAbutment/ foundation springs

Superstructure mass

Pounding Element

Bearing

AbutmentL L

L

T

T

Longitudinal

Tran

sver

se

Bearing

L

T

Tran

sver

se

Longitudinal

Developed in OpenSees Software

Bridge Fragility Curves in Texas

27

Slight Moderate Extensive Complete

Prob

abili

ty o

f dam

age

PGA (g) PGA (g) PGA (g) PGA (g)

0 0.2 0.4 0.6 0.8 1

0

0.2

0.4

0.6

0.8

1

0 0.2 0.4 0.6 0.8 1

0

0.2

0.4

0.6

0.8

1

0 0.2 0.4 0.6 0.8 1

0

0.2

0.4

0.6

0.8

1

0 0.2 0.4 0.6 0.8 1

0

0.2

0.4

0.6

0.8

1

MCSTEEL MSPC MCRC-Slab MSRC MSSTEEL SSPC MSRC-Slab

Steel and Pre-stressed

Concrete

Reinforced Concrete

Fragility Comparison

28

21.6Continuous steel girder bridges

Slight Moderate Extensive Complete

Prob

abili

ty o

f dam

age

PGA (g) PGA (g) PGA (g) PGA (g)

0 0.4 0.8 1.2 1.6 2

0

0.2

0.4

0.6

0.8

1

0 0.4 0.8 1.2 1.6 2

0

0.2

0.4

0.6

0.8

1

0 0.4 0.8 1.2 1.6 2

0

0.2

0.4

0.6

0.8

1

0 0.4 0.8 1.2 1.6 2

0

0.2

0.4

0.6

0.8

1

Texas HAZUS

Risk Assessment Framework

29

Potential earthquake Shaking

estimation

Damage estimation

Loss estimation

Monte Carlo Simulation

P(Damage State)1.0

0.0

SlightModerateExtensiveComplete

IMMonetary loss

Prob

abili

ty

Structural fragility models

Repair cost analysis

Ground motion model

Shaking level

Damage Estimation

30

Earthquake with M5

P(Damage State)1.0

0.0

SlightModerateExtensiveComplete

IM

A

PGA = 0.5 gFrom ground

motion models

Damage ProbabilitySlight 0.4

Moderate 0.2Extensive 0.1Complete 0.1

No damage 0.2

Slight damage

Hypothetical Earthquake with M5 in Dallas, Texas

One RealizationDamage Repair Cost

Slight 0.03 x Construction Cost

Moderate 0.08 x Construction Cost

Extensive 0.25 x Construction Cost

Complete 1.00 x Construction Cost

Damage Estimation

31

Hypothetical Earthquake with M5 in Dallas, TexasPGV (cm/s)

Loss Estimation

32

Earthquake with M5

Mean = 4.2 M Std. = 0.8 M

1.0 3.0 5.0 7.0 9.0

0.1

0.2

0.3

Scenario-based Loss Estimation

33

$ 4.2 M

Earthquake with M4

Earthquake with M5

Earthquake with M5.8

Mean of Monetary Loss

$ 16.2 M$ 0.04 M

Scenario-based Loss Estimation

34

$ 4.2 M

Earthquake with M4

Earthquake with M5

Earthquake with M5.8

Mean of Monetary Loss

$ 16.2 M$ 0.04 M

$ 0.09 M $ 31.3 M $ 284.3 M

Using HAZUS Fragility Curves

Scenario-based Regional Loss Estimates

35

21.6

Overestimation when region-specific info is NOT used

M4

M5

M5.8

3 times

8 times

17 times

36

Rapid Post-Event Consequence Assessment

• TxDOT implementation of ShakeCast software

• Input TxDOT bridge inventory & vulnerability (from research)

• Automatically retrieves ShakeMap from USGS minutes after event

• Integrate new GMM and Vs30 map

• Real-time report of inspection priorities

• Sends notifications to personnel

(ongoing project funded by TxDOT)

Probabilistic, Time-Dependent Risk Assessment: Oklahoma

Ground Motion Hazard

37

• Annual seismicity rates from Grigoratos et al. BSSA model calibrated through 2017• After 2017, injection rates assumed constant • Ground shaking from Zalachoris and Rathje (2019, EQS) GMM • Event-based annual PSHA from 10,000 simulations using OpenQuake (GEM Foundation)• Building inventory from 2010 census and 2018 replacement costs• Fragility curves for “low-code” buildings in the US (from GEM/USGS)

2015

Simulated Seismicity Monetary Loss Curves10% Annual Probability of Exceedance

38

Conclusions

• Seismic hazard and risk approaches for tectonic earthquakes can be adapted for induced earthquakes

• Key improvements required:‒ Semi-empirical models to forecast spatial and temporal

variations in seismicity‒ Ground motion models for induced earthquakes in the region of

interest‒ Detailed Vs30 maps using regional/local data‒ Fragility models to predict infrastructure damage for the

expected ground shaking characteristics and local construction practices