accuracy and uncertainty of texnet absolute...

?

??

?

Accuracy and Uncertainty of TexNet Absolute Seismic Event Location in West Texas

Anthony Lomax ALomax Scientifc

Mouans-Sartoux, France anthony@alomax.net

Alexandros SavvaidisBureau of Economic Geology

The University of Texas at Austinalexandros.savvaidis@beg.utexas.edu

2019 Annual Meeting23-26 April 2019, Seattle

?

?

??

?

?

6km

Where are the earthquakes?

critical to monitoringand explaining

human-inducedseismicity

P, S

P, SP, SP, S

P, S

Problems: stations only at surface + velocity model error

?

?

??

?

?

modelerror model

error

stations

Problem: stations only at surface → horizontal rays at depth

?

?

ray paths?

?

?

?

stations

?

?

ray pathsmodelerror model

error

??

?

?

Problem: velocity model error → incorrect times and rays

stations

ray paths modelerror

Surface stations + model error → large hypocentral error

modelerror

stations

West Texas, Delaware Basin study area

Delaware Basin

An increase in induced seismicity inthe central U.S. since 2009 led to

establishment of the TexNet seismic-monitoring program in Texas.

One of the largest petroleumproducing areas in the U.S.

→ Hydraulic fracturing→ Waste water disposal

PB02

West Texas, Delaware Basin study area

TexNet seismicity 2017-2018M = 0.3 → 3.7

PB3D model (Huang et al., 2019)

12km

25km

25-50km

Delaware Basin

PB02

West Texas, Delaware Basin study area

TexNet seismicity 2017-2018M = 0.3 → 3.7

PB3D model (Huang et al., 2019)

12km

25km

Deepening of hypocenterswith distance from stations!

Delaware Basin

~4+ kmepicentershift

Single event absolute errors – Diferent models, station sets

PB02

Random N/2 stations excluded in turn, PB02 may be excluded25-50km station spacing

4 km

PB02

~4+ kmepicentershift

PB02

~5+ kmdepth shift

Single event absolute errors – Diferent models, station setsRandom N/2 stations excluded in turn

4 km 6km

25-50km station spacing

6km

Best: calibration with ground-truth sources

explosion

dense, nearbynetwork

→ P, S Station Corrections

6km

Best: calibration with ground-truth sources

explosion

dense, nearbynetwork

→ P, S Station Corrections

6km

Assume hydraulic fracturing may cause earthquakes

→ Probabilistic, Proxy Ground-Truth(PPGT) Station Corrections

→ Proxy Ground-Truth

12km

Probabilistic, Proxy Ground-Truth corrections from fracturing

PB02

~1000 events near PB02

distancetime

Based on literature + absolute error analysisParameters optimized with grid search

hydraulic fracturing wells

Association likelihood

Probabilistic, Proxy Ground-Truth corrections from fracturing

~900 associations fracturing jobs→events

12km

distancetime

PB02

Association likelihood

Association likelihood

12km

supports assumption: fracturing may cause earthquakes

test reverse time: earthquake→fracturing (a-causal)

Probabilistic, Proxy Ground-Truth corrections from fracturing

→ always fewer a-causal associations✓

6km

Assume fracturing may cause earthquakes → Proxy ground truth

→ PPGT Station Corrections

P, SP, S

Proxy ground truth corrections from fracturing wells Probabilistic, Proxy Ground-Truth corrections from fracturing

Referencelocations

PPGTcorrections

SSSTcorrections

12km

25kmPB02

TexNet seismicity 2017-2018, PB3D model

Referencelocations

wastewater injectionfracturing

Distance from PB02 (km)

Distance from PB02 (km)

Distance from PB02 (km)

Dept

h (k

m)

Dept

h (k

m)

Dept

h (k

m)

count

count

count

Proxy ground truth corrections from fracturing wells Probabilistic, Proxy Ground-Truth corrections from fracturing

Referencelocations

PPGTcorrections

SSSTcorrections

12km

25kmPB02

TexNet seismicity 2017-2018, PB3D model

Referencelocations

wastewater injectionfracturing

Distance from PB02 (km)

Distance from PB02 (km)

Distance from PB02 (km)

Dept

h (k

m)

Dept

h (k

m)

Dept

h (k

m)

count

count

count

Proxy ground truth corrections from fracturing wells Probabilistic, Proxy Ground-Truth corrections from fracturing

Referencelocations

PPGTcorrections

SSSTcorrections

12km

25kmPB02

TexNet seismicity 2017-2018, PB3D model

Referencelocations

wastewater injectionfracturing

Distance from PB02 (km)

Distance from PB02 (km)

Distance from PB02 (km)

Dept

h (k

m)

Dept

h (k

m)

Dept

h (k

m)

count

count

count

Proxy ground truth corrections from fracturing wells Probabilistic, Proxy Ground-Truth corrections from fracturing

Referencelocations

PPGTcorrections

SSSTcorrections

12km

25kmPB02

TexNet seismicity 2017-2018, PB3D model

Referencelocations

wastewater injectionfracturing

Distance from PB02 (km)

Distance from PB02 (km)

Distance from PB02 (km)

Dept

h (k

m)

Dept

h (k

m)

Dept

h (k

m)

count

count

count

✓

?

?

?

?

Accuracy and Uncertainty of TexNet Absolute Seismic Event Location in West Texas

Anthony LomaxALomax Scientifc

Mouans-Sartoux, France anthony@alomax.net

Alexandros SavvaidisBureau of Economic Geology

The University of Texas at Austinalexandros.savvaidis@beg.utexas.edu

Conclusions: ● Accurate absolute location of seismic events isdifcult but critical to monitoring and explaininghuman-induced seismicity. ● For the Delaware Basin, western Texas, relocationusing diferent velocity models and station subsetsshows 4–5 km absolute epicentral and depth error. ● We associate fracturing and seismicity to developProbabil ist ic Proxy, Ground-Truth stationcorrections to improve absolute event locations. ● Always fewer a-causal associations: supportsassumption fracturing may cause earthquakes. ● Spatial patterns and statistics of west Texasseismicity relocated with PPGT corrections suggestimprovements in absolute location accuracy.

2019 Annual Meeting23-26 April 2019, Seattle

✓✓