This document summarizes a seismic hazard model for the Middle East region. It includes 3 area source models, 9 fault source models, and a spatially smoothed seismicity model developed based on a declustered earthquake catalog. The models were constructed through a collaborative process involving multiple experts. The key elements summarized are: - 143 area source zones defined based on seismicity patterns and tectonic features. - Fault sources were selected based on being capable and having slip rates above 0.1 mm/year, with 3 confidence classes. - Maximum magnitudes were assigned through various methods with sensitivity analysis performed. - A logic tree incorporates the alternative source models and characterizations. - The models were developed to be stable

1. Seismic Hazard Model
For Middle-East Region
Laurentiu Danciu
Swiss Seismological Service,
ETH Zurich, Switzerland &
GEM Hazard Modeler
Karin Sesetyan
Mine Demircioglu
KANDILLI OBSERVATORY
and EARTHQUAKE RESEARCH
INSTITUTE,
Istanbul Turkey
EMME Final Meeting
September 30th – October 2nd
Istanbul, Turkey

2. Regional PSHA Model
• Regional seismic hazard assessment

3. Regional PSHA Model
• Harmonization across national-borders

4. Target

5. Stability
1. Provide assurance that the numerical hazard
results will be stable for the next years (50
years ?)
2. Unless significant new seismic information,
which could occur at any time, calls for a
major revision

6. Consensus

7. What Consensus stands for?
• 1) there is not likely to be "consensus" (as the
word is commonly understood) among the various
experts and
• 2) no single interpretation concerning a complex
earth - sciences issue is the "correct" one.
SSHAC : Recommendations for PSHA: Guidance on Uncertainty and Use
of Experts
Most likely for no consensus there is a
consensus!

8. SHARE Project- DB Stats
• 3 source models
• 960 End-Branches
• 12 Intensity Measure Types
• 7 Return Periods [50 to 10000Years]
• Mean, Median and Four Quantile
• 130 000 sites
‣ Hazard Maps: 504
‣ Hazard Curves: 9.36 mil
‣ Uniform Hazard Spectra: 5.46mil
‣ Disaggregation: ongoing
Dynamic Model
EMME Project- DB Stats
55000
sites

9. Overview
• Earthquake Catalog
• Maximum Magnitude
• Seismic Source Models
–Area Source Model
–Fault source Model
–Spatially Smoothed Seismicity

10. EMME Earthquake Catalog
• Historical part (-1900)
• Early and modern instrumental (~2006)
• Harmonized in terms of Mw
Total : 27174 events

11. EMME Earthquake Catalog
• Seismicity models require a
– Declustered earthquake catalog of independent events
– Completeness intervals for estimating the Poissonian (time-
independent) earthquake rates.
• Declustering Method.
– Windowing approach based on windows provided by Grünt
hal (1985)
• Gardner and Knopoff (1974)
– After Declustering: 10524 events

12. Declustered Earthquake Catalog
– After Declustering: 10524 events

13. EMME Earthquake Catalog
• 18 Completeness super-zones

14. Completeness plots per super-zones

15. Completeness plots per super-zones

16. Completeness plots per super-zones

17. Maximum Magnitude
• Largest magnitudes that a seismogenic region is
capable of generating.
• Upper-bound magnitude to the earthquake
recurrence (frequency-magnitude) curve.
• Maximum Magnitude assessment (Super-Zones)
– Historical seismicity record
– Location uncertainties
– Analogies to tectonic regions
– Added increment (0.30)

18. Maximum Magnitude

19. Weighted Mmax

20. Maximum Magnitude: Sensitivity

21. Maximum Magnitude: Sensitivity
5%
13%
18%

22. Yerevan City
Magnitude - Disaggregation
Maximum Magnitude = 8.00
aGR = 4.00
bGR = 1.00

23. Maximum Magnitude: Sensitivity
• Maximum magnitude impacts the activity
computation –if used to anchor the expert
fitting
• 5 to 20% increased hazard values
• Return period dependent
• The pair of Maximum Magnitude
Recurrence rates to carefully be revised

24. Source Model Logic Tree

25. Area Source Model
• Classical area source zones based on the tectonic
findings and their correlation and up-to-day
seismicity
• Derived from seismicity patterns
– Ensure the zonation adequately reflect this
pattern
• Surface projection of identified active faults
(capable of generating earthquakes)

26. Model Construction: Phase One
• Country based models
• Phase one:
– Overlapping sources at national borders
• [ trying to keep the original information]
– Remove duplicates (the same source
defined within countries)
– Eliminate zones too small to be analyzed
(spatially smooth seismicity take cares of
it)

27. Model Construction: Phase Two
– Simplify unnecessary or artificial complex
zonation
– Reshaping according to the known main
seismogenic features (i.e known faults )
– Local experts feedback
– Reconcile different interpretations
– New sources re-defined after technical
discussions among the national
representatives/local experts

28. Area Source Model
• 143 shallow crustal area source zones,
• 6 for modeling the deep seismicity, and
• 5 complex faults

29. Source Characterization
• Main Assumptions:
– Homogeneous, declustered catalogue
– Completeness defined for 18 super zones spanning the entire
region
– Maximum likelihood approach (Weichert 1984)
– Truncated Guttenberg-Richter Magnitude Frequency
Distribution
• 10a – annual number of events of magnitude greater or
equal to zero
• b-value
• Truncated at each assigned maximum magnitude
• For each source three magnitude-frequency-distributions were
derived
• A Matlab* toolbox was developed

30. Source Characterization: issues
• Sources with limited number of events
• Sources with less that 15 events were assigned with a default
activity rate corresponding to each region
• 30 area source zones with less than 15 events
• Estimation stability achieved for more than 30 events/source

31. Source Characterization: issues
• Light color shows area sources with less than 15 events

32. Source Characterization: issues
• Automated procedure was constantly under predicting the
occurrence rates of moderate to large magnitudes

33. Source Characterization: issues
• Automated procedure was constantly under predicting the
occurrence rates of moderate to large magnitudes

34. Source Characterization: issues
• Automated procedure was constantly under predicting the
occurrence rates of moderate to large magnitudes

35. Source Characterization: solution
• Expert fitting/adjustments
automated

36. Source Characterization: solution
• Expert fitting/adjustments
automated

37. bGR-value : Spatial Distribution
bGR-value is between 0.70 to 0.75
bGR-value is between 1.05 to 1.25

38. Sanity Check: aGR normalized:
• Yellow color shows area sources with very low annual activity per km2

39. Sanity Check: Seismic Moment
log(Mo)  c dMw
c 16.05
d 1.5
Kanamori and Anderson (1975)

40. Sanity Check: Global Strain Rate
Courtesy of C.Kreemer

41. Area Sources vs Global Strain Rate

42. Seismic Moment vs.Global Strain Rate

43. Source Parameterization
• Depth Distribution (three values and the
corresponding weights):
– Active shallow crust
– Nested Deep Seismicity
– Subduction Inslab
• Focal Mechanisms
– Rake Angle values (Aki’s definition)
– Percentage weights
• Ruptures Orientation
– Strike Angle (Azimuth)
– Dip Angle
• Rupture Properties
– Upper and Lower Seismogenic Depth
Complex Fault
Area Source
[Single Rupture]

44. Depth Distribution

45. Dip-Angle Distribution
• Yellow color shows area sources with
• dip angles smaller than 30o

46. Source Model Logic Tree

47. EMME Faults Dataset
• Fault source model derived from the faults database collected within WP02
– Total number: 3397 fault segments
– Total Km: 91551km

48. Fault Sources
• Criteria to select active faults to be used for hazard
assessment:
– Identified active faults [capable of earthquakes]: Northern
Anatolian Faults, Marmara Faults, Zagros Transform Faults
– At least 0.10mm/year (1m in 1000years - Neocene)
– Maximum magnitude equal to 6.20
– Fully parameterized:
• Geometry
• Slip-rates
– Confidence Classes:
• Class A: complete information provided by the compiler
• Class B: partial information provided by compiler
• Class C: limited information provided
• Class D: only top trace available

49. Fault Sources
• Confidence Classes:
• Class A (red): complete information provided by the compiler
• Class B (green): partial information provided by compiler
• Class C (blue): limited information provided

50. Fault Sources-SHARE projects
• Class A complete information provided by the compiler

51. Fault Sources- EMME
• Class B partial information provided by the compiler

52. Fault Sources- EMME
?

53. Fault Sources- Class C
• Class C fault trace and fault type info available
• Maximum magnitude
– Estimated from faults size
– Slip rate
• First Slip Rate Estimated as proposed by USGS

54. • First Slip Rate Estimated as proposed by USGS

55. 15.
1
8.5
15.
0
<3
15
.0 9.
2
16.
8
11.4
16.8
6.
8
13.
2
9.
9
1
3
1
3
4
11±5
10±2
6±1
3±2
25±5
~4
20.
0
0
5
10
15
20
5
5
15
20
Fault Sources- Class C -refinement
By Prof. Asif Khan

56. Active Faults

57. Active Faults
How to characterize the seismic potential of the faults?
- Convert slip-rates to seismicity

58. Fault Source Model Characterization
Procedure:
1. Generate a buffer region of 20km
for each fault

59. Procedure:
2. Remove earthquakes within buffer zone
3. Activity on faults computed from slip rates
4. Activity on the background – based on the “outside”
catalogue
Fault Source Model Characterization

60. Faults + Background Seismicity

61. Faults + Removed Earthquakes
3202 earthquakes removed
Magnitude range: 4.00 to 7.90

62. Faults Characterization
Activity rates are calculated from geologic
information:
•Slip rate
•Fault length / aspect ratio
•Maximum Magnitude
Recurrence Rate Model:
•Anderson & Luco (1983) Model 2:
•b-value assumed from the corresponding completeness
super zones
•Integration from Mmin = 5.00 to Faults Mmax
N2(M)=
d -b
b
æ
èç
ö
ø÷
S
b
æ
èç
ö
ø÷ eb-(Mmax-M
-1é
ë
ù
ûe-((d/2)Mmax )

63. Activity Rates - Background
• Smoothed spatially with a variable Kernel
– r : epicentral distance
– di: Variable epicentral distance to next neighbor
nv
• Optimization for distance parameter with
retrospective tests
vsF (r,di )  c(di )(r2
di
2
)1.5

64. Kernel Optimization: Retrospective Testing
 Optimize kernel using a
likelihood tests
 Split catalog in learning
and target period
 Optimize on 5 year target
period
 Use best likelihood-value
to generate model rates
Learning Period Target Period
1000 2002 2007

65. Activity Comparison

66. Fault Source Model
Combine seismicity from long term geological observations with observed seismicity

67. Fault Source Model

68. Source Parameterization
• Depth Distribution
• Focal Mechanisms
– Rake Angle values (Aki’s definition)
– Percentage weights
• Ruptures Orientation: Strike and Dip Angles
• Rupture Properties
– Upper and Lower Seismogenic Depth
Point Source
[Single Rupture]
Simple Fault• Fault Top Trace
• Focal Mechanisms
– Rake Angle values (Aki’s definition)
• Ruptures Orientation: Strike and Dip Angles
• Rupture Properties
– Upper and Lower Seismogenic Depth

69. 24th Sept 2013 Event in Pakistan

70. 15Years Seismicity Mw >= 6.5
2013-09-24 Awaran Pakistan
2013-04-16 East of Khash Iran
2011-10-23 Eastern Turkey
2011-01-18 southwestern Pakistan
2010-12-20 southeastern Iran
2009-01-03 Hindu Kush region Afghanistan
2008-10-05 Kyrgyzstan
2005-12-12 Hindu Kush region Afghanistan
2005-10-08 Pakistan
2004-04-05 Hindu Kush region Afghanistan
2003-12-26 Southeastern Iran
2002-06-22 Western Iran
2002-03-03 Hindu Kush region Afghanistan
2002-02-03 western Turkey
2001-01-26 Gujarat India
2000-12-06 Turkmenistan
2000-11-25 Caspian Sea offshore Azerbaijan
1999-11-12 western Turkey
1999-11-08 Hindu Kush region Afghanistan
1999-08-17 Western Turkey
1999-03-04 Southern Iran
1998-05-30 Hindu Kush region Afghanistan
1998-03-14 Eastern Iran

71. Before 24th Sept 2013 Event in Pakistan
EMME Results, before the earthquake

72. Source Model Logic Tree

73. Spatially Smoothed Seismicity
• Based on the
– Up-to-date seismicity
– Declustered catalogue
• Main Assumption:
– Earthquake's self-similarity: earthquakes occur at near clusters of
previous smaller earthquakes.
– Derived equally spaced [10 x 10 km] cells
– 53300 non-overlapping cells
– the earthquake rates determined for cells are spatially smoothed
using a one Gaussian smoothing kernel Frankel 1995]
– Kernel constant size equals to 25km

74. Smoothing Algorithms

75. Source Parameterization
• Depth Distribution (three values and the corresponding
weights):
• Focal Mechanisms
– Rake Angle values (Aki’s definition)
– Percentage weights
• Ruptures Orientation
– Strike Angle (Azimuth)
– Dip Angle
• Rupture Properties
– Upper and Lower Seismogenic Depth
Point Source
[Single Rupture]

76. Source Model Logic Tree
How do we weight them?

77. Summary
•Building a regional seismic hazard model is a collective
effort
•Aim at generating the up-to-date , flexible and scalable
database hat will permit continuous update, refinement, and
analysis.
•Data will be parameterized and input into the database with
a specific format.
Hazard
Software
“Black Box”
INPUT OUTPUT
“Easy Review” Box
Data
Interpretations
Assumptions

78. Summary
•Transparent computational procedure, with all input files
available as well as the software packages (Hazard
Modeler Toolkit, OpenQuake)
•Each dataset has certain degree of completeness, but
there is room for improvements;
•Specifically,
•The depth information of the events
•Maximum magnitude definition
•More parameterized faults
•Velocities from GPS data
•Revision of all source models
•What are the weakness points of each model?
•Road map to the final deliverable

79. Thank you!