Remote Monitoring of the Earthquake Cycle using Satellite Radar Interferometry Tim Wright, COMET, Department of Earth Sciences, Oxford University

Transcription

1 Remote Monitoring of the Earthquake Cycle using Satellite Radar Interferometry Tim Wright, COMET, Department of Earth Sciences, Oxford University

2 Outline Lecture 1: Satellite remote sensing Imaging radars the synthetic aperture Satellite radar interferometry How it works Components of interferometric phase Uncertainties associated with InSAR measurements Examples of applications Lecture 2: The Earthquake cycle Coseismic deformation Interseismic Deformation Postseismic deformation

3 Remote Sensing This is passive remote sensing where the Sun provides a natural source of illumination. Active remote sensing involves illuminating the ground from the observing platform in some way, e.g. with radar or lasers.

4 Wavelengths and Frequencies For a periodic wave, as the wave travels past some point the amplitude will oscillate with a period T, which must be the time it takes for the wave to travel a distance equal to the wavelength, i.e. T = λ / The frequency of the wave,, is the number of oscillations per second at some point, so that and c f = 1/T c = fλ

5 The Electromagnetic Spectrum

6 Active Remote Sensing with Microwaves

7 Side-Looking Airborne Radar θ ~ λ / W e.g. λ = 0.05 m W = 10 m θ ~ radians If at 800 km height, along-track footprint ~ 4 km

8 Trick the Synthetic Aperture All the radar echoes that illuminate a given patch of ground are used to construct a synthetic larger antenna

9 Synthetic Aperture Radar (SAR) ERS A SAR makes use of measurements of the range and Doppler shift of the radar returns to locate ground points. The signals from many returns are analysed together to image ground elements ~5x20m in size, much smaller than would be possible with a stationary antenna of the same size - hence the Synthetic Aperture.

10 Active illumination Polar-orbiting, sidelooking Satellite Radar

11 Active illumination Polar-orbiting, sidelooking European Satellites (ERS-1/2, Envisat) Stable orbits and precise pointing (at least until 2000) 10 by 2 m antenna C-band (5.6cm) wavelength ~10 year time series Coherent illumination source Satellite Radar

12 InSAR Interferometric SAR Phase is a function of distance from satellite to ground (range) Path difference results in phase shift 780 km 5.6 cm

13 InSAR how it works Phase is a function of distance from satellite to ground (range)

14 Image A - 12 August 1999 Interferogram = Phase A - Phase B Image B - 16 September 1999 Remove phase from topography satellite positions earth curvature

15 (-20) 567 mm range decrease (-10) 283 mm range decrease (-2) 57 mm range decrease (-1) 28 mm range decrease (0) 0 mm range change 17 August 1999, Izmit earthquake (Turkey)

16 A A 17 August 1999, Izmit earthquake (Turkey)

17 Components of interferometric phase φ int = φ geom + φ topo + φ atm + φ noise + φ def B γ φ = y πb cos hλ 4 2 γ Calculate phase ramp from satellite orbits h ~500 fringes across typical frame Subtract from interferogram Residual orbital errors: ~3 fringes / 100 km across-track ~1 fringe / 100 km along-track y

18 Components of interferometric phase φ int = φ geom + φ topo + φ atm + φ noise + φ def B γ Stereoscopic effect topographic fringes 1 fringe for each change in elevation h a rλ sinγ 10, 000 h a = 2B B h r y

19 Components of interferometric phase φ int = φ geom + φ topo + φ atm + φ noise + φ def Stereoscopic effect topographic fringes h a = 500 m B = 20 m 1 fringe for each change in elevation h a rλ sinγ 10, 000 h a = 2B B h a = 100 m B = 100 m

20 Components of interferometric phase φ int = φ geom + φ topo + φ atm + φ noise + φ def A foggy morning, near ancient Mycenae, Greece

21 Components of interferometric phase φ int = φ geom + φ topo + φ atm + φ noise + φ def Layered atmosphere 29/8/1995 to 29/7/ /8/1995 to 29/7/1997 Topography

22 Components of interferometric phase φ int = φ geom + φ topo + φ atm + φ noise + φ def Layered atmosphere Pass 1 Pass 2 Path delay Path delay

23 Components of interferometric phase φ int = φ geom + φ topo + φ atm + φ noise + φ def Turbulent atmosphere June to December July to December June to July Athens Earthquake September 1999

24 Components of interferometric phase φ int = φ geom + φ topo + φ atm + φ noise + φ def Size of φ atm (at sea level) ~ ± 3 fringes (90 mm) in extreme cases Methods for dealing with φ atm - Ignore (most common) - Quantify - Model based on other observations (e.g. GPS, meteorology ) - Increase SNR by stacking

25 Components of interferometric phase φ int = φ geom + φ topo + φ atm + φ noise + φ def Biggest source of noise is due to changing ground surface Coherence is convenient measure Im φ int Re

26 Components of interferometric phase φ int = φ geom + φ topo + φ atm + φ noise + φ def Biggest source of noise is due to changing ground surface Coherence is convenient measure Im a Im b Re Re Coherence = b / a

27 Components of interferometric phase φ int = φ geom + φ topo + φ atm + φ noise + φ def Coherent surface types Bare Rock Buildings esp. towns/cities Grassland Agricultural fields Ice Athens Incoherent surface types Leafy Trees Water

28 Components of interferometric phase φ int = φ geom + φ topo + φ atm + φ noise + φ def 1. incoherence Changes in the ground cover cause a random phase shift for each pixel 2. Unwrapping errrors Phase in interferograms is wrapped (each fringe is 2 π radians). Discontinuities or data gaps can cause phase unwrapping errrors

29 Components of interferometric phase φ int = φ geom + φ topo + φ atm + φ noise + φ def InSAR ONLY MEASURES THE COMPONENT OF SURFACE DEFORMATION IN THE SATELLITE S LINE OF SIGHT r = - n.u n where n is a unit vector pointing from the ground to the satellite r u φ def = (4π / λ ) r i.e. 1 fringe = 28.3 mm l.o.s. deformation for ERS

30 Applications of InSAR Earthquakes (lecture 2) Volcanoes Ice motion Land subsidence Thematic mapping

31

32

33 Flow of the Lambert Glacier, Antarctica (up to ~1000 m per year)

34

35

36 Outline Lecture 1: Satellite remote sensing Imaging radars the synthetic aperture Satellite radar interferometry How it works Components of interferometric phase Uncertainties associated with InSAR measurements Examples of applications Lecture 2: The Earthquake cycle Coseismic deformation Interseismic Deformation Postseismic deformation

37 100 km The Earthquake Cycle

38 The Earthquake Cycle 5m 100 km 200 yrs

39 The Earthquake Cycle 5m 100 km 200 yrs 20 secs

40 17 August 1999, Izmit Earthquake

41 The Izmit earthquake displacement field

42 A A 17 August 1999, Izmit earthquake (Turkey)

43 Elastic Dislocation Modelling Y. Okada, Surface deformation due to shear and tensile faults in a half-space. Bull. Seism. Soc. Am., 75, To define a rectangular fault dislocation, need 10 parameters: Location of fault x,y,z (x=y=0, z = -d) [1] Length, Width and dip of the fault (L, W, δ) [3] Slip components (u 1 = strike-slip; u 2 = dip-slip; u 3 = tensile) [3] 3D Displacements can be calculated for a point (x obs, y obs ) in the fault-centred reference frame, where the x-axis points along strike. [3]

44 Elastic Dislocation Modelling OKSAR3 (in today s practical) takes 9 friendly fault parameters: x, y-position of centre of fault s surface projection in a map projection [2] Strike, Dip and Rake of fault (Aki, and Richards convention) [3] Magnitude of earthquake slip vector (u 3 = 0, i.e. no opening) [1] Top and Bottom Depths (measured vertically), Fault Length [3] To define a rectangular fault dislocation, need 10 parameters: Location of fault x,y,z (x=y=0, z = -d) [1] Length, Width and dip of the fault (L, W, δ) [3] Slip components (u 1 = strike-slip; u 2 = dip-slip; u 3 = tensile) [3] 3D Displacements can be calculated for a point (x obs, y obs ) in the fault-centred reference frame, where the x-axis points along strike. [3]

45 Surface displacements of strike-slip faults Uplift Subsidence Subsidence Uplift

46 Displacements of normal faults Distance (km) Depth (km)

47 Determining best-fit elastic models Calculating the predicted displacements from a specified fault geometry (forward modelling) is relatively easy. The inverse problem (finding the model that fits a given set of displacements) is harder: Finding the fault geometry is a non-linear inversion problem. Determining slip distributions for a fixed fault geometry is a linear problem.

48 Surface Displacements and Source Parameters of the 2003 Bam (Iran) Earthquake from Envisat ASAR Imagery Gareth Funning 1, Barry Parsons 1, Tim Wright 1, Eric Fielding 2,3, James Jackson 2 and Morteza Talebian 4 1 COMET, Department of Earth Sciences, University of Oxford, UK 2 COMET, Department of Earth Sciences, University of Cambridge, UK 3 Jet Propulsion Laboratory, Caltech, USA 4 Geological Survey of Iran, Tehran, Iran

49 26th December 2003, M w 6.6 Death toll 26,000

50 Tectonic setting SRTM shadedrelief topography

51 Tectonic setting Nayband fault Dasht-e Lut Gowk fault NEIC Sabzevaran fault Jebal Barez mountains SRTM shadedrelief topography

52 The Bam area Main geomorphic features of the Bam area: 2: The Bam fault a prominent ridge running between Bam and Baravat Bam Baravat LANDSAT-7 ETM 541 false colour green=vegetation 10 km

53 The Bam fault Bam Baravat The Bam fault is marked by a 20m-high ridge

54

55 The Bam fault Bam Baravat Post-earthquake field surveys found only minor cracking at the foot of the ridge

56 The Bam fault Bam Baravat and fault ruptures observed in the north were also minor (< 5 cm offset)

57 The Bam fault? BUT More damage in Bam than Baravat Peak vertical acceleration of ~1g in central Bam Very small surface rupture on Bam fault Bam Baravat LANDSAT-7 ETM 541 false colour green=vegetation 10 km

58 Preliminary InSAR data There is a prominent band of incoherence running S of Bam First Bam interferogram (each colour cycle=2.8cm of deformation) Bam Baravat Constructed from Envisat ASAR data released for free by ESA 10 km

59 The Bam earthquake main fault Low coherence indicates vegetation and surface damage Bam Baravat Interferometric coherence Red = high Blue = low Constructed from Envisat ASAR data released for free by ESA 10 km

60 The Bam earthquake main fault Bam Baravat Surface rupture found in the field right-lateral offsets of ~20 cm 10 km

61 Descending track interferogram Track 120, beam mode I2, 03/12/ /02/2004 Wrapped Unwrapped

62 Ascending track interferogram Track 385, beam mode I2, 16/11/ /01/2004 Wrapped Unwrapped

63 Azimuth offsets Ascending Descending

64 Determining 3D displacements If the 3D displacement at a pixel is given by u = [u x, u y, u z ], then Ascending interferogram, d 1 = los A u Descending interferogram, d 2 = los D u Ascending az. offsets, d 3 = los AO u Descending az. offsets, d 4 = los DO u Which can be rewritten as a matrix equation, d = Lu, and solved for u. See e.g. Wright, T.J, B. Parsons, Z. Lu., Geophys Res. Lett. 30(18), p.1974, 2003

65 Bam earthquake 3D displacements East North Up σ = 0.01 m σ = 0.09 m σ = 0.01 m

66 Bam earthquake 3D displacements East North Up Data Vert. pure s-s

67 Single fault, uniform-slip model About 2m slip on 12 km long fault in top 10 km of crust Ascending model Descending model

68 Single fault model Large residuals, especially in SE quadrant (rms = 25 mm) Ascending residual Descending residual

69 Bam earthquake 3D displacements East North Up Data Vert. pure s-s

70 Is it a single fault? There is an extra amount of displacement in the SE quadrant Ascending interferogram Descending interferogram

71 P SH

72 Two fault model W S E 0 Depth (km) 5 20

73 Two fault model (uniform slip) Ascending model Descending model

74 Two fault model (uniform slip) Improved fit in SE quadrant (rms = 17 mm) Ascending residual Descending residual

75 Variable slip model S N Main fault, M 0 = 9.1 x Nm

76 Variable slip model S Main fault, M 0 = 9.1 x Nm Secondary fault, M 0 = 1.6 x Nm

77 Variable slip model Ascending model Descending model

78 Variable slip model Significantly improved fit (rms = 13 mm) Ascending residual Descending residual

79 P SH

80 two sources one source P SH

81 Two fault model Secondary fault appears to be a southward continuation of the Bam fault Geodesy Seismology Geomorphology LANDSAT-7 ETM 541 false colour green=vegetation

82

83 Arg-e Bam citadel stood for over 300 years and the human history of Bam extends back for ~ 2000 years In all of that time, there had been no reports of earthquakes in the Bam area (Ambraseys & Melville, 2002)

84 Arg-e Bam citadel stood for over 300 years and the human history of Bam extends back for ~ 2000 years In all of that time, there had been no reports of earthquakes in the Bam area (Ambraseys & Melville, 2002) 1 st rupture for this fault OR Geomorphic signature of the fault is buried by flood deposits

85 The Earthquake Cycle

86 Interseismic Strain Why are we interested? Earthquakes Constrain models of continental tectonics and the earthquake cycle Observations biased by distribution of GPS sites Why is it difficult with InSAR SMALL SIGNAL Noise from atmosphere, orbits and topographic errors Maintaining coherence

87 Interseismic Fault creep Hayward Fault, California (Burgmann et al, Science, 2000)

88

89 Seismicity: (Most events with M w > 6; Jackson & McKenzie, 1988) (Most events with M w > 5; Harvard CMT)

90

91 (McClusky et al, JGR, 2000)

92 The North Anatolian Fault ~1300 km long Slip rate from GPS = 24±1 mm/yr 20 th Century earthquake sequence Single strand in most places Good GPS in west, poor in east Orientation good for InSAR

93 Track 35 Processed 12 interferograms 1 yr < t < 4 yrs h a > 250m Wright, Parsons and Fielding, GRL 2001

94 Incoherent

95 Strong atmospheric signal

96 Good

97 Orbital error? If only have 100 km of data ambiguity between orbit error and deformation Extend profile to South impossible to fit linear orbital tilt to expected signal

98

99

100 Stack = I1 + I2 + I3 + I4 Σ (time intervals)

101

102

103 Savage & Burford (1973) Simple dislocation model

104 Simple dislocation model

105 Simple dislocation model

106 Simple dislocation model Determine uncertainties with Monte-Carlo simulation approach Find best fit model to 100 profiles perturbed using the estimated error Range of parameters found uncertainties on parameters Strong trade-off between slip and locking depth Slip = mm/yr Locking Depth = 5-33 km If GPS slip rate of 24 ± 1 m/yr is correct locking depth = 18 ± 6 km Wright, Parsons and Fielding, GRL 2001

107 Track 78 7 Interferogram stack 11.7 years of data Track 35

108 NAF Slip rate: mm/yr Locking Depth: 7-18 km 1σ

109 How wide is the shear zone at depth? 75 km 50 km 25 km 100 km For an 18 km elastic lid, a ~ 60 km wide shear zone at depth (π x lid thickness) cannot be distinguished from a deep, planar fault.

110 Triggered Slip on the Shahdad Thrust, Eastern Iran Iran S Dasht-e-Lut Shahdad Thrust F Gowk Fault

111 SAR Interferogram for the M w 6.6 Fandoqa Earthquake, 14 March 1998, Eastern Iran

112

113 Fielding, E.J., T.J. Wright, J. Muller, B.E. Parsons, and R. Walker, Geology, 32(7), , 2004

114 Postseismic Deformation M w ~ 7.1 Hector Mine Earthquake, October 16, 1999: Rapid Postseismic Relaxation of upper mantle? (Pollitz, Wicks & Thatcher, 2001)

115 Exploiting Envisat and the ERS archive Coseismic interferograms for upper-crustal earthquakes > M5.5 [< 8.0?]. Interseismic deformation for faults with reasonable rates and favourable orientations. Postseismic transients for large earthquakes. Slow earthquakes, triggered aseismic slip and other unexpected phenomena. If ground cover is suitable!

116 The practical class (1) How much slip was there in the Izmit earthquake?

117 The practical class (2) Interpretation and modelling of 3 coseismic interferograms