GEOPHYSICAL RESEARCH LETTERS, VOL. 32, L09302, doi:10.

1029/2005GL022601, 2005

Electrokinetic effect of the Loma Prieta earthquake calculated by an

entire-Earth FDTD solution of Maxwell’s equations
Jamesina J. Simpson and Allen Taflove
Department of Electrical and Computer Engineering, Northwestern University, Illinois, USA
Received 2 February 2005; revised 31 March 2005; accepted 6 April 2005; published 10 May 2005.

[1] We report what we believe to be the first three- work described here is to provide a more rigorous analysis
dimensional computational solution of the full-vector of this issue than previously published, thereby placing
Maxwell’s equations for hypothesized pre-seismic hypothetical pre-seismic EM phenomena on a firmer
electromagnetic phenomena propagated within the entire physical basis.
Earth-ionosphere cavity. Periodic boundary conditions are [3] This letter reports what we believe to be the first
used in conjunction with a variable-cell finite-difference three-dimensional (3-D) computational solution of the
time-domain (FDTD) space lattice wrapping around the full-vector Maxwell’s equations for hypothesized pre-
complete Earth-sphere and extending ±100 km radially seismic EM phenomena within the entirety of the
from sea level. This technique permits a direct time- Earth-ionosphere. Applying the finite-difference time-do-
domain calculation of round-the-world ULF/ELF main (FDTD) method [Taflove and Hagness, 2000], we
propagation accounting for arbitrary horizontal as well obtain a direct time-domain calculation of round-the-
as vertical geometrical and electrical inhomogeneities/ world ULF/ELF propagation accounting for arbitrary
anisotropies of the excitation, ionosphere, lithosphere, and horizontal as well as vertical geometrical and electrical
oceans. In this study, we model electrokinetic currents at inhomogeneities/anisotropies of the excitation, ionosphere,
depths of 2.5 km and 17 km near the hypocenter of the lithosphere, and oceans. Using this rigorous numerical
Loma Prieta earthquake and compare the FDTD- model, we accommodate the complete physics introduced
calculated surface magnetic field to analytical results by impulsive electromagnetic wave propagation through
and measurements previously reported in the literature. the conductive Earth and subsequent reverberation
We accommodate the complete physics introduced by through the entire Earth-ionosphere waveguide, and hence
impulsive electromagnetic wave propagation through the illustrate the importance of solving the full Maxwell’s
conductive Earth, and hence illustrate the importance of equations when modeling current sources within the
solving the full Maxwell’s equations when modeling Earth’s crust.
current sources within the Earth’s crust. Our calculated [4] Previously, we reported applying FDTD to efficiently
spectra agree qualitatively with those reported by Fraser- model the complete Earth-sphere at ELF using a 3-D,
Smith et al. (1990). Citation: Simpson, J. J., and A. Taflove spherical-coordinate, latitude-longitude space lattice with
(2005), Electrokinetic effect of the Loma Prieta earthquake periodic boundary conditions [Simpson and Taflove, 2002a,
calculated by an entire-Earth FDTD solution of Maxwell’s 2002b, 2003, 2004a]; see http://www.ece.northwestern.edu/
equations, Geophys. Res. Lett., 32, L09302, doi:10.1029/ ecefaculty/taflove/3Dmovietext@gif.avi. To minimize the
2005GL022601. effects of increasing space-cell eccentricity upon approach-
ing a pole, we employed an adaptive cell-combination
technique applied to adjacent grid cells in the east-west
1. Introduction direction, thereby allowing maintenance of the time-step
[2] Anomalous ultra low frequency (ULF: 0– 3 Hz) and at nearly the level permitted by the Courant stability
extremely low-frequency (ELF: 3 Hz– 300 Hz) electro- condition for the square equatorial cells. This technique
magnetic (EM) wave phenomena occurring prior to major permits a standard laboratory computer with 2 GB
earthquakes have been of particular interest for many years of random access memory to generate high-resolution
[Park et al., 1993; Johnston, 1997]. A number of physical (
40  40  5 km) modeling results for global, fully
mechanisms related to hypothetical earthquake precursors 3-D, impulsive ULF/ELF propagation within the
have been proposed to explain the origin of such obser- entire Earth-ionosphere cavity to within ±100 km of sea
vations [Park et al., 1993]. The existence and nature of level.
any EM signatures associated with these mechanisms [5] In this study, we apply our entire-Earth FDTD
currently remains unproven and unclear, however. For numerical model to calculate EM emissions generated
example, analysis of potential signatures were presented by electrokinetic currents located near the hypocenter of
by Molchanov et al. [1995] and Kulchitsky et al. the Loma Prieta earthquake. Such currents have been
[2004]; however, Molchanov et al. [1995] neglected wave proposed to explain one of the primary examples of
propagation, and Kulchitsky et al. [2004] provided a scalar hypothetical precursory EM phenomena to date: the
two-dimensional analysis which neglected the Earth’s anomalous magnetic fields reported by Fraser-Smith et
curvature. Our overall goal for the novel EM modeling al. [1990] prior to the Loma Prieta earthquake. We
compare the FDTD surface magnetic field calculations
Copyright 2005 by the American Geophysical Union. to analytical results and measurements previously
0094-8276/05/2005GL022601$05.00 reported in the literature, and find that our calculated

L09302 1 of 4
L09302 SIMPSON AND TAFLOVE: LOMA PRIETA EARTHQUAKE L09302

a pole, allowing the user to specify a maximum allowable

cell eccentricity.
[8] The wrap-around or joining of the east and west
edges of the lattice occurs along a specific line of constant
longitude, or meridian. This joining is, in effect, a periodic
boundary condition applied at each j-row of lattice cells,
whether trapezoids or triangles [Simpson and Taflove,
2004a].
[9] Given the above assumptions, Ampere’s Law in
integral form can be applied to develop an FDTD time-
stepping relation for each electric E field components of the
grid [Simpson and Taflove, 2004a]. Similarly, Faraday’s
Law in integral form can be applied to develop an FDTD
time-stepping relation for the magnetic H field components
of the grid [Simpson and Taflove, 2004a].
Figure 1. General layout of the of the 3-D FDTD
Maxwell’s equations space-lattice covering the complete
Earth-sphere as seen from a TM plane at a constant radial 3. Details of the Electrokinetic Study
coordinate. [10] In our present study, we focus on modeling EM
emissions resulting from proposed electrokinetic currents in
the Earth’s crust near the hypocenter of the Loma Prieta
spectra agree qualitatively with those reported by Fraser- earthquake along the San Andreas Fault. These current
Smith et al. [1990]. sources are believed to arise when fluid flows through a
porous medium in the presence of an electrical double layer
at a solid-liquid interface [Park et al., 1993]. Fenoglio et al.
2. The 3-D Latitude-Longitude FDTD Grid [1995] previously modeled electrokinetic currents using a
[6] Figure 1 illustrates the general layout of the FDTD Biot-Savart model to explain the behavior of the anomalous
space-lattice as seen from the transverse magnetic (TM) precursory EM data reported by Fraser-Smith et al. [1990]
plane at a constant radial coordinate [Simpson and Taflove, prior to the Loma Prieta earthquake. For their study, they
2004a]. The lattice is a logically Cartesian 2M  M  K - used a layered half-space geometry to characterize the
cell arrangement, where M is a power of 2. We see that the surface magnetic fields directly above the electrokinetic
grid cells follow lines of constant latitude, q = constant, currents. Subsequently, Majaeva et al. [1997] modeled
where q is the usual spherical angle measured from the north electrokinetic currents using a layered cylindrical model
pole; and along lines of constant longitude, f = constant, that also accounted for diffusion.
where f is the usual spherical azimuthal angle measured [11] In the present study, we report a significantly more
from a specified prime meridian. In this manner, each TM rigorous calculation of electrokinetic EM emissions relative
plane of the grid shown in Figure 1 is comprised of to those of Fenoglio et al. [1995] and Majaeva et al. [1997].
isosceles trapezoidal cells away from the north and south Using the entire-Earth 3-D FDTD grid briefly described
poles [Simpson and Taflove, 2004a], and isosceles triangular above, we solve the full-vector Maxwell’s equations to
cells at the poles [Simpson and Taflove, 2004a]. Similarly, account for the complete physics introduced by impulsive
each transverse electric (TE) plane at a constant radial wave propagation through a conductive Earth and subse-
coordinate is comprised of isosceles trapezoidal cells away quent round-the-world propagation, which has been shown
from the north and south poles, and a polygon cell at each to be an important factor at ULF/ELF [Simpson and Taflove,
pole [Simpson and Taflove, 2004a]. 2004a]. Our model utilizes topographic and bathymetric
[7] We choose to have the same angular increment in data from the NOAA-NGDC ‘‘Global Relief CD-ROM’’.
latitude, Dq = p/M, for each cell in the grid. Thus, the These data are mapped onto our 3-D space lattice with an
south-north span of each trapezoidal or triangular grid cell assumed resolution of 40  40  5 km at the equator. For
is Dsn = pR/M, where R is the radial distance from the the lithosphere, conductivity values are assigned according
center of the Earth. To maintain square or nearly square to Hermance [1995], depending upon the location of the E
grid cells near the equator, we select the baseline value of component (i.e., below an ocean or within a continent). For
the angular increment in longitude, Df, to equal Dq. the atmosphere, both day- and nighttime exponential con-
However, this causes the west-east span of each cell, ductivity profiles used by Bannister [1985] are assumed,
Dwe = RDf sin q, to be a function of q. This could be with early evening (the timing of the Loma Prieta earth-
troublesome for cells near the north and south poles where quake) occurring along the San Andreas Fault.
q ! 0 and q ! p, respectively. There, the geometrical [12] Fenoglio et al. [1995] modeled a current associated
eccentricity of each cell, Dsn/Dwe = Dq/(Df sin q), with fluid flow into a 200 m by 1000 m shear fracture at a
would become quite large, and the numerical stability and depth of 17 km. According to their model, repeated pulses
efficiency of the FDTD algorithm would be degraded. We are generated by successive 25-m fractures along the 200 m-
mitigate this problem by merging pairs of adjacent cells of long compartment. In the present study, we first simulate
the TM plane in the west-east direction, effectively halving one such pulse occurring along a 17-km-deep horizontal
the cell eccentricity [Simpson and Taflove, 2004a]. This electric field at 121.88 W, 37.04 N, near the hypocenter of
process can be repeated several times as the grid approaches the Loma Prieta earthquake. This is essentially the space-

2 of 4
L09302 SIMPSON AND TAFLOVE: LOMA PRIETA EARTHQUAKE L09302

Figure 3. Comparison of the time waveforms for the

Figure 2. Comparison of the FDTD-calculated time horizontal surface magnetic field component due to 17-km
waveforms for the horizontal surface H-field component deep electrokinetic currents associated with eight successive
resulting from one electrokinetic current source: (1) 17-km- 25-m fractures occurring 1.65 seconds apart along a 200 m-
deep; and (2) 2.5-km-deep. See color version of this figure long compartment: (1) as predicted by the Biot-Savart
in the HTML. model of Fenoglio et al. [1995]; and (2) as predicted by the
3-D FDTD Maxwell’s equations model. See color version
time Green’s function of the global Earth-ionosphere wave- of this figure in the HTML.
guide as excited at the hypocenter. To ensure a smooth onset
of the excitation, the current is assumed to linearly increase using a simple Biot-Savart model), with the predictions of
to its maximum value at 1000 Dt, where Dt = 3.0 ms. our FDTD Maxwell’s equations model. For both sets of
[13] Our present grid resolution at 40  40  5 km results, each current source pulse occurs after a time delay
requires us to model the 25-m current source onto a 40-km- of about 1.65 seconds. As shown in Figure 3, the Maxwell’s
long electric field component. This does not, however, upset equations model yields a vastly different temporal response.
our calculated surface magnetic field values after normali- We believe that this difference has two primary and inter-
zation. In fact, we can determine the resulting surface linked causes: (i) The FDTD Maxwell’s equations model
magnetic field due to successive 25-m fractures along the properly accounts for temporal integration due to EM
200 m-long compartment by modeling multiple current diffusion through 17 km of lithosphere; (ii) The FDTD
signals on the same buried horizontal electric field compo- Maxwell’s equations model properly accounts for reverber-
nent. This is due to the alignment of the current sources ation within the Earth-ionosphere waveguide, which has
along the same general direction, allowing each to contrib- low propagation loss below 0.2 dB/Mm for frequencies
ute linearly in the far-field to the overall dipole moment. below 10 Hz [Bannister, 1984]. In this manner, the
The total surface magnetic field due to any number of instantaneous response of the surface H-field is properly
cascading, parallel pulses occurring within the span of one
grid-cell component is therefore simply the superposition of
multiple time-delayed versions of the response caused by a
single electrokinetic current source.

4. Discussion of Results
[14] We now report the results of our full-vector, 3-D
FDTD Maxwell’s equations model. Figure 2 shows the
time-waveforms of the horizontal surface H-field compo-
nent generated by a single electrokinetic pulse at a depth of
2.5 km and 17 km. Using the Lemieux parallel processing
supercomputer at the Pittsburgh Supercomputing Center,
we calculated the initial 25 seconds using our FDTD
model, and thereafter applied Prony’s method [Academic
Technologies, 2002] to extend the tail of the response to
its ultimate decay. The time waveform generated by the Figure 4. Comparison of the spectra for the horizontal
17-km-deep source is subsequently used as an equivalent surface H-field component: (1) measured by Fraser-Smith
Green’s function in the discussion below. Interestingly, we et al. [1990] and A. C. Fraser-Smith (personal commu-
see that the late-time response of the deeper source is nication, 2004); (2) FDTD result for eight electrokinetic
higher due to the longer diffusion time within the lossy currents associated with successive 25-m fractures occur-
crustal medium. ring 1.65 seconds apart (corresponding to the comparatively
[15] Figure 3 compares the normalized time waveform of large spectral peak at 0.61 Hz) along a 200 m-long
the response of the horizontal surface H-field component compartment and at a depth of 17-km-deep; and (3) FDTD
due to eight successive 17-km-deep electrokinetic current result for one electrokinetic current at 17-km depth. See
pulses as reported by Fenoglio et al. [1995] (calculated color version of this figure in the HTML.

3 of 4
L09302 SIMPSON AND TAFLOVE: LOMA PRIETA EARTHQUAKE L09302

influenced by current sources occurring much earlier in (ACES), available at http://www.quakes.uq.edu.au/ACES/).

time. Our ultimate goal is to provide as rigorous as possible a
[16] Figure 4 compares H-field frequency spectra from physics basis for proposed earthquake prediction schemes
0.01 – 1.0 Hz for three sets of data: (i) measurements employing either land-based or satellite-based detectors of
reported by Fraser-Smith et al. [1990] and A. C. Fraser- ULF/ELF anomalies.
Smith (personal communication, 2004); (ii) FDTD calcula-
tion for one electrokinetic current pulse located at a depth of [19] Acknowledgment. This work was partially supported by
17 km, as shown in Figure 2; and (iii) FDTD calculation for National Computational Science Alliance under grant number
DMS040006N and utilized the Dell Xeon Linux Cluster.
eight successive electrokinetic current pulses (at intervals of
1.65 seconds) located at a depth of 17 km. All three spectra
are normalized at 0.01 Hz to illustrate their relative rates of References
decay with increasing frequency, and are graphed on a Academic Technologies (2002), MATLAB: The language of technical
computing, version 6.5.0.180913a, release 13, Northwestern Univ.,
common logarithmic frequency and amplitude scale. There Evanston, Ill.
is qualitative agreement of the measured and calculated Bannister, P. (1984), ELF propagation update, IEEE J. Oceanic Eng., 9(3),
results. Assuming that the hypothesized electrokinetic cur- 179 – 188.
Bannister, P. (1985), The determination of representative ionospheric
rent actually underlies the Fraser-Smith et al.observations, conductivity parameters for ELF propagation in the Earth-ionosphere
the deviations between the FDTD-calculated spectra and waveguide, Radio Sci., 20(4), 977 – 984.
the measurements have two probable causes: (i) assumption Fenoglio, M. A., M. J. S. Johnston, and J. D. Byerlee (1995), Magnetic and
electric fields associated with changes in high pore pressure in fault
of a lithospheric conductivity in the FDTD model that zones: Application to the Loma Prieta ULF emissions, J. Geophys.
exceeds the actual value; and (ii) a measurement artifact Res., 100, 12,951 – 12,958.
reported by Fraser-Smith et al. [1990] wherein instrumen- Fraser-Smith, A. C., A. Bernardi, P. R. McGill, M. E. Ladd, R. A. Helliwell,
tation saturation likely occurred at the lowest observed and O. G. Villard Jr. (1990), Low-frequency magnetic field measurements
near the epicenter of the Ms 7.1 Loma Prieta earthquake, Geophys. Res.
frequencies. In the first case, an excessive lithospheric Lett., 17, 1465 – 1468.
conductivity would cause the FDTD-calculated spectra at Hermance, J. (1995), Electrical conductivity of the crust and mantle, in
higher frequencies to be unduly attenuated, thereby bending Global Earth Physics: A Handbook of Physical Constants, AGU Ref.
Shelf, vol. 1, edited by T. J. Ahrens, pp. 190 – 205, AGU, Washington,
the calculated curves downward below the Fraser-Smith et D. C.
al. data. In the second case, the Fraser-Smith et al. data Johnston, M. J. S. (1997), Review of electric and magnetic fields accom-
point at 0.01 Hz would have too small a value, thereby panying seismic and volcanic activity, Surv. Geophys., 18, 441 – 475.
Kulchitsky, A. V., Y. Ando, and M. Hayakawa (2004), Numerical analysis
unduly flattening their measured spectra. on the propagation of ULF/ELF signals in the lithosphere with highly
conductive layers, Phys. Chem. Earth, 29, 495 – 500.
Majaeva, O., Y. Fujinawa, and M. E. Zhitomirsky (1997), Modeling of
5. Conclusions and Ongoing Work non-stationary electrokinetic effect in a conductive crust, J. Geomagn.
Geoelectr., 49, 1317 – 1326.
[17] We have presented what we believe to be the first Molchanov, O. A., M. Hayakawa, and V. A. Rafalsky (1995), Penetration
rigorous 3-D computational solution of the full-vector characteristics of electromagnetic emissions from an underground seismic
Maxwell’s equations for hypothesized pre-seismic EM source into the atmosphere, ionosphere, and magnetosphere, J. Geophys.
phenomena propagated within the entire Earth-ionosphere Res., 100, 1691 – 1712.
Park, S. K., M. J. S. Johnston, T. R. Madden, F. D. Morgan, and H. F.
cavity. Using the robust FDTD method, we simulated Morrison (1993), Electromagnetic precursors to earthquakes in the ULF
electrokinetic currents near the hypocenter of the Loma band: A review of observations and mechanisms, Rev. Geophys., 31,
Prieta earthquake and compared the calculated surface 117 – 132.
Simpson, J., and A. Taflove (2002a), Two-dimensional FDTD modeling of
H-field with analytical results and measurements previously impulsive ELF antipodal propagation about the Earth-sphere, Proc. IEEE
reported in the literature. Our entire-Earth FDTD grid Antennas Propag. Soc. Int. Symp., 3, 678 – 681.
allowed us to properly model complex, interlinked phe- Simpson, J., and A. Taflove (2002b), Two-dimensional FDTD model of
nomena due to EM diffusion through the lithosphere and antipodal ELF propagation and Schumann resonance of the Earth, IEEE
Antennas Wireless Propag. Lett., 1, 53 – 56.
EM wave reverberation due to round-the-world propaga- Simpson, J., and A. Taflove (2003), Global three-dimensional FDTD
tion within the Earth-ionosphere cavity. We found that the modeling of impulsive ELF propagation about the Earth, Proc. IEEE
FDTD-calculated spectra for the surface H-field qualita- Antennas Propag. Soc. Int. Symp., 4, 940 – 944.
Simpson, J., and A. Taflove (2004a), Three-dimensional FDTD modeling
tively agree with measured data recorded by Fraser-Smith of impulsive ELF propagation about the Earth-sphere, IEEE Trans. An-
et al. in the immediate hours prior to the earthquake tennas Propag., 52, 443 – 451.
[Fraser-Smith et al., 1990; A. C. Fraser-Smith, personal Simpson, J., and A. Taflove (2004b), Efficient modeling of impulsive ELF
antipodal propagation about the Earth sphere using an optimized two-
communication, 2004]. There are remaining uncertainties dimensional geodesic FDTD grid, IEEE Antennas Wireless Propag. Lett.,
in this comparison due to inexact knowledge of the 3, 215 – 218.
lithosphere conductivity and possible instrumentation Simpson, J. J., R. P. Heikes, and A. Taflove (2005), 3-D FDTD modeling of
artifacts acknowledged by Fraser-Smith et al. [1990]. ULF/ELF propagation within the global Earth-ionosphere cavity using an
optimized geodesic grid, Proc. IEEE Antennas Propag. Soc. Int. Symp.,
[18] Our ongoing work in this area includes developing in press.
an alternative hexagonal/pentagonal 3-D geodesic FDTD Taflove, A., and S. C. Hagness (2000), Computational Electrodynamics:
grid with superior computational efficiency [Simpson and The Finite-Difference Time-Domain Method, 2nd ed., Artech House,
Norwood, Mass.
Taflove, 2004b; Simpson et al., 2005]. We are also planning
to couple our EM models to emerging whole-Earth geo-
physical codes now under development for the study of 
J. J. Simpson and A. Taflove, Department of Electrical and Computer
seismic phenomena, including potential earthquake pre- Engineering, Northwestern University, 2145 Sheridan Road, M367,
cursors (APEC Cooperation for Earthquake Simulation Evanston, IL 60208, USA. (j-simpson@northwestern.edu)

4 of 4