Professional Documents
Culture Documents
Inverse Problems in Geophysics
Inverse Problems in Geophysics
Examples in Geophysics
- Traveltime inverse problems
- Seismic Tomography
- Location of Earthquakes
- Reflection Seismology
Forward Problem
Model m Data d
Inverse Problem
X
X
X
X
Gravimeter
?
Gρ (r ' )
Φ (r ) = ∫ dV '
r − r'
-> given a density model we can predict the gravity field at the surface!
One approach:
Use the forward solution to calculate many models for a rectangular box
situated somewhere in the ground and compare the theoretical (synthetic)
data to the observations.
But ...
-> Let us
sand
gold
Nonlinear Inverse Problems Introduction
ρ (d , m)θ (d , m)
Treasure Hunt – Non-uniqueness
σ ( d , m) = k
µ ( d , m)
X X X
X X
Gravimeter
Gravimeter
Exercise:
Gravimeter
Forward Problem
True Model m
Appraisal Data d
Problem
Inverse Problem
Estimated Model
~
m
3. Abel problem: find the shape of a hill from the time it takes for a ball
to go up and down a hill for a given initial velocity.
P(x,z)
dz’
ds
x
Find the shape of the hill !
z
P(x,z)
d
1 2 z’ ds
gz = v 0
At any point: 2 x
1
At z-z’: − mg ( z − z ' ) = m ( ds / dt ) 2
2
z
After ds / dz '
integration:
t (z) = ∫
0 2 g (z − z' )
dz '
z
P(x,z)
z
ds / dz' dz’
t(z) = ∫
0 2 g (z − z')
dz' ds
a
1 d t ( z ) dz
f (z') = −
π dz ' ∫
z' z − z'
∆
r 2 / c 2 ( z) − p 2 ⎛ r0 ⎞ 1 1 −1 ⎛ p ⎞
r1
The integral of the inverse problem contains only terms which can be
obtained from observed T(∆) plots. The quantity ξ1=p1=(dT/d∆)1 is the
slope of T(∆) at distance ∆1. The integral is numerically evaluated
with discrete values of p(∆) for all ∆ from 0 to ∆1. We obtain a value
for r1 and the corresponding velocity at depth r1 is obtained through
ξ1=r1/v1.
d1 = A1 (m1 ,..., mm )
This is the nonlinear
d 2 = A2 (m1 ,..., mm )
formulation.
#
d n = An (m1 ,..., mm )
If the functions Ai(mj) between model and data are linear we obtain
d i = Aij m j
or
d = Am
in matrix form. If the functions Ai(mj) between model and data are
mildly non-linear we can consider the behavior of the system around
some known (e.g. initial) model mj0:
∂Ai
d i = Al (m ) +
0
∆m j + ...
∂m j
j
m 0j
∆d i = d i − Ai (m 0j )
∆d = A∆m
Nonlinear Inverse Problems Introduction
ρ (d , m)θ (d , m)
Linear(ized) Inverse Problems
σ ( d , m) = k
µ ( d , m)
∆d = A∆m
Interpretation of this result:
d = g(m)
The only option we have here is to try and go – in a
sensible way – through the whole model space and
calculate the misfit function
L = d − g(m)
and find the model(s) which have the minimal misfit.
Data vector d:
Traveltimes of phases observed at
stations of the world wide
seismograph network
Model m:
3-D seismic velocity model in the
Earth’s mantle. Discretization using
splines, spherical harmonics,
Chebyshev polynomials or simply
blocks.
Seismometers
Data vector d:
Traveltimes observed at various (at
least 3) stations above the
earthquake
Model m:
3 coordinates of the earthquake
location (x,y,z).
Model m:
the seismic velocities of the
subsurface, impedances, Poisson’s
ratio, density, reflection
coefficients, etc.