Lecture 23 Exemplary Inverse Problems including Earthquake Location

Syllabus Lecture 01Describing Inverse Problems Lecture 02Probability and Measurement Error, Part 1 Lecture 03Probability and Measurement Error, Part 2 Lecture 04The L 2 Norm and Simple Least Squares Lecture 05A Priori Information and Weighted Least Squared Lecture 06Resolution and Generalized Inverses Lecture 07Backus-Gilbert Inverse and the Trade Off of Resolution and Variance Lecture 08The Principle of Maximum Likelihood Lecture 09Inexact Theories Lecture 10Nonuniqueness and Localized Averages Lecture 11Vector Spaces and Singular Value Decomposition Lecture 12Equality and Inequality Constraints Lecture 13L 1, L ∞ Norm Problems and Linear Programming Lecture 14Nonlinear Problems: Grid and Monte Carlo Searches Lecture 15Nonlinear Problems: Newton’s Method Lecture 16Nonlinear Problems: Simulated Annealing and Bootstrap Confidence Intervals Lecture 17Factor Analysis Lecture 18Varimax Factors, Empircal Orthogonal Functions Lecture 19Backus-Gilbert Theory for Continuous Problems; Radon’s Problem Lecture 20Linear Operators and Their Adjoints Lecture 21Fréchet Derivatives Lecture 22 Exemplary Inverse Problems, incl. Filter Design Lecture 23 Exemplary Inverse Problems, incl. Earthquake Location Lecture 24 Exemplary Inverse Problems, incl. Vibrational Problems

Purpose of the Lecture solve a few exemplary inverse problems thermal diffusion earthquake location fitting of spectral peaks

Part 1 thermal diffusion

x h ξ 0 temperature in a cooling slab erf(x) x

temperature due to M cooling slabs (use linear superposition)

temperature due to M slabs each with initial temperature m j temperature measured at time t>0 initial temperature

inverse problem infer initial temperature m using temperatures measures at a suite of x s at some fixed later time t d = G m data model parameters

distance, x time, t temperature, T

distance, x time, t initial temperature consists of 5 oscillations

distance, x time, t oscillations still visible so accurate reconstruction possible

distance, x time, t little detail left, so reconstruction will lack resolution

What Method ? The resolution is likely to be rather poor, especially when data are collected at later times damped least squares G -g = [G T G+ε 2 I] -1 G T damped minimum length G -g = G T [GG T +ε 2 I] -1 Backus-Gilbert

What Method ? The resolution is likely to be rather poor, especially when data are collected at later times damped least squares G -g = [G T G+ε 2 I] -1 G T damped minimum length G -g = G T [GG T +ε 2 I] -1 Backus-Gilbert actually, these generalized inverses are equal

What Method ? The resolution is likely to be rather poor, especially when data are collected at later times damped least squares G -g = [G T G+ε 2 I] -1 G T damped minimum length G -g = G T [GG T +ε 2 I] -1 Backus-Gilbert might produce solutions with fewer artifacts

Try both damped least squares Backus-Gilbert

Solution Possibilities 1.Damped Least Squares: Matrix G is not sparse no analytic version of G T G is available M=100 is rather small experiment with values of ε 2 mest=(G’*G+e2*eye(M,M))\(G’*d) 2. Backus-Gilbert use standard formulation, with damping α experiment with values of α

Solution Possibilities 1.Damped Least Squares: Matrix G is not sparse no analytic version of G T G is available M=100 is rather small experiment with values of ε 2 mest=(G’*G+e2*eye(M,M))\(G’*d) 2. Backus-Gilbert use standard formulation, with damping α experiment with values of α try both

distance time TrueDamped LSBackus-Gilbert estimated initial temperature distribution as a function of the time of observation

distance time TrueDamped LSBackus-Gilbert estimated initial temperature distribution as a function of the time of observation Damped LS does better at earlier times

distance time TrueDamped LSBackus-Gilbert estimated initial temperature distribution as a function of the time of observation Damped LS contains worse artifacts at later times

Damped LS distance Backus-Gilbert distance model resolution matrix when for data collected at t=10

Damped LS distance Backus-Gilbert distance model resolution matrix when for data collected at t=10 resolution is similar

Backus-Gilbert distance Damped LS distance model resolution matrix when for data collected at t=40

Backus-Gilbert distance Damped LS distance model resolution matrix when for data collected at t=40 Damped LS has much worse sidelobes

Part 2 earthquake location

z x P S s r ray approximation vibrations travel from source to receiver along curved rays

z x P S s r ray approximation vibrations travel from source to receiver along curved rays P wave faster S wave slower P, S ray paths not necessarily the same, but usually similar

z x P S s r T S = ∫ ray (1/v S ) d travel time T integral of slowness along ray path T P = ∫ ray (1/v P ) d

arrival time = travel time along ray + origin time

data earthquake location 3 model parameters earthquake origin time 1 model parameter

arrival time = travel time along ray + origin time explicit nonlinear equation 4 model parameters up to 2 data per station

arrival time = travel time along ray + origin time linearize around trial source location x (p) t i P = T i P (x (p),x (i) ) + [∇T i P ] ∆x + t 0 trick is computing this gradient

x (0) x (1) x (0) r r ΔxΔx ΔxΔx ray Geiger’s principle [∇T i P ] = -s/v unit vector parallel to ray pointing away from receiver

linearized equation

z x z x All rays leave source at the same angle All rays leave source at nearly the same angle Common circumstances when earthquake far from stations

then, if only P wave data is available these two columns are proportional to one-another (no S waves)

z x depth and origin time trade off shallow and early deep and late

Solution Possibilities 1.Damped Least Squares: Matrix G is not sparse no analytic version of G T G is available M=4 is tiny experiment with values of ε 2 mest=(G’*G+e2*eye(M,M))\(G’*d) 2. Singular Value Decomposition to detect case of depth and origin time trading off test case has earthquakes “inside of array”

x, km y, km z, km

Part 3 fitting of spectral peaks

counts velocity, mm/s counts velocity, mm/s xx typical spectrum consisting of overlapping peaks counts

what shape are the peaks?

try both use F test to test whether one is better than the other

what shape are the peaks? data 3 unknowns per peak data 3 unknowns per peak both cases: explicit nonlinear problem

linearize using analytic gradient

issues how to determine number q of peaks trial A i c i f i of each peak

our solution have operator click mouse computer screen to indicate position of each peak

MatLab code for graphical input K=0; for k = [1:20] p = ginput(1); if( p(1) < 0 ) break; end K=K+1; a(K) = p(2)-A; v0(K)=p(1); c(K)=0.1; end

velocity, mm/s counts Lorentzian Gaussian

Results of F test Fest = E_normal/E_lorentzian: P(F =Fest) = Lorentzian better fit to % certainty