Você está na página 1de 16

TASK OF ELECTROMAGNETIC EXPLORATION

Inversion Method on Magnetotelluric Data Processing

Arranged by:
Fuad Aulia Bahri

3713100007

M. Arif Budiman

3713100048

Lecturer:
Wien lestari, S.T., M.T.

Departement of Geophysical Engineering


Faculty of Civil Engineering and Planology
Sepuluh Nopember Institute of Teknologi
2016

Some inversion methods that are used for MT modelling


1. Monte Carlo
This method is one of inversion technique that concerned of experiments on random
numbers. Monte Carlo also be used for deterministic problem, for example evaluating
multidimensional integrals. Monte Carlo meant generating discrete Earth Models in a
uniform random between upper and lower bounds, which were chosen a priori. For
electromagnetic data, such as Magnetotelluric, this method we applied to inversion data for a
modelling have topography models and to modelling another. This diagram below explain the
process of Monte Carlo Inversion :

Picture.1. Monte Carlo FlowChart Process

As for the application of this method are:

Graphic, especially for ray tracing


light transport modelling in a multi-layered tissues (MCML)
Monte Carlo method on financial sector
simulation of protein structure prediction
Used to model the flow of carrier transport In the research of semiconductor

equipment
Genetic Mapping involving hundreds of genetic markers and QTL analysis

2. Occam
A simple model containing the essential properties of all possible models fitting the
field data. A large number of geoelectric models could match the observed data, some of
which may be very complex. When attempting to achieve a better fit between small portions
of the calculated and observed curve of (), (), the complexity of the obtained model
increases and the results are often unreliable. The model should be as complex as the
medium, but not more complex. The algorithm departs from a halfspace and produces a
stratified medium. The resistivities vary until an adequate fit between the field and calculated
curves is achieved. Roughness (or the inverse of softness) is defined in terms of the first and
the second derivatives of the electric resistivity with respect to the depth as
R1= (

dm 2
) dz
dz
2

R2= ( d2 m dz 2) dz
Where :
m (z)

= the resistivity or log resistivity

= the depth

R1 and R2

= roughness functions

3. Bostick

This Inversion method is the fastest and easiest way to estimate the variation of
resistivity against depth that is directly from curve of pseudo resistivity sounding. This
method are derived from analytic relation of resistivity, frequency, and investigation depth or
skin depth. But, this method can only be done as modelling and interpreting for preliminary
identification.
On this least-square inversion method, preliminary model is iteratively modified to
obtain a response model that fits the data. Their approximation or linearized non-linear
function between data and model parameters lead Bostick inversion method to be very
sensitive to the selection of the initial model. Therefore, the initial model is usually
determined from the results of the indirect modeling or inversion results Bostick. Biostick
Inversion:

Where,
z

= skin depth

= apparent resistivity

= frequency

1 D MT Processing Data Using Forwad Modelling (Matlab)


The script were taken from http://www.digitalearthlab.com/tutorial/tutorial-1d-mtforward/ that is written by Andrew Pettick, 2013. This script explain about how to processing
magnetotelluric data. But, this script is not using inverse modelling process, but using
forward modelling. This script also obtain constantly output because it doesn't have iteration.

Main Script

clear all;
clc;
figure(1);
close(1);
disp('====================================');
disp('1D MAGNETOTELLURIC MODELLING PROGRAM');
disp('====================================');
disp('
LAST UPDATED 29TH DECEMBER 2013 ');
disp('
DEVELOPED BY ANDREW PETHICK
');
disp('
WWW.DIGITIALEARTHLAB.COM
');
disp('====================================');
disp('');
disp('
licensed under WTFPL')
disp('');
data = load('data.txt');
dataFrequencies = data(:,1);
dataApparentResistivities = data(:,2);
dataError = data(:,3);

Input the
required MT Data

%data taken from telford


resistivities = [300 2500 0.8 3000 2500];
thicknesses = [200 400 40 500];

Syntax is for
calculating the
apparent resistivity

dataModelledApparentResistivities =
zeros(length(dataApparentResistivities),1);
for i = 1 : length(dataFrequencies)
frequency = dataFrequencies(i);
[apparentResistivity] = modelMT(resistivities, thicknesses,
frequency);
dataModelledApparentResistivities(i) = apparentResistivity;
end

%Calculate Misfit
misfit = zeros(length(dataApparentResistivities),1);
for i = 1 : length(dataFrequencies)
d = dataApparentResistivities(i);
m = dataModelledApparentResistivities(i);
e = dataError(i);
misfit(i) = ((m - e)^2)/((d*e/10))^2;
end
%Create array of frequencies from 10^-4 to 10^4Hz
%i.e. frequencies = 10^n, where n = -4,-3.9,-3.8....4
logFrequencies = -4:0.1:4;
frequencies = 10.^logFrequencies;

apparentResistivities = zeros(length(frequencies),1);
for i = 1 : length(frequencies)
frequency = frequencies(i);
[apparentResistivity] = modelMT(resistivities, thicknesses,

frequency);
apparentResistivities(i) = apparentResistivity;
end
earthModelText = repmat(cellstr(''), length(resistivities + 2),1);
earthModelText{1} = ['Misfit = ' num2str(sum(misfit)) '%'];
earthModelText{2} = ['Layer
' ' Resistivity
' ' Thickness'];
for i = 1 : length(resistivities)
resistivity = resistivities(i);
mainText = ['Layer ' num2str(i) '
' num2str(resistivity) ' Ohm m
'];
earthModelText{i + 2} = mainText;
if(i == length(resistivities))
earthModelText{i+2} = [earthModelText{i+2} 'Halfspace'];
else
earthModelText{i+2} = [earthModelText{i+2} num2str(thicknesses(i))
'm'];
end
end
%Plot for each resistivity
scrsz = get(0,'ScreenSize');
figure(1)
set(1,'Position',[50 50 900 400]);
yy = spline(dataFrequencies,dataApparentResistivities,dataFrequencies);
subplot(1, 4, [1,3])
loglog(frequencies,apparentResistivities,'-r','LineWidth',2);
hold on
loglog(dataFrequencies,dataApparentResistivities,'bs','LineWidth',1,'MarkerSize',3,'MarkerFaceColor','b');
loglog(dataFrequencies,dataModelledApparentResistivities,'rs','LineWidth'
,1,'MarkerSize',3,'MarkerFaceColor','r');
%set(get(AX(2),'Ylabel'),'String','Error (%)');
legend('Field Data','Modelled Data');
hold off
title({['Apparent Resistivity (Ohm m) vs Frequency(Hz)']});
xlabel('Frequency (Hz)');
ylabel('Apparent Resistivity (Ohm m)');
subplot(1,4,4);

text(0,0.5,earthModelText);
axis off
xlabel('Frequency (Hz)');
ylabel('B-Field (T)');

From that script, Authors will need function (fx) to input MT equation. This script
below is contents from model MT.M
Script Model MT.M
%
%
%
%
%

Digital Earth Lab


www.DigitalEarthLab.com
Written by Andrew Pethick 2013
Last Updated October 29th 2013
Licensed under WTFPL

function [apparentResistivity, phase] = modelMT(resistivities,


thicknesses,frequency)
mu = 4*pi*1E-7; %Magnetic Permeability (H/m)
w = 2 * pi * frequency; %Angular Frequency (Radians);
n=length(resistivities); %Number of Layers
impedances =
%Layering in
% Layer
% Layer 1
% Layer 2
% Layer 3
% Layer 4
% Basement
%

zeros(n,1);
this format
j
1
2
3
4
5

% Steps for modelling (for each geoelectric model and frequency)


% 1. Compute basement impedance Zn using sqrt((w * mu * resistivity))
% 2. Iterate from bottom layer to top(not the basement)
% 2.1. Calculate induction parameters
% 2.2. Calculate Exponential factor from intrinsic impedance
% 2.3 Calculate reflection coeficient using current layer
%
intrinsic impedance and the below layer impedance
% 3. Compute apparent resistivity from top layer impedance
%
apparent resistivity = (Zn^2)/(mu * w)
%Symbols
% Zn - Basement Impedance
% Zi - Layer Impedance
% wi - Intrinsic Impedance
% di - Induction parameter
% ei - Exponential Factor
% ri - Reflection coeficient
% re - Earth R.C.
%Step 1 : Calculate basement impedance
Zn = sqrt(sqrt(-1)*w*mu*resistivities(n));
impedances(n) = Zn;

%Iterate through layers starting from layer j=n-1 (i.e. the layer above
the basement)
for j = n-1:-1:1
resistivity = resistivities(j);
thickness = thicknesses(j);
% 3. Compute apparent resistivity from top layer impedance
%Step 2. Iterate from bottom layer to top(not the basement)
% Step 2.1 Calculate the intrinsic impedance of current layer
dj = sqrt(sqrt(-1)* (w * mu * (1/resistivity)));
wj = dj * resistivity;
% Step 2.2 Calculate Exponential factor from intrinsic impedance
ej = exp(-2*thickness*dj);
% Step 2.3 Calculate reflection coeficient using current layer
%
intrinsic impedance and the below layer impedance
belowImpedance = impedances(j + 1);
rj = (wj - belowImpedance)/(wj + belowImpedance);
re = rj*ej;
Zj = wj * ((1 - re)/(1 + re));
impedances(j) = Zj;
end
% Step 3. Compute apparent resistivity from top layer impedance
Z = impedances(1);
absZ = abs(Z);
apparentResistivity = (absZ * absZ)/(mu * w);
phase = atan2(imag(Z),real(Z));

After that, Authors were doing running on Matlab so this window will pop out like
picture below. Inside the window theres a line with red colour which is a calculation data
supposed to be. Then, theres a line with blue colour which is the real data. The farther the
distance between the blue line and the red line, the greater the error (misfit).

Picture.1. Output from MT Data Processing via matlab

The data from Matlab could be modified freely as Authors wish. Other than that,
Author also could input resistivity and thickness data like pictures below:

Picture.2. Input data for 1D MT forward modelling via wev

Picture.3. Output data for 1D MT forward modelling via web

MT Processing Data Using Inverse Modelling


Authors didnt get the script for inverse modelling, but Authors get a software for MT
processing based from inverse modelling from (Imam B. Raharjo, Dept. of Geology and
Geophysics, The University of Utah, Salt Lake City, UT., U.S.A., 2008). These pictures
below show alameda software that is used on inverse modelling 1D-MT.

Picture.3. Display of app resistivity vs period

Firstly > Start/Restart> Inver, one step / invert five steps


Here are the result from step by step inversion
1

Here are the result after many iteration

1
0

1
5

20

2
5

3
0

Picture.3. Display of plot resistivity vs depth

The following pictures are the display from 2D MT Modelling software to processed
2D MT Data.
Input the data click calculate

Picture.3. Data and calculation of apparent resistivity

Click graph to show the geometry

Picture.3. Acquisitions of MT Geometry

Click curves TE to obtain the curve

Picture.3. Acquisitions of MT Geometry

Also, theres software of geothermal project tool. This software can not be used on
MT calculation, but this software are supposed to calculate the probability from existing data
with Monte Carlo Inversion principle.

Pictu
re.3. Inversion of Monte Carlo

Picture.3. Inversion of Monte Carlo

Picture.3. Inversion of Monte Carlo

References
Basokur A T 1994 Definitions of apparent resistivity for the presentation of
magnetotelluric sounding data; Geophysical Prospecting 42 141149.
Cagniard L 1953 Basic theory of magnetotelluric method of geophysical prospecting;
Geophysics 18 605635.
Niwas S, Gupta P K and Gaur V K 2005 Normalized impedance function and the
straightforward inversion scheme for magnetotelluric data; J. Earth Syst. Sci. 114 5 523-531.
Pederesen J F and Hermance 1986 Least-square inversion of one-dimensional
magnetotelluric data: An assessment of procedures employed by Brown University; Surv.
Geophys.
8 187231.
Weidelt P 1972 The inverse problem of geomagnetic induction; Z. fur. Geophys. 38
257289.
http://www.digitalearthlab.com/tutorial/tutorial-1d-mt-forward/

(accessed

on

Thursday, March 10th 2016 on 22.30).


Imam B. Raharjo, Dept. of Geology and Geophysics, The University of Utah, Salt
Lake City, UT., U.S.A., 2008

Você também pode gostar