Professional Documents
Culture Documents
(for KRIGING)
Dr. Charlie Wu
November 7, 2018
1
Kriging estimation
Kriged Standard
estimate error
Semivariogram
Input data
(thickness, m) Sill1 Model 1
130 Input Data
120
150
110
Sill2 Model 2
180
170
Sill3
90
160 11 Model 3
10km
Nugget2
Nugget1
Nugget3
Figure 2. Nonstationarity and anisotropy causing ambiguous semivariogram models which lead to uncertain kriged By EccoSse
estimates.
software
Kriging
• A geostatistical method for spatial interpolation. Stochastic,
exact, smooth/abrupt, global/local.
• A Gaussian process to estimate the mean value of a normal
distribution, second order stationarity. kriging can assess
the quality of prediction with estimated prediction errors.
• Assumes the spatial variation of an attribute is neither totally
random (stochastic) nor deterministic (theory of regionalized
variables).
• Mature technique. Most mineral exploration projects has
100’s data points with spacing 100m or less.
• The spatial variation may consist of three components:
a spatially correlated component (variable);
a “drift” or structure;
a random error term.
• Implemented using a semivariogram.
3
Linear Model: semivariance=4*distance
Slope = 4m/km
Kriged est
Kriging Residuals 4 562
120m
542 estimate
Range = 30 km
Geostatistical modeling of a gold mine
Sample size 1000+, grid 50mx200m
W E
http://www.kingstonresources.com.au/home-
news/livingstone-gold-project-exploration-update/
SEMIVARIOGRAM EXAMPLE
W 200m
E
Gold
ppm 11 18 21 19 21 22 11 14 20 31 30 25 31 25
W d E
Lag x1 x2 x3 x4 x5 x6 x7 x8 x9 x10 x11 x12 x13 x14
Z(xi) ∑(Z(xi )-Z(xi+nd ))2 n(h) g(h)
(d) 11 18 21 19 21 22 11 14 20 31 30 25 31 25
Z(x i ) - Z(x i+d ) -7 -3 2 -2 -1 11 -3 -6 -11 1 5 -6 6
2 200 452 13 17
(Z(x i )-Z(x i+d )) 49 9 4 4 1 121 9 36 121 1 25 36 36
Z(x i ) - Z(x i+2d ) -10 -1 0 -3 10 8 -9 -17 -10 6 -1 0
400 781 12 33
(Z(x i )-Z(x i+2d )) 2 100 1 0 9 100 64 81 289 100 36 1 0
Z(x i ) - Z(x i+3d ) -8 -3 -1 8 7 2 -20 -16 -5 0 5
600 897 11 41
(Z(x i )-Z(x i+3d )) 2 64 9 1 64 49 4 400 256 25 0 25
Z(x i ) - Z(x i+4d ) -10 -4 10 5 1 -9 -19 -11 -11 6
2 800 962 10 48
(Z(x i )-Z(x i+4d )) 100 16 100 25 1 81 361 121 121 36
Z(x i ) - Z(x i+5d ) -11 7 7 -1 -10 -8 -14 -17 -5 SEMIVARIOGRAM
2 1000 60 894 9 50
(Z(x i )-Z(x i+5d )) 121 49 49 1 100 64 196 289 25 50
g(h)
40
Z(x i ) - Z(x i+6d ) 0 4 1 -12 -9 -3 -20 -11
1200 30 772 8 48
(Z(x i -Z(x i+6d )) 2 0 16 1 144 81 9 400 121 20
10
Z(x i ) - Z(x i+7d ) -3 -2 -10 -11 -4 -9 -14 lag
2 1400 0 527 7 38
(Z(x i )-Z(x i+7d )) 9 4 100 121 16 81 196 0 200 400 600 800100012001400
6
SEMIVARIOGRAM EXAMPLE
2 200 906 8 57
(Z(x i )-Z(x i+4d )) 49 0 256 81 100 64 256 100 40
g(h)
20
Z(x i ) - Z(x i+5d ) 3 14 11 14 4 -2 -15
2 250 0 lag 767 7 55
(Z(x i )-Z(x i+5d )) 9 196 121 196 16 4 225 0 100 200 300
7
allow us to calculate a “modeled semivariance” between any 2
points knowing the distance h separating the two points.
C0+C1
Range Sill
50
40 Partial
Sill
g(h) 30
C1 SemiVariogram
1 ∑( Z ( xi ) Z ( x j ))
2
20 g(h)=
2 n
10 j =i+ lag distance h
0 C0 Nugget
0 200 400 600 800
Lag (distance) 8
Kriged value Tp = W1 T1 + W2 T2 + W3 T3
3 W3
T3=16
ppm
W2
T2=14 2
ppm
W1 : weight factor
Di j : distance 1
gi j : semivariance W1 T1=10
ppm
Ordinary Kriging
3 equations for 3 data points
Semivariogram
• Variogram – a function of the distance
separating two locations. A cross-plot of
correlation (variance) of two data points as
the distance between two data points
changes.
• The semivariogram is half of the variogram
1 ∑ (Z ( x ) Z ( x )) 2
i j
2 n
h = lag distance j =i+ lag distance h
11
Calculating Semivariogram
1. Compute distance and direction of all data PAIRS
2. Sort pairs
c. find all pairs with (h - δ/2) < distance (xi, xj) < (h +δ/2);
count pairs
100m 100m
2
31 7 8
32 9 31 10 28 11 28 12
26 13 26 14 26 15 24 16 27 17 26 18
500m
24 19 27 20 21
24 22 26 23 27 24
25 25 24 26 25 27 24 28 22 29 22 30
27 31 26 32 24 33 34
19 35 36
500m
36 soil specimen (interval 100m) were planned and 31 were collected. Gold
concentration in ppm were shown in each specimen.
200m 100m
Semivariance at 100m=
33 1 29 3 31 4 29 5 28 6
100m 100m
2
(29-31)2
31 7 32 9 31 10 28 11 28 12 +(24-27)2
8 +(31-29)2
+(29-28)2 +(24-26)2
26 13 26 14 26 15 24 16 27 17 26 18
500m
+(32-31)2 +(26-27)2
24 19 27 20 21
24 22 26 23 27 24 +(31-28)2 +(25-24)2
+(28-28)2 +(24-25)2
25 25 24 26 25 27 24 28 22 29 22 30 +(26-26)2 +(25-24)2
+(26-26)2 +(24-22)2
27 31 26 32 24 33 19 35
34 36 +(26-24)2 +(22-22)2
500m +(27-26)2
+(24-27)2
+(27-26)2 +(26-24)2
/ (2 x 21)
=59/42 =1.4
14
200m 100m
Semivariance at 200m=
33 1 29 3 31 4 29 5 28 6
100m 100m
(33-29)2
31 7 8
32 9 31 10 28 11 28 12
+(29-29)2
+(31-28)2 +(27-24)2
26 13 26 14 26 15 24 16 27 17 26 18
500m
+(31-32)2 +(24-27)2
24 19 27 20 21
24 22 26 23 27 24 +(32-28)2 +(25-25)2
+(31-28)2 +(24-24)2
25 25 24 26 25 27 24 28 22 29 22 30 +(26-26)2 +(25-22)2
+(26-24)2 +(24-22)2
27 31 26 32 24 33 19 35
34 36 +(26-27)2 +(27-24)2
500m +(24-26)2 +(24-19)2
/ (2 x 18)
=125/36 =3.5
15
300m
Semivariance at 300m=
33 1 29 3 31 4 29 5 28 6
100m 100m
(33-31)2
31 7 32 9 31 10 28 11 28 12
8 +(29-28)2 +(24-24)2
26 13 26 14 26 15 24 16 27 17 26 18 +(31-31)2 +(27-26)2
500m
+(32-28)2 +(25-24)2
24 19 27 20 21
24 22 26 23 27 24 +(26-24)2 +(24-22)2
+(26-27)2 +(25-22)2
25 25 24 26 25 27 24 28 22 29 22 30 +(26-26)2 +(26-19)2
27 31 26 32 24 33 34
19 35 36 / (2 x 13)
500m =90/26 =3.5
16
400m
33 1 29 3 31 4 29 5 28 6 Semivariance at 400m=
100m 100m
31 7 32 9 31 10 28 11 28 12 (33-29)2 +(24-26)2
8
+(31-28)2 +(27-27)2
26 13 26 14 26 15 24 16 27 17 26 18 +(26-27)2 +(25-22)2
500m
+(26-26)2 +(24-22)2
24 19 27 20 21
24 22 26 23 27 24
+(27-19)2
25 25 24 26 25 27 24 28 22 29 22 30 / (2 x 9) =107/18 =5.9
27 31 26 32 24 33 34
19 35 36
500m
Semivariance at 500m=
(33-28)2 +(31-28)2 +(26-26)2 +(24-27)2 +(25-22)2 / (2 x 5)
=52/10 =5.2
17
Gaussian model
N-S
Gaussian
model
Spherical model
Semivariogram
E-W
Spherical model
18
Homework 7: Semivariogram
..xxxx5 ..xxxx3
NPM ..xxxx1 ..xxxx2 ..xxxx6 ..xxxx0 ..xxxx4 ..xxxx8
Sa mpl e ..xxxx9
A Sam ple B ..xxxx7
Sa mpl e C Sam ple D Sa mpl e E Sam ple F Sa mpl e G Sam ple H
Au Cu Au Cu Au Cu Au Cu Au Cu Au Cu Au Cu Au Cu
1 10.62 0.00
1 0.49 0.00
1 4.00 893.41
1 0.00 1
0.00 1.20 268.81
1 1
13.96 74.70 0.40 90.30
1 2.00 3.70
2 2.92 99.60
2 18.50 180.80
2 3.10 21.53 0.00 2 31.40 134.16
62.90 2 0.00 2
0.00 6.50 952.00
2 0.00 0.00
3 3.20 4.04
3 200.00 103.20
3 0.00 70.10
3 0.85 3
16.26 0.00 30.00 0.00 3
0.00 0.22 151.47
3 0.31 0.00
4 0.80 222.00
4 1.33 0.00
4 0.00 40.00 0.65 4
16.31 0.00 49.95 0.00 4
0.00 0.04 5.16 0.06 69.00
4
5 0.00 1.60
5 0.00 98.76
5 0.01 50.00 0.08 5 13.80 258.12
2.01 5 0.00 5
0.00 15.63 46.29
5 0.96 213.90
6 227.60 23.60
6 0.00 0.00
6 0.13 60.00 7.38 3.29
6 0.99 60.39 4.27 6
63.60 0.00 0.00 1.66
6 1.39
7 0.00 56.23
7 1.86 0.00
7 0.00 70.00 1.26 7
0.00 1.62 7
70.00 0.05 105.11 83.80 1041.89
7 0.92 37.53
8 36.00 172.00
8 0.00 0.00
8 9.22 130.65
8 1.42 8
0.23 0.00 80.00 0.00 8
0.00 0.45 0.00 0.42
8 0.00
9 0.27 0.00
9 0.00 44.10
9 4.53 21.31
9 0.52 9
47.80 3.89 9
49.86 0.00 9
0.00 0.00 0.00 2.66 26.14
9
10 0.43 10
0.51 0.00 10
0.00 53.34 10
1507.00 0.48 0.00 0.02 102.77 0.00
10 0.52 10.73 130.00
10 10 7.08 275.43
11 2.63 11
3.52 0.35 11
12.40 3.10 5.10 11 1.61 6.45 11 4.05 87.36
Seven datasets (A to H) of gold grade (in g/t) are shown above. All data
were measured from East to West with distance between two adjacent
data points is 100m. a) Please calculate empirical semivariogram (see
example on pages 6 and 7) using your dataset. b) Can you draw the
theoretical semivariogram and estimate Range, Sill and Nugget?
Empirical SV
Theoretical SV