You are on page 1of 31

Geostatistics

Mike Goodchild
Spatial interpolation
 A field
– variable is interval/ratio
– z = f(x,y)
– sampled at a set of points
 How to estimate/guess the value of the
field at other points?
Characteristics of interpolated
surfaces
 Representation
– raster, isolines, TIN
 Form
– rugged or smooth
– exact or approximate
– continuity
• 0-order
• 1-order
• 2-order
 Uncertainty
– variance estimators?
Linear interpolation
 Along a line
– geocoding with address ranges

x2,y2
address2

x,y
x1,y1
address
address1
In a triangle
30

40

20
In a rectangle
 Bilinear interpolation
20 (24) 30

(29)

40
30 (34)
Characteristics of linear
interpolation
 Exact
 0-order continuity
 Contours are straight
– but not parallel in bilinear case
IDW
 Advantages
– quick, universal, theory-free
 Disadvantages
– theory-free
– directional effects
• non-spatial
– characteristics of a weighted average
• when all weights are non-negative
0.5
1.5
2.5
3.5
4.5

0
1
2
3
4
1

10

13

16

19

22

25

28

31

34

37

40

43

46

49

52

55

58

61

64

67

70

73

76

79

82

85

88

91

94

97

100
Characteristics of IDW
surfaces
 Pass through each data point (exact)
– if negative power distance function
– 1/0b = 
 0-, 1-, 2-order continuous
– except at data points
 Underestimate peaks
– volcanoes
– unless peak is observation point
 Extrapolate to the global mean
 Noisy extrapolations
Kriging
 Geostatistics as theoretical framework
 Estimation of parameters from data
 Use of estimated model to control
interpolation
 Many versions
– not a simple black box
– highlights
– demonstration
The variogram
 Relationship between variance and distance
 Formalization of Tobler's First Law
 Estimated from data
– how well can a given data set estimate variogram?
– distribution of sample points is critical
• at peaks and pits
• samples the range of possible distances
• uniform spacing not desirable
• but often out of the user's control
Estimation
 Data points zi(xi)
 Interpolate at x
– stochastic process
– multiple realizations
• variance obtained from variogram
 A set of weights i unique to x
– chosen such that the estimate is
• unbiased
• minimum variance
Kriging prediction
Results of Kriging
 A mean surface
 A variance surface
– minimum at observation points
 Mean surface is smoother than any
realization
– is not a possible realization
• a mean map is not a possible map
– compare a univariate process
– average rainfall versus rainfall from a single storm
– conditional simulation
Kriging standard error
Kriging variants
 Co-Kriging
– interpolation process guided by another
variable (field)
– hard and soft data
– observations of interpolated data are hard
– guiding variable is soft
70

55

83

68

z = f (elevation)
Co-Kriging
 Linear relationship f
 Point observations are hard
– accurate, sparse
 Elevation observations are soft
– inaccurate (errors in measurement or
prediction)
– dense
Co-Kriging prediction
Co-Kriging standard error
Indicator Kriging
 Binary field
– c {0,1}
 Obtained by thresholding an interval/ratio
field
– c=1 if z>t else c=0
– estimate variogram from observations of c
– z is hidden
 The multivariate case
– sequential assignment
Indicator Kriging
 Assign Class 1, notClass 1
 Among notClass 1, assign Class 2,
notClass 2
 Continue to Class n-1
– notClass n-1 = Class n
Universal Kriging
 Simple Kriging is all second order
– trend results from random walk
 Stochastic process plus trend
– trend is first order
– remove trend before analysis
– restore trend after analysis
Advantages and
disadvantages
 Theoretically based
 Not a black box
 Statistical
– variance estimates
 Sensitivity to sample design

You might also like