Spatial

Interpolation
presented by:
Tim Haithcoat
University of Missouri
Columbia

Compiled with materials from:

Nigel M. Waters
University of Calgary
 Introduction
Spatial interpolation is the procedure of estimating
the value of properties at unsampled sites within the
area covered by existing observations
In almost all cases the property must be interval or ratio
scaled
Can be thought of as the reverse of the process used
to select the few points from a DEM which
accurately represent the surface
Rationale behind spatial interpolation is the
observation that points close together in space are
more likely to have similar values than points far
apart (Toblers Law of Geography)
2
 Introduction (Continued)
Spatial interpolation is a very important feature of
many GISs
Spatial interpolation may be used in GISs:
To provide contours for displaying data graphically
To calculate some property of the surface at a given point
To change the unit of comparison when using different data
structures in different layers
Frequently is used as an aid in the spatial decision making
process both in physical & human geography
As well as interrelated disciplines such as mineral prospecting &
hydrocarbon exploration
Many of the techniques of spatial interpolation are two-
dimensional developments of one dimensional methods
originally developed for time series analysis
3
 Various Classifications of
Interpolation Procedures

Point Interpolation/Areal Interpolation

Global/Local Interpolators
Exact/Approximate Interpolators
Stochastic/Deterministic Interpolators
Gradual/Abrupt Interpolators

4
 Point Interpolation/Areal Interpolation
Point based:
Given a number of points whose locations and values are
known, determine the value of other points at predetermined
locations 10 9
9.5
8 7
Point interpolation is used for data which can be collected at
point locations (e.g. weather station readings, spot heights, oil well
readings, porosity measurements)
Interpolated grid points are often used as the data input to
computer contouring algorithms
Once the grid of points has been determined, isolines (e.g., contours)
can be threaded between them using a linear interpolation on the
straight line between each pair of grid points
Point to point interpolation is the most frequently performed
type of spatial interpolation done in GIS
5
 Point Interpolation/Areal Interpolation
(continued)

Lines to points
E.g. contours to elevation lines

Areal Interpolation
Given a set of data mapped on one set of source zones
determine the values of the data for a different set of target
zones
Example: given population counts for census tracts, estimate
populations for electoral districts

6
 Global/Local Interpolators
Global Interpolations determine a single function
which is mapped across the whole region
A change in one input value affects the entire map
Local Interpolators apply an algorithm repeatedly
to a small portion of the total set of points
A change in an input value only affects the result within
the window
Global algorithms tend to produce smoother surfaces
with less abrupt changes
Used when theres an hypothesis about the form of the
surface (ex: a trend)

7
 Global/Local Interpolators
(continued)

Some local interpolators may be extended to include

a large proportion of the data points in set, thus
making them in a sense global
The distinction between global and local
interpolators is thus a continuum and not a
dichotomy
This has led to some confusion
and controversy in the literature

8
 Exact/Approximate Interpolators

Exact Interpolators honor the data points upon which

the interpolation is based
The surface passes through all points whose values are
known
Honoring data points is seen as an
important feature in many applications
Example: the oil industry
Proximal interpolators, B-splines and Kriging methods all
honor the given data points
Kriging, may incorporate a nugget effect & if this is the
case the concept of an exact interpolator ceases to be
appropriate

9
 Exact/Approximate Interpolators
(Continued)

Approximate Interpolators are used when there is

some uncertainty about the given surface values
This utilizes the belief that in many data sets there are
global trends, which vary slowly, overlain by local
fluctuations, which vary rapidly and produce uncertainty
(error) in the recorded values
The effect of smoothing will therefore be to reduce the
effects of error on the resulting surface

10
 Stochastic/Deterministic Interpolators

Stochastic methods incorporate the concept of

randomness
The interpolated surface is conceptualized as one of many
that might have been observed, all of which could have
produced the known data points
Stochastic interpolators include trend surface analysis,
Fourier analysis and Kriging
Procedures such as trend surface analysis allow the statistical
significance of the surface and uncertainty of the predicted
values to be calculated
Deterministic methods do not use probability theory
(e.g. proximal)
11
 Gradual/Abrupt Interpolators
A typical example of a gradual interpolator is the
distance weighted moving average
Usually produces an interpolated surface with gradual
changes
However, if the number of points used in the moving average
is reduced to a small number, or even one, there would be
abrupt changes in the surface
It may be necessary to include barriers in the
interpolation process
Semipermeable barriers (example: weather fronts)
Will produce quickly changing but continuous values
Impermeable barriers (example: geologic faults)
Will produce abrupt changes
12
 Point Based Interpolation
EXACT MODELS
Lam (1983) and Burrough (1986) describe a
variety of quantitative interpolation methods
suitable for computer contouring algorithms
Proximal
B-Splines
Kriging
Manual (or eyeballing)

13
 PROXIMAL (1 of 2)
All values are assumed to be equal to the nearest
known point
Is a local interpolators
Computing load is relatively light
Output data structure is Thiessen polygons with
abrupt changes at boundaries
Has ecological applications such
as territories and influence zones

14
 PROXIMAL (2 of 2)

Best for nominal data, although

originally used by Thiessen for computing
areal estimates from rainfall data
Is absolutely robust, always produces a result,
but has no intelligence about the system
being analyzed
Available in very few mapping packages,
SYMAP is a notable exception

15
 B-SPLINES (1 of 2)
Uses a piecewise polynomial to provide a series
of patches resulting in a surface that has
continuous first and second derivatives
Ensures continuity in:
Elevation (zero-order continuity)
surface has no cliffs
Slope (first-order continuity)
slopes do not change abruptly, there are no kinks in contours
Curvature (second order continuity)
minimum curvature is achieved
Produces a continuous surface with minimum
curvature
16
 B-SPLINES (2 of 2)
Output data structure is points on a raster
Note: maxima & minima do NOT necessarily occur at
the data points
Is a local interpolator
Can be exact or used to smooth surfaces
Computing load is moderate
Best for very smooth surfaces
Poor for surfaces which show marked fluctuations, this
can cause wild oscillations in the spline
Are popular in general surface interpolation
packages, but not common in GISs
Can be approximated by smoothing contours
drawn through a TIN model
17
 KRIGING
Developed by George Matheron as the theory
of regionalized variables and D.G. Krige as an
optimal method of interpolation for use in the
mining industry

The basis of this technique is the rate at which

the variance between points changes over space
This is expressed as a variogram which shows how
the average difference between values at points
changes with distance between points
18
 KRIGING
Variograms (1 of 2) sill
e
range
e (vertical axis): E(zi - zj)2,
i.e. expectation of the difference nugget
d
Example: the average difference in elevation
of any two points distance d apart
d (horizontal axis): distance between i and j

Most variograms show behavior like the

diagram
Sill: the upper limit (asymptote) of e
Range: distance at which this limit is reached
Nugget: intersection with the y axis
A non-zero nugget indicated that repeated
measurements at the same point yield different values
19
 KRIGING
Variograms (2 of 2)
In developing the variogram it is necessary to make
some assumptions about the nature of the observed
variation on the surface:
Simple Kriging assumes that the surface has a constant
mean, no underlying trend and that all variation is
statistical
Universal Kriging assumes that there is a deterministic
trend in the surface that underlies the statistical variation
In either case, once trends have been accounted for
(or assumed not to exist), all other variation is
assumed to be a function of distance

20
 KRIGING
Deriving the Variogram (1 of 2)
The input data for Kriging is usually an irregularly
spaced sample of points
To compute a variogram we need to determine how
variance increases with distance
Begin by dividing the range of distance into a set of
discrete intervals
Example: 10 intervals between distance of 0 and the
maximum distance in the study area
For every pair of points, compute distance and the
squared difference in z values

21
 KRIGING
Deriving the Variogram (2 of 2)
Assign each pair to one of the distance ranges, and
accumulate total variance in each range
After every pair has been used (or a sample of pairs
in a large dataset) compute the average variance in
each distance range
Plot this value at the midpoint distance of each
range

22
 KRIGING
Computing the Estimates (1 of 2)

Once the variogram has been developed, it is used to

estimate distance weights for interpolation
Interpolated values are the sum of the weighted values of
some number of known points where weights depend on
the distance between the interpolated and known points
Weights are selected so that the estimates are:
Unbiased (if used repeatedly, Kriging would give the
correct result on average)
Minimum variance (variation between repeated estimates
is minimum

23
 KRIGING
Computing the Estimates (2 of 2)

Problems with this method:

When the number of data points is large, this
technique is computationally very intensive
The estimation of the variogram is not simple, no
one technique is best
Since there are several crucial assumptions that
must be made about the statistical nature of the
variation, results for this technique
can never be absolute

24
 MANUAL or EYEBALLING (1 of 2)

Traditionally, not a highly regarded method among

geographers and cartographers

Dutton-Marion (1988) have shown that among

geologists, this is a very important procedure and that
most geologists actually distrust the more sophisticated,
mathematical algorithms.
They feel that they can use their expert knowledge,
modeling capabilities and experience and generate a
more realistic interpolation by integrating this knowledge
into the construction of the geological surface.
25
 MANUAL or EYEBALLING (2 of 2)

Attempts are now being made to use knowledge

engineering techniques to extract this knowledge from
experts and build it into an expert system for
interpolation
Characteristics of this method include:
Procedures are local as different methods may be used by the
expert on different parts of the map
Tend to honor data points
Abrupt changes such as faults are more easily modeled using
these methods
Surfaces are subjective and vary from expert to expert
Output data structure is usually in the form of a contour
26
 Point Based Interpolation -
APPROXIMATE METHODS

Trend Surface Analysis

Fourier Series
Moving Average/Distance
Weighted Average

27
 Trend Surface Analysis (1 of 3)

Surface is approximated by a polynomial

Output data structure is a polynomial function
which can be used to estimate values of grid
points on a raster or the value at any location
The elevation z at any point (x,y) on the surface is
given by an equation in powers of x and y
Example: a linear equation (degree 1) describes a
tilted plane surface
z = a + bx + cy
Example: a quadratic equation (degree 2) describes
a simple hill or valley:
z = a + bx + cy + dx2 + exy + fy2
28
 Trend Surface Analysis (2 of 3)

In general, any cross-section of a surface of degree

n can have at most n-1 alternating maxima and
minima
Example: a cubic surface can have one maximum an
done minimum in any cross-section
Equation for the cubic surface:
z = a + bx + cy + dx2 + exy + fy2 + gx3 + hx2y + ixy2 + jy3
A trend surface is a global interpolator
Assumes the general trend of the surface is
independent of random errors found at tech sampled
point
29
 Trend Surface Analysis (3 of 3)

Computing load is relatively light

Problems:
Statistical assumptions of the model are rarely met
in practice
Edge effects may be severe
A polynomial model produces a rounded surface
This is rarely the case in many human ad physical
applications
Available in grate many mapping packages

30
 Fourier Series
Approximates the surface by overlaying a series
of sine and cosine waves
A global interpolator
Computing load is moderate
Output data structure is the Fourier series which
can be used to estimate grid values
Best for data sets which exhibit
marked periodicity, such as ocean waves
Rarely incorporated in computing packages

31
 Moving Average/Distance
Weighted Average (1 of 2)
Estimates are averages of the values at n known
points:
Z = wizi/ wi
Where w is some function of distance, such as:
W = 1/dk w = e-kd
An almost infinite variety of algorithms may be
used, variations include:
The nature of the distance function
Varying the number of points used
The direction from which they are selected
Is the most widely used method
32
 Moving Average/Distance
Weighted Average (2 of 2)
Objections to this method arise from the fact
that the range of interpolated values is limited
by the range of the data
No interpolated value will be outside the observed
range of z values X X interpolated
X

X actual
Other problems include:
How many points should be included in the
averaging
What to do about irregularly spaced points
How to deal with edge effects
33
 A further examination of
Spatial Interpolation
Looking at areal interpolation techniques and
some applications
Areal interpolation : problem of transferring
data from one set of areas (source reporting
zones) to another (target reporting zones)
This is easy if the target set is an aggregation of the
source set, but more difficult if the boundaries of the
target set are independent of the source set
Also, applications that do not fall easily into
either point or areal interpolation categories
34
 Areal Interpolation
Non-Volume Preserving (Point Based)
Example: interpolating population counts from
census tracts to school districts

x x
x x x
x
x x x x
x x x x x
x

Source Zones Interpolated Target Zones

with Centroids & Population Density on Grid
Population Density Surface with Grid
35
 Non-Volume Preserving
PROCEDURE
Calculate the population density for each source census tract by
dividing population by area
Identify a centroid for each region
Assign to the point located at each centroid, the population density value
determined for its enclosing area
Using this set of points, interpolate a gridded population density
surface using any of the methods described previously
Convert each grid cells value to a population by multiplying the
estimated density by the cells area
Overlay the interpolated grid on the target map and assign each
grid value to each its target region (school district)
Calculate the total population in each target region

36
 Non-Volume Preserving
CRITICISMS of PROCEDURE
Choosing the center point is ill defined
Inadequacy of point based interpolation
methods
Most importantly, the total value of each zone
is not conserved
Example: if a source zone is divided into two
target zones, the total population of the target
zones after interpolation need not equal the
population of the source zone

37
 Areal Interpolation
Volume-Preserving: Overlay
Procedure
Overlay of target and source zones
Determining the proportion of
each source zone that is assigned
to each target zone
Apportioning the total value of the
attribute for each source zone to
target zones according to the area
proportions
Assumes uniform density of the attribute within
each zone
e.g. uniform population density if the attribute is total
zone population
38
 Areal Interpolation
Volume-Preserving: Pycnophylactic (1 of 2)
This technique has two objectives:
Create a smooth surface, no steps
Attribute values should not change suddenly at zone
boundaries
The total value of the attribute within each zone
must be correct
Does not require an assumption of homogeneity
within zones but rapid variation within zones
may affect the quality of interpolation
Output is a contour or continuously shaded map
39
 Areal Interpolation
Volume-Preserving: Pycnophylactic (2 of 2)
Procedure
Overlay a dense raster on a choropleth map
Divide each zones total value equally among the
raster cells that overlap the zone
Smooth the values by replacing each cells value with
the average of its neighbors
Sum the values of the cells in each zone
Adjust the values of all cells within each zone
proportionally so that the zones total is the same as
the original total
Example: if the total is 10% low, increase the value of
each cell by 10%
Repeat steps 3,4, and 5 until no more changes occur
40
 Areal Interpolation
Volume-Preserving: Boundary Conditions
At the boundary of the reporting zones,
pixels will have neighbors outside the study
area and therefore without values
Some decision must be made about the behavior
of the surface outside the study area
Examples:
Population density equals zero
(a lake or rural area)
population density unknown,
assumed equal to the values of
the outermost pixels of the
study area
41
 Special Cases of Spatial Interpolation
Mapping Populated Areas (1 of 4)

Objective: create a map showing populated

areas, given point population values for a
number of cities and towns
This problem arises frequently when populated
areas are represented as points
It arises for small reporting zones when
boundary files are unavailable, but data includes
centroid locations
Example: US or UK census data
Are several methods that could be used
42
 Special Cases of Spatial Interpolation
Mapping Populated Areas (2 of 4)

Simple Approach:
estimate the populated area using an empirical
relationship like:
A is proportional to p0.84
And draw a circle around the point of radius:
(a/)

Bracken & Martin (1989) have developed

methods for replacing ED centroids by disks,
the radius of each disk being estimate from the
distances to neighboring centroids
The method works very well with UK ED data
43
 Special Cases of Spatial Interpolation
Mapping Populated Areas (3 of 4)

Alternative Approach:
Establish a critical population density for defining
an urban area
Spread the population over each urban area so that
population density is highest in the center and
decreases gradually outwards
e.g. use a normal distribution function

44
 Special Cases of Spatial Interpolation
Mapping Populated Areas (4 of 4)

Alternative Approach (continued):

Interpolate densities to a raster, accumulating values
where the population spread from two urban areas
overlap

Draw contours at the critical value to define the

boundaries of the populated areas
Both of these methods fall within the general
heading of density estimation
A density is being estimated from a collection of
points
45
 Special Cases of Spatial Interpolation
Estimating Trade Areas (1 of 3)

In marketing, it is often desirable to plot the

boundary of a trade area for example, a store,
given information of the home locations of
customers
Simplest case: location of all customers and
non-customers is known
Simply draw a boundary contour between them
n
n nc n
c n
n c cc cc n n
X
n c cc c c c c c
n n n n c cc n
n
n n
46
 Special Cases of Spatial Interpolation
Estimating Trade Areas (2 of 3)

If location of non-customers is NOT known:

Calculate the average distance to all customers and
draw a circle
OR
Give each customer a small probability surface
Accumulate values as in the populated areas example
Set critical value for delimiting trade area
OR
Continued on next page ...

47
 Special Cases of Spatial Interpolation
Estimating Trade Areas (3 of 3)
OR
Divide the area into sectors, average the distance to
customers within the sectors and draw a distance arc
for each sector

X X

These techniques do NOT pick up islands or holes in the

trade area c c c c
c c c c n c Xc
c c c c c cc c c
c cc c c c
48
 A GIS Perspective on
Interpolation
Both point and areal interpolation try to
estimate a continuous surface
In the point case, the surface has been measured
at sample points
In the areal case, the surface of population
density is estimated from total population
counts in each reporting zone
 A GIS Perspective on
Interpolation
In other cases it is impossible to conceive of a
continuous surface
e.g., each point is a city & the attribute is city population
Example: if city A has a population of 1 million & city B 100
km away has a population of 2 million, there is no reason to
believe in the existence of a city halfway between A & B with
population 1.5 million
In this case, the variable population exists only at the
points, not as a continuous surface
In other cases, the variable might exist only along lines
Example: traffic density on a street network
 A GIS Perspective on
Interpolation
We must distinguish here between layer and
object views of the world
A continuous surface of elevations is a layer view of the
world - there is one value of elevation at an infinite
number of possible places in the space
The point map of cities is an object view of the world -
the space in between points is empty and has no value of
the population variable
The street map is an object view of the world - the world
is empty except where there are streets - only along
streets is traffic density defined
A GIS Perspective on
Interpolation

Spatial interpolation implies a

layer view of the world,
and it requires special techniques
(e.g. density estimation)
to apply it to objects
such as store customers.
 Expert Systems for Spatial
Interpolation Algorithms (1 of 2)
A good GIS should include a range of spatial
interpolation routines so that the user can
choose the most appropriate method for the data
and the task
Ideally, these routines should provide a natural
language interface which would lead the user
through an appropriate series of questions about
the intentions, goals, and aims of the user and
about the nature of the data

53
 Expert Systems for Spatial
Interpolation Algorithms (2 of 2)

A number of prototype expert systems for

guiding the choice of a spatial interpolation
algorithm have been developed
These may be written in the form of:
An expert system shell
In one of the artificial intelligence languages (such
as Prolog or LISP)
Or in a high level language (such as Pascal)

54
CONCLUSION: If computer contouring and surface generation
techniques are to be incorporated successfully in GIS, they must be
easy to use and be effective.
easy to use implies that those effective means that these
without a detailed knowledge of the techniques should be
mathematical & statistical informative, highlighting the
characteristics of the procedure essential nature of the data
should be able to choose the correct and/or surface and serving the
technique for displaying a particular purpose of the
data set for a particular purpose researcher/analyst.
Note: Statisticians argue that this is The researchers measure of
NOT an ideal goal as people may use success will be largely
techniques without a proper subjective & visual - does the
understanding of the underlying result look right?
assumptions.

This purpose may vary from an attempt to model all the real intricacies
of the surface to simply trying to highlight the general, spatial trend of the
data in order to aid in the decision-making process.