Peter Telek

GEOG 586 - Fall 2008

Project 5 - Page 1

Project 5: Interpolation Methods

Part I: Inverse Distance Weighted Interpolation (1)

Figures 1a-c shows three outputs as the number of points were varied from 6 to 12 to 24 for the
Inverse Distance Weighted Interpolation of 597 randomly selected elevation height values in
central PA. As the number of points increase, then contour "smoothness" decreases. I chose the
12 point (figure 1b) as the final data set for later work.

Figure 1d shows the output of a fixed radius of 2500 m. Since the spacing in the area of interest
is neither uniform nor dense, there are several points or areas that have a distance greater than
2500 meters. Therefore the interpolation is incomplete and has "holes" in the result.

According to ArcGIS, "Inverse distance weighted (IDW) interpolation determines cell values
using a linearly weighted combination of a set of sample points. The weight is a function of
inverse distance. The surface being interpolated should be that of a locationally dependent
variable.

The {power} option of IDW lets you control the significance of known points on the interpolated
values, based on their distance from the output point. It is a positive, real number. The default
value is 2.

By defining the higher {power} option, more emphasis can be put on the nearest points. Thus,
nearby data will have the most influence, and the surface will have more detail (be less smooth).
As the power increases, the interpolated values begin to approach the value of the nearest sample
point. Specifying a lower value for power will provide a bit more influence to surrounding points
a little farther away. Since the IDW formula is not linked to any real physical process, there is no
way to determine that a particular power value is too large. As a general guideline, a power of 30
would be considered extremely large, and thus of questionable use.

An optimal value for the power can be considered to be where the minimum mean absolute error
is at its lowest. ArcGIS Geostatistical Analyst provides a way to investigate this.

IDW relies mainly on the inverse of the distance raised to the power. If the distances are large, or
the power value is large, the results may be incorrect.

The characteristics of the interpolated surface can also be controlled by limiting the input points
for calculating each interpolated point. The input can be limited by the number of sample points
to be used or by a radius within which there are points to be used in the calculation of the
interpolated points."
 Peter Telek
GEOG 586 - Fall 2008
Project 5 - Page 2

Figure 1a. Inverse Distance Weighted Interpolation of 597 random height values in central PA
using ArcMap with following options: power = 2, variable size search radius, number of points
= 6, and output cell size = 500.
 Peter Telek
GEOG 586 - Fall 2008
Project 5 - Page 3

Figure 1b. Inverse Distance Weighted Interpolation of 597 random height values in central PA
using ArcMap with following options: power = 2, variable size search radius, number of points
= 12, and output cell size = 500.
 Peter Telek
GEOG 586 - Fall 2008
Project 5 - Page 4

Figure 1c. Inverse Distance Weighted Interpolation of 597 random height values in central PA
using ArcMap with following options: power = 2, variable size search radius, number of points
= 24, and output cell size = 500.
 Peter Telek
GEOG 586 - Fall 2008
Project 5 - Page 5

Figure 1d. Inverse Distance Weighted Interpolation of 597 random height values in central PA
using ArcMap with following options: power = 2, fixed size search radius, distance = 2500, min
# of pts = 0, and output cell size = 500.
 Peter Telek
GEOG 586 - Fall 2008
Project 5 - Page 6

Part II: Inverse Distance Weighted Interpolation (2) - Interpolation Errors

Figure 2a shows error calculation map (estimate - truth). The large differences are dark red
(negative) values and dark green (positive) are shown while the lightest colors show small or no
difference between estimated and truth values in elevation in central PA. Figure 2b is a zoomed
in area with error contours. The greatest errors appear to be furthest from the points. The least
difference between the random points and the truth are near and at the randomly selected points.
This makes sense since the closer you get to the point with IWD, the better the estimate; and
identical for the point itself if the algorithm is coded correctly.

Figure 2a. Error calculation map (estimate - truth). The large differences are dark red (negative)
values and dark green (positive) are shown while the lightest colors show small or no difference
between estimated and truth values in elevation in central PA.
 Peter Telek
GEOG 586 - Fall 2008
Project 5 - Page 7

Figure 2b. Zoomed in Error calculation map (estimate - truth) with error contours. The large
differences are dark red (negative) values and dark green (positive) are shown while the lightest
colors show small or no difference between estimated and truth values in elevation in central PA.
 Peter Telek
GEOG 586 - Fall 2008
Project 5 - Page 8

Part III: Kriging Using the Geostatistical Analyst

Note: I had to use an evaluation copy of Geostatistical Analyst for ArcMap v9.2 software for this
part of the project. It seemed to work.

Figure 3a-c are results from ArcMap using the isotropy case to create an interpolated map by
Kriging from the 597 randomly selected elevations in central PA used in Part I.

Anisotropy, directional influences on the semivariogram/covariance, is a types of directional

component that can affect the predictions in the output surface. For anisotropy, the shape of the
semivariogram may vary with direction. Isotropy exists when the semivariogram does not vary
according to direction.

In this example, the Pennsylvania hills or ridges appear to be oriented southwest to northeast.
Since there is a directional trend, I'd expect that an anisotropy result to be better than an isotropy
result. Figure 3g shows an elliptical shape oriented SW-NE for specifying searching
neighborhood for anisotropy case.
 Peter Telek
GEOG 586 - Fall 2008
Project 5 - Page 9

Figure 3a. Interpolated map by kriging from 597 randomly selected elevations in central PA.
 Peter Telek
GEOG 586 - Fall 2008
Project 5 - Page 10

Figure 3b. Contour lines of interpolated map by kriging randomly selected elevations displayed
over the actual elevation data in central PA.
 Peter Telek
GEOG 586 - Fall 2008
Project 5 - Page 11

Figure 3c. Specifying the isotropy semivariogram model to work with in kriging.
 Peter Telek
GEOG 586 - Fall 2008
Project 5 - Page 12

Figure 3d. Specifying the ansotropy semivariogram model to work with in kriging.
 Peter Telek
GEOG 586 - Fall 2008
Project 5 - Page 13

Figure 3e. Interpolated map by kriging from 597 randomly selected elevations in central PA.
 Peter Telek
GEOG 586 - Fall 2008
Project 5 - Page 14

Figure 3f. Contour lines of interpolated map by Kriging randomly selected elevations displayed
over the actual elevation data in central PA.
 Peter Telek
GEOG 586 - Fall 2008
Project 5 - Page 15

Figure 3g. Specifying searching neighborhood for anisotropy case. Notice the elliptical shape
oriented SW-NE.