GIS Fundamentals Lab 12

Lab 12: Sampling and Interpolation

What You’ll Learn:
-Systematic and random sampling -Stratified sampling
-Majority filtering -A few basic interpolation methods

Data for the exercise are found in the \Lab12 subdirectory.

What You’ll Produce: maps, each with four panels. One map will have original
data and various sampled methods surfaces and the second map will show
interpolated surfaces.

Background: Theory is covered in Chapter12 (Spatial Estimation) and 10

(Raster Analysis) of the GIS Fundamentals textbook.

Sampling and Interpolation in ArcGIS Pro

Analysis often requires generating sample points, either to visit in the field, or to
characterize large raster layers. In the first part of this lab we’ll create sample
points, and use them to extract data from a DEM.

We’ll apply both systematic and random sampling. We’ll also develop and apply
a stratification layer, because sometimes you want to stratify your sample, which
means you wish to increase sample density in some portion of your area, using a
map of zones, or strata.

Create a project and insert 4 NEW maps.

View them in the

Catalog pane, note
that it names them
Map, Map1, Map2,
and Map3 (figure at
left).

You can rename

each by right clicking
on the name in the
Catalog pane, and
selecting rename
from the dropdown
menu (figure at
right).

1
GIS Fundamentals Lab 12

Add the DEM layer, chirdem, a region in southeast Arizona, to all four maps
and apply a color scheme to highlight the topography. (see examples below: Stretch,
Standard Deviation=1.5 Gamma =0.5; note I put for a 50% transparent Hillshade over Original just to make
more realistic; after you add chirdem & set the colors in the 1st map you can “copy” the layer & “paste” it into
the other 3; this approach saves steps)

Rename the maps: a) Original, b) Systematic, c) Random, and D) Stratified

Add a Layout (landscape) to your project, and place all four maps in the Layout, in
the pattern shown. Use the tools described in previous labs to assign a
1:300,000 scale and equal panel size to all four maps, at 3.5” wide by 3.1” high
(refer to previous lab, right click on a map square placed in the layout, then
properties).

Also specify a scale 1:300,000, you don’t need to type in the 1, :, or the commas.

Position and
label the Maps
in the layout so
that you have
something that
looks like the
figure
below/right:

Note that we
only added the
hillshade to the
Original Map,
and not to the
others.

We’ll now
create
sampling
points and
display them
with their
respective
maps.

Make sure the data frame windows DO NOT overlap, and the frames are equally-
sized and spaced, and the data in the frames are all at the same scale. Save
your Project now; you don’t want to lose all this set up work.

2
GIS Fundamentals Lab 12

Systematic Sampling
Activate your Systematic Map, switching to Map View

We’ll first create a systematic (grid)

sampling, and then sample the
elevation surface.

Open Tools ->Toolbox -> Data

Management Tools -> Sampling ->
Create Fishnet (see at right)

In the window that appears, specify

an appropriate output directory and
name your output fishnet something
like fishnet1000.

Set the template to extent of the

chirdem DEM (see right).

Specify Cell Size Width and

Height as 833.

Make sure box in lower left is

checked to Create Label Points.

Leave other values at defaults and

click Run to execute.

(Video: Systematic Sample)

After it creates the fishnet and

labels, remove the fishnet1000
from the Table of Contents, but
keep the label points layer.

3
GIS Fundamentals Lab 12

We need to assign the elevation values to the

points.

We can do this with Tools->Toolbox →Spatial

Analyst Tools -> Extraction -> Extract Values
to Points

Use the geoprocessing tool Extract Value to

Points, with

the label points as your input points,

the chirdem as the input raster,

to an appropriately named output point layer,

and interpolating and appending raster values

(see figure at right, and video: Extract Raster to
Points)

Once we run this tool, each point in the

output will have the elevation coincident
with the point.
TOC.

Remove the fishnet1000_label layer.

Open the table of the Extract_fishnet1

layer, and verify there is an item named
RASTERVALU

Right click on the RASTERVALU column, and calculate the statistics as shown
before.

Note the mean value, and modify the

appropriate label in your Layout to
display the mean value, e.g.,

4
GIS Fundamentals Lab 12

Random Sampling
Activate the Random Map.

Generate a set of random sampling points via the geoprocessing tree, or the
geoprocessing tool search:
Tools ->Toolbox →Data Management Tools -> Sampling -> Create Random
Points

(Video: Random Sample):

Specify appropriate file for the output
location, name the point feature class
something like ran1000, leaving the
constraining feature class blank, and set the
constraining extent to chirdem.

Specify 1000 as “The Number of Points” and

45 meters as “The Minimum Allowed
Distance.”

We set the value to 45 to avoid points in the

same cells. The cells are 30 m wide, or
about 43 m on the diagonal. If we set all
points to be at least 45 m apart, no two
points will fall in the same cell.

Do not check the box Create Multipoint

Output. Click Run

Now sample the elevation values from chirdem as previously, via

Extract Values to Points. (see top of Page 4)

Use the random_points layer you just created as the input

Name the output something like ran1000_with_elevation. (remember to set interpolate

and append check boxes)

Remove the random_points data layer.

Calculate the mean value of RasterValu for your sampled points, and add it to
the title for the Random Map in your Layout.

How different is the mean value from your systematic sample?

5
GIS Fundamentals Lab 12

Stratified Random Sampling of a Raster Layer

Sometimes we want to vary the sampling frequency across a map. Here, we’ll
place more samples in areas that are steeper. First we’ll create three zones, or
strata, and then we’ll assign samples based on these strata. Samples will be
assigned proportional to both the area of the strata, and the relative steepness,
with more samples in steeper strata.

Our strata boundaries will be based on slope, filtered to create larger, more
generalized areas.

Activate your Stratified Map.

Calculate the slope for chirdem (in the Toolbox, Spatial Analyst
Tools→Surface→Slope), specifying slope in degrees, and a z-factor set to 1,
(careful here, sometimes the Z factor is automatically set to a different value),
and saving to a permanent dataset named something like “slope_deg.” (this output
layer is not displayed in on this page)

Now to create strata. Usually you stratify for some threshold of an attribute, e.g.,
slopes above which you can’t build, or elevations where you’re unlikely to find a
resource of interest.

We’ll reclassify slope_deg into three classes; 1, 2 & 3 respectively using the
categories of 0-1.5, 1.5-18, 18 and larger. The Reclassify tool is in the
ArcToolbox, under Spatial Analyst Tools. Refer to the previous labs if you don’t
remember how to
permanently
reclassify, to a new
data set (Lab 11, page 13;
Lab 10, pages 15-16).

Name your output

RCSclope. Although the
coloring may be
different, the reclassed
output should look
something like figure:

(Video: Reclass Hints)

6
GIS Fundamentals Lab 12

Now we want to reduce some of the “speckle” in

the reclassed slope layer (RCslope). Single
pixels or long, thin areas embedded in another
strata, and if we place a sample point there, we
won’t be stratifying very well. The speckle isn’t
bad here, but sometimes it is, and this is a good
opportunity to practice smoothing (Video:
RemoveRasterSpeckle).

Use Toolbox →Spatial Analyst Tools

→Generalization→Majority Filter.

Specify a FOUR as the Number of neighbors to use and the HALF replacement
threshold.

Your output should look something like the image below.

Notice how the small speckles are removed in the majority filter output, but the
long, thin reaches in valleys remain.
Again, this isn’t a great issue here, but in many cases, we want very general
strata, so we’ll apply another step.

7
GIS Fundamentals Lab 12

To generalize further and reduce the number

of isolated blobs, you may apply a median
filter, via

ArcToolbox ( ) →Spatial Analyst Tools

NeighorhoodFocal Statistics.

Specify the majority filtered layer as input, with

a rectangular neighborhood, and a 31cell
height and width.

Be sure to select the Majority statistics type

and ignore NoData in calculations. The tool
may take quite a bit of time to run, depending
on your computer.

Notice how this substantially generalizes the data, removing the long, thin
features. (results of Focal Statistics are not shown on this page)

Convert the final FocalSt_Majo1 raster to a vector layer in the ArcToolbox via
Conversion Tools→From Raster→Raster to Polygon). You usually want to
simplify the polygons, and then here name it something like “Strata.”

Apply a symbology like that shown below. Inspect the layer’s data table. Note
there are Shape_Length and Shape_Area columns, these are in the projection
units, here meters and square meters. (Note: if you ever get “Table Failed to load data” when
you try an open an Attribute Table, remove the layer from the map and add it back to the map. The table will
now be relinked)

8
GIS Fundamentals Lab 12

Now we must create the stratified sampling points. (read to Page 12 BEFORE your try to
do start pushing buttons; you need to understand the process)

We would like to have a total of approximately 1000 sample points, with 10 times
as many sample points in the steep areas (red, at right) as in the flat (green
areas). We’d like three times as many samples in the intermediate areas
(yellow) as the flat areas.

The weightings are 10 for the steep, 3 for the intermediate, and 1 for the flat.

The total of our weights is 10 + 3 + 1 is 14, so that means

• 10/14ths of our samples should be in the steep areas,

• 3/14ths in the intermediate, and
• 1/14th in the flat areas.

So our samples should be distributed as:

1000 * 10/14 = 714 samples in the steep areas (strata; gridcode = 3),

1000 * 3/14 = 214 samples in the intermediate (strata; gridcode = 2),

and

1000 * 1/14 = 72 samples in the flat (strata;

gridcode= 1).

You might have noticed when we created a random

sample in previous exercises, there was an option to
set the number of points assigned to a polygon
based on a field:

It randomly assigns the number of points into a

polygon that is specified by the field value.

There are several polygons for each type, and so it

makes sense to distribute the samples for that type
proportionally to the area of each polygon.

We weight the assignment based on the polygon

area relative to the total area for the strata.

For example, in my layer above, the largest category, gridcode = 3 polygon has
an area of 255.5 square kilometers, and
the total area for category 3 is 272.3 square kilometers.

9
GIS Fundamentals Lab 12

So, this largest category 3 polygon should get 714 * 255.5/272.3, or 702 sample
points.

We multiply the number of points for the strata (category) by the polygon area,
and divide it by the total area of that strata.

How do we get the total strata area, in square kilometers? Remember, by

• opening the table for the data layer, then
• creating a new float or double column, call it SqKm
• calculating the area into this new column (field calculator on the new
column, SqKm
or

by realizing the square meters can be converted to square kilometers by dividing

the existing Shape_Area by 1,000,000

or

by using the geoprocessing tool Add Geometry Attributes

(video:AddGeomAttributes).

Since we’ve already gone over the first two methods in previous labs, here we’ll
describe the Add Geometry Attributes geoprocessing tool.

You can find it in the Toolbox -> Data Management

group,
or via the Geoprocessing pane search tool.

Either way, start the tool to see a window:

The input features are in the strata polygon layer

you just created

You wish to calculate area, in Square kilometers

You can leave the length unit and coordinate

system blank, the tool can obtain them from what
you’ve entered; if you want to be careful explicitly
select the Current Map.

Run the tool, and note it adds a new attribute

named POLY_AREA with calculated square
kilometers.

10
GIS Fundamentals Lab 12

Review the strata layer polygon table. Note that

you have a gridcode, from 1 to 3 (flat to steep),
and the area for every polygon.

We need to find the total polygon area for each

of the three categories so we can apply our
strata assignment formula.

We do this with the Summary Statistics tool

(you’ve used it before; Lab 8; page 7),
summarizing by the gridcode, which is the strata
ID.

The input table is the from the polygon strata

layer, we specify an output table, and we are
asking to summarize the field POLY_AREA to
sums, based on the Case Field of gridcode.

This case field means in all cases where the values for gridcode are the same,
sum them, e.g., sum all the areas for gridcode 1, do a separate sum for gridcode
2, and another for gridcode 3.

In my case, this results in a calculated area of:

• 272.75 square kilometers for the flat (gridcode = 1) strata, and I want 72
sample points in this strata
• 182.35 square kilometers for the intermediate (gridcode = 2) strata, and I
want 214 sample points in this strata
• 217.93 square kilometers for the steep (gridcode = 3) strata, and I want 714
sample points in this strata

Your numbers may be slightly different, but should be with a few percent of these
areas if you used the methods we described above.

Write down your summary areas for all three of your classes, you will need
them later

We now want to calculate the number of samples per polygon.

(Video: Stratified Random)

Samples per polygon may be computed many ways, but we’ll:

1) Open the strata layer attribute table and add a new long integer field, call it
samples (see previous labs if rusty on how to do this: (note if Add if grayed out
check to see if you have “unsaved” edits, if so save them. If not save your Project and close ArcGIS
Pro and reopen it and reopen you Project.)

11
GIS Fundamentals Lab 12

2) Select all polygons for a given strata (gridcode). You’ve

done this before, as a quick reminder, you start with the
select by attributes tool:
3) Use the field calculator (to calculate a value into the Samples field (aka. column)) to
multiply the total number of points for this stratum (e.g., 714 for the steep
strata, gridcode 3) by the area of the polygon, divided by the total area of
the strata (in this my case 217.93). In other words, build the formula
POLY_AREA/217.93 * 714, where the 714 is the number of total points I
want for that strata (gridcode), and the 217.93 is the total area for that
strata (gridcode). (note: you may have used SqKm for the name of the total strata area)

4) Repeat this select calculate & process for strata 2 & 1, substituting the
strata area and total points appropriate for each strata.
To repeat, our general formula is:
POLY_AREA/(Total area of strata)
* (Total sample points for that
strata)

(For my data (your should be close) the

formulas are:
Gridcode = 1 POLY_AREA/272.75 * 72
Gridcode = 2 POLY_AREA/182.35 * 214
Gridcode = 3 POLY_AREA/217.93 * 714)
Note that your numbers for the strata
area and relative number of samples may
be different than those I’ve used here, but
they should be close

When you have finished with the last sample number calculation, be sure to
clear all selected features

Now, to create the stratified random points.

We use the same random assignment tool as

before,
ArcToolbox -> Data Management Tools ->
Sampling -> Create Random Points.
Specify an output location and file name,
something like StratRand.

The Constraining Feature Class is the vector

layer you created with a column that specifies
the number of samples for each polygon.

Make sure to select the Field radio button and

identify the proper field to specify the number of
points, the column you calculated (samples in
this example).

12
GIS Fundamentals Lab 12

As with the other two Maps, Calculate the mean elevation with the Extract
Values to Points tool, and include the mean in the label for the stratified Map on
your Layout.

Add the stratified random sample layer polygon layer and the points to your
layout.
The map part of your layout should appear similar to the figure below:

Add a North arrow, legend, title, your name, and scale bar, create a PDF, and
turn it in on the course Canvas site.

13
GIS Fundamentals Lab 12

Interpolation
Add a new map, either in your existing project or in a new project, name the new
map Thiessen, and change the Map display coordinate system to UTM NAD83
(2011) zone 18N coordinates.

As you learned in Lab 8, use the

Toolbox→ Conversions to convert
the Excel spreadsheet to a ArcGIS
Table (within a your existing Lab12
GeoDatabase); the Excel file is
Ozone, Sheet 1, once the ArcGIS
Table is added to your Table of
Contents

• right click on the added table

name in the Table of Contents
and then
• Select Display XY data, then
use Site_Latitude as the Y field
and the Site_Longitude as the X
field, and finally
• Set the Output features to a
new layer, named something
like OzGeog
• Leave the Coordinate System
set to GCS_WGS_184 at this
set and click RUN
• This new layer is “on the fly” changed to UTM but we want it to be
“permanently” set to UTM coordinates. To do this use the ArcToolbox to
reproject (Data Management→ Projections and Transformations→Project) the exported layer
to the UTM NAD83 (2011) zone 18N coordinates, naming it something like
OzUTM.
• Remove the OzGeog and the loaded Excel table from your Map.

Display your data with graduated symbols, and with a topographic map
background, to appear something like the figure on the right. (Video:
InXY&Thiessen)

The spreadsheet data you loaded into OzGeog then to OzUTM, contains data on
ozone levels, a toxic element that may reach harmful levels, especially in the
summer, and in areas with large numbers of internal combustion engines.

The ozone column in the data set reports 24-hour average levels.

Readings higher than 54 are a moderate health risk, and readings higher than 70
are damaging to most populations.

14
GIS Fundamentals Lab 12

Readings are made at individual stations, but we’d like to create maps of the
concentrations across the northeastern U.S.

To do so, we must interpolate the readings

across space.

Open the Thiessen Polygon Tool, Toolbox →

Analysis Tools→Proximity→Create Thiessen
Polygons

Set your point layer with ozone concentrations as

the input, set an appropriate output Feature
class, and output all fields.

After running, you should get a set of Thiessen

polygons. Color those with safe ozone levels (<54) as transparent, warning levels
(54 to 71) as semi-transparent (about 65%) yellow, and damaging levels (> 71)
as semi-transparent red (see figure) (hint right click on the color and use Color Properties to
select the transparency level.)

15
GIS Fundamentals Lab 12

Change the title of this Map to “Thiessen”.

16
GIS Fundamentals Lab 12

Geostatistical Wizard
We will now apply a series of interpolations using the interpolation techniques.

Create three new Maps, one labeled IDW, one labeled Kernel, and the last
labeled Kriging. (Video: GeoStatWiz)

Add the OzUTM data you used above to the IDW Map.

Many interpolation tools are provided through the Geostatistical Wizard, found
first by activating the Analysis tab on the main window (below left), then the
Geostatistical Wizard icon toward the right side of the tool ribbon (below right).

Double clicking on the Geostatistical Wizard (GeoWiz from here on) should
display a method selection window, with radio buttons next to various spatial
prediction methods:

17
GIS Fundamentals Lab 12

As you click each empty circle to choose a method, the parameters and
description of the method change. In the figure above, we’ve selected the
inverse distance method. Some of the names are different that those in the
textbook, e.g., the global polynomial method is the same as our trend-surface
method, and some of the methods are embedded within the choices, e.g., our
kernel interpolation is under their list of Interpolation with barriers, we would just
select this and apply without any barriers.

For now, select the Inverse Distance Weighting

method, using the OzUTM data you created, and
Ozone as your data field. Leave the Weight Field
blank.

On clicking Next, you will likely get a warning saying multiple samples exist at the
same location. If so, select “Use Mean,” then Next

You should see a window similar to that above, with a preview interpolation
image on the left and a set of parameters on the left. Most should be familiar
from the textbook readings, except perhaps the Sector type and Angle options,
which allow you to apply the minimum/maximum neighbor thresholds to sectors,
rather than globally, e.g., with 4 sectors, you would allow 10 points each in each
of the northeast, northwest, southeast, and southwest directions, and constrain
the fits in each direction to match well at the edges of the quadrants. Sectors

18
GIS Fundamentals Lab 12

often provide better fits when there are underlying local trends, but details are left
for a more advanced class.

Vary the parameters, e.g., powers from 1 to 6, different neighborhood types, and
numbers of maximum and minimum neighbors, and observe how it changes the
preview surface.

Set the parameters with a Power of 1, Standard Neighborhood, 10 Max

Neighbors, 4 Min Neighbors, a Sector Type of 1, and 0 Angle, and then click
Next. This should show and error graph and cross-validation window:

The panel on the left side shows the predicted vs. observed values, a regression
line through the plot, and a 1:1 reference line of perfect prediction. You can also
click on the Error tab above the window that shows the error for each
measurement, the difference between the point on the plot and the 1:1 line.

The right side shows the cross-validation statistics (explained in the readings).
Note the RMSE of the cross-validation for this set of parameters.

Click on the back button, and only change the Power parameter, changing it from
1 to 2 on the preview interpolation screen. Then hit next again and view the
graph and cross-validation RMSE. How have they changed?

19
GIS Fundamentals Lab 12

Click Finish and you will see a Method Report,

click Ok.

This should display a set of filled contours for

your predicted surface, with something like 10
classes.

Click on the Classes option and set to 3.

Then click on the color patches in the Class Breaks

table, and set the first to transparent, with an upper
value of 54,

The second class at 71 with a yellow/orange color,

and the third class at 89 or higher, with a red color.
Make the second and third classes semi-transparent,
with a 60% transparency.

You should get a surface approximately equal to that

on the next page:

20
GIS Fundamentals Lab 12

Compare this to the Thiessen polygon surface you produced earlier.

Activate the (empty) Map you created earlier and labeled Kernel, and add the
OzUTM data layer to it, symbolizing with proportional symbols as above.

Start the GeoWiz tool again, and choose the

Kernel Interpolation method:

Set the source dataset as the Ozone UTM point

layer you created.

Set the data field to ozone concentration and

leave the Weight Field and Barrier Features
datasets blank.

Go to the Next window and use the mean value if

you get the duplicate points warning.

21
GIS Fundamentals Lab 12

The next window shows the general options for the kernel estimator applied to
the ozone data.

The most important

choices are for the
Kernel Function,
Bandwidth, and Ridge
parameters.

Leave the Ridge at 50

for your fits, but explore
how the different
choices for Kernel
Function change your
error plot, RMSE, and
RMS-Standardized
(below right)

Try the Fifth-order

polynomial, Gaussian,
and Exponential Kernel
Functions.

Which has the lower

errors of prediction?

Choose a set of
parameters and
estimate ozone
surface for the study
region.

Create the filled

contours symbology
as before, using the
same colors and 54
and 70 thresholds as
in your previous Map.

22
GIS Fundamentals Lab 12

Conduct one final interpolation. Open/activate the empty Map you created and
labeled Kriging (or create it now).

Add the ozone data to your map and start the GeoWiz tool.

Select the Kriging/Cokriging, then Next, and

Use Mean for coincident points, then Next.

There are several variations on Kriging, and the options change substantially
depending on the Kriging methods you choose.

The point here is just to introduce you to the GeoWiz

tool methods, there are entire courses on
geostatistics that give a nuanced understanding.

For now, choose Simple Kriging, Prediction option:

Try at least both the Normal Score and None for

transformation type.

In the Normal Score method, accept the default

parameters for the first and second pages, and spend
some time digesting the 3rd page (example shown
below).

The covariance /
semivariogram
plot is
probably
the most
informative part
of this panel.
They show the
model fit in the
graph to the right,
and the
distribution of
binned samples
in the graph.

Nugget, sill, and

other information
is also provided.

23
GIS Fundamentals Lab 12

Hitting Next shows a

window that covers
model application,
with defaults for the
neighborhood and
configuration over
which to estimate
coefficients, and apply
the predictions.

Again, there are many

options for number of
neighbors, sectors,
and neighborhood
type.

The choices depend

on your expectations
about the behavior of
your variable of
interest and can be guided by diagnostics that are more appropriate for a more
advanced treatment. You may experiment a bit, then hit Next to display the by
now familiar cross validation page.

Inspect the
distribution of
predicted and error
points, and note
the RMSE and
standardized RMS.

Click on finish to
estimate a surface.

Again, symbolize
the estimated
surface equal to
the previous
methods. (Manual, 3,
54;71;89, transparent,
60%, 60%)

24
GIS Fundamentals Lab 12

Display all 4 Maps together in a layout.

To make it easier on yourself, remove the background topographic or other ESRI

Basemap, it is often rather slow and fussy to set the extent and scale when these
are displayed.

Add the vector shapefile STATES_GEOG and show only the state outlines (no fill
for polygons).

Set the Maps with the same scale and make each map box the same size (right-
click Map Properties, see previous labs for details).

Also include the usual elements: Title, name, legend, scale bar, north arrow:

25