Transcript Visually Comparing Maps

Spatial Data Mining
Practical Approaches for Analyzing Relationships
Within and Among Maps
“GIS technology is rapidly moving beyond mapping and
spatial database management to analytical capabilities that
assess spatial relationships within decision-making contexts”
(JKB)
Presented by Joseph K. Berry
W.M. Keck Scholar in Geosciences, University of Denver
Berry & Associates // Spatial Information Systems
2000 S. College Ave, Suite 300, Fort Collins, CO 80525
Phone: (970) 215-0825 Email: [email protected]
…visit our Website at
www.innovativegis.com/basis
Visually Comparing Maps
I bet you've seen and heard it a thousand times
before  a speaker waves a laser pointer at a
couple of maps and says something like "see how
similar the patterns are."
But just how similar are the maps?
…what proportion has the same
classification?
…where are they different?
…where are they the same?
and of course, your response should be
objective and repeatable
(Berry)
Approach for Comparing Discrete Maps
(Berry)
Coincidence Summary Results (Table 1)
(Table 1)
(Berry)
Proximal Alignment Results (Table 2)
(Table 2)
(Berry)
Approach for Comparing Map Surfaces
(Berry)
Statistical Test Results (Table 3)
…Statistical Tests of entire surface or
partitioned areas
(Table 3)
(Berry)
Percent Difference Results (Table 4)
…Percent Difference
between two map
surfaces
(Table 4)
(Berry)
Surface Configuration Results (Table 5)
The two superimposed maps at
the left side of figure show the
normalized differences in the
slope and aspect angles (dark
red being very different). The
map of the overall differences
in surface configuration
(Sur_Fig Index) is the average
of the two maps.
(Table 4)
Note that over half of the
map area is classified as
low difference (0-20)
suggesting that the two
surface maps align fairly
well overall.
(Berry)
Visualizing Spatial Relationships
Interpolated Spatial Distribution
Phosphorous (P)
What spatial
relationships do you
see?
…do relatively high levels
of P often occur with high
levels of K and N?
…how often?
…where?
(Berry)
Calculating Data Distance
…an n-dimensional plot depicts the multivariate distribution; the distance
between points determines the relative similarity in data patterns
…the closest floating ball is the least similar (largest data distance) from the comparison point
(Berry)
Identifying Map Similarity
…the relative data distance between the comparison point’s data pattern
and those of all other map locations form a Similarity Index
The green tones indicate field locations with fairly similar P, K and N levels; red tones indicate dissimilar areas.
(See Map Analysis, “Topic 16, Calculating Map Similarity” for more information)
(Berry)
Clustering Maps for Data Zones
…a map stack is a spatially organized set of numbers
…groups of “floating balls” in data space
identify locations in the field with similar data
patterns– data zones
…fertilization rates vary for the different
clusters “on-the-fly”
(Cyber-Farmer, Circa 1990)
Variable Rate Application
(Berry)
Evaluating Clustering Results
(Berry)
Map Surface Correlation/Regression
Histogram/Map View-- Data Space (joint magnitude of values)
are linked to Geographic Space (position of values)
(Berry)
Creating Prediction Models (Scatter Plot)
…a Scatter Plot identifies the “joint condition” at each map
location; the trend in the plot forms a prediction equation
(Berry)
Deriving a Predictive Index (NDVI)
…an index combining the Red and NIR maps can be used to generate
a better predictive model
Normalized Difference Vegetation Index
for the sample grid location
NDVI= ((NIR – Red) / (NIR + Red))
NDVI= ((121-14.7) / (121 + 14.7))= 106.3 / 135.7= .783
(Berry)
Evaluating Prediction Maps (Spatial error analysis)
…the regression equation is evaluated and the predicted map is compared
to the actual measurements to generate an error map
Error = Predicted - Actual
for the sample grid location
Yest = 55 + (180 * .783) = 196 …error is 196 – 218 = 22 bu/ac
Exercise #8c, page 30 – Create a regression model
relating Yield and NDVI
Note that the average error is 2.62 and 67% of the predictions are within +/- 20 bu/ac
Also, most of the error is concentrated along the edge of the field
(See Map Analysis, “Topic 16, Predicting Maps” for more information)
(Berry)
Stratifying Maps for Better Predictions
Stratifying by Error Zones
Other ways to stratify mapped data—
1) Geographic Zones, such as proximity to the field
edge; 2) Dependent Map Zones, such as areas of low,
medium and high yield; 3) Data Zones, such as areas
of similar soil nutrient levels; and 4) Correlated Map
Zones, such as micro terrain features identifying
small ridges and depressions.
The Error Zones map is used
as a template to identify the
NDVI and Yield values used to
calculate three separate
prediction equations.
A Composite Prediction map is
created by applying the
equations to the NDVI data
respecting the template map
zones.
(See Map Analysis, “Topic 16, Stratifying Maps for Better Predictions” for more information)
(Berry)
Assessing Prediction Results
Actual
Yield
Error Map
Stratified
Prediction
Error Map for
Stratified Prediction
none
80%
Whole Field
Prediction
none
(Berry)
The Precision Ag Process (Fertility example)
As a combine moves through a field 1) it uses GPS to check its location then 2) checks
the yield at that location to 3) create a continuous map of the yield variation every few
feet. This map 4) is combined with soil, terrain and other
maps to derive a 5) “Prescription Map” that is used to
Steps 1)–3)
6) adjust fertilization levels every few feet in the field.
Prescription Map
Step 5)
45.00
On-the-Fly
Yield Map
40.00
Map Analysis
Step 4)
Farm dB
Step 6)
35.00
Cyber-Farmer, Circa 1992
…come a long ways baby
Zone 3
30.00
25.00
Zone 2
20.00
15.00
10.00
5.00
Zone 1
5.00
10.00
Variable Rate Application
15.00
20.00
25.00
30.00
(Berry)
Spatial Data Mining
…making sense out of a map stack
Mapped data that
exhibits high spatial
dependency create
strong prediction
functions. As in
traditional statistical
analysis, spatial
relationships can be
used to predict
outcomes
…the difference is
that spatial statistics
predicts where
responses will be
high or low
(Berry)
Spatial Data Mining
Practical Approaches for Analyzing Relationships Within and Among Maps
Presented by Joseph K. Berry
W.M. Keck Scholar in Geosciences, University of Denver
Berry & Associates // Spatial Information Systems
2000 S. College Ave, Suite 300, Fort Collins, CO 80525
Phone: (970) 215-0825 Email: [email protected]
…visit our Website at
www.innovativegis.com/basis