Lab 5:
Pattern Analysis
Spatial Statistics Tools for
Pattern Analysis
This exercise involves using
ArcGIS Spatial Statistics tools to analyze the spatial patterns of bike-route
densities of the neighborhoods within the Metro urban growth boundaries (UGB).
Follow the instructions below and answer all questions at the end of this
exercise.
A. Data preparation
- Start ArcCatalog
and create a personal (or file) geodatabase call
“lab5” in a folder you have read/write access.
- Start ArcMap,
connect to I:\Students\data\GIS\RLIS\2007_Nov\ESRI Shapefiles\,
and add UGB_FILL.lyr and NBO_HOOD.lyr in the BOUNDARY folder
and BIKE_RTE.lyr in the TRANSIT folder to the current data frame.
You may have to re-connect the layers to their source data is the path is
broken (go to layer properties, source tab, set data source). You will
find the neighborhood map includes neighborhoods outside UGB and there are
several different types of bike routes.
- Click on the Arctoolbox icon in ArcMap to
open the Arctoolbox and select “Analysis Tools
-> Extract -> Clip” tool. In the Clip tool dialog window, use
the pull-down selecting list to specify “nbo_hood”
as the input feature and “ubg_fill” as the clip
feature. Navigate to the lab5 geodatabase (lab5.mdb) and set “NBO_HOOD_Clip”
in the geodatabase as the output feature class.
Leave the Cluster Tolerance field blank. Click OK. When it’s done, the NBO_HOOD_Clip is added to ArcMap.
Open the attribute table of NBO_HOOD_Clip and compare
the difference with the attribute table of nbo_hood.
- Next, we want to select the
bike routes that are cyclist-friendly. Open the attribute table of bike_rte (i.e., bike_rte.shp), click on the “options”
button, and select the “Select by Attributes…” item. Use the SQL query
builder to create the following select statement:
"BIKEMODE" = 'Bike lane' OR
"BIKEMODE" = 'Regional multi-use path' OR "BIKEMODE" =
'Local multi-use path'
- Click Apply and close the SQL
dialog window. The selected bike routes in ArcMap
are high-lighted. Open the Clip tool and specify “bike_rte”
as the input feature, “ubg_fill” as the clip
feature, and “bike_rte_Clip” as the output
feature class. Click OK. When it’s done, the “bike_rte_Clip”
is added to ArcMap.
- Now, we can calculate the
bike-route density for the neighborhoods within UGB. This can be done with
a spatial join or the identity tool. However, the results from both
methods could be different. We use the spatial join method first.
Right-click on the NBO_HOOD_Clip layer in ArcMap and select “join…” from the “joins and relates”
menu. Select “join data from another layer based on spatial location” for
“What do you want to join to this layer?” field and set bike_rte_Clip as the layer to join. Read the
descriptions of the default joining method carefully and check the
checkbox next to sum. Specify the output as a feature class called “NBO_bikerte_sj” in lab5.mdb. Click OK to finish
spatial joining.
- Open the identity tool
in the “Analysis tools -> Overlay” toolset. Specify bike_rte_Clip
as the input feature, NBO_HOOD_Clip as the
identity feature, and NBO_bikerte_id as the
output feature class in your lab5 geodatabase. Use ONLY_FID as the value for JoinAttributes and click OK.
- Open the summary
statistics tool in the “Analysis tools -> Statistics” toolset.
Specify NBO_bikerte_id as the input table and NBO_rte_sum as the output table in lab5.mdb. Select Shape_Length as the statistics field. The Shape_Length is added to the table below. Click on the
blank cell in the Statistics Type column next to Shape_Lenght
and select SUM. Select FID_NBO_HOOD_Clip for the
case field. Click OK.
- After the tool finished, add
the NBO_rte_sum table to ArcMap
and open it. This table summarizes the total bike-route length for each
neighborhood in UGB. Frequency refers to the number of routes (i.e., line
features) in each neighborhood polygon. The values in FID_NBO_HOOD_Clip
are the OBJECTIDs of the neighborhoods. We can use this field to join the
route-length data to the neighborhood polygons.
- Right-click on the NBO_bikerte_sj layer and select “join…” from the joins
and relates menu. Select “join attributes from a table” for “What do you
want to join to this layer?” field. Select OBJECTID as the field the join
will be based on in the NBO_bikerte_sj table,
set NBO_rte_sum as the table to be joined, and
select FID_NBO_HOOD_Clip as the field the join
will be based on in the NBO_rte_sum table. Click
OK.
- We have finished the
calculation of bike-route length in each neighborhood. Notice that the
values of route length derived from the two methods are different. We
continue to calculate the bike-route densities using only the results
derived from the identity method. Exit ArcCatalog
before proceeding to the following steps. Open the attribute table of NBO_bikerte_sj. Click on the options button and select
“add field….” Enter “route_density” as the new
field name and set the field type to “float” (i.e., floating-point
numbers). Click OK to add the field to the attribute table of NBO_bikerte_sj. Right-click on the heading row of the
“NBO_bikerte_sj.route_density” field and select
“Calculate values….” Click YES to the message window popped up and enter
the followings in the dialog window.
[NBO_rte_sum.SUM_Shape_Length] / [NBO_bikerte_sj.Shape_Area]
- Click OK and answer YES to the
message window popped up. The route density is calculated.
B. Analysis
- You can now remove all joins
from the NBO_bikerte_sj attribute table.
Right-click on the NBO_bikerte_sj layer and
select “Remove all joins” from the “joins and relates -> remove join(s)”
menu. Use the route_density field to display the
NBO_bikerte_sj map in ArcMap.
Visually check if there are any conspicuous patterns on the map. We will
use several spatial statistics tools to quantify these patterns.
- Read ArcGIS Desktop Help on topics
related to the following tools:
Spatial autocorrelation (Moran’s I)
High/low clustering (Getis-Ord
General G)
Cluster/Outlier Analysis with Rendering
Hot Spot Analysis with Rendering
- Open the “Spatial
autocorrelation (Moran’s I)” tool (script) in the “Spatial Statistics
Tools -> Analyzing Patterns” tool set. Set the input feature class to NBO_bikerte_sj and the input field to route_density. Use the default values for other
fields. Check the display output graphically checkbox and click OK. While
the tool is running, make sure that the “close this dialog when completed
successfully” checkbox is unchecked. Read the descriptions and results on
the graphic output window careful. Note that “0 (zero)” is used for
Distance band or Threshold Distance to indicate that there’s no cutoff
distance for calculating weights (i.e., all features are used in the
calculation).
- Open the “High/low
clustering (Getis-Ord General G)” tool
(script) in the “Spatial Statistics Tools -> Analyzing Patterns” tool
set. Set the input feature class to NBO_bikerte_sj
and the input field to route_density. Check the
display output graphically checkbox and click OK. While the tool is
running, make sure that the “close this dialog when completed
successfully” checkbox is unchecked. Due to a bug in the script, the tool
appears to be still running before you close the
graphic output window. Before you close the window, read the
descriptions and results on the graphic output window carefully. After
examining the graphic output, click close to dismiss the window. Now, the
script should print out the statistics in the message window and stop
running.
- Open the “Cluster/Outlier
Analysis (Anselin Local Moran’s I” tool in
the “Spatial Statistics Tools -> Mapping Clusters” tool set. Set the
input feature class to NBO_bikerte_sj, the input
field to route_density, the output layer file to
“Local_I”, and the output to “Local_I” as a shapefile.
- Next, we will use the “Hot Spot Analysis (Getis-Ord Gi*)” tool in
the “Spatial Statistics Tools -> Mapping Clusters” tool set. Set the
input feature class to NBO_bikerte_sj, the input
field to route_density, the output to “Local_G” as a shapefile, the
Conceptualization of Spatial Relationships as Inverse Distance, and the
Distance Band or Threshold Distance to zero.
Questions:
- What do “AREA” and “Shape_Area” refer to in the attribute table of NBO_HOOD_Clip?
- There is a record in the NBO_rte_sum table having an FID_NBO_HOOD_Clip value of -1. What does this mean?
- Explain why the bike-route
lengths calculated using a spatial join and the identity tool are different.
- Did the bike-route densities
measured at a neighborhood scale exhibit a clustered pattern? In what
ways? What are the Moran’s I, Getis-Ord
General G, and their associated Z values?
- Print maps showing the
distributions of Local Moran’s I (i.e., LMiIndex),
z scores of Local Moran’s I (i.e., LMiZscore),
and the z scores of Getis-Ord Gi* (i.e., GiZScore) of
neighborhood bike-route densities within Metro UGB.