Home › Courses › GIS and Anthropology › Weekly In-Class Exercises

Week 10 - Viewshed and Cumulative Viewshed Analysis

Week 10 In Class Exercise

VIEWSHED ANALYSIS


There are several types of viewshed analysis available in ArcGIS.
We will visit two of these types using the Callalli study area.

Instead of using the data from before, I've prepared new files because we need a larger DEM and a Sites shapefile with time periods assigned.

Download the View.Zip file and uncompress to a path with no spaces.

As with other data used in this class, this Sites dataset is based on my dissertation data set, but the attributes are not the actual data so don't compare this Sites file with what is in my dissertation.
 

I. Setup the project in Arcmap.

  • Add the Sites shapefile.
  • Add the 30m ASTER DEM from \dem_large\COLCA_ASTER
  • Add the hydro layer from the Callalli geodatabase known as "Hydro_lines100k" if you have it.

Look at the coordinates. Is this in UTM or in Decimal Degrees? With Viewshed analysis linear units (like Meters) are necessary to accurately calculate distances.

Look at the Attribute table for Sites. You're familiar with most of these attributes from previous labs.

We'll use the "Period" field to differentiate sites from different time periods in this study.

The values are

Formative (2000BC - AD400)
MH Middle Horizon (AD400-AD1100), there are no MH sites in this dataset
LIP Late Intermediate Period: (AD1100-AD1476)
LH Late Horizon (AD1476-AD1532)

 

2. How do we control the viewshed calculation?

Please see the explanation on this webpage: ArcGIS Help - Viewshed

The optional viewshed parameters are SPOT, OFFSETA, OFFSETB, AZIMUTH1, AZIMUTH2, VERT1, and VERT2<. These will be used if they are present in the feature observer attribute table.

The only two attribute variables we are concerned with at this point are OFFSETA (height of viewer) and RADIUS2 (the maximum radius of the viewshed calculation).
The viewer's eyes are at 1.5m above the ground, and we want to know the visible area within 5 km of each site.

  • Make sure nothing is selected.
  • Open the Attribute Table for the Sites layer
  • Choose Options > Add Field…
  • Call it OFFSETA, make the type FLOAT and the Precision 2 and the scale 1.

Precision sets the number of significant digits, while the Scale is the location of the decimal point from the right side of the value. This allows the field to store 1.5.

  • Use Calculate Values... command to put 1.5 in the field for the whole table.
  • Choose Add field... again
  • Call it RADIUS2 and leave it as the default, short integer

Since we want to limit the value of the view to 5 km, what should the value be in that goes in this field? What are the spatial units of this map?
You can use Calculate Values... command to fill the RADIUS2 field with the value 5000 all the way down.
 

II. Quick Viewshed analysis

  • go to Spatial Analyst toolbar > Surface Analysis > Viewshed…
Input: Colca_Aster
Observer Points: Sites
Use the defaults and allow the output to go into Temporary.


Look at the result. What is this showing? Is this informative?

  • Put the Colca_HS (hillshade) layer above the Viewshed output at 60% transparency.

This help you to visualize what is happening.
 

What are some problems with this output?
If we use this quantitatively do you see any potential issues with edge effects?


To quickly resolve problems in this dataset we have to subset the data:

  • Zoom out so you can see the project (click the globe)
  • Right-click the Sites layer in the table of contents,
  • Choose Select > Select All
  • Choose the selection arrow (white arrow on blue shape) and draw a selection box around all sites EXCEPT the two sites in the top-left part of the study area.
  • Right-click the layer again and choose Export Data… and save out the shapefile and call it "Sites_sub". Add the new layer and remove the old one.
Can you think of another way to solve the Edge effects problem? Two simple solutions: (1) throw out samples close to the edges, (2) try to acquire a larger DEM in the first place.

III. Viewshed analysis - differences between time periods

What is missing from the previous analysis was a theoretical question guiding the inquiry. Obviously, you can select a single site and run a viewshed and find out the resulting area. But how is this different from just going to the site and looking around and taking photos?
You could also select a line and find the viewshed along a road, for example. However, computers can answer interesting questions by looking at general patterns with larger datasets. This is where computers give you an advantage because you're repeating many small operations.

We want to answer questions such as:

How do viewsheds change based on the type of site that is being evaluated?
Are Late Intermediate sites more or less likely to be constructed such that they can view one another?

We can calculate separate viewsheds for each group of sites in our table by manually selecting each group from the Attribute Table and running the viewshed calc.
This is a good application for ModelBuilder, except that Batch processing is not available in ModelBuilder until version ArcGIS 9.2 and we are still using ArcGIS 9.1 in the lab.
 

1. Calculate viewshed for each major group:

  • Open the Sites_sub Attribute table, Sort Ascending on the "Period" field.
  • Select only the Formative sites
  • Use Spatial Analyst > Surface Analysis > Viewshed…
  • Save the result as "Form5km" in the view folder
  • MAKE SURE the INPUT SURFACE IS COLCA_ASTER.
  • Repeat the above process two more times with LH and LIP, calling the output GRID files LH5km and LIP5km respectively.

2. View the results

Have a look at the output GRID Attribute table. The VALUE field indicates how many of the sites can view a given cell, and the COUNT field indicates how many cells fall into that category.

In other words, in Form5km GRID the VALUE = 3 row shows that 1716 cells are visible from 3 sites but no one cell is visible from all four Formative sites in this group.

We could calculate the number of cells in view per site, or other types of averages.
For example, in the Sites Attribute table there is a SiteType field that indicates that some of the LIP sites are Pukaras. These are hilltop fortresses that have notoriously good views. However… there's a problem. The problem lies in the bias sample we have because sites tend to aggregate around resource patches or other features.
In these areas our dataset has high "intervisibility" because it reflects aggregation within the 5km buffer.
In other words, many sites can see a particular cell in those situations where there are many sites in a small area!

We can get around this problem by calculating viewshed for random site locations as representative of "typical" views in the area. We will then compare our site views against this typical viewshed.

IV. Cumulative Viewshed Analysis

Another way to calculate viewsheds is more powerful yet because it allows you to compare the result against a "typical" viewshed for the area.

In version 9.1 of ArcGIS we still need to use the command line GRID program to accomplish this task. Thus we have to convert out files to coverage or GRID to view them in the older GRID program.

I created a Shapefile with 500 random points was created in Hawth's Tools (available from Arcscripts and SpatialEcology.org, as you learned previously in this class) and there is a minimum of 50m spacing between each point.

Download Random.zip and unzip into the same folder as the Colca DEM folder.

Look at the Attribute table and note that the OffsetA and Radius2 fields were added, as above.

We must convert the Rand500 point file to Coverage in order to use it with GRID.
Choose
Toolbox > Conversion > To Coverage > Feature Class to Coverage….

Save the coverage as "rand500_cov" into the /view/dem_colca/ folder.
That way everything is in one folder.
 

  • Type
    &wo and the complete path to your /view/dem_colca/ folder
  • Issue the following command
    Randvisib = visibility ( colca_aster , rand500_cov, point , frequency )


When it completes return to Arcmap and add the resulting RandVisib grid.

Type 'q' to quit GRID.

Now use the Spatial Analyst > Zonal Analysis ...
Tool to answer the question: Sites from which of the three time periods have the highest mean visibility with respect to the RandVisib grid?