Lab 4: Summarizing fields, making graphs, and assembling a layout

A map is good for showing what is next to what, and for showing contrast between neighbors. But it's hard to tell from a map whether one region is, for example, twice as dark as another. A graph can complement the map, making the quantitative values visible. I'll arrange the map together with its graph on a page layout, along with other accessories such as a legend.

This lab took a huge amount of time for various reasons:

ArcMap's graphing is awkward and hard to use, especially trying to change the marker colors and trying to control the order of categories in a bar graph. The help screens often produce errors and crash the program, and the other documentation under C:\arcgis\Digital_Books 8p3 didn't answer my questions. I had to do a lot of tedious trial and error to understand the different options.
The Data Dictionary supplied to us didn't include all of the layers and didn't explain what a lot of the categories meant. For example, it didn't explain why the "roads" layer was different from "streets", and which field might represent the size of a road. (I finally figured this out after a lot of web searching and examining the different fields.)
There seem to be some errors in the data, such as missing values of H_A; see discussion below. It took some experimentation to deal with these records.
Mainly, it takes a lot of exploration to decide what the data mean and how to represent it best. Which fields have meaningful correlation with each other? How many different categories can be symbolized on a map before it becomes unreadable? How should data be summarized: display the maximum and minimum, or the standard deviation, or just the average? I spent a lot of time trying out different answers to these questions. (It would probably have helped if I'd read the background on the fire model first before diving into the database.)

1. Summarizing a field

I'll look at the wildland vegetation layer and the Fuel Model field, which is a set of standard classes developed by the Forest Service's Northern Forest Fire Laboratory. (These classes describe only the wildlands; the developed areas are recorded as "Residential" or "No Fuel Model".) The Fuel Model is one of the inputs to the BEHAVE computer model:

BEHAVE inputs	BEHAVE outputs
vegetation type fuel model development stage crown potential tree height slope class weather conditions (wind, temperature, moisture)	rate of spread flame length fire intensity (heat per area) crown potential

I'll concentrate on fire intensity, which is the field called HSUBA ("H-sub-A", H_A, heat per unit area) in the attribute table, by summarizing the fire intensity for each of the fuel models that occur in the study area. Summarizing will mean losing the details of what BEHAVE calculated for each polygon, but we can't display all the exact values on a map anyway, since they would have to be represented by just a few discrete colors. I wanted to know how much the data varied within each Fuel Model, so I requested average, maximum and minimum values of HSUBA. However, some of the minimum values seemed suspiciously low; it turned out that (as seen by sorting on HSUBA) four polygons have HSUBA recorded as 0, which seems wrong; it must be missing data for some reason.
Bad HSUBA

I decided to eliminate these polygons, as well as the Residential polygons which weren't covered by the BEHAVE calculation, by selecting only records with HSUBA > 0, and re-doing the Summarize request with the box checked to summarize only on the selected records.
Summarize dialog box

(Note to myself: when re-opening a map, look for the summary tables in the Source tab of the map table of contents, not the Display tab.)

2. Making a graph It took some exploration before I understood the different options for graphs. This simple tutorial from Statistics Canada discusses some appropriate, and inappropriate, uses of different types of graphs.

At first I wanted to visualize not just the average of the data, but also its range. A high-low-close graph seemed ideal for this, but it can only be made with a horizontal x-axis, which doesn't leave enough room to show the names of the fuel models. The only way to show the category names is to use one of the horizontal bar graphs. I tried graphing the maximum, mean and minimum:
bar graph of max, mean, min HSUBA
I finally decided that this was too complex, and the maximum, mean and minimum were all fairly close together anyway, except for some of the lower heat categories. The important information to communicate is that Chaparral is way above all the others, with twice as much heat as Southern Rough in second place. Therefore, I decided to show only the average, and to focus attention on the larger averages with a Pareto (i.e., sorted) bar graph:
pareto bar graph
Meanwhile, what should be on the map? It's hard to distinguish eight different kinds of area using only black and white! I decided to focus attention on the top four categories by using shades of gray, and to represent the others using various kinds of hatching from the symbol menu. The polygon borders were set to no color to make the map less complicated to look at. (It would be nice to match the colors of the bars on the graph to the symbology on the map, but the dialog box for selecting marker colors is extremely tedious to use; it claims that patterns can be applied to bars, but they don't appear.) I re-ordered the Fuel Model categories in the Symbology tab so they matched the order in the bar graph.

Finally, I think it needs some street names as familiar landmarks to orient the reader. I have to use the "streets" layer from the Census, even though it has less positional accuracy than the "roads" layer from the USGS 7.5' Quadrangles, because it's the only one that has street names. It will be more recognizable if major roads are thicker, so I symbolized the streets according to their CFCC (Census Feature Class Code), using standard symbols for highways and residential streets from the menu. I don't want to label all the streets, just a few important ones, so on the Labels tab of the Layer Properties dialog, I selected "Method: Define classes of features and label each class differently", and constructed a query to select the names of streets I thought were familiar.
label dialog box
Also, adding a "halo" to the labels (under Text Symbol, click Symbol...) makes them more readable when they fall on top of other lines or dark patches.

3. Assembling the layout

The map will go on the top half of a portrait page, with the graph, legend, etc. below. (Another option would be to use a landscape page and cover most of it with the map, and have the graph as an inset in the southwest corner, but I decided that wouldn't look as clean.) I put in some guide lines and turned on Snap to Guides to make everything sit neatly. Some final touches to the layout:

Zoomed the map so the features fill the visible area, to get rid of extra white space so the map can be as large as possible on the layout.
Added definitions of "chaparral" and "southern rough" in a text box (I didn't know what they were before looking them up, so maybe the reader doesn't either). Another text box gives some general explanation of what to look for in the map.
Added a note on the date of the map: the vegetation polygons were delineated from aerial photographs in 1993. The front lines in the war against our enemy, eucalyptus, may have shifted. As of October 2005, the university continues to fight.

Here's the final layout!
my map layout