Andy Mitchell

The ESRI Guide to GIS Analysis: Vol. 1: Geographic Patterns and Relationships

Redlands, CA: ESRI Press, 1999. ISBN 1-879102-06-4

CHAPTER 1: WHAT IS GIS ANALYSIS?

1.What are the three types of features in a GIS? P/ 12

Point data: individual locations, such as businesses

Line data, such as rivers or roads

Polygon data, such as land parcels or political jurisdictions

2. Explain vector vs. raster representation of geographic information. P. 14

Vector: the most common type for mapping. Data are stored as information on x.y coordinates, line lengths, and angles, so the map may be recreated on any scale by redrawing according to these directions.

Raster: used in orthophotography and where there is continuous spatial data.. Data are stored as a matrix of cells (often very fine ones). Increasing map size causes graininess as cells are drawn further apart.

3. Explain “projections” and “coordinate systems”. P. 15.

Projections: There are several ways to draw the map on a spherical ball onto a flat surface.There is no perfect way to do this and all projections involve distortions. There is a tradeoff among accurate representation of distance, area, and shape.

Coordinate system: For any given projection, a coordinate system provides the definition of the origin and also the units from the origin to locate any point.

4. What are attributes, and what are the types of attribute values? Pp. 16-17

Attributes are variables, often represented as different layers on a map.

Categories are nominal values such as types of crime.

Ranks are ordinal data such as poor to excellent ratings for “scenic value”

Counts and amounts are interval data, such a population sizes.

Ratios are percentages, such as average no. of people per household in each census tract.

5. What are data tables and what are some common operations in working with them? Pp. 18-19.

Data tables are literally the tables containing attribute data. GIS software will have a table view as well as a map view.

Selecting is an operation which finds all features which have certain attributes.

Calculating is an operation used to create a new variable (ex., creating a ratio variable from two count variables).

Summarizing is the operation which gives descriptive statistics like means, medians, miximumx, and minimums.

CHAPTER 2: MAPPING WHERE THINGS ARE

1. What is Mitchell’s rule of thumb about the number of categories that can be represented on a map and still be readily understood? Pp. 30-31.

About 6 - 7 values per variable. He suggests grouping if you have more categories.

2. Which is easier to distinguish – colors or symbols to represent points? P. 32.

Colors.

3. Differentiate clustered, uniform, and random distribution of points. P. 35.

Clustered: features appear in groups, near each other.

Random: features are equally likely to appear anywhere on the map.

Uniform: features tend not to be near each other, but are also not likely to be found in some areas of the map.

CHAPTER 3. MAPPING THE MOST AND LEAST

1. What methods are used to show most and least on a map, and what are the problems of each? What tips for good usage? Pp. 56-63

Graduated symbols: ex., small to large dots showing population size. Problem if points overlap. Tips: too small gradations are hard to discern; too big overlap too much. Note not just dots, but also various sizes of lines for line data.

Graduated colors: light to dark blue, to show ocean depth. Colors may not intuitively relate to magnitude. Tips: pink to dark red is more contrasty than light green to dark green; if you have positive and negative values, use two color schemes.

Contours: the closer contour lines are together, the steeper the grade at that elevation. Can be hard to read and requires interpolating values between observed points.

3-D Perspective: like contour but more visual; front areas may hide rear. Also, the z-value (perspective height) cannot be too small or too great or contrasts will be obscured. You may also have a light source with origin and angle, and this can also affect contrasts.

Charts. Superimposing bar or pie charts on areas can show patterns too. Can be too cluttery. Bar charts are more for amounts, pie charts for percentages. You can also use pies of graduated sizes to represent a variable.

2, What is classification for and explain the standard classification schemes. Pp. 48-51.

Classes are groupings of values, such as income ranges. Classification is assigning cases to these classes by a classification scheme below:

Natural breaks: nominal variables (ex., race) have natural breaks, and some “natural” breaks are human-defined, such as above and below the official poverty level. Use natural breaks if they exist, as a rule.

Quantiles: rank the cases from low to high and set cutting points which divide the cases into equal groupings (ex., 20% in each group, for quintiles)

Equal intervals: divide the cases by equal ranges of the unit of measurement, such as $0 - $10,000; $10,000 - $20,000; $20,000 - $30,000; etc. Easier to interpret than quantiles but some ranges may have few or even no cases.

Standard deviations: Assume a normal distribution and compute the standard deviation of the variable. About 95% of cases will be within 2 sd’s of the mean, 99% within three sd’s. Used to identify outliers.

3. How many classes are best to display data? P. 54.

This has to be approached by trial and error. Too few classes will average apples and oranges. Too many classes will lead to hard-to-discern complexity. The optimal number of classes will make the pattern stand out on the map. See example on p. 54.

CHAPTER 4. MAPPING DENSITY

1. What are some examples of mapping density? Pp. 70 ff.

Mapping density is mapping concentrations of something: concentrations of people (population), fish, businesses/square mile, logging road density

2. What are some examples of policy decisions using density maps? P. 70-71

- Assigning police to patrol areas based in crime incident density

- Locating government service agency offices based on population density

- Locating fish monitoring devices in lake areas of high fish density

3. What is a dot density map, and what is a major drawback of this type of map? Likewise, what is a density surface map and what is a major drawback of density surface maps? P. 70

A dot density map shows location and can show density, but it has the problem of overlaying dots at the same location, obscuring the true density.

A density surface map is a type of contour map , with shaded contours linking points of approximately equal density. Using a color gradient to display increasing density (ex., darker reds), it is easier to see the true density of any area. See map p. 70, upper left. However, like all contour maps, it may falsely suggest equal density throughout any given contour plane.

4. Is dot density mapping the same as mapping locations? P. 73

No, Location mapping is a special case of dot density mapping, where there is one dot per location (ex., per business). But it may be that one dot represents FIVE businesses in a zip code, for instance. In such cases the dot may be located at random within the zip code and may obscure actual locations (ex., locations along a highway). See maps in lower left of p. 73.

5. Can a choropleth (shaded area) map be a density map? Compare such maps with density surface maps in terms of pros and cons for measuring density. P. 72

A choropleth map is a density map if the color gradient used to shade the areas (ex., census tracts) based on some density value (ex., population per square mile). See map in lower left corner of p. 72.

Both chorpleth and density surface maps measure density, and both have the drawback of suggesting equal density thoughout a given shaded area. When the choropleth areas (ex., census tracts) are related to the variable whose density is being measured, they may be preferable. When there is no particular relation, the unconstrained definition of density areas by a density surface map may be preferable. Of course, one can have a density surface map AND overlay boundaries like census tracts or school districts.

6. What is Mitchell’s point about mapping features vs. mapping feature values? P. 72

Make sure you are mapping the variables you really want. In the example, is it number of businesses or is it number of employees?

7. What does “calculating density on the fly” mean? P. 75

It means ArcView, ArcInfo and some other GIS’s do not require an actual density field to be a column in the database. Instead one specifies the count field (ex., # of businesses) and the area field (ex., zip code feature).

8. What three mistakes in dot mapping can obscure the pattern one is seeking?pp. 76-77.

1. Too few features per dot (ex., 2 businesses per dot rather than 5 businesses per dot) leads to too many dots, overlapping, and obscuring the pattern.

2. Dot size too large, leading to overlapping and obscuring.

3. Overlaid boundaries too fine (census blocs instead of census tracts), causing boundary lines to overlay dots and obscure patterns.

9. What causes the difference between the two maps at the top of p. 77?

The one on the left maps 1 dot = 500 households by county and has county boundaries.

The one on the right maps the same thing BY ZIP CODE and still has county boundaries.

This locates the dots more accurately WITHIN counties (but still at random within any given zip code).

10. How exactly are density contour lines drawn by ArcView?

The map is treated as a raster surface, divided by a grid. The number of features (ex. businesses) within a given radius of each cell center. Each grid cell is colored in according to its summed count divided by area size.

The smaller the grid cell size, the smoother the contour lines but the longer to process and the larger the file size to store. (Cell size is defined by the length of one of the sides of its square). Generally, one picks a cell size which is 10 to 100 cells per density unit (ex., per square mile)

The larger the search radius, the more generalized the patterns. Too large or too small a radius will make patterns hard to see. In counting features within the radius, one may have a simple count or may weight to give more emphasis if as the feature is closer to the center of the grid cell.

11. When using graduated colors to show different density values, what are the four methods of graduation?

1. Natural breaks

2. Quantiles: equal number of cases in each of the categories: can force more cases into the highest category, obscuring the true centers of density.

3. Equal interval: The ranges represented by each category are equal (ex., 0-500, 501-1000, etc) even if this means some categories are unequal or even empty in count.

4. Standard deviation: plus or minus 1, 2, 3 standard deviations: highlights the extremes.

12. When using graduation, how many categories should you ask for? P. 83

Several is good. Over 15 is really hard to interpret. Under 4 may well not show patterns well. By the way, usually darker=more dense.

13. Is a contour map the same as a density surface map? P. 84.

No. A contour map shows actual contour lines. A density surface maps shows the same information as shaded contours. The contour lines could be narrower or more widely spaced than the shaded contours of the density surface map. See the map at the bottom of p. 84.

14. How can it be that there are no features where the highest density is shown on a density map? P. 85

Density is calculated based on counts within a radius of each grid cell. The grid cell itself might have zero count. This would happen, for instance, for grid cells located between other grid cells with high counts and within the search radius of grid cells with zero count.

Chapter 5. FINDING WHAT’S INSIDE

1. What are some examples of mapping to “find what’s inside”? P. 89 ff.

- Mapping interior features: business points, road lines, school district polygons

- New features you draw, like new sales areas dividing an existing polygon

- Merging existing polygons to make a large one

- Service areas: features w/in a radius, driving distance, or driving time

- Buffer zones, like around a school for purposes of speed limits

Note: You can include only polygons wholly inside an area (ex., inside a flood plain); areas which are wholly or partly inside; or polygons which are wholly inside plus the inside portions of the partly inside polygons).

2. Differentiate list, count, and summary output. Pp. 92-93.

- List: table of all parcels, with ones inside a flood plain highlighted in yellow (as in ArcView)

- Return the count of the number of parcels inside a flood plain

- Sum the tax assessment value of all parcels inside a flood plain; sum the area of parcels by each of several types of land use within a flood plain.

3. What are the three ways of “finding what’s inside”? What question does each method address? Pp. 96-98.

- 1. Overlay a polygon (ex., flood plain) on a feature map (ex., tax parcels). En Question: Is a given feature inside the polygon?

- 2. As (1) but also let the GIS use the polygon overly to select automatically the features included within the polygon. Question: Return a list of features inside the polygon, or return a summary of features inside the polygon.

- 3. As (2) but then also create a new third layer combining the features of the first two layers. Question: Return a list or summary for multiple polygons in the overlay.

Note: When overlaying, you can put the polygon boundary layer on top of or underneath the feature layer, depending on which you want to emphasize. You can also control the transluscency of the shading of polygons, and the thickness of border lines to get different effects.

4. What are three ways of displaying an overlaid polygon? P. 100.

- Use a dark line border

- Fill it with hatching or a transluscent shade.

- Place a screen over all areas outside the polygon, with the polygon itself transparent (thus showing the feature layer underneath). See p. 100, lower right.

5. What is the difference between counts , frequencies, and summaries? Pp. 102-103

- counts are the number of features regardless of value (ex., businesses, regardless of type) within an area

- frequencies are the number of features of a given value or of each value (ex., retail businesses) within an area

- summaries are sums, averages, medians, or standard deviations of numeric attributes (ex., number of employeess at businesses in an area).

6. What happens when you overlay a polygon laye on top of a feature layer, select features within a polygon (within a flood plain), and create a new third layer? Pp. 106-107

The GIS uses the x,y coordinates of the features to match them to the polygon(s) (ex., census trracts) and create a new merged table containing both the feature information and the polygon information. Rows will be the features, but there will now be columns for census tract id and census tract population. You can now sum features by tract and join the summary table to the tract table, in which rows are tracts. You can then divide the sum of features by population to create a new features per population variable.

7. What happens when you overlay a polygon layer on another polygon area? Pp. 108-109.

The boundaries of the first layer are used to subdivide the boundaries of the second area, giving you a merged layer with a lot of polygons.

Note: when you do this, you may get slivers, which are small areas where borders are offset. It is best to merge these slivers with adjacent larger polygons. This happens only with vector overlays, not raster overlays.

8. How can bar or pie charts be integrated in mapping? Pp. 11-12

Pie or bar charts can show the frequency distribution of values of a feature or percentages (ex., percent of area by land use type). Charts can be placed on top of their respective polygons.

CHAPTER 6: FINDING WHAT’S NEARBY

1. What are some examples of proximity maps (mapping what’s nearby)? Pp. 116 ff.

- Service or catchment areas (p. 116): concentric zones (by radius, road distance, or time distance) around some feature

- Color-coded points coded by which feature they are nearest (p. 117, bottom left)

- Spider diagram: same as above but with lines to the feature (p. 117 right)

2. What is “cost” in the context of proximity mapping?

Cost is a value associated with distance. Cost can be used as the definition of nearness. Examples would be operating cost per mile, travel time per mile, or simply miles. Distance by cost will be different from straight-line (radial) distance.

Cost is usually in the context of cost over a network, where each line segment has an associate time or distance cost.

However, you can also have cost over a surface (without a network). An example would be travel cost in an open hiking area, based on slope (steeper slope = more cost).

3. What are some uses of straight-line distance? P.122.

- Create a buffer around something.

- Create a distance surface around something (a concentric series of buffers, with a color gradient).

- Calculate feature to feature distance

- Select features within a given distance of another feature

4. What is Euclidean distance and how is it calculated?

It is the straight-line distance from one point to another. It is calculated as the square root of the sum of the squares of the sides (over and up) of the triangle connecting the points along the hypotenuse. (Pythagorean theorem).

5. How would you find features within 1,000 feet of two other features, X and Y? p. 127.

Create a 1,000 foot buffer around X and select features within the buffer

Create a variable and set all selected features to 1

Repeat for Y

Select features with both variables set to 1.

6. If you want to have a map’s table have a column for distances to a feature, do you have to calculate the distance for each point of interest? p. 129

No. The GIS’s distance calculation for all features in relation to a given feature(s) will insert two new columns in the table: one for the id# of the nearest given feature(s) and one for the distance from the current point to that feature.

7. What is a distance surface? p. 132.

It is a raster layer of concentric distance intervals from a point, line, or polygon. You can set the distance ranges and also the maximum distance for which intervals will be calculated.

8. What is a network? pp. 135 ff.

CHAPTER 7: MAPPING CHANGE

1. Explain some different types of change-focused maps Pp. 151-2.

* Change of location: one map with a series of connected points, with each point reflecting a change in location. Ex: bird migratory path from north to south US

* Change in character: a before and after map with one color per character value.. Ex. Land cover in 1914 and 1988

* Change in magnitude: choropleth map keyed to a change variable. Ex. Percent change in population by county.

* Change in location and character: one map, connected points, point symbol changes to reflect changing character value. Ex. Path of precipitation with different symbols for rain, sleet, snow.

* Change in location and magnituede: one map, connected points, but point size changes by magnitude. Ex: hurricane track with dot size based on wind speed.

2. If there are many change points (ex., many points in time), what determines how many intervals you have? Pp. 155-7.

* Wide enough intervals that anomalies and outliers get averaged in

* Use the fewest divisions necessary to show change (fewer is easier to visualize)

* Include start and end points, of course (total time period tracked should be consistent with the change process being studied - ex, cannot study climate change over a short period).

3. How does mapping change relate to use of color gradients in maps?

* Normally use a gradient with darker=more change

* But if you have a change-nochange situation, let no change be a contrasting color like white

* If there is both positive and negative change, use a two-color contrasting gradient with no change at the color midpoint.

4. What is a “tracking map”? P. 161

A tracking map shows the position of something at each of several points in time. For instance, it might be a choropleth map where the gradient colors show the progressive spread of some area (ex. A forest fire area on days 1, 2, 3, etc.). Could be done with lines or points, not just polygons as shown on p. 161, bottom.

5. How is movement related to position in a tracking map? P. 165

Since you locate the successive polygons, line segments, or points at equal time intervals, the further apart they are, the more rapid the movement. For instance, in a hurricane tracking map, long line segments indicate rapid movement in those time periods, short segments mean little movement in those other time segments.

6. How can orthophotography be integrated with tracking maps, and why?

This is simply superimposing successive polygon boundaries, line segments, or points on an aerial photograph of the are area, to show visually what the original area looked like (ex., as before a fire swept through successive polygons).

7. How can you map just the area that has been changed using the raster method? P. 170

The raster method divides an area into grid cells. Overlay one may, say 2000, on another map of the same area, say 1990. A GIS can match cells and automatically select the ones which differ...this selection is just the changed area.

8. How does the vector method differ? P. 170-171.

The two layers (2000 and 1990) are overlaid and a new layer is created with polygons for each intersection of the overlaying boundaries with the base layer boundaries. For each new polygon in the new layer, select it if, say landusetype2000 differs from landusetype1990.

9. What is the maximum number of category values Mitchell reommends in a single map? P. 172.

Six

10. What is a density surface change map? P. 173.

Recall a density map is a contour map where the contours are areas of like density, such as crime incident density. If you have density maps for two time periods, you can subtract the incidents in the second table from those in the first, and in the resulting difference table, use these positive and negative values to create a new density surface map. Use contrasting gradients (ex., reds and blues) to represent positive and negative density contours, reflecting increase or decreas in, say, crime incidents. This is illustrated on p. 173.