The main purpose of lesson 8 was to learn different Symbology options and classification schemes available in ArcGIS and apply them during map creation process. For the purpose of this exercise we used City of Philadelphia crime data on Census tract level. After joining this data to the tract boundary file using 6 digit FIPS code we were able to create series of thematic maps using different symbolization techniques. All maps contain 2 background layers: hydrology (rivers and streams) and road network (Interstates, US Highways, PA highways and residential streets) within Philadelphia boundaries.
Figure1. Graduate colors scheme.
Figure 1 shows percent of burglaries in particular census tract between 1998 and 2003 out of total number of burglaries in Philadelphia area. Since we want to use graduated colors scheme, mapping row numbers could have been misleading for user. Normalization by area is not good either since this would have given us number of burglaries per sq. foot, which is not useful. We used equal interval classification method with 5 classes, which means that all values were divided into 5 groups and each group had an equal value range. This method is usually good when values are equally distributed and there is better chance for relatively big number of values to fall into each particular category. In this case user would get better idea what happens in each region (as opposite to quantile method where each category includes equal number of cases. This method is better for skewed data distribution). In order to visualize our variable we used green-blue color ramp, which we were able to create using ArcGIS Style Manager. This particular ramp is good for people with vision impairments since it is using colors from separated regions in colorblind diffusion diagram.
Figure2. Diverging scheme.
Figure 2 shows the same variable (total number of burglaries) using diverging scheme. The main idea behind this method is breaking sequence into 2 schemas using similar colors for each of them and some neutral color in the middle. Another color ramp was created for this purpose. It uses dark-light blue color continuum from one side and dark-light brown color continuum from the other side with yellow in the middle. The best classification method that works with this kind of scheme is standard deviation. All values that fall into -1: 1 range can be considered as a middle values and therefore can be represented with the neutral color. All values smaller than -1 but bigger than -2 will be represented with light blue, while values smaller than -2 will be represented with dark blue. The same idea works for value bigger than 1, the only difference is that in this case we use different hues of brown color.
Even though both maps represent similar information, visual impression of the user would be somewhat different. An abrupt change of color in the second map gives better emphasis on the differences between different parts of the city, while smooth color transition in the first map doesn't create this clear impression. From the other side in the second map there is a big number of cases that fall into the middle category and represented by neutral color. In this case user gets impression about crime situation in general, but would probably miss some important information about local trends.
Figure3. Graduate symbol.
Figure 3 shows burglary data represented in a form of graduate symbol. I have decided to use circles as my symbol even though it is not always the best choice for user’s estimation of the particular phenomena. By my opinion in this particular case it gives the right message about the density of burglaries in Philadelphia. I have decided to use blue circle on light green background for better contrast and because this color pattern is more convenient for people with visual impairments. After experimenting with different settings I decided to make 5 classes with circle size range 3 to 15. In this case user can definitely see patterns on the map and there are not many overlaps that obscure symbols and create feeling of busy map.
Figure4. Proportional symbol.
Proportional Symbols (Figure 4) work by the same principle, the only difference from graduated symbols is absence of classes. ArcGIS arbitrary chooses size of symbols, the only thing we can choose is number of symbols to display in the legend (I chose 4). Flannery scaling factor didn’t really help in this case, it just made map busy and hard for interpretation.
Figure5. Pictographic symbology.
Pictographic Symbology (Figure 5) is another way to apply proportional symbols, the only difference is that we can choose meaningful symbol for our data. In our case we could choose burglary symbol from Crime Analysis category and show it as a proportional symbol. The only symbol modification that made sense in this case was removing yellow background; all other modifications didn’t really improve map appearance.
Figure6. Dot density map.
Figure 6 represents map that is conceptually different from all others. If the first 5 maps represent single variable (total number of burglaries in 6 years), the sixths map represents each year in separate. The symbolization schema used for this purpose is Dot Density, where each year has its own color and each dot represents 20 cases. In a case like that it is extremely important to choose the right color pattern and proper background color. After experimenting with these settings I decided to make blue background and light-dark red color scheme for the data. This choice is good for people with visual impairments (blue-red colors) from one side and makes emphasis on the most recent data (2003 has the darkest hue). Another important choice was about dot size. Too small dots would have made map very sparse while too big dots would have made it too busy. In both cases map would have missed some important information and would have been hard for interpretation. For this particular map I chose dot density=2, which, by my opinion, creates reasonable representation pattern. In this case dots overlap a little in the most dense part of the map which is downtown Philadelphia. The map pattern in this map is similar to patterns in 3 previous maps. We can see 2 main crime clusters - downtown and western Philadelphia, while level of crime activity declines as we move out of central city. This pattern reflects demographic situation in the city of Philadelphia where there is a big concentration of poor population in certain areas.
One of the most important things that we learned in this lesson is that there are no ‘good’ or ‘bad’ classification schemes. Each particular scheme may represent data either in a good-looking or in hard for interpretation way. A decision which classification method to use must come from the cartographer after examining all necessary data. Even though there is couple of general rules which method must be used in which situation, there is no universal recipe for each particular situation. Cartographer must take into account type of presented information, map area, intended audience and other relevant factors before making decision about method, and then spend some more time experimenting with different settings to make a final map attractive for users.