Module 4 - Color Concepts and Choropleth Mapping

While easily overlooked, color choices and data classification schemes are instrumental in constructing accurate and appropriate maps. This laboratory assignment highlighted how color selections impact a map's overall legibility and message to the viewer. Adjusting colors to meet the needs of an array of audiences, such as those who are visually impaired or even color blind is also very important to take into account when creating a map. Throughout this post I will highlight the different color ramps and maps that I created during this fourth lab in Communications in GIS.

I began the assignment by manually creating two color ramps - linear and adjusted. These ramps were constructed using various RGB values within the ArcGIS Pro 'Color Properties' option. I used different formulas and patterns to determine what RGB values to use, such as using a 33% increase in the interval between the darkest and next darkest and 33% less interval between the lightest and next lightest for the adjusted progression color ramp. Then, I selected the most closely resembling ramp within ColorBrewer, and assessed the differences among these ramps, noting the change in hues as well as RGB values.


Linear Color Ramp

        
Adjusted Progression Color Ramp
  
Colorbrewer Value Color Ramp













The ColorBrewer ramp selected was a multi-hue, likely resulting in the varying intervals between red, green, and blue values. The linear and adjusted progression ramps were essentially the same as one another, with the adjusted ramp having more varying shades between its darker values. Whereas the ColorBrewer ramp went from green to white instead of blue-green to a lighter shade of blue-green. Thus, the green values for the ColorBrewer ramp were much higher (proportionally to other values in the ramp) than the other ramp types, as this was the predominant color in the scheme. However, this quickly changes as the ramp progresses to the off-white final color value. One other noticeable difference is the contrast is greater in the ColorBrewer ramp. This is likely because it is showing transitions between two different colors and thus has to make larger changes between each interval, as opposed to the manually created ramps, where I adjusted the values based on the formulas and prior knowledge in the lab guide.

The next portion of the assignment involved assigning appropriate colors to land cover/use classes. Below is the map and explanation of my color selections.

Land Use / Cover Map with Appropriate Legend Color Values

I chose to implement the point that was made in the “Designing Better Maps” text that states that if logical relationships are among different categories, using those relationships with related colors in the color scheme improves the map (Brewer, 2016). I find that maps that use it are typically more aesthetically pleasing and visually accurate. Thus, I chose to make bodies of water have a blue color, vegetation with varying hues of green, and tree covered areas brown (to represent the bark). Otherwise, I chose to use typical color choices for residential/industrialized areas, such as red and purple. I manually adjusted the color properties to each be different enough from the other categories’ color values, while still retaining a true-to-life visual appearance. Many of the color selections I implemented were found in the “Using Color in Maps” guide, which suggests using green for vegetation, red for roads/cities, blue for water, etc. (Strode).

Understanding classification schemes is crucial when deciding how to best represent numerical data within ArcGIS Pro. The next part of the lab assignment required me to use different classification methods to compare the differences among them in mapping the Hispanic population in Texas. This involved normalizing the data to act as percentages rather than totals, as well as adjusting the method appropriately within the symbology pane.

Data Classification Comparison Layout

The classification methods assessed were:
  • Natural Breaks - This method categorizes the data by emphasizing gaps in datasets and grouping data in clusters with similar values. This is known, because the breaks are mostly located where there are already breaks in the raw data, and the histogram shows the data is slightly skewed.
  • Equal Interval - This classification method divides the data into groups of similar ranges of values. For example, the class breaks in my data all fell between 18-20% intervals, per class. This means that the count is disregarded in this scenario and it is more based on the data itself and grouping it accordingly. This provided a very right-skewed histogram.
  • Quantile - Last but not least, the quantile method categorizes data into groups that contain roughly the same counts per class. All classes in my data (using this method) contained 50-52 counties. The histogram is very even and not skewed for this method, proving my point further, as each class has roughly the same amount of values.
When looking at demographics and population statistics for such a large and diverse state, I think it is important to employ the most appropriate classification method to showcase the data. In this case, I would argue that the quantile method is the best. This is because the quantile method evenly distributed the classes, resulting in each class having the same number of counties on the map. This produced a very visually pleasing, accurate, and informative map. It shows the trend of hispanic populations having much larger ratios towards the border, where other methods only slightly showed this trend. Also, the quantile method produced an even, un-skewed histogram, which indicates that the data was evenly spread.

To complete my assignment, I constructed a map of the state of Colorado and showed population changes in each of its counties. I used a diverging color ramp and assigned RGB values based on a similar ramp I located within ColorBrewer. After employing the most appropriate classification method and adjusting the color ramp as necessary, the final map was complete. Below I have included my map as well as explanations of the choices behind it.
Showcasing Diverging Data using Population Change

I chose to employ the NAD 1983 UTM Zone 13N Coordinate System. This is because the state of Colorado contains 3 different State plane systems, thus I would be unable to map the entire state using only one system. The next most accurate coordinate system available for mapping the entire state was the UTM system, in which Colorado falls in zone 13. 

The formula used to normalize the raw data is as follows:
        (New Value-Old Value)/Old Value * 100
               With attribute table values -> (POP2014-POP2010)/POP2010 * 100

5 classes were used as I believed this was an adequate number to show the population changes without being too concise. I had initially chosen to utilize the quantile classification method, so there would be an even number of counties per each category. However, I was not satisfied with how this method assigned population growths to the light red color, which I believed should only be showing population decrease. This is because Colorado was quickly growing during the years the data covered, with most counties in the state experiencing growth. Thus, I manually adjusted the upper value limits for most categories to better showcase the data and have the color representing each county be more meaningful. 

Red was selected to represent significant population loss, light red for less significant population loss, white for little to no population change, light blue for less significant population gain, and blue for significant population gain. These colors are often used in population change maps and are more universally understood than other color combinations. I appreciate their visual contrast, which results in the viewer not becoming confused between colors on both the map and legend.