Mapping Quantities: Classified versus Unclassified Symbols
Broadly speaking, there are two approaches to symbolizing quantities on a map: classifying the values into “buckets” or not.
Choosing a map symbology approach is a matter of choosing which symbol will be applied to each feature and how those features’ attribute values will be “symbolized” (or represented) on the map.
Quick Definitions
For quantitative values, we first choose whether or not to classify the values included in our dataset.
Classified symbology systems involve grouping values into classes (or “buckets”). Each feature in a class is then assigned the same symbol. The span of values included in each class is called its range.
Unclassified symbology systems do not include grouping the values beforehand. Instead, the symbols vary to reflect each value included in the dataset. They might vary in size, color, or some other legible characteristic.
Example 1: Varying symbols by size (with points)
In the images below, we symbolize the same point data based on a generic quantitative attribute field. (The data used in these examples is available with the Standard Tutorial Data package available here.)
On the left, the values are classified into five classes. We read these classes from the legend, which indicates that the range for each class is roughly 100. The symbols vary in size from one class to the next, but features within the same class share the same-sized symbol. In other words, a location point with a value of 205 and another with a value of 290 are represented with identical symbols on the map.
On the right, the values are unclassified before symbols are applied to represent their values on the map. Because there are no classes, the legend does not describe ranges. Instead, the legend indicates the minimum and maximum values and the spread of symbol sizes. Each feature on the map is sized in proportion to its value relative to its spread. As a result, for example, a location point with a value of 205 will be proportionally smaller than another with a value of 290.
point features symbolized by size, with classified values
point features symbolized by size, with unclassified values
Example 2: Varying symbols by color (with polygons)
In the images below, we symbolize the polygon data, again, based on a generic quantitative attribute field.
On the left, the values are classified into five classes. We read these classes from the legend, which indicates that the range for each class is roughly 1000. Each class has a different color symbol applied on the map. As with the point-based classified symbology above, features with different values may be in the same class and thus are represented with the same color. For example, a feature with a value of 1002 is represented with the same color symbol as a feature with a value of 2000.
On the right, the values are unclassified before symbols are applied to represent their values on the map. Like we saw with the unclassified points, the legend does not describe ranges. Instead, it indicates the minimum and maximum values and the spectrum of symbol colors included on the map. Each polygon feature is assigned a color based on where its value sits along this spectrum. To continue our example, a polygon feature with a value of 1002 will be proportionally lighter than a polygon with a value of 2000.
polygon features symbolized by color, with classified values
polygon features symbolized by color, with unclassified values
When to Classify? (Some tips and rules of thumb)
There is no one-size-fits-all answer to whether and when classification is preferred. Here are some considerations to help:
It is difficult for most people to distinguish many shades of color along a long color ramp (or distinguish between small variations in symbol sizes, for that matter). Thus unclassified symbologies are difficult for most folks to read specific values in the middle of the full span of values. That said, they are great for showing variation as well as overall trends of highs and lows.
Classified systems are, by comparison, much easier to read when the total number of classes is kept to a reasonable number such that the difference between their symbols is legible. Classification can obscure micro-variations in a dataset’s distribution of values by grouping large ranges into a single class. If this is the case for your data, you will want to consider carefully how your classes are constructed. For example, classified systems are an excellent choice when the “break values” between classes indicate some contextually significant threshold.