3.3. Lesson: Classification

Labels are a good way to communicate information such as the names of individual places, but they can't be used for everything. For example, let us say that someone wants to know what each landuse area is used for. Using labels, you would get this:

This makes the map's labeling difficult to read and even overwhelming if there are numerous different landuse areas on the map.

The goal for this lesson: To learn how to classify vector data effectively.

3.3.1. Follow Along: Classifying Nominal Data

1. Open the Layer Properties dialog for the landuse layer

2. Go to the Symbology tab

3. Click on the dropdown that says Single Symbol and change it to Categorized:

   

4. In the new panel, change the Value to landuse and the Color ramp to Random colors

5. Click the button labeled Classify

   

6. Click OK

   You'll see something like this:

   

7. Click the arrow (or plus sign) next to landuse in the Layers panel, you'll see the categories explained:

   

   Now our landuse polygons are colored and are classified so that areas with the same land use are the same color.

8. If you wish to, you can change the symbol of each landuse area by double-clicking the relevant color block in the Layers panel or in the Layer Properties dialog:

   

Notice that there is one category that's empty:

This empty category is used to color any objects which do not have a landuse value defined or which have a NULL value. It can be useful to keep this empty category so that areas with a NULL value are still represented on the map. You may like to change the color to more obviously represent a blank or NULL value.

Remember to save your map now so that you don't lose all your hard-earned changes!

3.3.2. Try Yourself More Classification

Use the knowledge you gained above to classify the buildings layer. Set the categorisation against the building field and use the Spectral color ramp.

Catatan

Remember to zoom into an urban area to see the results.

3.3.3. Follow Along: Ratio Classification

There are four types of classification: nominal, ordinal, interval and ratio.

In nominal classification, the categories that objects are classified into are name-based; they have no order. For example: town names, district codes, etc. Symbols that are used for nominal data should not imply any order or magnitude.

• For points, we can use symbols of different shape.

• For polygons, we can use different types of hatching or different colours (avoid mixing light and dark colours).

• For lines, we can use different dash patterns, different colours (avoid mixing light and dark colours) and different symbols along the lines.

In ordinal classification, the categories are arranged in a certain order. For example, world cities are given a rank depending on their importance for world trade, travel, culture, etc. Symbols that are used for ordinal data should imply order, but not magnitude.

• For points, we can use symbols with light to dark colours.

• For polygons, we can use graduated colours (light to dark).

• For lines, we can use graduated colours (light to dark).

In interval classification, the numbers are on a scale with positive, negative and zero values. For example: height above/below sea level, temperature in degrees Celsius. Symbols that are used for interval data should imply order and magnitude.

• For points, we can use symbols with varying size (small to big).

• For polygons, we can use graduated colours (light to dark) or add diagrams of varying size.

• For lines, we can use thickness (thin to thick).

In ratio classification, the numbers are on a scale with only positive and zero values. For example: temperature above absolute zero (0 degrees Kelvin), distance from a point, the average amount of traffic on a given street per month, etc. Symbols that are used for ratio data should imply order and magnitude.

• For points, we can use symbols with varying size (small to big).

• For polygons, we can use graduated colours (light to dark) or add diagrams of varying size.

• For lines, we can use thickness (thin to thick).

In the example above, we used nominal classification to color each record in the landuse layer based on its landuse attribute. Now we will use ratio classification to classify the records by area.

We are going to reclassify the layer, so existing classes will be lost if not saved. To store the current classification:

1. Open the layer's properties dialog

2. Click the Save Style ... button in the Style drop-down menu.

3. Select Rename Current..., enter land usage and press OK.

   The categories and their symbols are now saved in the layer's properties.

4. Click now on the Add... entry of the Style drop-down menu and create a new style named ratio. This will store the new classification.

5. Close the Layer Properties dialog

We want to classify the landuse areas by size, but there is a problem: they don't have a size field, so we'll have to make one.

1. Open the Attributes Table for the landuse layer.

2. Enter edit mode by clicking the ^{Toggle editing} button

3. Add a new column of decimal type, called AREA, using the ^{New field} button:

   

4. Click OK

   The new field will be added (at the far right of the table; you may need to scroll horizontally to see it). However, at the moment it is not populated, it just has a lot of NULL values.

   To solve this problem, we will need to calculate the areas.

   1. Open the field calculator with the button.

      You will get this dialog:

      

   2. Check the Update existing fields

   3. Select AREA in the fields drop-down menu

      

   4. Under the Expression tab, expand the Geometry functions group in the list and find $area

   5. Double-click on it so that it appears in the Expression field

      

   6. Click OK

   7. Scroll to the AREA field in the attribute table and you will notice that it is populated with values (you may need to click the column header to refresh the data).

   Catatan

   These areas respect the project's area unit settings, so they may be in square meters or square degrees.

5. Press to save the edits and exit the edit mode with ^{Toggle editing}

6. Close the attribute table

Now that we have the data, let's use them to render the landuse layer.

1. Open the Layer properties dialog's Symbology tab for the landuse layer

2. Change the classification style from Categorized to Graduated

3. Change the Value to AREA

4. Under Color ramp, choose the option Create New Color Ramp...:

   

5. Choose Gradient (if it's not selected already) and click OK. You will see this:

   

   You'll be using this to denote area, with small areas as Color 1 and large areas as Color 2.

6. Choose appropriate colors

   In the example, the result looks like this:

   

7. Click OK

8. You can save the colour ramp by selecting Save Color Ramp... under the Color ramp tab. Choose an appropriate name for the colour ramp and click Save. You will now be able to select the same colour ramp easily under All Color Ramps.

9. Click Classify

   Now you will have something like this:

   

   Leave everything else as-is.

10. Click OK:

3.3.4. Try Yourself Refine the Classification

• Change the values of Mode and Classes until you get a classification that makes sense.

Check your results

3.3.5. Follow Along: Rule-based Classification

It's often useful to combine multiple criteria for a classification, but unfortunately normal classification only takes one attribute into account. That's where rule-based classification comes in handy.

In this lesson, we will represent the landuse layer in a way to easily identify Swellendam city from the other residential area, and from the other types of landuse (based on their area).

1. Open the Layer Properties dialog for the landuse layer

2. Switch to the Symbology tab

3. Switch the classification style to Rule-based

   QGIS will automatically show the rules that represent the current classification implemented for this layer. For example, after completing the exercise above, you may see something like this:

   

4. Click and drag to select all the rules

5. Use the ^{Remove selected rules} button to remove all of the existing rules

Let's now add our custom rules.

1. Click the ^{Add rule} button

2. The Edit rule dialog then appears

3. Enter Swellendam city as Label

4. Click the button next to the Filter text area to open the Expression String Builder

5. Enter the criterion "name" = 'Swellendam' and validate

   

6. Back to the Edit rule dalog, assign it a darker grey-blue color in order to indicate the town's importance in the region and remove the border

   

7. Press OK

8. Repeat the steps above to add the following rules:

   1. Other residential label with the criterion "landuse" = 'residential' AND "name" <> 'Swellendam' (or "landuse" = 'residential' AND "name" != 'Swellendam'). Choose a pale blue-grey Fill color

   2. Big non residential areas label with the criterion "landuse" <> 'residential' AND "AREA" >= 605000. Choose a mid-green color.

      

      These filters are exclusive, in that they exclude areas on the map (non-residential areas which are smaller than 605000 (square meters) are not included in any of the rules).

   3. We will catch the remaining features using a new rule labeled Small non residential areas. Instead of a filter expression, Check the Else. Give this category a suitable pale green color.

      

   Your rules should now look like this:

   

9. Apply this symbology

Your map will look something like this:

Now you have a map with Swellendam the most prominent residential area and other non-residential areas colored according to their size.

3.3.6. In Conclusion

Symbology allows us to represent the attributes of a layer in an easy-to-read way. It allows us as well as the map reader to understand the significance of features, using any relevant attributes that we choose. Depending on the problems you face, you'll apply different classification techniques to solve them.

3.3.7. What's Next?

Now we have a nice-looking map, but how are we going to get it out of QGIS and into a format we can print out, or make into an image or PDF? That's the topic of the next lesson!