data-defined attributes in QGIS' point cluster symbols

Question

I'm wondering how data-defined attributes work for the cluster symbol when using the point cluster layer styling in QGIS 3.*. How can you use the values of the points within the cluster for data-defined attributes?

The image below is the result of using the winddirection "D" attribute as the rotation parameter for the cluster symbol. It seems reasonable, however I don't know if this is the average, the sum, the mean or even the first value of the clustered points.

I would like to use the mean value of the clustered points, because I would like to get the prevailing winddirection. I tried using the expression mean( "D" ) but that results in all the points pointing the same direction. Documentation is a bit sparse and the only variables in the expression builder related to clustering that I can find are cluster_color and cluster_size which are both of no use in this case.

Here is the dataset if someone wants to try.

Answer 1

The cluster renderer is a relatively new feature of QGIS. As you've discovered, it only has two variables, cluster_color and cluster_size. It lacks variables like @currentcluster or similar, that would allow you to build an expression that refers to all of the points in a single cluster.

There is a recently-opened feature request for exactly this feature. I recommend watching that request for further developments.

---

In the meantime, your only option would be to generate your own point clusters.

One way is to generate a hexagonal grid, and symbolize it based on the average wind speed of the wind point it intersects.

Note: Unlike the cluster renderer, this method does not change cluster size based on zoom level. You would have to create a separate grid for every zoom level, and use scale-based visibility on each grid layer.

For example, my wind points layer is called 'wind' and the wind speed field is called 'speed'. Using the functions from the refFunctions plugin, I used this expression to calculate the average wind speed within each grid cell:

intersecting_geom_sum('wind','speed' )/ intersecting_geom_count( 'wind')

Add average wind speed and average wind direction as fields to the grid layer.

In case not every grid cell has a wind speed point in it, use this expression to define layer visibility:

"avg_wind_speed" > 0 and  "avg_wind_speed" is not null

Set up centroid fill symbology with two levels of marker:

• One level is simple points.
• The other level is a vector field. Use the fields "avg_wind_speed" and "avg_wind_direction" to control length and angle of the vector field.

In case not every grid cell has a wind speed point in it, use this expression to define layer visibility:

"avg_wind_speed" > 0 and  "avg_wind_speed" is not null

This gives a slightly random appearance to the point layout, although you can still see that the points are vertically aligned. If that's a problem, you could add a random data-defined offset to the centroid fill. Eg, define the x-coordinate as $x + rand(0,10) and likewise for the y-coordinate. Choose values for the rand() function that are appropriate to your layer units and scale.