Urban streets are shaped by the history, economy, and values of a community. Streets provide mobility to residents, but they are also gathering places for a range of activities: dining, shopping, arts, and recreation. Shared “check-ins” reveal how we are choosing to use streets now. By linking Sina Weibo check-in information to the road network, we learn the character of our streets, and how they vary by neighborhood and over time.

In these maps, the width of the roads represents the volume of check-ins on each street. By selecting a particular area, you can see the proportion of check-ins in different categories of establishments. By changing the day, you can see that both the check-in volume and check-in locations change over the course of a week.


Data Sources:
Sina Weibo Check-ins for September 23 – 28, 2013
Open Street Map

Methods:
The creation of a road network with proportional widths that smoothly change is not trivial. We created our graphics using the following process:

1. Create a fine grid that completely covers the road network (for the Shenzhen extent a 25m cell grid was appropriate).
2. Intersect the grid with the road network. This will create many additional vertices in the road network.
3. Convert all the vertices into points. The result is a road network that is made up of closely spaced points, rather than line segments.
4. Interpolate a surface from the point information of interest. In this case we used an Inverse Distance Weighting method, and the Sina Weibo Check-in information for each day in Shenzhen. This created a smoothly changing surface which was appropriate for the intended design.
5. Associate each point in the road network with the value in the interpolated surface. (There are many ways to do this – we used ArcGIS’s Extract Values to Points tool.)
6. Use graduated symbols for visualization.

The result is a road network that has a smoothly changing width. However, because the road network is still made up of individual points, an interactive visualization requires some additional steps. We were interested in being able to select areas roughly 2km by 2km in size. The steps required were as follows:

1. Export the map and grid with 2km cells to a graphics editor.
2. Using a path editing tool, unite all the points into a single shape.
3. Divide the shape (representing the road network) using the grid.
4. Export the resulting file to an svg format, and label each portion of the road network with the id of the original grid.
5. Use the 2km grid to summarize the check-in information by category for each grid cell.
6. Using D3 and JavaScript, on hovering over a cell, the path color is changed, and the id of the shape is used to look up the appropriate summary values. These are in turn used to visualize the nature and volume of activity in a given area through a radar chart.