GeoData Visualization vs. Analysis: A Bit-o-Fun with 3D

When Laurie was working on her blog post covering the geopolitics of oil, she asked Raj and I to help out with creating some maps. She had some nice data showing the known oil and natural gas reserves around the globe. Specifically, she wanted some 3D maps to really show the relative amounts of oil and natural gas in different geographies.

Creating the map presented us a classic cartographic decision – should we do a data visualization or a data analysis? While this is a very distinct difference, I think it gets largely lost by most GeoWeb users, and hopefully this little example will help illustrate the importance of the difference.

For Laurie’s maps she wanted to show the relative amounts of known natural gas and crude oil around the globe. The data set she had collected from the USGS provided polygons where petroleum and natural gas were located. The most straight forward way to map the data was to create a thematic map shading the polygons based on the amount of oil or natural gas contained in each. In cartographic circles this is called a choropleth map. Below is an example of a choropleth map in Maker! with the oil data from Laurie’s post:

Known Petroleum Reserves

You could also use the centroid of each polygon to make a proportional symbol map – where the size of the symbol is representative of the amount of oil or natural gas in the location. Here is an example of using the natural gas data set but rendered as proportional symbols:

Known Natural Gas Reserves

Proportional or graduated symbols are a great choice when you have small polygons with high values that might get lost in a choropleth map with many large polygons. For instance in the oil reserve map Venezuela has a very high oil reserve, but since the polygon is quite small it is easy not to notice it on the map.

Another technique for getting small places noticed is using extruded polygons, which are 3D. Bjorn Sandvik has done a great job promoting these techniques in Google Earth with his Thematic Mapping Engine. Below is an example of the oil reserve data as extruded polygons on a 3D globe:

3D globes have some shortcomings like accuracy and not being able to see the whole globe at once, but offer a great dynamic way to interact with data. For the geopolitics of oil, we decided to go with creating a thematic map using extruded polygons but on a 2D projection of the earth. That way we could see all the continents and curvature of the earth accuracy would be diminished. While not as cool as Bjorn’s virtual globe maps they got the job done:

A second option we looked at was doing a spatial analysis of the oil data. Instead of visualizing the data values for the polygons, we did an analysis of the spatial distribution of the data. In this case we thought it would be interesting to analyze the spatial density of the oil reserve locations. To do so we needed to convert the polygons to points based on the centroid of each polygon. Then we could run a kernel density analysis of the data. This sounds fancy, but it’s really just placing a grid over the data and tabulating counts for each cell with a bit of fancy math for smoothing to create a continuous surface. While this is not terribly complex, it is much different than just visualizing the data. The results of the density analysis for the oil data can be seen in the map below:

We ended up not using the density analysis because it was not really accurate since the source data had been polygons and not points. If the source data had been well locations weighted by oil production it would have been spot on.

The confusion between geodata visualization and analysis was one of the reasons we deprecated GeoIQ (sometimes called heatmaps for Google maps). Spatial density analysis has popped in lots of applications since the original work but I think it is still commonly misunderstood by users. We found many users that thought the hot spot was the highest single point value on the map, which often it was not. Few realized the hot spot was the cluster of points that were closest together with the highest aggregate value. I think with the right work flow or user interface it can be a great tool, but we are not quite there yet. The difference between the two only becomes apparent when users can have access to geodata visualization tools (i.e. thematic maps) and geodata analysis tools (i.e. density maps).