off the map

I have no doubt that Google employs some very clever people, but I am equally certain that none of its employees is a cartographer, having arrived at this conclusion because the company’s maps are the cartographic equivalent of a dog’s breakfast.

I’m not a cartographer either, but I do claim some expertise in the production and use of maps. When I sat GCE O-level geography, more than 50 years ago, part of the exam was a small section of an Ordnance Survey map on which we were asked questions, such as ‘Give five reasons why village X is located where it is’. This kind of question requires not only an ability to read a map but also an ability to interpret it, because a map with the richness of detail that is typical of Ordinance Survey products at all scales offers a vast amount of usable information.

Then, when I went to university, I spent a year studying cartography, which struck me as more interesting than the usual subsidiary subjects that are offered to science students. During this time, I learned how to use most standard surveying instruments, including a theodolite, a level and staff, and a plane table, to produce maps.

So what do I think are the principal shortcomings of Google’s maps? The way that roads are represented is hopelessly misleading: they make no distinction between dual and single carriageways, between single-track and two-track roads, and between metalled and unmetalled roads, although expressways are identified separately. There is also no indication of access restrictions, such as those that apply on many of the roads built by Hong Kong’s Drainage Services Department, which are gated to block access by motor vehicles and thus allow only pedestrians and cyclists to enter. Military barracks are not marked either, and a Google map that should include such a feature has roads inside the barracks that appear to be contiguous with roads outside.

The existence of footpaths is not recorded, even though there is an extensive network of concrete paths throughout the New Territories. And there is no indication of relief, apart from the occasional name attached to a prominent peak. This provides a clue as to why Google’s maps are of such poor quality. Clearly, they have been compiled directly from commercial satellite images, and information incorporated into a map is not verified by any system of checking on the ground.

There is also evidence that human input into the production process is either limited or nonexistent. Take a look at the satellite image corresponding to a map segment with a lot of roads. The ‘roads’ superimposed on that satellite image do not coincide precisely with the roads on the image. Not that a software-generated map is necessarily inferior to one produced by conventional means. The process of using stereoscopic pairs of aerial photographs to generate a relief map has been automated for more than 40 years, and the accuracy of the resulting map can be relied on.

Another absentee is a compass rose, which probably reflects the tendency among most users to hold their map so that north is at the top, regardless of the direction they are facing. This habit is inevitably encouraged by the presence of text labels on a map, which are most easily read if north is at the top. However, I once met two Ordnance Survey employees, who, to my surprise, brought a few examples of their organization’s maps with no text on them to the meeting. They too were surprised when I asked why such maps weren’t available for sale to the public, but there are many situations where text labels on maps are a distraction, and knowing what a particular landmark is called is of absolutely no importance when navigating across featureless terrain.

I do have some sympathy with Google’s position though, because in my experience most people are overwhelmed by the sheer quantity of information contained in an Ordinance Survey map, and for general use, not involving any detailed navigation, a pared-down product that includes only the most important information may be useful. However, that information needs to be accurate, or it is of no value.

Unfortunately, accuracy is not guaranteed. Here are two examples of egregious errors that I discovered in the process of exploring the northwest of the New Territories:

The red asterisk marks the correct position of the Hau Ku Shek Ancestral Hall (note the spelling error).

The bridge over the tributary of the Kam Tin River marked by the red asterisk does not exist, and the road actually follows the red dotted line. The footbridge about 100 metres upstream also does not exist.

Both mistakes could have been avoided by having someone checking on the ground, but this would have meant that the maps would have been more expensive to produce, and Google’s maps are a good example of a product that has been put together as cheaply as possible. As such, they have almost no practical value.