Slashdot Mirror
Nicaragua Raids Costa Rica, Blames Google Maps
Garabito writes "An error on Google Maps has caused an international conflict in Central America. A Nicaraguan military commander, relying on Google Maps, moved troops into an area near San Juan Lake along the border between his country and Costa Rica (Google translation of Spanish original). The troops are accused of setting up camp there, taking down a Costa Rican flag and raising the Nicaraguan flag, doing work to clean up a nearby river, and dumping the sediment in Costa Rican territory."
The first incident was apparently due to the local Indonesian authorities persisting in the use of 1933 Dutch maps and the Australians using more recent Indonesian maps. The Dutch map indicated that the Mota Bicu river formed the border. However, the 1992 Indonesian map used by the Australians showed the border as being 500 metres to the west of that position. Apparently, the Indonesian map reflects a post-1975 decision to make the border a fixed provincial border not dependent on the river as a landmark, with the result that as the river changed course over time and as the villagers moved with it, the village of Motaain would shift its location from East to West Timor and vice versa....
http://www.icrc.org/web/eng/siteeng0.nsf/html/57JQZ2
Given that Maps/Earth is a free service, and Google isn't exactly a charity, it would actually not at all surprise me if the quality of Google's offerings for a given area is strongly correlated to that area's level of wealth, IT development, and existing national mapping services and/or 3rd party information providers.
Consider, most of what Google does, it does either as an experiment/long term investment, or as part of its core ad-selling business. Now, their mapping services have been around for a while, and don't seem to be an experiment(and the concept of geographically localized advertising is obviously attractive), so it seems very likely that they are ancillary to the core business.
Consider: Where are ads most valuable, per impression, and consumer data/metrics most valuable? In wealthy, populous, areas with good internet penetration and lots of electronic commerce.
Where is good mapping data cheapest: Where some existing national, regional, and/or local mapping/planning authority exists, and has already collected decent records in a standardish format, at public expense and available for no or nominal money.
Therefore, you would strongly expect Google to have the best starting data in relatively wealthy, stable, well-governed areas, and have the greatest incentive to do the labor-intensive data cleaning process of sending out GPS-carrying surveyors and streetview cars and things in dense, wealthy areas. The further from either of those you go, the more likely it is that Google's "data" are whatever satellite or aerial photos they managed to pick up cheaply and georectify well enough that there aren't visually obvious gaps and tears. Because modern sensors are good, such data are actually likely to be perfectly OK for things like physical geography lessons; but there isn't actually a big black line painted along most national borders, satellites aren't going to see that. And, given that this incident occurred in what sounds like a relatively sparsely populated Latin American border region, I'm guessing that the place isn't crawling with streetview cars...
If what you care about are things like national borders, military installations/posts, and geographic features where some kind of army engineering corps is doing work, the national mapping service is probably actually the place to go. Unfortunately, they are probably not set up with a very nice user interface. Paper maps or some ghastly 80's GIS frontend, usable after a few months of specialized training, are a definite possibility. Google, on the other hand, has virtually no incentive to care about such things(at least in their free civilian offering, I don't know if they have a government/intelligence version); but has a decent interface, and produces results with a lovely air of apparent accuracy most of the time.
Consider some history: During British colonial rule(first via East India company, later direct) The Great Trigonometric Survey (1802-early 20th century) produced some quite accurate maps of the entire subcontinent, and some pretty hostile terrain, using nothing more than hand tools, dead trees, and pre-computer math. Surveying, like civil engineering, is nontrivial; but you can actually do an excellent job with quite primitive tools. Satellites and GPS enabled everything sure makes the job easier, and computers sure make the interface nicer; but there is nothing except disorganization stopping even a country with early 19th century technology from producing excellent maps.