Google Earth, Google Maps, and Trinidad

I've been working on a lot of land related issues recently and have been 'playing' with Google Earth and Google Maps. As one would expect, they aren't perfect - but armed with my own old Garmin 12 GPS I was to find out just how inaccurate they are. Or how inaccurate my GPS was.

Check Your Equipment

When I first started the GPS in conjunction with Google Earth, I was immediately dismayed. I was using the Naparima BWI datum, which should be fairly accurate - and I verified the waypoints I took on more than one occasion. There was the same error just about every time, so I did some digging on the Naparima BWI datum because the first thing you're supposed to do is check your own tools. I came across this page which had a lot of useful information on it:

Due to station destruction, points based on the Naparima 1955 datum are difficult to find and may be well away from the area of interest. Accordingly surveyors have adopted a pragmatic approach and often use old Cassini stations. This is achieved by applying a '‘transformation' to the Cassini point to obtain co-ordinates in Naparima 1955 – and vice versa. Unfortunately the transformation is always calculated on a local basis. This is undesirable as the values obtained are entirely dependent on the points used. Hence different surveyors using different points will obtain different answers. This leads to disputes regarding the correctness of the surveys...

...Two private surveyors interviewed, Mr Winston Mohammed and Mr Sasha Ian Addo, have developed an efficient method of working with GPS. They have achieved this by adopting ‘known’ co-ordinates for some stations and using the same transformation throughout Trinidad and Tobago. This is exactly the thrust of the recommendations made in this paper, but with appropriate attention paid to national standards...

... It seems beyond doubt that the most used datum is Naparima 1955 consequently there exists a significant amount of work in this datum. It is recommended that this datum continues to be used for land survey in Trinidad and Tobago...

There is much more on that page, but suffice to say that since I am working with land surveyors to some degree, I was re-assured about the use of the Naparima BWI datum in Trinidad and Tobago. And this ended up being an error - a recent trip to a land surveyor's office allowed me to find a letter from the Lands and Surveys division which now requires WGS 84 datum on cadastral sheets. So - back to WGS84, where I started.

I did all sorts of testing with the GPS for resolution, and used a few waypoints - rechecking them with different datum. The margin of error would best be described as trivial.

What was funny was that I came across a proposal a week ago that had a cadastral overlay on Google Earth itself. I chuckled because it was being used as a guideline - but there were landmarks I am familiar with that were notably off. I sent them a note.

So I decided to check the maps on Google Earth here in San Fernando, Trinidad and Tobago. The picture at top left is a screen capture (for academic purposes) of an intersection in San Fernando that I often use as a reference. You can look at a larger view here; the differential within Google Earth and imagery itself is off by approximately 80 meters. Other intersection checks throughout the area demonstrated the same - the overlay of streets is off by about 80 meters South South East throughout San Fernando. I haven't bothered with an exact calculation because terrain can alter this as well.

So - 80 meters. But being 80 meters off didn't explain the problem I've been seeing. Here's an example:

Approximately 261 meters off, South South West? This falls under 'not good'. GPS conditions were accurate, ye olde Garmin was reporting within 3 meter accuracy. I can use the GPS to get back to the same position using both Naparima BWI datum and WGS84 datum, but it shows up in a wildly different position. However, GPS distances are well approximated on Google Maps, though the differential may change with terrain. That area has many hills. Checking of other points within the South Trinidad region consistently showed error with Google Earth/Google Maps that ranged from 210 meters to 260 meters using WGS84 Datum - all, interestingly, toward the general bearing of South West (to varying degrees). San Fernando Hill's error is more Westerly than in Dow Village, so it is easy to hypothesize that the error is in fact due to terrain detail (topography inaccuracy) as well as the offset I found within Google Earth/Google Maps without the GPS.

So, no - GPS and Google Earth/Google Maps do not work well in Trinidad and Tobago, and they are unlikely to unless you enjoy being within 260 meters. They are good for looking over areas in a very general way - which they were designed for - but navigating using GPS with Google Earth or Google Maps is probably going to get you in a ditch - or a building.

Some further references on the accuracy of Google Earth that I found:

• Where to report Google Earth errors: Can't expect it to get better without some feedback.
• Interesting comments on accuracy of Google Earth in conjunction with GPS.
• Why Google Earth and Prime Meridian do not line up.
• Why GPS positions do not appear properly on Google Earth.

Me? I'm using my own differential for my own reference based on landmarks and official surveys... Which is still probably flawed, but I haven't fallen in a ditch yet. And, oddly enough, it is exactly where I started. But at least I do have a general reference where I did not have one before.