Google Map accuracy

better than if he used a handheld, worse than if you surveyed it.

Google uses public imagery that has been georeferenced by others. If the others made a mistake, you can see it when you flip to orthos from different dates.

I picked a coordinate on Google Earth at my house, right between the house and garage. I checked it with my handheld and it checked within less than ten feet.

But when it comes to street names, Google's accuracy is quite poor.
The street leading into Port Newark, Port Street, had been listed as Olympia Drive for years. Other streets are paper streets and still others are not shown where they are located.
NJ One Call uses Google and I manage the One Calls for Port Newark and other Port Authority facilities and the descriptions of One Call locations can be confusing, especially to outside non faciliaty owned utilities.

> better than if he used a handheld, worse than if you surveyed it.
>
:good:
I use GE as a rough check when filling out the Lat/Long for Elevation Certificates. Always hits close to the house for me. Even the ground elevation is in the ball park.

I used to not be surprised at 50-75 ft discrepancies between GE and known positions, or discontinuities where photos met. I think they have gotten better, but have inadequate data to prove that.

Apart from the accuracy/inaccuracy of google earth in the situation you describe there is also the aspect of how well can they 'see' a well?

What does a 'well' look like on a photo?

Over the last 5-8 years that google earth has existed we have seen situations where it has been way inaccurate, 200 meters for example. Those were probably isolated events, but I am not sure they have been corrected. Be aware and validate it.

I would not trust it unless you had at least some good quality GPS control on ground points to at least identify gross errors.

The currency of the photos is one critical factor.

Your ability to understand and interpret orthophotos is another. For example if the ground elevation has changed from what is modelled in the DEM's most often used to correct the imagery you can have imagery greatly displaced from the ground.

My experience in a well controlled area is that the imagery is probably in the 3 meter range. I would say that is the 'best case' situation, if you can actually see the object you want coordinates on.

Sometimes in this kind of exercise one has to wonder about what datum is being used. by you as well as GE.

Remember that NAD83 and WGS/ITRF datums have be cut loose and there can be several meters just in that transformation. This often shows up even at the noise level of consumer grade GPS, let alone resource grade units. GE doesn't really tell you the metadata for their coordinates. You are lucky to even get the photo dates. Each edition(date) of GE will have different quality.

So you have at best 3 meter positional quality readily identifiable ground objects where the ground has not changes, add to that a few meters of possible datum ambituity and that is my answer.

Of course it is relatively easy to actually test for yourself by comparing GE derived known points with actual survey data. So do that.

If I were submitting this as a proposal I would expect the contracting people to require you to submit some kind of statistical validation for the approach.

- jlw

Use the DS-World software from NGS to see how well the Google Earth imagery matches up to some of the landmark objects in your area.

Accuracy varies with each image. Some are georeferenced orthophotos, some are georeferenced raw photos. In ALL cases, GoogleEarth uses a spherical model of the WGS84 Earth and NOT an ellipsoidal WGS84 model. If the area of interest is near the image isocenter, accuracy can be good. If the area of interest is near an image edge, accuracy can be awful.

It can vary from image to image.

With GoogleEarth, accuracy is a matter of chance.

When panning across Google map, I have noticed some really bad edge-matching.