Mapping of differences between GPS location and Google Maps

Mapping of errors between GPS location and Google Maps

Update 2006/02/27: Never mind... I have found egregious/obvious stitching error in GoogleMaps, which alone can explain all the differences seen below.
Also, I have done some precision estimations with my GPS, which tend to indicate it's rather precise.
So for now at least, I'm putting the blame on GoogleMaps. I will probably design an algorithm to adjust for these differences and go ahead.

Here's my somewhat systematic attempt to map the differences between positions indicated on Google Maps (as plotted from coordinates read from a GPS) and the actual positions (visually located on the satellite photo):

Ref._Point: 		GPS_location NSats HDOP: 	Actual_location:
pr_sdu			-22.912068,-43.167055 7 1.1 	-22.912052,-43.167203
pr_perimetral_1		-22.895695,-43.178715 11 0.8 	-22.895736,-43.178844
pr_perimetral_2		-22.889542,-43.216043 10 0.9 	-22.889569,-43.216159
pr_mocangue		-22.873148,-43.134517 10 0.9 	-22.873131,-43.134781
pr_nit			-22.888042-43.115148 9 1.0	-22.887957,-43.115308
pr_nit_002		-22.898790,-43.111705 7 1.2	-22.898741,-43.111918
pr_nit_003		-22.902555,-43.107368 9 1.1	-22.902575,-43.107744
pr_nit_005		-22.913955,-43.082222 7 1.5	-22.914020,-43.082403
pr_nit_007		-22.921138,-43.054103 8 1.3	-22.921046,-43.054519
pr_nit_008		-22.916255,-43.048442 9 1.1	-22.916194,-43.048832
Notes: all GPS locations were taken on 2005/11/01 around 00:00 GMT-3, using a Holux GR-213 GPS in NMEA mode, connected to a Tungsten E PDA running Cetus GPS 1.1; the measures were taken aboard a car, with the car stopped. Also, the GPS was loaded with the default datum (WGS84)

Please note that I took care to take the best possible GPS readings: never less than 7 satellites in view, yielding never more than 1.5 HDOP (Horizontal Dispersion Of Precision).

Below we have some images generated with Google Maps, with markers showing both the GPS location (ie, the coordinates read from the GPS entered into Google Maps) and the "Actual" location (generated by marking directly on the photo the point where I actually was when I took the GPS coordinates).

Please note that the "GPS location" is always the marker more to the right, and the "Actual Location" is always the marker more to the left.

My conclusions so far are:
1) The offset between GPS and "actual" positions is always more or less in the West direction;
2) This offset is in the range of 10 to 30 meters, getting larger the more to the east the location is.

I have not yet found an explanation for those differences; If you think you can help, or want to know more about that, please reply on this Google Maps group thread , where I posted my original message about that and me and other folks are discussing it.

pr_sdu

pr_perimetral_1

pr_mocangue

pr_nit_007 (bottom pair), pr_nit_009(top pair) and pr_nit_011(middle pair)

Sat Nov 5 20:36:08 EDT 2005 Update: new positions measured, more to the west: 
Ref._Point: 		GPS_location NSats HDOP: 	Actual_location:
pr_rjo_1                -23.014223,-43.297820 8 1.1	-23.014267,-43.297730
pr_rjo_3                -22.970517,-43.217114 10 0.8	-22.970702,-43.217136
pr_rjo_5		-22.935178,-43.185828 7 1.3	-22.935107,-43.185904
That is, more to the west, the east-west difference between "GPS" and "Actual" locations first diminish, then almost zero out, then invert themselves (that it, "Actual" more to the east relative to the "GPS"). Here are the Google Maps with the markers, for the positions above:

pr_rjo_1

pr_rjo_3

pr_rjo_5

Last Change: Mon Feb 27 22:33:35 EST 2006