Accuracy Estimator For GPSTest; correcting the correlation between satellite image and actual Lat/Lon

[Jim bell]

  I am aware of the visual accuracy estimator in the app GPSTest, although I've never seriously used it.  One reason is that I have questioned how good it is, really.  After all, I think it relies on somebody's calibration of the correlation of the actual lat/lon and the satellite photograph of earth.   I assume that while good, maybe even very good, it cannot be perfect.   So, how imperfect can it be?

A couple of days ago, I got back from a town named Lincoln City, Oregon.  It possesses a point, 45N 124W, which happens to be in somebody's rather tiny backyard.  I checked Google Maps, and it shows the point as being about 1 m away from a wall, sorta a small 'cliff'.  I have long wondered, how accurate is the photographic data with respect to the actual lat/lon.

So, I measured the actual location of a nearby manhole cover, then entered the precise location into the Google Maps.  The location shown on the screen was about 5 feet east, nearly all east, on the display.  So, that told me that the actual location of that confluence would be 5 feet west of the seeming location. 

Five feet might be seen to be good, or not good, depending on your standards.  So I had an idea for an improvement.  Suppose reference marks were painted on a scene, and by scene I mean potentially the entire world.  AND suppose their exact locations were measured, to an accuracy  well better than the best visual accuracy that satellite photography would get.  And suppose those locations were reported to somebody with the ability to get the process 'fixed':  Could they use those reference marks to fix, to correct, this inaccuracy?

I have two ideas for a reference mark.  One would be white circles, painted on roads, possibly disused roads so that they don't quickly wear down.  Perhaps they could be painted on manhole covers?  There might be an advantage to painting on manhole covers.  If you look at a scene on Maps, you will notice that manhole covers are slightly glowing.  The reason, I believe, is thst while aged steel isn't especially reflective, it nevertheless is reflective ENOUGH so that manhole covers do indeed appear on the satellite photographs. What they are reflecting is the bright sky.  

It would probably be best if these locations were defined to be away from heavy traffic.  Maybe they should be on sidewalks, or on dead-end roads.  

The work could also be somewhat shared:  A person who wants to help but cannot paint or measure could bring up  Maps or Earth, and search for good candidates for painting and measurement.  He could make a well-ordered list of such locations, and the person doing the actual painting and measuring work could visit these locations over time.  Or, a person without a survey-grade GNSS coulld paint the circles or other marks, and measure their approximate locations, and then publish the results for somebody who has the GNSS device to later show up and do the measurement.  

One advantage of painting white on manhole covers is that past photographic imagery will already be tagged, slightly,  with those locations.  So, old imagery can be readily corrected for lat/lon.   The targets won't be nearly as visible, but they should be predictable and consistent.  

Another advantage of this system is that it is somewhat "self-help':  A person who is interested in improving location accuracy over a region can do his own work.  I happen to live in the Greater Portland Oregon metropolitan area.  I could imagine painting and measuring one location in 30 minutes or so.  At 16 per day, it might take 6 days to reference-mark 10,000 square miles.  

But if they are painted white, they would truly stand out.  I have questions, however.  At what density must these white dots be painted in order to correct the map image?   Once every 10 miles, horizontal and vertical, perhaps? 

Another would be two intersecting quarter-circles, which I believe is sometimes used as a type of 'bullseye'.  

My question would be:  Could the "somebody" who adjusts the "satellite view" layers with the GNSS values be convinced to adjust their software to employ these reference marks?  

Who would paint and measure these locations?  I, for one, would volunteer to paint and measure perhaps 100 such locations, over time.  But anybody and everybody could do it.  My ZED-F9P would easily locate to 1 centimeter, which I suspect is 30x better than might be sufficient.  My understanding is that the best satellite photography is about 30 centimeters resolution.