[I wrote this a couple years ago]
This post is about investigating Google Maps’
accuracy – specifically, how accurately you
can measure distances off of it. If you’re not
interested in that, no need to read any further.
Google will give you a straight-line distance
between any two points on the map, but it
turns out you can get a much more accurate
distance than the one Google gives you.
(Some? Most? of the time, anyway.)
A few feet east of the red marker
https://goo.gl/maps/dS1v59YwGTLDhX797
you see the shadows of two posts –
they’re four-inch steel pipes. Two more
posts just east of the marker at
https://goo.gl/maps/HQgWaFZLFsJo6Q3bA
Google says the southeast post of the west pair
is 3275.4 feet from the southeast post of the
east pair. Actual distance is 3281.1 feet,
correct within 0.1 foot. A 0.2% error is typical
for a Google distance; the problem is that when
Google calculates the distance it assumes a
spherical Earth, so in an extreme case
(a north-south line at the equator, or at
a pole) it’s off by 0.5%, that being how
unspherical Earth is.
Fortunately we can hope to eliminate that
particular error, because Google gives us
the latitude-longitude wherever we click
on the map. We take the lat-lons at the
ends of the line we’re measuring and
calculate the distance ourselves, on the
unspherical Earth.
Good a guess as any: Google says the
posts are at 37,80319465N 122.3174799W
and 37.7997694N 122.3069758W. (You
right-click on the map and it gives you the
lat-lon to five decimal places; left-click on
that lat-lon and it stores the lat-lon to 15 places.)
On the GRS80/WGS84 spheroid those two
lat-lons are 3281.35 ft apart; if we want to be
quite correct we’ll cut it down to 3281.33 ft
since we’re below the spheroid. So the error
is 3 inches – for all we know, just because
I can’t expect to get the cursor point in the
exact right spot, even at maximum close-up.
The lat-lons aren’t correct to 3 inches,
of course. They’re probably off by 6 feet.
But the error is the same 6 feet, in the
same direction, for both posts.
Two more pairs of posts
https://goo.gl/maps/cDM2zhggye2hPpSV7
https://goo.gl/maps/XiHpkpoHiJ1xAd2D9
As before, we get the lat-lon of the SE post
in each pair. Based on Google’s lat-lons,
the calculated distance from the west post
to the next one is 492.58 ft; actual distance
is 492.74 ft, again correct to better than 0.1 ft.
Calculated distance post #2 to post #3 is
918.39 ft; actual is 918.12, Calculated distance
post #3 to the east post is 1870.39 ft;
actual is 1870.24.
Judging by the shadows, the posts likely were
photographed together, on one picture –
no seams between them, where Google
stitched together two pics. We have no idea
how much area each pic covers, or how much
error is introduced by a seam. But as it happens,
we can easily enlarge the investigation.
See the white dot at the center of the concrete
rectangle, just southwest of the red marker?
https://goo.gl/maps/935UeVSad5b9b7WE8
That’s the concrete around a fairly new
survey marker; NGS gives the marker’s
NAD83(2011) lat-lon correct to something
like a centimeter. So we can see how far off
Google is, and how consistent its error is.
Turns out it’s more consistent than you’d
have thought possible. Google’s lat-lon for
the above mark is 6.1 feet from the actual
position, in azimuth 290 degrees.
Next survey marker is the concrete dot
a few feet south of the red marker
https://goo.gl/maps/8NvZi2pQiG9oywQA7
There, Google’s lat-lon is 6.28 feet from
the actual lat-lon, in azi 288 degrees.
Next marker is the dot a few feet SE of
the red marker
https://goo.gl/maps/isEiQmDkGb3BDgdG6
Error there is 6.46 ft in azi 292 degrees.
Next marker is at the center of the fence
rectangle just south of the red marker
https://goo.gl/maps/2zGQyYtMZMuorQvv8
Error is 5.68 feet, azi 289 degrees.
Next marker is next to the red
https://goo.gl/maps/TG6wUVxfnmfbu7dF8
Error 6.20 feet, azi 291 degrees.
Last of the new markers is 3-4 ft NE of the red
https://goo.gl/maps/2qp9okLpMM6K1ieL7
Error 6.20 feet, azi 293 degrees.
There’s another mark at the foot of Clay St,
downtown. It’s not visible on Google, but with
a little extra work we can learn Google thinks
it’s 6.10 ft from its actual, in azi 292 degrees.
And one more, also not visible, set in the
concrete block at
https://goo.gl/maps/um23MWQC9VVkZiTA8
Google thinks the north corner of the block is
6.98 ft in azimuth 292.5 degrees from its
actual position. Ignorant guess: maybe the
error is larger here because this is the only
marker that’s not at ground level – it’s a
couple feet up in the air, and if the camera’s
not vertically above it, Google can’t quite fix
the resulting error. Which would mean Google
doesn’t do as well when marks are at different
elevations, instead of all being near sea level
as my marks are.
Maybe eventually I’ll get around to enlarging
the investigation further, checking survey marks
farther away. NGS has numerous NAD83(2011)
marks in San Francisco and elsewhere, miles away
from the ones I checked. (I ought to do it before
an earthquake throws everything off.)
I left out some details in the discussion above;
if you have questions, I may have answers.
Or maybe not.