I’m uploading pics from recent fieldwork in a somewhat remote, rural area, and there’s an odd mismatch in Google Maps in the general area between the actual roads as shown in the satellite imagery, and the ‘superimposed’ grey bars representing the roads. See eg:
The ‘real’ roads, are the correct ones, there aren’t two roads running parallel here; the grey ones have all been shifted up to 50 m to the south/east. This is the case for quite a large area around here, and then moving further outwards it returns to normal where they’ve superimposed exactly. Is there a particular reason here the grey roads are so far off?
Also most importantly: for photos I took along the roads like this in the area, my phone is geotagging the photos to the wrong grey roads. I’ve been shifting the locations manually to match the correct spot on the real roads; this would be the best approach right? I ask because I’m assuming it’s the grey roads that have been overlain onto the map incorrectly, rather than the entire underlying map being ~50 m out, but wanted to double check
This happens a lot with Google Maps, especially with trails. Almost always, my GPS points, from a continuously-running Garmin etrex follow the trail or road shown on the aerial maps, so I’m pretty sure in almost all cases, they are correct. Phone coordinates are not as accurate, so it is a good thing to move the location to the nearest place along the road.
However, if your phone locations match the grey Google roads, then you may have discovered a case where the map has not been correctly located.
How about other iNat obs in your area? Most people don’t pay any attention to where their location lands, so if their obs also follow the gray road, like yours do, that is pretty good evidence the map registration is off.
The difference between your gray roads and the real roads looks suspiciously like the wrong datum has been used for one.
I use a geotagging app on my phone, and it’s generally pretty good. I’m checking through all my pics along roads now, and they’re all along the grey Google roads, so I’m more so suspecting this is the case now.
I think best practice here might be to still readjust to the imagery though? That way if eg someone tries to re-find a plant I photographed, they can match to landscape features (gates, river crossings, etc). Unsure though which approach to go with
One alternative to test if the Google roads layer or imagery is wrong is to display an observation on OpenStreetMap (on the observation web page, click “Details” under the map, then scroll down to “View on” an click “OpenStreetMap”). In my experience in Oregon, OpenStreetMap is more accurate for minor rural roads than Google (presumably because OSM is ground-truthed by many rural recreationists). If OSM and Google agree on the road location, then the imagery is off.
Note also that on the observation page map, in the Google bar on the bottom of the map you can “report a map error” to Google, and they might fix the problem.
Within the property itself (it’s very large, 20,000 ha) as you go towards the edges, the roads match up perfectly. It just seems to mostly be in the centre and towards the south, one weird big blob of mismapping
That depends on who you think is likely to re-locate your observations. Re-locating to the aerial imagery assumes someone is using the same imagery that you mapped to - I.e. an iNat or other Google user - and assumes that Google doesn’t correct the offset the next time they reload their satellite imagery. In general, I would trust my GPS locations unless there is an obvious problem with the device - that way anyone who uses GPS to relocate your observations will get close to your location, not the potentially mis-mapped location based on Google imagery. I would expect any researchers using your observations would download the observations with GPS points and use them directly or remap them in their own software.
There seems to be a tendency in this thread to figure out which map is right or wrong. The answer could be neither. When I worked a lot with maps at my last job, I just recall many headaches with differing map projections and coordinate systems. If some data points were not collected or projected in the same system as the base map on which they were plotted, they can appear “off”. Neither the base map nor the data points were in error; it was just a case, so to speak, of the mapping equivalent of “apples and oranges”.
makes sense, but I’m intrigued as to how this could happen for only a very specific part of the map, whilst the surrounding areas (including on the same property) are all ok. A very weird/idiosyncratic piecemeal job?
I doubt it is a matter of differing coordinate reference systems or reprojection in this case.
However, aerial photos are not so easy to orthorectify… From experience, Google (or Bing, or Yahoo, or…) aerial photos - improperly termed ‘Satellite’ views - are not always reliable; it is not uncommon to measure systematic shifts reaching several meters locally. And their vector overlays are not always reliable either, being sometimes too coarse.
TL;DR: aerial imagery can be inaccurate - the geolocation of observations pinpointed purely from such ‘maps’ can be inaccurate as a result - for pinpointing needs better use good maps of known accuracy, or good instruments.
Yes, if this were a North American example, I would have said that this particular offset looks very much like a datum-shift - the difference between the old NAD1927 (still seen on some older maps) and WGS1984.
It’s possible that a portion of the aerial imagery in this area was aligned to data based on an older geodetic datum, because maybe those were the only data available for the area.
I agree that the best course is to maintain your original phone coordinates and not try to second-guess them based on ambiguous data sets.
a nice read, but the apparent difference in scale of the problem is intriguing. The article says
in metro areas, the accuracy of aerial imagery could be on average about 0.5 meters. But rural areas could be on average 1 to 1.5 meters or worse. For example, here (https://bit.ly/2ZcitfC) is an article that examined the accuracy of Google Earth imagery across the city of Montreal, Canada. The authors found that accuracy could be as good as 0.1 meters in parts of the city but as poor as 2.7 meters in other parts.
So in their examples here, really no big deal at all. In my case, however, I’m getting a 30-40 m difference in places!