The ghost of my high school chemistry teacher is in my ear, patiently explaining the difference between accuracy and precision.
Relevant feature request ;) https://forum.inaturalist.org/t/rename-location-accuracy-circle-to-precision-circle/9774
Lo, I just reviewed this in class this past week in first-semester chem ahead of the upcoming cumulative final exam. ;)
Well, unless there isn’t. If I made the observation in an area I know extremely well, I can sometimes pinpoint to within a footstep or two where I was from the Google Maps satellite image. Yes, you also said there might be an error of only a few cm; but the marker icon covers a larger area than that anyway, even at maximum zoom. I still include an accuracy circle, though, although in some cases it could be smaller than the area you could see by looking around.
If your observation is of a shade tree alone in a pasture, you can pick out that exact tree on the satellite image, if you know that pasture well enough. If you then set the accuracy circle at the exact diameter of the tree’s crown, does that count as having some error because it isn’t a single point?
Sure, of course this is the case, where if you know a ‘landmark’ the accuracy essentially becomes a moot point because you can manually pick the spot on the map. But that’s not really what the thread has been about; your example is ‘you telling yourself/iNat where you were’ vs. the discussion being about accuracy in the case of ‘the GPS telling you/iNat where you were’. In the case of the latter, there will always be some degree of error.
I’ve worked with extremely advanced military maps and GPS systems, and trust me, you’d be hard-pressed to find an existing map and/or GPS that is accurate beyond a meter. Plus, the Earth is constantly shifting, so even if you had a “perfect” GPS and map, it would quickly become inaccurate.
Google maps change from time to time, it can be so that in a version of 2030 the spot you chose would be some metres away from the placement of observation (and who knows what version is right) so accuracy of some metres is a way to prevent that.
when it comes to making data useful, i think there are bigger things to worry about than whether or not observations are recorded with positional accuracy values. i would rather have good coordinates without associated positional accuracy than bad coordinates with positional accuracy. and no one should ever just assume that coordinates or positional accuracy values recorded for any given observation in iNaturalist are necessarily reliable.
i’ve given this example in other threads before, but i’ll provide a better visualization here than i have before (below). at a park in my area, there have been now 9 observations of a specific patch of Heteropterys glabra plants. it’s grown a little over time, but the entire patch probably is not bigger than 5m in diameter. here are the observations in iNaturalist: https://www.inaturalist.org/observations?place_id=65598&taxon_id=542500, and here is a visualization of those observations with their accuracy circles, based on the data recorded in the system (only 6 of 9 of the observations had positional accuracy recorded):
if the locations of the observations were reliable, then all of the circles would overlap, but that’s not the case. one of the non-overlapping circles has a relatively small positional accuracy value recorded – just 15m – but if you compare it with the other observations, it looks like the true positional accuracy value should be more like 30m. and more to the point of this particular thread, if you look at the 3 observations without positional accuracy recorded (no circles), they look like they more reliably represent the true location of the patch of plants than some of the observations with positional accuracy recorded.
so, yes, if you can record good values for coordinates and positional accuracy, then please do. and i’m not against educating certain people about what positional accuracy is and how to record it properly. but i don’t think everyone needs to be burdened with the technical intricacies of all of this. leave it to the scientists and analysts to correct for errors in the data they want to analyze, and leave it to the engineers to figure out new technical approaches to capture better data.
Yes, but there is always the possibility that the satellite imagery isn’t correctly positioned geospatially. I do a lot of mapping with OpenStreetMap and we have access to several “brands” of satellite imagery (Maxar-Premium, ESRI, Bing, etc.) and the actual position of a feature can be off by several meters.
To make matters worse, the positional accuracy of any satelite imagery varies depending on where you are on the globe, Alaska vs Thailand, Europe vs the east coast of the U.S. OSM mappers are encouraged to use GPS traces previously uploaded to determine an “offset” and then adjust the imagery to align with those traces where they follow a road, for example.
This means that finding that “shade tree alone in a pasture” and picking its coordinates off the sat imagery in Google Maps does not guarantee that those coordinates are accurate.
I was recently able to work out batch editing to update some of my old observations with location accuracy. You can apply set the radius in meters to 200 observations at once. You can also adjust particular observations to different numbers in the batch edit module.
Yes, I know you can batch edit the accuracy data, but I don’t know of a way to search for observations that lack these data. I wouldn’t want to overwrite the data of observations that already have them.
Doesn’t this do it ?
Yes, but you can’t batch edit from there. I have thousands of values to backfill. I’m not going to do that one at a time.
No, but you can export it and create a comma separated list of the observation numbers to add as a url parameter on the batch edit screen.
I didn’t know you could do that. I’ll give it a try. Thanks!
I tried that on my own observations, via this URL:
which shows 33 observations lacking accuracy.
However, when I try to export, the export page apparently ignores the
acc=false parameter, and exports all my observations. Here’s the export URL:
which displays and exports all 354 of my observations. I tried several variations on the URL, including adding
acc=false into the query edit box on the export page, but couldn’t find any way to get it to accept the
This is a known issue on the export screen where it drops certain parameters. Sorry, I did not know this was on the list of ones that are dropped when I posted above.
It is not obvious, but you can manually edit/paste into the first box on the screen it takes you to and re-add that parameter into the string it shows.
if that’s not working for that set of parameters, you can always try using the API (see https://api.inaturalist.org/v1/docs/#!/Observations/get_observations).
this may also help get info from the API in a more human-readable format: https://jumear.github.io/stirfry/iNatAPIv1_observations.html?user_id=twainwright&acc=false.
This topic comes up again regularly. I will repeat a request I have made before, which no one has ever followed up on. Can someone who understands this issue well do a short writeup about it, explaining what the issue is, why it is important, and exactly what information should be retained (presumably a single floating point number which is a distance in meters)?
I am a prolific observer with thousands of observations whose locations are handled automatically by my tool chain. But the accuracy information is not there. I will not do these by hand, but would be happy to urge (1) the JPG standards committee to adopt a standard way to record this, (2) my camera manufacturer to update the firmware to put the information in the EXIF data on the photo at the time the picture is taken, and (3) Adobe to make sure that Lightroom preserves this information. Yes, that’s a lot of organizations that aren’t going to move quickly. But it was probably 7 years ago that we first had this same discussion. It might be solved by now if we had started pushing for it then.
I don’t think that’s the correct use of that flag. I think that should only be used if you believe that the location of the observation isn’t anywhere within the accuracy circle (which should really be called the precision circle).