The need for assigning accuracy
We are seeing observations with a range of locational accuracy, a measurement of the location’s precision. Some people are posting observations from with very general locations (over 1 km), while others are posting observations that are mapped to within 5 or 10 meters of the actual observation. No matter what your accuracy is for an observation, it is important to assign the accuracy so we know how the data can be used. An observation lacking an assigned accuracy has very limited value.
The observations that you provide are used for a variety of conservation purposes. Some do not require a high degree of accuracy, like county lists. Other projects, like distribution modeling, require an accuracy of 30 meters or less. Observations without an assigned accuracy will not be used for environmental review, conservation planning, or distribution modeling.
Fortunately, it is easy to assign accuracy with iNaturalist. If you record the location with the iNaturalist app, then it assigns the accuracy automatically. However, if you assign the location using the metadata from a photo, or by manually mapping it through the web interface, then you need to add the accuracy by editing the observation.
The accuracy field, labeled “Acc (M)”, is located directly below the coordinates, and just above the map. Just enter the distance in meters that includes the entire area where the observation could have taken place. When you assign accuracy, a red circle around the point will appear. You can click on the circle and drag it in or out based on the map and it will automatically adjust the accuracy. It is also possible to edit accuracy for multiple observations using the batch edit tool.
If you are concerned about mapping a precise location, you may want to adjust the geoprivacy setting. You can learn more about geoprivacy here: http://www.inaturalist.org/pages/help#geoprivacy
Comments
Maybe this is a somewhat dumb question, but how do I know the geolocation accuracy of a photo taken by an iPhone? I could guess I suppose, but how is that any better? Is there any reason not to trust a phone's geo data and just add an arbitrary but somewhat precise accuracy of 1-3 meters?
Good question, thanks for commenting. I usually use the iNaturalist app; which documents accuracy. If I use the photo's metadata, I usually add an accuracy of 10 to 15 meters, which is great accuracy. If I am in a wooded area or a canyon I might double check the location on the satellite map. For almost all uses of the data, the important thing is that the accuracy is under 25 meters and not something like 100 meters, or 1,000 meters. So I wouldn't stress about the difference between 10 and 15 meters, I would just assign the accuracy that you feel confident covers the location. If there is a question, then round up.
This is a very important field yet it is easy to omit. I've been trying to get in the habit of double-checking my locations and assigning acc accordingly. DOR (dead-on-road) specimens are a good way to confirm locational accuracy since you can zoom in and examine location with respect to the road. My smart phone usually maps things pretty well (I can tell which lane it was in or if it was on the shoulder) and I am able to use road width for assigning accuracy (usually 10m). The hardest part is simply remembering to assign accuracy in the first place.
Okay, thanks for the thoughts on this. Glad it doesn't have to be a tight tolerance. Now that I think about it, a large number of my photos are taken with a DSLR without GPS capabilities, so all of my location info for those is completely manual anyway. Not too hard to assign an accuracy while working with the map.
I just noticed an interesting distinction when using the iNat mobile app. If you create an observation in iNat mobile, then import a photo from your camera roll (one you took outside iNat), and, finally, allow it to import EXIF data, you lose the accuracy measurement. Only photos taken within iNat get an accuracy value. Which must mean accuracy can't be pulled from a photo's EXIF data or calculated - or they made some decision not to trust that value if the photo comes from an import.
My experience is on an iPhone; not sure if Android is different.
Unfortunately that is correct, when you use the photo's metadata, it does not record accuracy. I believe this is also the case with the android app.
Something we noticed in Vermont... if you import photos rather than using the app the location can be off in some cases. It seems to be related to this: if you activate the phone and quickly take a photo before the gps gets a lock on your position, there can be high error... but as mentioned the accuracy is not recorded by a photo unless you use the app. So the app is a better solution when it works but there are times I don't have time to use the app and snap photos, or decide i want to make an observation later... another solution is if you can just wait 10 or 20 seconds before taking the photo.
I have noticed the same issue. I usually snap 2-34 pics in the hopes that it will "home in" on my location and maybe, just maybe, one of them will be in focus...
My normal procedure when I am able to use the app is to load the app in the 'map' mode and watch the uncertainty circle until it zooms in enough to be good accuracy - usually just a few seconds.
Also, on the add observation screen you can watch the accuracy go down as the phone zeros in on the location; at least with the iPhone. Usually it is pretty precise by the time you finish adding the photo etc.
Add a Comment