I am seeing many observations, many of which in urban areas, that display a very imprecise position. I mean with a radius often exceeding 1 km and sometimes being larger than 10 km.
I think that, on the one hand, there can be good reasons not to keep the gps switched on or to use different camera apps that are not allowed by the operating system to add the position in the exif data. Sometimes it is also possible that the gps malfunctions. Anyway, on the other hand, it would be desirable to have observations with a decently precise position.
So, how to suggest or encourage to ameliorate this aspect?
This could be particularly desirable for species that are frequently cultivated or to get a precise idea on which habitat the specie is growing in.
I agree that more precision (smaller accuracy values) would definitely be better/provide more useful data. One issue that I see specifically in cities/towns is that some observers just use the city/town name via a location search as the location. This gives an accuracy of varying (but often large) size based on Google results which attempt to encompass the whole location/city. Many of these are observations made sometime after the fact/not in the app, so the user may just be doing as good of a job as they can at remembering the location since they donāt have GPS coords for it. Or the user may not realize how accuracy works on iNat (to be fair, itās a bit obscure).
I think that some users, especially in the apple app, also just zoom and stick a pin and donāt refine further or know about setting accuracy manually. I see some observations from iOS users with an accuracy circle of 1,000 km and such.
In cases where records might be valuable, Iāve asked users if they can update/edit the accuracy, and had this happen a few times. But since there are different ways to edit this (depending on app/website), and it is a bit complex, I donāt always know how to give instructions, and donāt often get a response.
I use my iPhoneās GPS geotag for virtually all my observations. I have no idea how it calculates the accuracy bubble, but I do notice that it is wildly variable even when the pinpoint is accurate. I donāt often check the accuracy bubble when I submit an observation, so I seldom know what it is. Of course if someone asks me about a specific location I check it, and I shrink the bubble if thatās appropriate. But my point is that I donāt think most people are controlling the accuracy estimate when they submit.
There are also many obs where āAccuracy: Not recordedā is listed.
At least if some accuracy is recorded you know how much it can be trusted.
I donāt think we can fix this. A lot of people really donāt understand that they should define an accuracy or how. Many donāt care. I personally will add moderately large accuracy when Iāve wandered all over a park. And just wait until I get my photos from the 80ās uploaded; what I wrote on the slides often wasnāt very clear! Some of those will exceed 10 km and itās the best I can do.
The imprecise accuracies are frustrating when I want to relocate something, or judge whether a puzzling photo could have been taken in the area reported. I ran into an accuracy circle of over 1700 km recently. Not helpful.
It seems to me that we keep pointing out more things we want added to the on-boarding process for new users, forgetting that (1) most people wonāt remember them anyway and (2) added complications will discourage people from participating. Citizen science ā always messy.
i personally think we should go back to excluding anything over ~100 meters accuracy from research grade, or at least creating a better way to filter out those observations. I know this isnāt the majority opinion out there, but just throwing it out there because as an ecologist who does mapping and fine scale inventory, i have no use in my work for those low precision observations at all and not being able to filter them out (that i am aware of?) makes iNat a lot less useful to me professionally.
If your suggestion would ever become true I personally would be off of this plattform.
I rarely do use accuracy circles smaller then 100mā¦ because often I just donĀ“t know, when I am on a hike for 8 hours somewhere in the mountains without signal or snorkeling along a shoreline.
In some cases I donĀ“t want to be too specific to protect my own security or the one of the organisms I observed.
I do use accuracy circles of up to 2km without any shame regularly, if I just donĀ“t know better. I also use 500m+ if I want to obscure the location somewhatā¦ in some cases I did minimize them after I moved to a different town.
If you download iNaturalist observations via the API, each observation has a variety of metadata including accuracy data. This makes it easy to define your own version of āresearch gradeā after downloading the data.
This might be a bit off-topic, but is there any way to make a non-circular accuracy polygon? I have a lot of observations from before I joined iNat where I know the photos were somewhere along a road or trail between Points A and B. That is a fairly wide inaccuracy in one respect, but at the same time, the organism was clearly in the middle of a road thatās only 5 m wide. So a circle with the diameter AB actually includes a lot of ground that I know the organism was not in.
definite hardware limitations that are not really fixable.
I take ALL of my photos with my phoneās built-in camera app when I am out and about. I am almost entirely outside cellular range, so itās irritating and wasteful to pre-set a bunch of inat obs that are trying to upload and eating battery every time the phone gets a tiny whiff of a signal for a few minutes.
I upload observations after I return home when I have a reliable data connection. sometimes, I find that my photos did not record sufficient metadata. sometimes not even a date or time. so for those, I have to estimate the location the best I can. when the observation was deep in the woods, far from identifiable landmarks, the accuracy circle can get pretty big. sometimes the phone does get a location, but the accuracy of said location isnāt great, and being in the middle of the woods, canāt really be refined manually.
if there were a bunch of restrictions placed on these kinds of observations, Iād probably leave inat, too.
No, accuracy can only be expressed via a radius. You can always use the description to say āseen along trail Xā if you like.
One thing you can do is keep a tracking or hiking app going on in the background - anything thatās getting your location on a regular basis. It really helps you get accurate location info when you take photos because your device is not starting from scratch when you open the camera.
often I will start GPS tracking of my position using another phone application in the background, like eBird (two citizen science platforms at once!), which narrows in its wake the precision/accuracy on iNaturalist to 4m in almost all cases.
I find the attachment of coordinates directly to the image via the iNat photo-taking process to be too valuable to give up (Iām sure there is a way to do this with the default camera app, but I donāt care to have coordinates attached to every one of my non-organism photos.) however, what youāre describing sounds like the option for āauto-upload observationsā is turned on in your app. Iāve kept it switched off since I found out about it, since I also almost never want to drain the battery immediately sending things out over cell signal.
Yes, it is possible.
For folks who are interested in ways to add accurate GPS locations to observations taken out of cell range or outside of the app, there are lots of proposed workflows in other threads. This tutorial:
https://forum.inaturalist.org/t/geotagging-photos/66
is a great resource, but searching also turns up many others.
Iāve found that even within cell range, a GPS tracker in the background is necessary for reasonably correct location data (I use GPS Logger, the Android app).
The way phones acquire location within cell range (the cell tower + wifi networks method) is often wildly wrong unless Iām in a very dense urban or suburban environment. Iām talking up to 2+ kilometers across a coastal inlet away from where I really was, if my phone is using cell towers and wifi networks to estimate my location. Even within dense urban areas it can often be a whole city block off. I havenāt checked what it logs as the measured accuracy of those positions; hopefully it recognizes how inaccurate they are, at least.
I donāt mean you should feel shame or feel like you shouldnāt observe things with less location precision. I do it too. But those observations are not useful for a lot of what i do and thereās no way to filter them out, which is frustrating.
Yes, and i do this sometimes, but for the generalist type work i do, iād have to download a huge amount of stuff, and it isnāt really feasible.
Definitely turn off auto-sync/auto upload in all cases, itās really buggy and breaks the app. Unfortunately there is a weird bug on the iPhone where auto sync turns itself back on, so you have to keep making sure to turn it off any time it tries to activate.
I recognize most people have different views on geoprivacy than me and i donāt want to alarm anyone as the chance iNat would enact my preferred restriction is about 0%. I do think ability on the website to filter to better than a certain uncertainty distance would be incredibly powerful, but maybe itās a difficult or computation intensive query or maybe iām just the only one who cares about it. I am not sharing this stuff to make anyone feel like their low-precision observations arenāt worth uploading, iām more trying to encourage those who do find ways to share accurate and precise locations because they are really valuable.
I mean, sometimes when hiking in the bottom of a canyon 100m accuracy from the phone GPS is unrealistic. Though, if the uncertainty ellipse is actually more like 10m X 1 km I guess the uncertain area would still be about the same?
Understood! 
Yes, a lot of frustrations around the use of iNat come down difficulties (or impossibilities) in setting the right filters. I get that.
If it is an organism that is mobile, you donāt really need a 5m accuracy.
You probably donāt need 5 m accuracy, but if the full accuracy circle with diameter AB includes both the wetland on one side of the road and the upslope forest on the other side, that could be problematic for people who are trying to figure out habitat use or other fine-scale questions