Missing Location Accuracy data?

[stephen220]

Thanks for the input - I really appreciate it. Since this morning, I’ve noticed this same user has sent the same message to others over the years. Looks like you got one too! It’s even prompted other Forum discussions. I didn’t take it as abusive. But, as noted in comments of others above, it is misleading. It seems to encourage making up some numbers regarding location accuracy, which isn’t justifiable to me as a scientist. And some parts of the message are, at best, impolite - “There is no point in arguing that you used a GPS or that your data are always accurate”. Sounds like he’s not interested in a discussion and it’s his way or the high way

@Shauns

The same message does go on to say “Ideally we like observations to be 2-5m accuracy, but often one is uncertain or does not know quite where one was, in which case it might be 100s of metres. but please if you genuinely don’t know exactly where you were don’t make them too precise, even if it is to the nearest km or two.”

I understand and appreciate the desire to be accurate with location, but that’s not the issue he’s taking with my observations. I know where I was. I recorded the GPS coordinates. What I don’t know and don’t have data for are the accuracy of those GPS readings. With modern GPS receivers we can assume the accuracy is within certain limits, but I don’t have actual data for the GPS accuracy, i.e. I don’t have a number to put in the “location accuracy” field, which is what he’s asking for - none of the equipment I’ve used over the years records that information.

@Shauns

Said with that I have seen observations with a radius of hundreds or even thousands of kilometres for accuracy.

I’ve seen those too. I don’t know how those come about. I’ve not had that problem with any of the GPS systems I’ve used. And yes, I agree, those observations must be of limited usefulness.

[anon83178471]

@stephen220

large proportion of my >20,000 observations are useless

This is literally impossible, so long as you enjoy using the site and are not breaking any rules. I really don’t even believe there is a “useless” observation, unless somebody is using the site solely to break the rules and cause problems for other people. Please don’t ever feel that your contributions aren’t important or valid.

[tiwane]

I’m pretty sure the vast majority of my iNat observations have no location accuracy/precision recorded because I use the sync-with-track method you use, @stephen220. So I could either enter what I think is a correct accuracy value (but don’t know for sure), or just keep my locations without accuracy data.

I don’t know what external GPS device you use, but I’ve been using Garmin eTrex devices, and if you go to the “Satellites” screen on that device you’ll see an accuracy radius value. In my experience it usually reads 3-5 meters if I’m walking and not surrounded by canyons or trees. It’s larger if I’m in a moving vehicle or in a canyon or something.

For some people who are doing certain types of work then yes, maybe that accuracy value is necessary, as @charlie mentioned. But your observations are not “useless”, and the onus isn’t on you to add accuracy values. If you want to, sure, but first and foremost you should find iNat enjoyable for yourself. If you find it enjoyable to know that your observations are being used by others for research and conservation and that they need accuracy values for that to happen, then go ahead and add them (as long as they’re not arbitrary).

[deboas]

@stephen220

I know where I was. I recorded the GPS coordinates. What I don’t know and don’t have data for are the accuracy of those GPS readings.

If you wanted to add an accuracy value, just make sure it is large enough that you are quite confident that “where you were” is within the radius of the circle it describes - for example, with reference to features in the satellite image. But you’re under no obligation to add such a value.

@stephen220

With modern GPS receivers we can assume the accuracy is within certain limits

Not necessarily. If you’re in a valley, or under tree canopy, or there is heavy cloud, GPS coordinates can be highly inaccurate, especially when you first switch the equipment on and often for some time after that. Hence the usefulness of the accuracy field.

[stephen220]

Thanks @tiwane - that makes a lot of sense! The geotagging techniques I use these days (phone-camera Bluetooth connection and, most recently, in-camera GPS) do not show or record accuracy data, but looking at where my observations end up on the map, I can see the accuracy is within a few meters most of the time. Adding in some best guess of what the accuracy was doesn’t sit well with me.

In my own research, I appreciate there is a certain amount of error inherent in the data I use, and it’s my responsibility to take that into consideration when drawing conclusions from that data. I would think that researchers crowdsourcing data from community science platforms like iNat, where users are employing all manner of techniques to determine the locations of their observations, would appreciate that too

[stephen220]

Thanks, I appreciate the suggestions.

@deboas

If you’re in a valley, or under tree canopy, or there is heavy cloud, GPS coordinates can be highly inaccurate, especially when you first switch the equipment on and often for some time after that. Hence the usefulness of the accuracy field.

Yes! I’ve seen that happen sometimes, depending on the geography. Though I am amazed how well some of the latest devices cope with those issues, since they use not only the GPS satellites, but also GLONASS, Galileo and BeiDou combined for better accuracy.

[jdmore]

@stephen220

Adding in some best guess of what the accuracy was doesn’t sit well with me.

Given that device-supplied values (when available) are estimates based on satellite configuration and signal strength, I have no problem making my own (educated) estimate based on map/imagery position compared to where I think I actually was. I just measure the minimum radius that I am confident includes all possibilities for where I was. In some ways that is probably a better number than what the device supplies, though in most cases they align very well for me (3-5 meters for a 2009-vintage Garmin eTrex with WAAS enabled with good sky view). My phone-based observation positions have had similar accuracy when I work in airplane-mode and run a GPS app in the background to keep it active.

Of course, one does have to consider that a map or imagery alignment will also have some non-zero additive error, mostly unnoticeable for recent stuff. One gets a feel for that after checking several observation points that are easy to locate exactly on the imagery.

[ksanderson]

@stephen220

This seems to miss the point that the coordinates of each image were determined using a GPS-capable device. How accurate the GPS measurement was, I have no idea of knowing, as none of the GPS-capable devices I’ve used over the years seem to record that information. And surely there’s a certain level of error that’s inherent in all GPS measurements, however they are collected?

I also don’t understand why lacking “location accuracy” means the observation can’t be included in checklists for nature reserves etc. iNat Projects seem to have no problem including observations that lack accuracy data, provided the coordinates provided fall within the geographic limits of the project.

I recently wrote a blurb on this topic for my iNaturalist project, why the accuracy is missing for records, how to clean it up, etc. If you are interested you can see this here: https://inaturalist.ca/projects/mushrooms-of-pei/journal

As someone that runs a project, I do filter against accuracy. We are organizing the data in an atlas style format so in order for it to be included, we define an accuracy threshold so that we can confirm that it is most likely in the applicable atlas square. If I am doing any kind of review of the data in a GIS environment, say comparing against landcover types, I also set an accuracy threshold. There is all kinds of inherent error with GPS locations and data, so I for one do not have expectations of a super low accuracy rating when I set thresholds.

That said, observation data is observation data and there are plenty of uses where you would not need to worry about accuracy, so I wouldn’t think the data is useless, just in some situations it could be filtered out. I also assume the importance of the accuracy will vary by type of organism.

[ksanderson]

@stephen220

Adding in some best guess of what the accuracy was doesn’t sit well with me.

IMO you are over thinking the required level of detail of your accuracy, just err on the side of caution. If you feel like you are in the <10m range but don’t feel comfortable saying that, set it for 25m, 50m or 100m, whatever your are comfortable with. No one using this data is looking for absolutes, or they shouldnt be.

[stephen220]

@jdmore

I have no problem making my own (educated) estimate based on map/imagery position compared to where I think I actually was. I just measure the minimum radius that I am confident includes all possibilities for where I was. In some ways that is probably a better number than what the device supplies

Thanks @jdmore - I like this suggestion! This seems like the most scientific way of arriving at an accuracy estimate. I might start doing this going forward. I wouldn’t be comfortable applying it to old observations though, relying on my recollection of exactly where I was last month, last year etc

@ksanderson

I recently wrote a blurb on this topic for my iNaturalist project, why the accuracy is missing for records, how to clean it up, etc. If you are interested you can see this here: https://inaturalist.ca/projects/mushrooms-of-pei/journal

Thanks @ksanderson - that’s a very interesting and thoughtful read. And I do understand why accuracy measurements are important to you. Clearly using the app to make observations has advantages in terms of recording accuracy. Unfortunately I find myself almost never using the app, as a phone camera is not appropriate for most of the organisms I photograph (mammals, birds, macro subjects). Interchangeable lens cameras fitted with appropriate telephoto and macro lenses are the only ways to get identifiable images of many of these organisms, and the methods I have found for geotagging those images in real time (camera-phone Bluetooth connections, cameras with onboard GPS chipset) apparently do not record accuracy. Thanks again!

[ksanderson]

@stephen220

Thanks @ksanderson - that’s a very interesting and thoughtful read. And I do understand why accuracy measurements are important to you. Clearly using the app to make observations has advantages in terms of recording accuracy. Unfortunately I find myself almost never using the app, as a phone camera is not appropriate for most of the organisms I photograph (mammals, birds, macro subjects). Interchangeable lens cameras fitted with appropriate telephoto and macro lenses are the only ways to get identifiable images of many of these organisms, and the methods I have found for geotagging those images in real time (camera-phone Bluetooth connections, cameras with onboard GPS chipset) apparently do not record accuracy. Thanks again!

I never use the app, my phone geotags my images including accuracy, I use the browser to add all my observations. If your camera-phone bluetooth connection is using android, you need to make sure you have the ‘record accuracy’ or ‘high accuracy’ settings turned on, for it to record accuracy, otherwise it just adds the lat/long. Cant help with the onboard GPS chipset, but I would look for similar settings you may need to turn on to record accuracy. If you cant find a way you will just have to set it manually for every upload, or you could use a bulk exif editor to add accuracy to all your images for a value you are comfortable with.

[garrytre]

I’m a citizen scientist.

I can’t really comment about the culture of iNat and the uses to which it’s accumulated data is put.

I can comment on some frustration I have had with another, older, data set, and my use of it.

Recently I’ve wanted to find out if a certain lizard is restricted to above a certain elevation. And similar for a certain plant. And for two other plants, do they grow better, worse, or not at all on a certain type of rock.

For these, location, and lolcation accuracy, are critical. But a surprising number of records have arbitrary location accuracy of 250m, 1000m, 10000m. Some show no location accuracy. Some claim 10m accuracy, but other information in the record indicates 20km out. One with a transcription error claims 100m but is 700km out.

So for the research as I’ve indicated above, first I discard accuracy <250m, but how do I know the validity of the other claimed accuracy? And I pity the researcher who wrote a paper based on a certain plant growing on a salt alke, not noticing that the stated accuracy was 250km.

If I just need to know if something is present in South Australia, I’m pretty OK. But if I need to know something about habitat, I couldn’t use the OP’s observations.

So back to the iNat discussion. Given that most (?proportion) iNat obsrvations come with some sort of GPS data included, and transferred electronically, ie not transcribed on paper, I’d be happy to use them, even without loaction accuracy.

I’ve tested a few GPS’s that have an ‘accuracy’ readout. They aren’t as accurate as they claim, but within 10 or 20 m is good enough for me. For recording individual plants 3m apart, no.

For a different research project, not iNat, if I see an eagle in the distance, I happily write accuracy as 10km. But for an animal with a very restricted type of habitat, (a particular pile of rocks) I am very specific, to the extent of using ‘GPS averaging’.

[carrieseltzer]

@ksanderson

my phone geotags my images including accuracy

Can you describe the setup you have that does this? I’d love to be able to do this.

[ksanderson]

It differs a bit by device, might take a bit of trial and error.

Obviously your phone needs location services turned on.

You need to give the camera App location permissions, and it needs to have the Precise Location, Google Accuracy, High Accuracy, whatever its called on your specific device needs to be turned on.

For my iphone it is: settings → privacy → location services → select camera app → permission is given while using app, precise location turned on

For my android it is: settings → location → local services → turn on google location accuracy, settings → location → App Permissions → camera → Allow only while using, use precise location

After you think you have it set, take a picture, load into iNaturalist and see if the loader auto provides a position with accuracy. If it doesn’t, you missed a setting, when helping one person we needed to restart the phone.

Keep in mind, that this only adds the information to your images, the location it records has a lot of variability in its accuracy. A phone’s GPS takes time to position itself, more time than it takes if you just pull your camera out of your pocket and snap a picture. So if you take a picture quick the camera just goes with whatever location it has at that moment, which could be a cell tower, internet provider, etc. So, if you want consistent accuracy, you need to make sure your GPS is awake and positioned before taking a picture. Some folks will just open google maps and wait for it to locate itself, or use a GPS app of some kind and make sure it shows a location before taking a picture. My preferred method is to use the free Avenza App. Set up a default map that covers the area you typically hike in, then every time you go for a hike, turn on the app and start tracking, put the app in the background and off you go. The tracking will keep the GPS awake and almost always provide an accurate location to your images.

Hope that helps.

[fffffffff]

@ksanderson

turn on google location accuracy

How does it look on you phone? Could you please make a screenshot?

[ksanderson]

Unfortunately androids vary a lot from device to device and the settings are different, makes it hard to describe.

[tallastro]

Yes, Androids vary. My old phone did not save location information to the phone. My new phone does but not the accuracy. I also learned that the HEIF photo format does not seem to save location info or at least not save it to jpg when exporting. So I’m still using jpg for most photos.

Accuracy estimates from GPS can be tricky. Even if you have a value from your recreational grade unit, it’s calculated by a proprietary method. Now that I’m aware the accuracy isn’t getting into iNat, I’ll be adding it as part of my process. The values below seem like good conservative numbers based on my experience. The coordinates may often be better but they’ll almost never fall outside of these estimates.
Open sky GPS - 4m
Forested GPS - 10m
Device clock synced to GPS - add 4m

So… 14m will probably be my most used value.

Addendum after editing one record.
I noticed 2 things. First: when you add an Accuracy value to a phone GPS observation, it changes the SRC field from “GPS” to "Manual’. Hmm, may have to rethink adding values. Second: Pet peeve, coordinates are save to 6 decimal places. That puts it in the “pointing to a specific burner on the stove” range. (see https://xkcd.com/2170/)

[pisum]

@tallastro

Yes, Androids vary. My old phone did not save location information to the phone. My new phone does but not the accuracy.

does any Android phone’s camera app actually record an accuracy value or something equivalent to the photo file? others have said that the camera app in iPhones will capture a horizontal positioning error, but i’ve never heard of the equivalent in an Android camera app.

on my Android phone, the stock camera app definitely does not record an accuracy value, and the app for my Panasonic camera doesn’t record that either. the iNat Android app can record an accuracy value, but it stores it only in the observation, not in the metadata of the photo file itself (although the coordinates are stored in the file metadata).

@trscavo

I don’t know what you mean by “good”. Isn’t location accuracy a measure of “goodness”?

who knows exactly what a location accuracy value means for any given observation? is it supposed to represent a radius of uncertainty within which the observer is located? a radius of uncertainty within which the subject is located? the distance between the observer and the subject? radius within which a subject moved? radius within which the observer moved? is it the iNat app default accuracy value based on how far you zoomed in when you manually placed your observation on the map? is it a guesstimate of the error of the map used to place the coordinates? is it a guesstimate of the typical accuracy or error of your GPS device? is it an actual measurement from your GPS device? is it Android’s definition of horizontal accuracy (~68% chance that the device is within the radius)? is it Apple’s definition of horizontal accuracy (the radius of uncertainty for the location, where uncertainty threshold is not publicly defined, except that it is supposed that it is at least as accurate as Android’s definition)? is it something else or a mix of the above?

my point about capturing good locations is more about methodology. recorded locations – both coordinates and accuracy values – are only as good as the methods used to record them. bad method produce bad locations. good methods produce good locations. the presence or absence of an accuracy value tells you nothing about whether the methods used were good or bad. a low recorded accuracy value tells you nothing about whether the methods used were good or bad and whether or not you should really trust the values. you can look at this post to see what i mean: https://forum.inaturalist.org/t/location-accuracy-too-easily-bypassed/18547/31.

[whitneybrook]

I have a slightly different question, but I’m not sure if it deserves it’s own thread.

I have been monitoring casuals in some areas to watch for missing dates and accidental DQAs. I have noticed user(s) DQAing inaccurate location with a comment like “Location not precise enough for RG”. The accuracy on the most recent observation is 146.08 km (so a fairly large error).
Is there a location accuracy range requirement for Research Grade. If so, what is it?

If this true, Id like to come up with a comment that gives users the information they need to improve it.

Edit: fixed some grammatical mistakes for clarity

[cthawley]

There is no accuracy location requirement for RG. If you see users using the DQA on location accuracies that they personally feel are too high, please ask them (politely) to stop doing it, and I would also suggest voting against any DQA thumbs downses given for that reason if you are so inclined. That DQA field is for when an observation’s location pin is in the wrong place and its accuracy circle does not include the true location (like a observation with a photo where the Eiffel Tower is visible placed in NYC or an observation from the ocean placed in the middle of Kansas, etc.) .

I think it’s fine to leave a polite comment asking if OPs can improve the accuracy to ask them to do so - many users aren’t aware of this field or how it works.

If more discussion on this question/topic is warranted, I can split to its own thread.