I think a lot has to do with the volume of things being uploaded, more than personal weakness. For example, I use iNat to keep track of all the species at all the local parks and their phenologies. That means I’ve got a “just for the record” shot of all the common weeds, invasive bugs, etc., at all the different locations. For some with different appearances throughout the year, I’ll snap a quick record every few weeks at each place to get the phenology recorded. At this point I’ve recorded well over 100 observations of Spotted Lanternfly, which is ubiquitous and can be seen virtually everywhere. The idea that you could pull one of my 5-year-old shots of a pile of lanternflies and expect me to know which of my 100,000 lanternfly experiences that photo is from is ridiculous. Of course I need to rely on the automated GPS coordinates to know where I was. Sure, I can find on a map where I was yesterday, or which trails I tend to travel at each park. But when you say
that sounds to me like your use case for iNat is entirely different. Once you’ve submitted your 60th observation of the same leaf mine, or your 100th voucher observation of the same plant to link to the insects you found on it, it gets to the point where you’d need Rain Man powers to be shown one of those images and be able to say “that was the lanternfly from August 2021 when I was 10 feet down the road from the big hickory tree at the Preserve”. A user posting 100s of their best photos ever and a user posting 10,000s of mediocre-but-identifiable photos to track data have entirely different goals. No better or worse, just different. And one of those workflows just allows a person to keep track of all their observation locations in their head a lot better than the other. There’s even one user who’s observed over 45,000 of the same plant species in the southwest USA. If they can’t show you on a map where each one of those was observed from memory, I would hesitate to chalk that up to having been
To be fair, people make observations in very different ways. There are folks who blaze away - taking photos of all kinds of organisms (plants, bugs, birds, etc.). They might also want to report the locations of these observations with extreme precision - like down to a few meters (perhaps they’ve been fed some line like “observations are useless if the location isn’t precise down to a 10 meters or less”).
If I was taking lots of photos of all kinds of organisms (some of which I may not be very familiar with), I might not remember exactly where along the trail I took some of them. Most of the time, I’m not interested in that level of precision, so it’s not a problem for me. But if someone is convinced that extreme precision is necessary, or even desirable, then they may have to rely on some form of GPS technology for that. But that’s founded on a big assumption - that the GPS is always recording accurate lat/long and that the uncertainty estimate is accurate. I think the original poster is overestimating the reliability of the GPS capabilities in most smartphones, and their interactions with the iNaturalist apps. For example: I’ve seen cases where the iNaturalist app records a completely incorrect location for an observation because it recorded the location where the app re-connected with the internet rather than the location where the photo was originally taken.
I identify a lot of butterfly observations. Many of these observations are made with phone cameras. These photos often leave a lot to be desired - some are just blurry blobs. Because our region has a reasonably limited number of butterfly species, it’s often possible to identify these blobs to species if you know the species that are likely to be found at the particular date/location where the photo was taken. But for some groups with multiple look-alike species, you need a clear closeup photo if you want to have any hope of identifying the butterfly in question. In the tropics, where there are far more species, and there’s possibly less detailed knowledge of which species fly where/when, the need for clear photos goes way up. If, as suggested by the original poster, everyone abandoned their cameras in favour of smart phones, the accuracy/reliability of the location data might go up, but it would be pointless since you wouldn’t be able to identify many of the organisms photographed. But as I outlined in a previous post, that doesn’t mean that one can’t use some combination of camera, smartphone, handheld GPS unit, etc. There’s more than one way to skin the cat. Does this mean that the observer might mess things up and get the location data wrong? Sure. No system is foolproof. Fools are far more creative than the folks who design the systems they (mis)use.
This approach seems to have a “noisy” built-in error, since it just uses an approximate centroid for the whole route. This might be okay for people who only walk a few hundred metres whilst recording, but it’s much less so for someone walking several kilometres, and totally impractical for anyone (like me) who regularly walks from ten to thirty kilometres in a day.
I used to use landscape/beacon photos in the manner suggested above by jeanelz - which is certainly workable, but I found it to be far too much of a distraction when visiting locations I wasn’t already familiar with. So I decided to invest in a hand-held GPS tracker, which has certainly proved to be worth every penny.
One of the big advantages of a GPS tracker is that it continually saves data points throughout the route. This makes it possible to eliminate any weird outliers and fill in the gaps using a centroid based on a few nearby points (which will rarely be more than about fifteen metres away). It’s also very easy to visually sanity-check the whole dataset in a GPS viewer before linking it to the photos.
I have never owned a smart-phone, but I assume it would be impractical to use it as a GPS tracker, due to the continual drain on the battery. This also seems to imply that the overall accuracy may be a little more hit and miss, since the device must rely on whatever GPS data is available at the moment the photo is taken (i.e. there’s no error-correction via back-tracking).
For me, having a permanent backup of every route I have walked over several decades is invaluable. It means I can (re-)process my photos many years after the event without ever needing to rely on my gradually fading powers of recollection.
True. If you feel that extreme precision is necessary or desirable, then this isn’t the method for you.
My observations don’t go into iNat, and in fact, I usually don’t take photos of what I’m observing these days. I’m mostly just jotting down the species (or adding tick marks) on a pad as I walk. Having more photos of most of these species is a burden I don’t need. Same goes for GPS traces. I already spend too much time on my computer identifying other people’s observations, I don’t need to spend even more time managing photos, GPS traces, etc. I barely have time to manage the few photos I do take.
I have a hand held GPS device, but I only use it in circumstances where there’s a need for me to precisely retrace my steps to extricate myself from tricky habitats ( like bogs ). I rarely carry it. My focus is often on finding extremely rare target species (and not missing serendipitous rarities). Everyone has different priorities.
As far as I’m concerned, having a precision of a few hundred meters, or even a few kilometers is usually good enough (especially because the lat/long combined with the uncertainty is always accurate - the observation is ALWAYS inside the uncertainty circle, which cannot be said for observations on iNat). If someone else wants to record their observations with greater precision, more power to them (as long as accuracy is maintaned).
I will break up longer hikes into segments, especially if there are major habitat discontinuities, or if my route traverses Atlas square boundaries. But I report each segment based on a centroid and an uncertainty radius that encompasses the segment.
Using the iNat app: This can only happen if you take the fotos without coordinates, and then create new iNat records using the pre-recorded fotos while sitting at the wrong place. E.g. on the parking space. This is much more likely the result of somebody not using the app, but entering locations via names on the computer.
I am using the Android app. And it is really a great tool for good location data. Once you made sure the location got properly recorded (I agree that is an error source i forgot to mention), it comes with relatively reliable geocoordinates. It doesn’t make the transcription errors one finds so often in geocoordinates of museum specimen, and it doesn’t loose or switch lables either. You should really give it a try someday.
In case you fancy rare plants, this article about a species that was not scientifically reported for about 90 years might be of interest
The main reason it wasn’t found for 90 years was: The name of the type locality was wrongly recorded on the herbarium sheet, hinting to a different place less than 1 km from the real type locality. So yes, insisting on proper location data does make sense in this kind of environment.
I can’t say as they weren’t my observations. It may have been earlier (buggier) versions of the App. I’ve been doing ID work on iNat for around a decade or so. I just know that I’ve seen some strange locations, and when I asked the observer, they said something like “oh, I probably didn’t have a data connection in the field, and the app uploaded my observations when I pulled into my driveway”, or something along those lines. I’ve seen several variations on this over the years. I don’t try to keep track because I don’t use the phone apps. I’m sure there are other scenarios, like where the observer photographs an image on their computer screen using their smartphone. I see a lot of those, where the location pin is squarely on their house in suburbia.
Yeah no. As I explained earlier, a cell phone is a poor choice of camera for the things I report, and mostly, I can’t be bothered to take a photo at all. I have no time to enter my observations into iNat in any case.
But I think we’re talking about different things here. If a specimen just has a county or state listed as the location, it’s probably correct in most cases. It’s not very precise, but it’s accurate. The observation actually took place within the specified county/state.
When folk type in the name of a place, and simply use the default lat/long/uncertainty without making any adjustments, there’s a good chance that the result will be fairly precise, but inaccurate (ie. the true location does not fall within the uncertainty boundary). In some cases, the default lat/long/uncertainty default might be reasonable, and may encompass the area where most observations take place. But I’ve seen a number of examples where this simply isn’t the case. If you’re submitting an observation using the web interface and you enter “Pinery Provincial Park” as a location, the default lat/long that gets filled in by iNat coincides with the spot where the park name is displayed on google maps. The default uncertainty is just over 185m. That’s fairly precise, at least by my standards. This spot is on the main road near the park entrance. There may be a few people who actually make observations within this circle, but for the vast majority of the observations I see with this location, I can be fairly confident that the location is inaccurate. The default should be about 6km to cover the whole park.
I’d rather have an accurate location that is imprecise than a very precise location that is inaccurate. Whenever I see extreme precision specified on an iNat observation, I’m skeptical. Your GPS device may have overestimated the precision of the lat/long it has captured. Even if you’ve looked at the satellite imagery and pinpointed a landmark you recognize, you have to remember that the satellite imagery isn’t necessarily lined up accurately. Switch to a different satellite image and chances are your landmark will have moved.
I think this is something where you have to find a solution on your own. Personally, I was thinking I would put together a list of frequently referenced locations with their default lat/long/uncertainty values. I will then write some code to sort through downloaded observations and check if they use those default lat/long/uncertainty values. If a match is detected, the code will take some action (such as discarding the observation, or replacing the lat/long/uncertainty with more appropriate values).
The problem I have with most discussions on accuracy is the fact that your GPS does not locate the subject! The GPS in your phone or Garmin etc calculates your position - it has no way of knowing which direction you were facing or distance from you to the subject.
I try to imagine a scenario in which the difference between the location of the organism and the observer has any relevance.
Most herb, insect and other small animals observations are from nearby anyway. So larger distances between observer and organism seem to be just a problem with trees, mammals and birds.
In the case of mammals and birds a very small accuracy might not really contribute to reproducibility as these organisms are mobile to begin with.
And the information that the organism was observed in a distance should in most cases be obvious from the fotos.
So under which circumstances could that difference become relevant?
If it isn’t relevant - why worry about accuracy circles at all? I have spent a fair amount of time looking through my own images and you might be surprised by how many subjects don’t land within the accuracy circle.
Well, I think it’s an issue if for instance a fairly sedentary organism has a pin dozens of km away from it’s actual spot and the accuracy radius is only a few meters, but that’s not gonna be caused by a telephoto lens alone. In my experience the only things that are often photographed from more than 50ish meters away are large vertebrates and these are cases where extreme precision is unlikely to be of much use to researchers.
In my experience, accuracy circles have limited value for mammals, birds, large trees etc. My main point is that MANY people assume that those pins in iNaturalist represent actual subject locations - they don’t. How you interpret or use iNaturalist data should at least consider that. The other issue of course is that there is no consistency - some users manually locate observations, while others may never move or manipulate them and others may.
I personally don’t pay much attention to accuracy circles, not all that valuable to me. Everyone is free to disagree.
I think the fact that many birds and large mammals were probably a few meters outside of their corresponding circles doesn’t mean accuracy circles are useless in general or that people shouldn’t try to at least not make them much more precise or much less precise than they really are.
My android phone has only so so location accuracy and doesn’t produce error values (not uncommon) - so for me, moving the location is far more valuable than adding an accuracy value.
Also, we have a prolific user who uploads all his observations with huge accuracy circles - not obscured, but of very limited value.
In the article I linked above, a plant species was not refound for 90 years because the herbarium sheets of the holotype had a place name that was wrong. Distance between the place named on the herbarium sheet and the true type locality was about 1km.
So yes with smaller sedentary organisms a 1 km offset can make observations very difficult to reproduce. And an inflated 50m accuracy circle can certainly still lead to long searches.
When starting out with iNat, i also thought that it is important that the true location of the wild occurence is within the accuracy circle. I have for example made fish observations on fish-markets, and manually increased accuracy to include suspected fishing grounds.
But now i think reproducibility is best served when the GPS data recorded are not manipulated at all (unless they were obvious wrongly recorded). E.g. with the original fish geocoordinates on the market, one is at least able to find the fish market. While the inflated accuracy circles, that might (or might not) include the fishing grounds, are useless for locating the market and for any modeling.
Similar with distant bird observations in a bog. With the true coordinates of the observer one can at least be certain that there is a path or observation point. While the real coordinates of the bird on the tree in the bog might bring people trying to reach the exact coordinates into dangerous situations.
Yes that certainly is a problem. Strangely i didn’t find any official information on best practise until now.
I completely agree. That degree of precision is not something that I have any interest in recording. As long as something was within a few hundred yards, that’s about as precise as I would ever need or want the location to be. Of course, other may have different ideas about precision, but those ideas are often not founded in actual real-world usefulness.
what a great insight! Hilarious! But unfortunately, all too true.
