look at your image files on your device, and you’ll likely see there is no accuracy or error in the metadata.
it’s possible that some Android devices do capture accuracy or positioning error, but as far as i’ve seen, most do not.
the iNat app is requesting a location from the Android OS’s internal location services. from what i remember, as soon as Android returns a location that has accuracy <= 10m, the iNat app takes that value and records it in the observation record.
Yes, this should be the case. As far as I know the iNat app doesn’t add accuracy data to the photos it saves, and I believe most Android devices do not record it with their default camera app.
it may be worth noting, too, that Android’s definition of positional accuracy is that the true location has a 68% probability of being within that number of meters from the recorded coordinates.
iOS’s positioning error is not defined publicly, as far as i know, except to say that their is as least as good as Android’s definition. this is why this should be expected at least some of the time:
note that this is very different from the recommendation from many people that observers make sure the accuracy circle will always include the true location when adding or adjusting the accuracy manually.
when people adjust accuracy values, they generally will do this in relation to some sort of map, but the accuracy of the maps / satellite images is not perfect, and generally you won’t know what their underlying error is.
there’s also no telling whether the coordinates represent the location of the subject or the observer’s device or something else entirely.
all that inconsistency is why recording arbitrary accuracy values just to satisfy some other user’s arbitrary requirements is totally unnecessary, as far as i’m concerned. it’s unlikely those other users really understand what any particular location / accuracy value actually represents anyway.
I see a lot of people mentioning using the satellite view to “fix” their lat/long, and just want to point out that Google’s “satellite” view (usually actually airplanes, but whatever) itself has ~2-20 m wiggle room in accuracy depending on where you are in the GIS projection model, the angle of where the airplane’s camera passed relative to the objects in view, etc. So really, I wouldn’t worry about accuracy levels below about 20 m unless you’re actually using a real GPS device.
Augh. That makes sense. What a frustrating extra detail to consider. I often locate my bird observations down to a single tree or post, but it never occurred to me that the trees might be misplaced!
Yeah. But on the bright side, this probably isn’t wasted effort, since odds are that most people will also visualize it on the same Google-based map that you used to fine-tune it! So whatever bias may exist should likely stay consistent, it only really matters relative to “reality”, or perhaps when Google updates their map layers.
And let’s face it, the bird (and most other animals, for that matter) likely won’t be in that exact spot again if another person tries to find it the next day. Even plants, which are mostly sessile, might have died or gone dormant if a person comes back to that exact spot ten years later, or even just a few months later. Accuracy is important, but speaking as someone who has often gone looking to re-locate state-listed organisms for work, the best possible GPS location and tiny accuracy circle doesn’t help you find the organism much of the time; it just gets you to the location where someone else saw the orchid, the turtle, the moth, whatever, some time ago.
It can be important to be precise even with birds if they are nesting and you want the observation to be useful when considering, e.g., cutting of a forest.
An example of the importance of accuracy is when the Swedish Forest Agency determines whether to allow logging or not. They only consider observations with an accuracy of 100 m or less. Else, they get too much data(!) which is also more difficult to assess because of the uncertainty, and they have very little time for each submitted case.
Similar thinking applies to things like planning for a new road or other land use cases. But at least Swedish authorities in general have access to all iNat records of research grade together with observations from various other sources.
That is a great example of when accuracy really matters! But most bird observations are not of a nest and, for most species of birds, the nest site does not remain exactly the same from year to year.
I routinely collect herbarium and museum specimens, and never include accuracy data along with the coordinates. I don’t know of anyone else that does either–at least, I’m not aware of this practice. Are we asking citizens to be more accurate/precise in their data recording than scientists themselves? Or am I just a sloppy scientist and everyone else is including accuracy on their specimen labels?
I have been asked a few times to add accuracy circles (by a state wildlife agency personnel) so that my record could be included within their dataset.
…at least until Google updates the picture with one taken from another viewpoint. :(
Ahh. You already covered that one. It is really tiresome that in OSM new imagery means entire villages can be suddenly in the wrong place except for some later additions or ones based on GPS instead. Without those the fix would be easy, but…
For what it’s worth, I do include accuracy data along with the coordinates I put on my specimen labels. Comes from my previous training as a NatureServe/Heritage botanist. It is a standard field in the Symbiota data model used by many digitized collections, and I imagine in most other data models serving similar purposes.
It is also part of the Darwin Core data standard used by GBIF.
When I asked about inaccurate borders (my observations near the border often are wrongly listed in Belgium), somebody immediately asked if he could get a hunting permit for foxes.
I was shocked. So people do use Inaturalist to kill animals. No thanks.
Double reason not to list locations very precisely, if the border is already 500m off in naturalist, there is really no point in nailing it down to 5m.
Also, how on earth am I to remember after a 3-4 hours tour where exactly I saw that dragonfly and find it back on a map?
I have contributed many pretty rare ones and the professor who IDs them never wanted to know a 100m precise location, he is more interested in the altitude.
The camera says WHEN you took a picture, not WHERE. And phone locations, at least at home in Costa Rica, are often anything but precise.
So far I have dedicated my efforts to decent pictures, not survey tools
Thank you everyone for very helpful inputs.
Just to clarify: the point is that, adding accuracy makes your observation more valuable in some situations, and because it’s so easy to bulk-add them on the web, I would recommend doing so. However if you do find the extra process there’s no harm not to do it.
Accuracy of location is one of the default columns in many natural science databases like GBIF, and it is usually included in specimen registry of museums / herbariums too.
Also FYI - There is a very helpful thread on geotagging your photos if you haven’t looked yet. https://forum.inaturalist.org/t/geotagging-photos/66
It’s important not to just add specific numbers uncritically, however. It’s good to check that the resulting circle does in fact include where the organism was observed.
Yes I agree. I always make sure that my accuracy radius include the actual location 100% using the landmarks on the satellite map, even when if I have taken it from a GPS device.
Sometimes radius of 1m can be achieved if it was on a landmark visible on the satellite map e.g. a specific tree, rock, or a corner of a garden.
It’s not something that should be done always, but in case of particularly interesting plant / lichen that only occurs there for example, it can be helpful for others to locate the exact thing again. It happened to me a few times.
I’m not sure that bulk adding fake accuracy values to my observations is a good idea – fake information seems worse than no information.
Fake information is definitely worse than no information, however adding a radius usually doesn’t result in fake information.
If you are always within areas with good GPS reception, I would think that bulk adding 50m or 100m radius for instance will make your observations (slightly) more valuable in terms of location accuracy for research purpose - compared to observations with absolutely no accuracy, which might be filtered out at the first place, just like @conorflynn have said.
But, after reading this thread, it seems like some people don’t really agree with this, while some does, which was great to learn.
I tend to double-check the recorded GPS anyway, because the satellite has told my camera that it was up to a klick away from where I know we both were!
Yes it happens to me sometimes, even when using proper GPS equipment. This is one of the reasons I think it’s good to check the satellite imagery and adding the radius.
Should I go back and add a figure for accuracy for those observations?
I wouldn’t think so because Olympus Tough cameras’ GPS sometimes accidentally records a very inaccurate coordinate. You can’t really go back to 55,000 observations and assess whether those coordinates are correct or not.
However, I would recommend checking the satellite imagery and bulk adding the radius in the future, as long as a bit of extra work doesn’t bother you. It definitely improve the value of those observations (although some argue that it’s not that much), and during the process you might be able to pick up errors you wouldn’t have noticed otherwise.
I’m curious if all the herbarium and museum specimen records of the world are also “of little use” and “sloppy recording” since they also generally provide GPS coordinates in their data without any radius of uncertainty specified.
I would disagree with this statement. Majority of herbarium and museum specimen labels or online registry do have a section for location accuracy, which I was told to be quite important by three museum curators from three different places in the past. It’s a standard field internationally, as @jdmore have mentioned.
When I was volunteering in the marine invertebrate section of the Auckland Museum in NZ, I was actually working on adding accuracy radius by interpreting old labels without GPS, that only says about a beach, for example - and the curator told me that’s one of the most important work. Many of the accuracy radiuses ended up becoming like 10km, but many scientific databases do require this data to be filled.
It’s not really a fatal flaw not to have it, but I think the value of the specimen do fall, although this depends on a lot on what it is and what purpose it is used for.
For example, in case of certain rare plants, you can only find them in a tiny area that the accuracy of its locality is crucial to for others to locate it again. But in case of birds, it really doesn’t matter - unless it’s a nest observation, just like @cynips have mentioned.
but in my case I’d be adding an arbitrary number, which doesn’t seem any better than just leaving it blank.
I personally don’t think so, because adding arbitrary numbers means that you at least assessed the accuracy of your GPX track recordings. As far as I know, leaving it blank never indicates that it’s been recorded precisely, especially when your observations are exported to other databases like GBIF. On iNat I suppose it doesn’t really matter because we all know that you are the best!
My personal opinion is that, except for applications where high precision is required, observations without an accuracy value are likely usable.
Yes, I agree that they are definitely usable. However, bulk adding accuracy values is not the hardest thing to do on the web and it helps in many situations as well, especially because it means the observer at least assessed the accuracy of their coordinates, while leaving it blank means they didn’t.
You are asking us to go back and add extra steps to our workflow. When I’m uploading from my mobile device, I don’t see a default field for accuracy circles. I have to go back and edit the observation after it has been uploaded.
That’s one of the reasons I don’t like the app. I use web 100% of the time.
Google’s “satellite” view (usually actually airplanes, but whatever) itself has ~2-20 m wiggle room in accuracy depending on where you are in the GIS projection model, the angle of where the airplane’s camera passed relative to the objects in view, etc. So really, I wouldn’t worry about accuracy levels below about 20 m unless you’re actually using a real GPS device.
Thank you so much for this input. I never realized that.
Do you think they move satellite imagery around when updating the map? The point of very precise accuracy below 20m is that you can relocate the individual (in particular, plant or lichen) easily, by comparing it with where the landmarks are like rocks or trees.
If such landmarks visible on google map’s imagery do move around 2-20m when the map is updated, I guess I’ll stop doing it.
At least, in my experiences in past five years, it haven’t really happened and I still see the same things on the same position.
I guess what I am hoping for is that our devices do this for us without anyone have to add precision values manually. For many of us, our phones already do that, and we just need to take a look every now and then to make sure that the precision values make sense. It’s a shame that many Android devices seem not to. I don’t think this should become an additional burden on observers, but I also think it’s useful for observers to understand that this information adds value to their records and to pay some attention to what precision values are being recorded. For those who find this burdensome or unnecessary, please feel free to ignore the issue, and your records will still be useful for many purposes.
Observations with a recorded (and fine) level of precision are more useful, for a wider range of applications, than those without. Even if there is variation in how that is generated or what it means, as @pisum has pointed out, that’s not a reason to discard the importance of recording precision. It would be like finding a few Research Grade observations that are wrongly identified - sure, this is not ideal, but it doesn’t mean that we should conclude that the identifications of all Research Grade observations are useless.
Museums and herbariums are hopelessly under-resourced, and have the legacy of collections made at a time when (as @jasonhernandez74 mentioned) it was considered fine to put “Brazil”, “Congo” or “Atlantic Ocean” as the locality. These days we can and should strive to do better when collecting data using smartphones, because it opens up ways of understanding biodiversity that were not available in the past.
A few quotes from papers on the topic:
“Explicit knowledge of uncertainty around the location of a specimen is needed to adequately approximate the relation of an organism to its habitat … Otherwise, erroneous inferences can lead to committing substantial errors”
Uncertainty matters: ascertaining where specimens in natural history collections come from and its implications for predicting species distributions
“There is always a degree of uncertainty that should be associated with any coordinate.”
Spatial data for fungal specimens: retrospective georeferencing and practical recommendations for mycologists
“SDMs [Species Distribution Models] can be developed to their full potential only when they are trained using many high-precision occurrence records for a species”
Why georeferencing matters: Introducing a practical protocol to prepare species occurrence records for spatial analysis
As I understand it, mapping apps use a simplistic model of the Earth (a projected coordinate system) to “drape” satellite or aerial imagery onto a flat plane so it can appear on your screen. This model does not capture the bulgy shape of the Earth entirely faithfully and this can result in some small degree of difference between the position of a landmark on the map and its true position. But the differences are usually so small as not to be worth worrying about.
This is not what happened in the thread you started.
In any case, you are always free to obscure observations (or not share them at all) if you feel that sharing the exact coordinates would put the organism at risk. For obscured observations, the accuracy radius of the location (which is a different issue than whether map borders are accurate) then becomes fairly irrelevant.
Not being able to find the same animal on another day, maybe another year, isn’t really the point, though that is certainly possible with plants. For wildlife conservation, the benefit of recording sightings accurately and precisely is to identify the habitat patch that the species is using. Multiple species records from the same habitat patch such as a wood or pond can add up to a case for protecting the site, maybe as a nature reserve or just alerting the landowner to avoid causing habitat damage. Giving the location as the nearest town may be enough for coarse-scale mapping of species distribution. But suppose a property developer wants to build on old pasture or fell a wood. You won’t stop his plans by saying certain rare species are found in the vicinity. You need to be able to say which rare species have been found in that pasture or in that wood.
Of course, doing the species recording doesn’t guarantee the records get taken into account. But if the data are not gathered at sufficiently fine resolution in the first place, they can’t be used for site protection.
is this some sort of actual regulation or law? or is this just some arbitrary personal rule or bureaucractic prioritization thing, such as:
…
so if you have multiple occurence records with no accuracy value recorded from roughly the same location, you’re saying that is of absolutely no value? but if someone adds an arbitrary accuracy value to two of these records, then magically, it’s valuable?
if i’m some sort of rogue activist, i could just go around adding fake multiple records with accuracy values, and i can stop development? nobody’s going to go and actually survey / groundtruth the land to check the data posted by randos on the internet?
…
this is totally not true. bulk adding accuracy values just means someone added an accuracy value. it doesn’t say anything about whether the observer assesssed their location or accuracy or whether they just added the same arbitrary value to all their observations.
here’s a real example. there’s a pond at a park in my area that has some Nymphoides plants. those plants are isolated to a single pond that is labeled as “Arboretum Lake” on Google Maps and the “Meadow Pond” on OSM.
Below are all the verifiable observations of Nymphoides in Memorial Park. note that there are observations clustered at the pond, but also at the Arboretum headquarters building, and a few others scattered throughout the park.
A lot of people mention that scientists who care about fine-scale accuracy will filter for observations with accuracy <= 50m. so let’s do that. note that you’ve fitlered out a few of the outliers, but you still have a cluster at Arboretum headquarters, and there are still a few one-offs.
when i look at these outliers and compare them to other observations their observers have made, half (2) appear to be cases where people added an arbitrary accuracy value , and the other half (2) appear to be cases where iOS determined a location with a very low accuracy value (but was very wrong).
now let’s look at our observations that had no accuracy recorded. there’s one at Arboretum headquarters, and there’s another one-off. did you really improve the quality of the data by excluding these observations?
if you don’t like that example, here’s another example of a patch of Alligator Flag isolated to another pond:
pick any example of an organism that you know exists at a particular location, and do the same examination of the observations in iNaturalist. does excluding observations with no accuracy help in any of these cases?
here’s another example i’ve provided in the past with a different sort of visualization: https://forum.inaturalist.org/t/location-accuracy-too-easily-bypassed/18547/31.
…
you would think that the people who are the most concerned about accuracy values would also be the ones most concerned about having and understanding of what those accuracy values actually mean.
if you’re just going to wave off inconsistencies in locations / accuracies, why not just wave off the lack of accuracy values, too? what’s the difference?
to me a better analogy to the inconsistencies in the locations / accuracies would be like allowing folks to identify their observations using any taxonomy they want without noting which framework they’re using. for example, i could have 3 different people observe the same plant. one of them could call it Hibiscus lasiocarpos, another could call it Hibiscus moscheutos ssp. lasiocarpos, and the last one could call it Hibiscus moscheutos, and they would all be right according to their view of how things should be classified. the issue isn’t that anyone is wrong. the issue is that they’re inconsistent. would you use that data as is without dealing with those inconsistencies?
