does a large “accuracy” value associated with any given location mean the location is imprecise or inaccurate (or both)? does no “accuracy” value mean a location is imprecise or inaccurate? does a small value mean that a location is precise or accurate?
the answer is that you simply can’t tell in most cases without some additional information which often is not available in the observation because observers rarely record how the value was determined or exactly what they intend for the value to mean for any given observation.
people who suppose the “accuracy” value has one standard meaning in the system are just plain wrong. you could ask folks to adopt a particular methodology for recording coordinates+“accuracy”, but if you don’t also ask them to record what that methodology is, is the “accuracy” value really all that valuable in the grand scheme of things?
i’m thinking that in most cases, unless you can tell that a particular observer seems to be consistently messing up locations, there’s really no sense in asking them to change their existing methodology to do it some other way that you happen to like more. it’s likely that their way is no less valid than your way, and you’re just going to confuse folks even more.
A lot of observations I make in my home, which I kinda prefer to obscure. Also I always keep on forgetting to turn on geotagging when I’m out with my camera these days, but I know the general area of where I made the observations.
If observers are doing their best to follow iNat’s operational definition of the value (“Try to make the circle big enough that you are sure you were somewhere inside it.”), then their particular methods for arriving at that value don’t make much difference to me. I already recognize that some will be more successful than others in following that definition, for a plethora of different reasons, and I already consider this when assessing the usefulness of a particular data set.
In the old days, before GPS, we would define locations based on miles or km from a reference point like a town. Extracting lat-long or UTMs from paper maps was a chore and so were not often used. Those old records based on specimens still have value despite the less precise locality data. It’s great if you want to nail down your location within a few meters, but for many purposes that level of precision is not really needed and is sometimes not desirable. Personally I’m fine with whatever level of precision an observer wants to use … within reason of course. A circle around an entire state is not particularly useful.
It’s your right but just obscure the position, avoid to add a very imprecise one.
It’s not a “data users vs data uploaders” matter. As far as I am concerned, t’s mainly to make iNat as far as possible a depiction of reality.
Yes, and it is undoubtedly a very precious resource. But also users who check things and try to make the whole thing work well here are mainly doing it for free.
I think it depends if the radius has been intentionally (in many cases for good reasons) enlarged from a center that has been obtained with the gps (when the gps works well) or, differently, if also the center has been put manually without caring of the actual position and also with a very large radius.
To be honest, it is not unfrequent that users answer positively when asked to refine the position. Notwithstanding, there are also users who are totaly unresponsive.
Apart this, if certain users would provide a more precise position by their own and without any recommendation, it would be a saving of time.
i don’t know if you could call that an operational definition. the only place i know that this exists is as the tooltip text on the web edit screen when you hover over the accuracy field. as far as i know, that guidance is not on the web upload screen. when i hover there, all i see is “Accuracy (meters)”. it’s not on the Android app.
moreover, folks who know that that text exists are often interpreting the text you quoted to mean the location of the subject of the observation rather than what it literally says – the location of the observer. folks may claim that, well, observer and subject are interchangeable most of the time, and, well, the text is probably just wrong. but how can you both say that that’s the instruction to follow and not the instruction to follow?
all i know is that there are many ways that people actually input locations, and even if they use some mechanism within the system to do so, they don’t end up with a location that necessarily matches that text that you quoted.
for example:
the size of the circle when you click on the map in the Android app is determined by how far you zoom in on the map.
the size of the accuracy if you use the GPS function in the Android app is based on Android’s definition of positional accuracy, which defines that the device is only 68% likely to be within that circle.
if you set your location based on typing the name of a place in the web upload screen, you end up with a circle that goes around the box that Google Maps has defined as the bounding box of the place – which often is too big but could also be way too small to contain the observer, depending on the definition of the place in GMaps.
For the work I do, I would prefer that folks take this approach to obscuring data within iNat, over obscuring it. And I have suggested this approach a number of times. YMMV
Yes and no. The trouble with observations based on reference points is that sometimes the reference point names fall out of use, or end up applied to a different location at a later date. I’ve had to deal with this exact phenomenon. I know of specimens that are now in a museum where the location is a wild guess because the original collector used the name of a place that nobody alive today recognizes. In another case, I was able to make a correction because as a “local”, I remember the old name for a particular location that now goes by a different name, and now something-very-like-the-old-name has been applied to a different location. I’d much prefer to deal with an unambiguous set of Lat/Long coordinates with a generous accuracy circle drawn around them.
Users should be aware, though, that in many circumstances this method makes it easier to discover the true location, compared with iNat’s standard obscuration method. (I won’t be sharing why, so don’t ask…)
For what its worth, when I’m out with my DSLR I’ll usually take a pic of my observation (if I can) with my smart phone as well, just to log the position. Even if I don’t do it for every observation, even doing it every couple can help pinpoint locations since I’m usually walking along trails.
The location of a biological record should include a place name and a map reference so that they can corroborate each other. The trouble with relying on map reference alone is that a wrong map reference is still a place on the map and there may be no clue that it is wrong unless it is something like a tree recorded in the sea.
I agree, and that’s what we get with data from other platforms. It’s also what we ask for in direct submissions. With iNaturalist, most users don’t even realize they CAN enter a meaningful location descriptor, or that they can manually edit the location coordinates. In an ideal world, iNat users would enter a meaningful location descriptor that matches up with the location coordinates, but I don’t have much hope of that becoming a widespread practice. I’ve tried to suggest “best practices” in our project journal: https://www.inaturalist.org/projects/ontario-butterfly-atlas/journal/60560-how-to-change-the-place-location-name-associated-with-an-observation
But I don’t expect that many users have read that.
Perhaps I should have clarified that I generally don’t pinpoint my own observations (down to the meter). I don’t have time to enter my own observations into iNaturalist (too much time spent vetting other people’s observations). My observations are recorded in a spreadsheet, and go into our project database more directly. My DSLR doesn’t have GPS capability, but I make notes of the times when I’m at different locations so that I can correlate photos to different locations. If I walk a trail during a survey, I simply find it on satellite view and find a suitable “midpoint” that will serve as a location for all the observations I made during my survey. Then I pick an accuracy radius that covers the whole survey, and that’s the data I enter for all my observations. Accuracy figures on my observations are typically in the hundreds of meters, and sometimes several kilometers. If I am on a particularly long hike, or one that traverses particularly divergent habitats, I might break things down a bit, but generally there’s one set of coordinates and accuracy for each survey.
Ask the observers to add the information you are after.
iNat works well promoting appreciation of biodiversity. Accepting low quality photos and locations without validation is part of that.
The more people are involved the more public support is for the type of research that might rely on the collected data. What is more important, high participation or quality observations with low participation?
Observations in Kirstenbosch Botanical Garden default to the suburb of Wynberg.
I have to zoom in anyway to sort out Wild or Cultivated, but location notes could be first aid.
There are thousands of them. Are you suggesting I message each one individually, and then provide custom coaching on how to do this (different people enter their observation data in different ways)? I already spend several hours per day providing identifications. When I can, I try to nudge people in the right direction, and I’ve compiled my suggestions into journal posts for our project. There are only so many hours in the day.
FWIW, as somebody who obscures everything (I’m rooting for an option to obscure only the date, since my current system sometimes feels like overkill, but better safe than sorry.), I’ve had countless people message me asking about data for different reasons. And usually I am able to help them use the data. Sometimes that’s by giving the exact location and other times it wasn’t. It’s another step for them, which isn’t ideal, but the door is always open for those who care enough to take that step.
One of my favorite instances of this was when I helped with a bat survey in Campeche. That time I did give exact coordinates but I was able to give a lot of information that I didn’t know would be so useful to somebody. In fact, I gave even more information than what would be readily available on an un-obscured observation since we started conversing about things, after I was messaged just about the un-obscured location part of it.
And that makes sense where one is considering individual observations. For certain special cases, it may be perfectly reasonable for a researcher to contact an observer for details on individual observations.
But for someone like me who is aggregating and processing 10’s of thousands of observations each year, it’s simply not feasible to message observers to discuss individual observations. Even if it was practical, it would be pointless, since “using” their data would entail revealing location and date information on our project’s website. For a while, we tied ourselves in knots trying to approximate the level of obscuration provided by iNat on our website, but that has now become a moving target, so the safer/simpler option was to simply exclude obscured observations from our database.
To be frank, I don’t think our project missing very much. I look over the obscured observations that we exclude from our database, and most of them are of common species found in areas where they’ve been reported by other people dozens of times. Very occasionally, I note that we’ve had to exclude an interesting observation, but that’s just the way things go. So, I don’t argue the point with anyone. They can obscure their observations if they want to, and it doesn’t matter to me how iNat obscures them. Obscure the location, obscure the date, whatever. Effectively, these observations have ceased to exist from my point of view.
For the observations that are not obscured, I’m fairly tolerant - I don’t require pinpoint accuracy for including them in our database. For some of our historical observations, we’re lucky if our recorded locations are within 20 km of the true locations.
You win some, you lose some, I guess. There’s more un-obscured than obscured and by taking the un-obscured you are winning far more than what you’re losing. You always have the ability to go on a case by case basis and reach out when you feel inclined to do so.