Don't let an observation attain Research Grade if its location is very imprecise

#1

For plants: If the accuracy is > 500 meters, even better >100 meters remove Research Grade.
In other words if the plants is impossible to check in the field it should not get a Research Grade.
We often follow up plants to be sure of the ID and > 100 is almost impossible for a tiny plant.

Now I have to manually set location inaccurate e.g. 13 km off, this could be done automatically.

0 Likes

#2

Are you saying you set “location inaccurate” for large accuracy circles?

0 Likes

#3

Not much a plant guy but this sounds quite logical. I agree. My phones gps is usually accurate to 50 to a 100 ft.

0 Likes

#4

Edited the topic title to make it more explicit. I will say this was a conversation in the Google Group many years ago.

0 Likes

#5

I’m not sure it is appropriate to be clicking that a location is inaccurate because of an arbitrary distance buffer that you have chosen has been entered. I don’t think that is the spirit of the flag. Imprecise and inaccurate are not the same thing.

9 Likes

#6

we used to do this,but we stopped. There was a long discussion about it. my opinion is they should not be research grade or at least not on any range maps since that data is very problematic and deceptive at least in the case of plants. I guess the idea was because a few animals like albatrosses might be still valuable at that ‘accuracy’ all of the observations should be research grade, even the plants which don’t move (i’m sure the people who wanted it that way would portray it differently, though). At least we now have some filters to exclude this largely unusable data, and I believe they are excluded from some range maps after a point, which is a good thing. The exact current policy I am not sure of.

I get what you are saying but we get things like alpine plants with the pin 50 miles away from the nearest mountain area in a desert, ponderosa pines showing up in the middle of Lake Tahoe - an absolutely massive freshwater lake - and the more deceptive data points that are in the wrong place when it isn’t immediately obvious. You could argue that the circle is accurate but not precise, but the point itself… is not accurate… and the point alone, not the circle, shows up on the species maps.

2 Likes

#7

I wholly agree for plants, but for animals I would like to think the research grade should be kept since they can move around. I have some very large accuracy bubbles in Kenya (5000-7000m) because I took all pictures without a GPS and had to “guess” where they were in a general area (I genuinely got all turned around on all the unmarked paths we took transects on). Additionally I will have large accuracy bubbles for insects I find in my bug sweep samples because they are collected on private land and I do not know the specific location. I still feel like this data with animals is still valuable because if they will be searched for again a general area of 5000m is definitely enough to work with. I guess if there were to be large ranges excluded it would have to depend if the organism can move or not.

3 Likes

#8

This is inaccurate if the uncertainty buffer is 5m or 50000m. It’s not related to the size of the buffer.

What’s to stop me from applying a personal standard of 5 meters, for flagging things as wrong. And what do you do with all the records where no buffer is entered ?

1 Like

#9

the buffer was touching the edge of the lake on both sides because it was huge. There are ponderosas on the lake shore so technically that wasnt inaccurate.

make it 10 and i’m fine with it :)

0 Likes

#10

Before starting the discussion again, I’d like to refer to what has been said before regarding this topic on the Google group:
https://groups.google.com/d/msg/inaturalist/-JR7FgQWALQ/49ymGRk7AgAJ

I still think that accuracy should not be used to automatically exclude observations from becoming RG.

What should be displayed on maps is a different issue, but I believe there’s now a filter excluding observations with low spatial accuracy from being displayed on iNat maps.

5 Likes

#11

On the one hand it could be logical to exclude from research grade observations that feature an “imprecise” position, while, on the other hand, for some protected and/or endangered species it is preferable to provide an imprecise position or even an obscured one. This criterion could apply also to observations that are still unpublished for which a user could prefer not to provide a precise position before their publication.

0 Likes

#12

yeah, while i still think they should not be RG, at least for plants, I am totally fine with this as a compromise. The imprecise obs can still get RG but won’t show up on the species maps. That being said I don’t know what the cutoff is and i think it could be ‘tightened’ a little. I think doing that plus making the obscured observations toggleable, would make the awesome species maps even better.

1 Like

#13

I agree that really imprecise locations should be excluded from range maps as they can misrepresent the species range pretty easily.

However, locations with large uncertainty buffers can still be useful. For instance, many organisms’ distributions are tracked at the county level in the US (and other countries which might call “county” something else). A location with a large buffer is often still placed wholly within a county.

Having a large buffer doesn’t make a location imprecise. It does quantify the uncertainty for anyone who wants to use the data, so I don’t see any issues here. One real problem is with points that have no buffer. Even military grade GPS has a little bit of uncertainty! When I use data, I exclude points that have no buffer as it is likely the person uploading isn’t being very detail-oriented with their data.

Lastly, I don’t think it’s a good solution to have folks implementing their personal preferences for a buffer distance that would “disqualify” observations from being research grade by flagging the location as imprecise. That means one person is deciding whether data is useful for all other users. We have no idea what level of precision the next person may need.

If there’s a standard to be set, it should be implemented uniformly across the entire site, otherwise RG will just become a patchwork of different users and their own personal feelings about what accuracy is appropriate. This would be a large step towards making the data unusable for many applications as it would strongly bias the data. So please, let’s avoid that.

7 Likes

#14

I agree. The google group thread @jakob linked to has a lot of examples of where such data can be very valuable (e.g. a historical rhino observation in an area where rhinos are no longer found). The observations already come with an accuracy buffer, so researchers can filter based on this depending on their purpose. If there are still problems with how the data appear on species maps, then let’s focus on making the maps more customisable, but not on making otherwise Research Grade observations casual.

I also agree there’s a problem with observations that lack any accuracy value. I’ve noticed that the accuracy circle disappears when manually moving a pin. This is unhelpful behaviour - instead it should retain an accuracy circle depending on how zoomed-in the map is. @cthawley perhaps you’d like to make a feature request to fix this?

5 Likes

#15

If you want to only see observations with certain accuracy values, see this thread for some filters you can use:
https://forum.inaturalist.org/t/create-filter-by-location-accuracy/298/6 (Kueda’s post on March 4)

I think this will serve your purposes fully depending what exactly you’re doing?

I can see some potential uses for having these filters available in a project, if that’s what you would need then maybe make a feature request for that option.

I agree with some above that is very inappropriate to be using “location inaccurate” for observations that pass some arbitrary threshold that you personally determine. Unless you have some reason to think the observation is not encompassed by the accuracy circle, you should not be using this option. Even if the accuracy is 1000 km. Marking “location inaccurate” provides zero additional information, all it does is hide the observation from other people who may want to see it despite that accuracy. I don’t see why anyone else should be making the decision for me as to what observations I think are valuable.

3 Likes

#16

I could see where this could be detrimental. One example is that my camera depends on my phone for GPS. Usually, this works, but the phone app is buggy and doesn’t always supply my camera with coordinates. I usually take photos in a small private lands that I obscure anyway, so I can fill the area in.

However, older photos that were taken before GPS was widely available is more difficult. One of my pet projects is finding and digitizing the slides of my dad’s trip to Finland in the 1970’s. If I find any slides with good shots of wildlife with the date on them, I’d still have to entirely rely on his memory for location.

Not everyone has access to GPS even now, and requiring exact coordinates could exclude a lot of valuable data.

I do like the idea of a tool/filter that could help people narrow down ranges to those with exact coordinates. That may be the solution here.

0 Likes

#17

This seems arbitrary. Setting it as the default would result in a lot of good data being ignored. Precise data are impossible when you take photos with your camera (not a phone) and are walking on a long trail. iNaturalist doesn’t currently allow anything other than circles for georeferencing, but a 100 m circle covers a lot more space than a much longer established path, which people may mention walking along in the descriptions. Better to use filters to cull whatever spatial data doesn’t meet the standards for your specific study, but keep it open for others to decide their own cutoffs.

6 Likes

#20

does anyone know what the limit is for the range maps? Maybe just tightening that a bit would help.

1 Like

#21

~30 km depending on how far from the equator

3 Likes

#22

yeah that’s kind of absurdly high uncertainty. I think anything under 1 KM or at the bare minimum 10 KM should be excluded from the range maps. The reason people are marking them as mapped wrong is probably because some of them show up in really absurd places when they are mapped that coarsely. Since the species/range maps do NOT show the uncertainty buffer, in those cases they ARE inaccurate since it’s just a point in the wrong place.

Too be honest it’s hard to imagine anything other than fringe types of use for data that is that coarse. It’s hard to imagine much value to it at all at least in the case of plants. I know it’s probably not worth it to get re-engaged in this conversation, but it sad to see data precision dismissed because of a few edge cases where people have a use for highly imprecise data. And in the least, if something mapped to a continent level could be research grade, things that don’t have an exact date should too. That data is a lot more useful in a lot of ways than 100 km accuracy circle data.

0 Likes