spatial accuracy new excluding some obs from becoming RG?

[AfriBats]

Hi developers

It seems there's a new criterion under Data Quality Assessment, which says Position accurate to within 5km?.

Does that mean that all observations with positional accuracy of > 5 km are automatically excluded from becoming Research Grade? If that's indeed the case, I suggest to remove that criterion again, or set it to a cut-off value several orders of magnitude larger.

Posititional accuracy is one of the core data fields that is included in data made available through GBIF. The specific value above which an obervation is useless for a given research question depends on many factors, and cannot be set to a one-size-fits-all value such as 5 km. For instance, we just analyzed the spatio-temporal occurrence of migratory Lepidoptera across Europe, where a grid size of 100 km (and hence a positional accuracy much larger than 5 km) was deemed appropriate for our purposes.

Everyone using GBIF data should carefully evaluate and filter available data before using them in any further analyses, and positional accuracy is obviously one of the central fields in that respect. For that reason I don't see any problems sharing a RG observation with a large circle of accuracy as long as that value is correct. On the contrary, I'm much more concerned with observation that appear to be accurate, but in fact are not. For instance, I see may observations mapped to the Google midpoint of a protected area, e.g. that of Tarangire NP. Several users apparently search for that area, but don't move the map pin to the specific place where the observation was made within the park, or adjusting the circle of accuracy to encompass the entire park.

Jakob

[AfriBats]

The title of the thread should say "Spatial accuracy now..."

[Charlie Hohn]

I submit that it should be retained. At least for plants. Plant observations at that level of inprecision or lower are absolutely valueless and mess up the range maps and phenology data. Please keep them out.

Also most people don't know how to use the uncertainty buffer. But I think the main issue is not adding it rather than making it too small. 

On Sat, Nov 19, 2016 at 10:27 AM, AfriBats <jakob...@gmail.com> wrote:

The title of the thread should say "Spatial accuracy now..."

--
You received this message because you are subscribed to the Google Groups "iNaturalist" group.
To unsubscribe from this group and stop receiving emails from it, send an email to inaturalist+unsubscribe@googlegroups.com.
To post to this group, send email to inatu...@googlegroups.com.
Visit this group at https://groups.google.com/group/inaturalist.
For more options, visit https://groups.google.com/d/optout.

--

============================
Charlie Hohn
Montpelier, Vermont

[Tony Iwane]

Hey guys, here's the GitHub thread about this issue:https://github.com/inaturalist/inaturalist/issues/1183

Personally I agree with Charlie that decent accuracy should be part of RG, but Jakob you bring up a good point about the points at the centers of parks, etc. 

[AfriBats]

I propose to discuss such a pretty fundamental issue first before having it implemented. Again, I don't think that high accuracy should be made a prerequisite for an observation being elegible for RG. Think of a highly mobile species: if I observe a humpback whale somewhere off the coast, and can map that observation to with +- 100 km, it's still a valuable record for many analyses dealing with humpback whales. Observations of less mobile species, or sessile species such as plants or corals, are obviously less valuable if accuracy is low. My point is that 5 km is way too strict as a general criterion applied to all organisms, and the extreme examples given in the GitHub issue could be easily dealt with by using a much larger cut-off value.

Charlie: if I'd analyse plant data, I would also use a strict cut-off value with regard to spatial accuracy in most cases. But that value should be question-driven rather than applying a single, exclusive value that doesn't suit many other cases and organisms.

I think that false accuracy is a much bigger issue - in my example of Tarangire NP, many observations have no accuracy information whatsoever although the circle should be probably in the range of dozens of km....

[Scott Loarie]

Hi Jakob,

That issue was up on Github for discussion for 15 days before the
change was made- I recommend you 'watch' the iNat repo if you want to
catch all the issues that are opened (see attached)

But I agree that using a data quality flag might not be the best way
to handle imprecise locations and you raise some good points - happy
to continue the conversation there
https://github.com/inaturalist/inaturalist/issues/1183

> --
> You received this message because you are subscribed to the Google Groups
> "iNaturalist" group.
> To unsubscribe from this group and stop receiving emails from it, send an

> email to inaturalist...@googlegroups.com.

> To post to this group, send email to inatu...@googlegroups.com.
> Visit this group at https://groups.google.com/group/inaturalist.
> For more options, visit https://groups.google.com/d/optout.



--

--------------------------------------------------
Scott R. Loarie, Ph.D.
Co-director, iNaturalist.org
California Academy of Sciences
55 Music Concourse Dr
San Francisco, CA 94118
--------------------------------------------------

[Tracey Fandre]

I'm having a strange new problem myself.   I have always entered my observations using my "city, state" and tonight I entered a lot of things and many went research but a few went "casual" and the accuracy was anything from 98m to 14,844m.   Odd that some went through to research and others failed.  Seems like a bug.   I was able to fix them by pinpointing the accuracy.   

[Scott Loarie]

Hi Tracy, can you include specific URLs and Screenshots so I can
better understand your issue?

[AfriBats]

Hi Scott

That issue was up on Github for discussion for 15 days before the change was made

Reminds me a bit of the plans for an intergalactic express highway, which were on display for 50 Earth years on Alpha Centauri. More seriously: I assumed that such issues would be first raised here and only then taken up on GitHub. Does it mean that one would need to follow both the Google Group and GitHub to participate in discussions of iNat's development?

 

- I recommend you 'watch' the iNat repo if you want to
catch all the issues that are opened (see attached)

But I agree that using a data quality flag might not be the best way
to handle imprecise locations and you raise some good points - happy
to continue the conversation there
https://github.com/inaturalist/inaturalist/issues/1183

 

 Why not discuss it here where probably more iNat users are participating and potentially contributing now that this thread has been opened? I don't have a GitHub account...

Jakob

[Tracey Fandre]

I can tell you a few of the ones that I had this problem with last night, but since I went in and pinpointed the accuracy on them, they went research so the problem was resolved by editing.   I don't have one at the moment that is sitting "casual" that I know of.  One thing that would be nice would be to include "casual" in the search so we can pull those up and easily fix them.    Here are the URLs of a few that I fixed:    http://www.inaturalist.org/observations/4602805, http://www.inaturalist.org/observations/4602795, http://www.inaturalist.org/observations/4602793

It just doesn't make sense why a lot of them, when entered the same way went research right away and then some got rejected due to accuracy in location.     

[Tracey Fandre]

I found one that has the problem.  It is sitting "casual".   http://www.inaturalist.org/observations/4493281 is the observation.   It shows the accuracy at 14844m. 

[Tracey Fandre]

Okay, I'm seeing the problem and it is with me being lazy in some ways.   I will start pinpointing closer to where I saw the subject but one thing I suggest is that the general ring around the initial location i.e. Garland, TX not be so darn large.  It emcompasses more than Garland...goes into the surrounding cities.  If someone wants the area larger, they can physically do so, but if you want things pinpointed closer, a smaller circle when choosing location would be helpful.

[Tony Iwane]

Hey Tracey,

If you've got a smartphone or handheld GPS device, you can use it to help geotag the photos you take with an SLR or other non-GPS equipped camera.

I made a short video of the workflow that I use. It makes getting location sooo much easier. https://vimeo.com/175298466

Tony

On Saturday, November 19, 2016 at 7:25:48 AM UTC-8, AfriBats wrote:

[Tracey Fandre]

I'll check that out...thank you

[Tracey Fandre]

Tony, that is AWESOME!  I use lightroom and that would make it so much easier for me.   What app do you recommend for an iphone?  Not only will this be easier for me to input, but it will be better for researchers to have it exactly where I found it instead of a general location...and it solves the lazy on my part.

[Tony Iwane]

Thanks Tracey!

I use the Motion-X GPS app on my iPhone (it's the one used in the video) when I use my phone. It's a great app, well worth the measly 2 bucks it costs. FYI, GPS tracking does put a bit of a strain on your phone's battery. Generally not terrible, but something to be aware of. When I'm out for a long time I use a handheld GPS.

Tony

[Scott Loarie]

Hi Jakob,

We try to post most major issues to the Github Group for discussion,
but if you want to track the tickets directly then I'd recommend you
follow the Github repo also.

I didn't post issue 1183 here because I didn't think it was a major
issue. I'll try to be less conservative in the future. But for
discussing the finer points of dev change I also prefer Github to the
Google Group.

Since you've raised it here, everyone should be aware of the Github
ticket and I invite you or anyone else to comment on that ticket. I'd
be particularly interested in your feedback on the histogram I pasted
in.

Best,

Scott

[AfriBats]

Hi Scott

Thanks for getting back to this! I'm kinda reluctant to also join iNat's GitHub site as I'm already struggling to keep up to date with the Google group. Would be great if the development team keeps us informed here about major steps.

Since you've raised it here, everyone should be aware of the Github
ticket and I invite you or anyone else to comment on that ticket. I'd

be particularly interested in your feedback on the histogram.

I'm pasting your chart plus explanation here:

 

This 5km cuttoff will convert 131,038 currently verifiable obs to casual. Here's the number of currently verifiable obs in various positional accuracy bins (5-10km, 10-15km, ... , >50km):

We'd be throwing out of verifiable ~35k fewer if we moved the cutoff from 5km to 10km. I'm a bit torn, I'd like to stick with 5km because thats the arbitrary (e.g. TIGER) coastal buffer cutoff where I'm getting the not-in-place issues. But I'd also like to keep as many that are useful in the verifiable bin

 

I would say: throwing out a single, valuable observation is one observation too much if there is not a good justification why we should exclude observations from becoming research grade. I've mentioned some points in previous posts and will add a bit here:

First, I'm quite pedantic with regard to mapping my own observations as accurately as possible. It would be obviously nice to have everything mapped with GPS-accuracy, but that's obviously impossible for various reasons.

Personally, I love to see users sharing old observations, and I think they can have tremendous value for science and conservation. A historic rhino observation in a region where there are no longer rhinos is highly informative if accurately mapped to within dozens or even a few hundred kilometres of the actual locality. If accuracy > 5 km excludes such observations from becoming RG, natural history collections would need to exclude major parts of their collections from data sharing with GBIF.

Whether an observation is useless due to low accuracy is dependent on the organism and research question. This observation, mapped with an accuracy of ~ 8 km, is currently excluded from qualifying for RG: www.inaturalist.org/observations/4605571. Mouse-tailed bats are powerful fliers and cover such distances within 10 min. Is that observation not accurate enough to determine the wider distribution, climatic tolerance, and general habitat requirements? I don't think so. Would I include this observation to analyze the fine-grained habitat conditions around roost sites? Probably not because the point is not accurate enough.

There are loads of species where it doesn't matter too much if there are a few observations more or less, but I don't want to lose a single Rhinopoma-observation, for which there are currently 11 for entire Africa on iNat.

Here's another example of a rarely observed gecko, which I would have overlooked because it doesn't show up when screening verifiable observations with the Identify page: www.inaturalist.org/observations/4595488. A rather low accuracy of 18 km is still good enough to understand the general distribution of this species (group), isn't it?

Maybe I'm missing something why these observations should be excluded from being verifiable / becoming RG. If you're concerned about widely inaccurate observations, then simply set the cut-off value to 500 or even 1000 km (your GitHub examples). Anything below such a large value should be left to those using the data in further analyses or applications.

I'd love to see functionality implemented into the uploader which prevents users mapping observations to large places and a seemingly high accuracy!

Cheers, Jakob

[Charlie Hohn]

The bad precision observations seriously screw up the plant range maps and I want to be able to use the maps without downloading and processing the data. If you must increase the cutoff can you do it for animals only? Or at least give the observations the same icon as obscured so I can know to ignore them? These excluded plant observations are not just useless but actively obstructive to doing ecological research and they have been an annoyance to me for forever. 

Just because they aren't research grade doesn't mean afrobats or others who want them can't still download them. Same as observations with no date or no photo. We aren't deleting them. I don't see why a few peoples desire for things that are mapped wrong justifies making them research grade. 

--

You received this message because you are subscribed to the Google Groups "iNaturalist" group.

To unsubscribe from this group and stop receiving emails from it, send an email to inaturalist+unsubscribe@googlegroups.com.

To post to this group, send email to inatu...@googlegroups.com.
Visit this group at https://groups.google.com/group/inaturalist.
For more options, visit https://groups.google.com/d/optout.

--
Sent from Gmail Mobile

[James Bailey]

Whether X or Y radius is too inclusive or exclusive should be for researchers to decide. Sometimes the entire database of historical records for certain species, both rare and common, consists of only 10km or greater observations. That could be critical data that is being excluded. It is true that users can just download observations from iNat regardless of radius if they feel suited, but how many researchers using GBIF are going to come all the way over here? If I was a researcher I might not bother to look any deeper into iNaturalist -- I would take the (research grade) observations and data, and move on. 

GBIF and other sites already take many observations at 5km or greater radius. I think it is worth discussing if it is "fair" to impose this limitation here, on iNaturalist. The question then really is "are there enough researchers using 5km+ accuracy observations?". If yes, then imposing this limitation seems too arbitrary. If no, then I guess this won't affect many people.

If others feel leaving research grade on non-accurate observations is an issue, perhaps there are some steps that could be taken (ability to filter observations in searches by accuracy, or not allow 5km+ accuracy observations to show on range maps, etc).

Just my quick thoughts.

[Charlie Hohn]

can you think of any value of a plant observation with higher uncertainty than that? I can't and i am a plant ecologist. I really think they do more harm than good. It would be best if they didn't show up on the range maps, just like the not native/naturalized ones. Including the range maps you see in each observation's map. I also think they should not be shared with gbif but i actually care less about that than the range maps. 

If a researcher is not looking deeper than GBIF it's even more important to exclude those becuase those 'lazy' researchers aren't going to see the uncertainty buffer either and that means not just imprecise data but bad data.

i don't understand how fairness plays in, ecuase people aren't being excluded from sharing the observations. We're just keeping useless or harmful ones from damaging the scientific utility of inat..

i should stop here before i just piss everyone off but.. i care about this and i think it's really important for plant ecology. Please keep the bad data out. It truly does more harm than good pretty much all the time, at least for plants.

--

You received this message because you are subscribed to the Google Groups "iNaturalist" group.
To unsubscribe from this group and stop receiving emails from it, send an email to inaturalist+unsubscribe@googlegroups.com.
To post to this group, send email to inatu...@googlegroups.com.
Visit this group at https://groups.google.com/group/inaturalist.
For more options, visit https://groups.google.com/d/optout.

--

[James Bailey]

I agree that plant observations are not very useful when they are not within 5km radius. Other taxon groups can be useful at this accuracy though (birds, mammals, and insect collections notably).

As for "fairness", my point is only that we should consider that there might be researchers who find these to be useful. Deciding that these observations are not worthy is potentially unfair for researchers who could use that data. But these are probably limited cases...so, I guess these is nitpicking to do here.

[Colin Meurk]

well i really agree with those wanting to be more inclusive. as many have said, a lot of historical records have very low accuracy. i understand that specific questions and fine-scale mapping against topographic or some other soil parameter require very accurate geo-positioning, but i'm not sure if iNat is really the place to get such data (except for personal data that has been collected for that purpose - which is easily extractable); on the other hand for larger geographic analyses such as climate gradients or regional distributions (such as for eco-sourcing plant species for ecological restoration) a general accuracy of tens of kms may be quite adequate. but i don't see why this is an issue for RG.  is anyone that is taking the data from GBIF going to be doing fine-scale ecological analysis?  i wouldn't have thought so, but are likely to be interested in broad bio-geographic patterns.

while on the topic of RG, i have raised the matter before about disqualifying captured/cultivated (= planted records) from being RG.  this makes no sense to me as they are still valid ecological statements about a species and help to define its actual/realised ecological envelope/niche.  i don't recall any debate or justification for this exclusion.  if the species/cv is accurately identified and confirmed and the location given then it is a valuable legit record. of course maps of such records should be colour-coded or have specific flags for 'cultivated' so natural distributions are clearly distinguished from those aided by people.  i'd probably draw the line at indoor or artificially grown plants or inside-zoo specimens. but for example we allow volunteer plants in irrigated lawns to be given RG so what's the difference.

c

[Charlie Hohn]

There are hundreds of thousands of accurately located plant observations on inat. It's absolutely the best place I know of to find those. Other places don't display accuracy or just have a bunch of wrong stuff. 

--

You received this message because you are subscribed to the Google Groups "iNaturalist" group.
To unsubscribe from this group and stop receiving emails from it, send an email to inaturalist+unsubscribe@googlegroups.com.
To post to this group, send email to inatu...@googlegroups.com.
Visit this group at https://groups.google.com/group/inaturalist.
For more options, visit https://groups.google.com/d/optout.

--
Sent from Gmail Mobile

[Charlie Hohn]

i also think maybe having it not be research grade will communicate with people why they shouldn't upload that sort of data?

[Scott Loarie]

Hi all,

So the system is currently automatically excluding all obs with a location accuracy >5km from Research Grade.

it sounds like there's some resistance to this specific threshold and/or the basic idea of including location accuracy in data quality assessment. Of the people who have chimed in, preferences are:

Scott and Charlie: keep 5km accuracy threshold for RG
Jakob: keep threshold but increase to 50km
Colin and James:  don't consider location precision in data quality

As a compromise I propose we keep location accuracy as a component of the data quality assessment but increase the threshold to 50km as Jakob proposes. This 50km radius circle has an area about the size of the state of Rhode Island and it seems that (a) the number of observations with such large uncertainties is small enough and (b) the 'harm' they do in making the maps look really funky outweighs the 'good' they do in determining where species occur. Does this sound like a fair compromise?

However, I'd like to also propose that an acceptable use of the  'Does the location seem accurate?' = 'no' flag is to manually flag observations where the centroid doesn't make sense where the accuracy >0km but <50km range on a case by case basis. For example, if someone posts an observation of a salamander with a marker centered in the ocean but a 10km accuracy circle that does overlap the coast, I'd like to be able to manually flag the location as not seeming accurate.

This is because I'm working on a system to automatically detect observations that are out of range for further review. For example, if a salamander is endemic to California and Oregon but it is observed in Texas, this gets flagged for review. Currently observations with their centroids >5km from the coastline are getting flagged (e.g. its not in Oregon or California but rather the Pacific Ocean) and there's no easy way to incorporate location accuracy into this new system.  So while I'm okay with not having these observations automatically flagged via the <5km filter. I'd like to be able to manually flag them on a case by case basis using the existing 'Does the location seem accurate?' flag so they stop triggering this automatic system. Does this sound fair?

Best,

Scott

>>> email to inaturalist...@googlegroups.com.

>>> To post to this group, send email to inatu...@googlegroups.com.
>>> Visit this group at https://groups.google.com/group/inaturalist.
>>> For more options, visit https://groups.google.com/d/optout.
>>
>>
>>
>> --
>> Sent from Gmail Mobile
>
>
>
>
> --
> ============================
> Charlie Hohn
> Montpelier, Vermont
>

> --
> You received this message because you are subscribed to the Google Groups
> "iNaturalist" group.
> To unsubscribe from this group and stop receiving emails from it, send an

> email to inaturalist...@googlegroups.com.

> To post to this group, send email to inatu...@googlegroups.com.
> Visit this group at https://groups.google.com/group/inaturalist.
> For more options, visit https://groups.google.com/d/optout.



--

[Charlie Hohn]

Is there a way to retain the 5KM cutoff for plants only? I see a good argument for using a larger cutoff for mobile organisms, especially birds, large roaming mammals, etc. But I still don't see any real value in having research grade plant observations with uncertainties up to 50 (!) KM. I still contend that they are detrimental to data quality.

If i'm outvoted i'll just flag them when I see them. But I feel like if doable the different radius for plants would make a more balanced compromise than going to 50 KM (which in my mind is basically the same as nothing, at least for plants, because it calls all points into question). I see these a lot because i often look at range maps and go after outliers to see if I can verify them or not. Then I open up a possibly exciting outlier population and see that it's just poorly mapped. 50 KM in California is the difference between a seaside salt marsh, an alpine tundra, an oak woodland, dense pine forest, prairie, and the Mojave Desert. Might as well just map it to the state, or not map it at all.

 I can't think of much of anything these plant observations are good for except possibly phenology observations in extremely uniform and flat areas like the Great Plains (where we barely have any observations anyway). Anywhere else microtopography and microclimate (elevation and proximity to large bodies of water) make the observations pretty much useless for even that.

I know i probably won't convince any one, but i'd honestly exclude all observations without an uncertainty buffer from R.G. as well.

other possible ideas:

-a way to notify users when they have huge uncertainty circles, to prompt them to increase accuracy?

-different icons for anything >5 km, maybe the same as the obscured ones or similar

-removing anything over 5km from the range maps, at least for plants

-disallow mapping by place when it creates those giant circles? Many of them are wrong anyway... for instance pretty much everything mapped to 'eaton canyon' pops up in a golf course in the city, with the uncertainty circle not including most of the canyon at all.

>>> email to inaturalist+unsubscribe@googlegroups.com.

>>> To post to this group, send email to inatu...@googlegroups.com.
>>> Visit this group at https://groups.google.com/group/inaturalist.
>>> For more options, visit https://groups.google.com/d/optout.
>>
>>
>>
>> --
>> Sent from Gmail Mobile
>
>
>
>
> --
> ============================
> Charlie Hohn
> Montpelier, Vermont
>
> --
> You received this message because you are subscribed to the Google Groups
> "iNaturalist" group.
> To unsubscribe from this group and stop receiving emails from it, send an

> email to inaturalist+unsubscribe@googlegroups.com.

> To post to this group, send email to inatu...@googlegroups.com.
> Visit this group at https://groups.google.com/group/inaturalist.
> For more options, visit https://groups.google.com/d/optout.



--

--------------------------------------------------
Scott R. Loarie, Ph.D.
Co-director, iNaturalist.org
California Academy of Sciences
55 Music Concourse Dr
San Francisco, CA 94118
--------------------------------------------------

--

You received this message because you are subscribed to the Google Groups "iNaturalist" group.

To unsubscribe from this group and stop receiving emails from it, send an email to inaturalist+unsubscribe@googlegroups.com.

To post to this group, send email to inatu...@googlegroups.com.
Visit this group at https://groups.google.com/group/inaturalist.
For more options, visit https://groups.google.com/d/optout.

--

[AfriBats]

Hi Scott

Jakob: keep threshold but increase to 50km

Nope, if setting a threshold, I would set it much larger: 500 km or more looks good to me, but only to throw out clearly erroneous observations, and not the baby with the bathtub.

What we're discussing here is estimated or measured accuracy. This bit, together with the XY-coordinates, is shared with GBIF, so every researcher should carefully evaluate which level of accuracy is required for a given question and taxonomic group. If the new, automatic system mentioned below requires to automatically flag observations with low accuracy, it should be de-coupled from being eligible for RG and eventually getting shared with GBIF.

I bet that 99% of researchers interested in the kind of data iNat is generating will go through GBIF rather than assembling them through the various, original data providers, so maintaining a widely open door when transferring data to GBIF is of prime importance, at least to me.
 

 

As a compromise I propose we keep location accuracy as a component of the data quality assessment but increase the threshold to 50km as Jakob proposes. This 50km radius circle has an area about the size of the state of Rhode Island and it seems that (a) the number of observations with such large uncertainties is small enough and (b) the 'harm' they do in making the maps look really funky outweighs the 'good' they do in determining where species occur. Does this sound like a fair compromise?

In previous posts of this thread I've tried to explain why I consider inaccurate observations in several cases as highly relevant and not to be discarded. Except for the extreme cases mentioned in the GitHub thread, I'd like to see some examples where inaccurate observations mess up aggregated species maps. Take a wandering albatross. A humpback whale. A Triops shrimp. Knowing the Where and When within several hundred km can be highly informative in these cases. I'm not a botanist, but I'm sure there are also widely dispersing plant species, and those occurring in vast and rather uniform habitats, where low accuracy is also not an issue in some applications.

 

However, I'd like to also propose that an acceptable use of the  'Does the location seem accurate?' = 'no' flag is to manually flag observations where the centroid doesn't make sense where the accuracy >0km but <50km range on a case by case basis. For example, if someone posts an observation of a salamander with a marker centered in the ocean but a 10km accuracy circle that does overlap the coast, I'd like to be able to manually flag the location as not seeming accurate.

Fully agreed, I'm also using the flag "not accurate", although I'd prefer a different wording. Your salamander case would suggest to me that the actual point is simply misplaced, and that's something very different from a correctly mapped location with a large circle of accuracy. If the observer was unsure of the location, he/she would place it somewhere on land, and with a circle of accuracy encompassing the wider area that was visited (remembered).

 

This is because I'm working on a system to automatically detect observations that are out of range for further review. For example, if a salamander is endemic to California and Oregon but it is observed in Texas, this gets flagged for review. Currently observations with their centroids >5km from the coastline are getting flagged (e.g. its not in Oregon or California but rather the Pacific Ocean) and there's no easy way to incorporate location accuracy into this new system.  So while I'm okay with not having these observations automatically flagged via the <5km filter. I'd like to be able to manually flag them on a case by case basis using the existing 'Does the location seem accurate?' flag so they stop triggering this automatic system. Does this sound fair?

Looking forward to see that system! Sounds pretty focussed on terrestrial biota, or are you doing the reverse, i.e. flagging marine species that are mapped far inland? What about species that are neither terrestrial nor marine? Without knowing what you're actually planning, I would think that this approach requires quite a bit knowledge of the respective ecologies.

In any case, the manual flag certainly sounds fair and useful, but as I said above, I would word it differently - maybe "Location misplaced: y/n"? This would shift the view from the isolated aspect of accuracy to the combined information that we strive to evaluate: XY-location plus circle of accuracy.

Cheers, Jakob

[Colin Meurk]

I’m ok with the 50km buffer (without giving it too much thought) – it seems a reasonable compromise in terms of broad bio-geographic purposes.  RG of course doesn’t affect the distribution maps viewed within iNat so all data is presented regardless of accuracy.  I presume Charlie and others who are wanting to map at a finer scale in relation to environmental gradients can filter these coarser data accordingly.

My ongoing concern and wish, which doesn’t seem to have attracted any/much discussion, is to allow RG for planted/cultivated/domesticated organisms.  I’ve given my reasons before – but what is needed/desirable is for such records (maybe same for species that are detected out of zone referred to in your post Scott) to be flagged distinctively. E.g. we have colour-coded pins for the various taxonomic groups, and empty flags for ‘location unknown’; can’t we have say an empty, colour-coded/edged pin for species that are planted/domesticated out of natural range.  This happens a lot in NZ, and yet it is ecologically valid information to know that say kauri (Agathis australis) is planted and growing successfully in Stewart island (at bottom of country), and that it is coning or producing seedlings halfway down the country, but its natural distribution is confined to the top 1/3 of the North Island.  This would be a hugely valuable rendering of information contained within records.

Cheers colin

 

Colin meurk | Research associate
LANDCARE RESEARCH MANAAKI WHENUA
DDI: +64 3 321 9740 | M: +64 27 702 8325
W:www.landcareresearch.co.nz | E: meurkc.landcareresearch.co.nz

 

--
You received this message because you are subscribed to a topic in the Google Groups "iNaturalist" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/inaturalist/-JR7FgQWALQ/unsubscribe.
To unsubscribe from this group and all its topics, send an email to inaturalist...@googlegroups.com.

To post to this group, send email to inatu...@googlegroups.com.
Visit this group at https://groups.google.com/group/inaturalist.
For more options, visit https://groups.google.com/d/optout.

---

Please consider the environment before printing this email
Warning: This electronic message together with any attachments is confidential. If you receive it in error: (i) you must not read, use, disclose, copy or retain it; (ii) please contact the sender immediately by reply email and then delete the emails.
The views expressed in this email may not be those of Landcare Research New Zealand Limited. http://www.landcareresearch.co.nz

[Colin Meurk]

Thx afribats – am quite persuaded by your argument.  GBIF users should be able to determine for themselves which data is relevant to their needs and filter accordingly.

500 km will seem very large for terrestrial records on an island, but not so much for a continent.

colin

 

Colin meurk | Research associate
LANDCARE RESEARCH MANAAKI WHENUA
DDI: +64 3 321 9740 | M: +64 27 702 8325
W:www.landcareresearch.co.nz | E: meurkc.landcareresearch.co.nz

 

 

From: inatu...@googlegroups.com [mailto:inatu...@googlegroups.com] On Behalf Of AfriBats
Sent: Tuesday, 29 November 2016 11:53 a.m.
To: iNaturalist
Subject: Re: [inaturalist] spatial accuracy new excluding some obs from becoming RG?

 

Hi Scott

--
You received this message because you are subscribed to a topic in the Google Groups "iNaturalist" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/inaturalist/-JR7FgQWALQ/unsubscribe.
To unsubscribe from this group and all its topics, send an email to inaturalist...@googlegroups.com.

To post to this group, send email to inatu...@googlegroups.com.
Visit this group at https://groups.google.com/group/inaturalist.
For more options, visit https://groups.google.com/d/optout.

[Scott Loarie]

Sounds like no one is totally satisfied (myself included) with my proposal - but that a sign of a good compromise.

Charlie: I'd prefer not to complicate things with taxon specific criteria. Jakob, 50km is a compromise between what you want and what others (like me want, 5km) so I'm assuming you're ok with that as a compromise (you can always find the obs you're looking for in the casual bin)? Colin: I'd prefer not to mix in your concerns about planted/cultivated/domesticated concerns in this thread.

So in the effort to move things forward, just to re-iterate my proposal:

1. keep location accuracy as a component of the data quality assessment but increase the threshold to 50km 

2. also propose that an acceptable use of the  'Does the location seem accurate?' = 'no' flag is to manually flag observations where the centroid doesn't make sense 

I'll try to make this change in a day or two

To unsubscribe from this group and all its topics, send an email to inaturalist+unsubscribe@googlegroups.com.

To post to this group, send email to inatu...@googlegroups.com.
Visit this group at https://groups.google.com/group/inaturalist.
For more options, visit https://groups.google.com/d/optout.

---

Please consider the environment before printing this email
Warning: This electronic message together with any attachments is confidential. If you receive it in error: (i) you must not read, use, disclose, copy or retain it; (ii) please contact the sender immediately by reply email and then delete the emails.
The views expressed in this email may not be those of Landcare Research New Zealand Limited. http://www.landcareresearch.co.nz

--
You received this message because you are subscribed to the Google Groups "iNaturalist" group.
To unsubscribe from this group and stop receiving emails from it, send an email to inaturalist+unsubscribe@googlegroups.com.

To post to this group, send email to inatu...@googlegroups.com.
Visit this group at https://groups.google.com/group/inaturalist.
For more options, visit https://groups.google.com/d/optout.

[AfriBats]

Hi Charlie

Here some response to the points you've raised:

The bad precision observations seriously screw up the plant range maps

Could you give some examples where this is the case? If we're talking about range maps, i.e. the generalized distribution generated from point data, accuracy of several dozen km is usually inherent in these data. This paper is highly recommended http://dx.doi.org/10.1073/pnas.0704469104
 

I want to be able to use the maps without downloading and processing the data.

Could you elaborate a bit, or give examples, how you're using iNat data, and where imprecise data hamper your interests? If we're talking about proper analyses, I think there's no way around downloading data and filtering them according to your own criteria. And then it's easy to apply a stricter threshold for accuracy.

 

Just because they aren't research grade doesn't mean afrobats or others who want them can't still download them. Same as observations with no date or no photo. We aren't deleting them. I don't see why a few peoples desire for things that are mapped wrong justifies making them research grade.

My concern is not about having them deleted, or that I'm unable to download them. I'm talking about whether or not fairly inaccurate observations are shared with GBIF. To make this plain clear: I'm not talking about misplaced locations ("things that are mapped wrong"). I'm referring to observations that are correctly mapped, but with a large circle of accuracy. That's a big difference, and I invite you to read a bit about the concepts of accuracy and precision.

Best, Jakob

[Matt Goff]

Re: non-animal accuracy - 

For what it's worth, where I live in Southeast Alaska there are a few fairly detailed mapping projects for particular organisms, but really, this region is still at the level of asking what even occurs here at all. 

It's an archipelago over 1000 islands and s a thin strip of mainland between the coast and the crest of the coast range (where the border with Canada lies).

I'm very much an amateur but have been working on trying compile information and get a better sense of what occurs in the region and where. This is big scale stuff, like knowing that a species even occurs in the region at all. 

Of course it's nice to have a highly accurate location, but the reality is, for many things, an accuracy of 5km (and more) would be a significant improvement over what is known now. 

I would love to see highly detailed ecological/occurrence maps in the long run, but at this point even just getting reasonable data-based maps for most species that accurately indicate whether a species occurs on/around the major island (groups) would be a big improvement.

This being the case, I have to suspect that there are also other under documented places in the world where broader scale (bigger circles of accuracy) information is still important/useful, so I think it makes more sense to have a filter which allows one to specify minimum accuracy desired rather than the approach of in/out of research grade.

Thanks,

Matt

[Ben Phalan]

Does Charlie's concern relate to using the maps of 'Observations' generated within iNaturalist? If that's the case, it would seem that adding an ability to filter observations to any specified level of accuracy would address that issue.

[AfriBats]

Does Charlie's concern relate to using the maps of 'Observations' generated within iNaturalist? If that's the case, it would seem that adding an ability to filter observations to any specified level of accuracy would address that issue.

Excellent suggestion!

[Scott Loarie]

Hi folks, you can't currently filter observations based on spatial accuracy. Maybe thats something we can do down the road but lets move it to another thread. This discussion is about https://github.com/inaturalist/inaturalist/issues/1183

The proposed compromise is to up the threshold from 5km to 50km.

If you really absolutely can't live with that, then the alternative would be pulling spatial accuracy from the data quality assessment altogether.

I'd like to wrap this up by the end of the day.

Best,

Scott

On Mon, Nov 28, 2016 at 3:47 PM, AfriBats <jakob...@gmail.com> wrote:

Does Charlie's concern relate to using the maps of 'Observations' generated within iNaturalist? If that's the case, it would seem that adding an ability to filter observations to any specified level of accuracy would address that issue.

Excellent suggestion!

--

You received this message because you are subscribed to the Google Groups "iNaturalist" group.
To unsubscribe from this group and stop receiving emails from it, send an email to inaturalist+unsubscribe@googlegroups.com.
To post to this group, send email to inatu...@googlegroups.com.
Visit this group at https://groups.google.com/group/inaturalist.
For more options, visit https://groups.google.com/d/optout.

[Colin Meurk]

I can live with either your last suggestion (RG based on other criteria only – i.e. having a photo or some independent means of checking identification; and having some geographical location) or 50 km.  basically bigger/more inclusive ..

Thx colin

 

Colin meurk | Research associate
LANDCARE RESEARCH MANAAKI WHENUA
DDI: +64 3 321 9740 | M: +64 27 702 8325
W:www.landcareresearch.co.nz | E: meurkc.landcareresearch.co.nz

 

 

From: inatu...@googlegroups.com [mailto:inatu...@googlegroups.com] On Behalf Of Scott Loarie
Sent: Tuesday, 29 November 2016 12:48 p.m.
To: inatu...@googlegroups.com
Subject: Re: [inaturalist] spatial accuracy new excluding some obs from becoming RG?

 

Hi folks, you can't currently filter observations based on spatial accuracy. Maybe thats something we can do down the road but lets move it to another thread. This discussion is about https://github.com/inaturalist/inaturalist/issues/1183

 

The proposed compromise is to up the threshold from 5km to 50km.

 

If you really absolutely can't live with that, then the alternative would be pulling spatial accuracy from the data quality assessment altogether.

 

I'd like to wrap this up by the end of the day.

 

Best,

 

Scott

On Mon, Nov 28, 2016 at 3:47 PM, AfriBats <jakob...@gmail.com> wrote:

 

Does Charlie's concern relate to using the maps of 'Observations' generated within iNaturalist? If that's the case, it would seem that adding an ability to filter observations to any specified level of accuracy would address that issue.

Excellent suggestion!

--
You received this message because you are subscribed to the Google Groups "iNaturalist" group.

To unsubscribe from this group and stop receiving emails from it, send an email to inaturalist...@googlegroups.com.

To post to this group, send email to inatu...@googlegroups.com.
Visit this group at https://groups.google.com/group/inaturalist.
For more options, visit https://groups.google.com/d/optout.

 

--

--------------------------------------------------
Scott R. Loarie, Ph.D.
Co-director, iNaturalist.org
California Academy of Sciences
55 Music Concourse Dr
San Francisco, CA 94118
--------------------------------------------------

--
You received this message because you are subscribed to a topic in the Google Groups "iNaturalist" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/inaturalist/-JR7FgQWALQ/unsubscribe.
To unsubscribe from this group and all its topics, send an email to inaturalist...@googlegroups.com.

To post to this group, send email to inatu...@googlegroups.com.
Visit this group at https://groups.google.com/group/inaturalist.
For more options, visit https://groups.google.com/d/optout.

[AfriBats]

Hi Scott

The proposed compromise is to up the threshold from 5km to 50km.

If you really absolutely can't live with that, then the alternative would be pulling spatial accuracy from the data quality assessment altogether.

I've presented detailed arguments why a low threshold, be it 5 or 50 km, is unjustified with regard to RG and data transfer to GBIF, and I haven't seen a single good argument why 50 km would be a reasonable compromise. If this needs to be wrapped up and has either the 50 km or none alternative, I clearly vote for none.

Best, Jakob

[Matt Goff]

"2. also propose that an acceptable use of the  'Does the location seem accurate?' = 'no' flag is to manually flag observations where the centroid doesn't make sense"

I guess I was under the impression that this was already part of the purpose - but maybe I'm misunderstanding and/or using it wrong?

I've run into a couple of situations, one where the mobile app hadn't really got itself located, but the observation was made (so a huge accuracy circle for a spider observation that was centered out in the ocean), and another where someone put all of their observations from a single vacation into the same point in Southeast Alaska (in this situation, some of the observations were plausible for the point but made on the same day as other observations which were clearly from elsewhere - so it was pretty easy to deduce the observation locations, while plausible, were not accurate). In both cases I ended up marking them as inaccurate locations after the observer did not correct them. (For the vacation one, I left a comment and was told they were all added to the same point because it was easier, and the locations would be fixed later - which didn't happen).

I guess this is a longer way of saying that I agree with the proposed acceptable use of "Does the location seem accurate?" flag, whether there's a 50km cutoff for research grade or none. 

I don't have a strong opinion about the accuracy cutoff, since I am not so familiar with how GBIF is used, and it seems the bigger issue with research grade is what gets added to GBIF (especially if, in the longer run there is a way to filter observations for accuracy on iNaturalist). I am also not dealing with the issue that originally brought this question forward - that is, your (Scott) preference to have observations excluded from your analysis by having them be casual.

Thanks,

Matt

[Scott Loarie]

Ok - it sounds like there's enough concern from Jakob et al. about incorporating location accuracy into the data quality assessment that I will roll back

https://github.com/inaturalist/inaturalist/issues/1183 (e.g. there will be no location accuracy threshold for what can be RG)

Because folks like Charlie and myself still need a way to filter observations with location accuracies > 5km, I'll write a new ticket along the line of what Matt and Ben proposed for filtering observations by location accuracy directly.

Appreciate the feedback from all involved,

Scott

[Charlie Hohn]

not clear on why addressing plants separately is not a better compromise than throwing away the accuracy of plant data because of how albatross behave. Is there a programming limitation?

Afribats, I do a lot of mapping. I can classify a wetland based on which plants are there. I look at the dispersal of oaks in Vermont, very site specific, and if you find them a bit out of range it tells you a very important and interesting story. We also do modeling based on location and mocroclimate, etc. Well... we haven't used inat for this because the data is too inprecise. So it would be great if we could get that fixed or at least addressed. it's really sad to be taking one step forward and then two backwards in that regard. 
Other things: making a species list of a small park. Using the Identotron, these bad location observations mess up the identotron when they reach research grade. Finding invasive species - knowing there is a nasty invasive species somewhere in your state but not where is frustrating but completely useless. Trying to map phenology based on tags, we can make animations that show the flowering time of a plant as it moves north and up in elevation. But not with bad location data.

you say you are not a botanist, well i am, if another one comes in and says they want 50 km 'accuracy' observations of any plant species for any reason i will listen, but until then why not listen to someone who actually is a plant person? i don't mean to be argumentative, i just find your comment odd. I don't pretend to know what is right for mapping bats. I don't have a clue.

The filtering won't be as good as just having the data not RG, since it is bad data (at least for plants). It's no better than a misidentified species really. Instead people will see 'research' grade data with desert plants showing up in the species lists of temperate rain forests and alpine plants in warm swamps, etc, and not all of the core functionality of the page allows for filtering.  And now when i am looking at a white pine observation id have to click through several slow screens just to get the bad data out.  I'll take it though. Better than nothing.

I still don't understand why some of you are forcing us to accept bad data You can always download it on your own. 

[Charlie Hohn]

I'm sorry. It just doesn't feel like 'compromise' or good community for a few people to demand removal of a useful feature and remove the integrity of tens of thousands of plant observations because they want to be able to map a few albatross and can't remember where they saw them. Id go so far to say if you don't remember where you saw something within 10 KM you shouldn't map it, period. 

We've got nearly 50,000 plant observations of almost 1500 species just in Vermont. Yes we use this data. Yes we want it to be accurate. Please think carefully about why you are so insistent on compromising our ability to do science.

[g_go...@yahoo.com]

Anyone who cannot determine where he or she is within 5 kilometers (or 500), feel free to come see me or one of my associates:

Wolfgang Siebeneich

Hawthorn Glen

1130 N. 60th St.

Milwaukee, WI 53280

Courses offered in map and compass, orienteering, basic GPS, geocaching, etc.

[Matt Goff]

Charlie - 

As I mentioned previously, I live in an area (Southeast Alaska) which has been poorly documented. I am also an amateur with no formal botanical training (my academic background is math/stats), but I'm a pretty dedicated naturalist and spend a lot of time with plants (see http://wiki.seaknature.org/Vascular_Plants for the basic distribution information I've been trying to compile related to plants in this region).

I understand that the more accurately something can be mapped, the better. However, given the lack of records, sometimes all we have is a collection that says "Sitka." It may have come from the central town area, one of the nearby smaller islands, or somewhere else on Baranof Island (which is 90 miles long). Technically, the boundaries of "Sitka" currently include an area approximately the size of Delaware. Even though it encompasses such a large area, it is often still useful to know that the species has been found - it's not unusual to find we don't even have records documenting even that much.

I have no problem acknowledging that for the work you are doing in Vermont such fuzzy records are not useful, and really should be excluded for your purposes.  I imagine you have both a good historical record (with reasonably accurate locations) and a relatively high density of records overall. However, that is most definitely not the case where I live, and I would argue that just because fuzzy records are useless for your scientific work, it does not follow that they are useless for other work elsewhere, especially in less documented parts of the world.

In my experience, even fuzzy records of plants (or bryophytes, lichens, fungi) can be helpful. I hope the following couple of examples might be illustrative:

Consider Woodsia scopulina - the FNA distribution map clearly indicates it should be expected to occur (presumably in appropriate habitat) throughout Southeast Alaska: 

http://www.efloras.org/object_page.aspx?object_id=7222&flora_id=1

e-Flora BC has a nice map (on one or more databases): http://linnet.geog.ubc.ca/Atlas/Atlas.aspx?sciname=Woodsia%20scopulina 

Looking at that, I would wonder if the species really occurs in Southeast Alaska (it looks to be more of an interior species - I'm not sure the basis for FNA range map). 

However, there is single (non-georeferenced) record in our state herbarium is one from 1918, simply stated as "Haines": http://arctos.database.museum/guid/UAM:Herb:201653

To put that on the map with an accuracy circle would probably require at least a 10-20 km radius, if not more. It would be helpful if they had done so, as it's still useful information, given the overall lack of it for this area. (As an aside, I'm not sure why they haven't georeferenced it, as there other similar collections in the database that have been.)

Another example that came up recently on iNaturalist is Lycopodium lagopus: 

FNA map shows the range includes the very northeastern part of the region: http://www.efloras.org/object_page.aspx?object_id=6412&flora_id=1

e-Flora BC map has records that show it nearer the outer coast both north and south of the region: http://linnet.geog.ubc.ca/Atlas/Atlas.aspx?sciname=Lycopodium%20lagopus

There is currently a single observation in the state herbarium from the region (it's included on the eFlora BC map): http://arctos.database.museum/guid/UAM:Herb:248692

Here's the iNaturalist observation: http://www.inaturalist.org/observations/3219846

In this particular case, the location is pretty accurate and I am not convinced of the L. lagopus id. However, in principal, if someone posted a good, identifiable picture from a mobile app that didn't have good GPS signal (not unheard of with our northern latitude and steep terrain), I would probably ask them to refine their location, but even if they never followed up, I would still consider an accuracy of 50km (in order to cover the whole island) to be useful information (even while thinking it would be nice if it was more precise). 

Generally speaking, it's not been unusual to have new island records found (especially of alpine species) that previously were only known from >100 miles away, so any new island record is nice to know, even if the precision is only at island level.

While I try to map things as accurately as I can, I have also added observations that have a relatively large circle (>5km, though in my case, also <50km, since getting that far away isn't as simple as driving down the road). These are generally observations using photos I took years ago. I don't always remember the location details (especially for things I collected and bought home to photograph). It would be nice if they were more precise, but again, I think they are still valuable records, as they document that the species at least occurred in the area at that time.

So, to sum up, I disagree with a blanket claim that for plants (or any other non-mobile organisms) >5km (or 50km) accuracy is necessarily "no better than a misidentified species" (which seems to make it worse than useless), and should therefore be excluded by default from anything that utilizes research grade (including showing up in gbif). 

I think different research questions and work require different levels of accuracy, and it would be very helpful to have the option to select on this basis according to need. In iNaturalist, it seems like the way to most flexibly do that is through the filtering mechanism and flagging of observations that seem clearly out inaccurate (even if they have a large enough circle to include a plausible location - such as in the example that Scott posted in the github conversation), and I look forward to having that capability. 

Thanks,

Matt

[Colin Meurk]

Hello all

I have to say I basically agree with Matt Goff.  The scale and accuracy required depends on the question and the purpose.  We should not second guess or presuppose how that data might be used – SO LONG AS precise data is not obscured or degraded in some way.  If one is looking at a google map on iNat one is usually drawn to interrogating the outliers (which one can do easily for a broad visual evaluation of distribition); or one can generate one’s own maps from downloaded coordinates after first filtering out those that are beyond the precision tolerances (accuracy) of the particular purpose the researcher is interested in.

The important thing is ALL potential purposes are accommodated J.  c

 

Colin meurk | Research associate
LANDCARE RESEARCH MANAAKI WHENUA
DDI: +64 3 321 9740 | M: +64 27 702 8325
W:www.landcareresearch.co.nz | E: meurkc.landcareresearch.co.nz

 

 

From: inatu...@googlegroups.com [mailto:inatu...@googlegroups.com] On Behalf Of Matt Goff
Sent: Tuesday, 29 November 2016 4:49 p.m.
To: inaturalist
Subject: Re: [inaturalist] spatial accuracy new excluding some obs from becoming RG?

 

 

Charlie - 

From: inatu...@googlegroups.com [mailto:inatu...@googlegroups.com] On Behalf Of AfriBats
Sent: Tuesday, 29 November 2016 11:53 a.m.
To: iNaturalist
Subject: Re: [inaturalist] spatial accuracy new excluding some obs from becoming RG?

 

Hi Scott

Jakob: keep threshold but increase to 50km

Nope, if setting a threshold, I would set it much larger: 500 km or more looks good to me, but only to throw out clearly erroneous observations, and not the baby with the bathtub.

What we're discussing here is estimated or measured accuracy. This bit, together with the XY-coordinates, is shared with GBIF, so every researcher should carefully evaluate which level of accuracy is required for a given question and taxonomic group. If the new, automatic system mentioned below requires to automatically flag observations with low accuracy, it should be de-coupled from being eligible for RG and eventually getting shared with GBIF.

I bet that 99% of researchers interested in the kind of data iNat is generating will go through GBIF rather than assembling them through the various, original data providers, so maintaining a widely open door when transferring data to GBIF is of prime importance, at least to me.
 

 

As a compromise I propose we keep location accuracy as a component of the data quality assessment but increase the threshold to 50km as Jakob proposes. This 50km radius circle has an area about the size of the state of Rhode Island and it seems that (a) the number of observations with such large uncertainties is small enough and (b) the 'harm' they do in making the maps look really funky outweighs the 'good' they do in determining where species occur. Does this sound like a fair compromise?

In previous posts of this thread I've tried to explain why I consider inaccurate observations in several cases as highly relevant and not to be discarded. Except for the extreme cases mentioned in the GitHub thread, I'd like to see some examples where inaccurate observations mess up aggregated species maps. Take a wandering albatross. A humpback whale. A Triops shrimp. Knowing the Where and When within several hundred km can be highly informative in these cases. I'm not a botanist, but I'm sure there are also widely dispersing plant species, and those occurring in vast and rather uniform habitats, where low accuracy is also not an issue in some applications.

 

However, I'd like to also propose that an acceptable use of the  'Does the location seem accurate?' = 'no' flag is to manually flag observations where the centroid doesn't make sense where the accuracy >0km but <50km range on a case by case basis. For example, if someone posts an observation of a salamander with a marker centered in the ocean but a 10km accuracy circle that does overlap the coast, I'd like to be able to manually flag the location as not seeming accurate.

Fully agreed, I'm also using the flag "not accurate", although I'd prefer a different wording. Your salamander case would suggest to me that the actual point is simply misplaced, and that's something very different from a correctly mapped location with a large circle of accuracy. If the observer was unsure of the location, he/she would place it somewhere on land, and with a circle of accuracy encompassing the wider area that was visited (remembered).

 

This is because I'm working on a system to automatically detect observations that are out of range for further review. For example, if a salamander is endemic to California and Oregon but it is observed in Texas, this gets flagged for review. Currently observations with their centroids >5km from the coastline are getting flagged (e.g. its not in Oregon or California but rather the Pacific Ocean) and there's no easy way to incorporate location accuracy into this new system.  So while I'm okay with not having these observations automatically flagged via the <5km filter. I'd like to be able to manually flag them on a case by case basis using the existing 'Does the location seem accurate?' flag so they stop triggering this automatic system. Does this sound fair?

Looking forward to see that system! Sounds pretty focussed on terrestrial biota, or are you doing the reverse, i.e. flagging marine species that are mapped far inland? What about species that are neither terrestrial nor marine? Without knowing what you're actually planning, I would think that this approach requires quite a bit knowledge of the respective ecologies.

In any case, the manual flag certainly sounds fair and useful, but as I said above, I would word it differently - maybe "Location misplaced: y/n"? This would shift the view from the isolated aspect of accuracy to the combined information that we strive to evaluate: XY-location plus circle of accuracy.

Cheers, Jakob

--

You received this message because you are subscribed to a topic in the Google Groups "iNaturalist" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/inaturalist/-JR7FgQWALQ/unsubscribe.

To unsubscribe from this group and all its topics, send an email to inaturalist...@googlegroups.com.

To post to this group, send email to inatu...@googlegroups.com.
Visit this group at https://groups.google.com/group/inaturalist.
For more options, visit https://groups.google.com/d/optout.

 

---

Please consider the environment before printing this email
Warning: This electronic message together with any attachments is confidential. If you receive it in error: (i) you must not read, use, disclose, copy or retain it; (ii) please contact the sender immediately by reply email and then delete the emails.
The views expressed in this email may not be those of Landcare Research New Zealand Limited. http://www.landcareresearch.co.nz

--
You received this message because you are subscribed to the Google Groups "iNaturalist" group.

To unsubscribe from this group and stop receiving emails from it, send an email to inaturalist...@googlegroups.com.

To post to this group, send email to inatu...@googlegroups.com.
Visit this group at https://groups.google.com/group/inaturalist.
For more options, visit https://groups.google.com/d/optout.

--

--------------------------------------------------
Scott R. Loarie, Ph.D.
Co-director, iNaturalist.org
California Academy of Sciences
55 Music Concourse Dr
San Francisco, CA 94118
--------------------------------------------------

--

You received this message because you are subscribed to the Google Groups "iNaturalist" group.

To unsubscribe from this group and stop receiving emails from it, send an email to inaturalist...@googlegroups.com.

To post to this group, send email to inatu...@googlegroups.com.
Visit this group at https://groups.google.com/group/inaturalist.
For more options, visit https://groups.google.com/d/optout.

--
You received this message because you are subscribed to the Google Groups "iNaturalist" group.

To unsubscribe from this group and stop receiving emails from it, send an email to inaturalist...@googlegroups.com.

To post to this group, send email to inatu...@googlegroups.com.
Visit this group at https://groups.google.com/group/inaturalist.
For more options, visit https://groups.google.com/d/optout.

 

--

--------------------------------------------------
Scott R. Loarie, Ph.D.
Co-director, iNaturalist.org
California Academy of Sciences
55 Music Concourse Dr
San Francisco, CA 94118
--------------------------------------------------

--
You received this message because you are subscribed to the Google Groups "iNaturalist" group.

To unsubscribe from this group and stop receiving emails from it, send an email to inaturalist...@googlegroups.com.

To post to this group, send email to inatu...@googlegroups.com.
Visit this group at https://groups.google.com/group/inaturalist.
For more options, visit https://groups.google.com/d/optout.

--

============================
Charlie Hohn
Montpelier, Vermont

 

--
You received this message because you are subscribed to the Google Groups "iNaturalist" group.

To unsubscribe from this group and stop receiving emails from it, send an email to inaturalist...@googlegroups.com.

To post to this group, send email to inatu...@googlegroups.com.
Visit this group at https://groups.google.com/group/inaturalist.
For more options, visit https://groups.google.com/d/optout.

 

--

You received this message because you are subscribed to a topic in the Google Groups "iNaturalist" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/inaturalist/-JR7FgQWALQ/unsubscribe.

To unsubscribe from this group and all its topics, send an email to inaturalist...@googlegroups.com.

To post to this group, send email to inatu...@googlegroups.com.
Visit this group at https://groups.google.com/group/inaturalist.
For more options, visit https://groups.google.com/d/optout.

[Scott Loarie]

Ok, I reverted https://github.com/inaturalist/inaturalist/issues/1183 with 

https://github.com/inaturalist/inaturalist/issues/1206 should be deployed tomorrow

and made a ticket for adding positional accuracy to the API

https://github.com/inaturalist/inaturalist/issues/1207 

--

To unsubscribe from this group and all its topics, send an email to inaturalist+unsubscribe@googlegroups.com.

To post to this group, send email to inatu...@googlegroups.com.
Visit this group at https://groups.google.com/group/inaturalist.
For more options, visit https://groups.google.com/d/optout.

 

---

Please consider the environment before printing this email
Warning: This electronic message together with any attachments is confidential. If you receive it in error: (i) you must not read, use, disclose, copy or retain it; (ii) please contact the sender immediately by reply email and then delete the emails.
The views expressed in this email may not be those of Landcare Research New Zealand Limited. http://www.landcareresearch.co.nz

--
You received this message because you are subscribed to the Google Groups "iNaturalist" group.

To unsubscribe from this group and stop receiving emails from it, send an email to inaturalist+unsubscribe@googlegroups.com.

To post to this group, send email to inatu...@googlegroups.com.
Visit this group at https://groups.google.com/group/inaturalist.
For more options, visit https://groups.google.com/d/optout.

--

--------------------------------------------------
Scott R. Loarie, Ph.D.
Co-director, iNaturalist.org
California Academy of Sciences
55 Music Concourse Dr
San Francisco, CA 94118
--------------------------------------------------

--

You received this message because you are subscribed to the Google Groups "iNaturalist" group.

To unsubscribe from this group and stop receiving emails from it, send an email to inaturalist+unsubscribe@googlegroups.com.

To post to this group, send email to inatu...@googlegroups.com.
Visit this group at https://groups.google.com/group/inaturalist.
For more options, visit https://groups.google.com/d/optout.

--
You received this message because you are subscribed to the Google Groups "iNaturalist" group.

To unsubscribe from this group and stop receiving emails from it, send an email to inaturalist+unsubscribe@googlegroups.com.

To post to this group, send email to inatu...@googlegroups.com.
Visit this group at https://groups.google.com/group/inaturalist.
For more options, visit https://groups.google.com/d/optout.

 

--

--------------------------------------------------
Scott R. Loarie, Ph.D.
Co-director, iNaturalist.org
California Academy of Sciences
55 Music Concourse Dr
San Francisco, CA 94118
--------------------------------------------------

--
You received this message because you are subscribed to the Google Groups "iNaturalist" group.

To unsubscribe from this group and stop receiving emails from it, send an email to inaturalist+unsubscribe@googlegroups.com.

To post to this group, send email to inatu...@googlegroups.com.
Visit this group at https://groups.google.com/group/inaturalist.
For more options, visit https://groups.google.com/d/optout.

--

============================
Charlie Hohn
Montpelier, Vermont

 

--
You received this message because you are subscribed to the Google Groups "iNaturalist" group.

To unsubscribe from this group and stop receiving emails from it, send an email to inaturalist+unsubscribe@googlegroups.com.

To post to this group, send email to inatu...@googlegroups.com.
Visit this group at https://groups.google.com/group/inaturalist.
For more options, visit https://groups.google.com/d/optout.

--
You received this message because you are subscribed to a topic in the Google Groups "iNaturalist" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/inaturalist/-JR7FgQWALQ/unsubscribe.

To unsubscribe from this group and all its topics, send an email to inaturalist+unsubscribe@googlegroups.com.

To post to this group, send email to inatu...@googlegroups.com.
Visit this group at https://groups.google.com/group/inaturalist.
For more options, visit https://groups.google.com/d/optout.

---

Please consider the environment before printing this email
Warning: This electronic message together with any attachments is confidential. If you receive it in error: (i) you must not read, use, disclose, copy or retain it; (ii) please contact the sender immediately by reply email and then delete the emails.
The views expressed in this email may not be those of Landcare Research New Zealand Limited. http://www.landcareresearch.co.nz

--
You received this message because you are subscribed to the Google Groups "iNaturalist" group.

To unsubscribe from this group and stop receiving emails from it, send an email to inaturalist+unsubscribe@googlegroups.com.

To post to this group, send email to inatu...@googlegroups.com.
Visit this group at https://groups.google.com/group/inaturalist.
For more options, visit https://groups.google.com/d/optout.

[Charlie Hohn]

Matt; I'm not arguing that historic data with poor location info is worthless! Just that we shouldn't be continuing to add that type of data now. I guess there are fringe cases but using the app I've never had anywhere near >10 km error unless there was some kind of bug. If you're exploring new ground botanically it's a real shame you can't get good location data. It's so, so useful. 

I don't understand why it's being stated that there is consensus here even though there clearly isn't! (At most three people strongly disliked the feature, and at least three wanted it) So I'll just try to do damage control. Please do prioritize that filter. I guess once we have it I can go through the Vermont ones and mark the really bad ones as mapped wrong. Though that means less id help. 

Scott if it's truly out of the question to consider using a different criteria for plants can we please consider different map display for the bad data? Thanks. 

To everyone else out there, I know some people don't have smartphones but most do, please at least consider either using geotagged photos or one of the apps that lets you attach location to photos later. It's really sad that in 2016 we are going to remote islands and documenting edge of range plants without the capability or will to get a location. This data will be so, so useful later. Seriously. 

Sent from Gmail Mobile

[AfriBats]

Hi Charlie, Jakob here (AfriBats is linking to my iNat-project).

It seems you're very unhappy, and feel I've pushed this through and against the preference of several users. Well, I've presented several detailed arguments why I think that observations should not be excluded from being eligible for RG, and thus for being potentially shared with GBIF, based on low positional accuracy. Matt Goff has added good arguments and examples why one size doesn't fit all. The core issue is to remain inclusive while maintaining all possibilities to do fine-grained analyses with highly accurate data.

Frankly speaking, I don't find very helpful to rant about "bad data" without addressing the conceptual and practical arguments raised in this thread. Yes, having everything mapped to +- a few metres would be lovely. Along these lines, I strive to map my own observations as accurately as possible. And I flag observations as "inaccurate" if they are clearly misplaced (which, by the way, is not the same).

If you guys in Vermont are doing modelling stuff where accurately mapped observations are needed, I'm sure you don't do it based on the iNat interface, but by downloading the data and then filtering them acc to your project goals. There's a column that includes positional accuracy. Set it to < 50 m or whatever you deem appropriate. You don't lose anything, and others can still use less accurately observations from remote regions, historic observations, highly mobile organisms...

I hope you won't spend your time flagging every plant observation in Vermont as inaccurate if all it takes is a few clicks to filter those that you don't want to consider. We need your expertise and time to identify plants!

[Charlie Hohn]

Thanks Jakob,

There are times I was really opposed to a feature too, for instance the almost app breaking auto upload. But features never got deleted because I didn't want them so I don't understand why it happened here. It's nothing personal against you of course, but yes it is incredibly frustrating. Especially being told there was a 'consensus' to do exactly what I felt strongly was the wrong thing to do.

What i really want is the icon to look different to indicate the 'bad' data. Once it gets to GBIF, whatever. If people really want that kind of data, that's their own deal. If the iNat team were up for looking at something like that, i'd be satisfied. But it doesn't seem like they are. 

It's not a matter of lovely. It is literally bad data for doing any kind of natural community mapping or other look at ecology in a mountainous area. The point appears on the map in the wrong place. It's not just imprecise, it's inaccurate. I guess we aren't going to come to agreement there, so I guess i Don't have anything else to say. If the rest of the community is set on keeping that sort of data as research grade, I can't really do anything about it. But I'd like to hear from more than one or two people before features are deleted. So far if you include the github thread as many people wanted as didn't want it.

I don't understand why remote regions require low locational accuracy. I'd never go into remote Africa or the fjords and islands of Alaska without knowing exactly where I was. To do so could literally be fatal. Even in Vermont that can be the case especially in winter. I don't think asking people to map data within 10KM is that high of a bar at all. TBH I don't understand why anyone is entering that sort of data short of not understanding the site.

I thought we weren't supposed to add historic observations to iNat? I mean I don't have a problem with doing so myself, but I thought it was against the rules kind of.

One of the things i am trying to save/fix is the maps that go with each observation and show where others have seen the same species. A filter can't fix those. To be honest even the research grade thing is a compromise to me. I'd have them removed fro the map completely, like they are when you click inaccurate location, or at LEAST given a different symbol.

Again I am not asking for special treatment, i am just asking to not have special treatment given to others either. One final request, maybe keep the RG filter on until the other filters are available? That would at least be something. Right now it feels like me, fine scale ecological mapping, and plants are getting the short end of the stick at the expense of charasmatic megafauna. i don't see why some sort of actual compromise isn't possible.

--

You received this message because you are subscribed to the Google Groups "iNaturalist" group.
To unsubscribe from this group and stop receiving emails from it, send an email to inaturalist+unsubscribe@googlegroups.com.
To post to this group, send email to inatu...@googlegroups.com.
Visit this group at https://groups.google.com/group/inaturalist.
For more options, visit https://groups.google.com/d/optout.

[AfriBats]

Hi Scott

2 suggestions related to the accuracy issue:

Could the flag "Does the location seem accurate?" reworded to something like "Does the location seem misplaced?". I think this captures better what has been discussed here.

Also, it would be great to address the issue of pseudo-accuracy when people search for places with large extents (e.g. protected areas), and then map the observation to wherever the Google place is set - missing accuracy information altogether. Does the Google map interface include anything which tells you the kind or size of the mapped entity? If so, this might be used to include accuracy information.

Thanks, Jakob

[Charlie Hohn]

The google map does have 'accuracy' associated with the places but unfortunately it is often way off

--

You received this message because you are subscribed to the Google Groups "iNaturalist" group.
To unsubscribe from this group and stop receiving emails from it, send an email to inaturalist+unsubscribe@googlegroups.com.
To post to this group, send email to inatu...@googlegroups.com.
Visit this group at https://groups.google.com/group/inaturalist.
For more options, visit https://groups.google.com/d/optout.

[Kent McFarland]

If changing inat comes down to voting, I'd vote that the whole tone of this conversation changes so that maybe more of us would be willing to discuss such issues. Perhaps taking a break and coming back to it might be in order.
Thank you
Kent

On Tue, Nov 29, 2016 at 8:09 AM Charlie Hohn <naturalis...@gmail.com> wrote:

The google map does have 'accuracy' associated with the places but unfortunately it is often way off

On Tue, Nov 29, 2016 at 8:08 AM, AfriBats <jakob...@gmail.com> wrote:

Hi Scott

2 suggestions related to the accuracy issue:

Could the flag "Does the location seem accurate?" reworded to something like "Does the location seem misplaced?". I think this captures better what has been discussed here.

Also, it would be great to address the issue of pseudo-accuracy when people search for places with large extents (e.g. protected areas), and then map the observation to wherever the Google place is set - missing accuracy information altogether. Does the Google map interface include anything which tells you the kind or size of the mapped entity? If so, this might be used to include accuracy information.

Thanks, Jakob

Ok, I reverted https://github.com/inaturalist/inaturalist/issues/1183 with 

https://github.com/inaturalist/inaturalist/issues/1206 should be deployed tomorrow

and made a ticket for adding positional accuracy to the API

https://github.com/inaturalist/inaturalist/issues/1207 

--
You received this message because you are subscribed to the Google Groups "iNaturalist" group.

To unsubscribe from this group and stop receiving emails from it, send an email to inaturalist...@googlegroups.com.

To post to this group, send email to inatu...@googlegroups.com.
Visit this group at https://groups.google.com/group/inaturalist.
For more options, visit https://groups.google.com/d/optout.

--

============================
Charlie Hohn
Montpelier, Vermont

--
You received this message because you are subscribed to the Google Groups "iNaturalist" group.

To unsubscribe from this group and stop receiving emails from it, send an email to inaturalist...@googlegroups.com.

[Charlie Hohn]

ok...

On Tue, Nov 29, 2016 at 8:44 AM, Kent McFarland <kmcfa...@vtecostudies.org> wrote:

If changing inat comes down to voting, I'd vote that the whole tone of this conversation changes so that maybe more of us would be willing to discuss such issues. Perhaps taking a break and coming back to it might be in order.
Thank you
Kent

On Tue, Nov 29, 2016 at 8:09 AM Charlie Hohn <naturalis...@gmail.com> wrote:

The google map does have 'accuracy' associated with the places but unfortunately it is often way off

On Tue, Nov 29, 2016 at 8:08 AM, AfriBats <jakob...@gmail.com> wrote:

Hi Scott

2 suggestions related to the accuracy issue:

Could the flag "Does the location seem accurate?" reworded to something like "Does the location seem misplaced?". I think this captures better what has been discussed here.

Also, it would be great to address the issue of pseudo-accuracy when people search for places with large extents (e.g. protected areas), and then map the observation to wherever the Google place is set - missing accuracy information altogether. Does the Google map interface include anything which tells you the kind or size of the mapped entity? If so, this might be used to include accuracy information.

Thanks, Jakob

Ok, I reverted https://github.com/inaturalist/inaturalist/issues/1183 with 

https://github.com/inaturalist/inaturalist/issues/1206 should be deployed tomorrow

and made a ticket for adding positional accuracy to the API

https://github.com/inaturalist/inaturalist/issues/1207 

--
You received this message because you are subscribed to the Google Groups "iNaturalist" group.

To unsubscribe from this group and stop receiving emails from it, send an email to inaturalist+unsubscribe@googlegroups.com.

To post to this group, send email to inatu...@googlegroups.com.
Visit this group at https://groups.google.com/group/inaturalist.
For more options, visit https://groups.google.com/d/optout.

--

============================
Charlie Hohn
Montpelier, Vermont

--
You received this message because you are subscribed to the Google Groups "iNaturalist" group.

To unsubscribe from this group and stop receiving emails from it, send an email to inaturalist+unsubscribe@googlegroups.com.

To post to this group, send email to inatu...@googlegroups.com.
Visit this group at https://groups.google.com/group/inaturalist.
For more options, visit https://groups.google.com/d/optout.

--

You received this message because you are subscribed to the Google Groups "iNaturalist" group.

To unsubscribe from this group and stop receiving emails from it, send an email to inaturalist+unsubscribe@googlegroups.com.

To post to this group, send email to inatu...@googlegroups.com.
Visit this group at https://groups.google.com/group/inaturalist.
For more options, visit https://groups.google.com/d/optout.

[Mark Rosenstein]

For another use case entirely, look at underwater observations.  GPS does not work underwater, so there is no way to automatically tag photos.  Ideally, I record the location of the ship before a dive.  But often there is no time for that, as it requires taking a GPS-enabled device from inside the metal ship and waiting a few minutes on the top deck for a fix.  And while the captain knows where the boat is, many dive operators consider the locations of dive sites trade secrets, to prevent their competitors from coming to the same sites.  All of which leads me to end up only knowing that we spent a day diving "off the southeast corner of a particular island", but ultimately having about a 10km circle of uncertainty.  Sometimes I do have accurate locations, but often not.

As to whether accurate locations are useful for fish: some are highly mobile and cover large territories, but others are very habitat-specific and will not move more than a few meters their entire lives.  But the inability to use GPS underwater means that the best you can hope for is a description of the micro-habitat and the general area of the dive site.

Given all of this, I would like to see observations included with moderately large degrees of uncertainty.  I understand that for some other uses of iNat, a smaller limit is desirable.  I think the only way to satisfy both is to make sure that the tools carry this accuracy value through all parts of the system, and allow users to specify what they need at the time they do any analysis.

-Mark

[AfriBats]

Hi admins & developers

I see that the automatic accuracy flag is now removed, but those that have been flagged as being inaccurate (> 5 km) are still set to "casual", e.g. www.inaturalist.org/observations/4654458

This one had the same issue, but after adding and then deleting a new ID, it switched to RG.

It looks as if the roll back needs some additional tweaking to get these excluded observations back into the "Needs ID" and "RG" pools.

Thanks, Jakob

[Scott Loarie]

Hi Jakob,

Yes, these need to be re-indexed which is on the list. In the meantime if you manually add and remove a data quality flag on the obs in question you can manually refresh it

--

You received this message because you are subscribed to the Google Groups "iNaturalist" group.
To unsubscribe from this group and stop receiving emails from it, send an email to inaturalist+unsubscribe@googlegroups.com.
To post to this group, send email to inatu...@googlegroups.com.
Visit this group at https://groups.google.com/group/inaturalist.
For more options, visit https://groups.google.com/d/optout.

--

[AfriBats]

Hi Scott

It seems the re-indexing still needs to be done going by this observation: www.inaturalist.org/observations/4655018

I'd also like to bring up the issue of pseudo-accuracy again. Here's a case were 85 observations are mapped to the (Google) midpoint of Isolo NP, Madagascar (click on Map view): www.inaturalist.org/calendar/airalcorn2/2016/11/19

Could you think of a solution that adds an appropriate circle of accuracy to each point (rather than having a false accuracy of 0 m)?

Jakob

[Charlie Hohn]

maybe having the accuracy marked as unknown with a different icon or something would be better than blank. 

[Cullen Hanks]

I have really been out of it to miss this conversation.  I love this, and I'm really excited that accuracy is finally taken into account for RG.  I think 5K is reasonable, and while historical obs do have significant value, this is primarily a tool for documenting observations now and into the future.  

Question: What happens when an observer does not assign accuracy?  

It would be great to provide an incentive to populate the accuracy field.  In many cases, people will map their observations precisely, but won't bother to assign accuracy.  At TPWD, we treat these observations as very imprecise records, basically the county level.  I personally think that observations without accuracy should not be RG.

Cullen

--

[Cullen Hanks]

I have been reading through this discussion, I'm so happy that this is getting attention.  I have had a really hard time educating people about accuracy because there is not a feedback loop in the UI on iNat, and accuracy is not relevant to most observers.  But it is extremely important to us, and our ability to use this data.  Essentially, we can't use observations with high accuracy or no accuracy.

While I get that there are some cases where it is difficult to map precisely, by not raising the bar on accuracy, we are generating crappier data on iNat.  If we did require 5 km for RG observations, we would get more observations with that accuracy.  By not requiring that level of accuracy, then many of my observers in Texas are mapping things to a very course level.  They tell me, "it's RG, that's all I care about".  This is really unfortunate for our mission at TPWD.  

I'm sorry that I'm so late to this conversation.  But I would argue that it would make a lot more sense to not give imprecise observations the same status as precisely mapped and documented observations.  I agree that we want to be able to use imprecise observations, but we also want to provide an incentive for documenting observations well.  I get that some people can't, or aren't willing to take the time and effort to map observations precisely, but we should at least acknowledge those observers that do take the time and effort to map the locations of their observations well.

I'm not clear on where this is heading, but I really hope we can acknowledge and provide an incentive to map observations as precisely as possible, and to document accuracy.  Currently, I don't see this in the user experience.  While observations with imprecise observations have value, they don't have as much value as precisely mapped observations.  Can we please communicate this to our observers in the user experience!!!!!!!

My preference would be to give imprecise observations a second class RG status.  "Imprecise Research Grade", or IRG.  Alternatively, maybe we could highlight observations with an accuracy of less than 1 Kilometer by giving them a gold star.  Whatever, but we need an incentive in the UI to map precisely and document the accuracy.  

This issue has been giving me fits since I started promoting iNat in Texas, and I really hope that we can find a compromise.

Best,

Cullen

[Colin Meurk]

hi iNatters ... I (and colleagues at NatureWatchNZ) totally disagree with this desire to exclude observations from RG that don’t meet some arbitrary precision criterion. All records have to be taken at face value and there is error in all data. One cannot know whether someone (out of the 100 000 observers) has falsely ascribed accuracy where it doesn’t exist anyway. The required accuracy is a moving feast depending on the purpose. Just because your purpose is to map things at <5km accuracy doesn’t mean there aren’t other legitimate purposes. For example, if one wants to generate a precise location for monitoring, or map/list at a suburb, city, district, county, catchment, national park, state or even at country scale, the precision level required varies enormously. And historic records are valuable as part of data sets – at whatever scale is available.

The main reason for assigning RG is so it can be uploaded to GBIF – but the GBIF/TDWG standard (Darwin Core Archive) is capable of storing the meta-data so people can know at what scale the accuracy applies. Regardless of whether on iNat or GBIF, you can filter the data by whatever criteria you want (including accuracy or not stated). And I’m fine with encouraging people, by any incentive you wish, to improve the accuracy of their recording, but I don’t want someone else arbitrarily second-guessing what my needs are. Also I don’t want someone determining that planted/cultivated/domestic records are somehow devoid of ecological information – and therefore not worthy of being given RG status. We the users can make those decisions for ourselves. Thankyou. Colin

To unsubscribe from this group and stop receiving emails from it, send an email to inaturalist...@googlegroups.com.

[Charlie Hohn]

The problem is there is no easy way to get the bad-precision observations off the map or out of the records. i don't want to get too into this again because I found it really frustrating, but the closest to a compromise we got was that there should be filters so that if we wanted we could exclude the imprecisely mapped data... however as far as I am aware you can currently NOT filter the data by precision within iNat. my hope is that it does happen at some point and we can exclude the imprecise observations from projects and also from the ID help page so we don't have to spend time helping with ID if we don't feel the data is useful. With tens of thousands of plant IDs I want to review I'd like to eliminate these poor precision ones because i don't see the value in the data. If we had this set up, Cullen could just set it so you can't add poor-accuracy data to the Texas project. I suppose for GBIF it is ultimately up to them if they want this data or not. 

In any event this was already settled, and this research grade filter is no longer active. 

[Colin Meurk]

sorry to prolong this, but of course we aren't always available to respond at the time these important matters get raised. But do I understand then that the 5km filter was disengaged?  so i'm still a little confused; Charlie when you say you can't remove poor precision data from map - I presume you are talking about the iNat google map, rather than a map which is generated by downloading a search onto a CSV file.  because my recollection is (and sorry if I haven't checked) that all this (accuracy) meta-data is downloaded onto the CSV and therefore one can generate maps from any subset of this. I think I covered the point about relative merit/value of data with varying precision levels (according to the scale one is working at) - 'horses for courses'.  but the answer on the iNat web map would be for a filter based on accuracy to whatever level you want, and a couple of other less precise cut-offs.  I don't know qiute what your application is, but say I wanted to relate plant distributions to soil data, then 5km isn't going to cut it anyway.  one needs accuracy of only a few metres (and for say riparian envts, even less) to do this kind of analysis.

My colleague sent me the following comment: "if you are talking about the necessary data checking/cleaning that should be done before any occurrence data is passed as ‘quality’ and used for a serious purpose then ... [one would have to go a lot further than is currently available in generalised iNat data and processing]".  he sent me the following links (the middle 2 have abstracts that summarise the situation quite well).

http://www.gbif.org/resource/80528

http://www.sciencedirect.com/science/article/pii/S1574954116300577

http://onlinelibrary.wiley.com/doi/10.1111/ecog.02118/full

https://github.com/ropensci/scrubr

regards from down under  Colin

[Charlie Hohn]

Yes, you can download the data and sort it by location precison... sorry not to be clear about that. I am a spatial ecologist and am interested in using iNat 'on the fly' to see, for instance, which species grow in different wetlands and where in the state different oak species occur, things like that. the data is absolutely used for management and bioassessment, but not for publishing scientific papers at this moment. If I have to download each bit of data ever, it's a dealbreaker for me. Really for my purposes anything over 100 meter accuracy isn't that useful, but i can find uses for things as coarse as 5 to 10 km sometimes. Beyond that they are points on the map that show up in the wrong place and cause confusion. I was hoping to have them not be research grade, but the truth is, i don't really care if they are shared with GBIF or not. I just want them off the map (or shown with a different icon) so they don't get in the way of being able to tell where a plant species grows. To that end a filter would be fine. I'd also like to be able to exclude inaccurate mapping from my ID help. And, it would be great to be able to set a threshhold for projects. Things like invasive species maps are pretty much useless beyond 100 meters, knowing a newly detected inavasive plant occurs somewhere within 50 km of burlington, Vermont is literally worse than worthless.

I realize for whatever reason I won't get to have the research grade filter active, but it would be really nice to have the filter. Maybe i need to go bump the github issue.

[Cullen Hanks]

You can also make the case that observations that don’t have a date, or only have a year have value.  They do, but not as much as observations with a specific date.  If date was not required for RG status, you would see more observations without date, or just year.  For me, it is not an issue to filter out these observations.  My issue is that many of my observers in Texas don’t see any feedback relating to the acc field.  For our program, an observation with an imprecise date, like year, has more value than an observation with an imprecise location, like 10 km or more.  Yet I don’t think you can even add an observation with just a year, much less be RG.    

In many cases, their observations are precise, or they could be mapped precisely.  However, unlike with date, the UI does not provide any positive feedback for making the effort to document the location precisely.

The result is that we can’t use a significant percentage of these observations.  GBIF is not the only database to pull the data, and just because they pull the data doesn’t mean it will be used.  For example, our environmental review team reviews over a thousand major development projects throughout the state, some spanning many counties.  This includes the review of the environmental impact of new roads, pipelines, etc.  If an observation is RG AND it has an acc field assigned, then those observations are officially part of that review process.  If there is no acc, or if it ten kilometers, it will not be taken into account in that review process.  

Even if your observations are uploaded to GBIF, observations with imprecise locations will not have the same utility as observations with precise locations.

Time and time again, I talk to observers that have no idea what the acc field is, or why it is important.  Time and time again, I have observers that post observations with vague locations.  When I tell them how this effects how their data can be used, then they change.  The truth is that most observers don’t think about spatial precision because most observers have not been tasked with using spatial data.  

I tell people that it doesn’t matter what the accuracy is, the most important thing is to document the accuracy so people know how to use the data appropriately.  But I also tell people, the more precise the location, the more valuable the observation will be.

I am not suggesting that we exclude observations with imprecise locations from iNat or from GBIF.  However, we will be generating data with more value for future research and conservation efforts if we can do a better job of highlighting the value of precision and the acc field in the iNat UI.  This may not matter to many people, but I can attest that this has direct conservation implications on the ground in Texas.

Best,

Cullen

[Charlie Hohn]

i agree on all of this. Including about the timing! Phylogeny data is nice but not essential. I have several good solid observations (with good location too) that can't get RG because I don't remember exactly when i saw them. 

[Mark Rosenstein]

What would get me to provide accuracy information on my (land-based) observations is if my tool-set supported it.  Who is asking Canon & Nikon to have their cameras with GPS built-in to tag the accuracy along with the coordinates on each photo taken?  Who is asking Adobe and other software companies to work with this data embedded in image files?  Until the tools support it, I'm not recording accuracy.

Secondly, any solution will probably have to have different rules by taxon group.  The plant people have different concerns than those studying birds, etc.

-Mark

[Julien Renoult]

Hi, 

As a researcher in Ecology, I think position accuracy should not be included in a binary Data Quality Assessment. The arguments have been already put forward by Jakob. Each scientific question has its specific requirement in terms at data precision, and we cannot anticipate today what will be the scientific questions of tomorrow.  

For example, for one decade an increasingly important topic in community ecology has been to understand how local and regional factors interact in explaining species assemblages. Researchers have then realized how strong are regional effects on local assemblages: a local community is not only determined by local competition and specialization for accessing ressources but also by the distance to the last glacial refugee and ability to disperse (in particular see works by Jans-Christian Svenning). Most of these studies have used coarse-grain distribution maps (often with 100x100km or 200x200km grids). It is a pitty to imagine that all those precious observations in iNat that have a low location accuracy could not be used for this kind of analyses. 

My best

Julien

On Saturday, 19 November 2016 16:25:48 UTC+1, AfriBats wrote:

Hi developers

It seems there's a new criterion under Data Quality Assessment, which says Position accurate to within 5km?.

Does that mean that all observations with positional accuracy of > 5 km are automatically excluded from becoming Research Grade? If that's indeed the case, I suggest to remove that criterion again, or set it to a cut-off value several orders of magnitude larger.

Posititional accuracy is one of the core data fields that is included in data made available through GBIF. The specific value above which an obervation is useless for a given research question depends on many factors, and cannot be set to a one-size-fits-all value such as 5 km. For instance, we just analyzed the spatio-temporal occurrence of migratory Lepidoptera across Europe, where a grid size of 100 km (and hence a positional accuracy much larger than 5 km) was deemed appropriate for our purposes.

Everyone using GBIF data should carefully evaluate and filter available data before using them in any further analyses, and positional accuracy is obviously one of the central fields in that respect. For that reason I don't see any problems sharing a RG observation with a large circle of accuracy as long as that value is correct. On the contrary, I'm much more concerned with observation that appear to be accurate, but in fact are not. For instance, I see may observations mapped to the Google midpoint of a protected area, e.g. that of Tarangire NP. Several users apparently search for that area, but don't move the map pin to the specific place where the observation was made within the park, or adjusting the circle of accuracy to encompass the entire park.

Jakob

[Charlie Hohn]

ecological natural community research that uses a 100 km grid? I have a hard time understanding how that would work. I imagine improving the resolution of those models would be huge. Here in Vermont we have all kinds of little outlier populations of things like different oak species based on microtopography and other unknown factors... you completely lose that if you lump to vast grids. To say nothing for endemic-rich places like california. Though I admit i am not familiar with that particular research. Despite the coarse precision, it sounds really interesting!

The bottom line is that these coarsely mapped observations clutter up maps and can be harmful when included in some sorts of research as well as when determining range maps. If people want them retained for certain research types and if GBIF wants them, that's their choice, and i will just know not to place too much stock in the GBIF range maps. But can we find an actual compromise instead of just giving up on any way of dealing with these observations? Filters for mapping accuracy and/or map symbology on iNat that indicate the very poor accuracy observations would be a huge step towards dealing with this issue. These measures seem really simple to implement and it would be nice at least to understand why that isn't happening. And I know I am the loudest in discussing this but I am NOT the only one who feels this way, after all i wasn't the one who advocated for the feature in the first place. I know of several others who feel the same way I do.

As iNat grows, especially in terms of very new amateurs, the data changes. It's not a bad thing... just different. But if we want to retain relevance as a place for hard science and solid data as well as outreach and social media, these features will offer a lot of benefit with what seems like little or no downside. The filter could be really simple: exclude observations above a set accuracy, and include or exclude observations with the field blank. And as atlases are developed, excluding them from the atlas detection algorithm after a certain point might be needed too. Though i confess i am uncertain as to the exact end goal of atlases so if i am off there, disregard it.

[Colin Meurk]

As a supporter of both being able to discriminate precision and to see the big picture, I think I support Charlie’s approach of incorporating a feature/functionality in maps that give us the option.  I also seem to have been a lone campaigner (but again knowing many agree J )   for distinguishing ‘wild or spontaneous occurrences of naturally occurring populations’/’ditto for artificially introduced species’/’planted, domesticated or cultivated occurrences’.  yes, some observers may not know which category applies in many cases (the first would be the default) but that is the value of crowd-sourcing. There are experienced people out there who can, if they see something out of place can ‘change’ its status under the ‘Details’ options.

SO … I support Charlie’s call for having filters available on the distribution maps so one can switch to whatever locality precision one wants (we need to agree on the classes – but maybe some exponential series like 10, 100, 1000, 10 000 m would be good).  I haven’t checked iNat to see if you use ft or m?  that could create a slight complication as we use m in nz – and probably in Mexico and Canada versions?

AND  … filters (or ideally permanent colour/shading differentiated pointers/flags for observations) on maps for my 3 (wild/non-local spontaneous/cultivated) status conditions.

Thx c

 

Colin meurk | Research associate
LANDCARE RESEARCH MANAAKI WHENUA
DDI: +64 3 321 9740 | M: +64 27 702 8325
W:www.landcareresearch.co.nz | E: meurkc.landcareresearch.co.nz

 

 

--
You received this message because you are subscribed to a topic in the Google Groups "iNaturalist" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/inaturalist/-JR7FgQWALQ/unsubscribe.
To unsubscribe from this group and all its topics, send an email to inaturalist...@googlegroups.com.

To post to this group, send email to inatu...@googlegroups.com.
Visit this group at https://groups.google.com/group/inaturalist.
For more options, visit https://groups.google.com/d/optout.

[Mark Rosenstein]

Charlie,

You keep saying that low-precision data is not useful.  But that's just in your discipline.  There are other disciplines where an uncertainty of 1000 meters is the norm and would be considered excellent.  Underwater observations for example, where you can't take a GPS fix underwater, and one taken before or after a dive from a moving boat is likely not very close to an observation made during the dive.  Sure, there are microhabitats around a reef that are significant (side of tongue-and-groove formation facing prevailing current, for example) that must be noted by the observer, not from a GPS fix.

I'm not saying iNat shouldn't support the precision you want, but we shouldn't disregard observations from other disciplines that don't meet your requirements.  Any real solution to this problem will involve filters that allow you to look at data with a certain minimum precision, and me to look at data with a looser requirement, but not throwing out that data as lacking enough precision to be useful.

-Mark

[Cullen Hanks]

This is not binary, but the solution needs to be more than just a filter.  Observers should have feedback in some way that indicates that more precise accuracy will have more utility to more kinds of research and conservation applications.  Furthermore, since many observers do not even understand the value of documenting spatial precision, there needs to be feedback to observers that it is important so people know how to use data in an appropriate way.

RG is a powerful feedback tool, but not the only way.  I'm sure the advocates for vague locations can at least agree that there is greater overall utility in observations that are more precisely mapped and have a documented accuracy?

-Cullen

--

You received this message because you are subscribed to the Google Groups "iNaturalist" group.

To unsubscribe from this group and stop receiving emails from it, send an email to inaturalist...@googlegroups.com.

[Charlie Hohn]

@mark, i understand those observations are useful to some people, but am just pointing out that for some taxa, very inprecisely mapped data  is actively harmful to the goals other people and how a couple of very easy change to iNat would solve that problem. I remain confused as to why this is controversial. I understand that you can't easily use GPS underwater and have never advocated for aquatic observations to be thrown away because of poor location precision.  I was originally saying it should not get research grade (but the data would still be there!) but i recognize that doesn't work for some people so i am not advocating for that any more either. All i'm requesting is a way to filter them out and simple map symbology on the range maps that shows the observations that are poor precision so i can disregard them. Right now they show up with the same symbol on the site and can't be filtered out at all. And as for GBIF once i realized they are accepting data with that low level of mapping precision i personally kind of stopped caring about their maps too.

Also what Cullen said

If it makes you feel any better, i don't care about the precise spatial mapping of fish, albatrosses, donkeys, hawks, moose, or any other organism that can move large distances. Plants are a special case, especially trees. When a 100 year old sugar maple is growing on a hill, it tells you that that hill has had suitable conditions for sugar maple survival for a century! This is a huge and near miraculous data point! A population of pine trees on a sky island in Arizona tells you a wonderful story as well, but if the point pops up in a cactus strewn desert it is confusing, frustrating, and obstructive. An invasive buckthorn observation in a new town mobilizes an effort to get rid of it before it spreads. An emerald ash borer mis-mapped 30 km from where it really is could literally cause mass panic and mass amounts of money wasted if people didn't understand how the accuracy bubble worked. I am fully aware that if a bobcat walks through my field it doesn't matter how precise i am because by the time i upload it, it is probably 10 miles away. But aside from radio collar studies no one really tries to gather precise data on where bobcats are.

Since day 1 i have set out to use inat to track species range, phenology, natural community distribution, microclimates, microtopography, and how plants are responding to climate change. I spent lots of time trying to recruit others and help them use the site so they could join in what to me was a shared mission of unraveling stories about the spatial diversity of habitats in Vermont and California. At this point i have to just accept that I am trying to make iNat into something it isn't.  I can still use my data and I guess I can filter for other people's data who map things precisely... but when i click on 'red oak' and watch the dots pop up in the map, which was always a near-miracle to see, i have to accept that a lot of them aren't precise enough to tell me the story i am looking to hear. It has been wonderful how the site has grown so much! However it's in my mind tipped towards students and casual users rather than high level motivated users who add bulk amounts of accurate, precise data. Much of the new site development has been to make the site audience wider, rather than to attract more of the 'super-users' (and to be clear i don't just mean people with pHDs, many of the best of us are self-trained).  I *know* there are others that feel the same way, including other people on the site who have messaged me (i am not running around manufacturing dissent, to be clear) and many potential new users i have described the site to in an entirely positive recruitment pitch that declined to join because some of these simple features aren't included.

 I will probably stay on iNat as long as it exists and I am alive, and since i have control over my own data, I can still use that for whatever. But i find myself spending less time helping with IDs, because a huge number are students using the site under 'duress' or one-time users who don't ever respond to my help on how to tell species apart. I'm spending less time trying to recruit ecologists and hardcore plant nerds to the site, because i find myself justifying things about it to them that i myself don't agree with.

So yeah, i wish all the best to iNat's comunity, admins, curators, etc.. and look forward to staying here... but if anyone knows of other communities more interested in hardcore spatial mapping of plants and such (including in Vermont) (by hardcore i mean bother to use GPS or know where you are within reason) do let me know.

[AfriBats]

Hi Scott, Ken-ichi and developers

The continued and sometimes heated discussion of this thread indicates that many iNat users feel quite strongly about this issue. Which is not surprising given that the Where and When is at the very core of what iNat is recording.

Let me try to summarize what has been suggested, with the hope that there's capacity to move some of these aspects forward:

- agreement to not use a fixed threshold that would exclude observations from being shared with GBIF, EOL etc.
- re-indexing of observations that are currently set to 'casual' due to the interim rule of "accuracy must be < 5 km for RG"
- functionality to filter observations on iNat (eg on observation / identification pages) based on accuracy
- incentives or functionalities to encourage users to enter both precise and accurate locations
- addressing the issue of pseudo-accuracy when users map observations to the midpoints of large areas (eg protected areas) with no circle of accuracy
- suggestion to reword the flag "Does the location seem accurate?" to something like "Does the location seem misplaced?"

Colin: I see your point to be more inclusive concerning the wild/not wild flag, but I would suggest to open a new thread for this to keep this one focussed on the spatial issue.

Best, Jakob

[Charlie Hohn]

Thanks AfriBats! Here's one more:

-functionality to have different icons on range maps (on taxon page) for observations above a set unaccuracy. Perhaps if they are as inprecise as the obscured obs. or more (10 km) they would get the same round circle. Ot just not show up when you zoom in a set amount. Maybe this could be implemented in the Atlas feature instead, that would work for me.

[Colin Meurk]

Incidentally Charlie, how do you handle endangered spp that are automatically obscured? This can be important for rare orchids, lizards or birds’ nests but is also often annoying.

I do support encouragement of precision!

 

Colin meurk | Research associate
LANDCARE RESEARCH MANAAKI WHENUA
DDI: +64 3 321 9740 | M: +64 27 702 8325
W:www.landcareresearch.co.nz | E: meurkc.landcareresearch.co.nz

 

 

--
You received this message because you are subscribed to a topic in the Google Groups "iNaturalist" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/inaturalist/-JR7FgQWALQ/unsubscribe.
To unsubscribe from this group and all its topics, send an email to inaturalist...@googlegroups.com.

To post to this group, send email to inatu...@googlegroups.com.
Visit this group at https://groups.google.com/group/inaturalist.
For more options, visit https://groups.google.com/d/optout.

[Charlie Hohn]

some of them are in a project where i am able to see the true coords, otherwise i mostly ignore them. In terms of conservation, knowing a rare plant is within 10km doesn't do us any good. Which isn't to say I am opposed to obscuring those... it's just that I don't personally have much use for the data. But it's also possible to filter out obscured observations, and the icon is different, and for those two reasons they don't cause as much of an annoyance

[Cullen Hanks]

Hi Jakob,

Thanks for the summary, I think that hits the major points. I suggest you split the following point into two since they are two different issues that too often get confused with each other.

"- incentives or functionalities to encourage users to enter both precise and accurate locations"

to 

"- incentives or functionalities to encourage users to enter precise locations when possible.

- incentives or functionalities to encourage users to document the locational accuracy regardless of how precise the location is."

Best,

Cullen Hanks

Texas Parks and Wildlife

To unsubscribe from this group and all its topics, send an email to inaturalist+unsubscribe@googlegroups.com.

To post to this group, send email to inatu...@googlegroups.com.
Visit this group at https://groups.google.com/group/inaturalist.
For more options, visit https://groups.google.com/d/optout.

---

Please consider the environment before printing this email
Warning: This electronic message together with any attachments is confidential. If you receive it in error: (i) you must not read, use, disclose, copy or retain it; (ii) please contact the sender immediately by reply email and then delete the emails.
The views expressed in this email may not be those of Landcare Research New Zealand Limited. http://www.landcareresearch.co.nz

--

You received this message because you are subscribed to the Google Groups "iNaturalist" group.

To unsubscribe from this group and stop receiving emails from it, send an email to inaturalist+unsubscribe@googlegroups.com.