Set 'Does the location seem accurate?' to no when Accuracy: > 5,000m

[loarie]

Currently, my 'outside of atlas' experiments are revealing alot of obs out in the middle of the ocean but with huge uncertainties that do intersect land. I'd like to exclude these from the analysis (preferably by having them be casual).

Since the iNat Standard places have a coastal buffer of 0.05 degrees (which is ~5.5km at the equator) any observation with an accuracy < 5km that overlaps land will always have its center within the coastal buffer (and thus won't raise any 'outside of atlas' flags)

I propose that obs with accuracy >5km are automatically flagged as 'Does the location seem accurate?' = no. We currently keep observations with IDs of family or coarser from becoming RG, having observations with accuracies >5km be casual would follow similar logic (e.g. too coarse to be 'useful').

counter-argument would be that obs with accuracy >5km are still useful for some purposes
counter-counter-argument would be that you can still find them in the casual bin

thoughts @kueda?

e.g. http://www.inaturalist.org/observations/1513176:

[naturalistcharlie]

I strongly support this move. Those observations mess with y beloved range
maps :)

On Fri, Oct 28, 2016 at 3:58 AM, Scott Loarie notifications@github.com
wrote:

Currently, my 'outside of atlas' experiments are revealing alot of obs out
in the middle of the ocean but with huge uncertainties that do intersect
land. I'd like to exclude these from the analysis (preferably by having
them be casual).

Since the iNat Standard places have a coastal buffer of 0.05 degrees
(which is ~5.5km at the equator) any observation with an accuracy < 5km
that overlaps land will always have its center within the coastal buffer
(and thus won't raise any 'outside of atlas' flags)

We currently keep observations with IDs of family or coarser from becoming
RG, having observations with accuracies >5km be casual wouldn't be too much
of a stretch.

counter-argument would be that obs with accuracy >5km are still useful for
some purposes
counter-counter-argument would be that you can still find them in the
casual bin

thoughts @kueda https://github.com/kueda?

e.g. http://www.inaturalist.org/observations/1513176:
[image: screen shot 2016-10-28 at 12 51 23 am]
https://cloud.githubusercontent.com/assets/598026/19799097/9ec72f32-9ca9-11e6-8f82-ad3faeaa15f7.png

—
You are receiving this because you are subscribed to this thread.
Reply to this email directly, view it on GitHub
#1183, or mute the
thread
https://github.com/notifications/unsubscribe-auth/AG1tk714-uDyiRWYEsbxGME1LhEWBF72ks5q4as7gaJpZM4KjLNr
.

Charlie Hohn
Montpelier, Vermont

[loarie]

here's a pretty good example of massive iPhone accuracy issues during a bioblitz - click on 'map' http://www.inaturalist.org/calendar/cwoody1/2016/10/1

[kueda]

I would propose making this not a votable quality metric and instead make it part of a requirement of Research Grade status, like we do with whether or not something is human. The coordinates for http://www.inaturalist.org/observations/1513176 might actually be inaccurate, but we can't determine that automatically. The only thing we can determine automatically is that they are imprecise (and here we're running into the problem of choosing the incorrect name of positional_accuracy when we should have chosen coordinate_precision or coordinate_uncertainty).

So my amended proposal would be to make observations with positional_accuracy set to > 5km Casual Grade.

[loarie]

that approach sounds good to me

[naturalistcharlie]

me too!

On Fri, Oct 28, 2016 at 2:48 PM, Scott Loarie notifications@github.com
wrote:

that approach sounds good to me

—
You are receiving this because you commented.
Reply to this email directly, view it on GitHub
#1183 (comment),
or mute the thread
https://github.com/notifications/unsubscribe-auth/AG1tk2CienxkZ7b_6WuiaK2bB6v8-vsKks5q4kNzgaJpZM4KjLNr
.

Charlie Hohn
Montpelier, Vermont

[loarie]

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

[naturalistcharlie]

Personally i don't find the observations with more than 5km accuracy to be
useful for much anything at all. but i'm kind of a map guy so maybe others
see value in them? Maybe a way to filter them out?

On Fri, Nov 11, 2016 at 7:01 PM, Scott Loarie notifications@github.com
wrote:

This 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, etc.):
[image: screen shot 2016-11-11 at 3 56 22 pm]
https://cloud.githubusercontent.com/assets/598026/20233714/6a7f71d8-a827-11e6-8d23-b2ef6c0fc064.png
We'd be 'throwing out' many 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. But I'd also like to keep as many that
are 'useful' in the verifiable bin

—
You are receiving this because you commented.
Reply to this email directly, view it on GitHub
#1183 (comment),
or mute the thread
https://github.com/notifications/unsubscribe-auth/AG1tk1IfwAdHA_lw-kB7RSNOQtffyA45ks5q9QHegaJpZM4KjLNr
.

Charlie Hohn
Montpelier, Vermont