Make users more aware of the accuracy field when uploading observations through the web interface

Feature Requests
1

I have been uploading observations for a few years, and only recently did I learn about the “accuracy” field. A seasoned user pointed out that my observations lacked this field, and that it is actually quite important to include this piece of information. From other posts on the forum I can see that not knowing about this field is actually quite common (see e.g. here and instructions on bulk updating to correct this here). I think that this issue is especially relevant to users whose photos are already geotagged, so they will not be drawing any circles on the map and they will simply not see any numbers in that Acc box, or see any need to fiddle with it.

How do these people currently learn about and add info to the accuracy field? As far as I can tell, you have to click on the location field so that the map pops up (which these users will not do very often as they think their photos are already geotagged). Underneath this map there are multiple fields, one of them reading “Acc (m)”. There is no explanation or indication that this information is useful/important, and from forum posts I think it is clear that this current workflow makes many people overlook (the importance of) this field.

My suggestion is: let’s make the (lack of) accuracy data more obvious - while not too intrusive nor required - for users uploading photos through the web interface.

I am attaching some options for this that could be considered.

OPTION 1: add an indication in each observation box to allude to the (absence of) accuracy data. I am attaching some samples below. In this example, there is an ACC text at the right top of each observation box to indicate whether or not accuracy data are available. Most importantly, when the user hovers over this text, a message is shown what that means and the user can click it to learn more (redirect to a page that describes what this is, why it is important, and how to deal with lacking accuracy data). If the observation already has accuracy data, that icon turns green.

accuracy1a
accuracy1a326×562 89.6 KB

accuracy1b
accuracy1b730×544 166 KB

accuracy1c
accuracy1c313×548 89.7 KB

Alternative scenarios are to only show this icon if accuracy data are missing (so hide if present), and of course exact location and style are to be fully figured out.

OPTION 2: add a box to set accuracy data without having to open the map

This will make it more obvious to users that this exists, and a small text blurb can explain it a bit more right away. In this sample I am only showing it in the box on the left, as not to make the regular boxes too big, although those might also be an option.

accuracy2
accuracy21396×898 170 KB

OPTION 3: add an alert on the page if accuracy data are missing

This is a simple one: it adds a message e.g. at the top or bottom of the page, if one or more observations has accuracy data missing. The drawback is that you don’t easily see which one (perhaps add a button?) and it makes the page a bit less clean.

accuracy3
accuracy3889×490 194 KB

OPTION 4: add more info in the map popup

Alternatively, some additional warnings + info links can be added to the popup map. The ? is a link to learn more, and by adding text to warn people that this is missing (without requiring it when submitting the page) more attention is drawn to this.

accuracy4
accuracy41238×914 159 KB

That is it. I think any of these options would be an improvement over the current system and would help people like me catch this shortcoming in their uploads much, much earlier. I personally prefer option 1, which I think is not very intrusive and yet makes it very clear whether or not this is missing. I would love to hear what other people think.

2

Great suggestion @phoekman. Do make sure to vote for your own feature request!

1 Like
3

In case useful for those weighing your case, I have never heard of the accuracy field before seeing this post.

3 Likes
4

This does sound like a good idea. I thought I’d point out that another way to adjust the accuracy is to click on the map itself – that’ll produce a circle that you can drag in and out, showing the accuracy.

2 Likes
5

Thanks @psweet – that indeed is also possible but (I imagine) not really used intuitively by people whose observations are already geotagged. I imagine that this is most used by those observations that completely lack GPS data.

2 Likes
6

welcome to the community.

i’m not against educating people about the positional accuracy field. that said, what you’re describing here seems like the wrong workflow because it’s more about collection than education. and from a collection perspective, it’s not ideal either because by the time a user sees the proposed indicators, it’s really too late to provide a positional accuracy value that’s anything other than a guess.

frankly, i think the importance of capturing positional accuracy is overstated. if you capture it, that’s fine, but if you don’t capture it, that doesn’t necessarily mean the location data is necessarily worse or unusable. there’s so much variability in how positional accuracy data is captured. there’s no guarantee it will be reliable, nor that the same value captured in 2 observations will necessarily mean the same thing (due to different collection methods). i suspect folks who rely on that value directly (or on the presence of that value) to do data analysis may not really understand the subtleties of that data point.

here are a couple of posts i’ve made in other threads that provide a little more detail about what i’m talking about:

3 Likes
7

Accuracy field is usually blank when a position is coming from a mobile phone or from a camera with GPS or from an overlapping GPS track.
The idea sounds great just to let people know about accuracy field.
As for me, I think that we have more problems with manually positioned pictures coming from cameras. Sometimes people are using very broad localities with accuracy 100 km or even 1000+ km. A warning message could be added in this case as well.

2 Likes
8

Hi, I have recently started ID’ing observations by others, and have become increasingly aware that many new (as well as some experienced) observers, don’t add the accuracy level to the GPS coordinates of their observations. My understanding from Tony Rebelo, leading inat in south africa, is that the accuracy is important to the specialists reviewing and using the data in management of conservation areas.

Is it possible to add an ‘autoresponse’ comment to such users when accuracy level is not recorded, to ask them to add it?

thanks
Santie Gouws
South Africa

1 Like
9

this doesn’t seem like a bug. this is a feature request probably. seems like this could be merged into: https://forum.inaturalist.org/t/make-users-more-aware-of-the-accuracy-field-when-uploading-observations-through-the-web-interface/23494.

my personal opinion is that the need to record positional accuracy is overstated. for most, i think an automatic comment would be more likely to confuse or be ignored than to be helpful.

1 Like
10

hi, I was adding this comment to the features forum, not the bug forum.
I think it is about the content of the message to prevent confusion. I have been adding the following comment (which could certainly be improved upon) to such observations mainly in the Fynbos region in south africa, it is up to the observer to respond or not, but at least a nudge could be useful?

‘Hi, could you please make sure you add the location ‘accuracy’ to this and all your other observations. This information is useful to the specialists utilizing the data. Apart from the GPS coordinates, they also need the degree of accuracy of the measurement of those coordinates. When you look at the location on the map, whether on phone app or on your computer, it should show a little circle around the pin that is indicating the accuracy of the measurement of the location of the observation. If it isn’t there, then in the app, just click on the location, and zoom in. It will automatically start adding a location accuracy. On your computer, you need to click on edit, then on edit again where it shows the location, then you can add it there, the closer, the better, but if you are really not sure where exactly it was, then just make it 100m or 1km or more, whatever is relevant. Remember to save your changes.
Thanks!’

1 Like
11

i think this is factually wrong. maybe folks using data would like a (horizontal) positional accuracy / error value, but they don’t really need it, nor is it recorded consistently enough (in terms of methodology) that i think most people should make much use of it except when it’s exceptionally large (to throw out data).

your message seems to indicate that all of a sudden data will be more useful if you just add an accuracy value or make it arbitrarily smaller. but adding an arbitrary / guessed value after the fact might make the data worse.

2 Likes
12

Hi @dryfveer, I moved your post from the #bug-reports section of the forum to this existing feature request.

1 Like
13

The vast majority of my observations come from my dedicated camera and I use third party software to geotag my photos from a GPX track. Unfortunately the geotagging doesn’t add horizontal accuracy data, just coordinates, so most of my observations lack any horizontal accuracy data. Any horizontal accuracy I add to them would just be a guess and I agree with @pisum that this isn’t necessarily an improvement. I think emphasizing it would a) reduce the number of observations being added because of confusion and it being an extra barrier and b) pressure people into just adding something without thinking much about it.

In my experience (and I could be wrong) most observations with incredibly large accuracy circles come from mobile app users and I honestly can’t figure out how they get those.

3 Likes
14

i think extremely large accuracy circles made via app are often due to folks manually setting location when the map is at a low zoom level. the lower the zoom level on the map, the proportionally larger the circle is on the map – so the larger the accuracy value. (this is different from the web page, where the size of the accuracy circle can be adjusted independently of the map zoom.)

another possibility i think could be if folks choose a large place as the basis of their location. for example, if i choose Houston, TX, USA, as my place, that will give me a location centered on city hall, with an accuracy value a little greater than 50,000 meters.

5 Likes
15

Sorry, I should clarify that I know they’re likely doing this, I just have a hard time picturing someone not zooming in more when two continents are shown in the map. But clearly that’s what’s happening and it’s not clear that the central circle of the location chooser (represents precision/accuracy).

1 Like
16

Hi @pisum! Thanks for your insights. I’ve read the posts you quoted in this thread. You said,

"i would rather have good coordinates without associated positional accuracy than bad coordinates with positional accuracy. "

How would you determine whether coordinates are “good”?

You also said

" leave it to the scientists and analysts to correct for errors in the data they want to analyze".

What techniques would you recommend for finding inaccurate data and correcting errors?

What I’m getting at is… the accuracy field can be useful for throwing out data, as you say. Without this information from data providers, what other strategies can be used to understand data quality? Thanks!

1 Like
17

welcome to the community.

this is the question. ideally, you would be able to to go out and find the same organism at the same spot to independently verify locations. but that’s not realistic in most cases. so without being able to do that kind of verification, you’d have to just trust the data as is or else do some sort of statistical analysis that tries to correct for biases and errors…

i would expect there are techniques for working with “big data” sets that are designed to handle this kind of error correction, but i frankly don’t have the specialized knowledge to really say anything intelligent here. (others might be able to chime in with intelligent answers though.)

the best i can do is talk about some problems i’ve noticed about locations in iNaturalist in general:

  1. you should probably look at the precision with which individual coordinates are recorded and throw out or at least flag the ones that look like they’re recorded below a particular precision. for example, if i were to see a location recorded as lat 29.6, long -90.1, acc 5m, i might scratch my head there because to achieve 5m precision, you really should be recording coordinates (if as decimals) down to 5 or 6 places.
  2. if there are clusters of observations at unusual spots, sometimes those are actually the locations of the nearest cell phone towers or the center point that Google Maps defines for a particular place, not the location of actual observations. you’ll have to decide if those kinds of locations are accurate enough for your particular purposes.
  3. if the observations that are just totally out of range or located in unsuitable habitat, that’s another red flag. sometimes obviously land-based observations get set way out in the middle of the ocean or sometimes you get strange things in the poles.
  4. theoretically, you could evaluate an observation against other observations by that user around the same time (or by other users around the same time, if it looks like they were observing together) to see if there are any inconsistencies in their locations. for example, sometimes, i’ll see that 5 people are observing the exact same plant as part of a bioblitz, and a few locations will be wildly off vs the others. the ones that are wildly off should probably be thrown out.
  5. finally, methods and tools used to record locations vary quite a bit. some people are plotting out locations on maps based on landmarks or trials. some people are just picking the nearest major point of interest. some coordinates are based on what people’s cell phones determine for them. but even within locations determined by cell phones, it matters what kind of cell phone you’re using, what kind of location method your cell phone is using (satellites from set A, satellites from A+B, cell phone towers, wifi hotspots, etc.), whether the phone is using the iNat app to figure out the location or using its own means of sampling location data, how many samples it used to calculate your location, etc., etc…

i’m not sure this provides the answers you were looking for, but i tried to help.

2 Likes
18

@pisum made an excellent list of possible “flags”. I’d only add my particular aggravation:

  1. Any Obscured Location that is the choice by the observer is a indication that the location is suspect. Because iNat randomly plops even a “precise” location within a huge area any geolocation precision is useless - nothing like having the first obs of a newly described species from a Texas county obscured by the user, then “plotted” by iNat to Mexico. Or a upland invertebrate stuck on the border of the Colorado/Kansas flatlands. Same with the amazing number of insects found in the Gulf of Mexico.

A resolution is to encourage users to at least use a public large radius of uncertainty/error where the exact point at least includes a hint of the ecology as well as location. From interactions with “obscurers”, in some cases they have no idea that what they see (the exact location) is not what every user sees - that, to me, indicates a need to educate users about the limitations of obscured locations.

And before the blow back from whatabout women who may be stalked, etc., I find that many of the invertebrate obs are posted by the same user to BugGuide with much more specific location information.

19

Obscured locations are clearly marked as obscured: there is no need to try and find them as being incorrect: just filter by the correct fields when you download or analyse the data.

positional_accuracy
geoprivacy
taxon_geoprivacy
coordinates_obscured
positioning_method
positioning_device

One possibility is planted records not flagged as planted, and instances of species going wild.

Another method is to ask users to add observation fields that can assist, but that is very intensive for the researcher. Of course, for any serious analysis one would overlay the data with geology and land transformation and veld age, and weed out suspect data.

But probably the best return is to make users aware of the importance of accurately recording accuracy - by pointing out if you notice any issues when they post data.

But I dont think you can ever overstate the importance of capturing positional accuracy. It is almost as important as the position itself. Yes there are issues with different cameras and phones exaggerating accuracy: and users inadvertently exaggerating as well. But at least it allows one to immediately sort out accuracies like 1950km or 109km.

(PS: the centroid of google places (2) is a dead giveaway for poor localities)

For us in the Cape with very high beta turnover, lots of threatened species and very small nature reserves, large accuracy values means that observations do not show up for projects and places that they were explicitly collected for. Very frustrating.
Accuracy is one of the first thing we teach teams monitoring populations of threatened species.

4 Likes
20

For me, it happens when I take pictures with iPhone (then upload on mobile) in areas with poor to no cell service.

1 Like