Missing Spatial Accuracy & Positioning Method/Device Vlues in Data Export

[Derek Broman]

Most of the observations exported from my Project do not have a Spatial Accuracy measurement. I was hoping that the  'Positioning_Method' and 'Positioning_Device' columns would yield information to help sort through those observations, but unfortunately there's almost no information in those columns (just a handful of entries that say 'gps').

I assume the information that's supposed to appear in those columns has to do with how the coordinates were derived, such as Android App, iPhone App, GPS Coordinates, Google Map (clicking on the map), or Google Location (entering a name or address in the 'Where were you?' box).  Am I (generally) correct in that assumption?  If so, I've got a glitch.  If not, I sure would like to see that type of information (that is already present in an observation) be present in a data export.

I'm doing a quick analysis of my project (DFW Carnivores) data and I'm very much interested in using the information for habitat analyses in the future.  The Spatial Accuracy measurement is a fantastic metric to aid in quality control, but I've discovered that there's not always a value given, mobile device GPS can fail miserably, users don't always adjust the value to reflect their confidence in that location, and often times users type in a location in the Google Location box and leave it at that.  I have a volunteer helping me to interpret the data in the hopes to develop a technique to intuitively 'clean up' observations with those shortcomings, but without the Positioning Method/Device values, we don't have much to work with.  I'm certainly up for suggestions!

Thanks & Keep up the Excellent Work!

Derek

[Ken-ichi]

Hi Derek,

The positional_accuracy column will only have a value if one has been
set, so in many cases it will be blank, e.g. if someone adds an
observation on the website and doesn't use the accuracy circle, or if
someone adds an observation by importing a photo into one of the apps
and the photo has no accuracy metadata (they almost never do). If
you're seeing some observations that clearly have positional_accuracy
set but do not have that value in the CSV export, please send the URLs
of these observations

positioning_method and positioning_device were fields that we added
really early on and never ended up using much because they weren't
useful to anyone. They're are basically only set from the apps, since
that's the only situation in which we can be sure a GPS was used to
record the location. They should get set to "manual" when the user
drags the map marker or clicks the map, and we could theoretically set
it to "google" or something when the user uses Google's geocoder,
though it's not doing that now, I don't think.

-ken-ichi

> --
> 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 http://groups.google.com/group/inaturalist.
> For more options, visit https://groups.google.com/d/optout.

[Derek Broman]

Hi Ken-ichi. Thanks so much for the reply.

First, I apologize, I overestimated the number of observations without a spatial accuracy value (~164 of 659).

To address your questions/topics, thankfully I am not 'seeing some observations that clearly have positional_accuracy 
set but do not have that value in the CSV export'.  

If the positioning_method/positioning_device fields could get set to something like manual/google/mobile app/gps that'd be incredibly valuable to assessing quality and establishing credibility. Beating a dead horse but...Without a way to determine how these accuracy-less coordinates were obtained, I would likely have to omit those observations from analysis, especially if want to use fine-scale GIS layers.

I've attempted to determine what practices yield no value for that measurement, but there's almost always an exception to each scenario. Digging through a lot of the information, there appears to be 1 of 3 situations:

1. Photos are being uploaded to the website or app, the 'sync metadata' option is selected, and the photo has spatial metadata but no values for accuracy.

2. The iPhone app is not always calculating and inserting the accuracy measurement.  I haven't seen an observation submitted with the Android app yield no measurement.  I tried various scenarios with my Android app in the hopes to duplicate the issue, but the device GPS measures my position (and accuracy) even if I import a photo with no spatial metadata.

3. Users are manually deleting that measurement from their observation.

On the website, anytime I use the Google geocoder or move a marker/click the map I'm ALWAYS given an accuracy value (perhaps this is specific to Windows?). All exported data has XY coordinates, so that rules out the problem being uploaded photos with no spatial metadata.

Thoughts? Questions?

Cheers, Derek

[Charlie Hohn]

Hi Derek!

Here's what I think is happening a lot of the time from the iPhone app - #2 on your list:

-User takes a photo with the regular camera app, NOT the iNaturalist app.

-User then maunally adds photo to iNaturalist observation via the iNaturalist app and imports spatial data from the photo.

Sadly when a photo is taken with the iphone camera app, accuracy data is not recorded. I don't think inat can do anything about it other than to continue to make the app more and more user friendly so that people use it more rather than the camera app. But it will always happen sometimes, as when someone sees something suddenly and doesn't have time to use the app, or else when someone takes a photo of a landscape or scenery or etc and happen to later see an animal or plant there they want to record to iNat.

I think #1 on your list also happens: users create observations from photos and import the geolocation from the photos.

Ironically I think these are often more accurate than the observations people manually enter, because most people aren't cartographers and it's really easy to misjudge where you saw something if it wasn't near a prominent landmark or somewhere you know very well (such as your yard)

Precise spatial data is important to me as well. One proxy I do use is presence of uncertainty buffers - sometimes they are added by the app and sometimes they are added manually, but my idea is that any user who is attentive and location-conscious enough to manually enter an uncertainty buffer probably has a realistic idea about their locational uncertainty. Then you can just filter out the ones with large uncertainty.  Though, I have seen uncertainty circles appear when someone manually entered a place name (for instance 'eaton canyon') that seem too small and inaccurate.

I've found that for my use of the iPhone app the locations are very accurate, less than 1% of them seem off which is comparable to a Garmin GPS or whatever else I use for non-inaturalist field work.  In teh past importing from the camera app, however, did often add errors. This has lessened in the new iOS but may still be a factor,.

I think your best course of action is to encourage users of your project to use the app and/or enter an uncertainty buffer.

C

--

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

coyot.es/slowwatermovement

[Ken-ichi]

What Charlie said. I also add plenty of photos that don't have
accuracy by taking a bunch of photos as well as a GPS track, and using
the GPS track to add interpolated coordinates to the photos. The
software I use doesn't add any accuracy metadata.

Basically people can send us photos with coordinates via a number of
means, and there's usually no way for us to know where those
coordinates came from. We could do a slightly better job tracking that
info when we can (e.g. when using the website), but there's no
absolute solution to this problem. That's just the nature of the data,
and I think anyone trying to do rigorous spatial modelling with such
data needs to understand that and work around it by encouraging people
to enter accuracy data, like Charlie mentioned, or just assuming it
will happen and developing models that take that into account and
reject spatial outliers that lack precision.

-ken-ichi

[Cullen Hanks]

Great comments. I would just add that many people are not aware of the spatial accuracy field, it is not relevant to most observers. It might be possible to highlight it, for example, you could require it for an observation to be research grade. Alternatively, you could highlight the field in red when it is not filled out. Just a thought.

Other than that, every presentation I do touches on the acc field.

-Cullen