I use the Apple Camera app on an iPhone to take photos with embedded GPS coordinates. If I upload an observation with multiple photos, iNat uses the coordinates in the first photo as the location of the observation. The GPS coordinates in the other photos are ignored.
The problem is that the GPS coordinates in the first photo of an observation are often the least accurate. This is because the first photo is most likely to be the first of a chronological sequence of photos. For my particular hardware-software combination, it seems thereās a start-up cost in conjunction with the GPS receiver.
For instance, the first photo I take on any given day usually has poor accuracy (at least 10 meters, often significantly more) while subsequent photos often have optimum accuracy (less than 5 meters). Whenever I take my phone out of my pocket, the first photo in the sequence is likely to be the least accurate.
As an experiment, I switched to a new camera app whereafter the overall GPS accuracy worsened, especially for the first photo in a sequence. It seems that some apps make better use of the GPS receiver than others.
It would be good if iNat made use of all the GPS information in an observation. Barring that, the best workaround (at least in my case) is to take an extra, throw-away photo at the beginning of each sequence.
Iām curious if other cameraphone users have noticed a pattern with respect to GPS accuracy.
Not a solution per se but to speed up convergence (i.e. reach good accuracy faster), some settings of the cameraphone may be tweaked. For example, disabling the āduty-cycleā function (a power-saving feature) of the GPS chipset - possible at least on Android phones in ādeveloperā mode.
Other solutions could be implemented at the iNaturalist level, e.g. a (weighted) average of coordinates and related accuracies, for all geotagged photos in an observation.
You may want to consider one of the many external, either USB or Bluetooth, GPS receivers available now. They take over for the phones GPS when using apps. I sometimes find it necessary when in the mountains. The prices vary wildly, from under $50 to thousands. Iāve used a Garmin GLO with good results on my Android phone.
A good practice is to have a mapping app running in the background. I often have an eBird list running while I record, or use Gaia GPS, or Mapy.cz. Any app that keeps recording your location even when in the background will help ensure that you always have good precision on your photos.
If it doesnāt slow down your workflow too much, could you maybe initiate the observation with one of the later photos to āestablishā the coordinates, then add all the rest in the sequence?
most observations donāt have more than one photo. theoretically if you take a photo within the iNat mobile app, it does try to get a reasonably accurate location, if the user waits long enough after taking the photo for the device to get a good lock on the satellites. you could theoretically improve the chances of recording a good location a bit by having the phone start to acquire a lock on the satellites as soon as you start the app (rather than waiting until after you take a photo) ā or at least providing an option to do so ā but then none of this helps if you take the photo from outside the iNaturalist app. so in the end, the best solution ā the solution that applies most broadly ā is still probably educate users about how to get good locations on their devices in general.
Honestly for iNat observations the accuracy discrepancy between observations like youāre describing is not really important or relevant.
If you were doing a survey as part of a research project that required sub-meter accuracy, then it would be relevant, but for iNat the relevant accuracy range is in the tens of meters to around a hundred meters even for sessile species.
If itās really bothering you itās easy to edit the location afterward for more accuracy. That will be the fastest approach to as you can do it on a laptop via the browser very quickly.
This may be what is relevant for your uses of iNat, but Iāve lost count of how many times another observerās ~5m accuracy (which is what I usually achieve) has helped me re-find an unusual observation in the field for verification and further documentation. Tens or hundreds of meters would have made that a much more time-consuming - if not impossible - endeavor.
In general it is quite funny how accurate iNaturalist is to most public sources. The government run data base in my state in Germany for example lets you add precise points but filling out the accuracy field, the lowest possible value is 100m. This is kind of a historical remnant from when people didnāt have satellites telling them where they are and having to use maps or verbal descriptions instead.
But I agree that it would be nice if the App could take positional data from all photos and either average the position or just take the coordinates from the picture with the highest accuracy.
I think the right place to start would be in this file for the upcoming iNat Next app https://github.com/inaturalist/iNaturalistReactNative/blob/a4d2c7ca2498ef2429c0f1b205a14ea579aeb1a7/src/sharedHelpers/parseExif.js#L4
From a quick glimpse I think it already can use exif data from all imported images. It could probably be modified to then use the one with the best positional accuracy but I donāt have time to look into that right now and I am also not familiar with react native.
I agree this is the best, as long as youāre not in danger of running out of battery life.
Have you tried iNaturalist Next? If so, what you can do is import the photos and group them in a batch. The first photo you tap on will be the āfirstā photo in the observation and itās used for the observationās date, time, and location. Then you can drag and drop photos in the order you want when you edit the observation.
Thanks to everyone who commented, suggested a solution, or provided a workaround.
Given the number of observations that have no accuracy value at all, I think itās safe to say that many (most?) users are probably not aware of the GPS accuracy associated with their observations. I tend to agree with @pisum, the best solution is to educate users how to obtain good locations in the first place.
The web uploader displays the observationās locality at the top level. To see the latitude, longitude, and accuracy, the user clicks on the locality. But what if the uploader also displayed the GPS accuracy at the top level? In that case, my problem would be solved since the GPS accuracy of each observation would be obvious. Prior to upload, I would reorder the photos of those observations that actually needed it.
For users who know what GPS accuracy is, displaying it at the top level optimizes the upload process. It also provides a potential educational benefit for other users. By exposing GPS accuracy in the interface, user awareness will likely increase and the quality of location data will likely improve. I suspect the same concept applies to the phone apps currently under development.
This would be great! Maybe a new feature request? The apps already show this when you create an observation, but itās a bit more buried on the web uploader
In most cases iNat data is simply being used for presence/absence and not for going back to find specific observations.
I stand by my statement as being accurate for the general user of iNat.
Personally, I tend to be highly accurate with my observation placements, often going back and adjusting the locations to more accurately reflect where they were. This is in part so I can avoid doubling observations or marking the same individual again.
Personally I would not want to make any guesses or assumptions about the ātypicalā use of iNat, if such a thing can even be defined. What I do feel confident about, though, is that more accurate locations generally expand the universe of potential uses.
My suggestion is a pure GPS app rather than a mapping app, so that it is not always trying to pull cell data to refresh a basemap in the background. This would help with battery life too.
Do you have any GPS app suggestions? I download offline maps so I donāt think this is an issue for the apps I mentioned. If you run in airplane mode and with screen brightness turned down battery drain is minimal.