Given an obscured location, I’d like to determine the bounding rectangle of its possible location. The help page describes that it will be placed within a 0.2 x 0.2 degree cell. Are these cells fixed at regular intervals of .0, .2, .4, .6 or .8 degrees from a whole number?
For example, a location of -41.284333, 174.703537 would be within the bounding rectangle -41.2, 174.6; -41.2, 174.8; -41.4, 174.8; -41.4, 174.6?
The reason for wanting to determine this is to restrict the possible location of the observation to approx 400 km2, as opposed to approx 2300 km2 if we draw a circle with radius of a location accuracy figure of approx 27km from its obscured location. We’re building an application that assigns the observations to regions and this will limit the potential regions the observation is assigned to.
Thanks, we’re developing a weed alert tool. We won’t be displaying the observation on a map, just assigning the observation to a region or multiple regions where it might occur and reporting it if people subscribe to records from that region.
Yes, the rectangles are fixed and the observation could be anywhere within the rectangle. People have asked at different times about a GIS layer that has the boundaries of the grid cells, but I don’t know that anyone has ever posted one.
It’s also good to note that the area of the rectangles varies with latitude and isn’t fixed since they are defined by degrees.
Since you’re looking at weeds, presumably these generally shouldn’t be auto-obscured. So one thing you could do is, if you find any taxa that shouldn’t be auto-obscured in any given location but are, is to flag them for curators to unobscure there. You would still need to deal with user-obscured locations which the process you’re proposing sounds like it will deal with about as well as can be done.
Thanks @cthawley, we’re doing some complex spatial joins in the application so should be able to handle this calculation without a specific GIS layer.
We’re pulling the data from GBIF so aren’t accessing iNaturalist directly at the moment. We’re planning to use the informationWithheld field in the DwC-A download to determine if an iNaturalist observation is obscured. It contains text such as “Coordinate uncertainty increased to 27285m to protect threatened taxon” if the location is obscured (we’ll use a regular expression to match the distance which can vary). Let me know if you’re aware of any other way we can determine whether an iNaturalist observation is obscured from the GBIF DwC-A data?
We’ll let the curators know if any taxa shouldn’t be auto-obscured. Since we’re aiming to detect emerging weeds as well as known weeds, we’re pulling through all exotic species, so it could be some of those would be auto-obscured.
That field will definitely work to tell you an observation is obscured, though I think by using the uncertainty value from the text string you’re giving up some accuracy. That accuracy value varies because the size of obscuration grid boxes varies with latitude as they are defined by degrees.
From your initial post, it sounded like you were trying to determine the iNat obscuration grid cell boundaries, which is essentially the best accuracy you can get for obscured locations (without trying to defeat obscuration which isn’t a topic for the forum). This would let you say that the observation could be at any location in that cell. So, for instance, if a place of interest had any part of its area/boundary inside/overlapping the cell, the observation could be in that place. This is probably the more “accurate” approach, but takes more effort.
The observation accuracy from that statement is a maximal value. I believe it is essentially the diagonal length of the obscuration box/grid cell. So it represents the accuracy if an observation had its true location in one corner of the grid cell but the obscured coordinates generated by iNat happened to be in the opposite corner. The true difference between the points is almost certainly lower. You can use the iNat-generated, obscured coordinates with that accuracy value and be certain that the point is within that radius. This is much easier to implement, but the spatial area (circle) associated with the point will be much larger than the corresponding obscuration grid cell.
Depending on your application, you could also consider accessing some data from iNat itself. The GBIF dataset will only have compatible licenses in it. Based on some other posts, including those from staff, that’s probably about 70% of records on iNat, meaning you’d be missing out on 30%. If you are just looking for spatial coordinates to make a map for your use, and don’t need to republish data, or post the data or individual observations, it’s probably ok to use spatial data from any observation on iNat regardless of license. You could probably use the API to pull coordinates for your taxa of interest and only those from licenses not sent to GBIF and incorporate them into your dataset which might increase sampling. iNat data downloads also show if an observation has been obscured (though the specific field is different, and I don’t remember what it is off the top of my head).