Hello! We have a project that is gathering data for road mortality of certain animals. We are gathering the data this summer with “boots on the ground” research, but also adding any preexisting data we find on iNat. Our study area includes county roads in the United counties of Leeds and Grenville, in Ontario Canada. It would be wonderful if we could get notifications for anytime someone uploads an observation of a alive or dead animal, of the specific taxa we are interested in. Is this possible? Are tags the only way to do this? or can it use GPS?
Thanks for any help at all,
The project link:
Mitigating wildlife-vehicle collisions in Leeds and Grenville · iNaturalist
Project based additional observation fields are likely you’re only reasonable method - you’re describing something which sounds somewhat similar to the Watch for Wildlife - Wildlife Vehicle Collision Tracking Project - the only way to use the GPS data would be if you could export it and then use it alongside data sets you have for the various roadways to determine which observations are eligible - which in itself might be a bit problemed with users who choose to not report the precise location of thier observations.
I’m sure some bigger GIS heads might’ve something more specific to say, but I also don’t quite know how much background or resources you’re coming into this with.
I can’t find it now, but I remember seeing a very similar question asked in this forum before. I believe the answer was that for relatively simple features, one could create a place based on a kml file that includes the narrow bands you want. If I remember correctly, and I don’t know that I do, it was said that kml files of complex shapes (such as everything within 5m of a large network of roads) would not translate well into iNat places. I’m sorry I can’t give you a link to that discussion.
GIS person here, adding to the answers from @wconey and @dlevitis I know one of the big issues hurting iNat speed, is complex place boundaries, so unless your road is a straight line, you’re going to have to do it in two steps.
I think your best bet would be to download the raw data from GBIF of the specific taxa you were interested in (you could do this for a whole Province or county, whatever geography best matches your road network/area of interest.
You might need to filter/clean your observations, for example, if the precision of the GPS is too broad, it’s probably not useful for this application.
You would then need to choose your roads dataset in GIS, buffer by specified distance and clip the point dataset OR
Run a selection by location query with your Records as your target, and roads as selecting features.
Hope that helps : )
this is exactly how I’d do it.
You can check observations from this global roadkill project:
Also note that, because of the GPS accuracy, many roadkill observations are found far from roads. I would choose 500m rather than 5m to not miss all those not precise observations…
Thank you, that is a useful discussion to link to.
@clay, every option here involves a process of:
- Download observations from iNaturalist or GBIF (usually in CSV format)
- Run some GIS tools on those observations
- See if any new observations show up since the previous run
There won’t be any way to get in instant notification that a new observation has shown up. It would be nice, but that option isn’t available here.
How up to date you are depends on how often you run these steps. It is possible to automate the GIS part, but it is also important not to download your observations too often. Also, this is easier to set up if you download all observations (of your taxa and rectangle of interest), but you will cause less server load if you work out a way to only download new observations and then merge them with your existing ones.
As noted in the discussion @pisum linked to, not all observation coordinates are meaningful. In CSV data from iNaturalist, location error is stored in a field called public_positional_accuracy. This isn’t really accuracy, it is possible error in metres. Also, it is often undefined. Whatever your road “buffer radius,” you should screen out observations with a possible error higher than that.
A very common GPS measurement error is 10 metres, and I would suggest your radius at least that size. I have done work looking at flowering plant observations close to walking tracks. Having looked at common measurement errors in the observations, plus measurement errors in the walking tracks, I chose a 100 metre radius, because that was near enough for my project.
As others have suggested, someone with GIS skills can do this fairly simply, once you define the problem the right way around. You may need to pay for their time if they aren’t already a volunteer on your project. However, you shouldn’t need to pay for data or tools; OpenStreetMap hopefully has enough road geometry in your area that you can download through the Overpass API, and QGIS is an open source GIS application that can do the analysis you need. (Some alternative tools exist too; I name QGIS because it’s the one I’m most familiar with.) If you may have ongoing questions of this sort, it’s worth learning QGIS yourself. The learning curve is steep, but there are many things involving iNaturalist observations that suddenly become easy to investigate with it.
In addition to what others have said, and the approaches they have suggested, you are going to have a major issue with location accuracy, both in terms of the specific location and the area of uncertainty as well.
Not only is 5m usually going to be too fine of a scale for the kind of GPS data usually found on iNat, people also are often making observations of things that are not directly under them, so, even setting aside GPS accuracy issues, there is going to be a location offset from the observer location to the observation location as well in many cases.
All that said, the way I would go about it would be to download all the observations of the species in question for the region of interest, and do the selection work offline in a GIS software suite like ArcGIS/ArcPro or QGIS. I’d make a road buffer and simply select the observations within that buffer. If I was being more careful, I’d buffer each observation point with its uncertainty value, giving an area rather than a point, then select the intersections with those buffered observations with the road buffer.