I work on ENM for some invasive species and use iNaturalist data. The global density map is necessary for weighting and normalization data. Raster map of 1-km resolution would be useful. Is there the kind of map?
as far as i know, nothing like this exists.
the easiest thing for most folks to do is just to download your observations and apply them to whatever grid you want to use for your purposes. the main issue with this approach is that there’s not a great way to download more than 200,000 records at a time from iNaturalist. so if you’re thinking of working with more records than that, it might be easier to get data from either GBIF, the DWCA export file to GBIF, or the AWS Open Data set, although these sources include only observations with certain licenses and also have other limitations.
iNat’s mapping interface (and also the API) does sort of provide a density map, although the grid cells are based on the the map projection and not physical units on the ground, and the density scale colors would not be easy to interpret as specific numbers.
GBIF’s API offers a little more granularity in their map visualizations, though it suffers from the same problem as the GBIF dataset, and it still wouldn’t necessarily be easy to interpret the colors as specific numbers.
you might be able to use the API’s UTFGrid data in a way to approximate your needs, but it just depends on what you’re attempting to do, and it probably would take a lot of upfront work to map the tile grid cells to your preferred grid of choice. so it may not be a great way to try to do super-precise modeling, although it might be good enough for quick real-time visualizations of very large numbers of records.
Thank you! A few words about: why a density map is needed. - For example, working on ENM of some species and having a map of the density of observations, we could better understand the cause for the absence of the species records in any region. If the density of records of other species in the region is high, then the absence of our species is more likely due to the ecological unsuitability of the region. If the density of records of other species in the region is also low, then the cause for the absence of the species could be as insufficient exploration of the region as ecological unsuitability. In addition, the global map of record densities makes it possible to normalize the occurrence of the studied species over the territory. Without this, it is impossible to compare the occurrence of a species in different territories. I think that such a map would be useful for many ENM and SDM researchers.
As for the mapping technique, it is not difficult if there is a complete coordinate base of the all iNaturalist records. The task is to assign the number of known records to each cell of a global 30-second raster represented in a general geographic projection in some GIS format, for example geotif. I believe that the iNaturalist team could easily prepare and present such raster map for common use.This could greatly improve the accuracy of future SDM and ENM.
nothing is ever difficult if you’re not the one doing the work, right?
there are endless of permutations of the data that you could come up with which someone might reasonably want to use as their basis for normalization. maybe some people want to exclude data with some level of geoprivacy applied. maybe some folks might want only certain taxa (ex. just animals or just plants). maybe some folks might want data for a given date range. maybe some folks want to exclude casual observations.
i suppose you could ask the iNat folks to create a mechanism to create a density raster for you on the fly, but that seems like a relatively niche use case but resource intensive thing.
There are only two points I would like to note:
There are an almost unlimited number of scenarios for using any open data, including for SDM and ENM purposes. Yes, data preparation can be done relatively easily (usually) for any of them. But in my mind, it’s rather odd to expect someone to do it instead of the researcher in every case. If I need it, that’s my deal and it’s my job to perform the necessary analysis.
The data is not “nobody’s data” except that published under a CC0 (public domain) license. All other iNaturalist data has some kind of restrictions in its use. Some cannot generally be used without permission from the observer, including for research purposes. Should iNaturalist staff address these issues as well? I don’t think so.
So, I suppose the best of the possible has already been done and made available for use by the scientific community. Certainly it has made a big difference in our ability to model ranges and ecological niches. And I’m sure it deserves the most cordial thanks. And I believe that it takes some serious reason to expect anything more from the project.
Actually, this dataset allows anyone to make any kind of observational frequency map with any kind of conditions. And to be assured of at least some reliability and clarity of legal rights to use them. Although, I’m not sure that project participants can be thought of as simple mechanical devices for recording biodiversity. Many species are observed by a relatively small number of users. And they may have a variety of reasons for their choice of observation locations. There are probably some statistical correlations that can be identified in this as well. But even that can only be done by the researcher themselves based on their objectives. And even so, I doubt that only based on iNatters observations it is possible to draw range boundaries or solve the issue of spatial autocorrelation of the data.
This could be a single GIS raster that would reflect the density of all iNaturalist records. In fact, such a raster reflects the number and activity of database correspondents across the globe. This is not a narrow solution. Without such kind of raster, it is impossible for any ENM specialist to normalize the occurrence of the species under study by region. For example, comparing the number of records of Halyomorpha halys in Europe and Southeast Asia, I cannot understand: is the decrease in the occurrence of the species in Asia due to the lower occurrence of the species, or because smaller number and activity of correspondents? The total global density raster would allow normalization for any species, because it characterizes the number and activity of correspondents for each cell of the Earth’s raster.
I have great respect and gratitude for the creators of iNaturalist. Together with ecological map portals, they make a big contribution to science and practice. My question and suggestion for creating a global record density raster is a recommendation from a scientist who sees ways to improve and enhance the use of the information provided by iNaturalist.
Unfortunately, it is impossible for the user to create a density raster, because the data download limit is currently 200,000 records, while the total number of records stored in the database is currently more than 127 million records. It means that only the iNaturalist team can make and provide such kind of the map for download.
As has already been said, you will have to access iNat data through something else (e.g. GBIF) and create your own density raster. Most “usable” iNat observations are found here https://doi.org/10.15468/ab3s5x in the iNaturalist Research-grade Observations dataset on GBIF. It is regularly updated and currently contains 57 millions observations. This dataset contains only RG observations from some more permissive licenses https://www.inaturalist.org/pages/help#GBIFdata.
You can use almost 60 million RG observations simultaneously (the archive is about 9 gigabytes, with compressed observations.csv file of about 36 gigabytes): http://www.inaturalist.org/observations/gbif-observations-dwca.zip. This is about three quarters of all RG observations and half of Verifiable. Anyone can do this without the help of the iNaturalist team. I just checked (downloaded and opened it) - it works. In doing so, you will have a clear (in both origin and legal terms), easy to cite source. Which will be updated regularly.
This is the way the team is already kindly providing us. As I tried to explain earlier, other ways of doing this in a global context are hardly possible.
There sort of is, but I don’t know how you’d get access to it without downloading all of the observations and recreating it.
Global heatmap of all iNat observations (the distribution of observations is more refined the closer you zoom in, until it changes to individual observatons at a close zoom):
It can be modified by including a username:
Or a taxon (the link is for stony corals, to change replace taxon_id= with appropriate taxon id number):
The functionality appears to exist, but I don’t know how you’d get at it for your purposes.
although you can get the iNat heatmap raster all the way to zoom level 20 or so through the API – the standard iNat mapping interfaces switch to pins at level 10, and somewhere between 11 and 15 is probably what someone would theoretically need for 1km granularity – a heatmap visualization is probably less appropriate than other kinds of visualization for this kind of purpose because it is created in part by effectively blurring out a density grid. it would be better to just get the underlying density grid visualization or, even better, the UTFGrid data.
Thank you! We have prepared preliminary raster using this data and try to use it.
This topic was automatically closed 60 days after the last reply. New replies are no longer allowed.