Atlas

Observations represent the primary species distribution data recorded on iNaturalist, but they only represent presence data (e.g. they tell you where a species is, but not where it is not). iNaturalist also has "listed taxa" which indicate places where species occur (these may be manually created or automatically generated from observations). But listed taxa also cannot be relied upon to understand where a species does not occur because it's unclear how to interpret the absence of a listing. iNaturalist does display "taxon ranges" for a small number of taxa but these generally come from external sources like IUCN and cannot be easily curated by the iNaturalist community.

We are now introducing a concept called an "Atlas." An Atlas defines the set of "atlas presence places" that completely encompass a species' range. As such, like taxon ranges, they can be used to convey information about where a species is not found. But unlike taxon ranges, it is easy for iNaturalist curators to build and curate atlases. As described below, Atlases define a structure that allows the existence of listed taxa (or lack thereof) to be interpreted as both presence and absence data. Atlases leverage the existing listed taxa infrastructure on iNaturalist. In the near future, we plan to use observations and atlases as inputs to model more data-driven iNaturalist taxon ranges for a wider variety of taxa.

Why Atlases

1. Let's be explicit about taxonomic concepts

On iNaturalist, we identify observations by linking them to species or other taxa. But what do we mean by a species? Unfortunately the species' name often doesn't offer a sufficient answer because the same name can mean different things. For example, when I say Black-Tailed Rattlesnake (Crotalus molossus) do I mean snakes in both the orange and yellow ranges below or just the orange areas? It depends on whether you recognize snakes in the yellow area as their own species, Crotalus ornatus, split off from snakes in the orange areas, Crotalus molossus (sensu stricto - meaning in the narrow sense), or whether you consider them all a single species Crotalus molossus (sensu lato - meaning in the broad sense).

When identifying observations to species, it's incredibly helpful to be explicit about which taxonomic concept we're referring to, and often the easiest way to be explicit about taxonomic concepts is by referring to a species range. For example "I mean snakes from both the Sonoran and Chihuahua deserts" means Crotalus molossus (sensu lato), while "I mean snakes from just the Sonoran Desert" means Crotalus molossus (sensu stricto). Ideally we can also separate species out by characters (e.g. "I mean snakes with scales like this"). But range alone can be an extremely efficient way to communicate what we mean by a species.

2. Putting observations in context

Observations are "presence-only data" meaning they are very good at telling us where species are but they aren't as good at telling us where species aren't. If we look at observations alone, we never know if certain areas have no observations because the species doesn't occur there or because no one has bothered to post an observation from there. This makes it hard to put observations in context.

For example, on iNaturalist there are a lot of observations of Gulf Coast Toad (Incilius nebulifer) from Texas. But what if someone posted a Gulf Coast Toad observation from Florida? If Gulf Coast Toad are known to occur in Florida, then observations from that state are to be expected as the iNat community grows. But if Gulf Coast Toads are not known to occur in Florida that means someone probably made an error (identification error or data quality issue like a misplaced location). Alternatively, perhaps we've made an exciting discovery of a previously unknown population. In either case the observation would need review.

Atlases enable both of the above by allowing iNaturalist curators to explicitly define the places that encompass a species' global distribution. For example, knowing that Mexico, Texas, Louisiana, Arkansas, and Alabama completely encompass the global range of what we mean by Gulf Coast Toad, we simultaneously know that (1) (1) we're treating toads that are often called Incilius nebulifer from Belize as a distinct species, Central American Gulf Coast Toad (Incilius valliceps), and (2) we ought to review that observation from Florida because thats either an error or a really exciting discovery...

In a nutshell, Atlases divide the Earth's land surface into a set of continuous, non-overlapping "atlas places" that are either "on" (i.e. "atlas presence places" exist) if the places encompass parts of a particular species' range, or "off" if the places don't overlap with the species' range. But probably the easiest way to understand atlases is to make one.

Getting Started With Atlases

Picking a group to curate and atlas

Before making an atlas, it's important to have a plan. Not every species is suitable to be atlased. Some species we just don't know enough about to be explicit about where they do and do not occur. Likewise, some groups of species have boundaries that are just too poorly defined. For example, maybe we know that a beetle genus has fifty species but all we know about most of them is that there's a single specimen in a drawer somewhere with a name attached to it. That means we don't know where one species population stops and the next begins, or whether their ranges overlap, because there just isn't enough underlying occurence data. Because of this latter issue, I recommend picking a small, well-defined, distinctive group of species and making atlases for each species rather than only atlasing a single species from a group.

The group should have: (1) an explicit taxonomic authority, (2) be distinctive enough that you can help explain to people which observations belong to group and which belong to sister groups, and (3) be relatively well known. Also, don’t be too ambitious - pick a group with a reasonable number of taxa that you can get your head around. And most importantly, pick a group you want to learn about and help other people identify. After all iNaturalist's taxonomic infrastructure is mainly to make it easier to identify observations. This is what sets it apart from many other archives of "species pages" on the Internet. I should also mention that each species can only have one atlas, so you can't atlas a species if it already has one.

I'm picking the sand crab genus Emerita which ticks all the above boxes: (1) they're covered by WoRMS one of iNat's taxonomic authorities, (2) they are easy to distinguish morphologically from their sister taxa, and (3) there are some good, accessible papers describing their distributions and distinguishing characteristics. Also, with just 11 species in the genus, it's not too ambitious a group to take on, and I'm super interested in learning about these crazy critters I've been digging up along the California coast since I was a kid.

Curating the taxonomy

Step one is to curate the taxonomy of the group by adding any missing species, swapping any synonymous taxa into valid taxa, and adding any common names.

Creating Atlases

Before creating atlases, make sure you know what sources you're planning on using for distributions. In this case, I'm using a figure from a 2002 paper that I've adapted below for 9 of the 11 species, and the WoRMS distribution for two recently described species.

The link to create an atlas is in the Curation tab on each species pages. Only curators can create atlases.

Leave the atlas as "Inactive" for now and click "Create."

Ignore any notices at the top of the atlas for now, and focus on the map. This map is made of 253 "country"-level "atlas places" but you won't be able to see them until you click on the map unless they are "on" (i.e. they are "atlas presence places" displayed in red or green). In my Emerita analoga example, clicking within Mexico displays the Mexico atlas place in black while the United states is already displayed as an "atlas presence place" in green. This means that default checklist listed taxa exist for Emerita analoga (or any descendant taxa) in the U.S. (or its descendant admin level 1 "state" and admin level 2 "county" places).

Clicking within Mexico displays two options in the right panel: "Add this place" and "Explode this place." "Add this place" will create a listed taxon for Emerita analoga in Mexico which will turn Mexico green as an atlas presence place. We'll discuss exploding places shortly.

In my example clicking on the U.S. altas presence place reveals three options: "Remove this place," "Show listed taxa," and also "Explode this place."

"Show listed taxa" fetches and displays links for up to ten listed taxa driving the atlas presence place. Remember, the atlas presence place might be driven by listed taxa linked to descendant places or descendant taxa. For example, maybe a listed taxon for a subspecies of Emerita analoga in a California county is driving the United States atlas presence place for Emerita analoga, or maybe it's a listed taxon of Emerita analoga on the United States checklist itself. Drilling into the listed taxa helps determine which is the case.

"Remove this place" is a powerful but also dangerous tool. It will remove up to 100 relevant listed taxa driving the atlas presence place at once. Remember, this might mean removing not just a listed taxon for the species in the country, but up to 100 listed taxa for descendant states and counties and descendant subspecies and varieties. Only use this tool if you are confident that the species does not occur in that place. If you misuse the "Remove this place" tool, you'll have to manually add back all the listed taxa you destroyed which can be alot of work.

Clicking "Explode this place" for a country swaps out the country atlas place and replaces it with its descendant admin level 1 "states." Similarly, you can explode an admin level 1 "state" down to its descendant admin level 2 "counties" or collapse them back together. By exploding the U.S., the atlas is now made up of 303 atlas places (253 countries - 1 US country + 51 US states = 303 atlas places). If I click on other land areas on the map, countries are returned, but because I've exploded the U.S., clicking within the U.S. returns states. Likewise, any US states with relevant listed taxa will be lit up as atlas presence places. By exploding and collapsing atlas places, I can make the atlas as fine or coarse grained as I wish. If I know alot about a species' distribution I might explode the atlas places down to "county" to create altas presence places that capture the boundaries of the species distribution in great detail. Alternatively, I might keep the atlas collapsed down to coarsely describe its range as consisting of the US and Canada. It's up to you to choose the structure of your atlas.

It's often useful to use overlays as a guide when building an atlas. You can display observations, the iNat "taxon range" if it exists, or overlay GBIF observations.

Once you've finished fleshing out your atlas, add a comment describing what sources you used and the reasoning you made behind any choices that might seem arbitrary. Atlases are communally-curated so disagreements should be discussed with respectful comments.

The top blue notice indicates that the atlas is inactive. Now its time to activate it by clicking "Edit" in the top right corner.

Select "Active" and "Update" the atlas.

Once an atlas is active, it will log any changes to the atlas structure (i.e. exploding or collapsing places) in the Recent Atlas Alterations table. Active atlases prevent the automatic creation of listed taxa from Research Grade observations or the automatic creation of listed taxa on ancestor places resulting from the manual creation of listed taxa on descendant places. But they don't prevent the community from creating or destroying listed taxa relevant to an atlas. However, the creation or destruction of any relevant listed taxa appears in the Recent Listed Taxon Alterations table so that curators can keep abreast of any changes that might drive the addition or disappearance of any atlas presence places.

A yellow warning at the top of an atlas indicates that there are observations that fall outside of the atlas presence places. Clicking "view" will display the culprits on the observations search page.

Review these observations and work with the community to determine if they result from errors (e.g. identification errors) or are legitimate range extensions. If the latter, update the atlas (leaving an atlas comment to explain yourself) and if it's a noteworthy discovery (e.g. a country record for a fairly well-known species) make sure to help report this discovery to the iNat community and staff.

Each day, iNaturalist checks each active atlas to determine if there are any "out of atlas" observations. If there are, the atlas is "marked" and displayed on the atlas index page. This is a good place to visit if you want to try to help curate observations that need review.

Deciding which places to include in your atlas

The geographical areas that you choose to include in your atlas should be based on actual evidence that the taxon has been seen, heard, or collected from that locale. Such evidence can include sightings recorded on sites like iNaturalist, BugGuide, WoRMS, eBird, etc.; collection records from museums or online databases like GBIF; and specific locales mentioned in the scientific literature. Use your best judgement and feel free to exclude locales that seem especially dubious or unsubstantiated. (See the sections on dealing with vagrants and "non-established" observations below.) Resist the temptation to fill in "holes" in your atlas. If there is no evidence that a taxon has been reported from a particular area, we shouldn't assume that it's there even if the taxon is reported from the surrounding areas. We can always go back and add that area later once an actual observation has been made. Also remember that it isn't necessary to create an atlas for every taxon. If there isn't much distribution data available for a taxon, it may be best to not create an atlas at all.

Remaining Issues

Fail whaleshark on warning link from long URLs

Sometimes the URL in the link to observation/search used in the "Warning: there are observations not represented by this atlas" is too long and triggers a 500 error. As a temporary workaround, you can replace www.inaturalist.org/ with api.inaturalist.org/v1/ if the error occurs.

Place-related issues

There are some gaps in coverage in the Atlas places that cause some observations to fall through the cracks. We're aware of and working on the following:

  1. iNaturalist's Antarctica place is screwed up (something having to do with the date line) so its not displaying on the Atlas maps

  2. The data provided by CONABIO for Mexico is missing some small islands causing some observations such as this one to trigger flags on atlases such as this one.

  3. There are some gaps in the seams between some adjacent places that allow some observations like this one to fall through.

  4. Adding positional accuracy to the API is needed to ignore observations with >5km accuracy. These may have centroids that fall in the ocean and are therefore interpreted as "out of range."

  5. Our atlas places are terrestrial which don't work for marine taxa. After consulting with various iNat users involved in marine efforts and OBIS, the following is proposed as a set of marine places. Essentially these include places representing the EEZ 200nm waters around each country and FAO Fishing Zones for open ocean.

Vagrancy

Vagrants sometimes show up far out of the normal range for a taxon. While we recognize that the distinction between "core range" and areas where a species is a vagrant is at best poorly-understood and at worst arbitrary, occasionally it may be undesirable to build atlas presence places around all vagrant observations. For example the observation of a Manatee off Massachusetts is so far outside of the normal range for Manatees that rather than include Massachusetts as an atlas presence place, an alternate solution is to flag these observations as vagrant and allow searches for out-of-atlas-observations vagrant observations. This is something we're working on and when it's ready, atlases will be able to ignore observations that users have marked as vagrant.

"Non-established" observations

Sometimes species show up far outside of their natural ranges because they were moved around by people intentionally or otherwise. Unfortunately, like vagrancy, it isn't always possible to tell whether an observation results from an established population in the vicinity or whether it recently escaped from cultivated population. The Brown Anole in North Carolina that is flagging this atlas is a good example. Our preference here, rather than to build new infrastructure for flagging non-established organisms, is to expand the concept of the existing captive / cultivated flag to include "observations likely not resulting from an established population."

Taxonomic uncertainty

This atlas is being flagged in part by this observation in TN. The new 2016 Peterson Eastern Herps (which follows the same SSAR taxonomy as iNaturalist) says that P. alleghaniensis is sometimes impossible to distinguish from P. spiloides, but, also, P. alleghaniensis doesn't occur in TN. There are clear issues with drawing species lines between indistinguishable populations and the circular range-based identifications they incentivize. But those issues aside, our preference is to be conservative and not make identifications in conflict with iNat's taxonomic authorities that trigger out-of-atlas flags unless there is evidence (morphological or genetic) to back up claims of range extensions.

Atlas UI not yet internationalized and translated

This is underway.

Displaying establishment means and observed/unobserved on atlas presence places

On taxon maps throughout the site, establishment means (native/non-native) is indicated by a dotted line and color indicates whether there are research-grade observations associated with a listed taxon (green) or not (orange). Currently, atlas maps are not displaying whether there are observations associated with atlas presence places or not. They are showing establishment means as follows: if any relevant listed taxa have native (or endemic) establishment means the atlas presence place displays as green, if all relevant listed taxa are introduced they display as red. There is currently not good atlas UI for setting establishment means.

Revised on June 17, 2019 19:57 by bouteloua bouteloua