range map depiction/production

[lelliott]

Say I'm a near-retired guy and I'm interested in looking into the use of iNaturalist observations for species range documentation/prediction. I have used Arcmap, Erdas IMAGINE, and eCognition extensively, but no longer have access to those licenses. I'm fairly fluent (marginally fluent?) in qgis, gdal, r, grass, saga, python, etc. Are there some references around that can direct my interests in a productive direction? I'm particularly interested in the handling of the absence of negative observations (nobody reporting what they didn't see), and the lack of reported effort (can't tell how long the observer was in appropriate habitat and cognizant of the target group). I know there is a wide range of literature on species habitat prediction/suitability. Is there one in particular that addresses data similar to that available on iNaturalist (or ideally, actually using iNaturalist data).

Sorry about the vague question, it just seems like iNaturalist is unique in its dataset, and I'd like to harvest meaningfully.

[Scott Loarie]

Hi lelliott - 

At the moment, iNaturalist generates 'presence only' data so the most relevant types of distribution models for this type of models are so called 'presence-only' models. There's lots of papers that have been written about those approaches ranging from MAXENT models to the 'Inhomogeneous Poisson Process', here's a nice paper on presence only modeling I was reading recently https://web.stanford.edu/~hastie/Papers/ECOGRAPHY_E0321.pdf

I was also interested in this paper by the same authors which combines both presence-only data and presence-absence data into the same modeling framework:

http://www.stat.berkeley.edu/~wfithian/biasCorrection.pdf

iNat currently can't be used to collect presence-absence data, but thats a direction a lot of people have pushed us to move into. And since its easy to argue that presence-absence data improve distribution models, thats certainly one good reason one could give for why the collection of presence-absence data is important. And its kind of compelling that the above paper proposes a single modeling framework that can make use of both presence-only and presence-absence data (were iNat able to produce both)

Best,

Scott

On Mon, Aug 1, 2016 at 4:05 PM, lelliott <lfel...@gmail.com> wrote:

Say I'm a near-retired guy and I'm interested in looking into the use of iNaturalist observations for species range documentation/prediction. I have used Arcmap, Erdas IMAGINE, and eCognition extensively, but no longer have access to those licenses. I'm fairly fluent (marginally fluent?) in qgis, gdal, r, grass, saga, python, etc. Are there some references around that can direct my interests in a productive direction? I'm particularly interested in the handling of the absence of negative observations (nobody reporting what they didn't see), and the lack of reported effort (can't tell how long the observer was in appropriate habitat and cognizant of the target group). I know there is a wide range of literature on species habitat prediction/suitability. Is there one in particular that addresses data similar to that available on iNaturalist (or ideally, actually using iNaturalist data).

Sorry about the vague question, it just seems like iNaturalist is unique in its dataset, and I'd like to harvest meaningfully.

--
You received this message because you are subscribed to the Google Groups "iNaturalist" group.
To unsubscribe from this group and stop receiving emails from it, send an email to inaturalist...@googlegroups.com.
To post to this group, send email to inatu...@googlegroups.com.
Visit this group at https://groups.google.com/group/inaturalist.
For more options, visit https://groups.google.com/d/optout.

--

--------------------------------------------------
Scott R. Loarie, Ph.D.
Co-director, iNaturalist.org
California Academy of Sciences
55 Music Concourse Dr
San Francisco, CA 94118
--------------------------------------------------

[meu...@landcareresearch.co.nz]

there is a presence/absence custom field that can be added - that will be downloadable in a CSV.  but what we need is an 'absence'-specific coloured map flag so one can see the definitively absent part of distribution of a species. perhaps an empty non-coloured flag would be appropriate.  on a related matter of more sophisticated depiction of distributions; can we please have a different coloured flag for 'cultivated/captive' records. ideally one would also have a means of differentiating naturalised or escaped records - but that would apply to most exotic spp records (as well as indigenous spp that have spread outside their natural range due to initial human agency).

[Scott Loarie]

I'd prefer not to use observations with a custom field as a way of recording absences.

iNat observations are are partially so useful and applicable because they are interpreted as species occurrence records - ie Species X was at location Y at time Z. Because museum specimens are species occurrence records (Lewis & Clark collected Species X at location Y at time Z which is now stuffed in a drawer at the Smithsonian) they have a long history of use by the scientific community and there's a huge amount of infrastructure like GBIF and standards like Darwin-Core thats dedicated to compiling and serving species occurrence data.

Expanding the interpretation of iNat observations to potentially represent either occurrences (Species X was at location Y at time Z) or absence points (Species X was absent at location Y at time Z) would require making a ton of changes to make sure that those observations were being interpreted properly on things like maps, frequency distributions and what get sent to GBIF etc. So for the moment I think its fair to say that posting an observation to indicate absence would be a misuse of the current observation data model.

But that said, I share a strong desire to see iNat as a platform for collecting absence data (in addition to occurrence data) mainly because this information is so powerful for species distribution model. 

But I'd suggest a different data model for recording that data. Occurrence records are dimensionless in space and time. By that I mean that even if the time Z is instantaneous and the location Y is infinitely small it still is meaningful to say that a species occurred at that instant in space and time. The same isn't true for absence data. I can say that a species is absence from an instant in space and time, but its sort of meaningless because the species might be present a second later or just a few inches the the left. 

To me, absence data makes more sense as a sampling context that wraps occurrence data. For example, I could say that within this 10x10x10 meter volume of space between 5:00pm and 6:00pm Species X was absent. If I recorded an observation of a skink from within that volume/time interval and also recorded that it was my intent to record observations of all reptile species from that volume/time interval, then it would even be fair to imply absence (or at least failure to detect) for geckos, snakes and all other reptile species from that volume/time interval.

This is essentially how eBird does it. The unit of sampling is a checklist which is more or less defined by a time interval and a distance traveled. The 'observations' are the species presences detected within that time period and distance traveled (these are what are peeled off and sent to GBIF as occurrences). The checklist combined with the 'observations' can be used to interpret absences for any bird species not represented by the observations  (things are a bit more complicated in eBird by abundance counts, but the idea of the checklist as a sampling context wrapping observations holds with in both cases).

If we want to handle absence data, my preference would be to build a 'survey' data model (serving more or less the role of eBird checklists) that was defined by a volume, time interval and one-or-more target species. If someone creates a survey that is not accompanied by an observation of the target species, it could be interpreted as an absence record. This would add the ability to record absences on top of our existing infrastructure without changing/complicating the existing interpretation of observations (as dimensionless species presences) and would be completely compatible with how eBird does it.

-Scott

--

You received this message because you are subscribed to the Google Groups "iNaturalist" group.
To unsubscribe from this group and stop receiving emails from it, send an email to inaturalist...@googlegroups.com.
To post to this group, send email to inatu...@googlegroups.com.
Visit this group at https://groups.google.com/group/inaturalist.
For more options, visit https://groups.google.com/d/optout.

[Charlie Hohn]

I would LOVE to see plots functionality in the website and app. Say you take a 10 by 10 meter plot (or whatever) and record all plants or inverts or whatever in it. Scientists and land managers do these sorts of things often. Having something like this would offer some real solid expansion of the scientific value of inat to balance the recent efforts to make it available to the newest amateur. I think such a feature would encourage more scientists and "serious professional amateurs" into our community. 

Sent from Gmail Mobile

[Colin Meurk]

Yes – tick that!  along with the time/effort expended

c

 

Colin meurk | Research associate
LANDCARE RESEARCH MANAAKI WHENUA
DDI: +64 3 321 9740 | M: +64 27 702 8325
W:www.landcareresearch.co.nz | E: meurkc.landcareresearch.co.nz

 

--
You received this message because you are subscribed to a topic in the Google Groups "iNaturalist" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/inaturalist/YWfoTemzLdo/unsubscribe.
To unsubscribe from this group and all its topics, send an email to inaturalist...@googlegroups.com.

To post to this group, send email to inatu...@googlegroups.com.
Visit this group at https://groups.google.com/group/inaturalist.
For more options, visit https://groups.google.com/d/optout.

---

Please consider the environment before printing this email
Warning: This electronic message together with any attachments is confidential. If you receive it in error: (i) you must not read, use, disclose, copy or retain it; (ii) please contact the sender immediately by reply email and then delete the emails.
The views expressed in this email may not be those of Landcare Research New Zealand Limited. http://www.landcareresearch.co.nz

[lelliott]

I agree. The spatial/temporal specificity that accompanies observations does seem to make it trickier to determine absence. And I agree, recording absences as an observation seems to me to confuse the issue. On the other hand, if 15 observations of Texas spiny lizards are reported for Travis County during the month of July by multiple observers, and 0 Texas horned lizards are reported, it means something. I'm just not sure how to use the data quantitatively. Fithian et al (2014) starts to get at this, it seems, but not quite. Part of the issue is how to fuzz the spatio-temporal values in a way that allows for meaningful aggregation. And how it's done will also be partly a function of the taxon, as they differ relative to detectability, identifiability, density, and observer interest.

The plot/survey data model would be better, but that still leaves regular observations as an increasingly vast dataset.

Maybe Fithian and Hastie could look at Pokemon Go data as a model? Having never actually played the game, I don't know, but it seems like it might have similar constraints.

[Charlie Hohn]

One form of absence that works for unmoving species like plants is absence of a previously present organism or species. This is where a pokemon go-like augmented reality angle could work 'someday' with observations with very precise locations. You could be on a walk and iNat would note that someone else saw a blooming trout lily in the area near that time in the past, and you could look for it and if it is not there somehow record it as absent. Similar things have been looked at by iNat in the past - trips functionality, and there was that other side-app I helped test a while back.

============================
Charlie Hohn
Montpelier, Vermont

[AfriBats]

Hi Scott

Recording absences in a useful way is obviously a very tricky business, and immensely depends on detectability of the respective group of interest. Maybe iNat could gradually implement the collection of absence data for selected taxonomic groups with 1) a high detectability and 2) with a good sample coverage / completeness of checklists for a given place.

Maybe that's something where iNat could collaborate with Map of Life (http://mol.org). Their aim is to "provide ‘best-possible’ species range information and species lists for any geographic area".

Cheers, Jakob

[Charlie Hohn]

Oh, that Map of Life site looks every neat albeit redundant with iNat. I love mapping stuff, so I hope they are able to use some of my data through GBIF. It looks like maybe they will, though it also looks like many if not most common Vermont plants aren't in their database. 

--

You received this message because you are subscribed to the Google Groups "iNaturalist" group.

To unsubscribe from this group and stop receiving emails from it, send an email to inaturalist+unsubscribe@googlegroups.com.

To post to this group, send email to inatu...@googlegroups.com.
Visit this group at https://groups.google.com/group/inaturalist.
For more options, visit https://groups.google.com/d/optout.

--

[AfriBats]

There's definitely overlap b/w the 2 platforms, but also stuff specific to each. That's why I'm suggesting a collaboration...

[Colin Meurk]

Important additional information to make presence/absence maps more meaningful is a measure of the effort that goes into the observation (time, method, specific search for target etc)

We have existing fields for this

 

Colin meurk | Research associate
LANDCARE RESEARCH MANAAKI WHENUA
DDI: +64 3 321 9740 | M: +64 27 702 8325

W:www.landcareresearch.co.nz | E: meu...@landcareresearch.co.nz

 

--
You received this message because you are subscribed to a topic in the Google Groups "iNaturalist" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/inaturalist/YWfoTemzLdo/unsubscribe.
To unsubscribe from this group and all its topics, send an email to inaturalist...@googlegroups.com.

To post to this group, send email to inatu...@googlegroups.com.
Visit this group at https://groups.google.com/group/inaturalist.
For more options, visit https://groups.google.com/d/optout.