In the UK iNaturalist observations which reach ‘Research Grade’ are being downloaded into iRecord (the UK system) and thence passed to UK expert data validators for checking. It would appear that a few elements of iNaturalist records are causing problems for the UK validators; these problems being poor resolution (records with too large a circle of accuracy) and also the fact that many (if not most) iNaturalist users are using a ‘nickname’ as their iNat user name which makes it almost impossible for UK validators to contact them at times where clarification of details of the record are required.
One of the UK Validators has recently posted on the “National Forum for Biological Recording” Facebook Group and I requested permission to post his text here for all UK iNaturalist users to read and hopefully to absorb - see below:
I have posted this on the iRecord forum, but thought more (different) eyes may see it here…
Just wanted to give a heads up to an issue for fellow iRecord verifiers. I know there has been much discussion about the incorporation of iNaturalist records into iRecord and appreciate the differing views this has elicited and I don’t want to dredge up discussion on the merits or otherwise of doing this. I do however want to specifically draw attention to the fact that some of the incoming records that are blurred, or of very low precision, on iNaturalist are appearing on iRecord with 10km squares allocated that appear to be incorrect, or at best unreliable. Worryingly some such records have already been verified as correct. Apologies for the long post below.
Because each iNaturalist record is captured using lat-long co-ordinates rather than grid ref, each record is allocated a precise central point and a precision (often based on GPS accuracy, but sometimes set from the map by the user). Sometimes this precision radius is HUGE - even covering the whole of the UK. There are lots of imprecise records on iNaturalist. How/why this occurs probably relates to numerous factors - the important thing to note is that it does happen.
There are two other reasons why iNaturalist record precision may be low. iNaturalist has its own list of ‘sensitive’ species (a curated status based partially on IUCN red list criteria I think). Some of these species have their detailed locations automatically obscured to public view, with the centroid marker being plotted randomly within a lat-long ‘rectangle’ covering approximately 25 sq km. Alternatively the recorder may chose to obscure any record’s location, and the record public resolution is blurred in a the same way (random dot in 25 sq km rectangular area).
When the new iNaturalist UK announcement was made this spring it inferred to iNaturailst users that those who affiliated their accounts to the UK portal would facilitate better access to obscured or publicly obfuscated data. However I don’t think this appears to be happening with at least some of the records flowing through iRecord so far. I hope someone from BRC will correct me if I am wrong.
Records coming into iRecord for verification are allocated a grid reference based on the iNaturalist co-ordinates and precision. Records with a precision greater than 500m are being allocated to hectads. Which is all fine. However this appears to be happening for records with MASSIVE precision radiuses. There is a record on there with a precision radius of 12,467 km, and allocated to a precise hectad. Who knows where the location of the record was - could have been anywhere, or the precise ‘dot’ may be perfectly correct - but we don’t know from the information fed into iRecord or when viewing the original iNaturalist record. In my opinion such records should never have been fed through with a hectad grid ref to start with.
In cases where the public resolution is blurred, records when fed into iRecord for verification are allocated a hectad grid reference. Again this would be fine if the original record precision is sufficiently good and the hectad allocation is being done on the originally provided location - not the randomly generated public view marker. However I don’t believe this is happening based on what I have been asked to verify. I have been presented with an Adder record in a purely marine hectad - I have a strong suspicion where the record is from as I know the recorder (and it’s not in the sea!) but the random public marker for the record is in the middle of the sea and that is the hectad allocated to the record. There are marine ladybirds out there too. I have also been presented with moth records from a single observer on the same day (which I believe are highly likely to be from the same trap) covering 4 or 5 different hectads - because they blurred their location and, I believe, the hectads are being allocated from the randomly generated marker dots, not the original location.
I don’t know for certain what is happening - I can only base my assumptions from what I am presented with to verify - but I find this extremely worrying and wanted to alert other iRecord verifiers about this. I have previously emailed my concerns to BRC ands posted on here but haven’t had a response. As far as I am aware records with ridiculously low precision values are still feeding through into iRecord and being allocated unreliably to specific hectads.
Until the situation is clarified, and rectified if indeed there is a problem, I would encourage any iRecord verifiers to be very wary of any low precision records and any iNaturalist records allocated to a hectad grid reference. I have added the ‘coordinate uncertainty in metres’ field to the verification table view (which you can do by clicking the spanner tool in the top right of the table) - this shows the radius of uncertainty associated with the iNaturalist record (which can also be viewed in the record details panel). I can sort by this when I start verifying and quickly deal with any imprecise records. I have chosen to reject any records with a radius of greater than 5000m and then look at any other hectad records and assess these on their merits based on location name, recorder history etc. and make a judgement on the hectad allocation.
I hope BRC can give some clarification on this issue for verifiers. With respect to the records which are blurred by the iNaturalist user, if the hectad is not being generated from the original location provided by the user then in my opinion these records should not be fed through to iRecord at all.
One Facebook correspondent answered the above post with teh following:
I have noticed that some records come through into iRecord with locations that are too imprecise to be of much use for most purposes (particularly when they are for relatively common, well recorded species).
Far too many of the records imported from iRecord also lack a recorders name and just have a ‘username’ attached.
I have changed my own User Name on iNaturalist from “SteveMcBill” to “steve_mcwilliam” so that my observations imported into iRecord are more obvious in terms of who they belong to and who made them. I would suggest that other UK recorders do something similar.and change their names on iNat to the format ‘forename’ undescore ‘surname’ in order to help this issue.
I have also found that uploading records from a phone using the iNat App tends to produce poor location accuracy and I now use my camera with GPS on and upload the images on my desktop computer via the iNat website - this allows me to adjust the positional location on the map and to amend the accuracy circles to a 2 metre size.
Cheers,
Steve
