There is the paper about ispot in genera Crowdsourcing the identification of organisms: a case-study of iSpot https://oro.open.ac.uk/42171/
There is also Biodiversity Information Science and Standards 6: e94578
doi: 10.3897/biss.6.94578
Conference How the Citizen Science Platform iSpot Ensures Data Accuracy During and After Collection
Both of these have been mentioned several times in news and other sections on ispot front page. The data that some of this is based on is still very slowly trickling in, only about half of the observations that were sent to irecord many years ago has been verified or otherwise which itself shows how long the systems take. Basically for groups of organisms such as plants, birds, mammals ispot identifications are more or less as good as those from schemes and socs, for more difficult groups the standard may not be as good and part of this might be that species of genus level ID has been given whereas this might not be justified from the photos.
I’m trying to ignore a nagging notion « Never mind the quality, feel the width ». Not sure what you mean by this, are you referring to the data aggregators? They do indeed want as much data as possible but also note whether it is verified or not so that any users of the data can decide what to use it for. There are plenty of different data use scenarios which require different data quality, as an example those checking biodiversity data for planning may rely on verified data but also check unverified data incase there are any important species noted in the area then check those records carefully and potentially carry out their own targetted surveys.
Thank you MikeD; Over 22,00 reads for this Terms of Use link, but it may be the first time for me. I have read and understand the terms and am happy to continue.
I look forward to some enjoyable reading of the papers etc. you have given us and, if I have anything else to contribute, I will do so.
Sorry for any confusion over Never mind the quality… I did not know until I checked on Google it had been a TV sitcom in 1960s; it was a common phrase in our house when I was a child for any stuff we had gathered together in the (not always successful) hope it would meet a perceived need.
Thanks Mike for explaining everything! I had looked through some of the links on the iSpot site and I couldn’t find anything about where the data went (though it could very easily be that I simply missed it while trying to skim read)! I can rest a bit easier now lol
I stopped using GBIF because so many iNat ZA observations are not verifiable, but I know some of the really good observers and are happy to follow them up.
A tricky problem with iSpot crops up when an agreement locks the correction.
See this one (even with Tony’s agreement) that can only be found by following the tag,
Yes, I agree there is a strong element of “Never mind the quality, feel the width”. Organisations like to quote how big their dataset is. I don’t think I have ever heard an LRC stress how accurate their data are. A Lancashire entomologist once said to me he felt wrong records in a dataset are like a turd in a swimming pool. The problem may be small in comparison to the whole, but you don’t really want to use it.
I had always assumed that when using large datasets that statistical techniques would be applied, especially when the quality of the data is not known. Establishing the quality of the data should be the first step in making use of it.
After all you would not take water out of a swimming pool, without filtering and treating if you intended to use it as drinking water.
I read today that every observation on iNaturalist that reaches ‘Research grade’ (which I think means has three agreements) gets transferred to iRecord. I’m not sure whether those records are then checked by an expert first (although I know this is the case for some taxa) before they are then put into the NBN Atlas. I think that I’m right in saying that GBIF operates independently from NBN Atlas. But it does look as if there is room for some of the detritus from the swimming pool to get into the water supply.
I think iNaturalist considers a record Research Grade when a species-level identification has a two-vote lead over other suggestions. So one vote for species A, three votes for species B makes a Research Grade record of species B. Or one agreement with no disagreements will do it. It doesn’t take into account people’s experience - all votes are equal. So someone posts a photo and uses the computer suggestion to call it species A, two of his mates click the agree button, then an expert says no, it is species B: that still gives a Research Grade record of species A.
Thanks for the info on iNat; I now remember why I stopped using it.
I am interested in accuracy in recording; I will continue to post, agree where I can, make helpful comments where I think they may be read; (i.e there are some posters to whom I have added some comment which they ignored, so I assume they didn’t find it helpful; I ignore their posts now).
I can do no other than avoid dipping even my toe into the contaminated swimming pools.
Excuse me, GBIF is far superior to NBN for quick and easy analysis
Someone has written a guide here https://www.ispotnature.org/communities/uk-and-ireland/view/project/834743/
Somewhere in that guide it suggests how you might Preclude iNat’s contributions or see, at ONE click only the UK September 2022 data etc.
I am using NBN quite a lot but, honestly, I find it really really unwieldy.
Did YOU know that iSpot is also uploading to NBN? see iSpot | NBN Atlas (View Records)
Superficially true but iSpot records become Likely (the Equivalent of RG) if the user has 2 icons, many of the s295 (OU) students achieve that on their first (and only) day
Currently iSpot records are being uploaded to NBN - (so far 140 thousand) some of those have no agreement. My stated view, for a long time, is that NO ID in iSpot gets the LB without an agreement - Maybe that should be a four iconned one.
We have had discussions with NBN about somehow carrying over the reputation so that people could select only observations above a certain level but this requires a lot of modification of their system which they have no resources for at present. There are actually other more serious issues which need fixing before this such as not being able to accept our ‘hide precise location’ observations as they don’t have a system for that either (apart from a small number of vulnerable taxa).
If a small group of ispotters are interested in these reputation issues then I could, for example, send a list of ‘low but still likely’ reputation observations and think about how to deal with them.
I don’t think it should be for the NBN to weed out the low quality iSpot data. iSpot shouldn’t be supplying it in the first place. It is all very well saying the user should establish the data is of suitable quality (six posts above) but how are you supposed to check the quality of someone else’s records? And it is far more time-efficient for the data to be checked once by the supplier than for every user to have to do a check.
I don’t think the NBN is the final user, as far as I can tell they are a repository for records which others can use.
.
In the dim and distant past of being involved in research projects, physics not biology, we never trusted any dataset provided or extracted from a third party until it had been tested to an inch of its life, so to speak. A suite of programs would be written to test every aspect of the data, that also went for third party analytical software. Serious users verify data, casual users have to apply their intelligence and use multiple references.
“A” level is as far as I got in physics, but if you are checking every possible aspect of someone else’s data, are you really using their data? It sounds more like you are repeating their experiments. The biological recording equivalent would be not to accept anyone else’s dots on the map until you have visited the site and refound the species for yourself.
Not repeating the experiments or other data collection, but checking its integrity. There are so many ways data can go wrong of not be was is expected. When you use data it is you own responsibility to check that it is what you expected and of the quality expected, because it your findings that are at stake not the providers.
I hope I’m not betraying my ignorance but isn’t biological recording (in the sense of analysing dots on a map) more in the realm of statistics than an exact science? I think that some of the data sets analysed by the British Trust for Ornithology assume that some birds will be wrongly recorded but they also assume that the variance from reality will, at least to a degree, be offset by the random nature of mis-recording. In other words, if you are dealing with a species for which there are many records, it is acceptable to use a less than perfect data set. Certainly in the case of birds in the UK, any really unexpected records are subject to intense analysis by the birding community - I would say that there is perhaps an unwarrented degree of scepticism about many records of rare birds!
In my experience even exact sciences are not exact sciences. You might even say quantum mechanics is all statistics. No measurement is valid without error bars.
.
Does anybody really think that all the observations done for the RSPB garden watch are entirely accurate? There will have been, to use a technical term, a shed load of statistics done to get results with a good degree of certainty.
Not a simple answer about proportion of records with locations hidden as there was a mistake in the early years of ispot and a much higher proportion were hidden than should have been. I have asked several times if users with lots mistakenly hidden would like us to help unhide them but had no response. Anyone can edit their own observations to hide or unhide but if you have large numbers you want to unhide then we might be able to help with that.
I can’t remember the current figure, perhaps 15% but it was above 50% for a period in those early years so it ended up with about 1/4 of the observations I wanted to send in I could not.
