I am exploring the use and reliability of iNaturalist data on fish species for a regional report card (ie very good, good, moderate, poor, very poor). To date there has been reporting based on scientific studies of freshwater fish but NOT marine and this is a data gap and of high interest to our community. So far I have established 10 years of baseline data to assist with expected number of species and I then propose to compare observed (research grade observations) against expected number. ie 280 species observed in 2022/23 against 400 expected would be 0.7 and good grade. This is simple (also known as POISE) but one of the challenges is some zones are different sizes and have lots of data and some have little which is a potential bias. I am planning or workshop in mid March and would appreciate any advice from the iNaturalist community
IDs of vertebrates here tend to be highly accurate – but iNaturalist data is a) presence-only and b) opportunistic.
we definitely have divers and other dedicated observers of marine life, but due to the higher barriers to entry they’re much less common than observers of, say, terrestrial flowering plants.
There are indeces like this but they usually require a close-to-full species list of a set taxa, or at least a list of all the dominants. Without further protocol i think there is too much bias for this to work using iNat. For instance one school project could add hundreds of dandelion and clover observations that would drive down the score of an area because these ruderal species usually rank low on ecological health indices. There could be an exemplary bog 100 feet away but since no one went into it, it isn’t represented.
I sometimes use iNat to compile a rough species list that can be used to estimate ecological condition and function of an area, but one has to be at least somewhat systematic. For instance if you said ‘Go into the wetland, stay within the same vegetation community, and keep adding new vascular plant species until you go 5 minutes without finding a new one, then stop’, you can create a pretty decent report card for that vegetation type. But just taking a sample of everyone who walked on a trail near a wetland won’t give you good data for that. The opportunistic data of the iNat fire hose is good for a lot of things, but i don’t think good for an ecological health index, even on regional levels.
I agree with other posters. The process-generating iNat data is highly biased compared to most planned scientific protocols for data gathering (not that those aren’t biased, just less). iNat data is really useful as an indicator of what is present, but comparisons to other datasets will be quite challenging. I don’t know of any studies that have looked at biases in iNat marine data specifically, but I am sure that they will exist. Some hypotheses:
- More observations near popular vacation/dive spots
- More observations of charismatic fauna
- More observations of fauna that are easier to ID
- More observations of fauna that are easier to observe (present in shallow water, not too small to need specialized equipment to photograph, more tolerant of human presence/cameras)
- More observations of diurnal fauna
I am sure that a marine biologist could come up with more.
Another issue in any comparison is whether iNat’s taxonomy meshes with that of whatever database/system it is being compared to.
One thing that might be useful is going through the observations in the target areas in a systematic way and assessing their accuracy and/or adding IDs. It’s possible that some of the species known to exist are already observed, just not IDed, or are IDed incorrectly. This would also let you distinguish whether differences between iNat data and baseline data are due to the iNat observation process or the identification process which would be very useful.
Depending on how ambitious you are, you could have a planned IDathon as part of your workshop where experts would take a pool of observations, add IDs, and assess accuracy or compare data before and after expert IDing to assess the iNat data.
I don’t know if this is closely related to what you want, but the Community Science Department at the California Academy of Sciences holds Snapshot Cal Coast events to record coastal data on iNat pretty often. A few years ago they asked Gio Rapacciuolo work on making usable data from it. It’s well beyond my experience or understanding, but you can see his webinar on YouTube and maybe it can give you some ideas.
Thank you for all your very useful comments and the link to the California study. As background there is data for freshwater fish based on a POISE index of observed vs expected species (plus translocated\exotic species) every 3 years to give scores on A to E with some confidence in our area. There are many more marine fish species and very few scientific methods count all species and generally record limited species\abundance once a year. So currently marine fish scorecards are ‘under development’ and fish are very important for community. So we are having a workshop to see what is possible for citizen science now and in the future. We have data of about 3500 research grade observations of 400+ species for the small area per year. This is often collected using a ReefBlitz or Bioblitz method. https://inaturalist.ala.org.au/projects/dtphw-report-card-2023-fish. Our challenge is to have an indicator where the community can relate to fish as well as scientifically acceptable and comparable over time.
This is a summary of the workshop for naturalists interested din fish and report cards. https://www.youtube.com/watch?v=7NyA_BfWPI4&t=7s