How do I handle images of 1 plant across time? For instance, in spring I often see a plant that Seek can’t identify beyond “dicot”. I’ll upload to iNat in hopes for an ID. Then summer will hit and I’ll take more pictures. Then autumn, I’ll take more pics. Should I be creating a new observation or adding the images to the original observation? I feel like adding it to the original observation makes more sense because a user can see all stages of growth for identification.
I would record each as a separate observation, use the “similar observation set” field to associate them, and put an explanation and link to the original observation in the notes for each subsequent observation. These make great sets of observations.
Welcome to the Forum. I don’t have much to say, other than documenting a plant, or whatever, over time is extremely valuable. Some folks have done it with insects, too. Others may have more concrete details about how to do this.
You should record each as a different observation.
Sounds like a good method, but what is the “ similar observation set” field?
Observing multiple times is currently possible, although I choose to reduce or avoid doing so. I see value in having most records be first records of an individual (or in any case not overestimating wild abundance), vs. if everyone observed multiple of the same individuals it might overestimate wild abundances. That said, I recognize iNat abundance can never be an exact estimate of wild abundance. Also if I were to observe an abundant species like a honey bee multiple days in a week in the same location, I view those as justified to upload multiple times since it could be different individuals, and in any case where there’s one there’s a colony, so no risk of overestimating wild abundance.
There are a couple of tricks that can be used to create a “similar observation set” or “linking group”, but the easiest is to make multiple observations, one at each time point, and then add a link to the related observations in the description. I tried one like this here, although I did it sloppily and it never did get conclusively identified.
If I make an observation of a plant flowering I will often go back a while later to the same plant to photograph the fruit. I use the “Related observation” observation field as well as manually editing the location data so that they match exactly
Here is the forum post where I first learned of it and how it can be used.
That makes sense, but iNat really isn’t suited for abundance type analysis, so I don’t see that as a concern.
Bioblitzes, for example, result in hundreds of ‘duplicate’ observations (same organism, different observers), and it’s common for a photo to be taken of a flock of birds treated as a single observation.
iNat is best suited to presence/absence type analysis.
Actually, this was the observation I was looking for - Genus Narina from Pringle Bay, 7196, South Africa on August 06, 2017 at 09:37 PM by magriet b. <h3>Ambush life in a tub</h3> <h4>6 Aug 2017:</h4> I collected an adult ambush bug with 10 eggs … · iNaturalist.ca
I thought this was a good way of handling at least an insect life cycle.
I know the current system results in some duplicate obs., and doesn’t disallow them. But at scales like county, city, region, country, etc, overall I estimate abundance for most individual taxa remains an underestimate. I actually hadn’t thought bioblitzes necessary result in many overestimates of wild abun. (different from duplicates occurring alone), at least at large enough scales like a large city. In my view massive overestimates would be more problematic. I choose/recommend users try to reduce known-duplicates (not including where it can’t be told if it’s the same organism). The abundance data, while not being a direct estimation of wild abun., is still related to wild abun. Some things about wild abun. can be inferred from iNat abun., although if an author were to use iNat data quantitatively/statistically they should 1) acknowledge any stat. sampling biases involved (e.g. more obs. in big cities, near where observers live, etc.), and 2) try to control/reduce their influence on the estimation of wild abun., such as through some statistical method (which alters the number). I’ve heard some users think iNat abun. is completely useless and unrelated to wild abun., but disagree. And actually essentially any kind of research based sampling/observation will also have some sampling biases, yet is used, often acknowledging them and trying to correct for them statistically as well.
I’d be very wary of using frequency of observations in an area as a proxy for actual abundance in the wild. It’s tempting to do so, and it makes sense at first glance, but it ignores the fact that iconic and easy to see species/individuals wind up being massively over represented.
An example, in the area I work there is an island where macaques of various species were dumped back in the 90s to attract tourists. Tour boats bring people to that island on a regular basis, and they have a massively higher observation record than their actual population deserves.
Same with the black kites in our region. They are numerous, but there are many other birds that are more numerous, but they’re most difficult to see, or to get photos of, compared to the slowly circling black kites, so the kites get over-represented in the observations.
Colorful, obvious, iconic, etc species are vastly over represented in many areas and skew the observation results.
Even using location to filter doesn’t really help as people are often not terribly precise in their locations, and people tend to go to the same places, so you get a lot of observations in one area, but not in an adjacent area. And that area people go is usually chosen for some specific reason (eg. for nature tourism because it has something special and unique to see).
The example of bioblitzes was just that, an example, meant to illustrate a point and demonstrate how observations in general get skewed. It wasn’t meant to be taken to mean that bioblitzes by themselves are responsible for the majority of duplicate type observations altering abundance data.
I partially agree/disagree. In both comments I caveated that iNat abundance isn’t the same as wild abun. (for a taxon with numerous obs. at least), and that to use it to infer anything about wild abun. requires noting the sampling biases (multiple different kinds), and ideally also trying to use statistical measures to correct for them. I disagree that iNat abun. has no potentially informative relation to wild abun. (which some implied elsewhere). So for me, knowingly observing a same organism many times contributes to further skew. Although currently it’s also allowed, so observers can decide. You’re also referring to relative abundances between species (vs. I only meant an individual species), that more abundant and conspicuous organisms located nearer to observers are observed more. I agree, I consider that a sampling bias.
You seem to imply a species’ iNat. abun. (at least for a certain small area) can be larger than the actual wild abun. This is possible to occur in some examples, although I’d assume not overly common. Because for example for a social bee or bat observers don’t even see most of the individuals in a colony, so the individuals they do observe (even if observed multiple times) often remain undercounts. I consider undercounts “better” vs. overcounts, since it means the iNat. abun. is at least true (although wild abun. is higher). For example, Rhinolophus malayanus was observed in China recently (outside of iNat). That one record should imply a population is present. Using that record, we could roughly estimate the locations where it occurs (between the last recorded location and the one in China) and the numbers (imprecisely, but at least as an undercount).
I mostly ID bees, for which iNat new locality records are somewhat common, and often consist of a single bee (but which implies populations/colonies are present there). In some cases the expanse between them incudes entire states, e.g. Pseudoanthidium nanum was found in East coast and Midwest US, went undetected in some states and then was found in Oregon. All of them are thought to originate from a single introduction event from Europe (transcontinental in US). Anyway if someone were writing different articles on these species, they can note the differences in sampling biases/circumstances. So they can note your macaque example may be or likely is an overcount, whereas the specific bat and bee species examples are known undercounts, where “nothing is misleading” about iNat abun. (n=1), since we know each are individuals never observed twice, and where each implies colon(ies).
It’s more that it gives a very incorrect view of the relative abundances of species.
And in the case of the macaques here, given that we know exactly how many there really are on the island, yes, the observations massively overrepresent them in terms of absolute abundance as well.
That true, I’m also including that all of that (any bias, any incorrect view) would be factored in. For example, I observed hundreds of individual bees in a few locations (not all of which were uploaded, but assume they were), which represent around 20-30 species per location. If I were to author a publication on it, I’d report the abun. and species richness, and also relative abun., such as rank-abundance analysis. Now, the abun. per species is an undercount per species. Also, there is some distortion to relative abun., correct, although not as much as you may imply. For example, Bombus impatiens and Apis mellifera were by far the most abundant, which actually can be inferred to be true of wild abun. despite that we can’t even know wild abun. precisely using any method(s).
There are also sampling biases related to how some are observed or collected, such as parasitic bees may be less abundant, but are also less visible because they visit few if any flowers. Now, these kinds of ecological monitoring/surveying studies have also been conducted without using iNat and reported on in literature. Essentially, it’s understood that no sampling method (including non-iNat methods like using visual observation, and nets and pan traps) will be precise. There will always be sampling biases unless we can count every single individual somehow and only once each. The results aren’t a “very incorrect” view or what that may imply, but are informative when viewed/caveated correctly at least. For this reason, I suggest against multiple observations of same individuals, but that’s just my optional recommendation.
Basically as a general point, large populations where we can’t observe all the individuals are always going to be difficult to estimate wild abun. for (e.g. even in the literature or survey monitoring). But we have to start somewhere, so what we can learn can be informative when viewed as I described. It can also possibly be combined with modeling or other statistical estimates, such as based on average colony size (e.g. inferred from captive colonies), etc. I assume in the future abun. estimations will get more precise (though still imprecise). I should also caveat I mostly mean abun. per specific locations which could be sampled by surveyors; estimating abun. for an entire State, country, etc. of course would be much more imprecise. Nonetheless there are also statewide surveys, which using iNat generated a surprisingly high sample size.
Also a general point, very high sample sizes (vs very low) can in some circumstances improve abun. estimates, although we also have to think about possible overestimation etc. If I observe or collect 30 bees in a meadow little can be told about relative abun. or total abun. per species. If I instead sample 30 per day for 3 months, the indivdiual per species and relative abun. indicate much more, and species are added to the list of richness/diversity which would be unknown prior. Even in that example, some species themselves would probably still go undetected though.
It’s one of the observation fields that exist already. It will ask for an observation ID to link them all together. This is the number at the end of the URL. I usually try to use the first observation in a series to get this number from. Once linked together in this way, it lets you pull up all the observations with this field (example). I think it was originally designed for raising insects through metamorphosis, but I like to use it for plants observed during different seasons and as a way to track phenology over the years.
Observing the same individual at different times can provide a useful record of phenology, of molt progression, of survival, etc. It’s great. To be maximally useful, the records should be tied together in some way. I’ve just been using notes about what other observations are involved, but I’ll to figure out this similar-observation field.
I agree with the several responses welcoming your suggested approach, @tnich - and the use of an observation field like “similar observation set” or “Observation group”. I use the latter with the iNat observation number of the initial observation to connect an individual tree with the insects I find depending on it, e.g. in galls. I add a link to it as a comment. - I’ve limited myself a lot on plant observations when it seemed unlikely to be identifiable due to (apparently) insufficient features. I feel your procedure is the solution that lets me learn more about plants. The number of such observations will be limited simply because it takes some effort to keep track of the individual plants.
That said, I can understand bdagley’s concern about watering down his abundance analysis method because I’m also more interested in aiding research than enjoying pictures. Though, it seems to me that the situation is quite different for plants than for insects. I can see that meaningful abundance information may be gleaned from iNat data for insects. The huge diversity of insects and the limited time interval during which many of them appear should have a randomization effect. - I’m not sure if anybody uses iNat data to analyze plant abundance in a similar way.
I’ve planted a bunch of native plants on our land in addition to a bunch of naturally occurring ones that came in on their own. I know where they are and watch them throughout the year and it’s been incredibly helpful in my ability to identify these species elsewhere during the off season. I haven’t put them all in inat and honestly someone would probably tag them as no further ID needed and knock the taxonomy back because I couldn’t identify them either without knowing where they were. It would be a neat project though. I’ll bet it would be possible to pick out differences in the first sprouts of even difficult taxa like solidago.