How can we measure biodiversity on INat!

Hi all! Any thoughts regarding how to quantify biodiversity in comparing a sanctuary to a park? Trying to do some research regarding the impacts of herbicides and how it plays a role in altering biodiversity…Thanks!

1 Like

I think you would need to perform observations (photos, bird watching, insect netting) with a repeatable frequency along a transect or in a defined small area to have properly comparable data.

If you only want to use iNaturalist, then you’ll need to come up with appropriate place definitions (bounding boxes) and decide on a set of statistics that account for different numbers of users/observations in each place. Some organisms imply the presence of others, like

  • a cicada killer wasp hunting even if no cicadas have been observed
  • a picture of an oak gall but no one has submitted the oak tree separately.

One thing to take into account is that iNat data are an indication of the prevalence of observations rather than the prevalence of organisms. Observations by their nature (excuse the pun) require humans to make them. Therefore a place more well-travelled by humans may end up with more observations of the same organism than a difficult-to-reach place (assuming an even spread of the same organism in those two places.) Just another data bias to contend with!


I did it this way, in a study for the planning of a municipal area (49.5 km²), in which, after identifying the habitats present, we wanted to prioritize the different areas:

  1. I have extracted all the data via GBIF;
  2. I have divided the data by species;
  3. I have divided the species by classes (arachnids, insects, mammals, birds, fish, amphibians, reptiles and some families of vascular plants of particular interest, such as orchids)
  4. for each single species I created buffers around the observation, which vary according to the type (500 m for birds and mammals, 200 m for plant species, 100 m for arthropods, fish, amphibians and reptiles);
  5. for each single species I merged the buffers of the observations, thus creating some sort of “ranges” for the species;
  6. I then used an algorithm to transform the vector of the areas / species into a grid, “counting” the number of species present in each cell of the grid (5X5 m in size);
  7. These data, for each group, were subsequently normalized (interval 0-1), added together and again normalized (always interval 0-1) thus creating a spatial index that measures the wealth in species in the area (you get a “heat map”).

Clearly this is not an absolute index, but it allows us to compare fairly close areas (say with a radius of up to 20 km) and in fairly similar conditions.

It is also very likely that this procedure overestimates the number of species in the areas of greatest interest, compared to the number of species in the areas of least interest, since naturalists are attracted, like bees to honey, to the most naturalistically interesting places, so that these interesting places collect in themselves a higher number of observations than the others.

In any case, given the purely purpose of comparison between areas, to identify the areas on which to concentrate conservation efforts, this is in any case an improvement over the “empirical” procedure previously used, which consisted simply of asking the naturalists in the area which are the most important areas by taxonomic group.


Start with the paper about iNat or any citizen science bias, knowing your enemies will help to avoid mistakes in your research.

1 Like

One approach to this problem is to use what are called “saturation curve.” A saturation curve is a way of asking the question, “at approximately what point, as the number of observations in this place increased, would the number of species detected stop increasing appreciably?” The idea is that the first observation in a place will always be a species that has not been observed there before, but as the number of observations increase a smaller and smaller proportion of new observations will be locally novel taxa. The more biodiverse the place, the slower the decline in novel findings will be. In practice this is both complex and fraught, but I know of no better way to use iNat data to estimate the biodiversity of a place.


I’m pretty new to this and have only a few observations, but I think that people like me are more likely to post an observation for something “cool”. Some ants or a boring weed? Pass. A cool bug? Yeah I’ll try and snap a picture. I’m sure that this bias has been addressed elsewhere.

Ants can be cool too. Maybe I’m biased, because I live in a state with over 100 documented species.


Great point to consider. My plan to address this concern is to gather the data myself and select a similar sized region in both locations. While this will exclude certain species that could be present, its the best way to make sure I get all kinds of species (versus just what people choose to post). Then I could use posts from others in the regions to support and/or bolster my own images, of course disclaiming the method of data collection as well.


Great point. I was thinking about choosing a similar sized region in both areas. The project is strictly tracking plants so hopefully the data I select will be ample. Then I can bolster my own research with what others upload in a similar region. I saw how to select a certain location in my project description, but how do you specify certain bounding boxes? Would love if you could help me out with that! Thanks again for your insight – super helpful:)

It seems to me a good idea, since in this way, since the observer is always the same, the problems described above are reduced (eg skipping common species or that the single observer considers not important).
Other useful points could be:

  • if the two areas of investigation are large enough, define for each the same number of sub-areas, chosen at random, in which to carry out the investigation;
  • devote the same time to both areas;
  • make observations in all seasons of active vegetation growth (some species, such as geophytes, appear only at certain times of the year).

That is one one the big problems we have in biodiversity research here in Canada. On a recent podcast the guest was asked about the challenges of biodiversity research in such a large country and she pointed out that most research observations are carried out within 2 km of a road. Canada has huge areas without roads of any sort.

1 Like

A pompillid wasp even if no spiders have been observed.
A host-specific plant pathogen even if the observer has not uploaded a corresponding observation of the host plant.

1 Like

Yet I still haven’t seen a tarantula in New Mexico :slightly_frowning_face:

If the impact of herbicides is the underlying question, I don’t think you can answer it by comparing two sites, unless you have data to show the two sites were virtually similar before the herbicides were used. Even then, you would have to disentangle the effects of herbicide from other differences such as frequency of mowing.

If sanctuary means nature reserve, it was probably designated because it was richer than average in wildlife, so you can’t assume the park is poorer because it has been herbicided. It was probably always poorer in species.


This topic was automatically closed 60 days after the last reply. New replies are no longer allowed.