I am trying to calculate the change in diversity of species encountered throughout time. My dataset is comprised to detections of tagged fish for 48 consecutive weeks. My thought was to bin the detections by time intervals (weeks, two weeks, 4 weeks, etc), to look at the change in diversity of species detected over time. In addition, I think I need to normalize the species counts by the number of tagged individuals. For example, if I observed 10 individuals of Species 1 during week 1, I would divide 10 by the total number of individuals of that species that are tagged to account for issues in the detectability of certain species.

I have two questions. 1. Is it appropriate to treat time intervals like sites when computing diversity indices (Simpson's, Shannon's, etc)? 2. Is it appropriate to use normalized species counts in this diversity indices?

It is my impression that these indices expect count data (integers), or proportion of site (coverage) data, which is not what I would be inputting.

I have attempted to look for similar situations in other studies to see if there is a precedent for handling a dataset like this in this way, but I have not been able to find the answers I am looking for. If anyone has any articles or books they would recommend please let me know.

I hope I explained my situation adequately. I can provide an example dataset if needed, but I figured these questions could be answered without. Please let me know if it does not make sense or if this question would be more appropriate on a different board. I don't know if this matters but I am working with this data in R and have been using the vegan package.

  • $\begingroup$ Ad question 1: I am not sure what do you mean. If you want to see dependence of biodiversity with time, well, just group it by time. What do you exactly mean by "treat time intervals like sites"? $\endgroup$ Commented Mar 31, 2015 at 15:42
  • $\begingroup$ While I think your desire to normalized by counts is a great one, you might consider using sample coverage instead doi.org/10.1890/13-0133.1. It's not at all crazy to bin your data and analyze changes in diversity through time, but there is lots to think about in terms of auto-correlation structure in your time series if you're fitting a model. Addiitonally, if you have enough information to parameterize an occupancy model, you may be able to deal with uncertainty better that way than trying to estimate uncertainty as suggested by @Piotr Migdal's links below. $\endgroup$ Commented Sep 17, 2020 at 17:48

1 Answer 1


Diversity indices use normalized populations, i.e. $p_i=n_i/n$. But as you are using small data, it's crucial to calculate uncertainties (as you will get sampling error, and you want to distinguish real changes from statistical fluctuations).

You can group data as you like, if it is the question.



Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.