Zero Inflated Model(s)

Hello all,
I'm working on a paper on the spatiotemporal distributions and abundances of rock pool fish. Over a couple of years, every month, the same pools were checked for fish presence.

The pools were fairly small in size so it was quite easy to identify whether fish were present or not. During the colder months, there were zero counts for fish species. This is expected, going off other literature.

In this case, would a Zero Inflated Model be needed, or not, being as I believe the counts of species to be pretty realistic?

Thank you all.

to check this, you should look at the histogram of the counts. If the histogram e.g. is bimodal (with one peak at zero and a second peak at a value > 0), this is a clear indication that there are actually two different processes producing your data, which should be modelled with zero-inflated or zero hurdle models. If the counts however resemble a Poisson distribution, there is no need for a zero-inflated model.

You can also think about this point from the biological point of view: Do you think that there are actually two different processes generating your data? The first (Bernoulli) process determining if there are fishes or not (success/failure), and the second determining how many fishes there are, given the success. A classical example for the need of a zero-inflated model are numer of nestlings: Often there is one process determining if the brood/nest is successful or not (e.g. flooding events) and other processes determining how many nestlings survive if the nest is successfull (e.g. food availability). Of course, also the second process can generate zeros.
Thank you, mmercker, this makes complete sense! I've looked up Zero Inflated Model stuff on the internet and nothing has spelt it out this simply before. Thank you very much!