search

EcologyR / Bloomers

Detect and identify potential bloomers in microbial communities
Other
0 stars 0 forks source link

formulas to calculate bloomers #9

Open onadeulofeu opened 1 year ago

onadeulofeu commented 1 year ago

Gibbons_2017.pdf

Pakillo commented 1 year ago

Interesting paper @onadeulofeu !

While you keep reviewing the literature out there, I think we could start with a simple metric based on the definition of bloomer you gave me: a taxon that can increase abundance fast and become dominant in the community for a certain period of time.

We could easily translate that into code: detecting fast rates of change in abundance (in %) and also detect taxa that reach certain threshold in abundance (e.g. 20%, 40%, no idea, should be based on the ecology of these systems). And then decide for how long that dominant abundance should be kept to be considered a bloomer (3 days? 4 days? 1 week? Again based on the ecology of these systems)

Then we could relate this to temporal variation in environmental factors (temperature, nutrients, etc)

iremendoza commented 1 year ago

Hi! I find this idea of Paco but not so easy to implement, because you also need to define a time period for the increase. In the case of daily censuses, maybe is easier, but not for monthly censuses. In any case, the use of wavelets may be a solution:

A basic reference for this is here: https://link.springer.com/article/10.1007/s00442-008-0993-2

Here they also use this method for a blooming species: https://link.springer.com/chapter/10.1007/978-90-481-9541-1_7

There is an R package for working with wavelets (although I have never used it before): https://cran.r-project.org/web/packages/WaveletComp/WaveletComp.pdf

I can try to use these functions if you find this interesting.

onadeulofeu commented 1 year ago

Hi! Thanks Irene! I'll take a look at it! I was trying to define functions that did not depend on the time-scale of the sampling to make it more flexible, but we could try something like this :)

onadeulofeu commented 1 year ago

https://ami-journals.onlinelibrary.wiley.com/doi/full/10.1111/1462-2920.16086 similar to what they do here.

iremendoza commented 1 year ago

I believe that both methods (wavelets or dissimilarity analyses among correlative censuses) are equivalent in terms of time-scale, but we could discuss this further. In any case, my advice is to do it the simplest as possible. I am sure that @ibartomeus thinks something similar ;-)

iremendoza commented 1 year ago

Hi Ona! Could you advance with this? Let us know if you need help.