Hello! What would you like to model exactly? I am not sure about your question.

If you have a list of species abundances for different sites and ecological variables (e.g. temperature, humidity, PH,...) I would suggest you to have a look at ordination methods (principal component analysis, canonical analysis, redundancy analysis, NMDS). If you are a R user, the package "vegan" is very useful for community analysis.

Cheers

Thanks Valérie Coudrain for your best answer....but like your said I have species abundance from different habitat types in a region (e.g. a city) and ecological variables (measured in one season)...

Besides, We already use the principal component analysis, canonical analysis, redundancy analysis, NMDS,UPGMA, optimum tolerance value to ecological variables etc...

I want to modeling the number of species will change or not change through the time. For that prediction, I want to state a model by using natural mortality or birth rate, anthropogenic effect and also the effect of ecological variables, so on.

Well, then I will be much interested by the answers you'll get, as I am also dealing with changes in species communities through time and looking for a good way to analyze my data

Do you have any information about the STELLA system modeling?

No sorry, I've never heard about it. I'll have a look

Hi!

Can you precise a bit better the data set you handle, and the exact hypothesis you want to test? Because first you described a biodiversity- ecosystem variables problem (univariate), and then you talked about population dynamics…

As far as I understood you want to model how a set of ecological variables influence the number of species. Here you will have a community matrix of counts (or any other abundance measure) per habitat (with or without replications). On the other hand you have a set of variables on the same site.

If you are interested on understanding how these variables influence biodiversity in general, then birth rates and so on are not necessary. You must simply decide what biodiversity variables are of interest to you. For example let’s say species richness, and Number’s of Hill using exp2 to consider richness and evenness. You will then obtain two vectors, the first a counts data, and the second continues variable. Both will change with several environmental characteristics so you have a “regression problem”. Each of your two biodiversity (response variables) will be explained with your environmental (explicative variables). Depending on the amount of data available, number of replicates, absence of pseudo replication you will choose an adequate regression function (lm, glm… whatever) with interactions or whatever. You can use a step algorithm (generally backwards) to eliminate progressively variables not supporting important effects on the studied variables based on AIC (BIC).

Otherwise if you have too many variables (and several are categoric) reducing considerably your degrees of freedom, you can create new variables based on a PCA to “fusion” several environmental variables that covariate together. Then carry the step…

Then if you have birth/death rates of a hole community of species on several sites (incredible sampling effort and expensive data base ), and you want to model biodiversity (and not specific patterns of abundance or population dynamics of these species). Then I suggest using functional group approach, and summarize species with variables like: fitness, individuals/(area or resource abundance) and minimal viable population, to associate fitness of species (birth/death rates) sharing similar functional responses under similar stressors to establish winners vs losers on a changing environment.