A colleague contacted me yesterday with some interesting modelling work based on the neutral community model of Stephen Hubbell.

Hubbell’s work fascinated me when it was first published. I have always assumed that it is easily misunderstood. If Hubbell is taken to imply that all species are equal, then the work appears to be eroding the basic subject matter of ecology. Ecologists tend to look for informative differences between species. However this is not exactly what the theory is based on. It is rather more subtle. The theory concerns the equality of individuals (in this sense it almost has political implications). Hubbell writes.

“The theory treats organisms in a community as essentially identical in their per capita probabilities of giving birth, dying, migrating, and speciating. This neutrality is defined at the *individual *level, not the species level. All that is required is that all individuals of every species obey exactly the same rules of ecological engagement. So, for example, if all individuals and species enjoy a frequency-dependent advantage in per capita birth rate when rare, and if this per capita advantage is exactly the same for each and every individual of a species of equivalent abundance then such a theory would qualify as a *bona fide *neutral theory by the present definition.”

I have illustrated this idea for students with a very small simulation in R . R code, like matlab or octave, is very terse and efficient so a couple of lines can do a fair amount of work. Lets assume a very simple reproductive rule applies to all individuals. The rule is that all have the same discrete generation time, in other words they all survive one time step of a simulation. At the end of the time step each individual produces two offspring. The lottery then applies. The offspring are randomly mixed and placed back on the finite space occupied by the previous generation. So half find a home and half “die”. The process is then repeated. This was exactly how I first implemented the model. I then re-implemented the idea in fewer lines by making the probability of selection depend on the proportion of individuals in each species, which amounts to the same thing.

This is all the code needed to run the model. (Try this if the quotation marks are not right lottery.doc)

nspecies<-8

gridsize<-20

time<-100

RUN<-function(){

X<<-matrix(0,nspecies,time)

palette(rainbow(nspecies))

mat<-sample(1:nspecies,gridsize*gridsize,replace=T)

X[,1]<-table(mat)

image(matrix(mat,gridsize,gridsize),col=sort(unique(mat)))

for (i in 2:time){

a<-X[,i-1]

mat<-sample(1:nspecies,gridsize*gridsize,prob=a,replace=T)

X[,i]<-table(factor(mat,levels=1:nspecies))

image(matrix(mat,gridsize,gridsize),col=sort(unique(mat)))

gc()

}

matplot(t(X),type=”l”,lwd=2,col=rainbow(nspecies))

}

#However to get a little tcltk interface add this

############################################

library(tcltk)

Run<-function(…){

nspecies<<-as.numeric(tclvalue(“nspecies”))

gridsize<<-as.numeric(tclvalue(“gridsize”))

time<<-as.numeric(tclvalue(“time”))

RUN()}

tt <-tktoplevel()

d1<-tklabel(tt,text=”Numero de species”)

d2<-tklabel(tt,text=”Tamaño del grid”)

d3<-tklabel(tt,text=”Tiempo”)

s1 <- tkscale(tt,from=0, to=20, variable=”nspecies”,

showvalue=TRUE, resolution=1, orient=”horiz”)

s2 <- tkscale(tt, from=2, to=100, variable=”gridsize”,

showvalue=TRUE, resolution=1, orient=”horiz”)

s3 <- tkscale(tt, from=10, to=500, variable=”time”,

showvalue=TRUE, resolution=10, orient=”horiz”)

but <- tkbutton(tt, text=”Run”,command=Run)

tkgrid(d1,s1)

tkgrid(d2,s2)

tkgrid(d3,s3)

tkgrid(but)

The model is to be played with (just download R from CRAN and paste all the code into the console), rather than watched but here are some example runs on You Tube,

The moral of the story is that under a neutral model some species start a random walk to extinction if co-existing in a small space with other species. However we cannot tell which will dominate and which become extinct from their characteristics, because all individuals, regardless of species are governed by exactly the same rules. The fate of species is determined by the cumulative luck or misfortune of the individuals that have been given the label. The larger the grid the more species that can be supported for a longer time, but extinction risk is always present if a long losing streak hits. The element of Hubbell’s theory that this model does not reproduce well is the log-normal (-ish) relative abundance distribution of the surviving species, although come to think of it I have not tried running the model with a really large number of species and a really large grid. R runs out of colours for the illustration, but this is not important.

Many interesting questions remain. Could a simulation model based on this sort of idea be used to predict extinction over a real landscape? Models of this sort seem far too abstract to provide real insight into the real world. But in some sense the process that Hubbell describes in a more formal mathematical way in his book probably does apply. But, how can these sort of ideas be communicated accurately to prevent them from provoking the sort of unproductive SLOSS debate that has often resulted from theory being pushed too far?

So this model does not have interactions between species — symbiosis, competition — correct? But it does have Space? I feel like this is so simple, it would be easy to play around, adding more factors — say overlaps between different spatial zones.

My immediate thought is — is there any Abstract Math that can explain what general patterns you would see? Robert Ghrist comes to mind.

Or maybe it’s even simpler and one could look to finite automata theory.