Hi Pablo,

did you get any further than this. I see your post is quite some time ago, but maybe some comments are worthwhile considering. First of all you are on the right track stating the importance of the first introduction of the species. And, as you said, this is rarely a decision made by the species but rather an unitentional (or intentional) incidence triggered by humans. Humans are also responsible for shaping the environment and they do that with a speed higher then the "normal" alteration by different species (or communities). This goes exactly in the direction of your third point, that niche models assume equilibrium which is in most cases not fulfilled.

Thinking about the way one should model invasion processes I would clearly include humans and their behavior in these models. Furthermore you would need a stochastic component to do that, since the "unintentional" introductions are rather hard to follow.

BTW, there is a model of ours in "Ecological Modeling 220:1339-1346" in which we used behavioural data of humans (the way and frequency they travel for work and leisure) to model the spread of invasive ('neophyte') species in a human dominated area.

Hope this is of some sort of help.

Ciao

Ronald

These posts include different aspects, which need to be differentiated:

1) you talk about "ecological niche models" - I much prefer the more generic name "suitability models" or "suitability maps", as this makes clear that anthropogenic factors can (should? must?) be included.

2) usage of the suitability models for identification of introduction points: your point is correct -humans facilitate these releases, and the suitability of the habitat is secondary: examples are horticultural plants, which are planted in otherwise unsuitable habitat, but because of gardening, they can grow there. They might spread afterwards, but this depends on several factors

3) usage of suitability models for modeling spread is essential. A recent paper [1] showed that climatic niche shifts are seldom, so in most cases the climatic niche of the origin of the species can be used to predict the suitable area in the region of introduction.

4) mechanistic framework: one has to be careful with using mechanistic, as in most cases the information needed is not available, and assumptions have to be used. In many cases, using statistical techniques for certain aspects (e.g. dispersal, but possibly not germination and establishment) is more useful and promising.

5) I now see that you seem to be talking about animal invasions - much more difficult, and I was talking about plants.

To sum up: the use of suitability models in predicting the spread of species is developing quickly, and quite a bit of work is done to integrate the static maps with dynamic modules, like dispersal.

Cheers,

Rainer

[1] Petitpierre, B. et al., 2012. Climatic Niche Shifts Are Rare Among Terrestrial Plant Invaders. Science, 335(6074), pp.1344-1348. Available at: http://www.sciencemag.org/cgi/doi/10.1126/science.1215933 [Accessed March 15, 2012].

Sorry Rainer,

I cannot agree with you. Suitability models are not adequate to model spread in most of invasions, but particularly animals. In fact, the species changes their habitats when introduced. Moreover, invasion by definition is a dynamic process and one cannot model dynamisc with static models assuming equilibrium.

I think that mechanicist models are much better, as you control and know exactly what is going on, while with suitability maps you only have correlational models that must be interpreted (with an obvious degree of speculation). Effectively, mechanicistic models are much more complicate to fit, but I think it's totally worth regarding the information you obtain. And it's our task as researchers to get this kind of data. But, of course, it is much more easy and quick to do the other. Further, in suitability models you also need to have a lot of assumptions that are systematically violated in modeling non-equilibrium species.

Further, remember that most models assume that animals are free to select their prefered habitat, but if, as in invasions, the habitat is selected directly by humans, then you cannot model the underlying process.

This is why suitability models fail to predict the future expansion of invasive animals (and this is true, you see model on invasion that after a while did not predict the areas colonised by species), and thus cannot be validated.

Other question is if you merge a suitability model with a dispersal kernel. Then, maybe you could have a better performance, but again care should be taken.

I prefer the approach of Ronald in modeling dispersal linked to humans, and then model survival in these sites.

Of course, I know that suitability models are adequate for predict areas of equilibrium species, but I'm not debating about this issue.

Thanks for the comments.

Hi Pablo,

please see my point 5: I was thinking about plants, and you are talking about animals - different story, different approaches needed, different problems, different framework needed.

You mention a few times equilibrium: this is needed to identify the climatic envelope - If the plant is not yet in equilibrium, you will underestimate the area suitable - for invasions I would say better then nothing.

If we wait till we have all data to paremetrize our completely mechanistic models, it will be to late.

"And it's our task as researchers to get this kind of data." - plus teaching, writing paper, supervising and get funding, write applications, reports, ... We can identify the data needed - but usually not collect all data needed.

Predicting into the future: all models only work for a certain time frame - and the further you look into the future, the worse your predictions get. In addition: the more dynamic a system is, the more difficult it is usually to predict anything - and then there are humans at play...

In aliens, you have dispersal by humans and "natural" dispersal. Both need to be accounted for. I haven't mentioned anything who is the dispersal vector.

"Suitability models are not adequate to model spread": I completely agree and I never said they are adequate - they play a role (an important role!), but they need to be combined with simulation models, whatever type the simulation models are.

And again: animals can move, which makes it more difficult then plants.

Cheers,

Rainer

Now I agree much more with you. I know that mechanicistic models are much more difficult (I'm spending several days in constructing the model, not to talk about six years of monitoring in the field), but I remark totally worth given the deeper insight one gains. And I don't think it will be too late. If the suitability model doesn't work, then you also be too late (or maybe you think you've strong evidence for conservation but you don't have it at all).

Thanks for the comments