A couple of questions from the last nichey post prompted this post. Geoff said that:
I’m not even sure what is meant by an optimal environment for a species/genus/whatever.
while Andrew said that:
it wouldn’t surprise me if a lot of species tend to live at the margins of their “ideal” habitat.
We need a bit of abstraction to address these questions. In a laboratory, a plant would be expected to show a humped response to the main variables of temperature and water availability. The parameterisation of this function can be termed the ‘fundamental niche’ of the species, and may be equated with a physiochemical optimum unaffected by competition.
In the wild there will be gaps in this function. This can be called the ‘realized niche’, and may be due to interspecific competition, but also affected by chance dislocation of the species’ distributions that are unrelated to any physiological capacity to occupy that space. If this were not the case, then no species could ever ‘invade’ another region.
Now we can imagine a ‘niche model’ of a species as consisting of an estimate of the humped function that represents it fundamental or realized niche. That function — to a first approximation the range of temperature and rainfall where the species has been observed — is the basis of original ‘envelope’ methods such as BIOCLIM.
But consider the logic of this model, calling the range model R and the sightings of the species S. If a species occurs at a point, then that implies it lies within the range R. However, we know from basic logic that:
S → R ≠ R → S
But R → S is what we need for predicting the distribution of the species. We need to know that if the location has certain characteristics M then the species occurs, i.e. M → S. Thought of another way, the ‘gaps’ in the fundamental niche cause inaccuracy when we try to use it to predict.
To discover the optimal description M such that M → S we need to do a different kind of search that optimises the accuracy of prediction of the model. It has been shown conclusively that methods based on the correct logical form of implication are more robust predictors of the occurrence of a species, than an envelope model composed of climatic variables temperature and rainfall.
I found that in practise, a single variable such as monthly rainfall (ie seasonality) was much better predictor of the distribution of a species than the range of temperature and rainfall. In some cases, non-climatic factors such as soil types were optimal. The notion of an optimal environment only makes sense in the lab. In the wild, it really only makes sense to talk about the optimal determinant of a species.
Why this should be the case is kind of conjectural, but it would make sense from an entropy or information viewpoint for species to ‘spread out’ over possible environmental determinants, providing they stay with their fundamental niche. This would include ‘marginal’ locations.
In this view, the effects of climate change on species would be very mixed. Some species with a determinant that is not primarily climatic would be very unaffected. Other species would be affected over only part of their range, depending on the determining factor. This has been seen, as the primary response of species to climate change has been range expansion to the north in the northern hemisphere, due to a temperature determinant up there. Some unlucky species with highly seasonal determinants, or isolated non-climatic determinants, may be severely affected.