Species-fragmented area relationship
Readings
Species-fragmented area relationship
<2016-09-01 Thu 07:36>
- Reading for QAECO reading group
- relationship between increasing number of species and increasing area of habitat
- (so I suppose inverse is true, e.g. climate change reduce area of habitat, reduce number of species?)
- predict number of extinctions due to loss of habitat
- the issue is fragmentation
- habitat loss generally results in a fragmented (sub) habitat
- SAR assumes contiguous regions (think a circle getting smaller)
- fragmentation is when holes appear (think a doughnut)
- (meta) population capacity is essentially the carrying capacity of the habitat
- fragmentation causes this to decline
- e.g. due to disconnect between fragments
- SFAR (species-fragmented area relationship) fits better than SAR in simulations
- Figure 1
- the authors suggest the SAR does not fit the number of species well, but it looks like it does to me? The lines are pretty close to the points?
- I think perhaps the figure legend is not described well — the lines are not SARs but just general fitted lines?
- Species traits
- when including the fragmentation part, the authors suggest a large amount of variation in the ratio of extinction and colonisation
- this has an effect on their \(b\) parameter
- how are the models fit? stupid pnas, need to check in supplementary material
- Case study uses only 14 points
- they suggest that species number is reduced due to fragmentation
- a better visualisation would be to use points proportional to \(\lambda\) in Figure 4A
- They suggest hat SFAR is better than SAR, due to a difference in AICc of 2.44 (!)
- furthermore, the least-squares value of the area coefficient is negative, meaning that conditional on \(\lambda\), the number of species \(S\) is increasing as area \(A\) decreases
- what!?
- this same thing happens in eqn 3, where if \(z < 0\), if \(A_{\text{new}} < A\), $S\text{new}/S >1 $!
- they do discuss what they term hidden parameters: those used for estimating \(\lambda\)
- perhaps these should be estimated in a hierarchical model?