Friday, August 29, 2008

Elith, J., Graham, C.H., et al. (2006) Novel methods improve prediction of species' distributions from occurrence data. Ecography 29, 129-151

Elithfest! Sean's post just reminded me of Jane's other paper that I thought everyone should know by now - but I am always surprised how many people have missed it. It is an excellent comparison of all sorts of algorithms to model taxa in different circumstances. Key messages:
1. Presence-only data can give decent results
2. There are a few newer techniques (MARS, MAXENT) that can outperform the old staples (GARP, BIOCLIM) under certain circumstances.

Elith, J., Leathwick, J. R. & Hastie, T. 2008 A working guide to boosted regression trees. Journal of Animal Ecology 77, 802-813.

A straight forward users guide to a machine learning process known as boosted regression trees (BRT's). BRT's are useful data mining tools, they learn the relationship between a set of variables and an output, for example a field ecologists might want to know which environmental factor's predict the presence of a species. Or I have used BRT's for sensitivity analysis on a simulation model with lots of parameters. Unlike regression approaches (GLM's or GAM's) BRT's (and machine learning in general) deals well with non-linear responses and higher order interactions.

The paper is written for ecologists so it explains BRT's in a way that doesn't require any mathematical ability (which is why I like it) or knowledge of machine learning. Best of all there is a fairly straight forward R package for BRT's, called gbm.

Tuesday, August 26, 2008

Bodin & Norberg (2005) Information Network Topologies for Enhanced Local Adaptive Management. Environmental Management Vol. 35, No. 2, pp. 175–193

Not sure how much network stuff people read, but this is a nice link up of social networks and ecological managment. In a nut shell they find that if you have no friends you won't do very well, if you have a few friends (low density of links) then the system as a whole is pretty resliant to change and the community as a whole does well because they can learn from each others experiences. If the the community has too many links the agents synchronise there actions, making the ecological system unstable, swinging from good to bad and back again. So it is good to have a few friends, but not too many. Perhaps we need to start managing which agencies can talk to each other?

Monday, August 25, 2008

Linquist - But is it progress? On the alleged advances of conservation biology over ecology.

Linquist, S. (2008) But is it progress? On the alleged advances of conservation biology over ecology. Biology and Philosophy, 23, 529-544.

Linquist, a former postdoc with the UQ Biohumanities department argues that dropping island biogeography theory for complementarity-based algorithms will lead to inefficient protection of biodiversity. Unfortunately, statements like "In practice, these [complementarity] algorithms are biased in favor of selecting networks of small reserves with non-overlapping species" ignore a huge body of work post 1988 - also demonstrated by missing references to larger planning frameworks (Margules & Pressey). Linquist also ignores target setting or discussions about adequacy as well as the literature about spatial constraints to reserve design.

Schweiger et al. - A comparative test of phylogenetic diversity indices

Schweiger, O., Klotz, S., Durka, W. & Kühn, I. (2008) A comparative test of phylogenetic diversity indices. Oecologia, 157, 485-495.

From the prolific folks at UFZ in Germany, this is an interesting overview of all the methods that calculate phylogenetic diversity indices, along with their properties. For those who don't know these indices: PD indices work like a normal diversity index, but give stronger weight to taxa that are separated further in a phylogenetic tree.

Wednesday, August 20, 2008


Hi all! We created this blog to share interesting ecology/conservation papers that might slip through our alerts. Please don't restrict yourself to new papers, overlooked older contributions might just be as valuable.