(Part 1, Part II, Part III)
I finally started doing some “real” work. Last night I got data on abundance of butterflies from some regularly surveyed transects. So, this morning I set to work, and by lunchtime I was totally confused about what I wanted to do.
The data consisted of a quarter of a million observations of abundances of butterflies along transects that someone has walked along. The transects are split into sections according to the type of habitat. We want to relate abundance to habitat type, for different species.
The first thing I wanted to do was to get a feel for the data: plotting where the sites are in the UK, and drawing a few tables, so that I knew where the data was. After a bit of playing around I decided to restrict my analyses to species 75 in 2007 – that species has the most data and I could extend the analysis further later.
I then started to fit habitat models – showing that the species prefers some habitats over others. That wasn’t too bad, although the results changed somewhat when I introduced longitude and latitude:
Estimates of habitat effects: black has no lat-/long-itude, red has both. We can see that (a) the estimates are different (i.e. the species prefers some habitats), and (b) adding latitude has an effect.
At this point I hit a big problem. I had totally lost sight of why I was doing the analysis. So I had to stop and discuss this. This turned into a long but useful discussion about what we were doing, which produced this:
What we concluded was that we want to estimate a distribution map for each species, and ask how that depends on climate. We need to use the observation data to do that, which means modelling the “true” absences. hence, we need to build observation models, using either the observed presences or the abundances. It all got a bit complicated, but I think we know what we’re doing now, and hopefully everything will work out.
I’ve now got some more data, so I can have some real fun!
