Workshop Blogging I: Getting Started

All this week I am at a workshop in a wet, flooded York. In a fit of enthusiasm I decided to blog the discussions and work, to give the outside world an idea about what we are actually doing.

The workshop is part of the UKPopNet, which organizes workshop to bring ecologists together to work intensively on a group of problems. This group is working on distributions of butterflies in Europe, and how they will change with climate change. This can be used for conservation planning: for endangered species: dynamic planning can even be tried, where habitat is created at times when it is needed by a species to survive, and not before when it would not be able to live on the site anyway, e.g. because it is too cold.
So, what happened today? Once we had sorted out the room, wireless connections etc. we introduced ourselves. Several of us have been to previous workshops, so we know what to expect. We then discussed what we wanted to do this week. We sorted the problem out into three sub-projects: analyzing two sets of data (one just observations of presences of butterflies, the other organized sampling), and then developing dynamic models of how the species will spread.
The discussion spread to the problems of moving between different spatial scales, from 100m up to the national scale. This is an issue we will keep on coming back to: the more smaller scale allows us to work on finer details, but extending the models to the national scale will need a big computer and a lot of time.
After lunch we split into two groups, one (which I joined) to discuss the data, and the other to discuss the models of the future.
Our group started by discussing our two main sources of data on the distribution of butterflies in the UK: recordings of presences of butterflies by observers in the UK, and estimates of abundance organized by the UK BMS. The former data is extensive because it is recordings from the many butterfly enthusiasts. The problem is that it is not organized according to any sampling scheme: many recorders just go off to somewhere nice, and note down what they see. Even worse some only report species once they have seen a rare species, and at least one person only reported the species he was interested in. Hence, we cannot be sure about absences: were the species not there, or were they just hiding? Did nobody just go to that area?
In addition, the details vary: some observations are at the 10km x 10km square scale, some at the 2km x 2km squares, and others at the 100m x 100m level. A couple of butterfly atlases have been produced, so the whole country is covered during these times, but other times are more patchy: observations will be in nice places near the observer’s home. And some data only say a butterfly was seen at some date in that year, others are at the monthly or daily scale. So, lots of details to sort out.
The second set of data is more organized: volunteers go out every week to a site and walk along a route counting the butterflies they see. This has been done since the 1970s, but cover less sites. It should give better information about absences of species, and crucially tells us about how many butterflies there are present.
Much of the discussion was around what exactly the data is, how was it collected, what are the problems. We decided roughly what the model would be that we would use: at the moment, we are concentrating on the observation model, i.e. trying to remove the biases caused by how the data was collected. After going through this, we agreed on what data we needed, e.g. the location, time, identity of the collector etc. Once we have this, we can start to develop the models, and see the problems in the data.
After a coffee break (during which it was pointed out to me that fitting a spatial dynamic model of the real population dynamics would be a real pain), we discussed the BMS data. The data is more organized, and we ended up in a long discussion about the phenology, i.e. the times when the butterflies are flying: this affects the models of the abundances, especially as the weather can have a huge effect (as butterflies only fly around when the weather is good; a great convenience for the many butterfly researchers). I wanted to ignore the phenology (because the models are going to be large anyway, so why make them stupidly huge?), others wanted to explicitly include the timing (because it does affect the estimated abundance). In the end we decided to do both, and see what happens: that could be a paper in its own right.
During this the other group arrived. Once we had talked ourselves to a stop, they described their discussions. Basically, they had discussed how to model dispersal, and disagreed on the two approaches they could take, so decided to do both: either building complex mechanistic models, or simpler descriptive models that can be fitted to data.
After this we decided to stop, have a few chats and then disappear off to the pub.