Day 4 contained the beginnings of the students towards developing (and presenting) their own research and ideas. The different 'controversies' now have groups of students thinking about them and developing ideas. These will be finessed over time into short, exciting research proposals. The participants of the school also had an opportunity to present research findings in parallel sessions at the end of the day. The talks were stimulating, well presented and produced some interesting early results. Most notable for me was a study on drought in China and the influence of pre-existing pollution on MCB efforts.
I think I have got to the bottom of the issue around detection that has been kicking around the school. The idea goes like this - the climate system is noisy and slow ramps would not, for at least a decade, show any response that is outside of natural variation. The counter argument (I assumed) was to consider paleoclimates, data assimilation etc to reduce the detection lag. It turns out that the real counter argument is that it doesn't matter than you can't attribute, especially given any response is within the noise. I think this is really interesting and points to issues I have with both stances. The first relies too much on models and overhypes risk, the second (I think) shows an absence of consideration of realpolitik. It may not matter scientifically that attribution is impossible, I suspect it matters politically. I need some more time to unpick this, but that's as good a summary as I can distil at the moment.