I have been worrying a lot about what to do next in the El Niño Project.
It would be great if we could carry out some interesting machine learning or neural networks approach to El Niño by December 1st, because shortly after that I'm supposed to [give a talk about this stuff at the Neural Information Processing Seminar](http://forum.azimuthproject.org/discussion/1364/networks-in-climate-science/#Item_1).
However, I feel very unconfident of my/our ability to do anything very interesting in this direction by December. I could probably do it if I did nothing else between now and then. However, I'll be teaching 2 courses, including a seminar on network theory, and I really want to explain - in the seminar, and on the blog - the great new research my grad students have done. I also need to apply for some grants before mid-November. I would also like to finish writing a book - which is almost done, but everything takes longer than expected.
If people here were organized and willing to work on this project together, it might stil get done in time. But I don't feel that I can say "write code that uses a random forest method to predict El Niños $n$ months in advance" and expect one of you to do that in a week. It just doesn't seem to be working that way. I could say a lot about why not, and how we might become better organized, but I won't now.
For now, I probably need to reduce the ambitiousness of my short-term plans.
So, I think I will try to better understand the network formed by connecting the Ludescher _et al_ sites in the Pacific with edges weighted by (various measures of) correlation between temperature histories. Graham Jones has already written a lot of the software needed to do this! He has already created a lot of graphs and plots that I haven't blogged about yet! There is a lot more that could be done... but I really just need to give a 1-hour talk and not get laughed off the stage.
The first step, then, is for me to blog about the work he's done.
If anyone else wants to produce some product that can help me give a good talk by December 1st, please let me know! You can say things like "I want to do this", or "tell me what to do".
I do not believe that wavelet or differential equation based models of El Niño will fit into this talk. Needless to say, our longer-term goals for after December can continue to be extremely ambitious, and they can include such models. It's just this deadline that's giving me an ulcer.