cc: Tim Osborn , Keith Briffa , Ed Cook date: Mon, 18 Jun 2001 17:24:45 -0400 from: "Jonathan M. Adams" subject: Re: possible reconstruction of climate transfer functionsfromtree to: Ed Cook Hi Ed, Good to hear that you are interested! The approach I'm proposing using a MANN (modular artificial neural network) model should be ideal for coping with the difficulties you point out. The model seeks out clusters that may be associated with a particular set of soil conditions, a local/regional genotype, a particular local/regional custom of tree treatment such as pollarding, or whatever. I am just starting writing the idea down as a proposal and will send you a draft in the next couple of days. As I don't know much about dendrochronology, it is bound to be full of mistakes so I'll appreciate your patience in correcting what I write. Best regards, Jonathan Ed Cook wrote: > > Hi Jonathan, > > I would love to work on this project with you, Keith, and Tim. Let's see > what we can do to pull it together. It is not without its difficulties, in > part related to the fact that tree-ring chronologies of different species > growing under the same regimes of climate, soil, etc. may lead to > distinctly different neural network models due to different intrinsic > biological/genetic responses to the same forcings. Maybe this is what we > are looking for. > > Ed > > >Ed, I've been in touch with Keith Briffa and Tim Osborne at UEA, and > >they are very interested in this neural network approach (am cc-ing this > >note to Tim). I am now thinking that it would be good to have a proposal > >involving yourself, the UEA people, and us folks at URI to look at tree > >rings in relation to a range of GIS layers such as soils as well as > >annual climate etc.. I might also need to bring in someone at PennState > >who is a real whizz at assembling data layers for GIS. Since the total > >for the project must not exceed $250,000, we'd all have to be working > >part time on this (I don't think there is enough to fund anyone full > >time). > > > >The type of model I have in mind, a MANN model, looks for and finds > >regional clusters of relatively uniform behaviour, and then does a > >separate neural network analysis for each of the clusters it finds. It > >would be just ideal for the tree ring data. > > > >I don't think computing capacity will be a problem. My colleague YQ Wang > >(who is the real neural net programmer in this) assures me this is well > >within the capacity of the computers in the GIS group here. I think we'd > >need to do separate model runs for the USA, Canada and Europe using > >whatever environmental GIS layers are available for each area. > > > > > > Jonathan > > > > > > > > > >> Artificial neural networks have been used to estimate climate from tree > >> rings by a few people like Connie Woodhouse and Joel Guiot. This approach > >> certainly has considerable promise when there are unknown or ill-specified > >> nonlinearities in the "true" model. As far as I know, no one has tried to > >> use GIS info to try to develop a regional model of tree ring widths using > >> neural networks. This is an interesting topic. Whether or not it is too > >> computationally demanding to do on a large regional scale is hard to say. > > ================================== > Dr. Edward R. Cook > Doherty Senior Scholar > Tree-Ring Laboratory > Lamont-Doherty Earth Observatory > Palisades, New York 10964 USA > Email: drdendro@ldeo.columbia.edu > Phone: 845-365-8618 > Fax: 845-365-8152 > ==================================