Friday, June 15, 2012


date: Tue Dec 2 09:26:07 2008
from: Tom Melvin <>
subject: Collaboration
to: BenSmith

Dear Ben,

I am half way through my current project and have reached the stage where it would be good to get your help/advice/assistance. I am amazed that time has gone so quickly and I seem to have got somewhat distracted with work on process based standardisation necessary to provide a basis for comparing tree-ring based output with the ecosystem based output from GUESS as well as trying to sort out the "Divergence" problem.

One objective in this current project was to arrange for GUESS to grow a "model" forest at the site of my dendroclimatic samples in such a way that the variance in growth of the modelled forest matched the variance in growth of the measured samples from the site i.e. the dendro sample would appear to be a typical sample from the modelled trees.

I have adjusted the input/output of GUESS to read specified sequences of climate data and adjusted GUESS to output files (selected by logical on/off) containing values for individaul trees e.g. tree rings or carbon generated per unit foliage (added a new field "indiv.ann_assim" which includes the respiration costs).

So far I have used the stable public code (2003) and with default parameters for GUESS and my initial testing was for north Finland (65N 30E) and high altitude in the Alps. I downloaded some recent papers (Smith 2008, Miller 2007, Hickler 2006 & 2008, Wramneby 2008, Muller 2007 and Zaehle 2007). Presumably there are several in the pipeline.

I would like to use the new baseline version if is available. I would also like to use the species parameters (Millar 2008) for Scandinavia and similar for the Alps if there are any.

My process-based standardisation (PBS) model (an "allocation" routine running in the reverse direction) required many changes to the allocation algorithms generally found in tree-growth models in order for the PBS to follow the growth of trees; represented by series of ring-width measurements and end up with a sapwood area that matches that measured from the tree cores; e.g. the need for variable sapwood senescence rates and a height/diameter relationship that varies with tree size.

One problem I found using GUESS was that, because of changes with tree age (or size) the initialisation period (needed to generate a random distribution of tree ages across patches and output that is not age-dependant) needs to be serveral time longer than the oldest trees (e,g, 1500+ years).

Another problem is that correlation (north Finland) between output from GUESS and monthly temperature is much earlier in the year, May, as compared to July, than for the dendroclimatic output.

I would like to visit you sometime during the next two/three months (if possible and convenient) to discuss these many other points with you and members of your team.


No comments:

Post a Comment