Subject: Results of modeling
Date: Fri, 21 Oct 2005 08:56:08 -0400
From: "Leverett, Bob"
Will, Jess, John, Don, and Bob
The attached spreadsheet contains the results of all the
modelings to date accomplished with the RD 1000. One of the tabs
gives the results of a multiple linear regression using 4
independent variables: height, CBH, diameter at 50 feet and
diameter at 100 feet. In the regression, to be consistent, I
should have used DBH in inches instead of CBH in feet to the
same result since the relationship is fixed and linear. I just
like seeing CBH in feet in my data. Note that the multiple
regression coefficient based on 24 trees is a whopping 0.984. I
wasn't expecting that. I didn't include the Thoreau Pine in the
regression because it wasn't modeled with the RD 1000. I also
left out the large 3-trunked pine in the Conway cemetery. It was
modeled far too crudely for the results to be reliable.
I doubt that the coefficient will remain so high as more trees
are included. Odd forms and natural variability are bound to
bring down the correlation. However, it looks to me that we will
likely stay above 0.90 using 4 the independent variables. I
should note that this exercise is mainly to whet the appetite.
After we've modeled a dozen more trees, I'll run the exercise
again. I'll also develop an alternate model that will use
diameters at say 33% and at 67% of total tree height as
independent variables. This is as opposed to diameters at 50 and
100 feet. For the Jake tree this would translate to measuring
diameters at 55 and 112 feet. For the Bullard Woods tree it
translates to measuring diameters at 44 and 89 feet. I'll still
have the measurements at 50 and 100 feet to enable us to compare
models. The 33% and 67% approach obviously allows us to model
shorter trees and treat them all consistently.
On Saturday, I hope to remodel the Tecumseh tree along with a
couple of other biggies in the Elders Grove. Gary, are you game?
Eventually, I hope to model all 18 tagged trees in the elders
grove.
Bob
Database Developer and Systems Analyst
Information Technologies
Attachment Converted: WP HM Models
|