MXD Age Based Variance
Posted by Jeff Id on January 22, 2014
I woke up the other day wondering if age of a tree would cause a different latewood density response to environmental factors. I don’t remember reading anything about it so it seemed worthwhile to check.
I am working with 124 tree samples from the Polar.mxd file here. This is the same data from my previous post. I took each MXD curve which we showed previously has a long term age related signal and filtered that curve with a 51 year lowpass Gaussian filter. The top of figure 1 shows the original curve (black) and the filtered curve (red). Subtracting the red curve from the black curve gives the HF signal in the second pane of Figure 1 immediately below. Basically a flat squiggly line.
I took all of the 124 trees and did the same thing. Then I took the average for each year and calculated the confidence interval from the flattened high frequency curves. The two sigma CI expands in a trumpet bell shape (dark gray) like the previous post so we know that differences in long term trend weren’t causing the whole shape. The second pane below is the simple standard deviation (blue line). Visually it has no appreciable trend over time meaning that as the trees age and gain mass, the response to 50 year and shorter signals isn’t a function of the age of the tree. The trumpet bell shape of the dark gray region is therefore due to less samples in the older years and we don’t need to correct MXD variance for tree age. It’s kind of a lame result, but at least we know the MXD answer.