Wednesday, May 20, 2015

PhD Tree

After the tree post last week, Brian Witte was kind enough to send me a fascinating e-mail, and I'm using it here. What follows is all Brian.

I saw your post on making computer models of trees and I wanted to share a story or two.  Brace yourself.

My friend did a PhD in Forestry in the late 90's at University of Washington.  The initial observation used to form hypotheses for the dissertation was this image:

The image you're looking at is farmland in eastern Washington State, near the Columbia River.  The circles are center-pivot irrigation used to grow vegetables.  The square blocks, however, are a tree plantation.  The trees are hybrids of different species of cottonwood, and they're being grown as a source of fiber for a paper mill.  Cottonwoods are notable both for their fast growth and the ease of cloning them.  Any twig can be dipped in rooting hormone, half-buried in the ground, and it will sprout into a new tree.  Using that approach, a single big tree can be chopped up and used to plant several contiguous acres of genetically identical trees.  Each of those rectangular blocks in slightly different shades of green are, genetically, one single tree that was chosen for its rapid growth and high quality fiber.  Here's a photo of a cottonwood plantation, with trees grown like corn:

Grown on this spacing, with plentiful drip irrigation, and cloudless summer weather of eastern Washington, the trees take only 7 years to reach 100' tall, with a trunk diameter of 2 feet. 

Cool story, right?  So where does modeling come in?  You'll notice that among those blocks of trees, some appear to be different shades of green.  When you broaden the range from visible light to full-spectrum imaging (the AVIRIS technology mentioned in the caption), the difference is even more pronounced.  Remember that each block is comprised of genetically identical trees. Looking at ONE tree is hard.  Looking at a giant block of trees allows you to eliminate individual variation over a tree's surface and measure averages.  Each block represents the average interaction of several thousand trees.  The blocks of trees were all planted at the same time, they all get the same water and fertilizer, and they all are rooted in the same soil.  So what accounts for the different appearance?

My friend, Kim Brown (now Dr. Brown), went out to those blocks of trees and measured every conceivable variable: chlorophyll content, leaf size, leaf shape, tree height, rate of photosynthesis, rate of respiration.  She even took the data she collected on the trees to France where she collaborated with a computer modeler to figure out what happened to incident rays of sunlight as they impacted the trees.  How much light was absorbed, how much reflected and at what angle, how much re-radiated in a different wavelength...even what happened to a ray of light as it entered a leaf and bounced around inside.  Here's an image from a related paper by the modeler :

It turns out that the different appearance of the blocks of trees came down to the angle at which the leaves hang, relative to the trunk.  Some cottonwood species have leaves with short, stout petioles (the little stem that connects a leaf to a branch), while others have long, limber petioles.

So yes, modeling trees is hard.  I still haven't seen a tree reproduced in a game that I would call convincing (plants have their own uncanny valley when a botanist like me is watching.  And don't get me started on the ents in Lord of the Rings with their totally unrealistic leaf physics).

As for the impact of long vs short petioles?  That would be for another time.

Site Meter