Friday, 30 November 2018

You can't make an omelette without .....

Destructive harvesting in Brazil

A few updates on some of our activities over the past month or two. Most excitingly, this involved a destructive harvest experiment in Caxuianã, where we first scanned in 2014. Andy and Matheus led the experiment to scan, cut down and weigh four large tropical canopy trees. The aim of this work is to provide the best TLS estimates of tree volume we can, and to compare these with the harvest values. We also wanted to make measurements of wood density, as a function of tree diameter and with height. Wood density is critical in going from volume to mass of course. But it can vary a lot within and between species and even within a single tree. In the hyperdiverse tropics, the relative lack of wood density measurements may be one of the big uncertainties in forest biomass estimates based on allometry - size-to-weight relationships - that convert tree diameter (or height from satellites) to biomass. That's ALL the estimates then.

Here is a 46 m tall, 20 ton tree coming down. The team did an amazing job of clearing around the tree to give a fantastic, clear view with the TLS - some of the best data we've ever been able to get for a tall tree like this. This is a highly-skilled job as you can see.


I very much have mixed feelings about this - it's hard to watch a tree like this coming down and knowing you're responsible for it. The flipside is that if we can show how TLS can work in weighing trees like this, we can massively improve our estimates of global forest carbon stocks, particularly in the tropics where few trees of this size have ever been weighed - and none have been done like this with TLS in the Amazon.


The tree weighed over 20 tons wet, probably half that dry, of which half again (~5 t) is carbon. Early results show the wood density is a lot more variable than we'd ever expected - a problem for converting volume to biomass with a single value, but really interesting in scientific terms. Regardless of the mechanical properties, it's easy to see why these trees are so commercially valuable just on the basis of the beauty of the wood.


Prof. Lola da Costa (far right) and the chainsaw team weighing a large. and beautiful, piece of hardwood trunk.

Matheus (front left) and the team with the results of cutting up the tree above. Giant heavy beermats.
An interesting aside - when we came back, Feng Yin pulled down the Sentinel 2 satellite data covering the period the harvesting was going on - a few days before and over ensuing couple of weeks. Amazingly, the gaps due to felling 3 trees can be seen from space at 10 m resolution! The NDVI images highlight this even more. A sobering thought - 3 trees and some clearing leaves a hole that can be seen from space.
Sentinel 2 images showing the site and the tree harvest area circled: dates are 26/09, 1/10, 6/10, 26/10.

NDVI generated from the images above, highlighting the holes in the canopy.

Early analysis by Andy and Matheus already shows some amazingly clean point clouds from the TLS, which will allow for high quality reconstructions and estimates of volume from the TLS. The wood density is also showing some very interesting variations!


Huge thanks to Lola da Costa who organised the local teams - without him this wouldn't have worked at all. Also huge thanks to Patrick, Lucy, Ingrid and the local crew.


New papers

Our leaf angle distribution (LAD) paper came out in AFM, which shows a new way to get at LAD from TLS and comparing this to a photographic method. LAD is an important parameter in understanding canopy photosynthesis. But it's very hard to measure so it's often assumed to be simply distributed to make life easy (and because we don't have any other info). This work was largely led by Matheus and Jan Pisek, applying Matheus's leaf-wood separation to TLS data from Kew, and comparing to Jan's photographic method. It shows that the TLS method does surprisingly well even for small leaves and odd morphologies.
Figure 8 from Boni Vicari et al. (2019) showing the TLS point cloud, separated leaves and extracted leaf angles.
Here is a 46 m, 30 ton tree coming down. The chainsaw team did an amazing job of clearing around the tree to give us a fantastic clear view with the TLS - some of the best data we've ever had of the upper part of a tree like this. This is a highly-skilled job as you can see.
Two other papers accepted for publication are Andy's paper on treeseg, his tree extraction tool for TLS data; and Matheus's paper on leaf-wood separation, as part of his lidartf tool development. These are both potentially really useful tools for anyone working with TLS in forests. More on both those soon. 

Lastly, my Tansley Insight piece in the New Phytologist, on the potential for TLS in plant science, has come out. I've had some very nice feedback from people on it, mostly to do with figure 1, the montage of trees sketched by my daughter's primary school class! People really seem to find it relatable - which was the original idea, but you never know if these things will work or fall flat.