Tag Archives: lidar

Oops there goes another rubber tree

1 Reply
2.10

Rubber trees bend, and they also break. This is one of the CATAS plantations in China being battered by strong winds.

Frank Sinatra credited an ambitious ant with shifting a rubber tree, which was a rather implausible achievement. The presumptuous ant might have been taking undeserved credit for a much more normal occurrence, which is rubber trees blowing over in strong winds. That happens all the time with no help from ants.

Rubber tree plantations are found in the tropics, these days predominantly in Asia (despite the rubber tree Hevea brasiliensis being originally a South American species, as its name suggests). Many of the main regions of production are also subject to periodic hurricanes which can cause serious damage to the trees and therefore economic losses to the producers. We can’t prevent hurricanes, but could we manage plantations to make them more wind resistant?

This is a question which we approach in a new paper (led by my collaborator Yun Ting at Nanjing Forestry University*) using terrestrial laser scanning and some novel computational methods. The study site was a research station in southern China where a number of different clones have been planted, each of which has a distinctive branching architecture. We already know which of these is most vulnerable to blow-down from the evidence of past hurricanes. What management options are likely to help prevent damage to the trees?

At present simulations are the only way to approach this problem. It’s not just that it’s far too risky to carry out detailed environmental measurements in the middle of a hurricane (not least because the trees often fall over), but the equipment is likely to blow away too. It’s hard enough to find equipment that’s capable of withstanding hurricane conditions anyway. And then there’s the issue of not knowing when or where a hurricane will strike. Our approach can potentially circumvent all these limitations.

The study involved a large team all of whom had incredibly specialised skills and I don’t pretend to fully understand all of them. The first step was using a terrestrial laser scanner to record the full three-dimensional structure of several rubber trees. This creates a high-resolution point cloud which could be used to reconstruct virtual trees which aim to capture every leaf and branch on the original tree. These were then used to create plantations which were subjected to hurricane-force winds in a simulation environment developed for testing the aerodynamics properties of aircraft and other vehicles. This allowed us to evaluate what the implications of crown structure were for the wind forces experienced by trees during a hurricane. No-one had to put themselves at any risk.**

image3

Computer visualisations of two rubber trees reconstructed from terrestrial laser scanning data. Notice how the one on the left has broader, spreading branches, while the one on the right has a more upright form.

There are many caveats to this study; don’t bother pulling it apart because we already know a good few of them. Most of the assumptions and simplifications were enforced by the computational demands of such high-resolution simulations. For example, plots were relatively small in size (nine trees at a time).*** More problematic is that adding natural flexibility to the trees wasn’t an option. What we are measuring is therefore the wind resistance generated by something approximating a tree made of platinum. Of course we know that this isn’t realistic, but I’d suggest that you look at the qualitative outcomes rather than expecting the quantitative predictions to be precisely accurate. This is a first demonstration of the approach, not a complete realisation of every possible feature. It’s good enough to reveal some interesting differences.

Wind velocity around clone 1
Wind velocity around clone 2

What did we find? Well, clones with larger, denser crowns put up greater wind resistance and generate higher degrees of turbulence, making them more likely to be damaged when exposed to hurricane-force winds. Reassuringly, these are the same ones that blow over more commonly in the real world. So we did all this sophisticated work… and found out something that everyone already knew.

That’s not really the point though. Having shown that the simulations generate reasonable outcomes that match with experience, we can now start to tweak the models and explore the impact of management strategies. Manipulating tree spacing or thinning of tree crowns can all be done virtually, more quickly and with less effort than establishing another plantation and waiting for the next hurricane to discover whether it was effective. What we have is a framework for trying out almost any combination of tree sizes, shapes and arrangements.

Rubber plantations are one of the simplest types of forest, which was a deliberate choice. We can imagine further applications of our method in more complex habitats, where a mixture of tree species could be put together to see how real-world forests cope in the face of one of nature’s greatest destructive forces. That’s our eventual aim, anyway. High hopes, as Sinatra would have put it.

 

 

* This was a massive collaborative study in which my own role was very minor. I haven’t even visited the field site.

** No trees were harmed in the preparation of this paper. It’s not even in a print journal.

*** When I suggested some minor amendments to a late draft I thought Yun Ting was about to cry. It’s hard to convey quite how much computational effort goes into generating just one of these figures. Simulations are not easy research!

 

How much stuff is in a forest?

2 Replies

Forest are complex, messy habitats. The tired old saw of not being able to see the wood for the trees has an element of truth, because the tangle of material prevents easy measurement and assessment. And that’s without even mentioning the leaves.

Terrestrial laser scanning (also known as ground-based LiDAR) is a great technology for getting round this because it allows us to create full three-dimensional reconstructions of forests, from which we can extract parameters that would be otherwise unavailable to a surveyor on the ground. In this aspect it’s a major advance over traditional techniques of forest surveying, though challenges remain in turning the vast quantities of data into relevant and meaningful measurements that we can use to understand how forests form and what the implications are.

DSC_8209

One of our laser scanners in action in a UK woodland. Photo by Joe Ryding. If FARO want to give us a discount for all this free advertising then I’d be very happy to hear from them.

In this post I’d like to focus on some of the findings of our recent paper in Journal of Applied Ecology that might have been overlooked in the media hullabaloo. For a reminder of the major findings, and what I think the implications are for forest conservation, see my earlier post. In any paper, however, there are always some hidden insights that couldn’t be elaborated on in the space available.

In this post I’d like to ask: how much stuff does a forest actually contain? In answering this, our data is perhaps non-intuitive. We split the whole forest into 1 cm cubes then asked whether each contained wood, leaf or neither. Terrestrial laser scanning can’t see inside stems, so instead we measured the surface area of trunks and branches*. Rather than using LiDAR to simply recreate the metrics we can obtain by other techniques, I think the future will be in learning to use these outputs more directly as indices in their own right.

foliage_alternative

Vertical distribution of foliage in 40 UK woodlands, split into those with either high or low deer densities, and which were either managed or unmanaged.**

Our scans found that woodland plots contained an average (median) density of leaves of 523 cm3/m3. What that means in real terms is only 0.052% of the total forest volume. In other words, although a forest looks like it’s full of foliage, actually it’s mostly empty space. This is necessarily a minimum estimate because we probably failed to detect many leaves higher in the canopy because our laser beams were blocked by other material getting in the way. Even so, if we only detected half of all the foliage in the forest, it would still be less than a tenth of one percent of the total volume.***

If forests are mostly empty space, this has a number of interesting implications, one of which is for the movement of species. For large, lumbering mammals like ourselves, we perceive forests to be difficult to move through, but for small insects they contain vast distances which need to be traversed. Organisms that can fly find this easier, others will depend on physical linkages to help them move around. The amount of stuff in a forest, along with its density and distribution, will have major influences on the mobility of organisms of all sizes.

Another implication is that we can treat the three-dimensional surfaces of forests as area in much the same way as in less complex habitats. This is actually an old idea, with Southwood coining the term ecospace to describe the effective area presented by a habitat, and proposing that this increased area might be responsible for differences in diversity between habitats. For the first time we have a straightforward way to measure the amount of leaf in the forest, which is a metric of how much habitat space there is for the many organisms that feed on or move across them. We could potentially compare this between sites.

We can do the same thing with stems, remembering that our measure is of surface area rather than woody volume. The area of stems in these woodlands was an order of magnitude lower than of foliage, at 49 cm3/m3. This is interesting because it tells us that for species that forage on tree trunks and branches, there is approximately ten times less area available to them than there is for those utilising leaves as habitat.

The Species-Area Relationship (the closest thing ecology has to a law) tells us that the number of species doesn’t increase linearly with area****. Instead, we would expect a 90% reduction in area to mean a roughly 50% reduction in species richness. Is that true? Do we find half as many species of insects, lichens or gleaning birds on bark as we do on leaves? I’d be interested to look. Unfortunately right now we don’t have an easy way to tell live and dead wood apart, which is important because they form habitats for completely different species.

In short, capturing the three-dimensional structure of these forests is just the beginning, and there are all sorts of new avenues which we can explore with these data.

* A range of algorithms exist to convert these into volumes using cylinder fitting, which is the standard approach used if you want to determine either the amount of timber or carbon in a forest. These weren’t part of our objectives so we didn’t, and actually as a measure of habitat structure as experienced by other organisms, surface area might be more relevant.

** Note that these figures are flipped relative to those in the original paper. This is a more intuitive way of looking at the patterns, whereas in the paper the orientation was set to match the statistical analyses.

*** Strictly this is ‘number of 1 cm cubes per cubic metre which contain foliage’. Leaves are flat and much thinner than 1 cm, of course, so the true volume of foliage is much, much lower than this.

**** The actual relationship is S = cAz, where z is a parameter determining the sensitivity of species richness S to area A, and c is a mathematical constant representing the theoretical richness of a single unit of area.

Should we eat Bambi?

8 Replies

 

Fallow_deer_arp.jpg

Fallow deer in Avon Valley Country Park, Bristol, UK. By Adrian Pingstone (Own work) [Public domain], via Wikimedia Commons. Note that deer kept in parks are not part of the problem; this just happens to be a nice picture of them. Leave these deer alone.

 Our new paper has just come out in Journal of Applied Ecology, in which we’ve used terrestrial laser scanning to examine the three-dimensional structure of 40 lowland British woodlands. We compared woodlands in areas with high and low deer densities, and which were either managed or unmanaged. One of the main findings will surprise no-one: in areas with lots of deer, there is much less foliage at heights below 2 m. What makes our study unique is that we were able to quantify this as a 68% reduction. The interesting results don’t stop there though; there were other differences between high- and low-deer woodlands, extending right the way through the canopy. High-deer woods were on average 5 m taller for some as-yet-unknown reason.

In another post on the Journal of Applied Ecology blog I’ve described these findings in more detail and explained their specific implications for forest management. Here I’m going to use my own site to step over the line into more controversial territory and ask what this means for the broader issue of conservation policy. Note that these are my personal opinions, and go some way beyond what was said in the paper itself*.

First, there are vast numbers of deer in the UK. Their populations have boomed over the last century for a number of reasons. These include a lack of natural predators (wolves and lynx disappeared centuries ago) along with a range of other factors that might include an increase in woodland area, planting of over-wintering crops, and perhaps also impacts of milder winters on their survival. The UK is not alone in this; similar patterns have been observed elsewhere in Europe, throughout North America, and in Japan.

One of the features of the deer we find in British woodlands is that they are overwhelmingly made up of non-native, invasive species. In our study most (85%) were fallow deer, pictured above, which were introduced in the 11th century for sport hunting in deer parks. They are joined by Reeves’ muntjac, a small Asiatic species that probably escaped into the wild in the 19th century. It’s worth emphasising that I bear no grudges against the large, noble red deer of Scotland, nor the scarcer native roe deer, which we seldom detected in our surveys.

Our work has shown that in areas with high deer populations (more than 10 per square kilometre and often much higher) there is a loss of complex understorey vegetation. This dense ground-level foliage includes the regrowing seedlings and saplings of canopy trees, as well as providing habitat for a wide range of birds, small mammals and insects. That many woodland birds have been in decline over the last century is probably not coincidental, and consistent with patterns seen elsewhere in the world.

There are several options to keep deer out of woodlands, but most are either infeasible or ineffective. Fencing is an option, but it’s enormously expensive to establish and to maintain. Moreover, the longer a woodland is left without deer, the greater the amount of palatable foliage that will build up, and hence an ever-increasing incentive for deer to find their way in. It’s also difficult to keep small deer out while allowing passage to the other animals we would wish to have free movement around the countryside. For this reason fencing can only ever be a local and temporary option. Deterrents are also unlikely to be effective. Chemicals soon wash away, and deer quickly learn to ignore attempts to scare them. These tricks might work briefly but they won’t keep deer away for decades, which is what we need to do if we would like complex forest structures to develop.

What then can we do? Let’s tackle that thorny euphemism, ‘control’. What this almost invariably means is finding a way to reduce deer populations. In such circumstances  well-meaning people will always suggest sterilisation, although this would be prohibitively expensive to apply to many thousands of deer, and probably as stressful as any other action. The next option then is that most unpopular of conservation moves, a cull. These are still expensive to carry out and to maintain over long time periods, at least in open landscapes where deer can constantly wander in from elsewhere. What do we have left?

Venison_escalope

Venison escalope in Switzerland. By Kueued (Own work) CC BY-SA 4.0-3.0-2.5-2.0-1.0, via Wikimedia Commons. Serving suggestion only.

We can eat them. It’s almost the same as a cull, except the meat doesn’t go to waste, and the market would help fund deer control. Fallow deer were actually introduced to the UK by the Romans a thousand years before the Normans brought them here, but died out — probably because they were eaten. Together we can do it again.

Venison was a traditional meat eaten throughout the UK only a century ago, as it remains in many parts of Europe. If wild-caught, free-range British venison were to appear in our butchers, on supermarket shelves and restaurant menus, we would only be restoring it to its former popularity**.  Another benefit is that it would provide a source of income for rural communities, many of which are among the most deprived in the UK. The same approach could be taken with wild boar, though my suspicion is that this would not be as effective at controlling their populations***.

Will this work? Nothing in ecology (or life) can ever be guaranteed. When we intervene in complex systems there is always the chance — indeed the likelihood — of unforeseen consequences. The only way to guard against this is through careful monitoring and intervention. If the aim is to restore forest structures then we don’t yet know how low deer populations need to be brought down, over what time periods, and whether market forces will be successful in achieving this. What we do know for certain is the effect of doing nothing. If there’s a chance that eating deer might work then, if you’ll pardon the pun, it’s worth a shot. And venison is delicious.

Eichhorn M.P. , Ryding J., Smith M.J, Gill R.M.A , Siriwardena G.M. and Fuller R.J. (2016). Effects of deer on woodland structure revealed through terrestrial laser scanning. Journal of Applied Ecology, in press. DOI 10.1111/1365-2664.12902

* One of the reasons I’m writing this here, rather than in the paper itself or the journal blog, is that the authors of the original paper wouldn’t all necessarily agree with my prescription.

** In North America hunting is a popular rural pastime and, contrary to the perception of outsiders, has little to do with taking down large animals for display. The majority of people hunt for the table. Why then has this not led to effective control of deer populations? This is probably down to two factors. The first is the vast area of North America, much of which is wooded, and with low densities of people. The second is that game laws regulate hunting so as to maintain populations of deer (at least in part) rather than for the conservation benefits.

*** What we refer to as wild boar in the UK are actually a breed of pig. Although they have escaped and naturalised (and trash habitats in places like the Forest of Dean), when you buy wild boar sausages in the shops it doesn’t necessarily imply wild-caught boar. Most meat still comes from farms. The reason I don’t think we will control wild boar by hunting alone is that they breed at an incredibly fast rate, whereas deer populations grow much more slowly.