[A portion of boreal forest in northern Quebec. A study found that commercial logging in Quebec and neighbouring Ontario has caused the removal of 35.4 million acres of forest. Credit: Renaud Philippe for The New York Times]
I was prompted by the article in the New York Times at reference 1 to turn up the paper – for which they provided the link – at reference 2. A paper which appears to have been led by the Climate Action Beacon at Griffith University on the Gold Coast of Queensland, but with participation from workers from Ontario and Quebec.
A paper where they go to some trouble to demonstrate that Canada’s once huge and unspoilt boreal forest is being badly damaged by clear felling, despite this felling being followed by replacement (by one means or another) and that the woodland caribou who like to live undisturbed in old forests are in trouble. A consequence both of the volume of felling and of the fragmentation of what is left.
I imagine that a fair proportion of what is felled goes up in smoke, so I am more bothered by the massive reduction of the amount of carbon being stored in these forests and the massive disturbance of our biome generally, but that is not the matter of present interest. Although thinking with my fingers I wonder if it is a massive reduction: perhaps I will have a play with Excel and attempt to model the total amount of carbon captured by a regularly harvested forest, taking into account the life cycle of the harvested timber, a lot of which winds up in paper and cardboard. A lot of which winds up in the acid forest soils, half digested.
There is a long history of logging in Canada, dating back to the time when the French first arrived in Quebec in the 16th century. Today’s industrial logging in Canada, mostly clear cutting, followed by reinstatement by one means or another, is now taking around 400,000ha of forest a year, with Brazil and Russia taking even more. For which see references 3 and 4.
Canada is big in the world of forests and is a big exporter of forest products. Maybe 50,000 people are employed in the forests themselves, with a lot more in forest dependent industries like paper. But despite their apparent size, forests do not figure large in the GDP, perhaps one percent, much less than mining.
The core of the present work looks to be the crunching of a large amount of land-use and forest data down to an array of 30m pixels covering parts of Ontario and Quebec – say 1,000 of them to the square kilometre – perhaps 1,000 million of them altogether – perhaps three orders of magnitude more pixels for a two-dimensional map of part of Canada than the three-dimensional voxels – say 1mm cubes – of an fMRI scan of the brain. Quite a lot of this data is derived from the US Landsat program of references 5 and 6.
I imagine that one useful product of having so many pixels is that it is easy to generate smart-looking coloured maps. Depending on the size of the map, more or less ready made.
Age of forest is one of the important data items for a pixel. Important but complicated, with a pixel potentially containing a number of trees of varying ages. Not to mention secondary growths of varying sizes and ages.
Notwithstanding, using disturbance, harvest and other data, they derived an age for each of the 30m pixels. Old pixels were aggregated to the one age class of greater than 100 years, partly because it is hard to be precise about the age of old forests using satellite data. It also happens that a lot of Canadian forest is managed on a 100 year cycle. All this gave a binary (old) age indicator for each pixel.
Then, starting from the 21 populations of caribou touching the study area and their ranges, and using a different range of data, particularly about various kinds of human disturbance, they also derived a binary indicator for each 30m pixel: was this likely to be good ground for caribou? What I think they call ‘critical caribou habitat’.
They then applied a nifty looking technique called morphological spatial pattern analysis (MSPA) to each of these two sets of data, age of forest and caribou friendliness, a technique to be found at reference 7 and 8, from which last the figure above has been taken.
Suppose one has a connected patch of forest defined on a raster array – the sort of thing that might be derived from a satellite image – allowing here both point and edge connections – and one wants to know something about the geometrical properties of that patch. Perhaps something like that top left in the figure above. MSPA breaks the pixels patch down into seven classes: core for the interior and then six others. With one of parameters of the analysis being the width of the edge, here taken to be one pixel – and as the edge gets thicker the core gets smaller, eventually to vanishing point. In the present case, core being the sort of forest likely to be attractive to a caribou. A sort of analysis which might have been of interest when I was looking at a raster version of LWS, for which see reference 9.
Along the way, I find that a lot of work has been done on the loss of geometry entailed in the move from the real world or from vector geometry to a raster format. How much damage is done by reducing the world to a rectangular array of pixels?
So, from the starting point summarised above, we get the MSPA summary below.
Suggesting that older forests and core caribou habitats are badly fragmented, the older forests particularly so, with core making up less than half the total.
This summary tells us that there are a lot of small patches, although it would have been interesting to see a version which counted area as well as counting patches (something which Excel’s pivot tables manage quite easily).
I have not worked through from critical caribou patches to the conclusions about their population fragility. Indeed, I am confused about the relation between older forest – which I understand caribou to prefer – and critical caribou habitat.
Another ingredient in the mix is the mapping product called MapTiler to be found at references 10 and 11. I have failed to find an accessible description of what this product does, beyond offering support for people who want to do mapping, but Bing offers: ‘MapTiler is a platform for web and mobile developers to create and customize maps with various data sources and tools. You can choose from different map styles, languages, and frameworks, and access documentation, code samples, and tutorials to speed up your development. MapTiler also provides a data processing tool that can turn an image into a map on your computer or process satellite imagery of the entire world on a cluster of computers ... They also offer free detailed maps of the entire world for your applications and map hosting for your websites and products’. Which, from what I have managed to glean on my own, seems a fair enough summary.
Presumably it takes care of the business, of the details of projection from the surface of a sphere onto the plane. This chunk of Canada being plenty big enough for that to be an issue.
The conclusion is that the once huge boreal forests of Canada are now much degraded, with more than half of the first growth gone and replaced with new, younger trees. And from the caribou’s point of view, even more degraded, even more fragmented. Not the caribou friendly environment it once was. On the plus side, there are more deer, which eat trees and much prefer the younger trees.
But we should not overdo it. The boreal forest was never an uninterrupted, homogenous sweep of trees. It was always something of a mosaic broken up by features of land and water and with lots of damage by wildfires and plagues of insects and the diseases associated with them. I learn that, for some of these forests, one might expect a wildfire every hundred years or so – which is clearly going to have an effect on the age of the trees.
A map which failed
[The orange patches show areas that have been logged in Ontario and Quebec since 1976. Turquoise indicates areas where the forest is at least 100 years old. Credit...Griffith Climate Action Beacon, Griffith University]
A map which is included in the article in the NYT (reference 1) and which is said to come from Griffith. But it does not appear in the paper at reference 2 – although various maps which may well draw on the same source material do. I can’t find it among all the likely looking stuff at Griffith and Google Image search, in this particular case, does not add anything. I suppose I could write to someone, but whether they would bother to reply is anyone’s guess. So, for the present, without a proper key to the colours, the map above does not tell us much more than that there has been a lot of logging and a lot of disturbance.
[Jeremie LeBlond-Fontaine/Getty]
I thought I had struck gold with reference 3, but, sadly, no, although there is a pretty snap of a woodland caribou – an image which Google Image locates without fuss. Maybe if I worked my way through all the many references in the references below, I would find the map by myself.
The Bing version of Google’s intelligent assistant Bard – sometimes called the copilot – is very prominent and accessible in Microsoft’s Edge, so I thought I would ask him about the feeding habits of caribou: do they live in forests or do they feed in them, or both.
Bing did rather well on this, coming up with nicely presented paragraph about how caribou live on lichen and relatively soft vegetation likes rushes, rather than trees, and gave some references in which this story was confirmed.
Then a bit later on, I started to puzzle about how you ascribe an age to a stand or patch of trees (as in the turquoise of the failing map above), focusing, probably mistakenly, on the constituent trees rather than forest management, that is to say times of harvest and planting. Bing came up with another nicely presented paragraph about how you could take an average of the ages of the trees in the stand, using the size of the trees as weights. He gave some more references, references which, in this case, did not bear on the question at all. When pushed, Bing made grovelling noises and came up with some more references, which again did not bear on the question. Even more grovelling noises, followed by confession that he could not do better.
While when I tried Bard, he came up with a rather longer story, explaining why ascribing an age to a patch was a tricky business. If you must do it, best to think about what you want to use this age figure for. What is the context? He did not offer any references, although I dare say he would have done had I poked him.
So these things are indeed helpful. But I continue to fret about how much untruth Microsoft and Google are going to push out into the world, before they learn how to stop these helpers telling porkies. Assuming that is, that that is going to happen anytime soon.
I associate to the various people I have known for whom the printed word in a newspaper is the revealed truth. It does not occur to them that it might be an untruth, let alone a lie.
Much of the mapping work described here depends on maps which are organised in layers. It occurs to me that all these layers are each, individually geometrical and map friendly. It makes sense to map them, might well make sense to combine them. While in other contexts, a layer might not be map-like at all, perhaps containing text data or categorical data, data which is associated with a place on the map but which is not itself map-like.
One such other context being the LWS-R of reference 9. By design, all the data there is on the map, whether or not it is map like.
On the other hand, it is also the case that the computer package concerned might be able to map text data onto a map in an agreeable way: think of all the labels you get on Ordnance Survey maps. And think of all the other stuff – for example the roads – the shape of which is not, strictly speaking, geometrically accurate at all.
I also associate to the drawing packages in which a contractor could hide whole cities in inconspicuous dots on the drawing he was being paid to create. Or perhaps curate.
All good stuff, deploying lots of interesting data and interesting techniques.
A paper aimed at the specialist, which means that this lay reader anyway took some time to get to grips with it. And even then, some way off a proper understanding.
And while I am convinced that the boreal forest of Canada – and no doubt that of Russia – is being badly damaged by industrial logging – I am not so sure about the caribou. And the carbon was missing altogether.
It would be interesting to get a better grip on this damage relative to that being done to the Amazon, very much the bad-boy in the world of eco-destruction these days – with reference 3 suggesting that there is a lot more at stake, in terms of carbon, in Canada, than there is in Brazil.
No comments:
Post a Comment