Lines in the sand

Dr. Suzanne Prober, Prof Richard Hobbs, Prof Hugh Possingham and I have recently had a paper entitled ‘Lines in the sand: quantifying the cumulative development footprint in the world’s largest remaining temperate woodland‘ published in the journal Landscape Ecology. You can view the article online here or contact me to send you a pdf and/or any appendices.

In the paper, we quantify cumulative anthropogenic development footprints for the Great Western Woodland and expose the large proportion of this that is made up of unmapped linear infrastructure. We highlight the crucial importance of explicitly accounting for the ecological impacts of linear infrastructure in impact evaluations – impacts that typically pass under the radar of impact evaluations.

We also present an analysis of key drivers of development footprint extent, both at the regional and landscape-patch levels, and provide key insights, such as the mitigative effect of pastoralism on development footprints in mining landscapes, and investigate the implications for edge effects. Our approach and methodology provide information and insights that are useful for cumulative and strategic impact assessment as well as landscape planning and conservation, and can be applied to other relatively intact landscapes worldwide.

Figure 3 JPEG way.png
Fig. 3 Contribution of different anthropogenic disturbance types to total direct development footprint, with some examples. a) Contribution of different types of infrastructure to total footprint. b) An example of ‘hub’ infrastructure: an abandoned gold mine. c) Aerial view showing both hub and linear infrastructure of a mine and associated exploration grids. d) Aerial view of exploration grids passing through shrubland and woodland vegetation, the white dots are drill pads. e) A mapped track leading to Helena-Aurora Range, one of the banded ironstone formations where mining is currently proposed. The track was probably initially built for mineral exploration purposes and is now used by miners, conservation agencies, and tourists. f) A ground-truthed unmapped track with abandoned exploration drilling sample bags to the left. An abandoned hydrocarbon drum was found further along this track.

Advertisements
Posted in blog, research, sustaining ecology | Tagged , , , , , , , , | Leave a comment

Being quoted

I just came across an interesting book recently published by Random House in the UK called Linescapes: Remapping and Reconnecting Britain’s Fragmented Wildlife, by Hugh Warwick. In the book, the author discussed some of my research, published a few years ago in an article called ‘Under the radar: mitigating enigmatic ecological impacts‘ (contact me if you’d like a copy).

I really like the discussion, even though he’s misquoted me a little (still, the point I was making is unchanged). What a nice feeling to see my research quoted as an example of ‘recent thinking’ and incorporated into the global discourse in such an obviously well-researched and thoughtful book. I hope to have more of my reasearch on the ecological implications of the lines that we draw through wild landscapes published soon too!

Click on this link to see the discussion: https://books.google.com.au/books?id=REEtDQAAQBAJ&pg=PT150&dq=keren+raiter&hl=en&sa=X&ved=0ahUKEwjck5aRmOjUAhWDabwKHbWuDKYQ6AEIKDAA#v=onepage&q=keren%20raiter&f=false

linescapes.png

Here is the book’s blurb: It is rare to find a landscape untouched by our lines – the hedges, walls, ditches and dykes built to enclose and separate; and the green lanes, roads, canals, railways and power lines, designed to connect. This vast network of lines has transformed our landscape.
In Linescapes, Hugh Warwick unravels the far-reaching ecological consequences of the lines we have drawn: as our lives and our land were being fenced in and threaded together, so wildlife habitats have been cut into ever smaller, and increasingly unviable, fragments.
I’d love to hear any comments from anyone who has read it!

Posted in blog | Tagged , , , | Leave a comment

Jungkajungka Woodlands Festival

jungkajungka_poster_final.jpgDon’t miss being a part of the inaugural Jungkajungka Woodlands Festival, held over Easter in Norseman, Western Australia—the Heart of the Great Western Woodlands. This event is organised by the Wilderness Society in collaboration with the Shire of Dundas, GondwanaLink, and with support from a number of other organisations.

This is Ngadju country and the festival organisers acknowledge the traditional owners of this part of the Great Western Woodlands and thank them for co-hosting the festival.

I’ll be talking as part of the presentations on Saturday afternoon on Woodlands Knowledge.

For more information, including how to register (for free), go to:

https://www.wilderness.org.au/events/jungkajungka-woodlands-festival-wa

or click here to download the full program.

Posted in blog | Tagged , , , , , , , | Leave a comment

The Great Western Woodlands: a biological wonderland, a poem, a movement

I’m pleased to share that I have had one of my poems used as the voiceover for the Wilderness Society’s Great Western Woodlands campaign that they launched last month. The poem is called Biological Cornucopia and is one of a suite of  poems that I wrote about the enigmatic ecological impacts of mining and associated linear infrastructure development in the Great Western Woodlands, the subject of the PhD that I completed last year.

My PhD was focused on the conservation of large, relatively intact landscapes in the face of widespread development such as resource extraction: a challenge of global conservation significance. In particular, ‘enigmatic’ ecological impacts that commonly evade consideration in conservation strategies invariably pervade such landscapes. Keren investigated the significance and ecological implications of linear infrastructure (e.g. roads, tracks and railways) largely associated with mining activity in the largest and most intact remaining temperate woodland on earth. Keren discovered significant effects on attributes of key ecosystem processes, including predator activity and water movement and recommends ways in which these impacts could be ameliorated.

Great cudos to Amy Matheson for excellent editing, and to the amazong team at the Wilderness Society for their great work on the campaign.

If you’re inspired to experience the Great Western Woodlands, consider joining the inaugural Jungka Jungka Woodlands Festival to be held in Norseman in April.

Posted in blog | Leave a comment

The Great Western Woodlands Campaign Launch: 3rd February 2017

GWW-esa-event-cover.jpg

All are invited to join the Wilderness Society for a special celebration of the Great Western Woodlands (GWW), one of West Australia’s most significant natural spaces. They’ll be launching a new campaign for the Great Western Woodlands and showcasing the new GWW website and video.

The GWW is the largest unfragmented woodland left on earth. It is vast, beautiful and unprotected. The GWW has remained home to a remarkable richness and diversity of plant and animal life. With over 3,300 flowering plant species, there are more native plants in the Great Western Woodlands than in the whole of Canada! This is a great opportunity to learn more about the GWW, why it’s under threat, and how you can help preserve it for all Australians.

This event will also exhibit work from local artists with a focus on natural spaces and the Australian outback including some  stunning photos of the GWW. I’ll be giving a presentation about the incredible environmental values of the GWW and my research in it, and also presenting a poem that I wrote about my experience doing field research in the GWW and reflections on the magnificent Helena Aurora Range and the proposal to mine it.

RSVP by Wednesday the 1st of February and…

gww_tagline-solid-red.png

Friday the 3rd of February 2017.
6.30 – 8.30pm.
North Perth Town Hall.
26 View St, North Perth WA 6006.

Food and drinks provided.

RSVP at: http://wilderness.nationbuilder.com/wa_gww_launch_event

Posted in blog | Leave a comment

In defense of science

As a government scientist in Australia, I stand in solidarity with those around the world who are restricted from enabling informed decisions

Being A Better Scientist

I (Pleuni Pennings) endorse the following, which was drafted by Graham Coop (UC Davis), Michael Eisen (UC Berkeley) and Molly Przeworski (Columbia):

We are deeply concerned by the Trump administration’s move to gag scientists working at various governmental agencies. The US government employs scientists working on medicine, public health, agriculture, energy, space, clean water and air, weather, the climate and many other important areas. Their job is to produce data to inform decisions by policymakers, businesses and individuals. We are all best served by allowing these scientists to discuss their findings openly and without the intrusion of politics. Any attack on their ability to do so is an attack on our ability to make informed decisions as individuals, as communities and as a nation.

If you are a government scientist who is blocked from discussing their work, we will share it on your behalf, publicly or with the appropriate recipients…

View original post 6 more words

Posted in blog | Leave a comment

Mixed-model analysis of a percentage outcome using logit transformation in R

Warning: this post is terribly technical. Readers not inclined towards statistical analysis and R coding may wish to skip this one!

I was recently contacted by a biologist in the UK regarding one of the analyses that I did in my PhD. This researcher was trying to model percentage data with mixed effects models and was struggling to find examples of where this had been done before to base his analysis on, except for the analysis that I performed to identify drivers of disturbance at a patch level in the Great Western Woodlands. I remember that at the time, I really struggled to figure out the best way to perform the analysis, it was so complicated and there were issues that made it different from all the other examples I was trying to go off. I remember assuming that it was probably really straightforward and that everyone else does these sorts of analyses all the time and that I was a bit slow for not being able to figure it out straight away… so it is somewhat gratifying to hear that it’s not a common problem that biologists solve, and to have someone else reaching out to me for my experience on it!

I’ve decided to write up my response as a blog post, in case anyone else is interested… and to remind myself of what happened there! Comments and suggestions for improvement are invited!

In general terms, the analysis is aimed at understanding what factors explain how much anthropogenic disturbance (e.g. land clearing) will occur within the landscapes of the Great Western Woodlands, where there is an abundance of mining activity and pastoralism within an otherwise relatively undisturbed region.

More specifically, the analysis relates to predicting the percentage of area disturbed (the ‘development footprint’) at a patch level (this was the smaller of two scales of analysis used in the investigation, and the only one to be discussed here), by a set of potential predictor variables. The predictor variables included both categorical and continuous variables. Categorical variables were: pastoral tenure, greenstone lithology, conservation estate, ironstone formation, schedule 1 area clearing restrictions, environmentally sensitive area designation, vegetation formation, type of tenements (exploration/prospecting tenement, mining and related activities tenement, none), primary target commodity (gold, nickel, iron-ore, other), and sample area (n=24). The numerical variables were: number of mining projects, number of dead tenements, sum of duration of all live and dead tenements, distance to wheatbelt, and distance to nearest town.

The next image (below) shows an example of these patches and their attribuets.

pic for blog 2.jpg

Example of polygons constructed to assess predictors of development footprint at the patch level

I used ‘nlme’ package in R for the analysis. Early in the data exploration (which I won’t detail here) I realised that the data for the proportional area disturbed was very skewed, and that log or other transformations weren’t enough to achieve a normal-ish distribution. A logistic transformation was the clear solution, given the percentage nature of the data (i.e. that it was bounded by 0 and 1); see the following image with scatter plots and histograms.

pics-for-blog-post

However, I couldn’t just use a GLMM with a gaussian distribution and a logit link, as this would have produced the problem that predicated values will be outside the 0-1 range. I had to first do a logistic transformation, model linearly, and then back-transform the modelled values.

The following line of code shows how I logit-transformed the proportion of area disturbed (‘dist_perc’), after adding 0.001 to each value (I had to do this as logit only works on values between 0 and 1, exclusive). I also show how I ‘untranformed’ the data and plotted it to produce a nice, normal, histogram. Note that dp was the dataframe I worked with, stands for ‘disturbance by patch’.

> dp$ldistp = qlogis(dist_perc+0.001)                   # qlogis does a logit transformation on (a set of) values.

> plot(dp$ldistp)                                         # scatter plot looks much better

> range(dp$ldistp)

> hist(dp$ldistp)                                         # Looks good.

> dp$s.dist = scale(dp$ldistp)                   # s.dist is the logit-transformed, scaled disturbed percentage

> back_transformed = (plogis(dp$ldistp))-0.001           # plogis ‘undoes’ the logit transformation.

 

After completion of the data exploration and dealing with other variables that needed to be transformed, confirming no outliers, dealing with collinearity etc and scaling issues, I modelled the transformed variable using maximum likelihood and performed a manual backward-selection and a few other checks to reach the final model, which I then modelled using REML.

I have included the whole script here. It includes the date exploration stages, as well as a plethora of comments, and also a few tries of alternative analysis techniques using lmer, lmerTest and lme4 (and modelling with a logit-link before I figured out that this wasn’t a useful approach). I take no responsibility for errors! Check line 621 for the final model, followed by back-transformations of the outcomes.

Posted in blog | Tagged , , | 1 Comment