Modelling Spatial Patterns of School Choice

A couple of weeks ago I visited King’s Department of Education to give a seminar I entitled Agent-based simulation for distance-based school allocation policy analysis. The aim was to introduce agent-based modelling to those unaware and hopefully open a debate on how it might be used in future education research. This all came about as I’ve been working on modelling the drivers and consequences of school choice with Profs Chris Hamnett and Tim Butler here in King’s Geography Department.

Hackney School Admissions Brochure

In their recent research, Chris and Tim looked at the role geography plays in educational inequalities in East London. Many UK local education authorities (LEAs) use spatial distance as a key criterion in their policy for allocating school places: people that live closer to a school rank get allocated to it before those that live farther away. This is necessary because it’s often the case that more people want to send their children to a school than there are places available at it. For example, you can read about the criteria the Hackney LEA uses in their brochure for 2012.

Using data from several LEAs, Chris and Tim showed empirically how this distance criterion is related to school popularity. School popularity is indicated for example by the ratio of school applicants to the number of places available at the school (A:P) – some schools have very high ratios (e.g. up to 8 applications per place) and others very low (e.g. down to around one application per place). Furthermore, this spatial allocation criterion is an important influence on parents’ strategies for school applications, dependent on the location of their home relative to schools and their ability to move home.

These allocation rules, combined with parent’s strategies, produce patterns and relationships between schools’ GCSE achievement levels, A:P ratio and the maximum distance that allocated pupils live from the school. In Barking, for example, we see in the figure below that more popular schools have higher percentages of pupils achieving five GCSE’s with grades A* – C, and that these same popular schools also have the smallest maximum distances (i.e. pupils generally live very close to the school).

Empirical Patterns in Barking Schools

This spatial pattern can also seen when we look at maps of the locations of successful and unsuccessful applicants to popular and less popular schools in Hackney. For example, looking at the figure below (found in Hamnett and Butler 2011) we can see how successful applicants to The Bridge Academy (a popular school) are more tightly clustered around the it than those for Clapton Girls’ Technology College (not such a popular school).

Map of successful and unsuccessful applicants to two schools in Hackney

The geography of this school allocation policy, combined with differences in parents’ circumstances, suggests this issue is a prime candidate for study using agent-based modelling. Agent-based simulation modelling might be useful here because it provides a means to represent interactions between individual actors with different attributes (in this case schools and parents) across space and time. Once the simulation model structure (e.g. rules of interactions between agents) has been established, it can then be used to examine the potential effects of things like opening or closing schools (i.e. changes in external conditions) or changes in school allocation policy rules or parents’ application strategies (i.e. internal model relationships and rules).

I developed an initial ‘model’ as a proof of concept and which you can try out yourself. Things have progressed from that proof of concept model, and the model now represents changes in cohorts of school applicants and pupils through time, including the potential for parents to move house to be more likely to get their child into a desired school.

In the seminar with the Department of Education guys I presented some ouput from the recent modelling. I showed how the abstract model with relatively few and simple assumptions can start from random conditions to reproduce empirical spatial patterns in school applications and attainment outcomes like those described above (see the figure below)

School model screenshot

I also presented early results from using the simulation model to explore implications of potential policy alternatives (such as closing failing schools). These ideas were generally welcomed in the seminar but there were some interesting questions about the what the model assumptions might entail for maintaining existing policy assumptions and intentions (what we might term the rhetoric of modelling).

I’m exploring some of these questions now, including for example issues of how we define a ‘good’ school and how parents’ school application strategies might change as allocation rules change. These will feed into a research manuscript that I’ll continue to work on with Chris and Tim.

Answering forest management questions

Although I’ve been working on new ideas since leaving Michigan and returning to London about a year ago, there’s still lots to do to examining alternative forest management strategies.

Several years ago we set out to develop a simulation model that could be used to investigate the effects of interactions between timber harvest and deer browse disturbances on economic productivity and wildlife habitat. We’ve already published several papers on this work, but just before Christmas we submitted a manuscript to Ecological Modelling entitled ‘Modelling for forest management synergies and trade-offs: Tree regeneration, timber and wildlife’. In the manuscript we report error analyses of the full simulation model (which uses the USFS Forest Vegeation Simulator) and use the model to investigate scenarios of different timber and deer management actions. Our results indicate that greater harvest of commercially low-value ironwood and lower deer densities significantly increase sugar maple regeneration success over the long term.

I expect we’ll also report some of these results at the Fourth Forest Vegetation Simulator (FVS) Conference to be held in April this year in Fort Collins, CO. Our abstract, entitled ‘Investigating combined long-term effects of variable tree regeneration and timber management on forest wildlife and timber production using FVS’, has been accepted for oral presentation. It would be great to be there myself to present the paper and discuss things with other FVS experts, but I’m not sure if that will be possible. If it’s not, Megan Matonis will present as, handily, she’s currently doing her PhD in that neck of the woods at Colorado State University.

In the meantime, Megan and I are in the process of finishing off a different manuscript describing the mesic conifer planting experiment we did in Michigan. In that experiment we planted seedlings of white pine (Pinus strobus), hemlock (Tsuga canadensis), and white spruce (Picea glauca) in northern hardwood stands with variable deer densities and then monitored browse on the seedlings over two years. We found that damage to pine and hemlock seedlings was inversely related to increasing snow depth, and our data suggest a positive relationship between hemlock browse and deer density. These results suggest that hemlock restoration efforts will not be successful without protection from deer. Hopefully we’ll submit the manuscript, possibly to the Northern Journal of Applied Forestry, in the next month or so.

All of this work has been pursued with management in mind, so it was nice this week to receive a call from Bob Doepker, a manager at the Michigan Department of Natural Resources with whom we worked to co-ordinate data collection and establish key research questions. Bob had some questions about the details and implications of our previous findings for deer habitat, tree regeneration and how they should be managed. It was good to catch up, and no doubt our ongoing work will continue to contribute to contemporary management understanding and planning.

Agent-based models – because they’re worth it?

So term is drawing to an end. There’s lots been going on since I last posted here and I’ll write a full update of that over the Christmas break. I’ll just highlight here quickly that the agent-based modelling book I contributed to has now been published.

Agent-Based Models of Geographical Systems, is editied by Alison Heppenstall, Andrew Crooks, Linda See and Mike Batty and presents a comprehensive collection of papers on the background, theory, technical issues and applications of agent-based modelling (ABM) in geographical systems. David O’Sullivan, George Perry, John Wainwright and I put together a paper entitled ‘Agent-based models – because they’re worth it?’ that falls into the ‘Principles and Concepts of Agent-Based Modelling’ section of the book. To give an idea of what the paper is about, here’s the opening paragraph:

“In this chapter we critically examine the usefulness of agent-based models (ABMs) in geography. Such an examination is important be-cause although ABMs offer some advantages when considered purely as faithful representations of their subject matter, agent-based approaches place much greater demands on computational resources, and on the model-builder in their requirements for explicit and well-grounded theories of the drivers of social, economic and cultural activity. Rather than assume that these features ensure that ABMs are self-evidently a good thing – an obviously superior representation in all cases – we take the contrary view, and attempt to identify the circumstances in which the additional effort that taking an agent-based approach requires can be justified. This justification is important as such models are also typically demanding of detailed data both for input parameters and evaluation and so raise other questions about their position within a broader research agenda.”

In the paper we ask:

  • Are modellers agent-based because they should be or because they can be?
  • What are agents? And what do they do?
  • So when do agents make a difference?

To summarise our response to this last question we argue;

“Where agents’ preferences and (spatial) situations differ widely, and where agents’ decisions substantially alter the decision-making con-texts for other agents, there is likely to be a good case for exploring the usefulness of an agent-based approach. This argument focuses attention on three model features: heterogeneity of the decision-making context of agents, the importance of interaction effects, and the overall size and organization of the system.”

Hopefully people will find this, and the rest of the book useful! You can check out the full table of contents here.

Citation
O’Sullivan, D., J.D.A. Millington, G.L.W. Perry, J. Wainwright (2012) Agent-based models – because they’re worth it? p.109 – 123 In: Heppenstall, A.J., A.T. Crooks, L.M. See, M. Batty (Eds.) Agent-Based Models of Geographical Systems, Springer. DOI: 10.1007/978-90-481-8927-4_6

ABM, Prezi and the New Term

I’ve not been in the office much over the last month or so, but that’s all about to change now that the new academic term has arrived!

Since I last posted, I attended and presented work at the Royal Geographical Society Annual Conference, one presentation on our managed forest landscape modelling in Michigan and one on the narrative properties of simulation modelling. Both presentations were in the environmental modelling and decision making session, but despite being the graveyard session (last of the conference!) we had some interesting questions and discussion. I tried out Prezi for my narratives presentation (brought to my attention by Tom Smith). It certainly requires a different approach than the linear style PowerPoint enforces. Whether Prezi is a more useful tool probably depends on the message you’re trying to communicate – if your story isn’t particularly linear then Prezi might be useful.

These last few days I’ve been up in Edinburgh visiting folks at the Forestry Commission’s Northern Research Station to discuss the socio-ecological modelling of potential woodland creation I’ve been working on recently. I also got to talk with Derek Robinson at the University of Edinburgh about some of these issues. Everyone seemed interested in what I’ve been doing, particularly with the ideas I’ve been bouncing around relating to the work Burton and Wilson have been doing on post-productivist farmer self-identities, how these self-identities might change, how they might influence adoption of woodland planting and how we might model that. For example, I think an agent-based simulation approach might be particularly useful for exploring what Burton and Wilson term the ‘‘temporal discordance’ in the transition towards a post-productivist agricultural regime”. And I also think there’s potential to tie it in with work like my former CSIS colleague Xiaodong Chen has been doing using agent-based approaches to model the effects of social norms on enrollment in payments for ecosystem services (such as woodland creation).

I was away on holiday for a couple of weeks after the RGS. On returning, I’ve been preparing for King’s Geography tutorials with the incoming first year undergraduates. The small groups we’ll be working will allow us to discuss and explore critical thinking and techniques about issues and questions in physical geography. Looking forward a busy autumn term!

Philosophy of Modelling and RGS 2011

I just updated the Philosophy of Modelling page on my website. It’s not anything too detailed but I was prompted to add something by my activities over the last few weeks. I’ve been working on both making progress with my ‘modelling narratives’ project and a paper I’ve started working on with John Wainwright exploring the epistemological roles agent-based simulation might play beyond mathematical and statistical modelling (expected to appear in the new-ish journal Dialogues in Human Geography).

It’s only a few weeks now until this year’s Royal Geographical Society annual meeting (31 Aug – 2 Sept). I’m making two presentations, unfortunately both in the same session! It seems my work sits squarely within ‘Environmental modelling and decision making’, as the both abstract I submitted were allocated to that session on the Friday afternoon (Skempton Building, Room 060b; last session of the week so people might be flagging!). The first presentation will deal with the ‘generative’ properties of agent-based modelling [.pdf] and what that implies for how we might study and use that modelling approach, and the second will summarise the Michigan forest modelling work we’ve completed so far. Both abstracts are below.

This also seems a good point to highlight that King’s Geography Department are hosting a drinks reception on the Thurdsay evening from 18:45 at Eastside Bar, Princes Garden, SW7 1AZ. Free drinks for the first 50 guests, so get there sharpish!

Millington RGS 2011 Abstracts

Model Histories: The generative properties of agent-based modelling
Fri 2 Sept, Session 4, Skempton Building, Room 060b
James Millington (King’s College London)
David O’Sullivan (University of Auckland, New Zealand)
George Perry (University of Auckland, New Zealand)

Novels, Kundera has suggested, are a means to explore unrealised possibilities and potential futures, to ask questions and investigate scenarios, starting from the present state of the world as we observe it – the “trap the world has become”. In this paper, we argue that agent-based simulation models (ABMs) are much like Kundera’s view of novels, having generative properties that provide a means to explore alternative possible futures (or pasts) by allowing the user to investigate the likely results of causal mechanisms given pre-existing structures and in different conditions. Despite the great uptake in the application of ABMs, many have not taken full advantage of the representational and explanatory opportunities inherent in ABMs. Many applications have relied too much on ‘statistical portraits’ of aggregated system properties at the expense of more detailed stories about individual agent context and particular pathways from initial to final conditions (via heterogeneous agent interactions). We suggest that this generative modelling approach allows the production of narratives that can be used to i) demonstrate and illustrate the significance of the mechanisms underlying emergent patterns, ii) inspire users to reflect more deeply on modelled system properties and potential futures, and iii) provide a means to reveal the model building process and the routes to discovery that lie therein. We discuss these issues in the context of, and using examples from, the increasing number of studies using ABMs to investigate human-environment interactions in geography and the environmental sciences.

Trees, Birds and Timber: Coordinating Long-term Forest Management
Fri 2 Sept, Session 4, Skempton Building, Room 060b
James Millington (King’s College London)
Megan Matonis (Colorado State University, United States)
Michael Walters (Michigan State University, United States)
Kimberly Hall (The Nature Conservancy, United States)
Edward Laurent (American Bird Conservancy, United States)
Jianguo Liu (Michigan State University, United States)

Forest structure is an important determinant of habitat use by songbirds, including species of conservation concern. In this paper, we investigate the combined long-term impacts of variable tree regeneration and timber management on stand structure, bird occupancy probabilities, and timber production in the northern hardwood forests of Michigan’s Upper Peninsula. We develop species-specific relationships between bird occupancy and forest stand structure from field data. We integrate these bird-forest structure relationships with a forest model that couples a forest-gap tree regeneration submodel developed from our field data with the US Forest Service Forest Vegetation Simulator (Ontario variant). When simulated over a century, we find that higher tree regeneration densities ensure conditions allowing larger harvests of merchantable timber, and reducing the impacts of timber harvest on bird forest-stand occupancy probability. When regeneration is poor (e.g., 25% or less of trees succeed in regenerating), timber harvest prescriptions have a greater relative influence on bird species occupancy probabilities than on the volume of merchantable timber harvested. Our results imply that forest and wildlife managers need to work together to ensure tree regeneration and prevent detrimental impacts on timber output and habitat for avian species over the long-term. Where tree regeneration is currently poor (e.g., due to deer herbivory), forest and wildlife managers should pay particularly close attention to the long-term impacts of timber harvest prescriptions on bird species.

Summer 2011 Papers

Since I last posted, THREE of the papers I’ve mentioned here previously have become available online! Here are their details, abstracts are below. Email me if you can’t get hold of them yourself.

Millington, J.D.A., Walters, M.B., Matonis, M.S., Laurent, E.J., Hall, K.R. and Liu, J. (2011) Combined long-term effects of variable tree regeneration and timber management on forest songbirds and timber production Forest Ecology and Management 262 718-729 doi: 10.1016/j.foreco.2011.05.002

Millington, J.D.A. and Perry, G.L.W. (2011) Multi-model inference in biogeography Geography Compass 5(7) 448-530 doi: 10.1111/j.1749-8198.2011.00433.x

Millington, J.D.A., Demeritt, D. and Romero-Calcerrada, R. (2011) Participatory evaluation of agent-based land use models Journal of Land Use Science 6(2-3) 195-210 doi:10.1080/1747423X.2011.558595

Millington, J.D.A. et al. (2011) Combined long-term effects of variable tree regeneration and timber management on forest songbirds and timber production Forest Ecology and Management 262 718-729
Abstract
The structure of forest stands is an important determinant of habitat use by songbirds, including species of conservation concern. In this paper, we investigate the combined long-term impacts of variable tree regeneration and timber management on stand structure, songbird occupancy probabilities, and timber production in northern hardwood forests. We develop species-specific relationships between bird species occupancy and forest stand structure for canopy-dependent black-throated green warbler (Dendroica virens), eastern wood-pewee (Contopus virens), least flycatcher (Empidonax minimus) and rose-breasted grosbeak (Pheucticus ludovicianus) from field data collected in northern hardwood forests of Michigan’s Upper Peninsula. We integrate these bird-forest structure relationships with a forest simulation model that couples a forest-gap tree regeneration submodel developed from our field data with the US Forest Service Forest Vegetation Simulator (Ontario variant). Our bird occupancy models are better than null models for all species, and indicate species-specific responses to management-related forest structure variables. When simulated over a century, higher overall tree regeneration densities and greater proportions of commercially high value, deer browse-preferred, canopy tree Acer saccharum (sugar maple) than low-value, browse-avoided subcanopy tree Ostrya virginiana (ironwood) ensure conditions allowing larger harvests of merchantable timber and had greater impacts on bird occupancy probability change. Compared to full regeneration, no regeneration over 100 years reduces merchantable timber volumes by up to 25% and drives differences in bird occupancy probability change of up to 30%. We also find that harvest prescriptions can be tailored to affect both timber removal volumes and bird occupancy probability simultaneously, but only when regeneration is adequate. When regeneration is poor (e.g., 25% or less of trees succeed in regenerating), timber harvest prescriptions have a greater relative influence on bird species occupancy probabilities than on the volume of merchantable timber harvested. However, regeneration density and composition, particularly the density of Acer saccharum regenerating, have the greatest long-term effects on canopy bird occupancy probability. Our results imply that forest and wildlife managers need to work together to ensure tree regeneration density and composition are adequate for both timber production and the maintenance of habitat for avian species over the long-term. Where tree regeneration is currently poor (e.g., due to deer herbivory), forest and wildlife managers should pay particularly close attention to the long-term impacts of timber harvest prescriptions on bird species.

Millington, J.D.A. and Perry, G.L.W. (2011) Multi-model inference in biogeography Geography Compass 5(7) 448-530
Abstract
Multi-model inference (MMI) aims to contribute to the production of scientific knowledge by simultaneously comparing the evidence data provide for multiple hypotheses, each represented as a model. With roots in the method of ‘multiple working hypotheses’, MMI techniques have been advocated as an alternative to null-hypothesis significance testing. In this paper, we review two complementary MMI techniques – model selection and model averaging – and highlight examples of their use by biogeographers. Model selection provides a means to simultaneously compare multiple models to evaluate how well each is supported by data, and potentially to identify the best supported model(s). When model selection indicates no clear ‘best’ model, model averaging is useful to account for parameter uncertainty. Both techniques can be implemented in information-theoretic and Bayesian frameworks and we outline the debate about interpretations of the different approaches. We summarise recommendations for avoiding philosophical and methodological pitfalls, and suggest when each technique is best used. We advocate a pragmatic approach to MMI, one that emphasises the ‘thoughtful, science-based, a priori’ modelling that others have argued is vital to ensure valid scientific inference.

Millington et al. (2011) Participatory evaluation of agent-based land use models Journal of Land Use Science 6(2-3) 195-210
Abstract
A key issue facing contemporary agent-based land-use models (ABLUMs) is model evaluation. In this article, we outline some of the epistemological problems facing the evaluation of ABLUMs, including the definition of boundaries for modelling open systems. In light of these issues and given the characteristics of ABLUMs, participatory model evaluation by local stakeholders may be a preferable avenue to pursue. We present a case study of participatory model evaluation for an agent-based model designed to examine the impacts of land-use/cover change on wildfire regimes for a region of Spain. Although model output was endorsed by interviewees as credible, several alterations to model structure were suggested. Of broader interest, we found that some interviewees conflated model structure with scenario boundary conditions. If an interactive participatory modelling approach is not possible, an emphasis on ensuring that stakeholders understand the distinction between model structure and scenario boundary conditions will be particularly important.

Changing the ‘Targets and Timetables’ Climate Change Narrative

Earlier this week I was in Leipzig, Germany, to meet the Ecological Modelling research group at the Helmholz Centre for Environmental Research (UFZ) and one of my PhD supervisors, Dr. George Perry. While there I was lucky to meet and talk with some renowned ecological modellers: Thorsten Wiegand, who’s work includes spatial point process modelling (although some of his discussion with George about that was a bit technical for me!); Volker Grimm, proponent the ‘Pattern-Oriented Modelling’ approach (look out for a new review of this in Phil Trans. of the Royal Society in the near future), and Andreas Huth, notable forest dynamics modeller.

At UFZ I gave a presentation I entitled “Future Forests: Managing and Creating Forests for Biodiversity, Recreation, Timber and Carbon” in which I talked about some of the work I did in Michigan and the new project I’m working on now in the UK. The talk seemed to go down well and the research group had some very good questions, both about technical aspects of the modelling and the issues it is applied to (i.e. forest ecosystem management and woodland creation, including the Woodland Carbon Code). Thanks to Juergen Groeneveld for organising this (and his hospitality at UFZ).

Much of the data I presented regarding the Michigan work was collected by Megan Matonis. Her analyses of that data, on which I helped and supervised, are now available to read in a paper that is currently in press with Forest Ecology and Management (email me if you can’t access the online version).

Another interesting activity at UFZ was hearing Roger Pielke Jr. talk about the need to ‘change the climate change narrative’. In his talk he suggested that understanding all carbon policy can be boiled down to a single sentence;

‘people engage in economic activity that uses energy from carbon emitting generation’.

He emphasised that he thinks the “Targets and Timetables” approach to reducing anthropogenic carbon emissions is flawed. As an example, he used the case of the UK and the Climate Change Act of 2008 which set the aim of an 80% cut in the country’s carbon emissions by 2050 compared to 1990 levels, with an intermediate target of 34% by 2020. However, Pielke argues that given the ‘iron’ law of climate policy (that we cannot mitigate emissions by reducing GDP, both because people will pay only so much to mitigation now, and because increasing GDP is seen as a virtue by way of its effects on povety reduction) we cannot hit these types of targets.

Previous decarbonisation of the UK economy has been achieved by replacing the contribution to GDP from high-emitting manufacturing with low-emitting financial services. He wonders how long can this go and presented his estimate that for the UK to actually hit its 2020 target it will have to build more than 40 nuclear power stations in the next 10 years. In this context, he suggested that the building of a third runway at Heathrow was an insignificant concern (in terms of the new emissions it would generate) when there are still 1.5 billion people globally who do not have access to electricity. His argument is that we do not know how to achieve the targets and the timetables we have set ourselves.

Pielke argues that we must change the climate change narrative from

“We need to use less energy and fossil fuels are cheap

to

“We need more energy and fossil fuels are too expensive“.

This would allow these 1.5 billion people to access the electricity they aspire to whilst driving the growth of alternative, cleaner, sources of energy. I like this argument – and his one about making small steps towards these change to reach bigger changes – but it seems to run counter to his point about the insignificance of another runway at Heathrow (which by increasing capacity for flights would continue the narrative of cheap fossil-fuelled energy). Opening a third runway but only allowing non-fossil-fuelled aeroplanes to use it is ultimately most consistent with the change in narrative he argues for.

And of course, while at UFZ, George Perry and I took the opportunity to discuss past, current and ongoing work over beers and dinner. Mainly we discussed the idea surrounding the narrative properties of generative simulation models and on which I plan to submit a manuscript to a journal for publication soon. But we also thought about other areas of research including land use modelling (continuing our work in Spain) and landscape disturbance-succession modeling (including the use of the LFSM I’ve developed with paleo-estimates of wildfire regimes).

All-in-all a very interesting and productive trip!

Launching effective interdisciplinary human-environment research

After a while bouncing around various outlets, the paper that emerged from the CHANS Workshop at US-IALE 2009 in Snowbird has been published. Presented as a meeting review in the ESA Bulletin, Research on Coupled Human and Natural Systems (CHANS): Approach, Challenges, and Strategies discusses what the CHANS approach is and what the current challenges and strategies in this field are. For example, we suggest the following are the keys to launching effective CHANS research projects:

Identify the goals and final products of the project

  • Goals and products could include answers to scientific questions, hypothesis testing, a simulation model or decision-support tool, policy or management recommendations, or education.
  • Identify and articulate analysis boundaries and scales of interest: spatially, temporally, and in terms of physical processes.
  • A preliminary conceptual model may help initiate discussion among potential collaborators; the conceptual model need not be correct in what it is illustrating, but rather serve to “break the ice” and generate discussion.

Build a team around the identified goals and products

  • Identify project manager(s) and submanagers, where the submanagers may be discipline specific and responsible for a particular component of the project.
  • It may also be beneficial to assign to one person responsibility for overseeing and maintaining the project timeline. It might be advantageous for this person not to be a manager or submanager to minimize potential conflicts.
  • Once the team is together, reexamine the initial goals and final products.

Methods necessary to accomplish project goals and products should now be developed.

  • It is important to recognize that the final products may change in response to the project team’s vision and analysis. Team members must be prepared to be flexible, to reevaluate the project’s conceptual framework and methods as a partnership matures.
  • Potential challenges of complexity and uncertainty should be discussed at this point; where in the project may they later manifest themselves? How may they be overcome?
  • Each team member should be recognized as “a tool in a toolbox,” each providing a unique contribution that works in tandem with the other tools (e.g., the compass and ruler) to build the entire project.

We conclude; “The CHANS approach is emerging from its infancy, characterized by the use of rudimentary language skills in describing deeply complex systems. With proper support, it stands to contribute to a better understanding of the multifaceted interactions between human and natural systems, and thus inform societal choices in pursuit of sustainability.”

Long-Term Bird-Timber Trade-offs

Not surprisingly, during my time at Michigan State University many of my posts on this blog focused on the work I was doing there on forest ecosystem management. I’ll continue to write papers and use and develop the simulation model I initiated at MSU, but now I’m back in London I’m sure the emphasis on this blog will switch to the primary work I’ll be doing here. Before it does, here’s a post on the work I’ve done recently related to the Michigan study and which I’m about to submit for review.

I’ve written here previously about how I’ve been working on modelling the long-term impacts of poor tree regeneration on forest structure and estimating bird presence in forests given their structural characteristics. In my last few months in Michigan, I integrated these two issues as part of the development of the integrated ecological-economic simulation model. The aim was to assess trade-offs between between protecting bird species of conservation concern and ensuring the productivity of industrial forests given the variable tree regeneration densities we have seen across our study area and given the timber harvest options available. I was particularly interested in how the variations in tree regeneration we have seen across our study area [we have a paper on these currently under review – more details soon] might influence long-term forest sustainability. Simulation modelling is an excellent way to look at these types of issues over long time periods.

To examine the trade-offs I integrated bird occupancy models I had developed for four bird species (black-throated green warbler, eastern wood pewee, least flycatcher and rose-breasted grosbeak) with our our model of forest gap regeneration and FVS. I then used the model to simulate various scenarios of regeneration and timber harvest prescriptions. For example, I simulated different densities of trees regenerating in the forest gaps created by timber harvest and different proportions of these trees as either sugar maple or ironwood. These are the sorts of variables that Megan Matonis found to vary across our study area and that are most likely driven by white-tailed deer herbivory. With the simulation model we could then look at how these different scenarios influence forest structure and, in turn, bird occupancy probability. We also looked at how different timber harvest prescriptions interact with these different densities and compositions of regenerating trees.

Using our model for a simulated century we found that the four bird species we examined responded uniquely to changes in forest structure (in turn due to the variation in regeneration composition and density and timber harvest prescriptions). We also found that 100-year average timber volume removals, which varied with harvest prescriptions and regeneration, were related to bird occupancy for three of the four species, positively for two and negatively for one. These results suggest that timber harvest prescriptions can be tailored to influence both timber removal volumes and bird occupancy probability, but only when regeneration is adequate. This is illustrated by the figure below for one of the bird species.

Plot illustrating tradeoffs

Mean annual timber removed is plotted on the horizontal axis and mean bird occupancy probability on the vertical axis. The different colours of points are the different densities of regeneration (darker is higher) and the different shapes are the different timber harvest prescriptions. When regeneration is poorer (lighter colours), differences in the volume of timber removed are smaller between prescriptions (horizontal axis) than differences in bird occupancy probability (vertical axis, relative to the uncertainty bars).

These results imply that management actions that promote high tree regeneration rates (for example, by reducing deer herbivory) will benefit both bird populations and timber production in the long-term. Consequently, we suggest that where tree regeneration is currently poor, forest managers should pay closer attention to the long-term impacts of timber harvest prescriptions on bird species.

As I highlighted above, this work is very near being submitted for publication. I’ll post here as the review and publication process progresses (and maybe try to use fewer hyphens in the title).

Multi-Model Inference in Biogeography

Earlier this month I, along with George Perry, finished a review on multi-model inference for the journal Geography Compass. Geography Compass publishes state-of-the-art reviews aimed at students, researchers and non-specialist scholars. The manuscript is currently under review so we’ll have to see what the reviewers think before the paper is published. Once it’s available I’ll re-post here. In the meantime, here’s the abstract to whet your appetite;

Multi-model inference (MMI) aims to produce scientific knowledge by simultaneously comparing the evidence data provide for multiple hypotheses, each represented as a model. Stemming from the method of ‘multiple working hypotheses’, MMI techniques have been advocated as an alternative to null hypothesis significance testing. These techniques will likely be particularly useful in research fields such as biogeography where formal experimentation is difficult and data are often influenced by uncontrolled factors. In this paper we review two complementary MMI techniques – model selection and model averaging – and highlight examples of their use in the biogeography literature. Both techniques can be implemented in a Bayesian framework and we outline the debate about different interpretations of probability. We summarise recommendations for avoiding philosophical and methodological pitfalls, and suggest circumstances in which each technique will likely be most useful in. We finish by advocating a pragmatic approach to MMI, one that emphasises the ‘thoughtful, science-based, a priori’ modelling others have argued is vital to ensure valid scientific inference.