This week one of the papers I have been working on as a result of my PhD research has been accepted for publication in the Journal of Artificial Societies and Social Simulation (JASSS). The paper, written with Raúl Romero-Calcerrada, John Wainwright and George Perry, describes the agent-based model of agricultural land-use decision-making we constructed to represent SPA 56 in Madrid, Spain. We then present results from our use of the model to examine the importance of land tenure and land use on future land cover and the potential consequences for wildfire risk. The abstract is below, and I'll post again here when the paper is published and online.
An Agent-Based Model of Mediterranean Agricultural Land-Use/Cover Change for examining Wildfire Risk
Abstract Humans have a long history of activity in Mediterranean Basin landscapes. Spatial heterogeneity in these landscapes hinders our understanding about the impacts of changes in human activity on ecological processes, such as wildfire. Use of spatially-explicit models that simulate processes at fine scales should aid the investigation of spatial patterns at the broader, landscape scale. Here, we present an agent-based model of agricultural land-use decision-making to examine the importance of land tenure and land use on future land cover. The model considers two ‘types’ of land-use decision-making agent with differing perspectives; ‘commercial’ agents that are perfectly economically rational, and ‘traditional’ agents that represent part-time or ‘traditional’ farmers that manage their land because of its cultural, rather than economic, value. The structure of the model is described and results are presented for various scenarios of initial landscape configuration. Land use/cover maps produced by the model are used to examine how wildfire risk changes for each scenario. Results indicate land tenure configuration influences trajectories of land use change. However, simulations for various initial land-use configurations and compositions converge to similar states when land-tenure structure is held constant. For the scenarios considered, mean wildfire risk increases relative to the observed landscape. Increases in wildfire risk are not spatially uniform however, varying according to the composition and configuration of land use types. These unexpected spatial variations in wildfire risk highlight the advantages of using a spatially-explicit ABM/LUCC.
Previously, I wrote about Orrin Pilkey and Linda Pilkey-Jarvis' book, Useless Arithmetic: Why Environmental Scientists Can’t Predict the Future. In a recent issue of the journal Futures, Jerome Ravetz reviews their book alongside David Waltner-Toews' The Chickens Fight Back: Pandemic Panics and Deadly Diseases That Jump From Animals to Humans. Ravetz himself points out that the subject matter and approaches of the books are rather different, but suggests that "Read together, they provide insights about what needs to be done for the creation of a genuine science of sustainability".
Ravetz (along with Silvio Funtowicz) has developed the idea of 'post-normal' science - a new approach to replace the reductionist, analytic worldview of ‘normal’ science. Post-normal science is a "systemic, synthetic and humanistic" approach, useful in cases where "facts are uncertain, values in dispute, stakes high and decisions urgent". I used some of these ideas to experiment with some alternative model assessment criteria for the socio-ecological simulation model I developed during my PhD studies. Ravetz's perspectives toward modelling, and science in general, shone through quite clearly in his review:
"On the philosophical side, the corruption of computer models can be understood as the consequence of a false metaphysics. Following on from the prophetic teachings of Galileo and Descartes, we have been taught to believe that Science is the sole and certain path to truth. And this Science is mathematical, using quantitative data and abstract reasonings. Such a science is not merely necessary for achieving genuine knowledge (an arguable position) but is also sufficient. We are all victims of the fantasy that once we have numerical data and mathematical argument (or computer programs), truth will inevitably follow. The evil consequences of this philosophy are quite familiar in neo-classical economics where partly true banalities about markets are dressed up in the language of the differential calculus to produce justifications for every sort of expropriation of the weak and vulnerable. ‘What you can’t count, doesn’t count’ sums it all up neatly. In the present case, the rule of models extends over nearly all the policy-relevant sciences, including those ostensibly devoted to the protection of the health of people and the environment.
We badly need an effective critical philosophy of mathematical science. ... Now science has replaced religion as the foundation of our established order, and in it mathematical science reigns supreme. Systematic philosophical criticism is hard to find. (The late Imre Lakatos did pioneering work in the criticism of the dogmatism of ‘modern’ abstract mathematics but did not focus on the obscurities at the foundations of mathematical thinking.) Up to now, mathematical freethinking is mainly confined to the craftsmen, with their jokes of the ‘Murphy's Law’ sort, best expressed in the acronym GIGO (Garbage In, Garbage Out). And where criticism is absent, corruption of all sorts, both deliberate and unaware, is bound to follow. Pseudo-mathematical reasonings about the unthinkable helped to bring us to the brink of nuclear annihilation a half-century ago. The GIGO sciences of computer models may well distract us now from a sane approach to coping with the many environmental problems we now face. The Pilkeys have done us a great service in providing cogent examples of the situation, and indicating some practical ways forward."
Thus, Ravetz finds a little more value in the Useless Arithmetic book than I did. But equally, he highlights that the Pilkeys offer few, rather vague, solutions and instead turns to Waltner-Toews' book for inspiration for the future:
Pilkey's analysis of the corruptions of misconceived reductionist science shows us the depth of the problem. Waltner-Toes’ narrative about ourselves in our natural context (not always benign!) indicates the way to a solution."
Using the outbreak of avian flu as an example of how to tackle complex environmental in the 'risk society' in which we now live, Waltner-Toes:
"... makes it very plain that we will never ‘conquer’ disease. Considering just a single sort of disease, the ‘zoonoses’ (deriving from animals), he becomes a raconteur of bio-social-cultural medicine ...
What everyone learned, or should have learned, from the avian flu episode is that disease is a very complex entity. Judging from TV adverts for antiseptics, we still believe that the natural state of things is to be germ-free, and all we need to do is to find the germs and kill them. In certain limiting cases, this is a useful approximation to the truth, as in the case of infections of hospitals. But even there complexity intrudes ... "
Complexity which demands an alternative perspective that moves beyond the next stage of 'normal' science to a post-normal science (to play on Kuhn's vocabulary of paradigm shifts):
"That old simple ‘kill the germ’ theory may now be derided by medical authorities as something for the uneducated public and their media. But the practice of environmental medicine has not caught up with these new insights.
The complexity of zoonoses reflects the character of our interaction with all those myriads of other species. ... the creatures putting us at risk are not always large enough to be fenced off and kept at a safe distance. ... We can do all sorts of things to control our interactions with them, but one thing is impossible: to stamp them out, or even to kill the bad ones and keep the good ones.
Waltner-Toes is quite clear about the message, and about the sort of science that will be required, not merely for coexisting with zoonoses but also for sustainable living in general. Playing the philological game, he reminds us that the ancient Indo-European world for earth, dgghem, gave us, along with ‘humus’, all of ‘human’, ‘humane’ and ‘humble’. As he says, community by community, there is a new global vision emerging whose beauty and complexity and mystery we can now explore thanks to all our scientific tools."
This global vision is a post-normal vision. It applies to far more than just avian flu - from coastal erosion and the disposal of toxic or radioactive waste (as the Pilekys discuss for example) to climate change. This post-normal vision focuses on uncertainty, value loading, and a plurality of legitimate perspectives that demands an "extended peer community" to evaluate the knowledge generated and decisions proposed.
In all fairness, it would not be easy to devise a conventional science-based curriculum in which Waltner-Toes's insights could be effectively conveyed. For his vision of zoonoses is one of complexity, intimacy and contingency. To grasp it, one needs to have imagination, breadth of vision and humility, not qualities fostered in standard academic training. ... "
This post-normal science won't be easy and won't be learned or fostered entirely within the esoteric confines of an ivory tower. Science, with its logical rigour, is important. It is still the best game in town. But the knowledge produced by 'normal' science is provisional and its march toward truth is seemingly Sisyphean when confronted faced with the immediacy of complex contemporary environmental problems. To contribute to the production a sustainable future, a genuine science of sustainability would do well to adopt a more post-normal stance toward its subject.
"The future is ours, not to predict, but to create."
- Al Gore, 16th June 2008
Hear, hear. Spoken in the context of climate change, this might be interpreted as a slight against Global Circulation Models used by scientists. Rather, I think this should be interpreted as an indication that Gore understands that we need to move past discussions about whether we can use such models to 'prove' whether climate change is actually happening, and instead act to mitigate against undesired change.
Sven Erik Jørgensen introduces a recent issue of Ecological Modelling that presents selected papers from the International Conference on Ecological Modelling in Yamaguchi, Japan (28 August - 1 September 2006). The paper provides an overview of the model types available for ecological modelling, briefly highlighting the shift from a dominance of bio-geo-chemical dynamic models and population dynamics models in the 1970s toward the application of a wider spectrum of models. The emergence of new model types has come as a response to questions such as:
How can we describe the spatial distribution which is often crucial to understand ecosystem reactions?
How do we model middle number systems?
How do we model hetergenous populations and databases (e.g. observations from many different ecosystems)?
How do we model ecosystems, when our knowledge is mainly based on a number of rules/properties/propositions?
Jørgensen suggests there are at least 10 types of model currently available for modelling ecological systems (purely mathematical and statistical aside):
(Bio-geo-chemical and bio-energetics), dynamic models
Static models
Population dynamic models
Structurally dynamic models
Fuzzy models
Artificial neural networks
Individual-based models and cellular automata
Spatial models
Ecotoxicological models
Stochastic models
Hybrid models
Of these, my particular interest is in spatial models, individual-based models and cellular automata models (with a passing interest in population models). This is largely because of my background in geography and landscape ecology, but also because of the heterogeneity in patterns, processes and behaviour often exhibited in socio-ecological systems.
Jørgensen offers a short description of each type, before listing their advantages and disadvantages. Here are a couple with my comments in italics:
Individual-Based Models (IBMs)and Cellular Automata (CA) First, counter to Jørgensen, I would argue that CA models should be placed with the 'spatial models' - the ability of CA to represent space for me outweighs their potential to represent (limited) heterogeneity between cells. This aside, their grouping does make sense when we consider that these models can be relatively easily combined to represent individuals' interactions across space and with a heterogeneous environment (via the CA).
Advantages
Are able to account for individuality - agreed, especially for IBMs
Are able to account for adaptation within the spectrum of properties - yes
Software is available; although the choice is more limited than by bio-geo-chemical dynamic models - but excellent free modelling environments such as NetLogo make this type of modelling widely available
Spatial distribution can be covered - yes
Disadvantages
If many properties are considered, the models get very complex - and may require the adoption and development of new techniques to present/analyse/interpret output (e.g. POM, narratives)
Can be used to cover the individuality of populations; but they cannot cover mass and energy transfer based on the conservation principle - I see no reason why the principle of energy and mass conservation could not be achieved by models of these types
Require many data to calibrate and validate the models - yes, this often the case, and in some cases (again) may require new approaches and types of data to calibrate and evaluate models
Spatial Models Advantages
Cover spatial distribution, that is often of importance in ecology - yes, particularly Landscape Ecology, an entire discipline that has arisen since the 1970s and '80s
The results can be presented in many informative ways, for instance GIS - GIS is a means to organise and analyse data as well as present data
Disadvantages
Require usually a huge database, giving information about the spatial distribution - this can certainly give rise to the issue of 'model but no data' and increases the costs of performing ecological research by adding space to time. We have found that our large (~4,000 sq km) Upper Michigan study area demands high time and resources needed for data collection.
Calibration and validation are difficult and time-consuming - maybe more so than non-spatial models, but probably not as much as some individual-based models
A very complex model is usually needed to give a proper description of the spatial patterns - not necessarily. A model should be only as complex as the patterns and processes it seeks to examine and the inclusion of space does not imply patterns or processes any more complex than a system with less variables or interactions that is non-spatial.
This isn't a bad review of the types of ecological modelling being done. However, more incisive and useful insight could have been made with respect to landscape ecology and those models that are now beginning to attempt to account for human activity in ecological systems. [And it definitely could have been better written.] Maybe I'll stop criticising sometime and write one myself eh?
The theme of the meeting was the understanding of patterns, causes, and consequences of spatial heterogeneity for ecosystem function. The three keynote lectures were given by Gary Lovett, Kimberly With and John Foley. I found John Foley's lecture the most interesting and enjoyable of the three - he's a great speaker and spoke on a broader topic than the the others; Agriculture, Land Use and the Changing Biosphere. Real wide-ranging, global sustainability stuff. He highlighted the difficulties of studying agricultural landscapes because of the human cultural and institutional factors, but also stressed the importance of tackling these tricky issues because 'agriculture is the largest disturbance the biosphere has ever seen' and because of its large contribution to greenhouse gas emissions.
Presentations I was particularly interested in were mainly in the 'Landscape Patterns and Ecosystem Processes: The Role of Human Societies', 'Challenges in Modeling Forest Landscapes under Climate Change' and 'Cross-boundary Challenges to the Creation of Multifunctional Agricultural Landscapes' sessions.
In the 'human societies' session, Richard Aspinall discussed the importance of considering human decision-making at a range of scales and Dan Brown again highlighted the importance of human agency in spatial landscape process models. In particular, with regards modelling these systems using agent-based approaches he discussed the difficulty of model calibration at the agent level and stressed that work is still needed on the justification and evaluation phases of agent-based modelling.
The 'modeling forest landscapes' session was focused largely around use of the LANDIS and HARVEST models that were developed in and around Wisconsin. In fact, I don't think I saw any mention of the USFS FVS at the meeting whilst I was there, largely because (I think) FVS has large data demands and is not inherently spatial. LANDIS and HARVEST work at more coarse levels of forest representation (grid cell compared to FVS' individual tree) allowing them to be spatially explicit and to run over large time and space extents. We're confident we'll be able to use FVS in a spatially explicit manner for our study area though, capitalising on the ability of FVS to directly simulate specific timber harvest and economic scenarios.
The 'multifunctional agricultural landscapes' session had an interesting talk by Joan Nassauer on stakeholder science and the challenges it presents. Specific issues she highlighted were: 1. the need for a precise, operational definition of 'stakeholder' 2. ambiguous goals for the use of stakeholders 3. the lack of a canon of replicable methods 4. ambivalence toward the quantification of stakeholder results
Other interesting presentations were given by Richards Hobbs and Carys Swanwick. Richard spoke about the difficulties of 'integrated research' and the importance of science and policy in natural resource management. He suggested that policy-makers 'don't get' systems thinking or modelling, and that some of this may be down to the psychological profiles of the types of people that go into policy making. Such a conclusion suggests scientists need to work harder to bridge the gap to policy makers and do a better job of explaining the emergent properties of the complex systems they study. Carys Swanwick talked about the landscape character assessment, which was interesting for me having moved from the UK to the US about a year ago. Whilst 'wilderness' is an almost alien concept in the UK (and Europe as a whole), landscape character is something that is distinctly absent in the new world agricultural landscapes. Carys talked about the use of landscape character as a tool for conservation and management (in Europe) and the European Landscape Convention. It was a refreshing change from many of the other presentations about agricultural landscape (possibly just because I enjoyed seeing a few pictures of Blighty!).
Unfortunately the weather during the conference was wet which meant that I didn't get out to see as much of Madison as I would have liked. Despite the rain we did go on the Biking Fieldtrip. And yes, we did get soaked. It was also pretty miserable weather for the other fieldtrip to and International Crane Foundation center and the Aldo Leopold Foundation (more on that in a future blog), but interesting nevertheless.
Other highlights of the conference for me were meeting the former members of CSIS and eating dinner one night with Monica Turner. I also got to meet up with Don McKenzie and some of the other 'fire guys', and a couple of people from the Great Basin Landscape Ecology lab where I visited previously. And now I'm already looking forward to the meeting next year in Snowbird, Utah (where I enjoyed the snow this winter).
Distinction between deterministic models and stochastic models
He suggests that, generally, ecological models fall in two seemingly exclusive categories, deterministic models and stochastic models, and that either category of model can use physical or empirical approaches, or a combination of both (see figure). However, the classification He presents in the paper is developed according to how models represent
spatial processes,
temporal processes,
site-level succession, and
the intended use of the model.
Models are classified by succession based on whether the model uses succession pathways (i.e., a Markov state-and-transition approach), vital attributes (as I utilised in my PhD modelling), or by coupling landscape models with more detailed stand-level vegetation succession models. The fourth classification criteria above highlights that there are numerous applications of forest landscape models, and that design is strongly related to the desired applications. He suggests applications of forest landscape models generally fall into one of three categories:
spatiotemporal patterns of model objects,
sensitivities of model object to input parameters, and
comparisons of model simulation scenarios.
After developing and presenting the classification, the paper goes on to discuss two dilemmas facing those using forest landscape models. The first is the validation of model results, which has been discussed on numerous occasion elsewhere (including this blog). The discussion on circular reasoning is more novel however, (and related in some ways to what I have written with regards models of human agents):
"It is often difficult to separate expected results from emergent results. A caution against circular reasoning is the caveat often encountered in this situation, where researchers discuss biological or environmental forcing (causes) of their modeled results, whereas the forcing (causes) is actually built in the model formulation to derive such results. It should be pointed out that most model simulations do not lead to new understanding of the modeled processes themselves. The primary and subsequent results simply reflect the relationships used in building the models, which in turn reflect current understanding of the processes. The findings of these models are simply the spatiotemporal variations of the spatial process (discussed in Section 5.1), not the mechanisms that drive the potential changes of the spatial process. Emergent results are generally those resulted from the interactions and feedbacks of model objects."
The paper concludes by summarizing likely development of forest landscape modelling in the future:
Model development will move from the foci of theoretical and exploratory purposes to the foci of strategic and tactical purposes with increasing model realism, responding to the needs of forest management and planning.
Multiple spatial and temporal resolutions will be implemented for different processes
Standardized module components may emerge as handy utilities that are ready to be plugged into other models. Since component-based models provide non-developers or end users with access to model components, a component-based model can be more rigorously tested, evaluated, and modified than before, and thus, model development processes can be driven not solely by original developers, but by the broader scientific community
Synchronization of multiple ecological processes can be made possible with multiple computer processors. This will help deal with the limitation that ecological processes are simulated in a sequential order as determined by the executable program.
Model memorization will be improved so that a forest landscape model not only memorizes vegetation, disturbance, and management status at the current and previous model iteration, but also the entire temporal sequence. This would allow more effective studies of legacies of forested landscapes responding to various disturbance and management activities.
Abstract Previous model classification efforts have led to a broad group of models from site-scale (non-spatial) gap models to continental-scale biogeographical models due to a lack of definition of landscape models. Such classifications become inefficient to compare approaches and techniques that are specifically associated with forest landscape modeling. This paper provides definitions of key terminologies commonly used in forest landscape modeling to classify forest landscape models. It presents a set of qualitative criteria for model classification. These criteria represent model definitions and key model implementation decisions, including the temporal resolution, number of spatial processes simulated, and approaches to simulate site-level succession. Four approaches of simulating site level succession are summarized: (1) no site-level succession (spatial processes as surrogates), (2) successional pathway, (3) vital attribute, and (4) model coupling. Computational load for the first three approaches is calculated using the Big O Notation, a standard method. Classification criteria are organized in a hierarchical order that creates a dichotomous tree with each end node representing a group of models with similar traits. The classified models fall into various groups ranging from theoretical and empirical to strategic and tactical. The paper summarizes the applications of forest landscape models into three categories: (1) spatiotemporal patterns of model objects, (2) sensitivities of model object to input parameters, and (3) scenario analyses. Finally, the paper discusses two dilemmas related to the use of forest landscape models: result validation and circular reasoning.
Keywords Forest landscape models; Spatially explicit; Spatially interactive; Definitions; Model characterization; Model classification
KML files are used in Google products, such as Google Earth or Google Maps, to display geographic data. The data Mark has posted on the King's server are freely accessible to all for non-commercial use. you can visualise the data in Google Earth and, in many cases, links to the actual downloadable GIS files also provided. Many of the datasets are works in progress and new data will continue to be posted in the future, so keep checking back.
The availability of data such as these, and projects such as Pete's, really show how Google Earth can be used for so much more than virtual tours of other places or previews of you next holiday destination... [Speaking of which, I'm off to Utah snowboarding next week so hopefully I'll have some new pics to post on my own Google-enabled photos page.]
I like this video. Less because of the message toward the end about the importance of ensuring western countries continue to train adaptable workforces in an increasingly flat world. More because of how it illustrates the speed and unpredictability of change. In hindsight it might seem obvious that this is how the world should end up - contingency matters in the real world after all. But in these contingent, historical, systems how do we generate a model for the future that we can trust with any useful degree of confidence?
During my modelling antics these last couple of days I seem to have been using many of the add-ins I've have installed with the software I use regularly. I thought I'd highlight some of them here as they are really useful tools that can expand the modelling and data manipulation possibilities of a standard software install.
Much of the modelling I do is spatial, so I'm regularly using some form of GIS. I'm most familiar with the ESRI products, but have also tinkered with things like GRASS. Two free add-ins that are really useful if you use ArcMap regularly are the Patch Analyst and Hawth's Tools. Patch Analyst facilitates the spatial pattern analysis (making use of FRAGSTATS) of landscape patches, and the modelling of attributes associated with patches. Hawth's Tools is an extension for ArcMap that performs a number of spatial analyses and functions that you can't do with the standard install of ArcMap. Most of the tools are written with ecological analyses in mind, but it's also be useful for non-ecologists with functions such as conditional point sampling, kernel density estimation and vector layer editing.
Although it is generally frowned upon for statistics (use R - see below), Microsoft Excel isn't a bad tool for organising small and medium-sized data sets and for doing basic systems modelling (spatial simulation is a little trickier). Developed by some guys at CSIRO, Pop Tools is a free add-in for PC versions of Excel that facilitates analysis of matrix population models and the simulation of stochastic processes. It was originally written to analyse ecological models, but has been used for studies of population dynamics, financial modelling, and the calculation of bootstrap and resampling statistics. Once installed, PopTools puts a new menu item in Excel's main menu and adds over a hundred useful worksheet functions. Regardless of whether you intend to do any modelling in Excel or not, the ASAP Utilities add-in is a must for automating many frequently required tasks (including those you didn't even know you wanted to do in Excel!). There are selection tools (such as 'select cell with smallest number'), text tools (such as 'insert before current value'), information tools (such as 'Find bad cell references (#REF!)') and many more.
If you're going to be doing any serious statistical analyses the current software of choice is R, the open-source language and environment for statistical computing and graphics. R provides a wide variety of statistical (linear and nonlinear modelling, classical statistical tests, time-series analysis, classification, clustering, ...) and graphical techniques. If you need to analyse or manipulate large datasets R is for you - you are only restricted by the memory available on your computer. For computationally-intensive tasks, C, C++ and Fortran code can be linked and called at run time. R is also highly extensible by installing optional packages that have been written by users from around the world.
Many of the packages I use are from the Spatial and Environmetrics Task views. For example, I use spdep for calculating spatial autocorrelation, VR for computing spatial correlograms or confidence intervals for model parameters, and hier.part for hierarchical partitioning. This week I started thinking about how I will use the yaImpute package to impute the stand vegetation data we have collected at specific points in our study area across the entire landscape ready to initialise our spatial simulation model. Download the R software and the individual packages from a CRAN mirror near you.
Of course, this is just the tip of the iceberg and only a few of the most useful add-ins for the most commonly used software. For a much more complete list of more technical software and programming tools for ecological and environmental modelling see Andrea Emilio Rizzoli's 'Collection of Modelling and Simulation Resources on the Internet' or the list of Ecological Modelling links by T. Legovic' and J. Benz.
Forest Ecology and Management Special Issue: Forest Landscape Modeling
In June 2006 the China Natural Science Foundation and the International Association of Landscape Ecology sponsored an international workshop of forest landscape modelling. The aim of the workshop was to facilitate a discussion on the progress made in the theory and application of forest landscape models. A special issue of Forest Ecology and Management, entitled Forest Landscape Modeling - Approaches and Appplications [Vol. 253, Iss. 3], presents 12 papers resulting from that meeting. In their editorial, He et al. summarise the papers, organising them into three sections that describe current activities in forest landscape modelling: (1) effects of climate change on forest vegetation, (2) forest landscape model applications, and (3) model research and development.
The LANDIS model is used in several of the papers on climate and human management of forest systems. Advances in the representation of processes that propagate spatially, including fire and seed dispersal, are discussed in several of the papers examining model research and development. He et al. conclude their editorial by reiterating why landscape models are a vital tool for better understanding and managing forested regions of the world:
The papers represented in the special issue of forest landscape modeling highlight the advances and applications of forest landscape models. They show that forest landscape models are irreplaceable tools to conduct landscape-scale experiments while physical, financial, and human constraints make real-world experiments impossible. Most of the results presented in this issue would not have been possible without the use of forest landscape models. Forest landscape modeling is a rapidly developing field. Its development and application will continually be driven by the actual problems in forest management planning and landscape-scale research. We hope that the papers contained in this special issue will serve both researchers and managers who are struggling to incorporate large-scale and long-term landscape processes into their management planning or research.
As I mentioned before, the Global Land Project website is experimenting with the use of webcasts to enable the wider network to "participate" and use the GLP webpage as a resource. For example, several presentations are available for viewing from the Third Land System Science (LaSyS) Workshop entitled 'Handling complex series of natural and socio-economic processes' and held in Denmark in October of 2007. One that caught my attention was by Tom Veldkamp, mainly because of its succinct title: Advances in Land Models [webcast works best in IE].
Presented in the context of other CHANS research, Veldkamp used an example from the south of Spain to discuss recent modelling approaches to examine the effects of human decisions on environmental processes and the feedbacks between human and natural systems. The Spanish example examined the interaction of human land-use decision making and soil erosion. A multi-scale erosion model, LAPSUS, represented the interactive natural and human processes occurring olive groves on steep hillslopes; gullying caused by extreme rainfall events and attempts to preserve soils and remove gullies by ploughing. Monte Carlo simulations were used to explore uncertainties in model results and highlighted the importance of path dependencies. As such, another example of the historical dimension of 'open' systems and the difficulties it presents for environmental modellers.
The LAPSUS model was coupled with the well known land use/cover change CLUE model to examine feedbacks between human land use and erosion. The coupled model was used to examine the potential implications of farmers adopting land use practices as a response to erosion. Interestingly, the model suggested that human adaptation strategy modelled would not lead to reduced erosion.
Veldkamp also discusses the issue of validating simulation models of self-organising processes, and suggests that ensemble and scenario approaches such as those used in global climate modelling are necessary for this class of models. However, rather than simply using 'static' scenarios that specify model boundary conditions, such as the IPCC SRES scenarios, scenarios that represent some form of feedback with the model itself will be more useful. Again, this comes back to his point about the importance of representing feedbacks in coupled human and natural systems.
For example, Veldkamp suggests the use of "Fuzzy Cognitive Maps" to generate 'dynamic' scenarios. Essentially, these fuzzy cognitive maps are produced by asking local stakeholders in the systems under study to quantify the effects of the different factors driving change. First, the appropriate components of the system are identified. Next, the feedbacks between these components are identified. Finally, the stakeholders are asked to estimate how strong these feedbacks are (on a scale of zero to one). This results in a semi-quantitative systems model that can be run for several iterations to examine the consequences of the feedbacks within the system. This method is still in development and Veldkamp highlighted several pros and cons:
Pros:
it is relatively easy and quick to do
it forces the stakeholders to be explicit
the emphasis is placed on the feedbacks within the system
Cons:
it is a semi-quantitative approach
often feedbacks are of incomparable units of measurement
time is ill defined
stakeholders are often more concerned with the exact values they put on an interaction rather than the relative importance of the feedbacks
I agree when Veldkamp suggests this 'fuzzy cognitive mapping' is a promising approach to scenario development and incorporation into simulation modelling. Indeed, during my PhD research I explored the use of an agent-based model of land use decision-making to provide scenarios of land use/cover change for a model of forest succession-disturbance dynamics (and which I am currently writing up for publication). 'Dynamic' model scenario approaches show real promise for representing feedbacks in coupled human natural systems. As Veldkamp concludes, these feedbacks, along with the non-linearities in system behaviour they produce, need to be explicitly represented and explored to improve our understanding of the interactions between humans and their environment.
4.0.2 and 3.1.5 are both bugfix releases, addressing various issues found in 4.0 and 3.1.4.
NetLogo 3D Preview 5 brings the NetLogo 3D series up to date with NetLogo 4.0.2. It includes the majority of NetLogo 4.0's features and all of 4.0.2's fixes.
Seeing the Wood for the Trees: Pattern-Oriented Modelling
A while back I wrote about the potentially misplaced preoccupation with statistical power in species distribution models. Our attempts at drawing out some relationships between our deer distribution data and descriptors of land cover is proving taxing - the relationships evident at a more coarse spatial resolution (e.g. county level) than we are considering aren't found in our stand-level data. As a result we moving toward taking a modelling approach that is driven less by our empirical data and more by inferences based on multiple information sources. Particularly I'm drawn toward emphasising an approach I first encountered in my undergraduate landscape ecology class taught by George Perry - 'Pattern-Oriented Modelling'.
A prime example of the POM approach is its use to model the spread of rabies through central Europe. The rabies virus has been observed to spread in a wave-like manner, carried by foxes. Grimm et al. (1996) describe how they developed a cellular automate-type model that considers cells (of fox territory) to be in either a healthy, infected or empty state. Through an iterative model development process, their model was gradually refined (i.e. its assumptions and parameters modified) by comparing model results with empirical patterns.
The idea underpinning this iterative POM approach is
"... if we decide to use a pattern for model construction because we believe this pattern contains information about essential structures and processes, we have to provide a model structure which in principle allows the pattern observed to emerge Whether it does emerge depends on the hypotheses we have built into the model."
This approach has been found particularly useful for the development of 'bottom-up' agent-based models. Often understanding of the fine-scale processes driving broad-scale system dynamics and patterns is poor, making it difficult to both structure and parameterise mechanistic models. However, whilst the logical fallacy of affirming the consequent remains, if a model of low-level interactions is able to reproduce higher-level patterns, we can be confident that our model is a better representation of the system mechanics than one that doesn't. Furthermore, the more patterns at different scales that the model reproduces, the mode confident we can be in it. Thus, in POM
"multiple patterns observed in real systems at different hierarchical levels and scales are used systematically to optimize model complexity and to reduce uncertainty."Grimm et al. (2005)
Grimm and Berger outline the general protocol of a pattern-oriented modelling approach (whilst reminding us that there is no standard recipe for model development):
Formulate the question or problem
Assemble hypotheses about essential processes and structures
Assemble (observed) patterns
Choose state variables, parameters and structures
Construct the model
Analyse, test and revise the model
Use patterns for parameterisation
Search for independent predictions
Several iterations of this process will be required to refine the model. In initial iterations, steps 2 and 4 may need to be largely inferential if the state of knowledge about the system is poor. However, by moving iteratively back through these steps, and in particular exploiting steps 6 and 7 to inform us about model performance relative to system behaviour, we can improve our knowledge about the system whilst simultaneously ensuring our model recreates observed patterns. For example, during the development of the landscape fire-succession model in my PhD, I compared the landscape-level model results of different sets of (unknown) flammability probabilities (parameters) of each vegetation type required by the model with empirically observed wildfire regime behaviour. By modifying parameters for individual vegetation types I was able to reproduce the appropriate wildfire frequency-area distribution for Mediterranean-type environments that had previously been found (I'm currently writing this up for publications - more soon).
But what does this all have to do with our model of the relationship between deer browse and timber harvest in Michigan's Upper Pensinsula? Well, right now I think we're at steps 2,3 and 4 (all at the same time). As our deer and land cover relationships are weak at the stand-level (which is the level we are considering so that we can integrate the model with an economic module), I am currently developing hypotheses (i.e. assumptions) about the structure of the system from previous research on different specific aspects of similar systems. Furthermore, we're continuing to look for spatial patterns in both vegetation and deer distribution so that we can compare the results of our hypothetical model.
For example, one thing I'm struggling with right now is is how to establish the probability of which individual trees (or saplings) will be removed from a stand due to a given level of deer browse (which in turn is dependent upon a deer density). This is not something that has been explicitly studied (and would be very difficult to study at the landscape level). Therefore we need to parameterise this process in order for the model to function. We should be able to do this by comparing several different parameterisations to empirically observed patterns such as spatial configuration of forest types classified by age class or age/species distributions at the stand-level. That's the idea anyway - we'll see how it goes over the next months...
In the meantime, next week I head back to the study area for the first stage of our seedling experiment. We're planting seedlings now across a gradient of browse and site conditions with the intention of returning in the spring to see what has been browsed and count deer pellets. This should improve our understanding of the link between pellet counts and browse pressure and provide us with some more empirical patterns which we can use in our ongoing model development.
In their feature Formulae for the 21st Century, Edge ask 'What is your formula? Your equation?' Scientists, Philosophers, Artists and Writers have replied. Some gave their favourite, or what they thought to be the most important, formulas from their fields.
My favourites are the most simple - model parsimony, Occam's Razor and all that. Here are a couple (click for larger images).
This got me thinking about why I like quotes so much too - because they're models. Take the essence of an idea and express it as elegantly as possible. That's what scientists and mathematicians do, but equally it's what writers and artists do. Take it far enough, and being a bit of critical realist, I would say that all human perception is a model. But these elegant models are more useful than our sensory apparatus alone (which, along with our subconscious does plenty of filtering already) - they observe whilst simultaneously interpreting and synthesizing.
So what's my model? I'm not sure - it would have to involve change. My personal models are continually changing, vacillating. Sometimes I believe time has an arrow, sometimes it doesn't. Sometimes the world is equations and energy, sometimes it's story and sentiment. Sometimes life is light, sometimes life is heavy. Even when my model is relatively stable it's usually paradoxical (or should that be hypocritical?) and ironic. I'll try to parse it down to it's most parsimonious state and then find some words and symbols to express it elegantly. Then I'll post it here. I can't guarantee that will be any time soon mind you...
Integrated, multi- and inter-disciplinary studies are becoming increasingly demanded and required to understand the consequences of human activity on the natural environment. In a recent paper, Sandra Lavorel and colleagues highlight the importance of considering the feedbacks and interactions between several systems when examining landscape vulnerabilities to fire. They present a framework for integrated fire research that considers the fire regime as the central subsystem (FR in the figure below) and two feedback loops, the first with consequences for atmospheric and biochemical systems (F1) and the second that represents ecosystems services and human activity (F2). It is this second feedback loop in their framework that my research focuses.
To adequately quantify the fire-related vulnerability of different regions of the world the authors suggest that a better understanding of the relative contributions of climate, vegetation and human activity to the fire regime is required. For example, they suggest that an examination of the statistical relationships between spatio-temporal patterns evident in wildfire regimes and data on ecosystem structure, land use and other socio-economic factors. We made a very similar point in our PNAS paper and hope to continue to use the exponent (Beta) of the power-law frequency-area relationship that is evident in many model and empirical wildfire regimes to examine these interactions. One statistical relationship that might be investigated is between Beta and the level of forest fragmentations, thought to be a factor confounding research on the effects of fire suppression of wildfire regimes.
But the effects of landscape fragmentation can also be examined in a more mechanistic fashion using dynamic simulation models. Lavorel et al. mention the impacts of agricultural abandonment on the connectivity of fuels in Mediterranean landscapes which are attributed, in conjunction with a drier than average climate, to the exceptionally large fires that burned there during the 1990s. My PhD research examined the impacts of agricultural land abandonment on wildfire regimes in central Spain. I'm currently working on writing this work up for publication, but I plan on continuing to develop the model to more explicitly represent the F2 feedbacks loop shown in the figure above.
The potential socio-economic consequences of changing fire regimes are an area with a lot of room to improve our understanding. For example, some regions of the world, such as the Canadian boreal forest, are transitioning from a net sink for carbon to a net source (due to emission during burning). If carbon sinks are considered in future
