By Michael Bode (This article first appeared in the Spring 2013 newsletter of the Resource Modelling Association. The newsletter and society journal – Natural Resource Modelling – contain some great articles on modelling that contains both biological and economic processes).
There is a famous, possibly apocryphal story about the genesis of Operations Research in the Second World War. According to the story, mechanics repairing flak-damaged Lancaster bombers noticed that the damage tended to occur with unusual frequency in particular sections of the aircrafts’ undercarriage. Additional armor was going to be added to these sections of the aircraft, in the hope that more would survive the frequent daylight bombing raids that were being planned. However, Bomber Command was one of the first sections of the Allied military to apply scientific analysis to their tactical military decisions, and the analysts in their Operations Research branch (the famous BC-ORS) made the counter-intuitive recommendation that armour be added to the undamaged parts of the plane. Their insight was based on an understanding of survivorship bias: these planes had returned precisely because the damaged areas were unimportant.
The use of Operations Research in conservation management began with an analogous story of unexpected revelation. In the late 1980s, environmentalists and state governments in eastern Australia could congratulate themselves on the large tracts of land that were gazetted as National Parks. Although it had been long recognised that terrestrial protected areas were often placed on land with fewer economic uses, conservation scientists began for the first time to quantitatively assess the cumulative value of these protected areas, of the land that had survived clearing and degradation. Specifically, they began to work out how many species were protected within the national reserve network, how many remained unprotected, and whether the National Park system could have done a better job. Their results showed that measuring the percentage of land protected failed to account for the “diversity” in biodiversity. While extensive, Australia’s National Park network reflected a history of inefficient decision-making. By protecting the land that no one else wanted, conservation managers had armoured the wrong sections of the conservation airplane – parts that didn’t need the extra protection. Fortunately, their research pointed towards approaches that would offer better protection to biodiversity in the future. These were the foundational ideas of “systematic conservation planning”, and through this field, the techniques and approaches of Operations Research (OR) have gone on to shape how we use a large proportion of the Earth’s surface.
Through the lessons of systematic conservation planning, quantitative analyses and OR have become central to conservation land management decisions. In the last decade, “cost-efficiency” analyses and “return-on-investment” tools have been used to plan the allocation of government budgets among threatened species, to decide which invasive species pose the most pressing risk to conservation values, and to direct monitoring and information gathering about the distribution and state of biodiversity. The methods used to find solutions to these questions are taken straight from the OR toolbox: simulated annealing, linear and dynamic programming, Bayesian analysis and Markov processes. With their help, prospective and retrospective analyses have shown that in comparison with ad hoc decisions, an OR approach to conservation decision-making can realise dramatic efficiencies, and greatly increase the benefits achieved by ever-limited conservation budgets.
Nevertheless, given the scale of the potential conservation benefits, and given the wide acknowledgement that conservation management is substantially underfunded, the uptake of quantitative decision-support tools has been less rapid in other sectors of conservation management. The backgrounds of conservation managers (generally in ecology or natural history), often fail to provide strong quantitative skills, and creates managers whose natural history training makes them inherently adverse to the simplification frequently required by decision modelling. Yet, the benefits of OR approaches are becoming increasingly apparent, and in recent years have been widely publicised. These benefits can be broadly categorised into three main areas.
Solving complicated problems
Ecosystems are notoriously difficult to understand, and difficult to predict. Even a small landscape can involve millions of individuals, from thousands of species, made interrelated and interdependent by a long history of coexistence and coevolution. It’s not at all surprising that ecology provided some of the first compelling examples of how unpredictable nonlinear dynamical systems can be.
However, as well as being complex, conservation systems are simply complicated: they often involve choosing between a very large numbers of options, based on a wide range of objectives (e.g., biodiversity outcomes, ecosystem services, economic and social costs). Even if stakeholders are able to express their values unambiguously, and even if the system dynamics are clearly and quantitatively understood, its still difficult to apply this information to a very large problem without systematic decision-support. From this perspective, OR is simply the extension of reasonable principles to large problems. Anyone can use mental calculations to identify the largest value from a list of five numbers. However, once that list contains five thousand different numbers, or five million, it makes sense to reach for a computer.
The 2002 rezoning of the Great Barrier Reef Marine Park provides an illustrative example of this particular OR benefit. Management wanted to ensure that the new network of marine protected areas protected a set percentage of every type of habitat found in the reef ecosystem. Datasets had been created which described how much of each habitat type could be found in each planning unit, and it simply remained to choose a set of planning areas that was as small and cohesive as possible, while meeting a protection target for each habitat type. Of course, this process had to be applied to literally trillions of options, and so OR techniques were needed to extend a handful of straightforward principles to this very large scale.
Moderating the fallability of human decision-makers
The human brain’s decision-making difficulties are not limited to large problems with combinatorial complexity: there are situations where humans would fail to identify the right solution, even in a short list of five options. Thanks to the insights offered by behavioural psychologists, it is increasingly clear that human decision-makers are subject to a wide range of “cognitive biases”, which affect their ability to dispassionately consider certain questions, and make particular decisions. As an example, human decision-makers will consistently and disproportionately believe hypotheses and theories that identify ourselves as active and influential agents of change. We will believe that we can make a difference, even in the absence of evidence, and even in situations when events are completely out of our control. This cognitive bias, known as “the illusion of control”, will affect any conservation decision-maker’s ability to naturally assess the effectiveness of a management intervention, and will do so in a predictable and significant manner.
Psychologists have identified a large number of these cognitive biases, and have shown their effects in a multitude of decision-contexts. Many can be avoided entirely by the application of decision-support methods. Others can be mitigated by particular decision-making structures, and information-eliciting structures (e.g., Delphi methods). OR methods are impervious to these biases for obvious reasons. For example, the illusion of control is partly explained by a human desire to make a difference – to justify our own existence. Mathematical techniques and computer algorithms do not share such existential motivations.
The field, science, and management of conservation have increased exponentially in the 30 years since the term “conservation biology” was coined, and in the 50 years since the environmental movement became politically active. The amount of money and resources spent on conservation management has increased in parallel, and conservation is today undertaken with all the resources of national governments, billionaire philanthropists, transnational agencies, and international nongovernmental organisations. However, in the last few years, a series of high-profile audits of conservation actors has revealed a level of accountability that does not reflect this new status. These analyses have consistently found that despite the investment of enormous resources – literally hundreds of billions of dollars – over long time periods, conservation managers are unable to identify which of their potential actions are the most efficient. In most cases, it’s not even possible to identify whether commonly applied methods are even preferable to doing nothing. Moreover, because of misdirected, poor, or nonexistent record-keeping, it’s impossible to go back and re-assess the outcomes to identify the relative performance of alternatives. Put bluntly, despite spending millions of dollars, we still haven’t learned what works, and what doesn’t.
However, conservation managers and agencies are learning from the mistakes that these audits have revealed, and the resulting changes are moving them in the direction of OR. To apply OR techniques to conservation problems, managers are forced to undertake a number of actions that indirectly ensure that their decisions are both transparent and auditable. OR approaches are by necessity explicit about their objectives. They require a clear and process-based explanation of the system dynamics, and quantitative predictions about the relative performance of alternatives. Because learning is almost always embedded in an OR framework, and because this framework is honest and clear-eyed about uncertainty, resource allocation will help identify which (if any) of the proposed management actions are effective.
There are numerous persuasive scientific reasons for using an OR approach in conservation management, and a great deal of accumulated scientific evidence to show that biodiversity will be more effectively protected by such systematic methods. Yet despite all of this scientific evidence, and the support of many conservation scientists, OR will likely become the norm in conservation based on the arguments of accountants and government ombudsmen, and for the relatively banal rationale of transparency and accountability.
Can there be too much transparency?
As of yet, this difficult issue has not yet been considered by conservation scientists. First, the majority of our focus and interest is on the biological, ecological and evolutionary aspects of conservation. It is revealing that conservation science is much more concerned with the evolutionary impacts of protected areas – that is, the impacts of protection on speciation across the coming thousand years – than with how human responses to protected areas might undermine their effectiveness in the next five years. Second, most of our science, and all of our management techniques are based on the assumption that the conservation landscape is a static one. Anecdotally, conservation organizations take strategic considerations into account in an ad hoc manner, generally based on the experience of individual managers. However, systematic conservation planning, which is so effective at calculating how to represent myriad conservation features in a protected area network, currently does not anticipate the possibility that opposing or reactive forces might adjust their behaviour in response to that protected area network.
Moreover, and perhaps more problematically, it is unclear how many of the most important conservation actors can avoid or deal with this problem. If transparency can undermine conservation outcomes, then the obvious response is to avoid transparency, or even engage in active dissembling. This might be possible for some privately-funded conservation organisations, whose objectives, decision-rules and strategies can and are kept confidential. However, for government agencies, and for nongovernmental conservation groups who rely on donations from large numbers of donors, stakeholders have a reasonable expectancy of transparency that cannot, and perhaps should not be circumvented. Transparent conservation therefore appears to be a conundrum.
Given that the problem is game theoretic, it is perhaps possible to look towards game theory for some ideas about whether this Gordian knot can be untied. In the absence of the ability to keep information in confidence, game theory would suggest that the next best option is to identify the objectives of the opposing actors, and then optimise our decisions on the assumption that other actors will behave rationally. Some advice and direction can be taken from bioeconomic modelling, where policy is constructed with the understanding that other actors in the system will alter their behaviour in response to (i) the conservation actions, (ii) their understanding of conservation objectives, and (iii) their own personal objective function. But the most useful, and the most directly relevant advice will only come if conservation scientists observe and research these phenomena in conservation contexts; if conservation turns its attention – at least temporarily – away from ecology and towards the dynamics of the broader, socio-economic landscape.