Tuesday, February 28, 2012

Does monitoring make the man?

AliĆ©nor Chauvenet and co-authors published an interesting article in Animal Conservation last week. They used mark-recapture data on Hihi to parameterize a stochastic population model and evaluate the benefits of supplemental feeding of a translocated population. This is a really solid piece of work: an interesting species with a nice simple management decision. This will definitely make it into my Population Dynamics course next year as an example. They have this to say about Adaptive Management in the introduction:
In situ food experiments (on–off or temporal and/or spatial variation in quantity) can help assess the consequences of altering management actions (Armstrong & Perrott, 2000). However, managers rarely take this risk as translocated populations are generally small (Shaffer, 1981)
and such experiments could result in the loss of precious translocated individuals. Alternatively, models can be used to study past and future variation in management regimes and assess the importance of such variation on a species’ survival and/or reproductive rates. The goal of this type of modelling exercise is to inform and update management decisions as an iterative process, that is, perform adaptive management (Holling, 1978; Walters & Hilborn, 1978; Walters, 1986). Ideally, adaptive management requires an a priori development of possible management options, which are evaluated and refined following targeted monitoring (Ewen & Armstrong, 2007). In many cases, however, new management options arise well into a project. If relevant monitoring has been ongoing, then population modelling can inform the likely response of populations based on past data, and new management can be incorporated into the adaptive management framework (Williams, 2010).
They assume that adaptive management requires experimentation, and seem to believe that introducing new management actions into an AM process is a relatively new idea. It isn't. At least in the Decision Theoretic school, the possibility of changes to the available actions or shifts in objectives is considered regularly as part of the iterative cycle - so called "double loop learning". Such double loop learning is also not dependent on monitoring data - you may learn things outside of any monitoring program, e.g. from independent research, changes in policies enabling new actions etc.
So, a long term monitoring dataset, clear management decision, nice models for forecasting the future. But is it Adaptive Management? I have to say I can't tell. It is clear that at least one decision, to cap the ad libitum feeding program in 2010, was made by trading off one objective - high adult survival, against another objective - cost. What isn't clear is whether the models developed by Chauvenet et al were used to evaluate the future consequences of that decision. One quote makes me think not:

Investigating other management scenarios, such as ones looking at the impact of reducing or increasing supplemental feeding by x% would be highly informative but data did not allow such models to be built. However, a new management regime has been put in place on Kapiti Island recently. In late 2010, managers reached the end of their ad libitum capacity and were forced to make a decision as to the future of management for the population. They came to the conclusion that capping the quantity of supplemental food to 75% of the 2009 amount was the best solution for both hihi and managers. As a result there may be a possibility for further model parameterization, that is, new scenarios, in the near future.

So they suggest that the model cannot be used to evaluate the effect of the new management regime until after the new regime has been in place sufficiently long to have data on its effectiveness. Hogwash, I say! They know the parameters of the model in the absence of feeding, and with ad libitum feeding. Surely a reasonable null hypothesis draws a line between those two points to get an idea of how capping feeding will affect population size. By making that prediction prior to changing management, or even now, they would be able to use subsequent observations of the population to test the validity of their population model.
So, I have to say that it doesn't look like Adaptive Management, although I think it is clearly one of those decisions that could benefit from a rigourous Decision Theoretic approach to AM.

Thursday, February 23, 2012

Soft systems thinking seems squishy

Georgina Cundill from Rhodes University in South Africa and some co-authors have an essay in the latest Conservation Biology entitled "Soft Systems Thinking and Social Learning for Adaptive Management". I've gotten interested in the literature on social learning as a result of some recent interactions with colleagues in political science, but I'd never heard of soft systems thinking before. The motivation for the paper is simple: "It is now generally accepted that social and political processes can determine whether management initiatives succeed irrespective of the quality of the science that supports them ..." and so "hard systems thinking" (which is what I do) will fail. They define AM by reference to Carl Walter's seminal book, but by assuming that AM is a monolithic concept they muddy the waters considerably. For example, when they assert that
Adaptive management often starts with a conceptual model or set of objectives or hypotheses to be tested, and then experimentation is used to validate, refute, and, ultimately, modify and refine the model and to make informed trade-offs among goals that may conflict ...
they are largely referring to actions that define the Experimental-Resilience school, but slip in decision theoretic ideas of objectives and trade-offs that are are rarely, if ever, the focus of Experimental Resilience approaches. In contrast, their definition of "hard systems thinking" as

decision making in pursuit of goals or objectives. Here we refer to this approach as objective-based management. This approach is evident in the step-by-step process of adaptive management
that begins with the identification of objectives.
which is the basis for Decision Theoretic approaches to AM.
So what does shifting to soft systems thinking add? I struggled to find a clear definition to quote - but the idea seems to be that a soft systems approach includes the people as part of the system. Wow, that sounds like a socio-ecological system! As a result, the system cannot be engineered towards an optimum, because the purpose of the system is an emergent property of the interactions among the people involved.
The idea of social learning is less squishy - they define it as

the collective action and reflection that takes place among both individuals and groups
when they work to understand the relations between social and ecological systems; it is conceptualized as a process of transformative social change in which participants critically question and potentially discard existing norms, values, institutions, and interests to pursue actions that are desirable to them. 
So if you're talking about a socio-ecological system, then social learning is the process by which the social components of the system respond to new information.


They then describe a new methodology for AM derived from these processes. The methodology consists of 4 assumptions and 4 actions. Their assumptions are so ambiguous as to be almost tautologically true of any socio-ecological system, so I won't repeat them. So what actions do they recommend?

  1. Situate and engage rather than defining objectives, figure out what the problem is from as many different perspectives as possible, and determine who is interested in the problem. 
  2. Raise awareness and encourage Enquiry and Deconstruction clarify and refine different frames of reference among the stakeholders, leading to the development of shared frames of reference.
  3. Take collaborative actions based on co-created frames of reference, and that are agreed upon by all the actors.
  4. Reflect on learning to continue the process of modifying frames of reference of all the actors. 
They go on to outline challenges to implementation, which include the observation that all of this is context specific (making general procedural recommendations impossible), and that conservation scientists lack any kind of training in the skill sets relevant to these sorts of social processes. 
It seems to me that the only real difference is whether one is taking a proscriptive stance on decision making versus a descriptive stance. Hard systems approaches are proscriptive - they describe what you should do, in which order, and provide recipes for carrying out each step. In contrast, this soft systems/social learning approach is describing what actually happens when a group of actors tries to manage a resource. 

My personal view is that fruitful progress involves collaboration between hard and soft systems thinking. Without understanding the social dynamics, hard systems approaches risk spending resources (people's time, mostly) without gain, and soft systems approaches are, well, too soft to provide useful guidance in all circumstances. There are situations where it is OK to be as hard as a rock, and situations where the best strategy is to be soft and squishy. What we need are frameworks to help us divine the appropriate mix of strategies in any particular situation.

Wednesday, February 22, 2012

Resilience

My 4th floor colleague Craig Allen and his collaborators have a new article on "Managing for Resilience" to appear in the next issue of Wildlife Biology. It is a pretty good up-to-date description of what I call the Experimental-Resilience school approach to Adaptive Management. They contrast command-and-control management of single species with "managing for resilience" - which necessarily involves ER style AM.
So what is resilience? In their words resilience is the

measure of the amount of change or disruption that is required to transform a system from being maintained by one set of mutually reinforcing processes and structures to a different set of processes and structures.
which is a definition that I like, but is hard to operationalize. If you can write down a system of equations describing the evolution of a system, this definition is equivalent to the "robustness" of an equilibrium point, which is a quantity that can be mathematically defined and calculated, so that is typically how I think of it. Of course, writing down the system of equations isn't so easy ... They go on to state that "...[f]or a system to be resilient implies that it maintains certain key properties ..." where a key property is one that is central to its identity. This is much more difficult - what is the identity of an ecosystem? How can you tell if an ecosystem has changed its key properties? The paradigmatic examples involve pretty obvious shifts, like woody plants invading a grassland, algae taking over a coral reef, and the classic clear/turbid lake example. Tough luck if you're managing a woodland park with lots of birds and understory plants.
One of the things that continues to bug me about ER AM is that a series of normative goals (all of which I happen to agree with) are deemed to be necessary because they contribute to resilience, which in turn contributes to those goals. For example

We expect that managing for resilience will sustain diversity, permit natural perturbations, facilitate the action of natural processes and integrate both social and ecological dimensions of sustainability.
But earlier they state "[c]omplex systems theory suggests that the conservation of function is strongly dependent on diversity ...". But this is completely circular - having diversity increases resilience and resilience sustains diversity. So it appears to me that resilience is an attempt to attach some kind of scientific objectivity to the normative goal of maintaining diversity, whether it is diversity of functions or species.
I like the idea of resilience as stated in the first definition, but I think it faces an uphill battle for implementation. This article doesn't advance the cause very much, because it falls into the trap of using resilience as support for a normative goal. Until we can calculate resilience, and predict, credibly, the effects of loss of resilience in a range of systems I don't think we'll have much success convincing the rest of humanity to forego maximizing production.

Adaptive Monitoring?

David Lindenmayer is a ..., well, there's no other way to put this, a god in conservation and ecology of Australia's southeastern forests. I was lucky enough to work with him a bit during my PhD at the University of Adelaide. He and 3 co-authors just published a paper in TREE "Adaptive Monitoring in the real world: proof of concept". This is a follow up piece to an article from a couple years ago defining adaptive monitoring. So what's the big deal? Why isn't this adaptive management, which after all, contains monitoring of responses as a core component in order to reduce uncertainty.
They define Adaptive Monitoring like this:
A monitoring program in which the development of conceptual models, question setting, experimental design, data collection, data analysis, and data interpretation are linked as iterative steps. An adaptive monitoring program is one that can evolve in response to new questions, new information, situations or conditions, or the development of new protocols but this must not distort or breach the integrity of the data record. The adaptive monitoring approach can be applied to all kinds of monitoring including question-driven, passive and mandated monitoring programs.
So models, questions, etc etc all linked into iterative steps, with the ability to evolve in response to new questions, information or conditions. Why isn't it Adaptive Management? Well, it doesn't directly involve decision making about anything other than the monitoring plan itself, at least on the surface, so that rules out Decision Theoretic AM. They claim that experimentation isn't required, which would rule out Experimental-Resilience AM. However, one of the three things they claim needs to be done to make Adaptive Monitoring is to build partnerships with ... wait for it ... policy makers. In other words, it is meant to influence decision making.
What I find interesting is that both of their case studies involve monitoring plots in different "treatments", which makes them suspiciously experimental sounding. In addition, in both case studies there are management decisions being made on the basis of the results of the monitoring, so I would suggest that the difference between AManagement and AMonitoring is a question of badging and marketing, rather than fundamental differences. Both case studies would benefit enormously from a decision theoretic approach to the whole management system, rather than a narrow focus on monitoring.

Friday, February 10, 2012

AM officially adopted for Delaware Bay Horseshoe Crabs

Conor McGowan has pointed to the Delaware Bay Horseshoe Crab fishery as a counter example to my stance that Decision Theoretic AM is less appropriate when social indeterminism is high. The AM plan has just been accepted by the Atlantic States Marine Fisheries Commission. Congratulations are due to everyone involved - it is a great example for an AM plan. I look forward to digesting the details in Addendum VII.

Wednesday, February 1, 2012

Bad modeler, bad.

Roger Pielke Jr. posted this excerpt from a report by Richard Deniss detailing how economic modelling can be abused (pdf):
The problem has become, however, that in an era in which segments of the media no longer have the time or inclination to examine claims before they are reported bad economic modelling is preferred by many advocacy and industry groups to good economic modelling for three main reasons:

1. it is cheaper
2. it is quicker
3. it is far more likely to yield the result preferred by the client

That said, bad economic modelling is relatively easy to identify if readers are willing to ask themselves, and the modeller, a range of simple questions. Indeed, it is even easier to spot when the modeler can't, or won't, answer such simple questions.
I think the same can be said for ecological modelling, except that in most cases, the client is us and our strong normative stance in favor of non-human species.