The Principle Challenge

I came across a really cool article on automating phenology measurements - using remote sensing data and photographs to track development of vegetation at many different scales. This is potentially a really useful way to monitor stuff - quantitative measurements of vegetation that can be reliably and automatically generated over time. However, my favorite part of the entire article was this quote:

For land management, the principle challenge relates to prediction. Managers need to know how today's management decisions will impact tomorrow's ecosystem processes.

Ra! Ra! Sis Boom Bah! Yes! And guess what, that means using models. All the fancy remote sensing in the world is no good unless you can use that data to meet the principle challenge. The trouble is, managers are often reluctant to recognize that models can be helpful. In my recent experience, if models are "known" to have flaws (see quote by George Box), or produce a range of predictions because of statistical error in parameter estimates or inherent variation (demographic stochasticity), then they are labeled useless. Better to use gut instinct to make decisions.

I do believe models are useful even when they are not (and they never will be) perfect.


Jeffery T Morisette et al. (2009) Tracking the rhythm of the seasons in the face of global change:
phenological research in the 21st century. Frontiers in Ecology and the Environment 7:253-260

Zen Buddhism and Adaptive Management

I've been entertaining a visitor this week, Dr. Scott Field, currently a lecturer at the Naval Postgraduate School in Monterey. 5 years ago we worked together on a project analysing fox monitoring data from Eyre Peninusula in South Australia. This week Scott's been conducting the 2nd Quinquennial fox monitoring analysis - and he has results! Nice ones. Moral of the story - analyze your data. It helps. Really.

Scott's recent work has focused on conflict resolution in International Relations, and he had a few extremely interesting comments about Adaptive Management. One thing that causes problems is people rejecting analyses and predictions that could help them with decision making. The difficulty arises if they are not able to understand the underlying methods used to derive those analyses, then they have no way to discuss them. This leads people to react emotionally, rather than critically. Thus, it is critically important to conduct analyses in groups and provide lots of opportunities for feedback and interaction. This still isn't going to solve the problem when you are dealing with sophisticated analyses of complex data. But recognizing that issue will help me to not respond in kind.

The broader notion that we've come up with is that Buddhism has alot to offer the practice of adaptive management - hey, stay with me for a bit. One of Buddhism's "Four Noble Truths" is that Attachment leads to suffering. Attachment can be to things, but also to ideas. So the clear connection to AM is that when someone is attached to an idea about how the world works, then it is hard for them to expose that idea to data and analysis - the risk is that they might have to give up their idea, leading to suffering. Thus finding a good solution to an environmental problem involves giving up attachments to ideas. The trouble is there is no way to force someone else to give up their attachments.

There are other components of Buddhism that have lessons for AM, but I'm just learning about them so won't write any more just now.

Blending math and biology - nothing new there!

One of the things we're pretty proud of here at UNL is our developing collaboration between Biology and Mathematics - we've done joint REU projects, and the crown jewel is the NSF funded Research for Undergraduates in Theoretical Ecology project. We want to do more. Then I saw this quote from R.A. Fisher on the jacket of Alan Hasting's textbook on Population Biology:

I can imagine no more beneficial chance in scientific education than that which would allow [biology and mathematics] to appreciate something of the imaginative grandeur of the realms of thought explored by the other.

Wow. So he was already recognizing this gap in 1930! We've got some catching up to do.

Wrapping around vs. setting up

The 2003 Biological opinion on the Missouri River Mainstem operations required that habitat creation and other "reasonable and prudent alternatives" to avoid jeopardy be conducted using Adaptive Management, because there were significant disagreements about the effectiveness of the RPA elements. However, because of the urgent nature of the problem, the RPA elements, especially habitat creation activities, were started immediately without working out the details of how adaptive management would be used. This is in stark contrast to other adaptive management processes, such as on the Platte River and in the Everglades, where planning and organizing the governance of the adaptive management plans took years - or even decades. This makes the Missouri River an interesting case study in how important those governance bits are.

As a result, the teams that I am working with are trying to take an existing batch of actions, including experiments at various scales, different monitoring programs, and restoration actions, and wrap an AM framework around them. This is exciting on one hand, because stuff is really getting done - including experimental habitat restoration complete with monitoring. On the other hand, it is incredibly difficult, because now there are alot of people with vested interests in projects and programs who are reluctant to embrace change. This is understandable, because a changed program management structure may not include "your" program, or at a minimum may require you to interact with people that you didn't interact with before. Thus a huge part of getting AM off the ground on the Missouri River involves managing people's expectations - and just plain communicating with them. Often.

So far, I'd have to say that I have a much greater appreciation for the governance piece than I did before!

Objectives! and PIGS

I spent the day meeting with USFWS, USACE, and PNNL folks - discussing feedback we received at last weeks presentation of draft Adaptive Management plans to the COoperating for REcovery team. There were quite a few memorable moments of humour - mostly involving new metaphors for the process. Even better is when we can combine a metaphor with an acronym. Our best effort today in that regard was Previously Implemented Goalseeking process, or PIG for short. Much of the stress associated with AM on the Missouri River has to do with the fact that the PIGs have bolted, and we're trying to get them back in the corral. Not get rid of them, just get them all in the same corral.

And as always, it comes back to objectives - and here's a quote to drive the point home:

If you don't know where you are going, any road will get you there.
- Lewis Carroll

Open data, free analysis?

One of the major headaches associated with trying to help organizations develop Adaptive Management plans involves getting all the data relevant to the question. The bigger the organization the harder this is. For some situations, transparent access to the data might be one way to build trust with stakeholders - if they know they can see the data for themselves and have someone else analyze it they might be more willing to accept the analysis that has already been done. Especially once they find out how much it costs to hire a statistician. Maybe I'm dreaming.

For the analyst it creates a different kind of headache, because now you're going to have to defend the myriad choices you made enroute to getting an answer. Because in many cases there are no right answers, only better and worse ones. I should have to defend those choices, but it increases the time required to finish an analysis.

I came across a post about open data publication in the world of genomics - interesting stuff. A world where academic labs of 20 people are "small". Hard to imagine when I'm wrestling with the transition from a lab of 1 (me) to a lab of 4 (me, a postdoc, and two grad students). But the critical point made was about the differences in incentives leading to differences in data publication - submission of raw or analyzed sequences to public databases. Genome centers of 1000 people get funded based on genomes produced - hence they "publish" their data to databases quickly. Academic labs get funding based on paper outputs, so they lag in submitting data to public databases until they have the papers in press - getting scooped sucks. So we could fix that problem by changing the funding model for small labs as well as large, but, and its a big but for me too - how do you get funding to do analysis?

I was frankly delighted to see that someone else also thinks analysis doesn't come for free. In my world, I regularly meet people who have data, and think its relevant to a management problem. But they don't have the expertise to turn that data into something relevant to their problem. Unfortunately it is hard for me to help - I'm only one person, and already completely swamped. The classic natural resources model of "put a student on it" doesn't work well for analysis, because it takes YEARS to develop the necessary skills. Frankly it took me decades. Grant milestones can't wait for that. Ideally, there is some method for a student to start developing the skills, absent pressure to deliver, then when they are ready they can practice by working on a real project. So - who pays for that early development? One solution is teaching assistantships - get them helping undergrads. Well - great, if your department works that way (mine doesn't).

So, what to do in an Adaptive Management world where data is guarded, analysts are scarce, and problems are immediate? The current solution is to make decisions without analysing the data. Publishing data - online, available in raw form - would mean that many additional hands and minds could come to the task of working out what it means, and what the best way to get there is. The USGS publishes stream discharge data in real time. That's not realistic for Tern and Plover fledging success data, but annually - it could be annually. A central database with relevant data would make many things easier for my Missouri River work. Models should also be a part of that database! Open source model for Adaptive Management.

Food for thought, but no answers today.

Making good decisions

The National Academies Press has a new book "Informing Decisions in a Changing Climate". A line from the first page:

Climate change will create a novel and dynamic decision environment. The parameters of the new climate regime cannot be envisioned from past experience. Moreover, climatic changes will be superimposed on social and economic changes that are altering the climate vulnerability of different regions and sectors of society, as well as their ability to cope. Decision makers will need new kinds of information and new ways of thinking and learning to function effectively in a changing climate.

The first two chapters are a goldmine of references on the social science of decision making. What's reassuring is the strong overlap with Department of Interior approaches to adaptive management that I've been pushing. The key learning piece for me is the emphasis here on the process and governance of decision support - something that I personally have started to recognize I'm incompetent at. Lots of good stuff.

The 3rd chapter is on learning - and Adaptive Management is one of the approaches they discuss. I found the discussion illuminating - it references the North American School of adaptive management, but not the more recent work on the Australian school, or the move by the Department of Interior to use AM. It does discuss several possible reasons why "active" adaptive management, the use of managment experiments to reduce uncertainty are difficult. They characterize decision making objectives as "unchanging" in AM, which I disagree with - at least in Australian School AM approaches we accept the ability and need to update objectives periodically.