BioOne sees sustainable scholarly publishing as an inherently collaborative enterprise connecting authors nonprofit publishers academic institutionsresearch libraries and research funders in the common goal of maximizing access to critical research

Is Adaptive Management Helping to Solve Fisheries ProblemsAuthor(s) Carl J WaltersSource AMBIO A Journal of the Human Environment 36(4)304-307Published By Royal Swedish Academy of SciencesDOI httpdxdoiorg1015790044-7447(2007)36[304IAMHTS]20CO2URL httpwwwbiooneorgdoifull1015790044-744728200729365B3043AIAMHTS5D20CO3B2

BioOne (wwwbiooneorg) is a nonprofit online aggregation of core research in the biological ecologicaland environmental sciences BioOne provides a sustainable online platform for over 170 journals and bookspublished by nonprofit societies associations museums institutions and presses

Your use of this PDF the BioOne Web site and all posted and associated content indicates your acceptance ofBioOnersquos Terms of Use available at wwwbiooneorgpageterms_of_use

Usage of BioOne content is strictly limited to personal educational and non-commercial use Commercialinquiries or rights and permissions requests should be directed to the individual publisher as copyright holder

LAND-247

Article Carl J Walters

Is Adaptive Management Helping to SolveFisheries Problems

Adaptive management has been widely recommendedas a way to deal with extreme uncertainty in naturalresource and environmental decision making The coreconcept in adaptive management is that policy choicesshould be treated as deliberate large-scale experimentshence policy choice should be treated at least partly as aproblem of scientific experimental design There havenow been upwards of 100 case studies where attemptswere made to apply adaptive management to issuesranging from restoration of endangered desert fishspecies to protection of the Great Barrier Reef Most ofthese cases have been failures in the sense that noexperimental management program was ever implement-ed and there have been serious problems with monitor-ing programs in the handful of cases where anexperimental plan was implemented Most of the failurescan be traced to three main institutional problems i) lackof management resources for the expanded monitoringneeded to carry out large-scale experiments ii) unwill-ingness by decision makers to admit and embraceuncertainty in making policy choices and iii) lack ofleadership in the form of individuals willing to do all thehard work needed to plan and implement new andcomplex management programs

INTRODUCTION

It has now been three decades since the concept of adaptivemanagement was first proposed as an approach to dealing withextreme uncertainty about the impacts of various policy choicesin renewable resource management (1ndash3) The concept arosefrom frustration in attempts to use computer modeling tointegrate scientific knowledge so as to make useful predictionsfor decision makers In many modeling case studies we keptfinding gross gaps in knowledge about various ecologicalprocesses that the modeling indicated to be important and noindication of progress in dealing with those troublesomeprocesses because they are ones that unfold at space-time scaleswhich are inconvenient or costly for scientists to study (anotorious example is recruitment of new individuals toharvested fish populations a complex process that typicallytakes place over spatial scales of thousands of kilometers andtime scales of years) We concluded from such cases that ifintegrative models cannot be reliably developed to comparepolicy choices then the only way to learn about those choices isthrough direct comparisons of their performance in the field iethrough planned experimental comparisons As this concept ofmanagement as experimentation was further developed we usedoptimization methods from the theory of optimal control tohelp determine when it might be worthwhile to investmanagement resources in potentially risky experiments ratherthan relying upon initial guesswork and subsequent monitoringto uncover good policies (4)

Early case studies taught us to use two main arguments tojustify adaptive management experiments which we calledlsquolsquoprobing for untested opportunityrsquorsquo and lsquolsquocoping with counter-intuitive dynamic responsesrsquorsquo Experimental policy tests are a

way to probe the dynamic responses of a system but moreparticularly such tests are justified only if the experimentalpolicy represents a possible opportunity to improve manage-ment and if historical data are inadequate to show whether thepolicy has already been tried (inadvertently or deliberately)Counterintuitive responses arise when scientists or managersattempt to base predictions on simple common sense argu-ments (like lsquolsquoreducing mortality rate of the fish should causetheir abundance to increasersquorsquo) when in fact the complexity ofecological systems implies that responses may depend onindirect and multiple causal pathways including pathways thatare easily overlooked even when prediction is approached withformal systems modeling techniques

The idea of an adaptive approach to management continuesto have wide intuitive appeal so that it is now routine to seeclaims and even legislative requirements (for example Califor-niarsquos Marine Life Protection Act) that it will be used on casesranging from restoration of endangered species to managementof large marine ecosystems In many cases the claim is simplythat the results of initial policy choices will be monitored so asto identify need for corrective action (so-called lsquolsquopassiversquorsquoadaptive management) but there have also been many caseswhere our original approach of using computer modeling toidentify critical uncertainties and to aid in design of diagnosticmanagement experiments has been followed

Unfortunately the practice of adaptive management hasbeen radically less successful than one would expect from itsintuitive appeal A decade ago I looked back at some 30 casestudies where we had worked with interdisciplinary multi-institution teams to develop adaptive management proposals Icould find evidence of field implementation of experimentalpolicies in only four or five of those cases (5) In a few caseseven the initial modeling step had failed to identify keyuncertainties but for most of the failures there was clearidentification of needed diagnostic management experimentsbut recommendations to do these experiments were simply notfollowed In (5) I suggested a variety of reasons for suchfailures mainly related to problems with institutional incentivesystems Far more elaborate and elegant analyses havesubsequently supported this finding and have suggested avariety of approaches to design of more effective institutions formanagement (6ndash10)

With more experience it is now becoming clear that there arethree main reasons for widespread implementation difficulties inadaptive management programs i) failure of decision makers tounderstand why they are needed ii) lack of leadership for thecomplex process of implementing an adaptive approach andiii) inadequate funding for the increased ecological (and ofteneconomic) monitoring needed to successfully compare theoutcomes of alternative policies This paper discusses each ofthese reasons and suggests what we might do to overcome them

Failure to Comprehend the Need for Management

Experiments

Proposals for management experiments are often greeted bydecision making groups (such as fisheries management councilsand stakeholders with strong political influence) with blank

304 Ambio Vol 36 No 4 June 2007 Royal Swedish Academy of Sciences 2007httpwwwambiokvase

LAND-247

stares of incomprehension or even outright hostility andcomments like lsquolsquowhy canrsquot you scientists make that predictionyou have gathered so much data alreadyrsquorsquo To some degreereactions like these are the long term result of scientists havingroutinely oversold their capabilities to provide useful advice soas to obtain grant funding or to gain credibility and henceauthority In highly structured settings such as fisheries stockassessments and water resource planning where particularmathematical models have been used routinely and parameter-ized with historical data by many scientists and have beensomewhat successful at providing policy advice it has becomereally difficult for practitioners and decision makers alike toadmit that their routine calculations and predictions might behighly suspect But beyond such obvious causes of misunder-standing many people in key decision making positions appearto expect single clear predictions and management prescrip-tions from scientists even when it should be obvious to allconcerned that the scientists should not be offering suchdefinitive advice the demand for single best prescriptiveanswers seems most often rooted in attempts by decisionmakers to lsquolsquopass the buckrsquorsquo to scientists for making difficultjudgments about how to cope with extreme uncertainty

Another serious cause of difficulty with decision makers isthat calls for adaptive management experiments often appear tocontradict conventional wisdom or obvious intuition Anexcellent example of this problem has been in the fisheries ofthe Gulf of Mexico (Walters Martell and Mahmoudi underreview) The Gulf supports several of Americarsquos largest fisheriesand in particular the largest regional shrimp trawl fishery Thatcolorful fishery has been notorious as an example of lsquolsquobadrsquorsquoecosystem management because of its obvious negative impacton marine benthic habitats and its large bycatch of finfish (theweight of fish discarded is typically several times the weight ofshrimp actually landed see [11]) Conservation groups haverepeatedly demanded that it be severely restricted or replacedwith a fishery based on use of more selective fishing methodsFurther support for such restriction has come from data showing

the trawl bycatch of small juveniles of another valued species thered snapper (Lutjanus campechanus) likely exceeds the landedcatch of that species by several fold (11 12) The red snapper hashistorically been quite severely overfished and still faces highfishing mortality rates especially from recreational fisheries thatare very difficult to regulate Trawl bycatch reduction initiallyvia so-called lsquolsquobycatch reduction devicesrsquorsquo (BRDs) has become afavored policy choice since it appears to have three types ofbenefits to restore the snapper stock to demonstrate concern forconservation of benthic habitats and biological diversity and toavoid imposing further restrictions on fishers who targetsnappers

While it is apparently self-evident and certain to many peoplethat BRDs will be beneficial to Gulf of Mexico fishes thescientific evidence based on analysis of historical populationtrends of red snapper suggests a counterintuitive possibilitynamely that there is compensatory mortality in juvenile redsnapper such that higher juvenile densities resulting from BRDprotection will lead to increased natural mortality rates andhence to loss of much or most of the surplus juveniles saved byusing BRDs (12) Recent developments with a complexecosystem model of the Gulf also support this more pessimisticprediction (Walters Martell and Mahmoudi under review) Infact the ecosystem model predicts two quite different possibleoutcomes of using BRDs depending on assumptions made inthe model about how other fish species besides red snapper reactto reductions in bycatch mortality (Fig 1) We get equally goodfits to the historical data with ecosystem model parametervalues that predict no recovery of red snappers as withparameter values that predict dramatic recovery The morepessimistic prediction (lack of recovery of red snapper) resultspartly from compensatory mortality changes and also frompredicted increases in juvenile snapper mortality rate due torecovery of some species that prey upon it like marine catfishesand have also been negatively impacted by trawling in the pastUnfortunately little historical data are available on suchpredators since they are not valued species for harvest andhave thus not been priority species for research

When scenarios like Figure 1 are shown to stakeholders inthe Gulf of Mexico the immediate reaction has been to ask themodelers which prediction is most likely as though we surelymust have a single best hypothesis about what will happenWhen we deny being able to make such a judgment based on thehistorical data and recommend instead the obvious policy ofclosing some experimental areas to trawling and thus obtainingdirect experimental evidence about which scenario is correct thereaction has then been to ask why we do not just get better dataon the species that the ecosystem model predicts might increasein abundance so as to prevent increases in red snapper juvenilesurvival Blank expressions reappear when we explain that thereis simply no way for us to get such data on the other speciessuch data were not collected historically as the shrimp fisherydeveloped and there is no way for us to predict just usingrecent data on abundances and life history characteristics of theother species how they will react to reduced trawling mortality

For decision makers in situations like the Gulf of Mexico tosupport our recommendation for management experiments likeclosed areas they would need to abandon intuitive simplisticarguments (BRDs will save fish) and make a considerableintellectual investment in understanding why we obtaindifferent or ambiguous predictions when we examine causalpathways and mechanisms more closely Virtually all failedadaptive management cases have had this character where inthe end it is just easier for decision makers to scoff at theelaborate analysis and models while trusting simpler intuitivepredictions and to defer difficult decisions about management

Figure 1 Two alternative fits of an ecosystem model to historicaltrends in abundance of an important fish species in the Gulf ofMexico red snapper Historical biomass estimated using stockreduction analysis which involves fitting a population model tohistorical catch and relative abundance trend data The alternativeecosystem model fits to the data are obtained when differentassumptions are made about interactions between juvenile redsnapper and various of its fish competitors and predators likemarine catfish Note how the two different ecosystem model fitspredict similar trends to the historical data (hence cannot bedistinguished on the basis of the data) yet diverge widely inpredicted effect of a simulated shrimp trawl closure introduced insimulation year 1990

Ambio Vol 36 No 4 June 2007 305 Royal Swedish Academy of Sciences 2007httpwwwambiokvase

LAND-247

experiments by instead calling for further research (datacollection modeling)

Lack of Leadership in Implementation

I have been involved in a large number (over 30) of cases wherethe Adaptive Environmental Assessment and Management(AEAM [3]) workshop process appeared to result in generalconsensus among stakeholders scientists and regulatoryagency staff about the need for an experimental managementprogram but where the recommended program was simplynever implemented In the few cases where implementation didoccur there was at least one singular individual (usually amiddle-level staff person from a regulatory agency) who made avery large personal investment of time and energy to make surethat the program actually succeeded

Implementation seldom takes place in an atmosphere ofcrisis where all concerned recognize a need for urgent policychange Instead regulatory agency staff in particular are usuallycomfortable with existing policies and programs and will notvoluntarily make additional work for themselves in terms ofsetting up new regulations and enforcement proceduresdesigning and staffing (funds equipment people) new moni-toring initiatives and organizing the oversight processes(committees administrative procedures) typically required forany new management program in todayrsquos highly bureaucratizedmanagement systems So unless one key individual pushes allthese people through apparently endless and largely wastedtime spent writing and talking to them the implementationprocess will fail at one or more of the necessary implementationsteps (and all it takes is for one step to fail)

In none of the successful cases would the key leader be calledan inspiring or charismatic personality Rather the leaders havebeen people who i) have a broad overview of the decisionmaking and implementation process along with intimateknowledge of all the people and technicaladministrative detailsinvolved in each step in the process ii) are very well organized interms of planning who what where and when specific activitiesand actions are needed iii) simply refuses to take no for ananswer on the many occasions when contributors to the processoffer excuses for inaction and iv) are willing to devote theirwhole career for extended periods of time (typically severalyears) to the implementation process Early in the developmentof the AEAM workshop process C S Holling wrote about thesorts of people who need to be involved in complex ecologicalanalyses and referred to leaders with these attributes as thelsquolsquocompleat emmanuensisrsquorsquo of the policy process (13)

Inadequate Funding for Monitoring Programs

Lack of adequate monitoring data and historical referenceinformation has become a universal complaint in naturalresources management independent of experimental manage-ment initiatives A few of the worst fisheries disasters such asthe collapse of the Newfoundland cod fishery that has beencalled one of the most severe social disasters in Canadianhistory have been attributed more or less directly to misinter-pretation of inadequate monitoring data on trends in stock size(14) So even before we start to talk about enhanced monitoringin support of large-scale management experiments we knowthat we will have to work with monitoring programs and datathat are already inadequate Further in many cases analysis ofcosts for various improvements in monitoring quickly demon-strates that increased investment in monitoring would not bejustified in the sense that monitoring costs could easily come tobe a large part of or even exceed revenues to resource users (Itcan easily cost more to monitor and manage a fishery than thatfishery is worth in the first place)

Management experiments generally involve replicated com-parisons of treatment alternatives (eg closed versus openareas) preferably over spatial experimental units but perhapsalso (or only) over blocks of time Replication or repetition isabsolutely necessary in order to demonstrate what if anydifferences in response are large enough to be repeatedly visibledespite the many other causes of variation in space and timethat are always present in complex open ecological systemsThe requirement for replication obviously has large impacts onpotential monitoring costs especially if it is assumed thattraditional monitoring methods and systems must simply beexpanded to collect the extra information In fact most largeexperimental management proposals in fisheries would beconsidered prohibitively expensive absent some special sourceof outside support since as much money as possible is alreadybeing spent on routine monitoring For example the largestsingle fisheries management experiment that has actually beenimplemented anywhere in the world today aimed at measuringeffects of fishing on the Great Barrier Reef Australia (15 16)would not have gone ahead without funding from sourcesexternal to the regional fisheries management agency

Almost all management agencies now rely mainly upon in-house or contract data collection largely by highly educatedstaff So fisheries scientists with doctorate degrees now routinelygo on long survey cruises where they mainly collect simple data(fish counts measurements) that could just as well (or better) becollected by much less costly technical staff with only modesttraining Further the measurements that we make could in mostcases be equally well done by resource users (fishermen) atradically lower cost because those people are already out in thefield and are skilled at the day-to-day operation of vessels andfishing equipment

Recognizing these points about data collection proposals forimproved overall monitoring and large-scale management exper-iments have included a strong recommendation to developcollaborative monitoring programs where resource users collectmost of the monitoring data at relatively low marginal cost (1718) Predictably some traditional scientists haveobjected stronglyto such proposals citing difficulties ranging from costs of qualitycontrol to examples of outright cheating (stakeholders recordinginformation that they think will be favorable to them andwithholding information such as tag recoveries that they thinkmay result in restriction of their activities) Fortunately newtechnologiespromise todeal effectivelywithmost suchobjectionsFor example fishermen can be provided hand-held computerswith GPS and mobile phone capability electronic chartingsoftware linked to the GPS and data entryerror correctionsoftware (essentially an electronic logbook system) Data enteredthrough such systems can be accurately geo-referenced and sentimmediately to central datamanagement systems for further errorevaluation and use in spatial data analysis

DISCUSSION

Of the three main causes of implementation failure describedabove easily the most important has been lack of leadership tocarry out the complicated administrative steps involved inmoving a new management vision into actual field practiceThere have been various suggestions about how to overcomethis problem mainly involved with techniques for promotingcollaborative work among stakeholders (19) Unfortunately allthe organizational theory in the world will not overcome theneed for that one singular individual who must make anextraordinary personal commitment to the organizationalprocess The single most important thing that proponents ofadaptive management must learn to do in the future is todevelop skill at identifying and nurturing such people Most of

306 Ambio Vol 36 No 4 June 2007 Royal Swedish Academy of Sciences 2007httpwwwambiokvase

LAND-247

the intellectual skills needed to originate an adaptive manage-ment proposal (critical scientific attitude knowledge ofmodeling and statistical design monitoring methods experience)are just not the right ones for carrying out the hard oftenboring and generally frustrating work of implementation

There are great opportunities for ecological scientists todayto participate creatively in the development of innovativemonitoring programs that utilize expertise and experience ofresource users (especially fishers) to collect data more cheaplyThe electronic logbook and data entry systems mentioned aboveare just one relatively simple example of the technologies thatcan now be developed for capture of ecological informationAnother major research opportunity is in the development offish tags and tag detectors that can be used by nonscientists tohelp collect better information on distribution movements andexploitation rates Still another is in the use of advancedacoustic and video techniques along with technologies likeremotely operated vehicles (ROVs) to carry out large-scalesurveys of fish and fish habitat

There remains the fundamental problem of how to convincedecision makers to support management experiments I believethat the key to this problem is for the scientific community toagree to stop providing simplistic point estimates and manage-ment advice That is when we are asked for our best estimatesand predictions we should agree to offer only a strategic rangeof possible outcomes for each policy choice (like the extremepredictions in Figure 1) and we should simply refuse to saywhich outcome we think is most likely This tactic would placethe burden of risk management squarely on the shoulders of thedecision makers essentially forcing them to confront the realuncertainty and think carefully about ways to deal with it Inrecent years the fisheries stock assessment literature has beenflooded with papers about how to calculate Bayesian probabil-ity distributions for key management quantities (20) Unfortu-nately while such analyses certainly look impressive from amathematical-statistics perspective they are generally based onhopelessly optimistic assumptions about statistical variabilityand knowledge of structural relationships and hence grosslyunderestimate how much uncertainty ought to be admitted (2122) Further the results of these analyses are typically presentedas probability distributions which are difficult to interpret andare accompanied not by recommendations to do managementexperiments but rather with admonishments about the need forcautious (so-called lsquolsquoprecautionaryrsquorsquo) management a risk aversemanagement judgment that scientists ought not to be making inthe first place

The history of fisheries management has largely involvedmyopic focus on single fish stocks and the direct impacts offishing on them Under that focus it has been permissible tobase policy on relatively simple dynamic models and to ignoremore complex interaction effects like described above for theGulf of Mexico shrimp trawl fishery Adaptive management hasbeen of little help in dealing with single stock management issueBut the Gulf of Mexico example is not an isolated or unusualone it is routine to see counterintuitive predictions (andinability to test these with fragmentary historical data) whenwe attempt to model trophic and habitat interactions in supportof ecosystem management questions such as how fisheries atone trophic level affect productivity of fisheries at other levelsand whether natural predators should be culled in order tomake fisheries more productive (23 24) Management agenciestoday are under very strong pressure to adopt ecosystem-basedmanagement approaches (25 26) With that pressure has comedemands for ecosystem modeling and those models will furtherexpose just how deep our uncertainty is about the efficacy ofeven apparently simple regulatory measures like makingfisheries more selective to reduce bycatch If we are honest

about admitting and embracing that uncertainty in developingpolicy advice adaptive management experiments will finallycome to play a central role in the management of fisheries andtheir ecosystems We will simply be forced to find ways toovercome the problems of selling experimental policies todecision makers leadership in implementation and high costsof monitoring

References and Notes

1 Walters C and Hilborn R 1976 Adaptive control of fishing systems J Fish ResBoard Canada 33 145ndash159

2 Walters CJ and Hilborn R 1978 Ecological optimization and adaptive managementAnn Rev Ecol Syst 9 157ndash188

3 Holling CS (eds) 1978 Adaptive Environmental Assessment and Management JohnWiley New York 377 pp

4 Walters C 1986 Adaptive Management of Renewable Resources McMillan PublisherCo New York 374 pp

5 Walters C 1997 Challenges in adaptive management of riparian and coastal ecosystemsConservation Ecology 1 (httpwwwconsecolorgvol1iss2art1)

6 Gunderson L Holling CS and Light SS 1995 Barriers and Bridges to the Renewal ofEcosystems and Institutions Columbia University Press New York 593 pp

7 Gunderson L and Holling CS (eds) 2002 Panarchy Understanding Transformationsin Human and Natural Systems Island Press Washington DC 450 pp

8 Parma A and NCEAS Working Group on Population Management 1998 What canadaptive management do for our fish forests food and biodiversity Integr Biol 1 16ndash26

9 Allan C and Curtis A 2005 Nipped in the bud why regional scale adaptivemanagement is not blooming Environ Management 36 414ndash425

10 Schreibner ESG Bearlin AR Nicol SJ and Todd CR 2004 Adaptivemanagement a synthesis of current understanding and effective application EcolManagement Rest 5 177ndash182

11 Ortiz M Legault CM and Erhardt NM 2000 An alternative method for estimatingbycatch from the US shrimp trawl fishery in the Gulf of Mexico Fish Bull 98 583ndash599

12 SEDAR 7 2005 Assessment summary report Gulf of Mexico red snapper (httpwwwsefscnoaagovsedar)

13 Holling CS and Chambers AD 1973 Resource science the nurture of an infantBioScience 23 13ndash20

14 Walters CJ and McGuire JJ 1996 Lessons for stock assessment from the northerncod collapse Rev Fish Biol Fish 6 125ndash137

15 Mapstone BD Davies CR Little LR Punt AE Smith ADM Pantus F LouDC Williams AJ et al 2004 The Effects of Line Fishing on the Great Barrier Reefand Evaluations of Alternative Potential Management Strategies CRC Reef Res CentreTechnical Report 52 pages indashxi and 1ndash205

16 Campbell RA Mapstone BD and Smith ADM 2001 Evaluating large scaleexperimental designs for management of coral trout on the Great Barrier Reef EcolAppl 11 1763ndash1777

17 Walters CJ and Collie JS 1989 An experimental management strategy forgroundfish in the face of large uncertainty about stock size and production Can JFish Aquat Sci 108 13ndash25

18 Hilborn R Dealteris J Deriso R Graham G Iudicello S Lundsten M PikitchEK Sylvia G et al 2004 Cooperative Research in the National Marine FisheriesService The National Academies Press Washington DC 132 pp

19 See reviews and advice offered on the Adaptive Management Practitioners Networkwebsite (httpwwwadaptivemanagementnetwhatisphp)

20 Patterson K Cook R Darby C Gavaris S Kell L Lewy P Mesnil B Punt Aet al 2001 Estimating uncertainty in fish stock assessment and forecasting Fish Fish 2125ndash157

21 McAllister MK and Kirchner C 2002 Accounting for structural uncertainty tofacilitate precautionary fishery management illustration with Namibian orange roughyBull Mar Sci 70 499ndash540

22 Clark CW 2006 The Worldwide Crisis in Fisheries Economic Models and HumanBehavior Cambridge Univ Press New York 264 pp

23 Yodzis P 2001 Must top predators be culled for the sake of fisheries Trends Ecol Evol16 78ndash84

24 Bakun A 2006 Wasp-waist populations and marine ecosystem dynamics navigatingthe lsquolsquopredator pitrsquorsquo topographies ProgOceano 68 271ndash288

25 Pikitch EK Santora C Babcock EA Bakun A Bonfil R Conover DODayton P Doukakis P et al 2004 Ecosystem-based fishery management Science305 346ndash347

26 Sainsbury KJ Campbell RA and Whitelaw WW 1993 Effects of trawling on themarine habitation the North West Shelf of Australia and implications for sustainablefisheries management In Sustainable Fisheries through Sustainable Habitat HancockDA (ed) Bureau of Rural Sciences Proceedings AGPS Canberra pp 137ndash145

27 First submitted 12 December 2006 Accepted for publication 12 December 2006

Dr Carl Walters is currently Professor of Zoology andFisheries at the University of British Columbia VancouverCanada Dr Walters is a Fellow of the Royal Society ofCanada (1998) and a Pew Fellow in Marine Conservation(2001) He was also the 2001-2002 Mote Eminent Scholar atFlorida State University and the Mote Marine Laboratory In2006 he received the Volvo Environment Prize and theAmerican Fisheries Society Award of Excellence His addressFisheries Centre University of British Columbia VancouverBC V6T 1Z4 CanadaE-mail cwaltersfisheriesubcca

Ambio Vol 36 No 4 June 2007 307 Royal Swedish Academy of Sciences 2007httpwwwambiokvase

LAND-247

stares of incomprehension or even outright hostility andcomments like lsquolsquowhy canrsquot you scientists make that predictionyou have gathered so much data alreadyrsquorsquo To some degreereactions like these are the long term result of scientists havingroutinely oversold their capabilities to provide useful advice soas to obtain grant funding or to gain credibility and henceauthority In highly structured settings such as fisheries stockassessments and water resource planning where particularmathematical models have been used routinely and parameter-ized with historical data by many scientists and have beensomewhat successful at providing policy advice it has becomereally difficult for practitioners and decision makers alike toadmit that their routine calculations and predictions might behighly suspect But beyond such obvious causes of misunder-standing many people in key decision making positions appearto expect single clear predictions and management prescrip-tions from scientists even when it should be obvious to allconcerned that the scientists should not be offering suchdefinitive advice the demand for single best prescriptiveanswers seems most often rooted in attempts by decisionmakers to lsquolsquopass the buckrsquorsquo to scientists for making difficultjudgments about how to cope with extreme uncertainty

Another serious cause of difficulty with decision makers isthat calls for adaptive management experiments often appear tocontradict conventional wisdom or obvious intuition Anexcellent example of this problem has been in the fisheries ofthe Gulf of Mexico (Walters Martell and Mahmoudi underreview) The Gulf supports several of Americarsquos largest fisheriesand in particular the largest regional shrimp trawl fishery Thatcolorful fishery has been notorious as an example of lsquolsquobadrsquorsquoecosystem management because of its obvious negative impacton marine benthic habitats and its large bycatch of finfish (theweight of fish discarded is typically several times the weight ofshrimp actually landed see [11]) Conservation groups haverepeatedly demanded that it be severely restricted or replacedwith a fishery based on use of more selective fishing methodsFurther support for such restriction has come from data showing

the trawl bycatch of small juveniles of another valued species thered snapper (Lutjanus campechanus) likely exceeds the landedcatch of that species by several fold (11 12) The red snapper hashistorically been quite severely overfished and still faces highfishing mortality rates especially from recreational fisheries thatare very difficult to regulate Trawl bycatch reduction initiallyvia so-called lsquolsquobycatch reduction devicesrsquorsquo (BRDs) has become afavored policy choice since it appears to have three types ofbenefits to restore the snapper stock to demonstrate concern forconservation of benthic habitats and biological diversity and toavoid imposing further restrictions on fishers who targetsnappers

While it is apparently self-evident and certain to many peoplethat BRDs will be beneficial to Gulf of Mexico fishes thescientific evidence based on analysis of historical populationtrends of red snapper suggests a counterintuitive possibilitynamely that there is compensatory mortality in juvenile redsnapper such that higher juvenile densities resulting from BRDprotection will lead to increased natural mortality rates andhence to loss of much or most of the surplus juveniles saved byusing BRDs (12) Recent developments with a complexecosystem model of the Gulf also support this more pessimisticprediction (Walters Martell and Mahmoudi under review) Infact the ecosystem model predicts two quite different possibleoutcomes of using BRDs depending on assumptions made inthe model about how other fish species besides red snapper reactto reductions in bycatch mortality (Fig 1) We get equally goodfits to the historical data with ecosystem model parametervalues that predict no recovery of red snappers as withparameter values that predict dramatic recovery The morepessimistic prediction (lack of recovery of red snapper) resultspartly from compensatory mortality changes and also frompredicted increases in juvenile snapper mortality rate due torecovery of some species that prey upon it like marine catfishesand have also been negatively impacted by trawling in the pastUnfortunately little historical data are available on suchpredators since they are not valued species for harvest andhave thus not been priority species for research

When scenarios like Figure 1 are shown to stakeholders inthe Gulf of Mexico the immediate reaction has been to ask themodelers which prediction is most likely as though we surelymust have a single best hypothesis about what will happenWhen we deny being able to make such a judgment based on thehistorical data and recommend instead the obvious policy ofclosing some experimental areas to trawling and thus obtainingdirect experimental evidence about which scenario is correct thereaction has then been to ask why we do not just get better dataon the species that the ecosystem model predicts might increasein abundance so as to prevent increases in red snapper juvenilesurvival Blank expressions reappear when we explain that thereis simply no way for us to get such data on the other speciessuch data were not collected historically as the shrimp fisherydeveloped and there is no way for us to predict just usingrecent data on abundances and life history characteristics of theother species how they will react to reduced trawling mortality

For decision makers in situations like the Gulf of Mexico tosupport our recommendation for management experiments likeclosed areas they would need to abandon intuitive simplisticarguments (BRDs will save fish) and make a considerableintellectual investment in understanding why we obtaindifferent or ambiguous predictions when we examine causalpathways and mechanisms more closely Virtually all failedadaptive management cases have had this character where inthe end it is just easier for decision makers to scoff at theelaborate analysis and models while trusting simpler intuitivepredictions and to defer difficult decisions about management

Figure 1 Two alternative fits of an ecosystem model to historicaltrends in abundance of an important fish species in the Gulf ofMexico red snapper Historical biomass estimated using stockreduction analysis which involves fitting a population model tohistorical catch and relative abundance trend data The alternativeecosystem model fits to the data are obtained when differentassumptions are made about interactions between juvenile redsnapper and various of its fish competitors and predators likemarine catfish Note how the two different ecosystem model fitspredict similar trends to the historical data (hence cannot bedistinguished on the basis of the data) yet diverge widely inpredicted effect of a simulated shrimp trawl closure introduced insimulation year 1990

Ambio Vol 36 No 4 June 2007 305 Royal Swedish Academy of Sciences 2007httpwwwambiokvase

LAND-247

experiments by instead calling for further research (datacollection modeling)

Lack of Leadership in Implementation

I have been involved in a large number (over 30) of cases wherethe Adaptive Environmental Assessment and Management(AEAM [3]) workshop process appeared to result in generalconsensus among stakeholders scientists and regulatoryagency staff about the need for an experimental managementprogram but where the recommended program was simplynever implemented In the few cases where implementation didoccur there was at least one singular individual (usually amiddle-level staff person from a regulatory agency) who made avery large personal investment of time and energy to make surethat the program actually succeeded

Implementation seldom takes place in an atmosphere ofcrisis where all concerned recognize a need for urgent policychange Instead regulatory agency staff in particular are usuallycomfortable with existing policies and programs and will notvoluntarily make additional work for themselves in terms ofsetting up new regulations and enforcement proceduresdesigning and staffing (funds equipment people) new moni-toring initiatives and organizing the oversight processes(committees administrative procedures) typically required forany new management program in todayrsquos highly bureaucratizedmanagement systems So unless one key individual pushes allthese people through apparently endless and largely wastedtime spent writing and talking to them the implementationprocess will fail at one or more of the necessary implementationsteps (and all it takes is for one step to fail)

In none of the successful cases would the key leader be calledan inspiring or charismatic personality Rather the leaders havebeen people who i) have a broad overview of the decisionmaking and implementation process along with intimateknowledge of all the people and technicaladministrative detailsinvolved in each step in the process ii) are very well organized interms of planning who what where and when specific activitiesand actions are needed iii) simply refuses to take no for ananswer on the many occasions when contributors to the processoffer excuses for inaction and iv) are willing to devote theirwhole career for extended periods of time (typically severalyears) to the implementation process Early in the developmentof the AEAM workshop process C S Holling wrote about thesorts of people who need to be involved in complex ecologicalanalyses and referred to leaders with these attributes as thelsquolsquocompleat emmanuensisrsquorsquo of the policy process (13)

Inadequate Funding for Monitoring Programs

Lack of adequate monitoring data and historical referenceinformation has become a universal complaint in naturalresources management independent of experimental manage-ment initiatives A few of the worst fisheries disasters such asthe collapse of the Newfoundland cod fishery that has beencalled one of the most severe social disasters in Canadianhistory have been attributed more or less directly to misinter-pretation of inadequate monitoring data on trends in stock size(14) So even before we start to talk about enhanced monitoringin support of large-scale management experiments we knowthat we will have to work with monitoring programs and datathat are already inadequate Further in many cases analysis ofcosts for various improvements in monitoring quickly demon-strates that increased investment in monitoring would not bejustified in the sense that monitoring costs could easily come tobe a large part of or even exceed revenues to resource users (Itcan easily cost more to monitor and manage a fishery than thatfishery is worth in the first place)

Management experiments generally involve replicated com-parisons of treatment alternatives (eg closed versus openareas) preferably over spatial experimental units but perhapsalso (or only) over blocks of time Replication or repetition isabsolutely necessary in order to demonstrate what if anydifferences in response are large enough to be repeatedly visibledespite the many other causes of variation in space and timethat are always present in complex open ecological systemsThe requirement for replication obviously has large impacts onpotential monitoring costs especially if it is assumed thattraditional monitoring methods and systems must simply beexpanded to collect the extra information In fact most largeexperimental management proposals in fisheries would beconsidered prohibitively expensive absent some special sourceof outside support since as much money as possible is alreadybeing spent on routine monitoring For example the largestsingle fisheries management experiment that has actually beenimplemented anywhere in the world today aimed at measuringeffects of fishing on the Great Barrier Reef Australia (15 16)would not have gone ahead without funding from sourcesexternal to the regional fisheries management agency

Almost all management agencies now rely mainly upon in-house or contract data collection largely by highly educatedstaff So fisheries scientists with doctorate degrees now routinelygo on long survey cruises where they mainly collect simple data(fish counts measurements) that could just as well (or better) becollected by much less costly technical staff with only modesttraining Further the measurements that we make could in mostcases be equally well done by resource users (fishermen) atradically lower cost because those people are already out in thefield and are skilled at the day-to-day operation of vessels andfishing equipment

Recognizing these points about data collection proposals forimproved overall monitoring and large-scale management exper-iments have included a strong recommendation to developcollaborative monitoring programs where resource users collectmost of the monitoring data at relatively low marginal cost (1718) Predictably some traditional scientists haveobjected stronglyto such proposals citing difficulties ranging from costs of qualitycontrol to examples of outright cheating (stakeholders recordinginformation that they think will be favorable to them andwithholding information such as tag recoveries that they thinkmay result in restriction of their activities) Fortunately newtechnologiespromise todeal effectivelywithmost suchobjectionsFor example fishermen can be provided hand-held computerswith GPS and mobile phone capability electronic chartingsoftware linked to the GPS and data entryerror correctionsoftware (essentially an electronic logbook system) Data enteredthrough such systems can be accurately geo-referenced and sentimmediately to central datamanagement systems for further errorevaluation and use in spatial data analysis

DISCUSSION

Of the three main causes of implementation failure describedabove easily the most important has been lack of leadership tocarry out the complicated administrative steps involved inmoving a new management vision into actual field practiceThere have been various suggestions about how to overcomethis problem mainly involved with techniques for promotingcollaborative work among stakeholders (19) Unfortunately allthe organizational theory in the world will not overcome theneed for that one singular individual who must make anextraordinary personal commitment to the organizationalprocess The single most important thing that proponents ofadaptive management must learn to do in the future is todevelop skill at identifying and nurturing such people Most of

306 Ambio Vol 36 No 4 June 2007 Royal Swedish Academy of Sciences 2007httpwwwambiokvase

LAND-247

the intellectual skills needed to originate an adaptive manage-ment proposal (critical scientific attitude knowledge ofmodeling and statistical design monitoring methods experience)are just not the right ones for carrying out the hard oftenboring and generally frustrating work of implementation

There are great opportunities for ecological scientists todayto participate creatively in the development of innovativemonitoring programs that utilize expertise and experience ofresource users (especially fishers) to collect data more cheaplyThe electronic logbook and data entry systems mentioned aboveare just one relatively simple example of the technologies thatcan now be developed for capture of ecological informationAnother major research opportunity is in the development offish tags and tag detectors that can be used by nonscientists tohelp collect better information on distribution movements andexploitation rates Still another is in the use of advancedacoustic and video techniques along with technologies likeremotely operated vehicles (ROVs) to carry out large-scalesurveys of fish and fish habitat

There remains the fundamental problem of how to convincedecision makers to support management experiments I believethat the key to this problem is for the scientific community toagree to stop providing simplistic point estimates and manage-ment advice That is when we are asked for our best estimatesand predictions we should agree to offer only a strategic rangeof possible outcomes for each policy choice (like the extremepredictions in Figure 1) and we should simply refuse to saywhich outcome we think is most likely This tactic would placethe burden of risk management squarely on the shoulders of thedecision makers essentially forcing them to confront the realuncertainty and think carefully about ways to deal with it Inrecent years the fisheries stock assessment literature has beenflooded with papers about how to calculate Bayesian probabil-ity distributions for key management quantities (20) Unfortu-nately while such analyses certainly look impressive from amathematical-statistics perspective they are generally based onhopelessly optimistic assumptions about statistical variabilityand knowledge of structural relationships and hence grosslyunderestimate how much uncertainty ought to be admitted (2122) Further the results of these analyses are typically presentedas probability distributions which are difficult to interpret andare accompanied not by recommendations to do managementexperiments but rather with admonishments about the need forcautious (so-called lsquolsquoprecautionaryrsquorsquo) management a risk aversemanagement judgment that scientists ought not to be making inthe first place

The history of fisheries management has largely involvedmyopic focus on single fish stocks and the direct impacts offishing on them Under that focus it has been permissible tobase policy on relatively simple dynamic models and to ignoremore complex interaction effects like described above for theGulf of Mexico shrimp trawl fishery Adaptive management hasbeen of little help in dealing with single stock management issueBut the Gulf of Mexico example is not an isolated or unusualone it is routine to see counterintuitive predictions (andinability to test these with fragmentary historical data) whenwe attempt to model trophic and habitat interactions in supportof ecosystem management questions such as how fisheries atone trophic level affect productivity of fisheries at other levelsand whether natural predators should be culled in order tomake fisheries more productive (23 24) Management agenciestoday are under very strong pressure to adopt ecosystem-basedmanagement approaches (25 26) With that pressure has comedemands for ecosystem modeling and those models will furtherexpose just how deep our uncertainty is about the efficacy ofeven apparently simple regulatory measures like makingfisheries more selective to reduce bycatch If we are honest

about admitting and embracing that uncertainty in developingpolicy advice adaptive management experiments will finallycome to play a central role in the management of fisheries andtheir ecosystems We will simply be forced to find ways toovercome the problems of selling experimental policies todecision makers leadership in implementation and high costsof monitoring

References and Notes

1 Walters C and Hilborn R 1976 Adaptive control of fishing systems J Fish ResBoard Canada 33 145ndash159

2 Walters CJ and Hilborn R 1978 Ecological optimization and adaptive managementAnn Rev Ecol Syst 9 157ndash188

3 Holling CS (eds) 1978 Adaptive Environmental Assessment and Management JohnWiley New York 377 pp

4 Walters C 1986 Adaptive Management of Renewable Resources McMillan PublisherCo New York 374 pp

5 Walters C 1997 Challenges in adaptive management of riparian and coastal ecosystemsConservation Ecology 1 (httpwwwconsecolorgvol1iss2art1)

6 Gunderson L Holling CS and Light SS 1995 Barriers and Bridges to the Renewal ofEcosystems and Institutions Columbia University Press New York 593 pp

7 Gunderson L and Holling CS (eds) 2002 Panarchy Understanding Transformationsin Human and Natural Systems Island Press Washington DC 450 pp

8 Parma A and NCEAS Working Group on Population Management 1998 What canadaptive management do for our fish forests food and biodiversity Integr Biol 1 16ndash26

9 Allan C and Curtis A 2005 Nipped in the bud why regional scale adaptivemanagement is not blooming Environ Management 36 414ndash425

10 Schreibner ESG Bearlin AR Nicol SJ and Todd CR 2004 Adaptivemanagement a synthesis of current understanding and effective application EcolManagement Rest 5 177ndash182

11 Ortiz M Legault CM and Erhardt NM 2000 An alternative method for estimatingbycatch from the US shrimp trawl fishery in the Gulf of Mexico Fish Bull 98 583ndash599

12 SEDAR 7 2005 Assessment summary report Gulf of Mexico red snapper (httpwwwsefscnoaagovsedar)

13 Holling CS and Chambers AD 1973 Resource science the nurture of an infantBioScience 23 13ndash20

14 Walters CJ and McGuire JJ 1996 Lessons for stock assessment from the northerncod collapse Rev Fish Biol Fish 6 125ndash137

15 Mapstone BD Davies CR Little LR Punt AE Smith ADM Pantus F LouDC Williams AJ et al 2004 The Effects of Line Fishing on the Great Barrier Reefand Evaluations of Alternative Potential Management Strategies CRC Reef Res CentreTechnical Report 52 pages indashxi and 1ndash205

16 Campbell RA Mapstone BD and Smith ADM 2001 Evaluating large scaleexperimental designs for management of coral trout on the Great Barrier Reef EcolAppl 11 1763ndash1777

17 Walters CJ and Collie JS 1989 An experimental management strategy forgroundfish in the face of large uncertainty about stock size and production Can JFish Aquat Sci 108 13ndash25

18 Hilborn R Dealteris J Deriso R Graham G Iudicello S Lundsten M PikitchEK Sylvia G et al 2004 Cooperative Research in the National Marine FisheriesService The National Academies Press Washington DC 132 pp

19 See reviews and advice offered on the Adaptive Management Practitioners Networkwebsite (httpwwwadaptivemanagementnetwhatisphp)

20 Patterson K Cook R Darby C Gavaris S Kell L Lewy P Mesnil B Punt Aet al 2001 Estimating uncertainty in fish stock assessment and forecasting Fish Fish 2125ndash157

21 McAllister MK and Kirchner C 2002 Accounting for structural uncertainty tofacilitate precautionary fishery management illustration with Namibian orange roughyBull Mar Sci 70 499ndash540

22 Clark CW 2006 The Worldwide Crisis in Fisheries Economic Models and HumanBehavior Cambridge Univ Press New York 264 pp

23 Yodzis P 2001 Must top predators be culled for the sake of fisheries Trends Ecol Evol16 78ndash84

24 Bakun A 2006 Wasp-waist populations and marine ecosystem dynamics navigatingthe lsquolsquopredator pitrsquorsquo topographies ProgOceano 68 271ndash288

25 Pikitch EK Santora C Babcock EA Bakun A Bonfil R Conover DODayton P Doukakis P et al 2004 Ecosystem-based fishery management Science305 346ndash347

26 Sainsbury KJ Campbell RA and Whitelaw WW 1993 Effects of trawling on themarine habitation the North West Shelf of Australia and implications for sustainablefisheries management In Sustainable Fisheries through Sustainable Habitat HancockDA (ed) Bureau of Rural Sciences Proceedings AGPS Canberra pp 137ndash145

27 First submitted 12 December 2006 Accepted for publication 12 December 2006

Dr Carl Walters is currently Professor of Zoology andFisheries at the University of British Columbia VancouverCanada Dr Walters is a Fellow of the Royal Society ofCanada (1998) and a Pew Fellow in Marine Conservation(2001) He was also the 2001-2002 Mote Eminent Scholar atFlorida State University and the Mote Marine Laboratory In2006 he received the Volvo Environment Prize and theAmerican Fisheries Society Award of Excellence His addressFisheries Centre University of British Columbia VancouverBC V6T 1Z4 CanadaE-mail cwaltersfisheriesubcca

Ambio Vol 36 No 4 June 2007 307 Royal Swedish Academy of Sciences 2007httpwwwambiokvase

LAND-247

experiments by instead calling for further research (datacollection modeling)

Lack of Leadership in Implementation

I have been involved in a large number (over 30) of cases wherethe Adaptive Environmental Assessment and Management(AEAM [3]) workshop process appeared to result in generalconsensus among stakeholders scientists and regulatoryagency staff about the need for an experimental managementprogram but where the recommended program was simplynever implemented In the few cases where implementation didoccur there was at least one singular individual (usually amiddle-level staff person from a regulatory agency) who made avery large personal investment of time and energy to make surethat the program actually succeeded

Implementation seldom takes place in an atmosphere ofcrisis where all concerned recognize a need for urgent policychange Instead regulatory agency staff in particular are usuallycomfortable with existing policies and programs and will notvoluntarily make additional work for themselves in terms ofsetting up new regulations and enforcement proceduresdesigning and staffing (funds equipment people) new moni-toring initiatives and organizing the oversight processes(committees administrative procedures) typically required forany new management program in todayrsquos highly bureaucratizedmanagement systems So unless one key individual pushes allthese people through apparently endless and largely wastedtime spent writing and talking to them the implementationprocess will fail at one or more of the necessary implementationsteps (and all it takes is for one step to fail)

In none of the successful cases would the key leader be calledan inspiring or charismatic personality Rather the leaders havebeen people who i) have a broad overview of the decisionmaking and implementation process along with intimateknowledge of all the people and technicaladministrative detailsinvolved in each step in the process ii) are very well organized interms of planning who what where and when specific activitiesand actions are needed iii) simply refuses to take no for ananswer on the many occasions when contributors to the processoffer excuses for inaction and iv) are willing to devote theirwhole career for extended periods of time (typically severalyears) to the implementation process Early in the developmentof the AEAM workshop process C S Holling wrote about thesorts of people who need to be involved in complex ecologicalanalyses and referred to leaders with these attributes as thelsquolsquocompleat emmanuensisrsquorsquo of the policy process (13)

Inadequate Funding for Monitoring Programs

Lack of adequate monitoring data and historical referenceinformation has become a universal complaint in naturalresources management independent of experimental manage-ment initiatives A few of the worst fisheries disasters such asthe collapse of the Newfoundland cod fishery that has beencalled one of the most severe social disasters in Canadianhistory have been attributed more or less directly to misinter-pretation of inadequate monitoring data on trends in stock size(14) So even before we start to talk about enhanced monitoringin support of large-scale management experiments we knowthat we will have to work with monitoring programs and datathat are already inadequate Further in many cases analysis ofcosts for various improvements in monitoring quickly demon-strates that increased investment in monitoring would not bejustified in the sense that monitoring costs could easily come tobe a large part of or even exceed revenues to resource users (Itcan easily cost more to monitor and manage a fishery than thatfishery is worth in the first place)

Management experiments generally involve replicated com-parisons of treatment alternatives (eg closed versus openareas) preferably over spatial experimental units but perhapsalso (or only) over blocks of time Replication or repetition isabsolutely necessary in order to demonstrate what if anydifferences in response are large enough to be repeatedly visibledespite the many other causes of variation in space and timethat are always present in complex open ecological systemsThe requirement for replication obviously has large impacts onpotential monitoring costs especially if it is assumed thattraditional monitoring methods and systems must simply beexpanded to collect the extra information In fact most largeexperimental management proposals in fisheries would beconsidered prohibitively expensive absent some special sourceof outside support since as much money as possible is alreadybeing spent on routine monitoring For example the largestsingle fisheries management experiment that has actually beenimplemented anywhere in the world today aimed at measuringeffects of fishing on the Great Barrier Reef Australia (15 16)would not have gone ahead without funding from sourcesexternal to the regional fisheries management agency

Almost all management agencies now rely mainly upon in-house or contract data collection largely by highly educatedstaff So fisheries scientists with doctorate degrees now routinelygo on long survey cruises where they mainly collect simple data(fish counts measurements) that could just as well (or better) becollected by much less costly technical staff with only modesttraining Further the measurements that we make could in mostcases be equally well done by resource users (fishermen) atradically lower cost because those people are already out in thefield and are skilled at the day-to-day operation of vessels andfishing equipment

Recognizing these points about data collection proposals forimproved overall monitoring and large-scale management exper-iments have included a strong recommendation to developcollaborative monitoring programs where resource users collectmost of the monitoring data at relatively low marginal cost (1718) Predictably some traditional scientists haveobjected stronglyto such proposals citing difficulties ranging from costs of qualitycontrol to examples of outright cheating (stakeholders recordinginformation that they think will be favorable to them andwithholding information such as tag recoveries that they thinkmay result in restriction of their activities) Fortunately newtechnologiespromise todeal effectivelywithmost suchobjectionsFor example fishermen can be provided hand-held computerswith GPS and mobile phone capability electronic chartingsoftware linked to the GPS and data entryerror correctionsoftware (essentially an electronic logbook system) Data enteredthrough such systems can be accurately geo-referenced and sentimmediately to central datamanagement systems for further errorevaluation and use in spatial data analysis

DISCUSSION

Of the three main causes of implementation failure describedabove easily the most important has been lack of leadership tocarry out the complicated administrative steps involved inmoving a new management vision into actual field practiceThere have been various suggestions about how to overcomethis problem mainly involved with techniques for promotingcollaborative work among stakeholders (19) Unfortunately allthe organizational theory in the world will not overcome theneed for that one singular individual who must make anextraordinary personal commitment to the organizationalprocess The single most important thing that proponents ofadaptive management must learn to do in the future is todevelop skill at identifying and nurturing such people Most of

306 Ambio Vol 36 No 4 June 2007 Royal Swedish Academy of Sciences 2007httpwwwambiokvase

LAND-247

the intellectual skills needed to originate an adaptive manage-ment proposal (critical scientific attitude knowledge ofmodeling and statistical design monitoring methods experience)are just not the right ones for carrying out the hard oftenboring and generally frustrating work of implementation

There are great opportunities for ecological scientists todayto participate creatively in the development of innovativemonitoring programs that utilize expertise and experience ofresource users (especially fishers) to collect data more cheaplyThe electronic logbook and data entry systems mentioned aboveare just one relatively simple example of the technologies thatcan now be developed for capture of ecological informationAnother major research opportunity is in the development offish tags and tag detectors that can be used by nonscientists tohelp collect better information on distribution movements andexploitation rates Still another is in the use of advancedacoustic and video techniques along with technologies likeremotely operated vehicles (ROVs) to carry out large-scalesurveys of fish and fish habitat

There remains the fundamental problem of how to convincedecision makers to support management experiments I believethat the key to this problem is for the scientific community toagree to stop providing simplistic point estimates and manage-ment advice That is when we are asked for our best estimatesand predictions we should agree to offer only a strategic rangeof possible outcomes for each policy choice (like the extremepredictions in Figure 1) and we should simply refuse to saywhich outcome we think is most likely This tactic would placethe burden of risk management squarely on the shoulders of thedecision makers essentially forcing them to confront the realuncertainty and think carefully about ways to deal with it Inrecent years the fisheries stock assessment literature has beenflooded with papers about how to calculate Bayesian probabil-ity distributions for key management quantities (20) Unfortu-nately while such analyses certainly look impressive from amathematical-statistics perspective they are generally based onhopelessly optimistic assumptions about statistical variabilityand knowledge of structural relationships and hence grosslyunderestimate how much uncertainty ought to be admitted (2122) Further the results of these analyses are typically presentedas probability distributions which are difficult to interpret andare accompanied not by recommendations to do managementexperiments but rather with admonishments about the need forcautious (so-called lsquolsquoprecautionaryrsquorsquo) management a risk aversemanagement judgment that scientists ought not to be making inthe first place

The history of fisheries management has largely involvedmyopic focus on single fish stocks and the direct impacts offishing on them Under that focus it has been permissible tobase policy on relatively simple dynamic models and to ignoremore complex interaction effects like described above for theGulf of Mexico shrimp trawl fishery Adaptive management hasbeen of little help in dealing with single stock management issueBut the Gulf of Mexico example is not an isolated or unusualone it is routine to see counterintuitive predictions (andinability to test these with fragmentary historical data) whenwe attempt to model trophic and habitat interactions in supportof ecosystem management questions such as how fisheries atone trophic level affect productivity of fisheries at other levelsand whether natural predators should be culled in order tomake fisheries more productive (23 24) Management agenciestoday are under very strong pressure to adopt ecosystem-basedmanagement approaches (25 26) With that pressure has comedemands for ecosystem modeling and those models will furtherexpose just how deep our uncertainty is about the efficacy ofeven apparently simple regulatory measures like makingfisheries more selective to reduce bycatch If we are honest

about admitting and embracing that uncertainty in developingpolicy advice adaptive management experiments will finallycome to play a central role in the management of fisheries andtheir ecosystems We will simply be forced to find ways toovercome the problems of selling experimental policies todecision makers leadership in implementation and high costsof monitoring

References and Notes

1 Walters C and Hilborn R 1976 Adaptive control of fishing systems J Fish ResBoard Canada 33 145ndash159

2 Walters CJ and Hilborn R 1978 Ecological optimization and adaptive managementAnn Rev Ecol Syst 9 157ndash188

3 Holling CS (eds) 1978 Adaptive Environmental Assessment and Management JohnWiley New York 377 pp

4 Walters C 1986 Adaptive Management of Renewable Resources McMillan PublisherCo New York 374 pp

5 Walters C 1997 Challenges in adaptive management of riparian and coastal ecosystemsConservation Ecology 1 (httpwwwconsecolorgvol1iss2art1)

6 Gunderson L Holling CS and Light SS 1995 Barriers and Bridges to the Renewal ofEcosystems and Institutions Columbia University Press New York 593 pp

7 Gunderson L and Holling CS (eds) 2002 Panarchy Understanding Transformationsin Human and Natural Systems Island Press Washington DC 450 pp

8 Parma A and NCEAS Working Group on Population Management 1998 What canadaptive management do for our fish forests food and biodiversity Integr Biol 1 16ndash26

9 Allan C and Curtis A 2005 Nipped in the bud why regional scale adaptivemanagement is not blooming Environ Management 36 414ndash425

10 Schreibner ESG Bearlin AR Nicol SJ and Todd CR 2004 Adaptivemanagement a synthesis of current understanding and effective application EcolManagement Rest 5 177ndash182

11 Ortiz M Legault CM and Erhardt NM 2000 An alternative method for estimatingbycatch from the US shrimp trawl fishery in the Gulf of Mexico Fish Bull 98 583ndash599

12 SEDAR 7 2005 Assessment summary report Gulf of Mexico red snapper (httpwwwsefscnoaagovsedar)

13 Holling CS and Chambers AD 1973 Resource science the nurture of an infantBioScience 23 13ndash20

14 Walters CJ and McGuire JJ 1996 Lessons for stock assessment from the northerncod collapse Rev Fish Biol Fish 6 125ndash137

15 Mapstone BD Davies CR Little LR Punt AE Smith ADM Pantus F LouDC Williams AJ et al 2004 The Effects of Line Fishing on the Great Barrier Reefand Evaluations of Alternative Potential Management Strategies CRC Reef Res CentreTechnical Report 52 pages indashxi and 1ndash205

16 Campbell RA Mapstone BD and Smith ADM 2001 Evaluating large scaleexperimental designs for management of coral trout on the Great Barrier Reef EcolAppl 11 1763ndash1777

17 Walters CJ and Collie JS 1989 An experimental management strategy forgroundfish in the face of large uncertainty about stock size and production Can JFish Aquat Sci 108 13ndash25

18 Hilborn R Dealteris J Deriso R Graham G Iudicello S Lundsten M PikitchEK Sylvia G et al 2004 Cooperative Research in the National Marine FisheriesService The National Academies Press Washington DC 132 pp

19 See reviews and advice offered on the Adaptive Management Practitioners Networkwebsite (httpwwwadaptivemanagementnetwhatisphp)

20 Patterson K Cook R Darby C Gavaris S Kell L Lewy P Mesnil B Punt Aet al 2001 Estimating uncertainty in fish stock assessment and forecasting Fish Fish 2125ndash157

21 McAllister MK and Kirchner C 2002 Accounting for structural uncertainty tofacilitate precautionary fishery management illustration with Namibian orange roughyBull Mar Sci 70 499ndash540

22 Clark CW 2006 The Worldwide Crisis in Fisheries Economic Models and HumanBehavior Cambridge Univ Press New York 264 pp

23 Yodzis P 2001 Must top predators be culled for the sake of fisheries Trends Ecol Evol16 78ndash84

24 Bakun A 2006 Wasp-waist populations and marine ecosystem dynamics navigatingthe lsquolsquopredator pitrsquorsquo topographies ProgOceano 68 271ndash288

25 Pikitch EK Santora C Babcock EA Bakun A Bonfil R Conover DODayton P Doukakis P et al 2004 Ecosystem-based fishery management Science305 346ndash347

26 Sainsbury KJ Campbell RA and Whitelaw WW 1993 Effects of trawling on themarine habitation the North West Shelf of Australia and implications for sustainablefisheries management In Sustainable Fisheries through Sustainable Habitat HancockDA (ed) Bureau of Rural Sciences Proceedings AGPS Canberra pp 137ndash145

27 First submitted 12 December 2006 Accepted for publication 12 December 2006

Dr Carl Walters is currently Professor of Zoology andFisheries at the University of British Columbia VancouverCanada Dr Walters is a Fellow of the Royal Society ofCanada (1998) and a Pew Fellow in Marine Conservation(2001) He was also the 2001-2002 Mote Eminent Scholar atFlorida State University and the Mote Marine Laboratory In2006 he received the Volvo Environment Prize and theAmerican Fisheries Society Award of Excellence His addressFisheries Centre University of British Columbia VancouverBC V6T 1Z4 CanadaE-mail cwaltersfisheriesubcca

Ambio Vol 36 No 4 June 2007 307 Royal Swedish Academy of Sciences 2007httpwwwambiokvase

LAND-247

the intellectual skills needed to originate an adaptive manage-ment proposal (critical scientific attitude knowledge ofmodeling and statistical design monitoring methods experience)are just not the right ones for carrying out the hard oftenboring and generally frustrating work of implementation

There are great opportunities for ecological scientists todayto participate creatively in the development of innovativemonitoring programs that utilize expertise and experience ofresource users (especially fishers) to collect data more cheaplyThe electronic logbook and data entry systems mentioned aboveare just one relatively simple example of the technologies thatcan now be developed for capture of ecological informationAnother major research opportunity is in the development offish tags and tag detectors that can be used by nonscientists tohelp collect better information on distribution movements andexploitation rates Still another is in the use of advancedacoustic and video techniques along with technologies likeremotely operated vehicles (ROVs) to carry out large-scalesurveys of fish and fish habitat

There remains the fundamental problem of how to convincedecision makers to support management experiments I believethat the key to this problem is for the scientific community toagree to stop providing simplistic point estimates and manage-ment advice That is when we are asked for our best estimatesand predictions we should agree to offer only a strategic rangeof possible outcomes for each policy choice (like the extremepredictions in Figure 1) and we should simply refuse to saywhich outcome we think is most likely This tactic would placethe burden of risk management squarely on the shoulders of thedecision makers essentially forcing them to confront the realuncertainty and think carefully about ways to deal with it Inrecent years the fisheries stock assessment literature has beenflooded with papers about how to calculate Bayesian probabil-ity distributions for key management quantities (20) Unfortu-nately while such analyses certainly look impressive from amathematical-statistics perspective they are generally based onhopelessly optimistic assumptions about statistical variabilityand knowledge of structural relationships and hence grosslyunderestimate how much uncertainty ought to be admitted (2122) Further the results of these analyses are typically presentedas probability distributions which are difficult to interpret andare accompanied not by recommendations to do managementexperiments but rather with admonishments about the need forcautious (so-called lsquolsquoprecautionaryrsquorsquo) management a risk aversemanagement judgment that scientists ought not to be making inthe first place

The history of fisheries management has largely involvedmyopic focus on single fish stocks and the direct impacts offishing on them Under that focus it has been permissible tobase policy on relatively simple dynamic models and to ignoremore complex interaction effects like described above for theGulf of Mexico shrimp trawl fishery Adaptive management hasbeen of little help in dealing with single stock management issueBut the Gulf of Mexico example is not an isolated or unusualone it is routine to see counterintuitive predictions (andinability to test these with fragmentary historical data) whenwe attempt to model trophic and habitat interactions in supportof ecosystem management questions such as how fisheries atone trophic level affect productivity of fisheries at other levelsand whether natural predators should be culled in order tomake fisheries more productive (23 24) Management agenciestoday are under very strong pressure to adopt ecosystem-basedmanagement approaches (25 26) With that pressure has comedemands for ecosystem modeling and those models will furtherexpose just how deep our uncertainty is about the efficacy ofeven apparently simple regulatory measures like makingfisheries more selective to reduce bycatch If we are honest

about admitting and embracing that uncertainty in developingpolicy advice adaptive management experiments will finallycome to play a central role in the management of fisheries andtheir ecosystems We will simply be forced to find ways toovercome the problems of selling experimental policies todecision makers leadership in implementation and high costsof monitoring

References and Notes

1 Walters C and Hilborn R 1976 Adaptive control of fishing systems J Fish ResBoard Canada 33 145ndash159

2 Walters CJ and Hilborn R 1978 Ecological optimization and adaptive managementAnn Rev Ecol Syst 9 157ndash188

3 Holling CS (eds) 1978 Adaptive Environmental Assessment and Management JohnWiley New York 377 pp

4 Walters C 1986 Adaptive Management of Renewable Resources McMillan PublisherCo New York 374 pp

5 Walters C 1997 Challenges in adaptive management of riparian and coastal ecosystemsConservation Ecology 1 (httpwwwconsecolorgvol1iss2art1)

6 Gunderson L Holling CS and Light SS 1995 Barriers and Bridges to the Renewal ofEcosystems and Institutions Columbia University Press New York 593 pp

7 Gunderson L and Holling CS (eds) 2002 Panarchy Understanding Transformationsin Human and Natural Systems Island Press Washington DC 450 pp

8 Parma A and NCEAS Working Group on Population Management 1998 What canadaptive management do for our fish forests food and biodiversity Integr Biol 1 16ndash26

9 Allan C and Curtis A 2005 Nipped in the bud why regional scale adaptivemanagement is not blooming Environ Management 36 414ndash425

10 Schreibner ESG Bearlin AR Nicol SJ and Todd CR 2004 Adaptivemanagement a synthesis of current understanding and effective application EcolManagement Rest 5 177ndash182

11 Ortiz M Legault CM and Erhardt NM 2000 An alternative method for estimatingbycatch from the US shrimp trawl fishery in the Gulf of Mexico Fish Bull 98 583ndash599

12 SEDAR 7 2005 Assessment summary report Gulf of Mexico red snapper (httpwwwsefscnoaagovsedar)

13 Holling CS and Chambers AD 1973 Resource science the nurture of an infantBioScience 23 13ndash20

14 Walters CJ and McGuire JJ 1996 Lessons for stock assessment from the northerncod collapse Rev Fish Biol Fish 6 125ndash137

15 Mapstone BD Davies CR Little LR Punt AE Smith ADM Pantus F LouDC Williams AJ et al 2004 The Effects of Line Fishing on the Great Barrier Reefand Evaluations of Alternative Potential Management Strategies CRC Reef Res CentreTechnical Report 52 pages indashxi and 1ndash205

16 Campbell RA Mapstone BD and Smith ADM 2001 Evaluating large scaleexperimental designs for management of coral trout on the Great Barrier Reef EcolAppl 11 1763ndash1777

17 Walters CJ and Collie JS 1989 An experimental management strategy forgroundfish in the face of large uncertainty about stock size and production Can JFish Aquat Sci 108 13ndash25

18 Hilborn R Dealteris J Deriso R Graham G Iudicello S Lundsten M PikitchEK Sylvia G et al 2004 Cooperative Research in the National Marine FisheriesService The National Academies Press Washington DC 132 pp

19 See reviews and advice offered on the Adaptive Management Practitioners Networkwebsite (httpwwwadaptivemanagementnetwhatisphp)

20 Patterson K Cook R Darby C Gavaris S Kell L Lewy P Mesnil B Punt Aet al 2001 Estimating uncertainty in fish stock assessment and forecasting Fish Fish 2125ndash157

21 McAllister MK and Kirchner C 2002 Accounting for structural uncertainty tofacilitate precautionary fishery management illustration with Namibian orange roughyBull Mar Sci 70 499ndash540

22 Clark CW 2006 The Worldwide Crisis in Fisheries Economic Models and HumanBehavior Cambridge Univ Press New York 264 pp

23 Yodzis P 2001 Must top predators be culled for the sake of fisheries Trends Ecol Evol16 78ndash84

24 Bakun A 2006 Wasp-waist populations and marine ecosystem dynamics navigatingthe lsquolsquopredator pitrsquorsquo topographies ProgOceano 68 271ndash288

25 Pikitch EK Santora C Babcock EA Bakun A Bonfil R Conover DODayton P Doukakis P et al 2004 Ecosystem-based fishery management Science305 346ndash347

26 Sainsbury KJ Campbell RA and Whitelaw WW 1993 Effects of trawling on themarine habitation the North West Shelf of Australia and implications for sustainablefisheries management In Sustainable Fisheries through Sustainable Habitat HancockDA (ed) Bureau of Rural Sciences Proceedings AGPS Canberra pp 137ndash145

27 First submitted 12 December 2006 Accepted for publication 12 December 2006

Dr Carl Walters is currently Professor of Zoology andFisheries at the University of British Columbia VancouverCanada Dr Walters is a Fellow of the Royal Society ofCanada (1998) and a Pew Fellow in Marine Conservation(2001) He was also the 2001-2002 Mote Eminent Scholar atFlorida State University and the Mote Marine Laboratory In2006 he received the Volvo Environment Prize and theAmerican Fisheries Society Award of Excellence His addressFisheries Centre University of British Columbia VancouverBC V6T 1Z4 CanadaE-mail cwaltersfisheriesubcca

Ambio Vol 36 No 4 June 2007 307 Royal Swedish Academy of Sciences 2007httpwwwambiokvase

LAND-247