june2000agenda

Following is a summary of the Black Duck Adaptive Harvest Management Working Group meeting held on June 5-7, 2000 at the University of Georgia in Athens. The intent of this summary is to provide an overview of the meeting along with key discussions and decisions. This summary does not include "minutes" of all of the activities and discussion that took place.

This summary is posted on the Black Duck AHM Working Group web site at http://coopunit.forestry.uga.edu/blackduck/.Components of the summary included in the appendices may also included other places on the web site and updated over time.

Michael J. Conroy with the Georgia Cooperative Wildlife Research Unit opened the meeting and welcomed all participants.David J. Case of D.J. Case and Associates served as meeting facilitator.

Name	Affiliation	email
Bob Blohm	FWS-MBMO	Robert_Blohm@fws.gov
Mike Conroy	University of Georgia	conroy@fisher.forestry.uga.edu
Brigitte Collins	BDJV Coordinator	Brigitte.Collins@ec.gc.ca
Chris Fonnesbeck	University of Georgia	chrisf@fonnesbeck.net
Daniel Bordage	CWS- Quebec	daniel.bordage@ec.gc.ca
Dave Case	DJ Case and Assoc.	Dave@djcase.com
Fred Johnson	FWS-MBMO	fred_a_johnson@fws.gov
Jerry Serie	FWS	Jerry_Serie@fws.gov
Kathy Dickson	CWS- Ontario	kathy.dickson@ec.gc.ca
Melody Miller	Indiana DNR, Miss. Flyway	mmiller@bluemarble.net
Myrtle Bateman	CWS Atlantic	Myrtle.bateman@ec.gc.ca
Nathan Zimpfer	University of Georgia	nlz5093@owl.forestry.uga.edu
Steve Wendt	CWS-National	steve.wendt@ec.gc.ca
Bryan Swift	NYDEC	blswift@gw.dec.state.ny.us

The Working Group discussed at length their overall purpose and approach. They agreed that:

The group listed the "non-technical" aspects of an adaptive approach to harvest management as:

The Working Group agreed the priorities for discussion at this meeting would be "scale" followed by "objective function." Discussion of "regulatory options" and "distribution of harvest" was felt to be premature.

Adaptive management has been recognized as a potentially useful concept in the development of an international harvest strategy for black ducks. Traditionally, the focus of adaptive management has been on reducing uncertainty about system dynamics, and on understanding the consequences of alternative management actions. This emphasis on management as a scientific enterprise can be problematic, however, if there is insufficient attention to the more subjective or policy-based aspects of resource management. Management policy ultimately is a function of: (1) beliefs about causation (i.e., the consequences of alternative management actions); and (2) preferences about outcomes (i.e., how society values alternative consequences). The first part involves objective science, while the second part involves subjective, value-based judgements about the goals and objectives of management. The purpose of my presentation was to clearly differentiate between these two fundamental aspects of management, and to explore their implications for the management of black duck harvests.

From a scientific perspective, harvest management is a sequential decision-making process in which the manager must periodically set hunting regulations. Each regulatory decision is dependent on the current state of the resource system (in this case, black duck population status), and the manager? goal is to make a temporal sequence of regulatory decisions that maximizes some performance metric (e.g., harvest) over an extended, if not infinite, time frame. There are four fundamental sources of uncertainty in this process that affect the manager? ability to make effective decisions: (1) partial system observability, reflecting the sampling error inherent in resource-monitoring programs; (2) uncontrolled environmental variation; (3) partial controllability, which is a lack of concordance between intended and actual harvests; and (4) uncertainty about the appropriate structure of system models and, therefore, about the predictions they make. An adaptive decision-making process that accounts for these uncertainties includes: (1) a monitoring program to recognize resource status; (2) a finite set of regulatory alternatives; (3) state and decision-dependent predictions of outcomes, which are based on alternative models of system dynamics; and (4) an unambiguous management objective. The adaptive approach is a three-step process:

(1) at the appropriate periodic interval, an optimal regulatory decision is identified based on resource status, and on ?rior?probabilities (i.e., current measures of credibility) associated with each alternative model of system dynamics;

(2) conditioned on that regulatory decision, model-specific predictions for population size in the next time step are determined;

(3) when monitoring data become available, predicted and observed population sizes are compared, and the probability associated with each model is increased or decreased depending on its relative ability to predict the actual change in population size.

The new or "posterior" model probabilities then become the prior probabilities, and are used to derive regulatory prescriptions for the next time step. Harvest strategies therefore "evolve" over time in response to changes in the characterization of structural uncertainty. This process eventually will identify the most appropriate model(s) of system dynamics and, therefore, lead to improved management decisions in the future.

This adaptive decision-making process provides useful analytical structure to the harvesting problem, explicitly links monitoring data and regulatory decisions, and accounts for the dynamic nature of both the resource system and our understanding of that system. However, the process cannot determine which alternative models (or hypotheses) to include, which regulatory alternatives to consider, acceptable management goals and objectives, nor the appropriate spatial, temporal, or organization scales of management. These management issues involve subjective, policy-based decisions.

One of the most problematic of these issues is the determination of appropriate management scales. All ecological systems exhibit variability on a broad range of temporal, spatial, and organizational scales, ultimately as a function of how individual animals respond to their environment. The manner in which individuals are aggregated or stratified (e.g., by spatially segregated populations of conspecifics) for management purposes is an arbitrary decision, but one that can strongly influence both the benefits and costs of management. Management approaches that account for important sources of ecological variation are expected to yield the highest benefits, but also are characterized by relatively high monitoring and assessment costs. Moreover, as the spatial, temporal, and organizational scales at which harvest management is delivered become progressively smaller, the marginal gain in management benefit likely will diminish. At the same time, management costs would continue to increase at a constant or even accelerating rate. Determining the optimal level of aggregation or stratification for harvest management depends critically on the availability of explicit performance criteria (i.e., costs and benefits), and on understanding patterns of ecological variation. Description of these ecological patterns, in turn, depends on sufficient data to investigate potential sources of variation and to suggest underlying causal mechanisms.

The application of adaptive management to black duck harvest management thus involves a number of challenges from both a scientific and human-dimension perspective. Adaptive management is technically complex, and requires a large dose of institutional commitment and patience. Stakeholders must be able to separate interests from positions if broadly accepted solutions to various issues are to be found. As managers, we must be cognizant of the fact that science is idealistic, while politics is practical. Therefore, there is an inevitable tension between the two that occasionally will require compromise on the part of both scientists and administrators. In the end, successful application of adaptive harvest management for black ducks will depend on whether stakeholders feel a sense of ownership in the process, and are prepared to trust those making both objective and subjective decisions on their behalf.

In this presentation I reviewed some key concepts underlying the black duck AHM models, starting with simple models of density independence and moving into density dependence, incorporation of stochastic environmental effects, and finally including the impacts of observational error on conclusions from modeling and population surveys.I then covered technical details of the models of black duck populations currently under development for use in AHM.The current AHM model describes a single population of black ducks initially observed via breeding population surveys, and projected through the year via modeling of alternative assumptions about harvest compensations and the impacts of mallards on reproduction and survival rates.At present there are 4 alternative models: HA11 (compensation plus mallard competition), HA10 (compensation plus no mallard competition), HA01 (no compensation plus mallard competition), HA00 (neither compensation nor mallard competition).

Mike Conroy, Nathan Zimpfer, and Chris Fonnesbeck provided attendees with copies of several programs that illustrated 1) simple population growth, 2) density dependent growth and compensatory mortality, and 3) the effects of random environmental variation and statistical error on observed populations trajectories. A simulation version of the single population black duck model was demonstrated, allowing users to see the predicted effects of harvest manipulation and other changes on black duck populations under each of the 4 models described above.A simple decision algorithm programmed in Netica (http://www.norsys.com/netica.html) was used to demonstrate decision making under uncertainty, and how long term objectives (such as cumulative harvest) may modify current decision making, both fundamental concepts of Adaptive Harvest Management (e.g, see http://www.consecol.org/Journal/vol3/iss1/art8/for a more complete conceptual and technical framework for AHM in waterfowl management).

Natural resource management typically involves the making of decisions under uncertainty, often with relatively weak understanding or predictive ability about the systems under management.Decision theory allows for the decision maker to make a rational decision in the face of uncertainty, essentially by evaluating the value of a decision (e.g., to allow harvest) to the decision maker or stakeholders under alternative possible outcomes, and computing the average or expected value of these outcomes.The decision maker then selects the decision that appears optimal, that is, provides the maximum expected objective value.Complicating this procedure is the fact that decisions must be made sequentially through time, with future outcomes and optimal decisions unknown, but possibly influenced by current decisions.For example, a waterfowl population may increase next year in part through future events (e.g., favorable environmental conditions) that are unpredictable, and in part as a result of current management decisions (e.g., restrictive hunting regulations). The combined effect of these present actions and uncertain future outcomes must be taken into effect in evaluating those management alternatives likely to lead to the highest long term benefit (e.g., sustainable harvest).Dynamic programming allows this complex problem to be solved, by decomposing the problem into a series of decision steps, beginning with the distant future and working backward to the present.In this way, present decisions are guaranteed to be optimal in the long term, having been found to be optimal at each future time step. Complicating this process is the fact that assumptions over which model is operating will affect which future outcomes are assumed most likely, and thus which decisions appear optimal.Rather than arbitrarily assuming that a single model is correct, the approach under AHM is to average the value of each decision over the uncertain models, weighted by a prior belief (often equal) for each model.Predictions under each model are then compared to observed outcomes (e.g., next year? breeding population estimate) and used to update the model weights. The value of this accumulation of knowledge can be calculated and expressed in units of the objective function, and may be used by managers to help decide which research and monitoring programs will be most helpful in reaching management goals.

Update on databases and analyses to support AHM (Conroy, Zimpfer, Fonnesbeck)

We described several databases and analyses currently being used to support and parameterize the AHM models for black ducks.First, we presented a descriptive analysis of band recovery distribution for black ducks that provides some preliminary ideas as to how population units might be delineated for further modeling based on spatial stratification.We then described an analysis of harvest trends in Canada and the U.S. (provided by D. Bordage) which suggests that on average harvests in the two nations have been at rough parity overall, in spite of some periods in which one country or the other had higher total harvests.Overall declines in total harvest likely are due to a combination of 1) declining black duck populations over some periods, 2) restrictive regulations in the 1980s, and 3) decline in hunter interest in participation, especially in Canada.Next we focused on analyses conducted in an attempt to re-calibrate the AHM models, originally developed using midwinter inventory (MWI) estimates, in terms of breeding population surveys (BPOP) for black ducks and mallards.Attempts to develop prediction equations between MWI and the Breeding Bird Survey (BBS) on one hand, and BBS and BPOP on the other, were only partially successful, and we are currently using adjustment estimates based on simple ratio-of-means estimators for overlapping sample periods. These have now been used to modify the AHM models, so that these now accept as state variable input the observed BPOP (currently based on the sum of helicopter and fixed-wing surveys in eastern Canada and the northeastern U.S.) instead of MWI for black ducks and mallards.We have also re-estimated age ratios for black ducks, based on U.S. harvest data adjusted for differential vulnerability using U.S. band recoveries. These data, together with revised abundance indices based on historical surveys, but for mallards excluding areas of the Mississippi Flyway and southern Atlantic Flyways thought to be more reflective of mid-continent conditions, have been used to provide updated estimates for the coefficients of the production sub-model, under two alternatives:no influence and negative influence of mallards on production rates of breeding black ducks. Both models contain a density-dependent effect, so that population growth is ultimately limited for all models.The revised population data have also been used to re-fit the survival sub-model, under four models assuming combinations of compensation/ additivity, and mallard effect/ no mallard effect.In addition, new model coefficients include more realistic estimates of spring-fall survival and relative vulnerability to harvest, based on new analyses of band recoveries.Work continues on re-estimation of the production model based on BPOP estimates, age ratios including the Canadian harvest data, and delineation of spatial subpopulations.

The issues of spatial resolution and the objective function were discussed at length by the Working Group.

Working group members discussed three aspects of spatial resolution-the number and delineation of breeding populations, the number and delineation of harvest units, and the frequency of decisions.

Members had diverse perspectives on both breeding populations and harvest units. A number of people felt there should be 3 of each to reflect the differences in population parameters from east to west in the breeding range. Others felt breeding populations and harvest units should be determined based on the analyses being conducted by the University of Georgia group. Some expressed the need to base decisions on costs and benefits of various options in terms of management performance.

In general, the Working Group agreed that further analysis of the data was needed and that decisions on breeding populations and harvest units must be based on the available data and analysis.

Objective Function
The Working Group discussed considerations or "criteria" for developing an objective function:

In order to proceed with this investigation, we need to continually access updated data from populations surveys, band recoveries, harvest surveys and other survey and monitoring programs on black ducks, mallards, and their habitats. These data were obtained from a number of sources maintained separately in the U.S. and Canada and are available via this website for the convenience of the project cooperators, and to provide a method for others wishing to test our methods or assumptions to do so using appropriate data. To the extent possible we provide users with information about how these data were collected, via brief descriptions of data format, definitions of data fields, and descriptions and/or links for the details of survey methodology. However, we make no representations as to the accuracy of these data, and in all cases refer the user to the original source(s) for the original data, methods, and descriptions.

The next meeting of the Black Duck AHM Working Group was scheduled for Friday, November 17 from 8 am to 4 pm in Sandusky, Ohio. The Black Duck Joint Venture meeting will be held November 14-16 in the same location.The November 17 meeting agenda will be a workshop designed to address the scale issue. Technical committee members will be invited to attend.A draft working model will be completed by December 2002 with a June 2003 completion date for the project.

The Working Group agreed that a meeting report is important to have. Dave Case will work with Mike Conroy and Chris Fonnesbeck to draft the report and post it on the secure portion of the web site. Once all Working Group members have approved the report, it will be posted on the public site.

Presenters should get their reports to Mike Conroy by July 10 with draft report to be completed by July 17. Daniel Bordage will help get a French version developed.

Appendix A. Agenda

1300-1500 Technical Key Elements of Black Duck AHM Models?M. Conroy/ N. Zimpfer / C. Fonnesbeck

CCosts/Funding Concerns

CCommunications

CMeeting summaries and communication within group

CNext meeting?bjectives, agenda, time, and location

COther items as needed

1200Lunch

1500 Adjourn

Appendix B. Relevant reports and other documents

Developing Models of Black Duck Populations in North America, Final Report .Valuable conceptual background for the development of the black duck models.Extensive literature review of factors potentially affecting black duck populations.

Adobe (pdf) format

2000 Annual Report for Development of an Integrated, Adaptive Management Protocol for American Black Ducks

Adobe (pdf) format

Protocol and Practice in the Adaptive Management of Waterfowl Harvests .A conceptual framework for the adaptive management of waterfowl harvest. F.A. Johnson and B.K. Williamshttp://www.consecol.org/Journal/vol3/iss1/art8/