ADAPTIVE MANAGEMENT AND THE REGULATION OF WATERFOWL
Byron K. Williams, U.S. Geological Survey, Division of
Biological Resources, 12201 Sunrise Valley Drive, Reston, VA 20192
Fred A. Johnson, Office of Migratory Bird Management, U.S. Fish
and Wildlife Service, 11500 American Holly Drive, Laurel, MD 20708-4016
(* originally published in 1995 in The Wildlife Society
Bulletin, Vol. 23(3):430-436)
Migratory bird hunting is an important form of outdoor recreation in the
United States. Each year, approximately 3 million people engage in 22
million days of migratory bird hunting, with an expenditure of $700
million (U.S. Deps. Inter. and Commer. 1993). Although the number of
waterfowl hunters has declined somewhat in recent years (Trost et al.
1987), duck and goose hunting still constitutes about one-third of all
migratory bird hunting activity (Martin, E. M. and P. J. Padding, 1994,
Preliminary estimates of waterfowl harvest and hunter activity in the
United States during the 1993 hunting season, U.S. Fish and Wildl. Serv.,
Off. Migratory Bird Manage. Adm. Rep., Laurel, MD, 1994).
In response to demands for increased hunting opportunities, waterfowl
harvest regulations have grown to include features such as special seasons
on more abundant species, geographic zoning, and species-specific bag
limits (U.S. Dep. Inter. 1988). These regulations, along with basic season
lengths and bag limits, constitute a sometimes bewildering network of
regulatory options. The impacts of these regulations on harvest and
population status is highly uncertain, even though North American
waterfowl are among the world's most thoroughly investigated biota
(Nichols and Johnson 1989).
Several factors have contributed to uncertainty about the effects of
waterfowl harvest regulations. One factor is the sheer number and
complexity of regulatory options offered in recent years. The introduction
of so many options has complicated the regulations process greatly and has
made it effectively impossible to assess either their marginal or
cumulative effects. A second factor is the large-scale confounding of
harvest and environmental effects that occurs when regulations "chase"
population and habitat conditions (i.e., liberal regulations are set
whenever populations appear to be abundant, and restrictive regulations
are used whenever populations are low). Such a strategy may be appropriate
under conditions in which population dynamics and the effects of hunting
are well understood. However, in this case the knowledge required is
unavailable to waterfowl managers, and "chasing" populations
makes acquiring the information needed to reduce key uncertainties
virtually impossible. A third factor concerns limitations of "reductionist
science" in which the behavior of a managed system is understood by
dividing it into subsystems more amenable to investigation (e.g., studying
the effect of harvest on annual survival rate). A reductionist approach is
likely to be most useful if one is dealing with processes that are
additive in their effects on population dynamics (i.e., survival,
recruitment, immigration, emigration), a situation seldom encountered in
These and other factors have resulted in an inability to recognize
effects of regulations on waterfowl population dynamics. Because this
uncertainty limits our ability to make regulatory decisions consistent
with long-term harvest and conservation goals, there are material benefits
in its reduction. We advocate an approach known as "adaptive resource
management" (Holling 1978, Walters 1986) that explicitly recognizes
uncertainty about management impacts and seeks to provide useful
information about system dynamics. Incorporation of uncertainty in the
formulation of management strategies sets adaptive resource management
apart from other, more traditional applications of strategic decision
We describe an adaptive approach to regulating waterfowl harvests. Our
objectives are to: (1) propose adaptive harvest management as an extension
of the current regulatory process, focusing on active pursuit of knowledge
for informed decision-making; and (2) discuss key issues and challenges
associated with adaptive harvest management.
Managing adaptively requires feedback between management and assessment,
when each activity influences the other. Thus, managers must assess a
managed system periodically and somehow adapt decisions to the system
state while accounting for uncertainty about the effects of those
decisions. Adaptive management is sometimes described as "embracing"
uncertainty (Walters 1986), in that it recognizes uncertainty as an
attribute of management, and uses management itself as a tool to
accelerate the reduction in uncertainty. A definition we find useful is:
"Adaptive harvest management describes the ability to make
a sequence of decisions, in the face of uncertainty, that is optimal with
respect to a stated objective, recognizing some constraints"
(D. R. Anderson, Natl. Biol. Serv., Ft. Collins, Colo., pers. commun.
1995). Less formally, adaptive harvest management might be described as
managing in the face of uncertainty, with a focus on its reduction. An
adaptive approach emphasizes uncertainty about regulatory effects and
incorporates uncertainty as a factor guiding management actions (Johnson
et al. 1993). It also implies that the performance of management can be
improved if uncertainties are reduced.
For management to account for resource status as well as uncertainty
about harvest effects, 3 critical components linking management,
assessment, and population dynamics are required: (1) a process of
decision making with clear, focused management objectives; (2) a
monitoring program that periodically determines the status of the
resource; (3) a process by which the effects of management decisions on
the resource can be assessed.
Decision making process.--The process of regulating waterfowl
harvests is well known, and involves a rather lengthy sequence of public
announcements, deliberations, and decision making (Blohm 1989).
Participation by the U.S. Fish and Wildlife Service (USFWS), state
wildlife agencies, the Canadian and Mexican governments, and the public
occurs at numerous times during the regulatory cycle. The process involves
assessment of waterfowl populations, publication of Federal Register notices, and numerous meetings by the Waterfowl Flyway Councils and USFWS
Regulations Committee. It culminates in selection of regulations at the
flyway level (season lengths, daily bag limits, and outside dates for the
earliest opening and latest closing dates for a hunting season) and
special regulations at the state level (e.g., split seasons, harvest
zones, special seasons, area closures). A major challenge for waterfowl
biologists is to identify the effects of such a profusion of harvest
A critical component of this process should be the unambiguous
specification of objectives for harvest management. Sound harvest
management requires clear objectives, if only to measure how well the
decision-making process has worked over time. This apparently simple
requirement has been difficult to fulfill. Indeed, many of the conflicts
arising in waterfowl harvest regulation can be traced back to disagreement
about goals and objectives for harvest management. At various times the
Flyway Councils and the USFWS Regulations Committee have promulgated
regulations to maximize average harvest, minimize harvest variability,
stabilize population size, and other variations. A predictable consequence
of this conflicting process is that regulations have failed to meet any of
the proposed objectives and thus have been unsatisfactory to many
participants and stakeholders. A rational regulatory process that
encourages regulations pursuant to recognized, agreed upon goals and
objectives is not possible in the face of such confusion and controversy.
Clarification of goals and objectives is an ongoing need, and must be seen
as a key responsibility of management.
Monitoring programs.--Informed decision making consists of
large-scale monitoring programs that deliver information about population
status and trends, harvest levels, and other important biological
attributes. This information is crucial to ascertain the impacts of
harvest regulation, and thereby to establish a coherent framework for
setting harvest regulations.
Key sources of information in the regulatory process are the long-term
cooperative monitoring programs of the USFWS, state wildlife agencies, and
federal and provincial governments in Canada. These programs yield
information about breeding-population status, harvest levels, production,
migration, and other population characteristics of value in regulating
harvests. Data collected each year and added to long-term databases
represent a foundation for harvest management of waterfowl. They are the
baseline on which much waterfowl research is conducted and are necessary
for modeling waterfowl populations.
Analysis and assessment.--Many organizations have played key
roles in adding to knowledge about waterfowl, including the Flyway Council
Technical Committees, the USFWS Office Migratory Bird Management, and
wildlife research programs at Patuxent Environmental Science Center,
Northern Prairie Science Center, the Cooperative Fish and Wildlife
Research Units, and other research institutions. These groups have made
important advances toward a sound understanding of waterfowl populations
and the impacts of harvest, including investigating patterns in monitoring
data, estimating key population parameters such as survivorship and
reproduction, and predicting harvest impacts on population dynamics.
Information accumulated through monitoring and assessment is folded into
models of population size and distribution as influenced by harvest
regulations. The goal of these efforts is to model the responses of a
population to harvest regulation, based on long-term monitoring and
research programs. By building on the databases they are designed to
represent, these models provide valuable information to management, and
thus represent a crucial link in the regulations process.
Regulating Waterfowl Harvests
The current regulatory process links decision making, modeling and
assessment, and population monitoring in an iterative cycle. Regulations
in 1 year influence harvest, which in turn influences population status in
the next year. Effects of harvest regulations on a waterfowl population
are reflected in monitoring data, which add to the cumulative body of
information that is used to update population models. These models in turn
guide the regulatory process in the next cycle. Model updating occurs each
year with new monitoring data, so that both the models and the information
base they represent constantly evolve.
In this scenario regulations have both direct as well as indirect
effects, and both are key to effective regulation of waterfowl harvests.
First, regulations directly affect a population by influencing the amount
of harvest, and through harvest, the subsequent population size. Second,
regulations have indirect effects by influencing the information available
for the subsequent regulatory cycle. For example, following on a period in
which regulations have been relatively constant for a number of years, new
insights often can be gained by deliberating changing regulations and
observing the response in harvest and population size. It seems intuitive
that "informative" regulations are in some sense better than
regulations that are not informative.
The current regulatory scenario is itself adaptive, in that it describes
a procedure whereby regulations are adapted to the available monitoring
data. For the management of waterfowl harvests, this means periodic
updating of databases, incorporating these data into improved population
models, and using this information for setting annual harvest regulations.
A typical application would involve using population models to explore the
impacts of a number of different regulations to identify regulations that
maximize harvest (or harvest opportunity) and limit the negative effects
on populations. Regulations thus identified should guide the
The Pursuit of Information With Regulations
Although the regulatory procedure described above is in some sense
adaptive, the strategy is far from optimal for attaining management
objectives. Its key limitations are a failure to account for uncertainty
about population responses to regulations, and a failure to recognize
value in attaining useful information to reduce that uncertainty. Thus,
information is simply an unplanned by-product of harvest regulations, and
the process is an example of passive adaptive management (Walters and
Holling 1990). Although a passive adaptive approach can lead to improved
management over time, improvements typically accrue very gradually. With
the single exception of the period from 1980-84 when waterfowl harvest
regulations were stabilized (Patterson and Sparrowe 1987), waterfowl
harvest management has been (and continues to be) passively adaptive.
The use of regulations to actively pursue understanding of regulatory
effects is an example of active adaptive management (Walters and Holling
1990). In this case the accretion of information and reduction of
uncertainty about population responses are incorporated explicitly as an
objective in the decision-making process. Some regulatory strategies are
likely to be more informative than others, because they lead to more
informative databases and improved models for describing the consequences
of regulations. Active adaptive management entails the use of such
strategies while pursuing more traditional harvest management objectives.
Hereafter, we use the phase adaptive harvest management to mean the
active pursuit of information through regulations. To avoid any confusion
about the role of information in adaptive harvest management, it is useful
to identify its role explicitly. An adaptive approach emphasizes resource
management per se, with value ascribed to information and understanding
only to the extent that they contribute to the objectives of resource
management. Thus, adaptive management does not recognize intrinsic value
in biological monitoring, research, or scientific assessment. From a
management viewpoint these activities, and the knowledge they produce, are
justified only to the extent that they serve management purposes. It may
be reassuring that adaptive management, by recognizing the importance of
reliable information on which to base management decisions, reinforces
strong cooperation between researchers and managers.
Adaptive Harvest Management
Adaptive harvest management can be described in terms of 4 components:
(1) an array of potential hunting regulations that are available to
decision makers for the control of waterfowl harvests; (2) a set of models
representing meaningful hypotheses about population dynamics and the
effects of harvest; (3) a measure of "uncertainty" for each
model, expressing the relative likelihood that it appropriately describes
population responses to regulations; and (4) an objective function (i.e.,
a mathematical expression of harvest management objectives) by which to
evaluate and compare regulatory options. These components are used to
identify the actively adaptive harvest strategy that is optimal with
respect to management objectives.
The actual mechanics of determining optimal regulatory strategies is
quite complicated and beyond the scope here (see Williams 1988, 1989;
Lubow 1993, 1995). In essence, the problem is cast in the framework of
constrained optimization of stochastic dynamic systems, with an objective
function based on expected long-term harvests weighted by the model
likelihoods. The system in this case consists of the set of population
models, each describing population dynamics in terms of population size,
environmental conditions, and regulations. The models represent different
hypotheses about the impacts of regulations, and the likelihood weights
represent uncertainty as to which hypothesis is most appropriate.
The optimization procedure accounts for both the current population
status and the degree of uncertainty about system dynamics in assessing
the influence of regulatory decisions on future population status
(Williams 1996a,b). It chooses regulations at each point in time based on
the expected sum of present and future harvests, recognizing that future
yields are influenced by regulatory decisions in the present. The goal, of
course, is to choose regulations at each point in time that are optimal,
in that they produce a maximum expected value of present and future
harvests. The process recognizes that optimal management can be realized
only by eventually identifying the most appropriate model of population
The key to an actively adaptive regulations process involves the active
pursuit of information. Regulations influence the removal of individuals
from the population through harvest and hence affect the size of the
population the next year. Monitoring programs record data on harvest and
population status from year to year, which then are used to improve the
models under consideration and update the likelihoods associated with each
model. This information is incorporated in an objective function
consisting of predicted long term harvests for each model, with
model-specific harvests weighted by the updated likelihoods. The
optimization procedure identifies harvest strategies that maximize the
weighted average of harvests, and these strategies subsequently are used
to set regulations. This sequence is repeated each year in an ongoing
cycle of monitoring, model updating, analysis and optimization, and
regulations setting. Information that discriminates among models accrues
with each cycle so that the most appropriate model for describing
population dynamics is identified over time.
International treaties relating to the conservation of migratory birds
clearly mandate that the opportunity to harvest waterfowl is of lesser
importance than the protection and maintenance of populations. Although
these priorities are useful to managers, the potentially competing
objectives (i.e., size of the harvest vs. size of the population) leave
room for debate about appropriate harvest strategies. Even if all
management strategies under consideration provide sustainable harvests,
average population size still can vary under the alternative strategies.
Waterfowl population goals have been identified in the North American
Waterfowl Management Plan (U.S. Dep. Inter. and Environ. Can. 1986). These
goals were established to ensure satisfactory levels of hunting
opportunity, but also for ecological and aesthetic purposes. Though the
Plan has a strong focus on waterfowl habitats, its population goals have
been formally endorsed by the federal governments of Canada, Mexico, and
the U.S. Therefore, waterfowl managers are obligated to consider these
goals in the development of harvest strategies. However, we believe that
the Plan goals are insufficient by themselves to identify unambiguous
harvest objectives and optimal harvest strategies, particularly under
conditions in which habitat cannot adequately support such population
We suggest an objective that attributes relatively high value to hunting
opportunity when Plan goals are met, and lower value to hunting
opportunity as populations fall short of Plan goals. Decline in the value
of hunting opportunity may be linear or non-linear, with the rate of
decline dependent on the relative importance of providing hunting
opportunity and achieving the Plans goals. In developing such an
objective, managers must specify the minimum population level at which
hunting opportunity can be allowed.
Managers also may wish minimize the temporal variability in hunting
regulations for sociological and administrative reasons (U.S. Dep. Inter.
1988). This can be accomplished most easily by either: (1) altering the
frequency of regulatory decisions from 1 to multiple years; or (2)
constructing a set of regulatory options with large differences in
predicted harvest rates. It is likely that either of these options would
lead to smaller amounts of hunting opportunity over the long-term and more
variability in population size than if a temporal constraint on
regulations were not imposed (Hilborn and Walters 1992).
The Array of Regulatory Options
A array of regulatory options representing, for example, restrictive,
moderate, and liberal seasons, must be developed and agreed upon. The
number of options must be limited to facilitate their assessment. In
addition, regulatory options should: (1) elicit different harvest and
population responses; (2) produce predictable harvest rates; (3) be
consistent with hunter preferences to the extent possible; and (4)
facilitate law enforcement.
The development of limited regulatory options provides an opportunity to
address the problems associated with the current regulatory complexity,
which is criticized by managers and hunters alike (Babcock and Sparrowe
1989). The complexity of migratory bird hunting regulations, particularly
those designed to increase hunting opportunities for lightly-harvested
species, was probably an inevitable by-product of the increasing
information available about waterfowl populations and demands for higher
levels of consumptive use. Goals to simplify regulations and to increase
harvest opportunities on lightly-harvested species conflict. Another
consideration is whether species-specific regulations have distributed
harvest pressure in the desired manner. Recent analyses (e.g., Rexstad et
al. 1991) suggest that the capability to shift harvest pressure among
various stocks of migratory birds may be limited. Finally, the ability to
provide expanded harvest opportunity for any species is dependent on the
quality of resource monitoring programs and on the ability to target the
species of interest with management actions. Thus, it is unrealistic to
expect to maximize harvest opportunities for every species.
Alternative Models of Population Dynamics
Many waterfowl models have been developed for purposes other than
harvest management (Williams and Nichols 1990), but there is a dearth of
models that describe mortality and reproductive processes over the annual
cycle. In adaptive harvest management, models must not only describe the
effect of hunting (e.g., additive vs. compensatory hunting mortality), but
also the effect of the environment (which may include population
abundance) on changes in population size. In addition, candidate models
should meet 3 other criteria: (1) models must describe different harvest
strategies (or there is no value in learning which model is best); (2)
models must describe different responses to harvest that are detectable by
the monitoring program (or the process will fail to identify the most
appropriate model); and (3) models should be consistent with historical
experience (i.e., empirical).
It is important to recognize uncertainties, other than those associated
with biological mechanisms, that further complicate the regulation of
harvest. A lack of knowledge about underlying biological mechanisms (often
called structural uncertainty) is exacerbated by our inability to
precisely observe population status and trends, and our inability to
completely control harvest with regulations. A limited ability to
recognize status and trends is described generically by the term partial
observability, to emphasize the fact that (1) observability of a
population is tied to monitoring precision and (2) population status can
be measured only within the limits of that precision. Limitations in the
control of harvest are described by the term partial controllability,
which expresses the fact that regulations can be used to target actual
harvest rates (and harvest impacts) only as precisely as control limits
Partial observability, partial controllability, and a lack of knowledge
about biological mechanisms all contribute to the uncertainties waterfowl
managers face in their attempts to regulate harvests. When simultaneously
operative, their joint effects may well be greater than the sum of their
individual effects. For example, monitoring imprecision diminishes one's
ability to "see" the system, and thus impedes the rate of
learning about underlying biological structures. Similarly, an inability
to recognize and control harvest rates in the presence of structural
uncertainty slows learning rates, essentially because population dynamics
can be attributable to: (1) an assumed rate of harvest; (2) an assumed
relationship between harvest and biological processes; or (3) both
factors. Finally, monitoring imprecision in the presence of partial
controllability can effectively mask the effect of partial control and
further confound the recognition of biological structure.
Implementation of an actively adaptive harvest strategy for waterfowl
faces formidable obstacles. In addition to the technical difficulties
discussed above, there are a number of potential institutional
impediments. Adaptive harvest management requires an explicit admission of
ignorance or disagreement (i.e., uncertainty about biological mechanisms)
and involves the use of mathematical methods (i.e., stochastic
optimization) not easily explained (Walters 1986). Competing interests
will make it extremely difficult to reach agreement on explicit objectives
and constraints. Administrators and politicians often have difficulty
focusing on long-term objectives, and even short-term sacrifices of lost
hunting opportunity or impacted resources may be unacceptable. Ultimately,
success in adaptive harvest management will require, more than anything
else, an institutional framework that embraces patience, persistence, and
Despite potential pitfalls, we believe adaptive harvest management
offers considerable benefits from both technical and administrative
perspectives. Some of these advantages are: (1) the opportunity to resolve
long-standing controversies about the effects of hunting regulations,
while pursuing more traditional management goals; (2) increased
objectivity and integrity in the decision-making process; (3) a clearer
focus on long-term management (i.e., sustainability); (4) a better
understanding of harvest management objectives through identification of
difficult trade- offs; (5) a clearly-defined role of data-gathering
programs in the regulations process; (6) a stronger link between migratory
bird management and research; (7) explicit accounting for all sources of
uncertainty (environmental, structural, partial management control,
partial system observability); and (8) treatment of management as an
adaptive process, which is more appropriate for dynamic systems than a
static strategy (i.e., managers make adjustments to hunting regulations
based on changing resource status and their understanding of population
BABCOCK, K. M., AND R. D. SPARROWE. 1989. Balancing expectations with
reality in duck harvest management. Trans. North Am. Wildl. and Nat.
Resour. Conf. 54:594-599.
BLOHM, R. J. 1989. Introduction to harvest - Understanding surveys and
season setting. Proc. Inter. Waterfowl Symp. 6:118-133.
HILBORN, R., AND C. J. WALTERS. 1992. Quantitative fisheries stock
assessment: choice, dynamics & uncertainty. Routledge, Chapman &
Hall, Inc., New York, NY. 570pp.
HOLLING, C.S., EDITOR. 1978. Adaptive Environmental Assessment and
Management. John Wiley and Sons, London, UK. 363pp.
JOHNSON, F.A., B.K. WILLIAMS, J.D. NICHOLS, J.E. HINES, W.L. KENDALL,
G.W. SMITH, AND D.F. CAITHAMER. 1993. Developing an adaptive management
strategy for harvesting waterfowl in North America. Trans. North Am.
Wildl. Nat. Resour. Conf. 58:565-583.
LUBOW, B. 1993. Stochastic Dynamic Programming (SDP): User's Guide.
Version 1.04. Coop. Fish and Wildl. Research Unit, Colorado State
University. 119 pp.
LUBOW, B. 1995. SDP: Generalized software for solving stochastic dynamic
optimization problems. Wildl. Soc. Bull. 23:738-742.
NICHOLS, J. D., AND F. A. JOHNSON. 1989. Evaluation and experimentation
with duck management strategies. Trans. North Am. Wildl. Nat. Resour.
PATTERSON, J. H., AND R. D. SPARROWE. 1987. Background and objectives on
stabilized duck hunting regulations: Canadian and U.S. perspectives.
Trans. North Am. Wildl. Nat. Resour. Conf. 52:177-185.
REXSTAD, E. A., D. R. ANDERSON, AND K. P. BURNHAM. 1991. Lack of
sex-selectivity in harvest of mallards under the point system. J. Wildl.
TROST, R. E., D. E. SHARP, S. T. KELLEY, AND F. D. CASWELL. 1987. Duck
harvests and proximate factors influencing hunting activity and success
during the period of stabilized regulations. Trans. North Am. Wildl. Nat.
Resour. Conf. 52:216-232.
U.S. DEPARTMENT OF THE INTERIOR. 1988. Supplemental environmental impact
statement: Issuance of annual regulations permitting the sport hunting of
migratory birds. U.S. Fish and Wildl. Serv., Washington, D.C. 340pp.
U.S. DEPARTMENT OF THE INTERIOR AND ENVIRONMENT CANADA. 1986. North
American waterfowl management plan. U.S. Fish and Wildl. Serv.,
Washington, D.C. 19pp.
U.S. DEPARTMENT OF THE INTERIOR AND U.S. DEPARTMENT OF COMMERCE. 1993.
1991 National survey of fishing, hunting, and wildlife-associated
recreation. U.S. Govt. Print. Off., Washington, D.C. 124pp.
WALTERS, C.J. 1986. Adaptive management of renewable resources.
MacMillan Publ. Co., New York, NY. 374pp.
WALTERS, C.J. AND C.S. HOLLING. 1990. Large-scale management experiments
and learning by doing. Ecology 71:2060-2068.
WILLIAMS, B.K. 1988. MARKOV: a methodology for the solution of infinite
time horizon Markov decision processes. Applied Stochastic Models and Data
WILLIAMS, B.K. 1989. Review of dynamic optimization methods in renewable
natural resources management. Natural Resource Modeling 3:137-216.
WILLIAMS, B.K. AND J.D. NICHOLS. 1990. Modeling and the management of
migratory birds. Nat. Resour. Modeling 4:1-89.
WILLIAMS, B.K. 1996a. Adaptive optimization and the harvest of
biological populations. Mathematical Biosciences 136:1-20.
WILLIAMS, B.K. 1996b. Adaptive optimization of renewable natural
resources: solution algorithms and a computer program. Ecological