Mark Twery
USDA Forest Service
Northeastern Forest Experiment Station
Burlington, Vermont
J. Michael Scott
Idaho Cooperative Fish and Wildlife Research Unit
College of Forestry, Wildlife, and Range Sciences
University of Idaho
Moscow, Idaho
Timothy Cohler (we are unsure of his involvement at this
time)
Santa Fe Institute
Washington State University
Pullman, Washington
1. The perspective and process described here will be
refined;
2. The specifics about ecosystem stewardship decision
support systems will be incorporated.
Management decisions will always be made by people in authority based on their perceptions of consequences of their decisions. Decisions on public values are made by the people or their representatives; decisions on private values are made by private owners, investigators, or their delegated employees. The question of "Who has authority to make decisions?" is the first, most fundamental, and often overlooked question in decisionmaking, as discussed later.
Broad decisions are often made at high levels in an organization or social institution and are referred to as "policies." Decisions made at local levels (often a detailing and expanding of policy decisions) are referred to as "management decisions." A continuum exists between the two levels since policies made at high levels can be interpreted different ways and produce different tradeoffs of values. The decision process and decision support structure needs are similar for both. Decisions at the local level may be more important because this level is where broad policies actually get implemented; however, these decisions must be consistent with broad policies to avoid inconsistent actions among regions and actions at different levels which contradict, counteract, and/or nullify each other.
For this paper, the person(s) with authority to make the decision(s) will be referred to as the "decision makers" even though they may be referred to as "decision makers" or "managers" at different organizational levels. Those people who have a legitimate interest in the decision(s) will be referred to as "stakeholders"; in representative democratic organizations, "stakeholders" are often represented by "decision makers." When a decision is made without a "stakeholder" being part of the decision, those making the decision "de facto" decide that the excluded stakeholder did not have legitimate authority.
Decisions will be successful where the actual consequences of the decisions meet expectations--or where a different, but equally satisfying result fortuitously occurs. Except for the occasional fortuitous result, the decisions will be most successful where the perceptions of the decision makers match reality. Two things are necessary for the decision makers to make good decisions:
1. there must be some understanding of the behavior of the
thing(s) to be managed--both what is known and what is not
known;
2. the understanding must be accessible to the decision
makers in formats which allow the decision makers to
understand the problem, issues, and tradeoffs; to weigh
outcomes of alternative actions relative to a chosen set of
objectives; and to make appropriate decision.
The process of gaining this understanding and making it accessible to decision makers has been emerging as an organized discipline for about 50 years under the various names of "policy analysis" and "decision theory." Decision support systems are specific tools and methodologies to aid decision makers in making decisions. This paper describes decision support systems helpful for Forest Ecosystem Management. To do this, it first gives perspectives on the understanding of systems and the ways knowledge becomes understandable. Then, it describes a general analytical approach to integrating scientific knowledge into the decision-making process. Finally, it describes ways to build upon experience in the analytical approach to improve the ability to make decisions.
Despite the scientific progress, many management actions have produced unforseen results or indirect consequences which were not desirable. Recognition and study of these complex interactions and the formal study of multiple interactions and consequences developed under the name "systems approach". The "systems" approach to analysis is now the basis for many fields of study from economics to ecology.
The basis for the systems approach is that many activities influence each other directly or indirectly (Figure 1). Consequently, when a change occurs in one activity, the result is a variety of influences on many proximally and distally related activities. (The "activities" discussed can be defined at different scales by the analyzer, as shall be discussed.) For systems where only a few activities occur and the interactions are known, behaviors can be largely calculated, explained, and predicted by a few simple relationships. For example, the behaviors of electrons, protons, and neutrons in the hydrogen and helium atoms can be calculated and predicted, since these contain only one or two of each particle. When the number of activities become more numerous, calculation and prediction becomes extremely complex even if the interactions remain known and simple. For example, predicting the behaviors of electrons, protons, and neutrons in more complex atoms and molecules becomes so complicated that even sophisticated computers have difficulty with them.
The question of whether there is a single direction of all future events which could be foretold if one had the capacity to understand and predict the behavior of all interacting "activities" has been debated for over one hundred years. This question becomes important to "ecosystem management" if one assumes that systems have a single "direction" which is interrupted by human intervention. The preponderance of evidence is that there is no single direction--or "natural" behavior--which is interrupted by human behavior. (Since we can not calculate the "single direction" if it did exist, the question becomes academic; and further discussion of this philosophical point is beyond the scope of this paper.)
The human brain is capable of consciously keeping track of less than about ten interactions at one time; consequently, people often have difficulty predicting the behaviors of complex systems. Even the computing power of sophisticated computers become dwarfed by the many interactions of complex systems. The human brain and scientific analyses overcome the problem of complex interactions by grouping "activities" together and generalizing about their aggregate behavior. These "grouped activities" can be referred to as "modules" and the interactions among "modules" can be referred to as "inputs" and "outputs" (Figure 2).
The "modules" can be designed in many ways, depending on convention and the objectives of the analysis. Many cultures, "schools of thought", and other social subsets recognize different "natural groupings" which can be considered the "modules." Even when using the same names, different subsets of people may not bound the "modules" in the same way; and confusion can occur where different people do not realize their perceptions of "flows" in systems are based on different definitions of modules and systems. The objectives of a particular analysis often dictate the "modules", with the proximate concerns to the analysis being split into smaller modules and more peripheral concerns being grouped together into larger modules. The number of modules considered at any time should be quite small to allow people to understand the flows among them. Problems can occur when an important module is ignored instead of included (even in grouped with another) in order to keep the analysis to a manageable size.
Nearly any module, delineated at any level, can be subdivided and the flows within it studied. For example, it may be appropriate to define "ecosystem" modules as the "biodeversity component", the "forest component", the "soil component", or others (e.g., Figure 4). Consequently, anything can be considered infinitely complex, not completely understood, and uncertain. When studying a system for any objective, it is important to create only a few modules and develop simple models of "inputs and outputs" within each module.
"Systems analysis" is the study of these "flows"-- inputs and outputs--among "modules." To study the flows among modules, it is convenient to allow the receiving ("customer") module to define the input it "receives"--the form of its input and amount of acceptable error. Each module will also act as a "server" by providing output in a form and amount targeted by its "customer(s)." The degree to which the targeted ouput(s) is achieved is the measure to which the module achieves its objective(s). Often a "base case" behavior within each module is used when developing the flows among modules.
The behavior of each module and the flows among modules often differ from what people expect. There can be time lags, important negative and positive feedbacks, redundancies, and compensating behaviors which create different results than people expect when considering only linear "cause-and-effect" relationships. Careful development of models of module bahavior and analysis of flows among modules is necessary to understand the behavior of systems. Any model of module behavior is necessarily a generalization of behaviors of submodules; consequently, its output contains variation. Careful study of the processes occurring within the module, as well as techniques such as "statistical process control", allow the primary causes of variation to be accounted for in analyses. Systems analysis often gives more rapid understanding of each module than could be gained by examining each module alone because the analysis of interacting modules shows the range of inputs and outputs within which a subject module can function without creating a conflict (contradiction) in flows among modules. A conflict in flows among modules means there is a misunderstanding within one or more modules or in the flow patterns among modules.
While a complete understanding of each module may be impossible, a partial understanding often contributes to better decision making than ignorance (or delaying decisions, which is itself a decision). Various techniques have been developed to show the relative certainty with which a particular outcome can be expected and to determine which "bottlenecks" of ignorance need to be resolved to make understanding most complete. Sometimes, the understanding of processes changes so completely that the "flows" into and out of a module (or among several modules) need to be completely altered--instead of refined, as described above. These dramatic changes are referred to as "paradigm shifts" and lead to changes in the flows among many modules which must be reanalyzed.
Treating biological interactions as "systems"-- "ecosystems"--was first proposed in the 1930's as a way of avoiding becoming bogged down in the complexity of ecological systems. Various interpretations of this "ecosystem" approach have since been made. Three areas of concern have arisen from this "ecosystem" approach:
1. Early study of systems sometimes stated that systems were not random, but needed a "direction" and led some ecologists to assume the "direction" of ecosystems was toward a stable, undisturbed system--the "climax." (The concept of a stable, climax condition is currently not the most accepted ecological paradigm; its origin can be traced to the nineteenth century Romantic concept of nature.)
2. Ecosystems can be bounded differently by scientists both spatially and temporally at different scales, depending on the purposes of the scientists. Consequently, the "ecosystem" could be simultaneously and correctly interpreted spatially as a single leaf, a large forest area, or the whole earth. Temporally, an ecosystem can be bounded as a single instant or a long time scale. When "old growth forests" are defined as ecosystems, the ecosystems are bounded in time, since such a forest exists in a given condition at one time and place.
3. Ecosystem is an extremely useful scientific concept but is confusing as a management or legal term, since the "modules" and systems are redefined depending on the purposes of the study. Management and legal directives to "protect the integrity of ecosytems" can be confusing because of #2, above.
To avoid this "folly", understanding of alternative courses of action must be presented to decision makers in ways which are both understandable and non-prejudicial.
Understanding begins with precise communication. Both the analysts and the decision makers need to agree on a common grouping of "modules" and flows. Similary, words are often interpreted differently by people of different background; and care is needed to ensure common understandings of word meanings. Often it is expedient to delete words which mean different things to different people and to create new words to avoid confusing or polarizing an issue.
The decision maker needs to have an understanding of the perceived problem, the perceived "goal state", and the impacts of different alternative courses of action. The understanding must be at the appropriate level of complexity (Figure 3). This level generally involves understanding the "flows" among a limited number of "modules." Attempts to make decisions at extremely detailed levels can lead to the inefficiencies of central planning. Understanding of the flows among "modules" is an education process for the decision maker, and various education aids can help-- graphics, films, "games", and other tools which promote an understanding of the behavior of the modules. Of course, many of these education aids can also promote disinformation.
At times, the perceived problem of a decision maker or the public may be only a perception, or the perceived goal state may be unachievable or may have many undesirable consequences. The problem may also be caused by conflicting objectives of the stakeholders or conflicting opinions of who are the stakeholders--people who have a legitimate interest in the decision(s). Such conditions often are generated in "crisis" atmospheres ("chicken-little" atmospheres). At such times, the primary course of action is to educate the decision maker and/or the public--so they will understand the conditions.
Analysts and scientists can model and display outcomes of alternative actions, but can not resolve differences in public values. Decision makers (as well as analysts, scientists, and others working with systems) maintain an understanding at a level of flow among relatively few groupings (modules). They need confidence that the inputs and outputs within each module are robust. This confidence is provided by experts who specialize in subdividing each module and treating these subdivisions as systems. The scientific process, including peer review, the scientific method, quantitative analyses, logic, and data collection-- are used to ensure the accuracy of these experts at all levels. In addition, the "feedback" nature of flows among modules and discrepancies between the expected and actual outcome can help decision makers detect false understanding within modules. Even so, the scientific process is not perfect and mistakes are made.
Because it is unlikely that human nature will change from being subject to prejudice, analytical processes have been developed which attempt to minimize the influences of these prejudices on scientific understanding and the decision making process. These analytical methods attempt to reduce the subjectivity in scientific research.
In policy decisions, both objective analyses and subjective decisions are needed; however, the subjective decisions should be made by decision makers with an understanding of the the probable outcomes of these decisions. It is important for decision support systems to provide objective inputs. Consequently, objective, analytical, systematic methods are followed similar to those followed for scientific research.
Normative/Rational Method--Non-iterative:
This method
assumes a "perfect" condition, in which the objectives and
relative weights (values) assigned to the various objectives
are well understood. In addition, the modules and links
(flows) among modules are well understood; consequently,
the outcome of various actions can be predicted with some
degree of certainty. Even where uncertainty exists, the
uncertainty is assumed to be quantified. The problem
becomes a simple one of determining the best way to satisfy
the most objectives based on their weight. Under these
conditions, various optimization models can lead to the best
solution.
Normative/Rational Method--Iterative:
This method occurs
where the problem, objectives, and relative weights of
different objectives are not well understood by the decision
makers. It does assume that the the modules and links
(flows) among modules are well understood; consequently,
the outcome of various actions can be predicted with some
degree of certainty. Even where uncertainty exists, the
uncertainty is assumed to be quantified. Various
alternative courses of action are proposed and their
consequences are described to decision makers. Through
updating the decision makers and allowing them to understand
the outcomes of various courses of action, they iteratively
develop a course of action which best meets their
objectives.
Bounded Rationality Method:
This method is similar to the
above method; however, instead of arriving at a solution
which is known to be the best for the set of objectives,
this method accepts the first alternative solution found
which minimally meets all objectives. The solution may or
may not be the best possible.
Garbage Can method:
This method often occurs where the
attempt is to develop a "Normal/Rational (usually iterative)
Method." Instead, it mixes and matches various elements of
decision tools and stops as soon as an adequate solution is
reached. Because of time, budget, personality, and
knowledge constraints, the ways of developing alternatives
and ways of comparing alternatives to objectives are not
completely analytical. Consequently, an approximation of an
analytical, objective process is attempted even though it
may be insufficient to support predictable outcomes for
decision makers.
Expert/Intuitive method:
Using this method, judgements are
reached either by a single person acting on his/her own best
judgement or by an assembled group of experts. These
judgements are "reached by an informal and unstructured mode
of reasoning without the use of analytical methods or
deliberate calculations"--a "seat of the pants" method.
This method assumes intelligent people will make wise
decisions, but does little to ensure experts restrict their
knowledge to their area of expertise, that all needed areas
("modules") of expertise are represented, and that a
systematic flow of information among modules (areas of
expertise) is maintained. The result is often the
prejudiced solutions based on charisma and "groupthink."
Muddling through:
In this process, decisions are made based
on expediency--whatever will provide short-term relief from
pressures (often known as "hip-shooting"). Any undesired
outcomes of the solutions are addressed as they arise, on a
case-by-case basis. Where resources are plentiful enough for
alternative choices not to affect people markedly, decisions
are often effective when made on the basis of "muddling
through."
Objective analyses strive for the "Normative/Rational" method; however, various factors can cause all or modular parts of the analyses to have varying degrees of the other methods. Even when a "normative/rational" method is followed perfectly, the result can be unsuccessful if there is a "paradigm shift" in one or more of the "modules" fundamental to the decision-making. In this case, the resulting decision has the same effect as if the decision had been made by "muddling through."
It is important to recognize that following objective, systematic, analytical procedures or using well recognized experts will not mean that the decision-making process is successful. The process will only be successful if the expected outcomes of the decision coincide with actual outcomes and all important factors have been included. The value of objective procedures and knowledgeable people is to ensure that the outcomes are as expected to best ability possible.
Human nature ensures biases such as those described above will always be present in the analyses. Recognizing this does not negate the attempt to ensure unbiased analyses (any more than recognizing no will be a perfect parent negates attempts to be a as good as possible). Since biases will always enter, it is important for the decision maker to understand the possible sources of bias in a particular process as well as the possible influences this bias may have on decision-making. Biases can be minimized both by keeping multiple disciplines involved and by having people experienced in systems analysis to monitor the linkages (inputs and outputs) among the modules.
1. It has been developed by many people over a period of
time. Consequently, it is free of the prejudices of
expediency or a single person's opinion;
2. Since the relation of objectives to alternative decision
choices are explicitly stated, it is relatively free of
hidden motives of decision makers or policy analysts
(experts and scientists);
3. The systematic procedures allows the process to be
studied and improved as the results of the decisions become
apparent;
4. The systematic, modular procedure allows many experts to
contribute according to their expertise;
5. The procedure can be repeated relatively easily as new
information or objectives are developed;
6. The procedure is based on objective assessments of
flows, rather than on subjective opinions, personalities, or
"groupthink" consequences.
In ecosystem management, the decision of "What is in the public's authority--and how are public values defined?" and "What is in the private authority?" needs to be determined before objectives can be determined. Different members of the public can have different objectives. For example, those closest to a National Forest may have different objectives relative to the forest that those far away, and it is unclear if the objectives are made by majority rule or protection of "rights" of the minority.
If decision makers had complete knowledge of the problem, they would define the problem and set the objectives. Often, however, the decision maker only perceives that the present condition ("problem state") is different than the expected condition ("goal state"); and the first role of the analyst is to refine this perception into a statement of specific problems and objectives.
A systems approach can be used by aggregating the various influences into a limited number of "modules" showing "flows" (e.g., Figure 4). The present condition of the various modules and the influences of changes in any module can be analyzed using simple "input-output" models to understand and convey to the decision maker the effect of changes in courses of action on various outputs from modules. The decision maker then chooses which inputs and outputs are of concern; achieving desired conditions of these "targeted" inputs and outputs become the objectives.
a. It is first important to determine if a problem really exists. At times the problem may be one of perception, and analysis can show that the solution is one of educating the decision maker (and the public) rather than changing a course of action.
b. It is important to understand and address the problem at the appropriate scale to avoid being overwhelmed by the complexity of the problem and resorting to overly centralized planning (micromanagement) which actually stifles management. Restricting the targeted inputs and outputs in any system to less than ten can avoid these problems. More detail within or across these modules should be handled by a lower level in a management heirarchy.
c. The criteria for making the decision should also be
refined by interactions between the decision maker and the
analyst. These criteria can be refined and expanded, but
can be roughly categorized into three basic approaches:
e. Once the legitimate stakeholders, problem, objectives, and policy strategies are agreed upon, criteria need to be developed for measuring the degree to which each objective is met. These criteria need to be analytically measurable, even if somewhat subjectively, so it can be objectively determine how well the objective is met. In ecosystem management, "change with time" needs to be incorporated into the measurable criteria. For example, "maintaining an average proportion of key habitat of 15 percent" will not be effective if the average is achieved by the proportion fluctuating over time between 0 and 50 percent, since the target species would become extinct during times of 0 habitat. The spatial scale of the the "ecosystem" also needs to be explicitly stated as a measurable criteria.
f. Then, one or more models needs to be developed to explain the relation of the objective to various courses of action. These models will be described in more detail under Step # 3. Developing criteria to measure how well objectives are met allows objective, analytical models to determine the degree to which each objective is achieved by each alternative management approach.
Various techniques and tools have been identified for helping determine and refine the objectives, including: scoping, cause and effect diagrams, stakeholder analyses, network diagrams, decision paths, decision frames, strategy tables, influence diagrams, decision heirarchies, objectives heirarchies, value models, tornado diagrams, graphics, other visualizations, computerized games and models, films, and pictures. The techniques and tools are used to give the decision maker and analyst an understanding of the problem. Like all tools, they can also be misused; and peer reviews or other objective processes are needed to ensure the tools are accurate and objective.
Eventually, the alternatives need to be combined to a manageable number of discrete, distinctly different alternatives. The combining needs to ensure that the breadth and creativity of the range of alternatives is maintained. Where several decisions are needed in the management, each decision can have several alternatives; and techniques have been developed for "mixing and matching" alternatives for each each decision to reach an overall alternative.
It is important to separate the roles of those suggesting alternatives from those of combining alternatives and those of analyzing the consequences of various alternatives. Those suggesting alternatives generally develop a bias in favor of their alternative which carries over to the treatment of other alternatives.
In addition to a robust, broad range of active management alternatives, the range of alternatives needs to be "bounded" by several specific alternatives:
a. A "do-nothing" alternative. "Do-nothing" has two
interpretations--ceasing all active management activities
and making no change in present management practices.
Ideally, both alternatives should be analyzed.
b. Alternatives which are believed to "maximize" each of
the objectives individually. Development of these
alternatives "bound" the problem on the extreme cases.
Often when the consequences of courses of action advocated
by single-issue groups are analyzed, the consequences seem
less desirable and support for the extreme positions
dissipates.
Methods of objectively analyzing the relation of the action to the the objective can be developed which are robust enough to apply to all alternatives. These methods often consist of models or series of models and other forms of analyses linked in a modular fashion. Many models of stand growth, habitat suitability, economic consequences, and other interactions have been developed in forestry. These can be used and linked to varying degrees to determine various relations. Many of these models were not specifically developed for policy analysis, and their use needs to be tempered with awareness (analytical or subjective) of the limitations and degree of accuracy of these models. Often, sensitivity, risk, and certainty analyses can help the analyst and decision maker. Use of "comparative risk profiles" and "probability distributions" is also helpful. Through time, the various models can be improved through adaptive management (discussed later and in other papers). Adaptive management will be most effective where process-based models are used.
In cases where it is necessary to substitute expert opinion for some "modules", techniques have been developed which allow expert opinion to be used objectively and analytically. The models and other methods should be as objective as possible, and peer review and other techniques should be used to ensure the objectivity.
At times of shifting paradigms, experts may not agree on the appropriate relation among actions and objectives. In these cases, separate models incorporating each opinion may be used and the influences on objectives can be compared. Where the differences in opinions have very strong influences on achieving the objectives, the more energy can be devoted to determining the correct relation.
It is important that techniques for analyzing the various alternatives be developed separately from (either beforehand or by a separate group [preferrably both]) the proposing of alternatives to avoid the possible sources of bias discussed above.
Linked models and other analyses often become complex and difficult to understand. Visualization and other systems which allow the decision maker to synthesize a large amount of information into a picture or similar graphic are extremely helpful in giving understanding to the decision makers and analysts.
Often analysis requires collection of data on the area of concern. It is important that the models and objectives be developed BEFORE the data is collected. Otherwise, the models and objectives are tailored to the data, rather than the data being tailored to the objectives and models.
First, the decision maker should have an independent critique of the decision support process. This critique will alert the decision maker of weaknesses caused by lapses into unobjective behaviors (e.g., "garbage-can methods" described above). Since such weaknesses will always occur, their presence does not nullify the analysis; however, awareness of these allows the decision maker to consider these weaknesses when studying the analysis.
Various tools have been developed which help the decision maker make appropriate decisions: the tools include linear programs, complex models with assigned weights, matrixes, decision trees, simulation models, comparative risk profiles, and probability distributions. Especially useful for allowing the decision maker to understand the relation of objectives, alternatives, and tradeoffs is the matrix approach; however, each tool is useful under certain conditions and can be described in more detail. The "analytical heirarchy process" is useful in helping decision makers assign relative weights to the various objectives.
At no time should the decision maker feel subservient to the policy analysis; however, the decision maker should respect the analyst's forecast of the consequences of each alternative.
Decisions at broad policy scales are made by studying flows among broad modules. At lower organizational levels, systems become more specific with more site-specific flows among more specific (but not more numerous) modules. At very low levels in organizations, decisions are often delegated to professionals. At this time the professional should recognize that he/she changes from a professional to a policy role when he/she begins making decisions. This change is common in organizations, but needs to be recognized. It is also important to realize that an individual advocating a policy position looses his/her objectivity--and therefore his/her effectiveness--for doing policy analysis.
The first action in implementing a decision is to ensure everyone involved in implementing the decision understands the decision and their changes in responsibilities and actions.
The more closely coordinated the analyst and the manager implementing the plan, the better the plan AND its implementation. Certain organizational structures--quality circles and other modern management techniques--integrate the planning and management very effectively.
Figure 1.
A "systems approach" recognizes many "activities"
interact (A), influencing the behavior of other
"activities." "Activities" can be the behavior of anything
from subatomic particles to planets or galaxies, depending
on the perspective of the analyzer.
Figure 1B.
An action taken (#1) does not lead to a simple
consequence; it puts in motion a series of interactions
among "activities" and leads to consequences which are not
always intuitively obvious because of "feedback" (e.g., #2),
"lag times" (e.g., #3), and other behaviors.
Figure 2A.
The human mind can only consciously keep track
of a few (less than 10) activities at one time. To study
the effects of different actions but avoid becoming
overwhelmed by complexity, "activities" can be "grouped"
into "modules." The detailed pattern of grouping of
"activities" is dictated by the objectives of the particular
analyses.
Figure 2B.
Different cultures, "schools of thought", and
other social subsets recognize different "natural groupings"
as modules (Compare A & B). Confusion often occurs when the
same name is used for different groupings by different
people. A socialization (communication) process is
necessary to ensure the analysts and decision makers are all
describing the same "modules."
Figure 3A.
The "flows" or "linkages" between "modules" can
also be grouped into single or a few "inputs" or "outputs."
The form of the "flow" is dictated by the recipient
("customer") module, while the amount is dictated by the
"server" module. The flows and groupings of each module
often change with time. Alternative policy and management
actions alter the behavior within one or more module, and so
affect the "inputs" and "outputs" other modules. The analyst
often uses systems to help decision makers realize and
understand what can be affected by a management decision,
and how significantly. The "inputs" and "outputs" most
significantly affected can become the concerns of
management, with targeted values of these "inputs" and
"outputs" being the management objectives. As a system
becomes extremely complex, there is a tendency to ignore
certain inputs and outputs, often leading to very unexpected
consequences. It is more effective to combine modules and
flows than to ignore some.
Figure 3B.
Models and experts are often used to determine
the behavior (flows in and out) of each "module." If
subdivided, each module can become infinitely complex, and
the models and experts necessarily generalize very complex
behaviors. Often "base case" analysis and "certainty" and
"sensitivity analyses" are performed to determine how
disagreements, variations, and uncertainty of behavior
within modules affect the flows to other modules. Poor
decisions can be made if the projected outcomes of actions
are incorrect because flows among modules are incorrect.
Incorrect flows can be caused by minor errors or by gross
misunderstandings (paradigm shifts) which require a
overhauling of relations among modules. Research, including
"adaptive management" and "continuous quality improvement"
can help improve the understanding of relations within and
among module.
Figure 4.
Systems can be defined in different ways,
depending on the objectives of the analyst. Use of systems
in policy (e.g., ecosystem management) requires that the
level of management be considered. One example is a global
approach shown here. At the national scale, the "forest"
module affects (and is affected by) soils, biota (all
species), climate, people, commodities, and the global
environment. The objectives of management could be
considered certain conditions of these factors. Once an
overall policy is agreed upon, more regional approaches
would analyze similar "modules" to determine the appropriate
behavior of each region to meet these goals. A similar
process would occur at increasingly local, site-specific
levels.