Adaptive management is an explicit and analytical process for adjusting management and research decisions to better achieve management objectives; and this process should be quantitative wherever feasible. Adaptive management recognizes that knowledge about natural resource systems is uncertain. Therefore, some management actions are best conducted as experiments in a continuing attempt to reduce the risk arising from that uncertainty. The aim of such experimentation is to find a way to achieve the objectives as quickly as possible while avoiding inadvertent mistakes that could lead to unsatisfactory results.

The concept of adaptive management is readily understood because it represents the common sense of "learning by doing". However, actually implementing adaptive management is neither simple nor intuitive. This complexity stems from the large number of interconnected potential scenarios, the related uncertainties, and the intricacy of necessary computations. Advanced technologies provide the decision support tools to help managers organize the relevant information, simplify the analysis of the scenarios, and assist in the search for optimal solutions.

Why the Current Attention On Decision Support Systems?

Adaptive management and decision support systems are gathering increased attention in natural resource management because three important trends seem to be changing the way managers and biologists must address resource issues. First and foremost, the stakes have gone up. Natural resource decisions increasingly are at the center of intense economic, political, legal, and value conflicts, as evidenced in the land management of many National Wildlife Refuges, National Forests, National Parks, and other public lands. Clearly, work with high profile endangered species and determinations of candidate species are other examples where the stakes are increasingly high, but many other less visible management actions are no less important nor complex. Second, the availability of certain kinds of data has exploded, and managers want to make the best use of that data and associated knowledge. Remote sensing, geographic positioning systems, and various monitoring technologies (datapods, coded wire tags, PIT tags, satellite telemetry) can generate previously unimaginable volumes of data, but often these data are somewhat indirect as indicators of the quantities and qualities that actually need to be known for management purposes. The third trend is that computer modeling is playing an increasing but sometimes controversial role. The complexity of the systems that need to be understood in an attempt to strike the correct balance in management decisions often necessitates computer modeling. However, the common phenomenon of "dueling models," where different constituencies present models that predict very different outcomes, has raised legitimate concerns about the reliability of models.

Rigorous decision analysis and scientific support can address all three trends by providing resource managers a set of tools to: (1.) quantify costs and benefits, and evaluate trade-offs; (2.) assimilate all the available data into the decision process; and (3.) assess the level of certainty in predictions, and take this into account in making decisions. Furthermore, the collaborative nature encourages broader buy-in by a variety of managers with different perspectives as well as involving research scientists. This facilitates a more participatory and cooperative atmosphere, in addition to helping managers arrive at better decisions.

The Reality of Tough Decisions

In many resource management situations, it is not uncommon for a decision making conflict to exist. On the one hand, we might be willing to take a chance that our knowledge about the system is "good enough". On this basis, we take an educated guess about the best choice for immediate and substantial progress towards reaching objectives. If the system is fairly static and we know a lot about how it functions, clear management choices may indeed exist. On the other hand, we often find ourselves working in situations where predicting future conditions is difficult at best, or where information about the system is not well documented. A flexible process of engaging in management experiments could be followed to increase our knowledge about the system as time unfolds; and in the long run, our management will thereby become even more effective.

Typically, options about where to fund additional research projects are determined based on subjective assessments. Using more rigorous decision analysis, the benefit of acquiring missing or replacing poor information can be simulated, thus determining how it might change the uncertainty associated with particular management actions. Using such analytical logic, the acquisition of new data and knowledge becomes an objective rather than subjective process. It avoids keying in on simply what appears to be the weakest link in the knowledge chain, and instead identifies the link that will increase the adequacy of management decisions the most. This is often defined as maximizing the probability of achieving objectives while minimizing the risk of an undesirable outcome.

Benefits of New Approaches

Innovative approaches are critical because the exhaustive evaluation of every combination of alternatives, and tracking the associated logic, may be impractical even with current computer technology. Modern procedures and search algorithms can reduce dimensionality, structure knowledge efficiently, and use probabilistic strategies to find completely satisfactory solutions, recognizing that "the best" solution may be unrealistic to ascertain. Where the number of potential decisions and associated outcomes is very large, decision support might be thought of as simply optimizing the course of action. In reality, however, the situation is more akin to that of winnowing the nearly incalculable number of possibilities to several, and subsequently suggesting one. And, potential courses of action can be evaluated on the risks associated with failure as well as probabilities of success. The usefulness of the recommendation is enhanced by including an explanation of the logical process of how the decision was reached.

Several of the direct benefits of approaching resource management and research from this perspective accrue from an enhanced ability to analyze and compare management scenarios. Through this analysis of alternative plans, the probability of short term catastrophe can be minimized, and the opportunities for long term success can be maximized. An added benefit of this type of framework is that justification of decisions becomes more straightforward and quantitative. Adaptive management integrates the setting of quantitative objectives, exploration of alternative management strategies, monitoring of progress, and evaluation of performance in terms of risks and benefits. Managers must still make the actual decisions on the ground, but decision support systems allow them to do so with greater confidence that the decisions are based on all the currently available knowledge, and that the decisions take correct account of the consequences of uncertainty.

When Should Adaptive Management Be Considered?

Adaptive management is most effective when situations have the following characteristics:

• The stakes are high: mistakes can be expensive, but rewards from determining optimum management would also be high.
  
  
• There is consensus about management goals, and how to recognize success. It must also be possible to objectively monitor progress towards reaching objectives.
  
  
• A spectrum of plausible management choices exists, and there is strong probability that at least some will work.
  
  
• Uncertainty in the system is considerable, complicating any attempt to make clear management choices by informal methods. And, there is reasonable promise that uncertainty would significantly be reduced (i.e., major decisions would be more straightforward) with well designed experiments.
  

Four capabilities are provided in effectively applying adaptive management and decision support in complex ecological systems.

• First, a computerized framework will represent vast amounts of knowledge and data.
  
  
• Second, the uncertainty associated with this knowledge and data is delineated and used to benefit decision making.
  
  
• Third, there must be the ability to examine a very large number of alternative, potential management scenarios.
  
  
• Last, complex resource management systems along with their logic and the degrees of scientific confidence associated with recommendations must be visualized and communicated to the actual decision makers.
  

These four capabilities can be handled in many ways. Our emphasis is on Bayesian analysis, data visualization, and various artificial intelligence techniques such as expert systems.