untitled

One unfortunate aspect of this terminology is that “active” seems a great deal more

attractive than “passive” adaptive management. It is not always clear that this is so.

Experimentation implies consciously ignoring the current “best” model to explore

alternatives. This can only come at a cost, perhaps a great cost, if the alternative

models prove to be inferior. For example, Anderson (1975) suggested that higher

harvests of mallard ducks could be sustained if the breeding stock is allowed to climb

considerably above the current levels. Choosing to explore this option would require

considerable reduction in harvest quotas for hunting enthusiasts and resultant loss

of revenue to agencies, tourist operators, and equipment suppliers. If Anderson’s hypoth-

esis is incorrect, the cost of learning would be a reduction in harvests (and profits).

Small wonder that this may not be an attractive option for everyone concerned! One

way to evaluate the wisdom of embarking on an active adaptive experiment is to eval-

uate the costs and benefits of sticking with the untested status quo hypothesis ver-

sus adopting an experiment to test an alternative hypothesis. This can be symbolized

by a decision matrix (Table 15.2).

For example, let us say that the current stocking rate of mallard ducks really is

most productive. If we increase the stock to a less productive level, this would lead

to a reduction in average harvest from 1.0 to 0.5 units. On the other hand, if

mallards really are more productive at a higher stock size, then increasing the stock

may increase yields by a given amount, say 20%. If each model is equally plausible,

based on our current data, then we can calculate the expected payoff by simply

averaging the possible outcomes across each row.

In this example, the expected payoff is 1.0 for the status quo option (=[1.0 +1.0]/ 2),

whereas the experimental option has an expected payoff of 0.85 (=[0.5 +1.2]/ 2).

This suggests that the experiment is too costly and /or too unprofitable to warrant

testing. In contrast, if the payoff in the lower right-hand cell of the matrix was 1.75,

then the expected payoff would be 1.125 and the experiment would be justifiable.

The point is that minor improvements in yield, accompanied by major costs, may

not justify experimentation.

Even when the decision to adopt an experimental procedure is justifiable, many

individuals may not value slight improvements in management efficiency if it inter-

feres with their personal recreational values. That situation may explain why moose

or trout populations close to population centers in North America are probably well

below the level of maximal productivity. Alternatively, some resource users would

not want to take any risk of losing income, simply because there are no other

economic options. Hence, the payoff matrix will often vary among different special

interest groups.

While passive adaptive management might be attractive to risk-averse decision-

makers, it also has its perils. For one thing, it can lead to lost opportunity for all

MODEL EVALUATION AND ADAPTIVE MANAGEMENT 265

Alternative models (hypotheses)

Policy options Current stock is optimal Higher stock is optimal

Maintain status quo 1.0 1.0

Increase stock 0.5 1.2

After Walters and Holling (1990).

Table 15.2Hypothetical
matrix of net benefit
from two alternative
stocking models,
showing that the best
harvest of mallard ducks
is obtained from current
stock values, versus
higher stock values.