by
Gene Namkoong and Mathew P.
Koshy,
Department of Forest Sciences, University of British
Columbia,Vancouver, BC V6T 1Z4, Canada
Email: [email protected]
Is there a rational way to make mistakes?
We obviously don't intend to make mistakes but we also have to admit that we are fallible humans and that we have little time, funds, and personnel with which to execute a conservation program. We also generally lack precise information on which of the genetic resources that reside in populations and species are most likely to suffer irreparable loss and we can't predict with accuracy, how effective our efforts to conserve will be. Therefore, we will not always select the most efficient course of action to save the most valuable of the genetic resources with the management resources we have available. There can be no question that we will fail to be most efficient, and no question but that we will make mistakes in allocating our efforts. But is there a way to direct our allocation of effort, before we have complete information, so that we can minimize the expected effects of the mistakes we do make?
If we first admit to ourselves as well as to our supervisors that we will make mistakes, we can begin to identify the kinds of mistakes we are likely to make. Among the kinds of errors we will make will be those in which we try to save a certain genetic resource that does not need our efforts, would have survived anyway, and thereby waste our own precious time, effort, and funds. We will also make mistakes wherein we will fail to make efforts where we might be able to avoid loss of a valuable population or species. We will put large efforts into saving one population or species and not put effort on a more valuable resource. Obviously, it is not our intention to do so, but because we are ignorant of what is more at risk than another, and what values are exposed to risk, it is inevitable that errors of omission and of commission will be made. We must therefore admit that some resources will be saved regardless of whether we act or not, and some will be lost, also regardless of our actions. But we might be able to put effort where it counts. The question is how can we use whatever information we do have to best advantage.
The second admission we must make to ourselves is that we are usually not entirely ignorant of the risks and values that are involved and that we may be able to minimize the chances of making costly mistakes. We have choices to make, and we are obligated to using the information we have, as well as we can, to use our time, effort, and funds, to manage our risks. To this extent, we can answer the question posed above in the affirmative. Yes, there are rational ways to approach decision making so that the anticipated costs of making mistakes are as small as we can make them with the information we have at hand or can obtain.
RISK
One approach is to first understand what we
feel about the nature of risk to genetic resources. As biologists or
geneticists, we often think about how species or populations are at risk
of extinction or of substantial change and reduction such that their
conservation (in situ or ex situ) is warranted. We may also think about
the safety and availability of genes for future contingencies and what the
sizes and location of populations need safeguarding. Since genetic
resources are subject to the forces of evolution; namely mutation,
migration, drift, and selection, we have to be concerned about how those
forces may affect regeneration and the future condition of the resource.
For these reasons, it is useful to have studies of the structure of the
genetic resource to better target conservation efforts. We can often
estimate what the population size is and perhaps can account for uneven
reproductive success, by estimating the type and distribution of mating
behaviour and seedling success. If we have information on pollen and seed
vectors, and on changes in environmental selective forces, we can also
estimate the effects of selection. Good biological studies reduce the
errors of estimating the susceptibility of a resource, but can never
totally eliminate error.
Even in the absence of studies, we can still obtain indirect information and can base estimates of susceptibility of the resource to different threats. This kind of information is always imperfect but even without field tests, or marker studies, we usually have some information about how the resource may be endangered by changing forest conditions. This kind of information about the resource tells us what its susceptibility is.
In addition to considering susceptibility, some genes, populations, or species may be more at risk than are others not only because their inherent properties dispose them to problems, but because environmental threats are more likely to expose their vulnerabilities. For example, large fires may be likely to occur or clear cutting may be expected, and would be a threat to species susceptible to such forms of removal. For species that are dependent on vegetative cover for regeneration or that are sensitive to high insolation, that threat may greatly increase the likelihood of population loss in that area. However, for some pioneer species with a seed bank or with a large immigrant pool, the threat may not lead to population loss or reduced regeneration. For some species with many sub-populations, the threat to a small centrally located one that can be easily re-colonized may not be serious to the species, but the same threat may be serious to one located in an isolated and peripheral site.
Threats to forests may be of several types, including grazing, over-harvesting non-timber forest products, selective excessive logging, as well as the more obvious types of large-scale destructive events. For each of these, the probability of occurrence and the probability that each of the threatened genes, populations, or species, would be significantly susceptible to the threats must be included in considerations of risk. In this terminology, the concept of risk combines both susceptibility and threat where susceptibility can be estimated from knowledge of the distribution of genes in populations or species, and their life histories, and threat can be estimated by expected forest practices.
As in all biological problems, these risk estimates are only estimates, and the probability that the risk is exactly as we may guess, is not actually a fixed number. There is finite probability that the actual risk is higher or lower for one resource vs. another, and especially in cases where the genetics of the resource is not well known, we may want to be more conservative in rating risk than if we had more information. The value of further research in this framework is that by further field-testing or with marker data, we can give a closer estimate of the actual risk. Information does not change the actual risk to the resource; it reduces our uncertainty about the risks that we might be able to reduce by management. Information may be useful to indicate how management can be more effective, but in these risk calculations, the value of information lies only in giving better risk estimates.
Our attitudes toward risk is often more complex than we can easily model. We may feel that if risks lie below a certain level, that we can ignore small differences such as between 5% to 25%, and that at very high levels, such as between 75% and 95%, they are equivalent. But we may feel very sensitive to differences between 40% and 60%. We may also feel sensitive to risks above 50%, but not below that and we would want to be able to adjust the importance of actions accordingly.
VALUATION
Another factor to consider in weighing a
course of action is whether the loss of the resource carries much of a
penalty to any of the forest users. The penalty may be in lost
opportunities to improve crops or in losses to income, or in losses to
ecosystem health and general forest productivity. This is one of the most
difficult questions for a conservation officer to face since we often lack
simple measures of value. Many of the values of conservation cannot be
measured well, if at all, in economic terms, and rarely would market
determinations be useful. In forests, there are multiple features of the
ecosystem that are valued including direct income, aesthetics,
environmental buffering, and symbolic and religious values. In addition,
different parts of society evaluate aspects of the resource in vastly
different ways and make it impossible to derive any single measure of
value. This is a profound problem that can only be approached in
democratic societies by discussion that respects all parties. This is not
a topic that can be usefully discussed here, but is necessary for any of
the following systems of analysis to be effective. For our purposes, we
assume that we can find a relative value for different genes, populations,
and species, and can somehow derive an agreed upon score for value.
We might note that at this time, at least one aspect of value is estimable by markets and that national conservation programs often use such values intuitively for estimating the value of conservation. On the other hand, many NGO's and government organizations concerned with the environment estimate value by ecological status. They may target the rare resource, or the resource that supports many other resources, or focus on those that may be most sensitive to threats, or those that indicate the presence of threats, or those that are most widespread. These kinds of resources may be termed keystone, indicator, or flagship resources, that would be targeted for attention. While more complicated and difficult, it is possible to derive estimates of combined, multiple values as long as scores can be derived for single values.
MANAGEMENT
Finally, another factor to consider in
decision making for conservation is how effective we think that management
can be. For some conservationists, the management options may involve only
locking up the resource in some kind of reserve, or leaving it alone. The
decision is to make a reserve or not and the problem for the
conservationist is to estimate the relative probability of the sufficient
survival of the resource with and without a reserve. For other
conservation programs, more options may be available but each with their
own cost and probability of achieving the various values produced. The
benefit of each option may then be estimated for both present and future
values.
Presumably, each management option would not only engender a cost and benefit, but it would change the risks to the resource. Obviously, we anticipate that the expected risk is lowered, and perhaps the uncertainty of the outcome is also reduced, but for each option, we assume that we can estimate what management can do for us. We also assume that with multiple resources to consider, that a total combined evaluation can be made for a finite set of management options.
DECISION ANALYSIS
With the kinds of information outlined
above, we can try to make decisions that will be logically consistent and
transparent. There may be many other factors that impinge on the decisions
we make, such as the need to retain political support for conservation.
However, we would not want our decisions to be only based on ephemeral or
popular whims that pass every few years that have no lasting impact on the
resources we wish to conserve.
One principle for conservation management that seems to be useful is to direct effort to those resources that are most valuable, at most risk, and most effectively manageable.
If a resource is easily replaceable then its loss may be of small impact economically, ecologically, or otherwise. If a resource is not at risk, then efforts may be wasted on trying to save something that would be saved without effort. And if management cannot do much to increase security because the techniques that can be applied are ineffective, then expending effort is also a waste of resources that could be more beneficially spent elsewhere. We would like to put our efforts into those cases where we can be most effective in relieving the highest risks to our most valued resources.
On the basis of that principle, we can analyze different management options and estimate what efforts would be feasible to consider. One way to arrange thoughts on decision making is to consider a sequence of actions and events and the likely outcomes that each sequence can take, and then evaluate what actions lead to the best or most acceptable results. This is not the only way, but it openly displays a decision process and can be illustrated by a simple example. Consider first that the reproduction of a potentially valuable population is susceptible to drastic reduction in regeneration capacity due to the threat of wildfire. We may have three action possibilities; increase seedling survival by clearing openings for seedlings, increase fire protection, or do nothing. The first two may have equivalent costs and be equally effective, so we leave that decision to the field manager and consider only the options of managing or not managing. We then consider what we expect would to happen under two action scenarios and the probabilities and costs if a fire occurs or doesn't occur.
Obviously, there are great uncertainties about the probabilities, but if we know anything, we can certainly estimate that the probability of decreasing risks by management is greater than if we did nothing. The issue is to estimate the degree by which we can reduce risk and if its cost would be worth it.
For illustrative purpose, let us assume that the value of a population is 100 units and the management efforts will cost 14 units. Also, let us assume that the probability of a fire occurring is 0.5 and the probability of the population surviving with no intervention is 0.5. However, if management can improve reproduction and increase minimum effective population size, the probability of survival in spite of the fire may be 0.7. If no fire occurs, we assume that the probability of survival is 0.95. Then the expected value consequent to management can be calculated as follows: 0.5 * 0.7* 100) + (0.5*0.3* 0) + (0.5*0.95*100)+(0.5*0.05*0) - 14 = 68.5 units. So the expected value of the management, when there is no information available to begin with, is 68.5 units.
On the other hand, if we don't manage the expected value will be 72.5 units (See Fig 1). This shows that with the assumed level of survival after management, the cost of management, and the value of the population, the better decision is to not go to the expense of management.
The above estimate of management is a priori and does not account for what a field conservator knows about the local situation. We can see what the effect of better management is on our decision by changing some parameters. Let us therefore say that based upon field evaluation, the conservator can give us better judgements on how management can be effective, say by supplementing the pollen or source. The probability a fire occurring may still be 0.5, but now by better management, the probability of saving a population is 0.8 in spite of occurrence of fire. Now the rate of fire doesn't change, but we can allocate effort better and now apply management when it is needed. We will still make mistakes and apply management and fail but now only with 0.2 probability. The expected value of management will be 73.5 units compared with 72.5 units for no management option indicating that management is a better option (Fig 2).
Let us assume a third case in which little information is available as in case 1 above, but can be collected at a cost. This third option is to collect additional information before making the decision to manage (Fig 3). In that case, the increase in research or other data analysis system would presumably change the probabilities for making the correct decisions. Suppose for example that research on the actual genetic structure of the population gives us more accuracy in predicting which populations are truly susceptible, and those that are not. In that case, the probability of suffering a loss when management is exercised and a fire occurs is lower because we manage when it is necessary. Now the probability of making the management effective may go to 0.9. Even when we make a decision not to manage, as it is based on more information, probability of survival after a fire incidence will be increased, say to 0.6. However, new information collection bears costs in terms of time and research. If we consider the cost of the additional information is 5 units, the expected value of managing after collecting additional information is 73.5 units. The increased expected value for new information option makes it a better decision to follow. We can now see what the judgement should be if extra information is useful even at a cost to the manager.
CONCLUSIONS
There are many options that can be
constructed for estimating the value of multiple management choices. We
can include sequential stages of events and decisions by adding more
branches to the decision tree and can include more management options by
adding to the number of branches at each node. It is possible to include a
decision to delay decision until further information can be obtained, such
as by research on the genetic structure or mating structure of populations
and some are sketched in by Koshy et al, (2000)2 . There are also ways to
evaluate choices when multiple values are involved in making decisions
about conservation.
These kinds of decision tools can help to rationalize how we can use information to increase our expected effectiveness. They are not tools for making the problems themselves any simpler or for resolving difficult problems in obtaining the estimates of susceptibilities and threats. It is obvious that the factors of biology that make up susceptibility are not independent and that the factors that constitute real threats are also no independent. Nevertheless, means exist for us to rationalize decision-making and to make honest mistakes, but to minimize their costs. Decision trees help to visualize the logical processes that we sometimes use intuitively, but make them more transparent to ourselves and to others.