Step 5. Identify Probable Causes

Ultimately, the case for a candidate cause must be evaluated as a consistent, credible argument. Then, each candidate cause is compared to every other candidate cause to determine which causes led to the specific effects. Your rationale for identifying one cause relative to others must be clear, reasonable and convincing to motivate and direct management action. Consequences of deferring a decision versus making an incorrect decision must be considered carefully.

Your analysis leads to the final position statement of the causal assessment. Therefore, it is most important that this step be undertaken with knowledge, evenness and care. It helps to be aware of what is known and what is not known. In particular the quality of the data and the causal analysis should be described when documenting the final conclusions.

Sometimes it helps to record what additional information would significantly strengthen the determination. Sometimes it helps to record unstated assumptions that you have made.  Above all, remain objective and question assumptions, biases and motives at every opportunity.

Determining the most probable causal agent from several candidates requires retaining and conveying a lot of information, especially the most compelling lines of evidence. These can be recorded on tables of evidence or on annotated conceptual models.

It is a good idea for the investigator to review the documentation of the measurements that underlie the causal association, that is, the evidence. You need to be comfortable with the information. If you are unfamiliar with a particular discipline, consult a specialist in that area.

This step is divided into two tasks to make the process of determining a probable cause more manageable.

• In the first task, evidence for each candidate cause is evaluated, candidate causes are categorized and the most compelling lines of evidence are noted.
• In the second task, the evidence for candidate causes is compared across all candidate causes. The product is the identification of the candidate cause or causes of the biological impairment and a description of the evidence for that decision.

In the best case, a probable cause or causes are identified, and the information is effectively communicated to managers and stakeholders. In some situations, no cause is identified or the confidence in conclusions will be too low to support management action. However, even then, the process will likely allow you to identify what information would allow a cause to be identified.

• Scores for each type of evidence displayed in a table,
• An evaluation of the consistency and credibility of the case based on the scores,
• A classification of each candidate cause as refuted, diagnosed, probable, unlikely or uncertain,
• A discussion of the reasons for the final conclusions including the most compelling lines of evidence, and
• A report describing the causal assessment.

Weigh the Evidence for Each Cause

You can be confident in deciding that a candidate cause did or did not lead to the impairment when all the evidence has high quality and makes sense qualitatively and quantitatively. When the supporting data are few or of poor quality, confidence is low. When the some types of evidence weaken and others strengthen the case for a candidate cause, uncertainty is greater and a determination is difficult.

Even when based on poor or minimal information, the causal analysis will still be useful as a screening assessment to reduce the number of candidate causes or to identify data needs. Therefore, "don't let the perfect be the enemy of the good!" Instead, use all of the evidence that you have to make what inferences you can.

This section describes how to evaluate the body of evidence for each candidate cause and how to identify probable, uncertain, and unlikely causes. When evaluating the evidence for a candidate cause, the quantity and quality of each type of evidence is evaluated separately; then consistency and credibility of the entire body of evidence is evaluated. Use of this consistent and transparent process will result in more defensible conclusions.

Evaluating the Quantity and Quality of Evidence

• High-quality data are always superior to questionable data or data of uncertain origin. You may choose to exclude or discount poor-quality data. However, if you do, first consider if the low quality data can help you to identify the type and quality of new data that would improve the assessment.
• Increasing the number of types or pieces of evidence decreases the likelihood that any one faulty study or data set will mislead you. Likewise, increasing the quantity of data increases confidence in the conclusions of a particular study or your own calculations and thus increases the quality of the evidence.

• Number of pieces of evidence,
• Number of types of evidence evaluated,
• Quality of the data,
• Quantity of the data,
• Proper sampling design,
• Relevance of the data from elsewhere to the case at hand, and
• Distribution of the data across candidate causes.

• Quantifying Uncertainty in Causal Analysis
• Data Quality

Evaluating Consistency and Credibility

The consistency and credibility of the overall argument is just as important as the quality and quantity of data and evidence marshaled to support the case. Whereas the quality and quantity of each piece of evidence was evaluated individually, now the types of evidence are considered together. Do they tell a consistent story? When the candidate cause is consistently supported or weakened by many types of evidence, the confidence in the argument for or against the cause increases. The number of types of evidence makes a difference. It is unlikely to find eight different types of evidence all supporting a cause by chance.

In contrast, consistent support for a cause by only one or two types of evidence could easily occur by chance alone. Sometimes there is a reasonable explanation for why a type of evidence does not agree with the rest of the evidence. So, if inconsistent evidence can be explained by a mechanistic, conceptual, or mathematical model, then the confidence in the argument for a candidate cause increases.

Evaluate consistency by bringing together the summary tables produced in Steps 3 and 4. Evaluate each specific effect individually. Although this makes for a complicated summary, it is important to do because different candidate causes may be eliciting different effects. Resist the temptation to add up the scores. Adding the scores erroneously implies that each type of evidence is equally important and is equitable only if the same types of evidence are available across all candidates.

Further, the symbols are not units. Consider a candidate cause with two types of evidence, each with a score of +, giving a sum of ++ (1+1=2), and another with three types of evidence with scores of +++, ++ and - - - (3+2-3=2). Both sum to 2, but the triple negative score may be strong enough to refute the candidate cause! Instead, please use the scoring tables to identify the most compelling pieces of evidence and to develop an overall sense of the case for each candidate cause.

• Consistency of Evidence
• Explanation of Evidence

Summarizing the Compelling Evidence

After scoring the body of evidence for consistency, make a preliminary evaluation of the potential for the candidate cause to have led to each specific effect. A strong case is based on evidence that demonstrates four or five characteristics of causal relationships using many types and pieces of evidence. The investigator records the most compelling evidence for or against each candidate cause. This evidence will be used to convince stakeholders and decision-makers.

Although there are fifteen types of evidence, they can be usefully thought of as potentially supporting the six characteristics of causal relationships listed in Table 5.2 below. Confidence in a cause is increased if the supporting evidence addresses all six characteristics. Bear in mind, however, that it is not necessary that you demonstrate all six characteristics to satisfy the decision-makers and stakeholders involved in the case.

Tables 5.3 and 5.4 list the Characteristics of Causal Relationships Supported by Different Types of Evidence.

Summarizing the Strength of Evidence for Each Candidate Cause

Table 5.5 summarizes options for categorizing the status of each candidate cause after the evidence is weighed. These results are used to compare evidence across causes.

Next, in Compare Evidence Among Causes, candidate causes are compared to determine if there is just one probable cause, more than one probable cause, or some other conclusion.

Compare Evidence Among Causes

At this point you should have all of the evidence organized and be ready to reach your final determination. The final determination is reached by comparing the evidence across candidate causes. Comparison supports your conclusions by:

• Ensuring that each candidate is treated fairly and that any biases in data collection and analysis are acknowledged,
• Identifying the candidate cause with the relatively strongest support when evidence is sparse, and
• Identifying the data or information that would most improve confidence in your conclusions.

If several specific effects were analyzed, conduct your comparison separately for each effect. Then, evaluate whether one cause is responsible for all of the specific effects, or if several causes are operating.

There is no magic formula. All of the candidate causes must be compared to determine if there is more than one probable cause and to determine the level of confidence in the overall determination. Typical combinations of status and confidence are described in the first section below, followed by a suggestions for documenting conclusions.

Typical Outcomes of Comparisons

One Candidate Cause is Diagnosed or Probable; Other Candidate Causes are Unlikely or Refuted

Celebrate! Document your conclusions and rationale.

You Have Compelling Evidence that Different Specific Effects Were Caused by Different Causal Agents; Other Candidate Causes are Unlikely or Refuted

Celebrate! Document your conclusions and rationale. Revisit how each specific effect is related to the impairment that originally triggered the investigation. You may be able focus management action on the causal agent(s) that will provide the biggest gains in improving condition. Revisit the conceptual models to see if the different causal agents can be traced back to a common source.

You Have Sparse Evidence Across All Candidate Causes

If the evidence for all the candidate causes is too sparse to confidently identify a probable cause, you may still be able to identify the candidate cause that has the strongest support relative to the others. To do this, consider what you know about ecology in general and about this particular ecosystem, impairment, and the candidate causes. All the evidence is important, as noted previously. However, the likelihood that the magnitude, intensity and duration of exposure were sufficient to cause the effect weigh heavily here. If one candidate cause emerges as having the strongest support, it may make sense to identify it and indicate uncertainty about the others. Consider the consequences of not identifying the cause with the strongest support: if not identified, it may be that no action will be taken at all. A thoughtful adaptive management approach can provide additional evidence for causal analysis while also improving some conditions at the site.

You Have Uneven Evidence Across Candidate Causes

If you have a strong case for one candidate cause, but the other candidate causes are uncertain because there are fewer data and less evidence to evaluate, then there may be bias in data collection, either from the site or from the literature. You must remain objective and question assumptions, biases, and motives at every opportunity. If the lack of data is from the field, look for data sets collected by other groups or agencies. You might also want to recommend changes to your monitoring program. If the lack of data is from the literature, consult other case studies and invest the time now to develop a useful literature summary so that you can strengthen future case studies.

You Have Insufficient Evidence Across All Candidate Causes

• Consider the specific biological effect again. Errors in the biological survey or assessment may have resulted in mischaracterization of the effect. For instance, bioassessment criteria for high-gradient streams may have been applied to a low-gradient stream. Defining the biological effect more specifically, or defining more than one effect, makes it easier to find relevant evidence.
• There may be other possible candidate causes that have not yet been considered. Re-examine your conceptual models. Consult experts outside your specialty. Talk to stakeholders and local people.
• Consider if jointly acting events cause the effect. For example, excessive high algal biomass plus three consecutive cloudy days might result in unusually low levels of dissolved oxygen. Multiple causes are discussed further below.
• Perhaps the data have not properly captured episodic events. Try to narrow the geographic scope of the assessment to make it easier to find potential sources. Investigate the types of sources and land-use activities to better characterize the possibility of episodic events.
• If all else fails and you are unable to isolate a probable cause, identify the cause or causes that are most likely by using best professional judgment and indicate what new data would strengthen a determination of the probable cause. Consult with decision-makers to determine if additional data collection is warranted.

The Evidence Suggests that Multiple Causes are Operating

When evidence supports more than one candidate cause, there are potentially multiple causes. Although this issue should have been addressed when defining the case and listing the candidate causes, it should be reconsidered here if the results are unclear. New evidence or new understanding may reveal relationships among agents that were not apparent in the beginning.

• It may be appropriate to consider whether the impairment was properly defined in Step 1. Define the Case.
  • The apparent multiple causes may actually be individual causes of multiple effects.  Consider partitioning the impairment if, for example one cause is inducing tumors in fish and another is reducing benthic insect abundance.
  • The apparent multiple causes may actually be operating in different areas of the aquatic system.  Consider partitioning the impairment in space.
• It may be appropriate to consider whether the candidate causes were properly defined or whether multiple stressors should have been evaluated in combination.
  • Listing Multiple Stressors as Candidate Causes

• Remediate a dominant and potentially sufficient cause.  An apparently dominant cause may be sufficient alone to induce the impairment and its actions may be masking the more subtle effects of other causes.
• Remediate a necessary cause. If one cause is necessary for occurrence of the impairment, then remediating only it is adequate.
• Remediate a feasible cause. If it is not clear how multiple causes interact, perform the easiest remediation and monitor the results.
• Remediate all causes. In some cases, it is feasible to remediate all of the multiple causes.

You Have Insufficient Data

This looks like an empty scoring table with only a few pluses and minuses or with comments about the uncertainty of the data. One option is to recommend the collection of additional data. Data collection is most likely when the costs of data collection are low, the costs of remediation are high, the situation is contentious, and the existing data do not suggest which is the most probable cause.

You Have no Data

This is highly unlikely. At the minimum you should have information on land use/land cover and sources within your watershed. Use this information to conduct a screening-level Stressor Identification to identify the most useful data to collect from the case. Then consult with decision-makers to determine if data collection is warranted.

Documenting Conclusions

• See example conclusions

• The Role of Stressor Identification in Various Water Management Programs

Is the assessment for permitting, meeting aquatic life criteria, or for providing information that may lead to solutions for more than minimum recovery? Is the level of confidence sufficient to make a determination? Decide and document your rationale. Then communicate your findings in this final step.

Communicating the Results

• The reason for the causal analysis,
• A list of the candidate causes and the information supporting their selection,
• The source of the data used in the analysis,
• Tables of the evidence derived from the data,
• Conceptual models of the causal pathways,
• The key evidence that strengthen the probable cause and weakens the other candidate causes,
• Determination of the probable cause or causes,
• Qualitative assessment of the overall confidence of the entire case, and
• Next steps or other recommendations.

• Example conclusions in table form
• Example conclusions in narrative form
• Example annotated conceptual model

Next Steps

Once you are at this step you need to decide if your:

• Confidence is Low? Iterate Process. If the cause is not sufficiently certain for the decision maker, there may be other sources of data, other ways to evaluate existing data, or other options for iteration can be explored.
• See other Iteration options
   
• Confidence is High? Identify Sources, Take Action, Monitor Results.  If the cause is confidently identified, then the next steps may include allocating the contributions of different sources of the cause, developing and implementing management options, and monitoring the effectiveness of actions. These important activities are outside the scope of this website. However, accurate and defensible identification of the cause is the key that directs management efforts toward finding solutions that have the best chance for improving biological condition.