Collect More Information on Remaining Causes

Even when the characterization of causes has not determined the cause with a satisfactory level of confidence, it is likely that the set of candidate causes has been reduced. It now may be possible to design experiments or observations that will potentially eliminate one or more of the remaining candidate causes, although such experiments are not always feasible. Additionally, it may be possible to identify critical pieces of positive evidence that would strongly support one scenario and none of the others.

Because additional data collection and analysis will result in additional costs, management should be consulted to discover whether determining the cause of the impairment is still required for decisionmaking. If so, the results of the SI process can be used to design a sampling and testing program that will generate potentially decisive evidence.

Reconsider the Impairment

When no cause is identified, it may be that there is actually no impairment, or that the actual impairment differed from the specific effects that were investigated. This situation is known as a false positive, or in statistical terms, a Type I error. It should be noted that both false positive and false negative errors (failure to detect an effect that exists) are inherent in any detection system, whether it is medical diagnostics, aircraft radar, or environmental monitoring.

A false positive might result from errors in a biological survey or in the analysis of data. The samples may have been collected improperly; therefore, the biotic community appears to be less abundant or species-rich than it truly is. The individuals performing the identifications could have misidentified organisms. There could have been errors in data recording or analysis. Any of these errors can artificially obscure or inflate the responses. A quality assurance program can minimize, but not entirely eliminate, these errors. If the causal analysis reveals weaknesses in the evidence for a real effect, a careful audit of the biological survey may be appropriate.

Other reasons for a false positive result may include sampling error and the natural variability of the biological indicators. In any monitoring program, sampling is stratified among perceived natural classes and subdivisions of systems (e.g., habitat type, salinity, sediment, elevation, biogeographic region), and often by season (sampling index period in defined season). A sample could have been taken outside an index period. A site could belong to a poorly characterized system type or could have been incorrectly classified (e.g., a cold-water system evaluated using warm water criteria). Any unrecognized misclassification can result in a false positive or a false negative outcome. Intensive monitoring and characterization of natural systems, combined with quality assurance and peer review of results, can reduce both types of errors.

• Revisiting the Impairment in the Case of DDT

Finally, natural variability of the indicators, not due to any measurement or analytical errors, can result in false positives or false negative outcomes. Environmental criteria can be defined by exceedance of a percentile or extreme value of some statistical distribution. This means that natural, or unimpaired, conditions also can exceed the criteria at some frequency. Ideally, acceptable error rates should be specified for decisions resulting from the biological assessment system.

If confidence in a finding of biological impairment is low (i.e., if the indicator barely exceeds the impairment threshold value), additional sampling may reduce uncertainty and increase confidence.

Collect Information on Additional Causes

If all of the most common causes have been eliminated or have been determined to be unlikely, then additional causal scenarios will need to be identified. The process for accomplishing this is similar to that described in List Candidate Causes. For example, new data may have become available during the first run of the Stressor Identification process. These data should be reviewed carefully, to determine if they contain any clues to suggest additional causal scenarios. Details of the available data should be considered, such as weather patterns, new construction, or land use information. If the descriptions of the effect or the scope initially were too broad, they may need to be refined or defined more clearly. Additional candidate causes can include new stressors or combinations of stressors that occur simultaneously or in a specific sequence (i.e., a causal scenario). After the additional candidate causes have been developed, key evidence likely to result in identification of the cause should be targeted.