Identifying Probable Causes for Impairment at Pretend Creek

1. Step 5: Identifying probable causes

2. What’s going on at Pretend Creek?

3. Detect or Suspect Biological Impairment Stressor Identification Define the Case List Candidate Causes Decision-maker and Stakeholder Involvement As Necessary: Acquire Data and Iterate Process Evaluate Data from the Case Evaluate Data from Elsewhere Step 5: Identify Probable Cause Identify and Apportion Sources Management Action: Eliminate or Control Sources, Monitor Results Biological Condition Restored or Protected

4. How we identify probable causes • Eliminate when you can • Diagnose when you can • Otherwise, analyze strength of evidence HOW? Apply a scoring system to the available evidence under each type of evidence

5. R refutes D diagnoses +++ convincingly supports (or weakens - - -) ++ strongly supports (or weakens - -) + somewhat supports (or weakens - ) 0 neither supports nor weakens NE no evidence The scoring system

6. Scoring of spatial/temporal co-occurrence

7. Scoring of laboratory tests of site media

8. Scoring of stressor-response from other field studies

9. Example: strength of evidence analysis

10. Weighing the evidence for each candidate cause • Evaluate the quantity & quality of evidence • Evaluate consistency & credibility • Summarize the compelling evidence

11. Evaluate quantity & quality of evidence • Quality & quantity of data influence scores • Lots of consistent evidence reduces quality concerns for any 1 type of evidence • Poor quality data may be discounted • Consider study designs, methods, relevance, variability, & other QA issues

12. Evaluate consistency & credibility • Prepare summary table of scores • Do not add up scores! • Evaluate consistency of evidence • Look for compelling evidence • If evidence is inconsistent, consider mechanistic explanations • e.g., lab data not consistent with field conditions due to differing bioavailability

13. Scoring consistency & credibility

15. Summarize compelling evidence • Make an overall evaluation of strength of evidence for each candidate cause • what evidence compels belief that candidate cause induced effect? • what evidence strongly casts doubt? • Consider the principle characteristics of causal relationships • these are what you’re trying to show • they summarize the 15 types of evidence

16. There is no magic formula… • All candidate causes must be compared to determine: • if there is more than 1 probable cause • your level of confidence in the results

17. …celebrate, then remediate for Candidate Cause 1 Comparing evidence among causes: best-case scenario You have compelling evidence for 1 candidate cause; others are weak or refuted...

18. Strong evidence for one candidate cause may be sufficient • Consider if weakness is due to lack of data – and try to fill holes Comparing evidence among causes: more (likely) scenarios You have uneven evidence across candidate causes...

19. Reconsider the impairment • Consider additional candidate causes • Consider episodic events • Consider gathering more data You have unsatisfying evidence across all candidate causes…

20. Consider disaggregating indices or metrics • Combine causes if they share causal pathways, modes of action, sources and routes of exposure, or if they interact • Remediate dominant cause • Design remediation to address multiple causes You have evidence suggesting multiple causes…

21. Gather data if possible • Consider other bases for remediation (e.g., BMPs, chemical criteria) and monitor biological responses • Use professional judgment as last resort You have insufficient data…

22. Next: let’s try it out! Scoring the evidence exercise

23. What’s going on at Pretend Creek?

24. Things to keep in mind when scoring… • Be consistent across candidate causes • Don’t double-count data • Think about the quality & quantity of the data you’re using • Document your thought process

25. How do I communicate results? • Make your logic clear • Present the critical evidence • Reveal uncertainties • Fit communication to your audience • For technical reviewers, include text & tables • For decision makers, may be helpful to use annotated conceptual model

26. industrial facility urbanization dam subdivision dairy farm ↓ dissolved oxygen ↑ metals ↑ temperature ↓ EPT richness Using models for communication: Pretend Creek

27. industrial facility urbanization dam subdivision dairy farm ↓ dissolved oxygen ↑ metals ↑ temperature DO higher at impaired site vs. reference ↓ EPT richness Using models for communication: Pretend Creek

28. industrial facility urbanization dam subdivision dairy farm ↑ metals ↓ dissolved oxygen ↑ temperature DO higher at impaired site vs. reference after rerouting industrial discharge had decreased metal concentrations & increased EPT taxa richness ↓ EPT richness Using models for communication: Pretend Creek

29. industrial facility urbanization dam subdivision dairy farm ↑ metals ↑ temperature ↓ dissolved oxygen DO higher at impaired site vs. reference after rerouting industrial discharge had decreased metal concentrations & increased EPT taxa richness temperature higher at impaired site vs. reference; negatively correlated with EPT taxa richness ↓ EPT richness Using models for communication: Pretend Creek

30. What comes after causal analysis? • If confidence in results is low… • plan studies to obtain critical evidence • experimental studies most likely to be convincing • If confidence in results is high… • identify sources • take action • monitor results

31. Example: fish kills in Virginia & West Virginia 1. Low dissolved oxygen in water 2. Gill damage from ammonia, high pH, or other mechanism prevents uptake of oxygen 3.Altered blood chemistry from nitrite exposure prevents fish from using oxygen 4. Viral, bacterial, parasitic, or fungal infections 5. Mortality from high pH 6. Mortality from pH fluctuations 7. Mortality from ammonia toxicity 8. Toxicity of unspecified substances 9. Starvation due to inadequate food resources There was no “smoking fish”… • Causal analysis resulted in: • Narrowed list of candidate causes • List of priority data needs