Transcript ppt 2.3mb - West Virginia Mine Drainage Task Force

Presented By:
Robert W. Gensemer
Co-authors: S. Canton, G.
DeJong, C. Wolf, and C. Claytor
Should There Be An
Aquatic Life Water Quality
Criterion for Conductivity?
WV Mine Drainage Task Force Symposium
Morgantown, WV
March 29, 2011
Why Conductivity?
 Coal mining and valley fill (CM/VF) activities in
West Virginia can be associated with increased
conductivity
•
Increased sulfate, bicarbonate
 Some have suggested an adverse relationship
between conductivity and benthic macroinvertebrate
communities
•
Primarily focused on “sensitive” mayflies
 Thus, aquatic life benchmarks (functionally “criteria”)
for conductivity are being proposed
2
D.S. Chandler
Conductivity Criterion –
Complications Exist
 Conductivity is a composite variable
• Surrogate measure for dissolved solids (cations & anions)
• Ionic toxicity exists, but varies with ion composition
– Composite variable cannot differentiate ionic balance differences
• Toxicity can be mitigated by hardness
 Patterns of macroinvertebrate community
composition vs. conductivity can be confounded
• Related to combination of abiotic and biotic factors
– Abiotic: e.g., water quality, habitat, temperature
– Biotic: e.g., competition, predation, colonization, biogeography
3
EPA’s Proposed Conductivity
Benchmark
 For central Appalachian streams
• 300 µS/cm
– Sensitive species assumed to be
‘extirpated’ if exceeded
• Limited to streams dominated by
sulfate and bicarbonate salts at
circumneutral to mildly alkaline pH
• EPA methods for aquatic life criteria
used, modified for use of field data
4
EPA’s Proposed
Conductivity Benchmark
 Assumption: “sensitivity” related to field
distribution
• Quantified as an extirpation concentration (XC)
– Instead of standard LC50 or chronic responses
• XC = concentration above which a genus is ‘effectively
absent’
• XC95 = 95th percentile of distribution of calculated
‘probability of occurrence’ of
a genus with respect to conductivity
XC95 (from EPA 2010)
5
EPA’s Proposed
Conductivity Benchmark
 Benchmark of 300 µS/cm
• Ranked distribution of XC95 values
• 5th percentile = 297 µS/cm (rounded to 300)
• Assumed to prevent
extirpation of all but
5% of the most
“sensitive” species
6
Primary Technical Concerns
 Assumed responses to conductivity not consistent
• 3 types of associations noted by EPA
 2 other types also present but not recognized by EPA
 These are all fundamentally different responses
•
i.e., not just varying levels of sensitivity
7
“Conflicting” StressorResponse Profiles
Percentage of genera with different types of stressor-response profiles with
respect to conductivity and probability of capture (data from EPA 2010a).
Stressor-Response Profiles
 Conflicting stressor-responses result
in conflicting answers:
•
•
•
•
•
Decreasing (Ephemerella): <300
Increasing (Hemerodromia): >300
Optimum (Psephenus): >75 and <2,500
Bimodal (Diplectrona): <200 and >2,000
No response/bimodal (Tvetenia): none
 How can a single benchmark value be
chosen from those numbers?
Primary Technical Concerns
 Incomplete analysis of causality
• Correlation ≠ causality!
• Limited experimental evidence
(few laboratory studies)
 Confounding factors dismissed
inappropriately
• Takes causality of conductivity “as a given”
• Important factors dismissed
– Habitat, flow, substrate characteristics, etc., widely known to
influence species composition
“Today's scientists have substituted mathematics for experiments, …
and eventually build a structure which has no relation to reality”
– Nikola Tesla
10
EPA Approach:
Causality
 Goal: “establish that salts are a general cause, not that they cause
all impairments, nor that there are no other causes of impairment, nor
that they cause the impairment at any particular site.” (emphasis added)
 Epidemiological approaches used
•
6 characteristics of causation
–
–
•
Co-occurrence, preceding causation, time order, interaction, alteration,
sufficiency
Weight of evidence scoring
Concluded that salts (measured by conductivity) are common cause
of aquatic macroinvertebrates impairment
 Our conclusion: This is an incomplete analysis
•
Weight of evidence scoring for each element relatively subjective
‒ Open to valid alternative interpretations
•
Limited experimental evidence
– Few toxicity tests
– No experimental verification of extirpation in whole communities
11
EPA Approach:
Confounding Factors
 Approach Used:
•
Do confounders alter the statistical relationship between salts and
macroinvertebrate assemblages?
–
•
Habitat, organic enrichment, nutrients, deposited sediment, high/low pH,
Se, temp, lack of headwaters, catchment area
Effect of confounders found by EPA to be “minimal and manageable”
–
–
Low pH → removed sites with pH < 6
Influence of Se → not enough data , should be investigated
 EPA’s confounding factors analysis took presumed impacts from
conductivity as a given
 Our conclusion: should have included rigorous, independent tests
to first determine if conductivity is indeed the best (or only?)
predictor of biological impairment
12
EPA Approach:
Confounding Factors
• What about alternative explanations for community structure
patterns?
Habitat:
1. RBP scores not best measure of macroinvertebrate habitat quality
2. RBP scores correlated with conductivity and biological response
3. Analysis focused on relationship with Ephemeroptera (mayflies)
 Excluded the rest of the benthic macroinvertebrate community
Relationship to other invertebrate taxa:
1. Relationships with Ephemeroptera used to reject other stressors as potential
confounders
2. Should include analyses for other invertebrates
 Again, excluded the rest of the community -- Protect all invertebrates, not just mayflies!
13
Our Approach:
ID Additional Confounders
 Independent analysis that considered additional
information
•
•
Identify key WQ and physical parameters most strongly
associated with biotic responses
Minimize use of composite variables (e.g., conductivity)
 West Virginia Department of Environmental Protection
(WVDEP) Watershed Assessment Branch Database
(WABbase) – same as used by EPA
•
Results for 3,286 sampling events
– 3,121 unique Station ID codes
•
A variety of site-specific data
–
–
–
–
Regional landscape
Water quality
Aquatic habitat conditions
Macroinvertebrate community composition
14
Our Approach:
Statistical Tests Used
 Principle Components Analysis (PCA)
•
Variable reduction procedure
–
–
–
Identifies redundancy among numerous variables
Do variable groups “move together”?
Can 1 variable be used as a surrogate for other variables within each grouping?
 All Possible Regressions (APR)
•
Identifies 1 variable or subset of variables that explains most variation
observed in biological response
–
Goal to identify smallest subset of variables that explains most of the variation
 Chi-square Automatic Interaction Detection (CHAID)
•
•
Evaluates relationships between dependent variable and independent
stressor variables
Selects subset of stressor variables that best predicts the dependent variable
–
•
Presents these variables in a decision tree
Decision tree:
–
–
Starts with dependent variable
Progressively splits into smaller branches (nodes) based on groupings of stressor
variables that best predict responses by dependent variable
15
Statistical Conclusions
Independent stressors most closely associated with key dependent
responses (genera-based total taxa and percent EPT):
Principal Component
Analysis
Genera-based Total Taxa
Total magnesium
Percent fines
Percent EPT
Percent fines
Total magnesium
Total suspended solids
All Possible Regressions
Chi-square Automatic
Interaction Detection
Undisturbed vegetation
Channel alteration
Sulfate
Sulfate
Channel alteration
Total magnesium
Embeddedness
Epifaunal substrate
Undisturbed vegetation
Epifaunal substrate
Fecal coliforms
Chloride
Total manganese
Epifaunal substrate
Fecal coliforms
Bank vegetation
pH
16
Statistical Conclusions
 A single composite parameter, like conductivity, cannot explain
the observed variation with respect to WQ and physical habitat
 Rather, some combination of ionic composition, substrate, and
channel features may be the most appropriate stressor
variables to consider
•
21% variation explained in Total Taxa
– Conductivity vs. Total Taxa (r 2 = 0.18)
•
14% variation explained in %EPT
– Conductivity vs. % EPT Abundance (r 2 = 0.08)
17
Alternative Approach:
Single Ion Criteria
 Illinois sulfate criterion
Ion Ranges
Hardness
<100 mg/L
Hardness
100 to <500 mg/L
Hardness
≥500 mg/L
Chloride
Chloride
Chloride
Chloride
<5 mg/L
5 to <25 mg/L
25 to <500 mg/L
≥500 mg/L
500
n = 696
500
n = 350
500
n = 23
500
n=0
500
n = 113
Eqn 1
n = 84
1 of 84 exceeded
criteria
Eqn 2
n = 270
2,000
n=1
2,000
n = 26
2,000
n = 15
7 of 15 exceeded
criteria
2,000
n=3
1 of 3 exceeded
criteria
500
n = 10
6 of 10 exceeded
criteria
Eqn 1: Sulfate = [-57.478 + 5.79(Hardness) + 54.163(Chloride)] x 0.65
Eqn 2: Sulfate = [1,276.7 + 5.508(Hardness) – 1.457(Chloride)] x 0.65
 <1% of WABbase samples exceeded the IL sulfate criteria
•
Majority of exceedances occurred with hardness levels >500 mg/L
 26% of these WABbase samples exceed the proposed conductivity
benchmark
18
Conclusions
 Relationship between conductivity and changes in
macroinvertebrate community structure not strong or
reliable enough to derive a benchmark
 EPA (2010) did not rigorously test primary hypothesis
that conductivity is best predictor of changes in
macroinvertebrate community structure
• Instead, their analysis takes it as a given that conductivity is the
best predictor
• Confounding factors prematurely dismissed
 Insufficient experimental confirmation of the proposed
benchmark
•
For similar reasons, IL, IN, and IA rejected the use of TDS or
conductivity-based criteria in lieu of criteria for individual ions
(sulfate or chloride)
19
Conclusions
 It is inappropriate and inadvisable to adopt a
conductivity benchmark at this time
• Many factors other than WQ are strongly related to
macroinvertebrate community structure
 To adopt this benchmark without additional
study runs a risk of expending financial
resources to reduce conductivity
• Little confidence that mitigating conductivity alone
would provide any measureable environmental
benefit
20
Acknowledgements
We would like to thank:
The National Mining Association
21
22
[email protected]