Rohit Mathur Atmospheric Modeling and Analysis Division, NERL, U.S. EPA
Karen Blocksom and Joseph Flotemersch EPA/ORD/NERL NWQM Conference, San Jose, CA May 11, 2006
description
Transcript of Karen Blocksom and Joseph Flotemersch EPA/ORD/NERL NWQM Conference, San Jose, CA May 11, 2006
Karen Blocksom and Joseph FlotemerschKaren Blocksom and Joseph Flotemersch
EPA/ORD/NERLEPA/ORD/NERL
NWQM Conference, San Jose, CANWQM Conference, San Jose, CA
May 11, 2006May 11, 2006
Effect of Taxonomic Resolution on Performance Characteristics of a
Method for Sampling Large Rivers
*Although this work was reviewed by EPA and approved for publication, it may not necessarily reflect official Agency policy.
Background• A new bioassessment method developed for
collecting macroinvertebrates in non-wadeable rivers (In press: River Research and Applications, vol. 22, July 2006)
• Study conducted to estimate performance characteristics using metrics: precision and sensitivity
• Evaluated at the lowest possible taxon (species) level
• Many states stop at genus or family, so relevant to evaluate performance characteristics at these levels as well
Sample Collection and Processing
• 19 sites across 5 rivers
• Deep and shallow sites across disturbance gradient
• 3 replicate samples per site in field
• Also habitat, basic water chemistry collected
• Up to three 300-organism sorts per sample in the laboratory
100 m
Flow500 m
Flow
Transect 1
2
3
4
5
6
A
A
A
A
A
A
BC
BC
BC
BC
BC
BC
Expectations (Family/Genus relative to Species level)
• Richness metrics will have lower values in general and hence, lower variability (higher precision) at the genus and family levels (Total and EPT taxa richness)
• Tolerance and functional feeding group metrics will have lower variability because designations will apply to broader groups and should be more consistent at this level (% Tolerant individuals and % Collector-filterers)
• Laboratory variability will be lower because less chance for error than at species level
• Sensitivity will be poorer because of broader groups at the family level – fewer correlations with abiotic variables
Precision - Methods
• Perform ANOVA with:
Site ID as factor
Multiple samples (A,B,C) as field replicates
Multiple 300-organism sorts as lab replicates
• Use RMSE from this analysis to calculate:
Detectable difference (DD) – minimum difference detectable based on sample size of 3
Relative DD (as percentage of metric range)
Precision - Field
Taxa ric
hness
EPT richness
% Tolerant
% Coll-f
ilterer
0
1
2
3
4
5
6
7
8
D
D
SpeciesGenusFamily
Precision - Field
Taxa ric
hness
EPT richness
% Tolerant
% Coll-f
ilterer
0
2
4
6
8
10
12
Re
lativ
e D
D
SpeciesGenusFamily
Precision - Laboratory
Taxa ric
hness
EPT richness
% Tolerant
% Coll-f
ilterer
0
1
2
3
4
5
6
7
D
D SpeciesGenusFamily
Precision - Laboratory
Taxa ric
hness
EPT richness
% Tolerant
% Coll-f
ilterer
0
5
10
15
R
ela
tive
DD
SpeciesGenusFamily
Results - Precision• Taxa and EPT richness
DD (variability) increased from family to species as expected
BUT relative DD similar or slight decrease from family to species
• % Tolerant individuals For lab, DD (variability) increased family to species No strong pattern for field DD
• % Collector-filterers Genus and species identical - same designations All measures similar among taxonomic levels
Sensitivity - Methods
• Spearman rank correlations
• Family, genus, and species levels
• Between mean metric values and potential gradients:
Taxa richness, EPT richness, % Tolerant individuals, % Collector-filterers
Conductivity, pH, DO, temperature, EMAP habitat and human influence variables, NWHI metrics
Sensitivity
FamilyFamily
0 1 2 3 4 5 6 7 8 90
10
20
30
40
50
60
70
80
Tax
a ric
hnes
s
GenusGenus
0 1 2 3 4 5 6 7 8 9
Human influence - Roads
SpeciesSpecies
0 1 2 3 4 5 6 7 8 9
r = -0.49 r = -0.62 r = -0.66
Sensitivity
FamilyFamily
200 400 600 8000
5
10
15
20
25
EP
T r
ichn
ess
GenusGenus
200 400 600 800
Conductivity (uS/cm)
SpeciesSpecies
200 400 600 800
r = 0.61 r = 0.59 r = 0.59
Sensitivity
FamilyFamily
0 1 2 3 4 5 6 7 8 90
102030405060708090
% T
oler
ant
indi
vidu
als
GenusGenus
0 1 2 3 4 5 6 7 8 9
Bottom Deposition score (NWHI)
SpeciesSpecies
0 1 2 3 4 5 6 7 8 9
r = -0.45 r = -0.41 r = -0.50
Sensitivity
FamilyFamily
0 1 2 3 4 5 6 7 8 90
10
20
30
40
% C
olle
ctor
-filt
erer
s
GenusGenus
0 1 2 3 4 5 6 7 8 9
Bottom Deposition score (NWHI)
SpeciesSpecies
0 1 2 3 4 5 6 7 8 9
r = 0.66 r = 0.67 r = 0.67
Results – Sensitivity
• As expected, generally more significant correlations at species level
• Strength of common correlations similar among taxonomic levels
• Variability around trend lines tends to be greater for genus and species level data
• In the context of relative DD, taxonomic level had little to no effect
• Sensitivity only slightly stronger at genus/species level than at the family level
• Overall, taxonomic resolution did not strongly affect performance characteristics of the method
Conclusions