relationships between indices and natural environmental variables, provides further
information on index performance [21, 101, 106, 116].
The effectiveness of predictive models used for reference site classification is
often evaluated using cross-validation by constructing models using only subsets of
the available data. Leave-one-out cross-validation is a data-efficient method in
which reference sites are excluded from the dataset one at a time. After each
exclusion, the classification and modeling process is repeated. Each left-out site
is then classified using the model constructed without that site. The proportion of
agreements between original and cross-validated classifications, relative to the total
number of reference sites, is a measure of the effectiveness of the model. This
technique can be used with any site classification approach [114, 134, 144]. Null
models, which are constructed by predicting metric values across all reference sites,
with no classification, are useful for evaluating all types of classifications. If
classification strength does not exceed that of the null model, then the classification
provides no advantage [145].
5 Expert Interviews: Challenges and Important
Considerations in Bioassessment
To provide a broader and more comprehensive perspective, we conducted interviews with three expert practitioners and developers of bioassessment programs.
Their responses to our questions, provided here in a question-and-answer format,
have been summarized with a focus on emerging issues relevant to bioassessment in
aquatic systems.
Expert 1 Michael Barbour, Ph.D.—Adjunct Senior Scientist, Mote Marine Laboratory, Sarasota, FL, USA, and retired Director, Center for Ecological Sciences,
Tetra Tech, Inc., Owings Mills, MD, USA
Q: What factors limit the potential for increased use of genetic information in
bioassessment surveys? Is it likely that molecular genetic analysis will replace
traditional taxonomic approaches, or will these processes be used in conjunction
with each other?
A: A major challenge in the use of genetic data for bioassessment will be in
determining how reference conditions are expressed and how to account for the
effects of natural environmental variability on reference populations. It is unlikely
that genetic analysis will replace traditional taxonomic approaches in the near
future. Evolving DNA methods, however, should help to decrease taxonomic
uncertainty and improve our evaluations of aquatic assemblages.
Q: What are the most important factors to consider in developing a
bioassessment program?
A: Adherence to the Critical Elements Process in the design and implementation
of bioassessment protocols should provide an objective means of evaluating the
rigor of regulatory assessment programs and a basis for comparing data quality
254
A.L. Garey and L.A. Smock
information on index performance [21, 101, 106, 116].
The effectiveness of predictive models used for reference site classification is
often evaluated using cross-validation by constructing models using only subsets of
the available data. Leave-one-out cross-validation is a data-efficient method in
which reference sites are excluded from the dataset one at a time. After each
exclusion, the classification and modeling process is repeated. Each left-out site
is then classified using the model constructed without that site. The proportion of
agreements between original and cross-validated classifications, relative to the total
number of reference sites, is a measure of the effectiveness of the model. This
technique can be used with any site classification approach [114, 134, 144]. Null
models, which are constructed by predicting metric values across all reference sites,
with no classification, are useful for evaluating all types of classifications. If
classification strength does not exceed that of the null model, then the classification
provides no advantage [145].
5 Expert Interviews: Challenges and Important
Considerations in Bioassessment
To provide a broader and more comprehensive perspective, we conducted interviews with three expert practitioners and developers of bioassessment programs.
Their responses to our questions, provided here in a question-and-answer format,
have been summarized with a focus on emerging issues relevant to bioassessment in
aquatic systems.
Expert 1 Michael Barbour, Ph.D.—Adjunct Senior Scientist, Mote Marine Laboratory, Sarasota, FL, USA, and retired Director, Center for Ecological Sciences,
Tetra Tech, Inc., Owings Mills, MD, USA
Q: What factors limit the potential for increased use of genetic information in
bioassessment surveys? Is it likely that molecular genetic analysis will replace
traditional taxonomic approaches, or will these processes be used in conjunction
with each other?
A: A major challenge in the use of genetic data for bioassessment will be in
determining how reference conditions are expressed and how to account for the
effects of natural environmental variability on reference populations. It is unlikely
that genetic analysis will replace traditional taxonomic approaches in the near
future. Evolving DNA methods, however, should help to decrease taxonomic
uncertainty and improve our evaluations of aquatic assemblages.
Q: What are the most important factors to consider in developing a
bioassessment program?
A: Adherence to the Critical Elements Process in the design and implementation
of bioassessment protocols should provide an objective means of evaluating the
rigor of regulatory assessment programs and a basis for comparing data quality
254
A.L. Garey and L.A. Smock