Introduction to the November 2005 PREDictive Assessment Tool for ORegon (PREDATOR)
Assessing the biotic integrity of Oregon streams

Why macroinverts? * Predictive modeling * Null models
METHODS (model requirements and how they were built

Predictive models, based on multivariate statistics, are one tool used to assess biological integrity. The other common assessment method is the multimetric approach.  DEQ uses the multimetric Index of Biotic Integrity (IBI) for assessing fish and aquatic vertebrate assemblages. In the past, DEQ used IBIs to assess macroinvertebrate assemblages in Oregon. But the IBIs were based on older and much smaller datasets than are now available for the development of predictive models. Using all the available data, the predictive models were constructed for much broader applicability and potentially greater sensitivity than the previously available IBIs

The PredictiveAssessment Tool for Oregon (PREDATOR) consists of three regional models that assess the biological integrity of wadeable streams across Oregon. The Oregon Department of Environmental Quality (DEQ) developed the models with the intent of supplying a scientifically rigorous bioassessment tool that is easy to use and understand by a large audience.  The overall goal is to promote better understanding of the conditions of Oregon’s streams.

STARTING NOTE: To actually use your data in these models, the samples must be collected with the same field protocols and identified to at least the taxonomic level used in building the models.

Why Macroinvertebrates?

Macroinvertebrates include freshwater insects, crustaceans, snails, clams, etc.  They occupy a central role in food chains and ecosystem processes, are easy to collect, are relatively inexpensive to process and analyze, and show strong responses to many stressors. Macroinvertebrates are the most commonly used aquatic communities for assessing stream biological integrity. 

Predictive Modeling

PREDATOR is a predictive model built from the concepts of the RIVPACS (River InVertebrate Prediction and Classification System) approach, first developed in Britain in the 1980s.  Excellent general overviews of the RIVPACS approach to predictive modeling are the Western Center for Monitoring and Assessment of Freshwater Ecosystems and the Centre for Ecology and Hydrology.   

RIVPACS-type predictive models use a large database of reference or least-disturbed sites to predict the macroinvertebrate taxa that would be expected to occur at a test site. The model compares the observed (collected) taxa at a site to the list of expected taxa. Model output is the observed to expected (O/E) ratio.  In simple terms, these are “taxa loss” models.  O/E scores lower than 1.0 represent a loss of native reference taxa. 

Null Models

In some cases, such as too few reference sites or environmental variables that do not allow for more accurate predictions, an effective predictive model can't be developed for an area. In such cases (like the Northern Basin and Range ecoregion of southeast Oregon), null models  provide a simple way to get a measure of O/E.  A null model does not cluster reference sites into multiple groups, and the expected taxa list (E) is simply those taxa that occur at greater than 50% of the reference sites.  The observed value is those expected taxa that were collected, or observed, at a site. Null models do not provide the rigorous analysis that predictive models, but would be useful for simple comparisons where other models aren't available.

The null model taxa list for Oregon's Northern Basin and Range ecoregion is:
Baetis, Chironominae, Optioservus, Orthocladiinae, Rhyacophila, Trombidiformes, Diphetor_hageni, Epeorus, Zaitzevia, and Brachycentrus

METHODS FOR MODEL DEVELOPMENT

There were five main steps to developing the PREDATOR models:

1)      Setting consistent sampling protocols and collection periods

2)      Selecting regional reference sites

3)      Grouping reference sites based on the macroinvertebrate communities

4)      Relating reference groups to predictor variables

5)      Assessing model performance

Home • PREDATOR • Macro taxonomy • Macro input file • Habitat input file • Run the models
Interpretation • Resources • Thanks

Copyright © 2006, The Xerces Society www.xerces.org; please send comments to info@xerces.org