5/19/2015  |   10:30 - 10:45   |  103C

UNDERSTANDING AGRICULTURAL LAND USE DISTURBANCE THROUGH A SERIES OF MODELS: LANDSCAPE TO WATER QUALITY TO INVERTEBRATES. In 2013, the U.S. Geological Survey and the U.S. Environmental Protection Agency sampled 100 streams across 11 States in the Midwest corn belt of the U.S. The ecological condition of streams were assessed in relation to flow, suspended sediment, nutrients, major ions and 230 dissolved pesticides and degradates collected weekly for 12 weeks prior to habitat and algae, invertebrate, and fish community sampling. The effects of various stressor metrics and time windows on macroinvertebrates assemblage metrics were assessed using response models developed for various hierarchical pathways: land use to the stressors, stressors to ecological condition, and then combining results from all models to highlight causal pathways and interactions. Boosted Regression Tree models are compared against alternative modeling techniques for predicting macroinvertebrate metrics across an agricultural disturbance.

Ian Waite (Primary Presenter/Author), U.S. Geological Survey, Portland, OR, iwaite@usgs.gov;


Travis Schmidt (Co-Presenter/Co-Author), USGS WY-MT Water Science Center, tschmidt@usgs.gov;


Mark Munn (Co-Presenter/Co-Author), U.S. Geological Survey, Tacoma, WA, mdmunn@usgs.gov;


Pete VanMetre (Co-Presenter/Co-Author), U.S. Geological Survey, Austin, TX, pcvanmet@usgs.gov;

5/19/2015  |   10:45 - 11:00   |  103C

DEVELOPMENT AND USE OF A PERCENT MODEL AFFINITY FOR ASSESSMENT OF PUERTO RICO STREAMS Puerto Rico currently lacks a stream monitoring program with direct assessment and reporting on biological conditions. This is partly attributed to a lack of development of biological assessment protocols applicable to Caribbean streams. Measures of macroinvertebrate community composition including percent model affinity (PMA) have been developed and successfully applied in temperate North America. Macroinvertebrate data collected 2006 to 2011 from riffle habitat at 41 reference stream sites, were used to develop a PMA for use on high gradient streams in Puerto Rico. The expected community composition was developed and represented by eight major organism groups. Overall, the PMA showed a moderate to strong response across environmental gradients related to land use, water chemistry, and physical habitat. The PMA was found to be correlated to a multimetric macroinvertebrate index that was recently developed for Puerto Rico streams. Use of the PMA for macroinvertebrate community composition in Puerto Rico would allow for determination of biological conditions without the need for extensive information related to invertebrate taxonomy, functional feeding group classification, and organism pollution sensitivity.

James Kurtenbach (Primary Presenter/Author), USEPA Region 2, kurtenbach.james@epa.gov;

5/19/2015  |   11:00 - 11:15   |  103C

EFFECTIVE VISUALIZATIONS OF COMPLEX BIOASSESSMENT INDICES BASED ON PREDICTIVE MODELS Effective visualizations can make complicated assessment indices accessible to general audiences. For example, indices based on predictive models are increasingly common in bioassessment applications because of their ability to set different site-specific benchmarks based on environmentally similar reference sites. However, the complexity of these indices may limit their adoption by audiences that lack training in statistics or stream ecology. The complex mathematics of an assessment tool need not prevent the use of effective tools in watershed management because effective data visualization can make an index more easily interpreted. We present a few visualization methods that transform the outputs of indices based on predictive models (both an O/E and multimetric index) into more easily understood graphics. These visualizations address common questions from data users, such as: Which reference sites are most relevant to my sites? How do expectations at my sites differ? And how close are my sites to meeting their biological expectations? Although our examples are specific for an index developed for California, we think these types of visualizations are broadly applicable to many types of indices.

Raphael Mazor (Primary Presenter/Author), Southern California Coastal Water Research Project, raphaelm@sccwrp.org;


Mark Engeln (Co-Presenter/Co-Author), The Southern California Coastal Water Research Project, mengeln@sccwrp.org;


Eric Stein (Co-Presenter/Co-Author), Southern California Coastal Water Research Project, erics@sccwrp.org;


Peter Ode (Co-Presenter/Co-Author), California Department of Fish and Wildlife, Peter.Ode@wildlife.ca.gov;

5/19/2015  |   11:15 - 11:30   |  103C

STRIVING FOR A BETTER MODEL TO ASSESS LAKE BIOLOGICAL CONDITION: A COMPARISON OF CART, RANDOM FOREST AND MULTIPLE LINEAR REGRESSION Ecological assessments with site-specific models of expected condition can improve assessments by accounting for natural variability at finer spatial scales than regional characterizations of reference condition. We compare ecological assessments using a multimetric index (MMI) and three approaches for calculating site-specific models for expected metric values: conventional multiple linear regression (MLR); classification and regression trees (CART); and random forest regression (RF). We used diatom assemblage data collected during the USEPA National Lakes Assessment in summer 2007 for our comparison. Boxplots of MMI showed both CART and RF models had greater separation power than the MLR model, which was assessed using overlaps between interquartile ranges and medians among reference, moderately disturbed and disturbed sites. The root mean squared error for the predictions of expected metric values at reference sites was slightly greater for CART model (0.33) than with random forest model (0.25). We conclude that CART worked sufficiently well that the trade-off between prediction accuracy and interpretability of CART versus RF models may become the deciding factor on which modeling approach to use.

Bo Liu (Primary Presenter/Author), Hebei University, liubo3@msu.edu;


Jan Stevenson (Co-Presenter/Co-Author), Michigan State University, rjstev@cns.msu.edu;

5/19/2015  |   11:30 - 11:45   |  103C

DEVELOPING GREAT LAKES BIOINDICATORS OF ENVIRONMENTAL CONDITION AND RECOVERY FROM DEGRADATION WITH REFERENCE TO WATERSHED BASED STRESS Bioassessment typically entails comparing a test site to the reference defined by characteristics of 'best available' sites and associated biota, and the complementary ‘degraded condition’ (sites whose environmental characteristics are deemed unacceptable (‘most disturbed’) by consensus. We derived taxon-specific bioindicators of reference-degraded conditions at Great Lakes coastal margins (assemblages of birds, aquatic vegetation, fishes, aquatic invertebrates and diatoms). Titan threshold analyses of taxon losses or gains often identified 2 thresholds on a stress gradient. At one, many sensitive species disappeared, suggesting biodiversity loss; at another tolerant taxa increasingly dominated. All assemblages were affected at approximately the same threshold, suggesting significant ecosystem functional alteration at these points. Biological indices can be calibrated to identify these critical points as “biological criteria”. We propose that the non-degraded/degraded threshold be a suitable operational target to define the boundary between impaired and non-impaired conditions needed to delist Beneficial Use Impairments at AOCs. The reference/non-reference threshold may be a suitable operational target to define the boundary between biodiverse and less biodiverse conditions.

Jan Ciborowski (Primary Presenter/Author), Department of Biological Sciences, University of Windsor, cibor@uwindsor.ca;


Katya Kovalenko (Co-Presenter/Co-Author), Natural Resources Research Institute, Univ. Minnesota Duluth, philarctus@gmail.com;


George Host (Co-Presenter/Co-Author), Natural Resources Research Institute – Univ. Minnesota Duluth, ghost@d.umn.edu;


Robert Howe (Co-Presenter/Co-Author), Department of Natural and Applied Sciences, University of Wisconsin Green Bay, hower@uwgb.edu;


Euan Reavie (Co-Presenter/Co-Author), Natural Resources Research Institute – U. Minnesota Duluth, ereavie@d.umn.edu;


Terry Brown (Co-Presenter/Co-Author), US Environmental Protection Agency, brown.terry@epa.gov;


Valerie Brady (Co-Presenter/Co-Author), Natural Resources Research Institute, University Minnesota Duluth, vbrady@d.umn.edu;


Nicholas Danz (Co-Presenter/Co-Author), Department of Natural Sciences, University of Wisconsin Superior, ndanz@uwsuper.edu;


Gerald Niemi (Co-Presenter/Co-Author), Natural Resources Research Institute, University Minnesota Duluth, gniemi@d.umn.edu;


Meijun Cai (Co-Presenter/Co-Author), Natural Resources Research Institute - U. Minnesota Duluth, mcai@d.umn.edu;


Lucinda Johnson (Co-Presenter/Co-Author), Natural Resources Research Institute, University of Minnesota Duluth, ljohnson@d.umn.edu;

5/19/2015  |   11:45 - 12:00   |  103C

MAPPING THE BIOLOGICAL CONDITION OF USA RIVERS AND STREAMS We predicted the probable (pr) biological condition (BC) of ~5.4 million km of stream within the conterminous USA (CONUS). National maps of prBC could provide an important tool for prioritizing monitoring and restoration of streams. The USEPA uses a spatially balanced survey design to estimate the proportion of streams that fail to support healthy BC, but does not infer BC at un-sampled locations. To model BC, we developed a GIS database of >100 anthropogenic and natural watershed metrics for streams within the CONUS. We combined these data with 1,883 USEPA-sampled streams that were previously assessed as having ‘good’ or ‘poor’ BC. prBC was best predicted (70% correctly classified) with random forests using elevation, % riparian naturalness, population density, air temperature, watershed % forest and % agriculture as predictors. National maps of prBC provided a unique assessment of model performance. Specifically, lower prBC was consistent with large-scale patterns of human-related land use. However, local prBC was sometimes unrealistic, suggesting that predictions could be improved with regional, rather than national, models. Models will soon be extended to also include 356,044 lakes.

Ryan Hill (Primary Presenter/Author), US Environmental Protection Agency, hill.ryan@epa.gov;
Ryan Hill is an aquatic ecologist with the U.S. EPA Office of Research and Development. He is interested in how watershed conditions drive differences in freshwater diversity and water quality across the United States. He has worked extensively with federal physical, chemical, and biological datasets to gain insights into the factors affecting water quality and biotic condition of freshwaters across the conterminous US. He has also worked to develop and distribute large datasets of geospatial watershed metrics for streams and lakes for the Agency (EPA’s StreamCat and LakeCat datasets). Ryan completed his PhD in Watershed Ecology at Utah State University in 2013 with Dr. Chuck Hawkins. He was an ORISE postdoctoral fellowship at the U.S. EPA from 2014-2019 before joining the EPA in 2019.

Marc Weber (Co-Presenter/Co-Author), US EPA, Pacific Ecological Systems Division, Corvallis, OR, weber.marc@epa.gov;


Scott Leibowitz (Co-Presenter/Co-Author), US EPA, Pacific Ecological Systems Division, Corvallis, OR, leibowitz.scott@epa.gov;


Anthony Olsen (Co-Presenter/Co-Author), EPA, olsen.tony@epa.gov;