A nonparametric Bayesian approach for quantifying herbicide exposure in streams

EPA Grant Number: R827898
Title: A nonparametric Bayesian approach for quantifying herbicide exposure in streams
Investigators: Qian, Song S. , Pan, Yangdong , Pratt, James R.
Institution: Portland State University
EPA Project Officer: Fields, Nigel
Project Period: November 26, 1999 through November 25, 2001 (Extended to November 25, 2002)
Project Amount: $166,519
RFA: Environmental Statistics (1999) RFA Text |  Recipients Lists
Research Category: Health , Ecosystems , Environmental Statistics


We propose a nonparametric Bayesian approach for detecting and quantifying low-level herbicide concentrations and their cumulative effects in agricultural basin streams using benthic algae. Regulations on pesticide usage depend largely on the knowledge of human/ecosystem exposure of a particular pesticide. However, characterizing low level pesticide concentrations in streams is a daunting task due to the high cost related to the necessary high sampling frequency and high method detection limits.


Using benthic algae species as indicators of low level pesticide exposure is an attractive alternative, since each species respond to herbicides according to their unique physiological characteristics and interactions with other species. Species composition will reflect and integrate the effect of ambient herbicide concentrations. The existing method of using algae composition data to infer environmental variables is stemmed from paleolimnology studies of historical lake acidification, where diatom species composition is used to reconstruct historical lake-water pH based on the correspondence of present day lake-water pH and diatom species composition. The method is based on the assumption that the species response curve (species abundance plotted against the pH gradient) is a unimodal, bell-shaped curve proportional to a Gaussian density function (the normality assumption). This assumption is, however, not justified when applied to environmental variables other than pH. The proposed method is a combination of the Bayesian nonparametric binary regression model developed by the PI and a maximum likelihood estimator. The method is nonparametric; therefore, the response curve is estimated from data. The Bayesian nature of the method allows continuous updating of the posterior model when new data are available. The proposed project includes method development, data collection, and a small scale laboratory experiment. We propose to use atrazine as the target herbicide, and demonstrate how the proposed method can be used to quantify the atrazine concentrations using algal species response curves.

Expected Results:

The proposed research integrates ecology, ecotoxicology, and statistical modeling. The proposed Bayesian model of algal responses to herbicide concentrations can be used to infer low-level atrazine concentrations in streams- based on algal species composition alone. Consequently, biomonitoring of herbicides in agricultural basin streams can be done routinely at a low cost. The model can also be used as a tool to assess or evaluate the effectiveness of new management and remediation practices to ameliorate adverse effects of herbicides on surface water resources.

Supplemental Keywords:

RFA, Economic, Social, & Behavioral Science Research Program, Scientific Discipline, Waste, Ecosystem Protection/Environmental Exposure & Risk, Remediation, Environmental Chemistry, Ecosystem/Assessment/Indicators, Ecosystem Protection, Ecological Effects - Environmental Exposure & Risk, Ecological Effects - Human Health, Ecological Risk Assessment, Ecology and Ecosystems, Environmental Statistics, Ecological Indicators, ecological exposure, agricultural basin streams, ecosystem valuation, nonparametric Bayesian approach, risk assessment, resource management, streams, ecological modeling, stream ecosystems, herbicide exposure, environmental risks, pesticides, modeling, surface water, statistical models, cost-benefit analysis, herbicides, modeling ecological risk, benthic algae, innovative statistical models, regulations