You are here:
Sensitivity analysis for an integrated avian fate and effects model
Citation:
Etterson, M., N. Galic, J. Carbone, AND V. Kurker. Sensitivity analysis for an integrated avian fate and effects model. SETAC North America, Sacramento, CA, November 04 - 08, 2018.
Impact/Purpose:
The MCnest model is used for ecological risk assessment of pesticides to birds. The recent release of V2.0 of MCnest includes the Terrestrial Investigation Model (TIM) as an exposure algorithm, in addition to the Terrestrial Residue Exposure Model (T-REX). Informed use of the model requires an understanding of the extent to which uncertainty in model parameters propagates to uncertainty about model risk predictions. The work described in this abstract will help risk assessors to understand this uncertainty when making inference about risk. It will also help to guide researchers in seeking to improve the quality of parameter estimates.
Description:
Ecological risk assessment models integrate information on exposure, toxicity, and life history to predict risk to ecological receptors. However, the underlying biological and ecological processes occur at different temporal rates and are often measured at different levels of resolution. These considerations of scale and measurement resolution influence the importance of different parameters for conclusions about risk that can be illuminated using model sensitivity analysis. We present a thorough sensitivity analysis of the MCnest model for each of these three categories of model parameters (life history, exposure, and toxicity). For toxicity we include two terrestrial exposure models, T-REX and TIM, which operate at different levels of complexity. Among life history parameters, nest survival rates and adult survival rates are the most influential parameters determining seasonal productivity. Among exposure and toxicity parameters, application rates, water solubility, and degradation rates were among the most important parameters. Different sensitivity metrics conveyed different information about the effects of perturbations in model parameters and were generally not correlated with each other. This highlights the importance of careful thought about what metric to choose to characterize model sensitivity.