"APPROXIMATING EFFECT LEVELS OF NOECS BY A BIC DECISION TREE FOR POPULATION-LEVEL ECOLOGICAL RISK ASSESSMENT ON CHEMICALS"

Meng Yaobin, Lin Bin-Le 

National Institute of Advanced Industrial Science and Technology

 EcoSummit2007 (beijing 2007/5/23)

---

Abstract

In order to bridge the chasm between No-Observed-Effect-Concentration (NOEC) and concentration-response relationship, this research built a decision tree with Bayesian Information Criterion (BIC) to approximate the effect level or response of an NOEC. NOECs records were collected from open literature, and the effect level of each NOEC was calculated from the original ecotoxicity test. Analysis of variation indicated that organism type, measurement endpoint, mode of action, replication number, dilution ratio, and ratio of EC50 to NOEC were significantly contributing to the effect level for an NOEC. In the decision tree developing process, these factors were recursively applied to branch the NOECs data set so as to refine the variation of the NOEC responses based on increasing context information. The best three decision trees were selected and applicable for approximating concentration-response relationships. This work will allow NOECs to help the population-level ecological risk assessment of chemicals.

Keywords

Bayesian Information Criterion, Concentration-Response Relationship, Decision Tree, No Observed Effect Concentration