| Summary: | 碩士 === 國立臺灣大學 === 環境工程學研究所 === 91 === A lot of uncertainty is associated with risk assessment of groundwater contamination. We attempt to integrate groundwater contaminant transport model into multi-pathway human health risk model. We develop the model for a case study,and then use random field generation and Monte Carlo method to analyze the uncertainty associated with risk assessment of groundwater contamination as well as quantify the sensitivity of parameters. The uncertainty should be reduced if the information increases. We study the relation between quantity of information and uncertainty by assuming that site information increases with increasing spatial resolutions. Generally, sampling is the primary way to access information. The study simulates a practical application by updating information with Bayesian approach. The Bayesian approach integrates the Monte Carlo method to quantify the effect of reducing uncertainty with increasing information.
The results show that the cancer risk is affected by parameters of groundwater contaminant transport model in this case study. The major influential parameters are hydraulic conductivity and fraction of organic carbon.
The study of hydraulic conductivity concludes that when the quantity of information increases with higher spatial resolution, the uncertainty of risk will be reduced with increasing resolution. The results of Bayesian approach with spatial resolution show that sampling data (likelihood) update the original data (prior). The average of variance coefficient (CV) of parameters will decrease with higher resolution if there are more sampling data in each section of site. The CV of cancer risk will be reduced with higher resolution if the posterior distribution is considered in the risk assessment model. When there is insufficient sampling in each section, the average of CV of parameters will not decrease with higher resolution and the CV of cancer risk will be reduced with smaller degree with higher resolution. Spatial resolution provides a method for quantifying the relation between uncertainty and quantity of information. With Bayersian approach, it can quantify the effect of reducing uncertainty with increasing information.
|
|---|
