Graduate Research Talk: Predicting Source Zone Characteristics from Down-gradient Plume Response: A Machine Learning Approach

Abstract

Toxic spills can result in the presence of contaminants in the subsurface, which can pollute ground-water and have detrimental effects on the environment. The distribution of a contaminant in the subsurface resulting from a spill is influenced by a number of factors such as the spill-rate, type of contaminant and various soil properties. A particular contaminant distribution resulting from a spill is known as a source-zone. Remediation of the contaminant requires knowledge about the source zone architecture. The amount and spatial distribution of contaminant in water observed away from the source-zone (the down-gradient plume response) can be used for predicting the source-zone characteristics. Machine learning techniques such as classification and regression can be used for this purpose. This talk will focus on the application of different feature extraction techniques on down-gradient plume response data and the subsequent performance based on the classification framework.