PREDICTION OF COUNTY-LEVEL NEW CONTAMINATION CASES FROM HISTORIC GROUNDWATER CONTAMINATION CASES THROUGH DATA DEPENDENT MODELING

摘要

Groundwater is one of the most important natural resources associated with the environment, public health, welfare, and long-term economic growth. Billions of populations rely on the groundwater for their daily activities. However, thousands of contamination cases have been documented in many groundwater reports. The primary contaminants are artificial products such as gasoline and diesel. To protect the important water resource, a series of efforts have been exerted including enforcement and remedial actions. Every year, the Texas Groundwater Protection Committee (TGPC) publishes a "Join Groundwater Monitoring and Contamination Report" described historic and new contamination cases in each county, which is an important data source for the prediction and design of prevention strategies. In this paper, a data dependent modeling (DDM) approach is proposed to predict the county-level new contamination cases (NCC). A case study with contamination information from Harris County in Texas was conducted to illustrate the modeling and prediction process with promising results. The one-step prediction error is 1.5%, while the two-step error is 12.1%. The established model is applicable for the use at county-level, state level, and even country level, while the prediction results could be a kind of reference during decision-making process.