The Toxic Release Inventory (TRI) is used for purposes ranging from estimation of trend to environmental justice. It was created to force public disclosure of releases of toxic chemicals to the air, water and land. Under a cooperative agreement with EPA's Office of Environmental Information, NISS studied the quality of TRI data.
Initially, attention was focused on anomalies in TRI data, many of which may have resulted from the data's being self-reported. These anomalies include order-of-magnitude annual changes in otherwise essentially flat time series, apparent shifts of the same amounts between reporting categories and clear indications of one year's reported releases being diverted from the previous year's.
Research issues included the automatic detection and characterization of such anomalies, which are quite different from ordinary statistical outliers, in order to both model the processes that generate them and construct inference methods that can accommodate them.
NISS was tasked with studying the quality of the data that was used in the U.S. EPA's Toxic Release Inventory.
Principal Investigator(s): Alan Karr