EXTENDED KALMAN FILTERING TO IMPROVE THE ACCURACY OF A SUBSURFACE CONTAMINANT TRANSPORT MODEL

摘要

Modeling the behavior of contaminants during thernsubsurface flow through soil is important in predicting thernfate of the pollutants, in risk assessment, and as arnpreliminary step of the mitigation process. A tworndimensional transport model with advection andrndispersion is used as the deterministic model of arnconservative contaminant transport in the subsurface.rnWith the system model alone it is very difficult to predictrnthe true dynamic state of the pollutant. Therefore, wernneed observation data to guide the deterministic systemrnmodel to assimilate true state of contaminant. AnrnExtended Kalman Filter (EKF), which is essentially a firstrnorder approximation to an infinite dimensional problem,rnis used to predict the contaminant plume transport. Arntraditional root mean square error (RMSE) of pollutantrnconcentrations is used to examine the effectiveness of thernEKF in contaminant transport modeling. The experimentrnshows that EKF can reduce 74 to 91% of prediction errorsrncompare to the numerical method while working with thernfull set of observation data and comparing to thernanalytical solution.