Abstract
A dispersion model has been presented which directly uses measured head data at an array of monitoring wells. The model divides the measured data into coherent and noncoherent components. The noncoherent components are formulated into a sub-grid dispersion coefficient, while the coherent components are used to estimate the seepage velocity. The model has been verified in a case where an analytic solution is available. The model has been validated by simulating the instantaneous release of a tracer at a site where 13 years of synoptic daily head measurements are available at several monitoring wells. The growth of the longitudinal dispersivity of the tracer with length scale compares favorably with field measurements. The results show that for large length scales the model may be considered nonempirical since the dispersion eventually becomes insensitive to the local dispersivity.
