Selecting MODFLOW cell sizes for accurate flow fields

Contaminant transport models often use a velocity field derived from a MODFLOW flow field. Consequently, the accuracy of MODFLOW in representing a groundwater flow field determines in part the accuracy of the transport predictions, particularly when advective transport is dominant. We compared MODFLOW groundwater flow rates and MODPATH particle traces (advective transport) for a variety of conceptual models and different grid spacings to exact or approximate analytic solutions. All of our numerical experiments concerned flow in a single confined or semi-confined aquifer. While MODFLOW appeared robust in terms of both local and global water balance, we found that groundwater flow rates, particle traces, and associated groundwater travel times are only accurate when sufficiently small cells are used. For instance, a minimum of four or five cells are required to accurately model total groundwater inflow in tributaries or other narrow surface water bodies that end inside the model domain. Also, about fifty cells are needed to represent zones of differing transmissivities or an incorrect flow field and (locally) inaccurate groundwater travel times may result. Finally, to adequately represent leakage through aquitards or through the bottom of surface water bodies it was found that the maximum allowable cell dimensions should not exceed a characteristic leakage length lambda, which is defined as the square root of the aquifer transmissivity times the resistance of the aquitard or stream bottom. In some cases a cell size of one tenth of lambda is necessary to obtain accurate results.