Evaluation of analytical and numerical approaches for the estimation of groundwater travel time distribution

SummaryIt is critical that stakeholders are aware of the lag time necessary for conservation practices to demonstrate a positive impact on surface water quality. For solutes like nitrate that are transported primarily by the groundwater pathway, the lag time is a function of the groundwater travel time distribution (TTD). We used three models of varying levels of complexity to estimate the steady-state TTD of a shallow, unconfined aquifer in a small Iowa watershed: (a) analytic model, (b) GIS approach, and (c) MODFLOW model. The analytic model was the least input-intensive, whereas the GIS and MODFLOW approach required detailed data for model development. The resulting TTDs displayed an exponential distribution with good agreement among all the three methods (mean travel times ranging from 16.2 years in the analytic model, 19.6 years in GIS model and 20.5 years in MODFLOW model). The greater deviation in the analytic model was attributed to the difficulty in estimation of a representative saturated thickness in an unconfined aquifer. The correspondence between the spatial travel time distributions generated by GIS and MODFLOW was a function of the landscape position, with greater correspondence in uplands compared to floodplains. In the floodplains the land surface slope is a poor approximation of the water table gradient that is captured by the MODFLOW model but not the GIS that uses the land surface as a surrogate for the water table. Study results indicate that except for cases where there are marked differences between water table surface and land surface, simpler approaches (analytic and GIS) can be used to estimate TTDs required for the design and optimal placement of conservation practices and communicating lag times issues to the public.

► Travel time distribution (TTD) of an unconfined aquifer estimated by three models.
► Models in order of increasing complexity: analytic, GIS, MODFLOW.
► All three models indicated exponential TTDs.
► Good agreement among three models suggests applicability of simpler approaches.
► Spatial TTD maps correlated well in uplands, but poorly in floodplains.