Spatial variability of radon production rates in an alluvial aquifer affects travel time estimates of groundwater originating from a losing stream

Assessing water travel times in hyporheic and adjacent alluvial sediments is important to quantify exchange rates, biogeochemical turnover, and pollution dynamics across groundwater-surface water interfaces and in floodplain aquifers. Heat and 222Rn are useful natural tracers for this purpose. While heat transport is commonly simulated via the convection-conduction equation, the transport of 222Rn is often simulated assuming purely advective transport and a homogenous distribution of 222Rn production rates across the model domain. In the present study, we explicitly model the production, transport, and radioactive decay of 222Rn after surface water infiltration into an alluvial aquifer with the numerical models MODFLOW-NWT and MT3D-USGS. Using field observations, numerical modeling, and laboratory experiments, we show that 222Rn production rates vary substantially in the alluvial aquifer of a lowland river. Distributions of mean river-to-groundwater travel time in the alluvial aquifer, estimated via a Bayesian approach, shifted considerably depending on whether 222Rn and time series of groundwater heads and temperature were jointly inverted and whether 222Rn production rates were allowed to vary spatially. With distance to the river, differences between the median river-to-groundwater travel time and apparent 222Rn ages increased by factors ranging from 1.4 at 4 m to 11.9 at 59 m, a finding that highlights the need to simulate 222Rn transport explicitly. The joint inversion of 222Rn, groundwater heads, and temperature reduced uncertainty associated with mean travel times, suggesting that the uncertainty introduced by spatially varying 222Rn production rates can at least partly be compensated by using a combination of natural tracers.