Comparison of naturally-occurring chloride and oxygen-18 as tracers of interbasin groundwater transfer in lowland rainforest, Costa Rica

δ18O data were used to test a two-end-member hydrologic mixing model previously published for an area of lowland rainforest in Costa Rica. The model distinguishes interbasin groundwater transfer from locally-recharged groundwater. The model was originally based on data for chloride (Cl) and other major ions and is fully consistent with new Cl data presented here. The mixing model was used to generate five predictions concerning δ18O at 13 groundwater and surface water sampling sites, on the premise that the model would be supported if δ18O data agreed with the predictions. Overall, the δ18O data are consistent with the predictions, giving reasonably clear support for 3 of the 5, and neither clearly supporting nor ruling out the other 2. The large intra-site temporal variability in δ18O precludes conclusive confirmation of the latter two predictions. This variability also affects the results of mixing calculations to quantify the proportions of each end member in water samples representing mixtures of the end members. Mixing calculations based on δ18O give different results and much higher uncertainties than those based on Cl. For δ18O data, intra-site variability is large relative to the difference between the two end member waters, leading to large uncertainty in δ18O-based mixing calculations. In this case, Cl appears to be the superior tracer for interbasin groundwater transfer, mainly because intra-site variability in Cl is small relative to the large Cl difference between end members.