Water use by anthropogenic activities in the face of climate change invokes a better understanding of headwater sources and lowland urban water allocations. Here, we constrained a Bayesian mixing model with stable isotope data (2018-2019) in rainfall (N = 704), spring water (N = 96), and surface water (N = 94) with seasonal isotope sampling (wet and dry seasons) of an urban aqueduct (N = 215) in the Central Valley of Costa Rica. Low δ 18O rainfall compositions corresponded to the western boundary of the study area, whereas high values were reported to the northeastern limit, reflecting the influence of moisture transport from the Caribbean domain coupled with strong orographic effects over the Pacific slope. The latter is well-depicted in the relative rainfall contributions (west versus east) in two headwater systems: (a) spring (68.7 ± 3.4 %, west domain) and (b) stream (55.8 ± 3.9 %, east domain). The aqueduct exhibited a spatial predominance of spring water and surface water during a normal wet season (78.7 %), whereas deep groundwater and spring water were fundamental sources for the aqueduct in the dry season (69.4 %). Our tracer-based methodology can help improve aqueduct management practices in changing climate, including optimal water allocation and reduced evaporative losses in the dry season.
Keywords: Aqueduct management; Bayesian mixing model; Costa Rica; hydrogen-2; isotope hydrology; oxygen-18; urban hydrology.