This study was conducted at an oil sands operation in the Athabasca Oil Sands Region (AOSR), northeastern Alberta, Canada. The mine comprises open pit excavation of bituminous sands at two sites (Mildred Lake, ML, and Aurora North, AN), with a single hot-water extraction circuit connecting extraction plants at each mine. Water samples were collected and analyzed regularly over an eight-year period to establish inventories of site-wide water isotope signatures including seasonal and interannual changes in the recycle water circuit, and to permit future application of an isotope balance model to constrain poorly quantified processes such as evaporation losses, dewatering of tailings, and tailings pond connectivity of the recycle water circuit. Sampling of precipitation inputs over an 8-year period was used to constrain a local meteoric water line for the area. Differences in evaporative isotopic enrichment of tailings ponds at ML and AN are attributed to use of Athabasca River makeup water at the former site versus basal dewatering sources at the latter, with similar atmospheric controls at both. A conceptual model is developed summarizing temporal variations in water balance and isotopic signatures within the recycle water circuit, including accurate simulation of the unique isotopic enrichment of cooling tower blowdown. This study provides foundational evidence for application of stable isotope mass balance to monitor and improve industrial water use efficiency and management. • Detailed summary of stable isotope variations at oil sands mine sites. • New dataset for precipitation, makeup water, and mine circuits. • Updated regressions defining local meteoric water line for district. • Contrasts isotopic variations for nearby mine sites with distinct sources. • Previously unpublished effects of cooling tower blowdown.

The oil sands industry in Canada uses soil–vegetation–atmosphere-transfer (SVAT) water balance models, calibrated against short-term (<ã10 years) field monitoring data, to evaluate long-term (ã60 years) reclamation cover design performance. These evaluations use long-term historical climate data; however, the effects of climate change should also be incorporated in these analyses. Although statistical downscaling of global climate change projections is commonly used to obtain local, site-specific climate, high resolution dynamical downscaling can also be used. The value of this latter approach to obtain local site-specific projections for mine reclamation covers has not been evaluated previously. This study explored the differences in key water balance components of three reclamation covers and three natural sites in northern Alberta, Canada, under future, site-specific, statistical, and dynamical climate change projections. Historical meteorological records were used to establish baseline periods. Temperature datasets were used to calculate potential evapotranspiration (PET) using the Hargreaves–Samani method. Statistical downscaling uses the Long Ashton Research Station Weather Generator (LARS-WG) and global circulation model (GCM) projections of temperature and precipitation. Dynamical climate change projections were generated on a 4 km grid using the weather research and forecasting (WRF) model. These climate projections were applied to a physically-based water balance model (i.e. Hydrus-1D) to simulate actual evapotranspiration (AET) and net percolation (NP) for the baseline and future periods. The key findings were: (a) LARS-WG outperformed WRF in simulating baseline temperatures and precipitation; (b) both downscaling methods showed similar directional shifts in the future temperatures and precipitation; (c) this, in turn, created similar directional shifts in future growing season median AET and NP, although the increase in future NP for LARS-WG was higher than that for WRF. The relative increases in future NP were much higher than the relative increases in future AET, particularly for the reclamation covers.