ALBERT

Description: Delineating spatial patterns of precipitation isotopes (“isoscapes”) is becoming increasingly important to understand the processes governing the modern water isotope cycle and their application to migration forensics, climate proxy interpretation, and ecohydrology of terrestrial systems. However, the extent to which these patterns can be empirically predicted across Canada and the northern United States of America (USA) has not been fully articulated, in part due to a lack of time-series precipitation isotope data for major regions of North America. In this study, we use multiple linear regressions of CNIP, GNIP and USNIP observations alongside climatological variables, teleconnection indices, and geographic indicators to create empirical models that predict the δ 18 O of monthly precipitation (δ 18 O ppt ) across Canada and the northern USA. Five regionalization approaches are used to separate the study domain into isotope zones to explore the effect of spatial grouping on model performance. Stepwise regression-derived parameterizations quantified by permutation testing indicate the significance of precipitable water content and latitude as predictor variables. Within the Canadian Arctic and eastern portion of the study domain, models from all regionalizations capture the inter- and intra-annual variability of δ 18 O ppt . The Pacific coast and northwestern portions of the study domain show less agreement between models and poorer model performance, resulting in higher uncertainty in simulations throughout these regions. Long-term annual average δ 18 O ppt isoscapes are generated, highlighting the uncertainty in the regionalization approach as it compounds over time. Additionally, monthly time-series simulations are presented at various locations, and model structure uncertainty and 90% bootstrapped prediction bounds are detailed for these predictions. This article is protected by copyright. All rights reserved.