Utilizing coupled isotope-flow modelling to estimate temporal evapotranspiration partitioning in remote northern Canadian watersheds

Abstract:

Identification of source water and loss within northern watersheds is of significant importance due to multifaceted changes in physiography as a result of climate change within this region. Understanding the dominant processes is paramount to assess and anticipate the effects of these changes on the flow regime. Evapotranspiration loss has been identified as a primary component of the hydrologic cycle in northern Canadian watersheds and has high seasonal variability; however research within these remote watersheds is very limited. Coupled flow-isotope models have been used to improve process identification at the catchment scale, particularly using stable water isotopes which are affected by evaporation. Using a coupled flow-isotope model to temporally partition evapotranspiration, by exploiting the difference in fractionation of evaporation and transpiration, improved understanding of evapotranspiration in these watersheds is accomplished. Simulating the difference of evaporative fractionation of oxygen-18 and deuterium helps to constrain the output and reduce uncertainty. To assess the temporal evapotranspiration partition established by the coupled flow-isotope model, a dual isotope transit time model will be applied on a monthly time-step to compare using effective precipitation input. Results indicate that the partition of evapotranspiration follows a seasonal trend, consistent with the time of abscission in the watersheds and length of the growing season. Transpiration is the dominant portion of evapotranspiration during the growing season, however as temperatures decrease, vegetation retains moisture and the dominant partition changes to evaporation. Utilizing stable water isotopes has been shown to have good potential at identifying the partition with limited spatial and temporal data, however, assessment of model results against long term or higher spatial resolution data would further reduce uncertainty.