Quantifying and partitioning evapotranspiration using Bayesian inversion of an isotope-enabled soil water balance model

Isotope ratios of soil water and atmospheric water vapor have been used to estimate soil evaporation fluxes and to partition evapotranspiration at local (plot, stand) scales, but the application of these methods has been limited by 1) challenges associated with data acquisition, and 2) the complexity of and lack of consensus about appropriate data interpretation methods. New initiatives that have expanded access to data, such as the U.S. National Ecological Observatory Network (NEON), are beginning to address the first of these limitations. In order to make progress toward the second, we link a model of soil water and water isotope balance, based on the widely used Noah land surface model, to a range of core NEON measurements and ancillary field-collected data using a Bayesian hierarchical framework. This model framework allows self-consistent treatment of the water and isotope cycles, including representation of uncertainties and differing assumptions, and simultaneous optimization of all model parameters conditioned on all data using Markov-Chain Monte Carlo sampling. We test the framework by applying it to estimate evapotranspiration partitioning at a dryland NEON site in central Utah and show that the posterior estimates give reasonable and useful constraints on flux rates and provide constraints on model parameters that could inform our understanding of soil properties and isotopic systematics in the system. This flexible framework for interpretation of water isotope data in evapotranspiration studies is amenable to application across ecosystems and at sites with different levels of data availability in support of cross-site syntheses and validation/testing of earth system models.