Hydrological Parameter Estimation (HYPE) System for Bayesian Exploration of Parameter Sensitivities in an Arctic Watershed

As part of a study on how vegetation water use and permafrost dynamics impact stream flow in the boreal forest discontinuous permafrost zone, a Bayesian modeling framework has been developed to assess the effect of parameter uncertainties in an integrated vegetation water use and simple, first-order, non-linear hydrological model. Composed of a front-end Bayes driver and a backend interactive hydrological model, the system is meant to facilitate rapid execution of seasonal simulations driven by hundreds to thousands of parameter variations to analyze the sensitivity of the system to a varying parameter space in order to derive more effective parameterizations for larger-scale simulations. The backend modeling component provides an Application Programming Interface (API) for introducing parameters in the form of constant or time-varying scalars or spatially distributed grids.

In this work, we describe the basic structure of the flexible, object-oriented modeling system and test its performance against collected basin data from headwater catchments of varying permafrost extent and ecosystem structure (deciduous versus coniferous vegetation). We will also analyze model and sub-model (evaporation, transpiration, precipitation and streamflow) sensitivity to parameters through application of the system to two catchment basins of the Caribou-Poker Creeks Research Watershed (CPCRW) located in Interior Alaska. The C2 basin is a mostly permafrost-free, south facing catchment dominated by deciduous vegetation. The C3 basin is underlain by more than 50% permafrost and is dominated by coniferous vegetation. The ultimate goal of the modeling system is to improve parameterizations in mesoscale hydrologic models, and application of the HYPE system to the well-instrumented CPCRW provides a valuable opportunity for experimentation.