Search

• Innovate a BaPMI method for macro-scale distributed hydrologic modeling. • Apply it to a typical large cold-region watershed, Athabasca River Basin in Canada. • Verify feasibility & advantages in characterizing macro-scale cold-region hydrology. • Quantify parametric & hyperparametric sensitivities for facilitating its applications. • Reveal variabilities of climate-driven cross-scale uppermost hydro-variational effects. At macroscales (e.g., large river basins, nations, continents, or the globe), hydro-hazard mitigation under climate change requires accurate, applicable hydrologic models, especially over cold regions of complicated hydroclimatic relations. As an effort to address such a challenge, we innovate a macroscale distributed hydro-modeling method, Bayesian principal-monotonicity inference (BaPMI), based on climate classification, representative-grids selection, statistical hydrology, surrogates-based Bayesian optimization, and variance-based sensitivity analyses. The method is applied to a typical large cold-region watershed, i.e., Athabasca River Basin in Canada, and reveals a series of findings such as the following representatives. BaPMI shows promising skills in quantitatively characterizing macroscale cold-region hydrology under overall impacts of heterogeneous climates. Bayesian optimization of which hyperparameters are insensitive decreases hydro-modeling computational time by up to 96.3% and, together with climate classification, largely enhances BaPMI applicability while ensuring modeling accuracies. Individual BaPMI hydro-model parameters can explain over ¾ of the variation of hydro-modeling accuracies and, due to their interaction, joint calibration is required for avoiding underestimation of the impacts. Climatic conditions dominating cross-scale uppermost hydrologic variations consist of night temperature (40%), day temperature (38%), and precipitation (22%) for all catchments; all river flows along the mainstem are associated with upstream climatic events such as glacier melts. Climatic impacts decline from upstream to downstream and from summer, spring, winter to autumn; this may relate to tempo-spatial heterogeneity of soil, vegetation, geology, human interference and other non-climatic factors. Without this study, macroscale distributed hydrologic modeling would lack one advanced method (i.e., BaPMI) that can immunize against effects of data uncertainty, subjective judgement and climatic collinearity, adapt to complicated hydroclimatic relations and diverse hydro-variable(s) distributions, reveal cross-scale uppermost hydrologic variations and, through inexpensive computations, enhance hydro-modeling accuracies to avoid climatic-impact underestimation. The findings of this study could facilitate BaPMI applications and advance macroscale (cold-region) hydrology. [ABSTRACT FROM AUTHOR]