An exploration of Bayesian identification of dominant hydrological mechanisms in ungauged catchments

Hydrological modelling of ungauged catchments, which lack observed streamflow data, is an important practical goal in hydrological sciences. A major challenge is to identify a model structure that reflects the hydrological processes relevant to the catchment of interest.

This study contributes a Bayesian framework for identifying individual model mechanisms (process representations) from flow indices regionalized to the catchment of interest. We extend a method previously introduced for mechanism identification in gauged basins, by formulating the inference equations in the space of (regionalized) flow indices and by accounting for posterior parameter uncertainty. A flexible hydrological model is used to generate candidate mechanisms and model structures, followed by statistical hypothesis testing to identify "dominant" (more a posterior probable) model mechanisms.

The proposed method is illustrated using real data and synthetic experiments based on 92 catchments from northern Spain, from which 16 catchments
are treated as ungauged. 624 hydrological model structures from the flexible framework FUSE are employed.

In real data experiments, the method identifies a dominant mechanism in 27% of 112 trials (processes and catchments). The most identifiable process is routing, whereas the least identifiable processes are percolation and unsaturated zone processes. In synthetic experiments, where "true" mechanisms are known, the reliability of method varies from 60% to 95% depending on the combined regionalization and hydrological error; the probability of making an identification remains stable at around 25%. More broadly, the study contributes perspectives on hydrological mechanism identification under data-scarce conditions; limitations and opportunities for improvement are outlined.