Uncertainty assessment in the regionalisation of hydrological signatures

Abstract:

Reliable information about the hydrological behaviour of an ungauged catchment is needed for a wide range of water-resource management purposes, and it has been a central topic of research in hydrology for the last decade through the Predictions in Ungauged Basins initiative. Such information derived from observed data is known as a hydrological signature, and has been used in a variety of studies for, e.g., change detection, catchment classification, and model calibration. When signature values are predicted for ungauged catchments, they are not only affected by uncertainties in the observational data for the gauged catchments used for the prediction, but also by uncertainties in the regionalisation procedure.

In this study we investigated a method for regionalisation of hydrological signatures to ungauged catchments that accounted for both of these uncertainty sources. This also enabled us to assess the role of the different uncertainty sources in defining the overall regionalisation uncertainty – e.g. for what signatures and conditions are the observational uncertainties more important than the regionalisation uncertainties and vice versa?

The study was made using an extensive dataset of catchments in England and Wales, incorporating gauging (stage-discharge) data from all the discharge stations. The uncertainties were assessed within a Monte Carlo framework that incorporated different types of uncertainties in the data as well as uncertainties in the regionalisation procedure.