The relative importance and characteristics of uncertainty in hydrology

Abstract

Quantifying uncertainty in hydrologic models is essential for their effective use and assessment. This thesis presents methods that can be used to estimate and evaluate uncertainty for hydrological modelling and contains four objectives. The first objective is to develop an uncertainty quantification metric known as the Quantile Flow Deviation (QFD) to characterise the relative contribution of different sources of uncertainty in hydrological modelling. Through this novel method it’s possible to identify the spectrum of uncertainty/error. The QFD method is applied to Australian and USA catchments to check the variability in the streamflow response. The results show that the contribution of model structure uncertainty is higher compared to structural identifiability/parameters and likelihood/objective functions.

The second part of the thesis characterises input uncertainty (rainfall and evapotranspiration) via QFD metric. Rainfall and evapotranspiration are key inputs for hydrologic models in studying catchment responses as well as climate scenarios. Four different catchments of Australia were analysed to observe the extent of input uncertainty. Both rainfall and evapotranspiration, however, are uncertain, with rainfall having a larger degree of uncertainty. Using uncertain inputs in hydrologic models, without due consideration of their associated uncertainties, results in biased outcomes.

The third part of the thesis demonstrates that an appropriate framework for estimating flow uncertainty is vital in quantifying potential flood damage, both for flood mitigation and management strategies. Design flood estimation is an essential step in evaluating the risk associated with the socioeconomic impacts of flood events in any location. However, the prediction or modelling of peak flows is subject to uncertainty associated with the selection of a hydrologic model structure and related model parameters. This research analyzes the spectrum of this model uncertainty via the QFD metric in association with other sources of uncertainty to explore the possible uncertainty in flood damage estimation.

Finally, this thesis investigates the impact of climate change uncertainty across different modelling domains focusing flood damage. As human induced climate change in many parts of the world, already create additional pressure on water resources systems, the only way of quantifying these changes is to use the outputs of Global Climate Models (GCMs). In this thesis GCMs output with three representative concentration pathways is used to estimate climate change uncertainty to understand likely changes in future water availability. Furthermore, how the uncertainty influences flood damage estimation with temperature increases (including those already occurred and likely before end of the century) is estimated.