Abstract

Water-related disasters such as floods and droughts highlight the urgent need for securing water resource systems for human and ecosystem utilizations. Increasing anthropogenic interventions along with climate variability and change have exacerbated the intensity and frequency of such water-related events, which will continue to increase in the future. Such pressures introduce substantial and unprecedented vulnerability to water resource management. Understanding the extent of potential vulnerabilities, however, is not trivial due to the uncertainty in current top-down impact assessments. To address current limitations, bottom-up frameworks have been proposed in the past decade to provide alternatives to top-down and scenario-led vulnerability assessments. The core idea behind bottom-up schemes is to analyze the potential impacts directly as a function of potential changes in streamflow conditions through a systematic stress testing scheme. To make such stress tests reliable, systematic methodologies are needed to synthesize streamflow, and other hydroclimatic variables, beyond the historical observations. Despite ongoing advances in stochastic streamflow generations under stationary conditions – with which the vulnerability assessment can be performed – little attention has been given to advancing the perturbation algorithms for altering the streamflow characteristics under nonstationary conditions; and in fact, only a few incorporate climate-related proxies into streamflow generation. This thesis aims to shed light on some limitations of bottom-up approaches and propose an improved stochastic streamflow generation framework for impact assessment in water resources systems under changing streamflow conditions. This takes place through: (1) Identifying uncertainties in current stochastic streamflow generation approaches as well as how and why these uncertainties matter to bottom-up impact assessment; (2) providing a guideline on the choice of the optimal scheme(s) for stochastic generation of streamflow series in various temporal and spatial scales; (3) proposing a methodology to incorporate the effect of large scale climate indices in stochastic streamflow generation; (4) identifying the types of changes in the streamflow regime through a systematic and globally-relevant approach; as well as (5) proposing a generic algorithm to shift a wide range of streamflow characteristics in streamflow time series, and to make a transient and non-stationary flow generation. This research results in an improved stochastic streamflow generation scheme capable of generating scenarios of change under nonstationary conditions. The skill of the proposed algorithm is assessed over multiple natural streams, showing good performance in representing the plausible changes required for the vulnerability assessment of water resource systems.