- Title
- Pacific decadal variability: uncertainties and implications for water resources
- Creator
- Zhang, Lanying
- Relation
- University of Newcastle Research Higher Degree Thesis
- Resource Type
- thesis
- Date
- 2019
- Description
- Research Doctorate - Doctor of Philosophy (PhD)
- Description
- The influence of low-frequency climate variability, as represented by the Inter-decadal Pacific Oscillation (IPO) and Pacific Decadal Oscillation (PDO), collectively known as Pacific Decadal Variability (PDV), has been explored in a wide collection of studies. Statistical characteristics of rainfall and streamflow, changes of drought, flood and bushfire frequencies have been found to be associated with PDV in some regions. PDV runs refer to periods during which a PDV index lies above (or below) some threshold. These alternating runs have been shown to be associated with predominantly dry (or wet) hydrological conditions in some regions. The goal of this thesis was to understand the statistical characteristics of PDV runs and assess the value of incorporating PDV information in urban water supply system decision making. In pursuit of this goal, this thesis had five objectives The first objective was to identify the regions in eastern Australia, both east and west of the Great Divide, where PDV significantly influences rainfall and streamflow. It was shown using relative differences that the Australian Water Availability Project (AWAP) gridded rainfall and streamflow product can produce largely unbiased estimates of PDV influence. Furthermore, this thesis showed that rainfall and streamflow over a large portion of eastern Australian are affected by PDV phases with relative differences in mean streamflow exceeding 50% over much of the study region north of Canberra. These regions can potentially benefit from PDV-informed water supply decision making. The remaining four objectives worked towards defining this benefit. Because of the limited instrumental record, there are many palaeo reconstructions of PDV indices from which PDV runs can be extracted. Using these reconstructions, two fundamental questions of hydrological importance were asked: are PDV wet and dry phases statistically different and have PDV phases been stationary over the last millennium? The results suggest that there is no substantial evidence to support the assumption that run lengths have statistically different distributions during positive and negative PDV phases. Analysis based on three millennium-long reconstructions suggests that it is more likely than not, that the PDV run length has been non-stationarity in the past millennium. PDV palaeo reconstructions are subject to reconstruction errors, both systematic and random. However, understanding the impact of such errors is limited by the fact that “ground truth” is only possible in the period where palaeoclimate and instrumental data overlap. To develop a deeper insight on the role of such errors, synthetic stochastic experiments were designed to simulate the effect on errors on reconstructed PDV indices and PDV runs. The results showed that reconstruction errors cause the PDV run length standard deviation to be overestimated, in some cases very substantially, with the autocorrelation of the errors having the most significant impact on this overestimation. A new composite index method was introduced to reduce the impact of reconstruction errors. This method was shown to perform better than the composite index method developed by Henley et al. (2013). Using the new composite index method and ten reconstructions of different data lengths, the mean of PDV run length was estimated to be 20.8 years and standard deviation to be 13.7 years. When stochastic models are used to incorporate low and high-frequency climate information in drought risk assessment, a number of parameters must be estimated from limited historical data. These parameters include parameters to simulate PDV run lengths and streamflow conditioned on PDV phase. The impact of uncertainty in these parameters on drought risk assessment was investigated. A case study based on the Warragamba River, the main water source for Sydney, showed that uncertainty in drought risk is considerable and that uncertainty in PDV run length parameters contributes only a small portion to the full uncertainty with most of the uncertainty coming from the streamflow parameters. The final objective assessed the value of PDV information in urban water supply system decision making. This was done using an urban water supply case study based on the Warragamba River and in which demand was growing. Multi-objective optimisation, in which present worth capital cost and expected restriction costs were minimised, was used to identify the optimal schedule of future augmentations for two cases in which streamflow was generated by a lag-one autoregressive streamflow model and a PDV-informed lag-one autoregressive streamflow model. The use of optimisation allows differences in performance to be attributed to the choice of streamflow model. The results showed that ignoring PDV led to a considerable over-investment in capital cost. Moreover, the Pareto front was found to be sensitive to the PDV phase at the start of the simulation.
- Subject
- Pacific decadal variability; palaeoclimate reconstruction; uncertainties; water resources management
- Identifier
- http://hdl.handle.net/1959.13/1406228
- Identifier
- uon:35606
- Rights
- Copyright 2019 Lanying Zhang
- Language
- eng
- Full Text
- Hits: 932
- Visitors: 1440
- Downloads: 541
| Thumbnail | File | Description | Size | Format | |||
|---|---|---|---|---|---|---|---|
| View Details Download | ATTACHMENT01 | Thesis | 13 MB | Adobe Acrobat PDF | View Details Download | ||
| View Details Download | ATTACHMENT02 | Abstract | 517 KB | Adobe Acrobat PDF | View Details Download |