<p>Lake Taupo is the effective source of the Waikato River. The Waikato Power Scheme relies on the outflow from the lake for moderated flows throughout the year. As the lake is maintained between a 1.4m operating range, it is the inflows to the lake that determine the amount of water available to the scheme for electricity generation. These inflows have not been modelled in any detail prior to this dissertation. This dissertation aims to develop a predictive rainfall-runoff model that can provide accurate and reliable inflow and lake level forecasts for the Lake Taupo catchment. Model formulation is guided by a fundamental understanding of catchment hydrologic principles and an in-depth assessment of catchment hydrologic behaviour. The model is a semi-distributed physically-consistent conceptual model which aims to provide a parsimonious representation of different storages and flow pathways through a catchment. It has three linear sub-surface stores. Drainage to these stores is related to the size of the saturation zone, utilising the concept of a variable source area. This model is used to simulate inflows from gauged unregulated sub-catchments. It is also used to estimate the inflow from ungauged areas through regionalisation. For regulated sub-catchments, the model is modified to incorporate available data and information relating to the relevant scheme‟s operation, resource consent conditions and other physical and legislative constraints. The output from such models is subject to considerable uncertainty due to simplifications in the model structure, estimated parameter values and imperfect driving data. For robust decision making, it is important this uncertainty is reduced to within acceptable levels. In this study, a constrained Ensemble Kalman Filter (EnKF) is applied to the four unregulated gauged catchments to deal with model structure and data uncertainties. Used in conjunction with Monte Carlo simulations, all three sources of uncertainty are addressed. Simple mass and flux constraints are applied to the four (soil storage, baseflow, interflow and fastflow) model states. Without these constraints states can be adjusted beyond what is physically possible, compromising the integrity of model output. It is demonstrated that the application of a constrained EnKF improves the accuracy and reliability of model output.Due to the complexity of the Tongariro Power Scheme (TPS) and the limited data available to model it, the conceptual model is not suitable. Rather, a statistical probability analysis is used to estimate the discharge from this scheme given the month of the year, day of the week and hour of the day. Model output is combined and converted into a corresponding change in lake level. The model is evaluated over a wide range of hydrological and meteorological conditions. An in-depth critical evaluation is undertaken on eight events chosen a priori as representation of both extreme and „usual‟ conditions. The model provides reasonable predictions of lake level given the uncertainty with the TPS, complexity of the catchment and data/information constraints. The model performs particularly well in „normal‟ and dry conditions but also does a good job during rainfall events in light of errors associated with driving data. However, for real-time operational use the integration of the model with meteorological forecasts is required. Model recalibration would be required due to the issue of moving from point estimation to areal rainfall data. Once this is achieved, this operational model would allow robust decision-making and efficient management of the water resource for the Waikato Power Scheme. Although there is room for improvement, there is considerable scope for extending the application of the constrained EnKF and techniques for incorporating regulation to other catchments both in New Zealand and internationally.</p>
Characterizing Curvilinear Features Using the Localized Normal-Score Ensemble Kalman Filter
