|dc.description.abstract||Available observations are often not sufficient as a basis for decision-making in water resources management, hazard mitigation, and water resources utilisation. The decision making process becomes more complex in a headwater catchment where measuring flows is virtually impossible due to mobile bed conditions and where the hydrology is strongly influenced by glacier systems. Conceptual runoff models are frequently used as tools for a wide range of tasks to compensate for the lack of measurements, e.g., to extend the runoff series or to compute design floods. Although some conceptual models do have snowmelt routines, their availability is limited, they require modification and rigorous calibration, and when applied to an area of dynamic geology and climate such as the Southern Alps they have failed to meet the expectation of the modeller in most cases. In this thesis, the adaptation of a physically based lumped conceptual model is described as a contribution to the predictive hydrology of wet temperate mountainous regions in general, and in response to the specific, urgent need for developing hazard forecasting systems in the Franz Josef area of the Southern Alps.
The "modified UBC" watershed model, a physically based, modular, and semi-distributed parameter watershed model, is designed for estimating runoff from a glacierized catchment using meteorological inputs of precipitation and temperature using the soil moisture accounting system. The model evaluates the watershed response to the various combinations of precipitation, meteorology and physical properties and predicts runoff from rainfall, snowmelt and glacier melt. The model has five modules, which form a logical sub-division of underlying physical process in a watershed. They are:
1. Climate Module,
2. Snow and Glacier Melt Module,
3. Moisture Distribution Module,
4. Runoff Distribution Module, and
5. Evaluation Module
The model accounts for evapotranspiration and soil moisture deficit, making it possible to continuously simulate streamflow as long as precipitation and temperature data are available. The model also provides information on the soil moisture budget, soil and ground water storage values, contributions of runoff from various portions of the watershed, and surface and sub-surface components of runoff. For calibration and verification purposes, the model calculates' performance statistics of volume and hydro graph shape reconstitution using the measured meteorological and streamflow records.
The modified UBC watershed model is first applied to and tested in the Hooker River catchment, then applied and recalibrated to the Whataroa River catchment (adjacent to the Franz Josef area), before applying the model to simulate the Waiho River flow. The model efficiency was measured to the extent possible by comparing with the measured flow records. These comparisons show the model capability in predicting runoff from glacierized catchments and its applicability in New Zealand conditions.
In the process of development and application of the modified UBC watershed model, it was necessary to estimate the missing values in temperatures, precipitation, and flow-records, as the model requires serially complete input and output records for model calibration. However, estimation of missing values in hydro-meteorological time series with adequate dependability is a complex problem. Five different methods were used in estimating the missing values in each of the input variables. The methods are Expectation-Maximisation, Multivariate Regression, Auto Regressive Integrated Moving Average (ARIMA), Multiple Linear Regression and a neural network model, aiNet. The outputs of all these models were then evaluated by comparing against the historical values and using the Analytical Hierarchy Process, a multi-criteria evaluation technique to obtain the best model for estimating the missing values in temperatures, precipitation and flow records. The comparison study demonstrated the potential of the aiNet model in the field of rainfall-runoff modelling and estimating missing records in the hydro-meteorological time-series.||en