A modified modelling approach to estimating the probable maximum flood and a revised modelling approach to flood risk analysis

Abstract

This thesis presents modified and revised methodologies for the estimation of extreme river discharge including the Probable Maximum Flood, and the analysis of flood risk. The models are related but independent - they provide a revised approach for flood impact catastrophe analysis. Both are developed in the New Zealand context and hydrological data used is from New Zealand river systems. However, their principles are generic and designed for wider application. Preliminary sections of the thesis review the international literature and examine the history of flood method in New Zealand. Alongside existing conventions are calls for the development and use of what may be referred to as, less explicit approaches, and the integration of physical explanations of behaviour into model development. Thus, modelling seeks to use descriptive techniques, incorporate physical process explanations and use methods that enable simple representations of nature to be examined. Also, a generalised approach is adopted whereby events and processes are viewed from different perspectives and conclusions drawn on the basis of collective rather than individual assessment. Extreme discharge events modelling uses the framework of Regional Flood Frequency Analysis. Recent trends show an increase in the use of regionalized methods that pool data to provide addition information able to be transferred from one location to another. The difficulty of maintaining data homogeneity is addressed by defining a hierarchy of influences that affect at-site conditions. This approach is used to pool data and examine the sample populations containing the largest observed discharge events for defined regions and clustered data sets. This preliminary analysis of New Zealand and Queensland, Australia data; shows that there is a strong linear correlation between the mean and maximum floods expressed as either discharge or specific discharge, and that there is a strong correlation with respect to the logarithms of discharge and catchment area. The correlation between the specific discharge variables and catchment area is weak for the New Zealand data but strong for the Queensland data. The modified methodology uses a physical model to define the hierarchy of influences that cumulatively affect the generation of peak discharge. Using this model and the results of the preliminary analysis, each of these influences are modelled. This involves defining principal flood regions, and for each flood region using relationships between the maximum and mean flood events expressed as specific discharges and between the maximum specific discharge and catchment area, to define the additional influences affecting discharge generation. The method used to produce estimates of extreme and PMF discharge events is to extrapolate envelope plots of the defined relationships by plotting probability distributions through partitions of the data fields thereby extending outwards the envelopes prescribing the empirical data. This produces a series of equations using the variables specific discharge (the specific discharge of the mean annual flood) and catchment area that describe, for example, the 1000 year, 5000 year, 10000 year and Limit (PMF) flood events. The PMF model is expressed in mathematical form as a series of equations with assigned confidence limits. The modelling of flood risk uses systems principles. This enables the physical mechanisms and event sequences of a flood event to be represented in a manner suitable for damage modelling. Systems principles provide means with which to visually present complex relationships and construct explanations of behaviour. The emphasis of flood risk is shifted to the damage outcome of flood events, the combinational events and mechanisms responsible for those outcomes, environmental variability, stationarity and establishing linked probability statements to define event likelihood's. Systems concepts structure the revised model in a physical and mathematically correct form to enable both magnitude and probability statements describing the impact of flood events to be produced using the underlying mathematical concept of best estimates. Existing conventions are adapted, weaknesses associated with these methods provided for or addressed, and principles associated with a number of recent developments such as vulnerability and uncertainty analysis incorporated into the revised flood risk analysis model. The flood risk analysis model produced is a scenario events model formatted to the Australia New Zealand Risk Management Standard AS/NZS 4360:1995. It is viewed as a component part of the risk management process more commonly referred to as floodplain management.