The Waimakariri River and its environment: frequency analysis of extreme events and the effects of flow modifications

Abstract

The Waimakariri River is known for its unpredictability; flood and drought events can occur at any time of the year. Knowledge about the return period of extreme events is crucial for water management and planning. Frequency analysis uses the stochastic nature of these extreme events as a basic concept to account for uncertainty by describing the expected occurrence of floods and droughts. While frequency estimates for the Waimakariri River have been produced in the past, sampling from the historical stream record has almost exclusively relied on annual maximum series (AMS), in combination with a few selected probability distributions based on theoretical considerations. Low flow frequency analysis, albeit an informative tool for water planning, is seldom utilised and no estimates are published for the Waimakariri River. However, low flow frequency estimates can provide valuable information for the management of instream biota under pressures from flow modification.

This thesis produced flood and low flow frequency estimates for the Waimakariri River, using the partial duration series (PDS) approach to sampling for the first time. Previous estimates utilised in floodplain management have solely relied on the use of AMS. Graphical and statistical testing methods were employed to determine the best fitting probability distributions, guided by the empirical streamflow record. Low flow frequency estimates were produced with ‘runs theory’ to reflect the multiple dimensions that characterise low flows. Current developments in the Waimakariri catchment are reflected by developing a discharge series subtracting consented water abstractions, which was compared to the current discharge series. Together with a rapid systematic literature review, the results from the low flow frequency analysis were used for a qualitative narrative assessment summarising the likely effects of flow modifications on the Waimakariri River environment.

Statistical results indicate that the PDS produces a better fit to the empirical flood and low flow data than the often applied AMS approach. The procedures for threshold selection and best fit decision making in this study identified methods for choosing the optimum PDS. Contrary to the published literature, the Gumbel distribution is not a good fit for the flood series of the Waimakariri River. Results indicate a better fit of four alternative distributions, i.e. the Log-Pearson III, Pearson III, Generalised Extreme Value and Generalised Pareto distributions. All dimensions of low flow series are equally well described by these four distributions. Design estimates for floods produced by the four alternative distributions are markedly higher than current estimates used in floodplain management. Low flow frequency estimates were produced for duration and deficit series, reflecting the current water allocation regime of the Waimakariri River. The longer low flow durations and higher deficits are expected to significantly impact not only on the reliability of river flow dependent catchment developments, but also on the flow dependent biotic and abiotic environment. Results showed that both habitat quantity and quality ensuring flows are significantly affected by flow modifications.