2 Data and Methods
2.1 Rainfall Data
2.1.1 Time Series Rainfall
The following rainfall time series' were used in this modelling analysis:
- 1981-2003 rain and flow gauge data provided by NSCC, (used for model calibration)
- 1 'Present' 150-year Stochastically generated rainfall time series, and
- 3 'Future' 150-year Stochastically generated rainfall time series.
The stochastically generated time series' are discussed in Cowpertwait (2003), and are derived from a 150-year series representing the "Present" climate scenario (which has the same statistical characteristics as the 30 years of gauged rainfall from the Wairau rain gauge). This was generated using a Neyman-Scott Rectangular Pulses (NSRP) model, which produced five 30-year series. These were then combined to form a single 150-year series of rainfall at 5-minute intervals.
Three "Future" 150-year series were then stochastically generated to reflect alternative future rainfall scenarios. The three alternative 'Future' rainfall data sets were based on the following assumptions with respect to potential climate change after consultation with NIWA staff:
Future 1: A 5% increase in mean monthly rainfall, leaving wet/dry periods unchanged.
Future 2: An increase in the proportion of dry days by 5% and the mean rainfall by 5%.
Future 3: An increase in the proportion of dry days by 10% and an increase in the mean rainfall by 5%.
For the Present and Future 1 series, only 100 years of the data was used, whilst for the Future 2 and Future 3 series, 110 years was used. This was due to limitations in computing capacity.
2.1.2 Design Storms
Beca (2003) undertook an analysis of the 'Present' and 'Future 3' stochastically generated 150-year rainfall series in order to develop Intensity Duration Frequency (IDF) curves for each of these scenarios. These IDF curves were then used to develop typical temporal 24-hour design storm profiles. Design rainfall depths taken from the IDF curves with a range of durations up to 24 hours were nested, and then normalised by the 24-hour rainfall depth. This was the same approach as was used to produce the design storm in the ARC's TP108.
The 10% AEP 24-hour depth was calculated as 121.26 mm for the 'Present' design storm, and 168.45 mm for the 'Future 3' design storm - a difference of 39%.
2.2 Modelling Approach
A computer model of the main components of the Wairau stormwater drainage network was developed using DHI's MOUSE 2003 software. In this case study, catchment runoff from all scenarios is routed through the same open channel network, a simplification of the Wairau Valley stormwater drainage system, characterised by large, open channels. One of the key features of the hydraulic model, specific to this catchment, is that the majority of the runoff was routed through stormwater ponds prior to entering the main network.
This hydraulic model simulates 8.6 km of open channel, 1.3 km of piped network and ten inline stormwater ponds.
For the hydrological models, the Wairau catchment was divided into 17 sub-catchments ranging in size between 22 ha and 216 ha. MOUSE has the ability to use several hydrological methods to estimate catchment runoff.
Dynamic Modelling
For the dynamic simulations, a calibrated Mouse RDII model was used in combination with a surface runoff model (in this case, Model A, a time/area method with a convergent time area curve number) in order to split the runoff into a 'fast' and 'slow' component (FRC and SRC respectively).
This model was calibrated using a number of rainfall events and corresponding flow data from the gauged records provided by NSCC. There are 14 parameters that can be manipulated in the RDII parameter set, most of which are of an empirical nature and difficult to estimate from geophysical measurements. Worley and AWT (2000) presented a calibrated set of RDII parameters for the Wairau catchment, which were then adopted for this project. A few minor adjustments were made to several of the parameters while calibrating the dynamic hydraulic/hydrological model to the gauge data.
Once calibrated, the RDII model was used to run 100 years of stochastic rainfall data from each of the "Present" and "Future 3" rainfall sets and 110 years of rainfall from the "Future 1" and "Future 2" scenarios. The LTS (Long Time-Series Simulation) module of MOUSE was used to reduce modelling time and summarise results from these simulations.
Design Storm Modelling
Design storms are commonly used with a unit hydrograph method of runoff estimation. Methodology recommended by Auckland Regional Council's TP108 (ARC, 1999) was followed for this analysis. Generally, the UHM method produces a single runoff hydrograph from a catchment, and fast and slow response components were simulated by splitting each of the sub-catchments into impervious and pervious sections, the impervious areas having a faster response time than the pervious components.
Design storms used in this simulation were developed from the stochastically generated 'Present' and 'Future 3' time series (Beca, 2003), rather than the Auckland regional design storm. This model was also calibrated using rainfall events and corresponding flow data from the gauged records provided by NSCC.
