IMPACTS OF SECONDARY FLOOD EMBANKMENTS ON THE WAIMAKARIRI FLOODPLAIN, NEW ZEALAND

Transcription

1 IMPACTS OF SECONDARY FLOOD EMBANKMENTS ON THE WAIMAKARIRI FLOODPLAIN, NEW ZEALAND Terry van Kalken (1), Tony Oliver (2), Ian Heslop (2), Tony Boyle (2) (1) DHI Water & Environment, e-centre, Oaklands Road, Albany, Auckland, New Zealand Phone , tvk@dhigroup.com (2) Environment Canterbury, 58 Kilmore St., Christchurch, New Zealand. Phone: ABSTRACT The City of Christchurch is located on the Canterbury Plains at the foot of the Southern Alps on New Zealand s South Island. The city is at risk from potential flooding from the Waimakariri River, a substantial braided gravel bed river typical of those which flow out of the mountains and onto the Canterbury Plains. The existing flood protection scheme, comprising a line of flood embankments along the rivers edge, provides protection up to a maximum flow of 4730 m 3 /s, approximately a 450 year return period flood event. However due to the nature of the braided river, there is a risk of embankment failure due to undercutting and piping failures for much smaller flood events. The potential flood damages to Christchurch in a Probable Maximum Flood are approximately NZ$5 billion. During extreme flood events in the Waimakariri River the most likely risk scenario involves breakouts to the north (affecting Kaiapoi) and to the south (affecting Christchurch). It is possible that the main river channel will continue to carry large flows, although complete avulsion (course or channel change) could occur to the north or to the south. A secondary flow channel, contained by a secondary line of embankments tying into natural high river terraces could substantially reduce potential flood damages to Christchurch City. This paper discusses the river and floodplain investigations undertaken for a range of design scenarios for both the current flood protection system and the secondary embankment proposal on the southern floodplain. The investigations were undertaken using a 1D/2D mathematical modelling approach, including a dynamic embankment breaching feature to represent the progressive failure of the river banks in the considered scenarios. The models were used to assess both the existing flood inundation extents for up to a 10,000 year return period event (6,000m 3 /s) as well as the impacts of the proposed scheme in terms of changes in water levels and flow patterns. The model results were published to large scale inundation maps and were used in the community consultation process prior to final design. Keywords: floodplains, modelling, embankment breach 1 INTRODUCTION The Waimakariri is a steep braided gravel bed river draining a mountainous catchment area of 3566km 2 on New Zealand s South Island. Major floods are generated by heavy rainfall in the upper catchment near the main divide of the Southern Alps. In its lower reaches the river, which occupies part of an extensive alluvial fan, has been artificially confined to a relatively narrow course to prevent extensive flooding to cities and towns on the lowland plains, including Christchurch, New Zealand s second largest city. Previous investigations have highlighted the flood hazard to Christchurch and environs. The flood hazard from the Waimakariri River is the most significant in New Zealand and, in a worse case scenario, potential damage has been assessed at approximately NZ$5 billion. Prior to European settlement of Christchurch, geological evidence indicates major recent (under

2 The Upper Waimakariri Basin near Arthur s Pass 200 years) overflows or avulsions (course or channel change) from the Waimakariri River into what is now Christchurch. Post European settlement, Waimakariri River floodwaters entered Christchurch a number of times (1858,1859,1865,1866,& 1868). The 1868 flood appears to have been the largest of those events with breakouts occurring on both north and south banks, with parts of the Kaiapoi settlement flooded to depths of nearly 2 metres. The 1868 flood, in particular, prompted more serious attempts at controlling the river. Authorities focused their efforts on the south side of the river. There was much political protest from residents on the north side of the river about this and the Waimakariri River Trust was established in This resulted in a flood protection system providing equal protection for both sides of the river. In spite of all this endeavour breakouts and embankment (or stopbank ) breaches still occurred. The last major breakout occurred in December The estimated return period of the event was about 100 years, or, equivalently, it had an annual exceedance probability (AEP) of 1%. Following this, a major scheme upgrade was promoted by the North Canterbury Catchment Board, the successor to the Waimakariri River Trust. The Waimakariri River Improvement Scheme 1960 was designed to pass (without overflow) the design flood of 4730 m 3 /s (estimated at that time to be the 100 year flood, but with the benefit of additional records now estimated to be the 450 year return period flood). Works included raising and strengthening stopbanks, new and strengthened groynes, erosion protection, and protection planting. Although these works have been completed, there is still a risk of stopbank failure at flows well below the design flow. Channel aggradation that has occurred increases the likelihood of breakouts occurring in this area. The Waimakariri floodplain and existing stopbank system are shown in Figure 1. Due to the residual risk to Christchurch, a secondary stopbank located some distance south of the primary stopbank has been promoted. The proposed design standard of this bank is approximately a 1 in 10,000 year return period flood. This high level of protection was chosen because of the severe consequences should Christchurch City be flooded.

3 Figure 1 Waimakariri River floodplain and location of primary embankments This paper describes the investigations that have been undertaken to assess the overall hydraulic impact of the proposed secondary bank from Halkett to Coutts Island. In many locations along the line the natural river terrace is of sufficient height to contain breakout flows. In lower areas of the terrace the secondary bank will be built. Breakouts from the river in this location would be contained by the proposed secondary bank and redirected into the South Branch of the Waimakariri (the Otukaikino Outlet), returning flood flows to the main river just upstream of the State Highway bridge. Figure 2 shows the indicative line of the proposed secondary bank system in its entirety. Figure 2 Proposed secondary stopbank system

4 2 EMBANKMENT BREACH RISK ASSESSMENT The Waimakariri floodplain is a classic alluvial fan, with aggradation occurring fastest near the head of the fan (at the apex of the cone). In its natural state, the deposition of the coarse river gravels in the main braided river channel can cause the stream to abruptly change course, in a process known as an avulsion (Ref Reinfelds and Nanson, 1993). The stream either creates a new channel, or re- occupies an old channel where it has flowed in the past. The particular topography of a fan will determine which areas of the fan are at greatest risk. Many avulsions have occurred on the floodplain in the past, easily evidenced from the high resolution floodplain land levels acquired as part of the study, see Figure 3. The construction of embankments along the Waimakariri River artificially constrains the flow width. Nevertheless, the risk of breaches occurring in the embankments arising from potential avulsions in the braided river channel is real, and can occur at flows significantly less that the design floods for the embankments. As a first step in the embankment breach risk assessment, a geomorphic mapping exercise was carried out of the entire Waimakariri floodplain. This highlighted locations where historical avulsions had taken place, and these have been used as the basis for defining future potential breakout locations. To define the probable peak breakout flows at each location, the river flows were grouped into flood ranges between 750 and 10,00m 3 /s. Associated with each flood range there are a number of probabilities. Firstly, there is the probability of the flood occurring. Secondly, is the probability of the existing system coping with the flood event, and if not, probabilities are associated with the locations the failures will occur. Finally, there are probabilities associated with the range of outflows that can occur. The assigned probabilities were derived giving due regard to the structural integrity of the existing system investigations, and from practical knowledge of how the system has functioned in previous events. The probabilities here relate to failures caused by erosion, overtopping or piping. Breakout flows floodwaters resulting from embankment breaching are related to the extent of the actual failure, and the magnitude of the flood in the river. For each of the scenarios three ranges of breakout flows were assigned, ranging between the maximum and minimum outflows that might be expected. To calculate composite risk over the full range of floods to individual sites on the floodplain, the zone, (or zones) that would relay floodwaters to the site are identified using the Geomorphic Maps. The probabilities of breakouts from these zones are then simply multiplied by the annual exceedance probability of the flood ranges. Across the full range of floods these are then summed to give the composite risk. 3 HYDRAULIC MODEL DEVELOPMENT 3.1 APPROACH Flows over a floodplain are multi-directional and thus are difficult to predict. To determine flood levels, flow patterns and velocities on the floodplain to the required level of accuracy, and in a realistic manner, a hybrid 1D/2D hydrodynamic computer model (MIKE FLOOD) was used. The MIKE FLOOD modelling system combines one-dimensional (MIKE 11) and two-dimensional (MIKE 21) modelling in one modelling system. It allows modelling part of the model domain in 1D detail (related to the main river channel) and other parts in 2D, and hence provides a computational efficient approach for dealing with different spatial resolutions and process descriptions in different parts of the model domain. Furthermore, the

5 combined approach facilitates implementation of fine scale structures, such as weirs and culverts, within a 2D domain. MIKE FLOOD can be used to simulate the detailed flooding pattern on floodplains in 2D detail while maintaining an efficient 1D description for the inchannel flow. (DHI, 2005). Outflows onto the floodplain from the river were located in the two-dimensional MIKE 21 model at the potential embankment failure points. The outflows were obtained from the risk assessment described earlier. The magnitude of these breaches has subsequently been confirmed by independent checks using a breach analysis computer simulation with the 1D MIKE 11 dambreak model. The MIKE 21 model is linked to a one-dimensional model of the Waimakariri River, to ensure that return flow conditions in the river are taken into account. This paper focuses on the development and application of the lower Coutts Island model D RIVER MODEL A MIKE 11 model of the Waimakariri River developed by Environment Canterbury was used for the study. The model extends from just upstream of the Halkett breakout point, downstream to the sea and includes a connection to the South Branch at the Otukaikino outlet. Cross sections derived from river surveys, extending between the stopbanks, have been included in the model. The river falls steeply with the model area, dropping from 130 m at the upstream boundary to below sea level at the downstream boundary over a 35km reach. A river discharge of 6,500m 3 /s for the 10,000 year ARI event was applied to the upstream boundary in the Waimakariri River and a constant sea water level of 1.7m was applied to the downstream boundary D FLOODPLAIN MODEL High-resolution topographic data of the floodplain was obtained by Airborne Laser Scanning or LIDAR, which provides a high-density digital elevation model. Both ground and non-ground points (i.e. trees and buildings) are recorded. This high-resolution survey over the study area provided average vertical accuracy of typically 0.15m, with an average point spacing (on the ground) of approximately 1.5m. Figure 3 shows a detail of the LiDAR levels, clearly showing the old river channels as well as the primary stopbanks. Figure 3 Detail of LiDAR survey on the Waimakariri Floodplain, existing river can be seen at the top

6 The MIKE 21 model of the McLeans-Crossbank area has been developed on a 10m grid based on the aerial laser topographic data, see Figure 4. The floodplain resistance has been derived directly from the LIDAR data through an analysis of the intensity (of the laser return) and the local variation in returns. A very detailed surface map of floodplain resistance was developed. Typically Mannings n values for hedges/trees were approximately 0.125, and grassed areas (depending on length of vegetation). Figure 4 MIKE FLOOD model of McLeans Coutts Island Area showing breakout locations 3.4 COUPLED 1D/2D MODEL To realistically simulate the return of flood overflows to the main river, via the South Branch of the Waimakariri, the MIKE 21 model was dynamically coupled to the MIKE 11 model of the main river. The coupling takes place via MIKE FLOOD, which incorporates facilities to specify coupling at MIKE 11 boundary locations, or via continuous lateral spilling. For the existing breach simulations, additional connections were included to the south east of the Harewood crossbank, and around the Otukaikino outlet, to allow floodplain flows to exit the MIKE 21 model domain. For the post construction stopbank scenario, the only outflows allowed in the model are via the South Branch outlet. 4 MODEL SIMULATIONS 4.1 HISTORICAL BREACH SIMULATIONS To provide confidence in the model predictions it is important to calibrate with historical events where possible. Within the last 60 years two breakouts have been recorded and mapped over this part of the floodplain. May 1950 event This event reached a peak flow of 2,500 3,000 m 3 / s (approximately 20 year return period) and breached the Harewood crossbank in three locations (Refer photo 1). The magnitude of the breach flows are not known precisely, but an historic map of the flood extent compared well with the modelled extent, using best estimates of breakout hydrographs.

7 Photo 1 Breaches in Harewood Crossbank May 1950 December 1957 This flood event is the largest recorded on the Waimakariri since at least 1930, when flow records commenced. The estimated peak discharge of approximately 4,000 m 3 /s is equivalent to approximately a 100 year return period (1%AEP) event. This flood breached the stopbank at Engelbrechts, flooding the Coutts Island area, before breaching Chaney s bank, flooding parts of Belfast and Kainga. Based on recorded flows and rainfall/runoff modelling a reasonable estimate of breakout flow (425m 3 /s peak discharge) has been made. A very good comparison of observed and predicted flood extent in the Coutts Island area was achieved. 4.2 EXISTING FLOOD INUNDATION Following calibration/verification of the model with the above historical events, river breakout flows onto the floodplain were modelled. For the purposes of the secondary bank investigation, two main design scenarios have been modelled i.e. the 500 and 10,000 year return period events. In the 500 year return period (0.2% AEP) event, it has been assessed that there would be one breakout through the stopbank system. This could be in the vicinity of Halkett, Baynons (North side), McLeans or Crossbank.. In the present study the 500 year breakout is assumed to be located at Crossbank. The flood extent is affected by the peak and volume of the breakout flow. Flood hydrographs from the 1957 event (the largest recorded) were used and scaled appropriately for both the river flood hydrograph and breakout (breach) flow hydrograph. This was a hour flood event (typical of large floods in the Waimakariri River) with the breakout flow continuing for approximately 24 hours.

8 The 500 year river flow is 5,100m 3 /s and the breach flow is assessed to be 750m 3 /s. In the 500 year event the flow return to the river via the South Branch is approximately 628 m 3 /s, the reduction is due to both natural attenuation and some loss from the floodplain. In the 10,000 year return period event, it has previously been assessed there would be a number of different breakout scenarios. The main river flow in this case is 6,500 m 3 /s. The combination affecting the study area to the greatest extent would be 1000 m 3 /s and 600 m 3 /s breakout flows at Crossbank and McLeans (respectively). At the same time, it has been assumed there would be a 1000 m 3 /s breakout at Halkett. Flood hydrographs were derived, by scaling the 1957 hydrograph (as above) to the appropriate magnitude. This scenario would not only flood most of McLeans and Coutts Islands, but also substantial parts of north and west Christchurch, including Christchurch Airport. For the 10,000 year existing scenario, the breakout flows from McLeans will travel down the floodplain and pond behind the Harewood crossbank. It is envisaged that this stopbank will breach once the water levels increase sufficiently to overtop the crest. This has been simulated in the model by including a dambreak structure in the MIKE 11 model, dynamically coupled to the MIKE 21 floodplain model either side of the embankment. 4.3 IMPACTS OF PROPOSED EMBANKMENTS To contain breakout flows (up to the 10,000 year return river flood period) in the Halkett, and McLeans to Coutts Islands area (i.e. to protect Christchurch City) it is proposed to construct a secondary protection system from Halkett downstream to Belfast. The secondary stopbank will not be continuous, and is only required in low areas of the terrace. Note that as part of the proposed secondary bank system, a cut will need to be made in the Harewood Crossbank to allow passage for the breakout flows from upstream. The primary banking system to the north east will also need to be realigned to provide an adequate opening to restrict flow velocities to acceptable levels. An 800 m (approximately) cut has been included in the model bathymetry to allow for this. The impact of the proposed stopbank in this area has been assessed for both 500 and 10,000 year events. For the 500 year event, (with a breach flow of 750m 3 /s) the flow returned to the river in the South Branch is increased slightly to 682 m 3 /s compared to the existing situation of 623m 3 /s, due to the fact that no water can now escape further downstream on the floodplain. The increased return flow results in an increase in water depth upstream of the Otukaikino Outlet of m generally and an increase in flow speed of up to 0.5m/s. The proposed Halkett secondary stopbank (upstream of McLeans Island) will return almost all of the breakout flows in a 10,000 year return period flood (1000m 3 /s) to the river upstream of McLeans Island. This results in the downstream river flows and potential breakouts at McLeans and Crossbank being greater than the present situation. Consequently the assessed breakout flows at McLeans and Crossbank were increased by 200 m 3 /s each, to take this increased flow into account. The maximum return flow to the river via the South Branch for the 10,000 year event is 1622m 3 /s, compared to the existing situation of 810m 3 /s, a 100% increase. The river flow downstream of the South Branch confluence is 5533m 3 /s, the reduction from the inflow of 6500m 3 /s occurring as a result of flow attenuation on the floodplain. The impacts of the stopbanks in terms of changes to water levels and flow speeds compared to the existing situation are shown in Figure 6. As expected the water levels along most of the floodplain increase, with the exception of the area upstream of the Harewood

9 crossbank where the 800m cut in the stopbank allows free passage of water through this area compared to the existing case. In other areas water level increases are generally less than 0.3m, apart from the outlet area where the water level increases to m upstream of the outlet compared to the existing situation Figure 4 Change in water depth (top) and flow speed (bottom) due to the proposed stopbanks 5 DETAILED ASSESSMENT AND COMMUNITY CONSULTATION The results of the 2D hydrodynamic modelling have been used directly as part of a community consultation programme to assess impacts of the proposed works on the local community. Water levels were extracted from the 2D model at and compared to surveyed floor levels to accurately determine the changes induced by the stopbanks. Large scale plots of the 2D model results were presented to clearly illustrate the before and after situations to non-specialist stakeholders.

10 6 CONCLUSION A secondary bank system to provide additional flood protection to Christchurch City is justified because of the high potential flood damages which may arise from breakouts in the primary system. The use of a two dimensional floodplain model with detailed topography of the floodplain has been used to assess the impacts of the proposed secondary banks. Although there are a number of uncertainties (e.g. hydrology, breakout flows, and location), particularly in such an extreme event, confidence in the model has been obtained with a good match of flood extent with the historical floods of 1950 and The simulations indicate there will be some increase in flooding on the land between the secondary bank and the river, under 0.5m in the 10,000 event, except near the South Branch confluence where the flow is concentrated and the outlet restricted due to high river levels. Here the flood depth increases are predicted to be m. This area, however, is already subject to flood depths of over 2 m in the existing 10,000 year return period event scenario. The investigation has shown the secondary stopbank system is a practical and viable solution, providing significant additional flood protection to Christchurch, with relatively minor local adverse effects. The use of a coupled 1D-2D model has optimized the use of available data computational times required for the simulations. In addition the presentation of flood maps generated by the models has aided the community consultation programme and aided the assimilation of the model results. AKNOWLEDGMENTS The study which is the subject of this paper has been carried out jointly by DHI Water & Environment Ltd. and Environment Canterbury. The authors wish to acknowledge the support of Environment Canterbury in the preparation of this paper. REFERENCES DHI (2005), MIKE FLOOD User Manual, DHI Software 2005, DHI Water & Enviroment, Denmark. Reinfelds, I. and Nanson, G, (1993) Formation of braided river floodplains, Waimakariri River, New Zealand, Sedimentology 40 (6), Environment Canterbury (1991), Draft Waimakariri River Floodplain Management Plan: Report R 91(9) July 1991.