View all publications

6 Case Studies

6.1 Introduction

The tools outlined in chapters 3, 4 and 5 can be linked in different ways to provide estimates of changes in rainfall, flood flows and inundation. This chapter provides a number of case studies that show how these tools have been used in real-world situations. As far as possible, each case study starts by defining the problem being studied, and then gives a technical summary of how the tools were used to answer that question using the three steps of: (1) estimating the change in rainfall, (2) converting rainfall to a flow rate, and (3) estimating the consequent inundation. The aim in each case study is to show the calculated expected change in flood hazard due to climate change.

6.2 Stoney Creek

Stoney Creek, a small stream in Central Otago, passes through an area being sub-divided for residential development. An assessment is needed of the frequency with which the stream would overtop its banks, both under the current climate and under a 2040 climate (which is appropriate to the 50-year design life of the project). There are no stream-flow measurements available and no rain gauges in or close to the catchment.

The three steps in the study are:

  1. use HIRDS and the Climate Change Effects manual factors to estimate current and future storm rainfalls (section 3.2.1)
  2. convert storm rainfalls to flood peaks using the Rational Method (see section 4.2.2)
  3. estimate the frequency of inundation using a flow resistance equation (Manning’s equation in this case), field inspection and local knowledge.

6.2.1 Step 1: Estimate rainfall

The time of concentration for this two-square-kilometre catchment is estimated as being 20 minutes. Accordingly, 20-minute rainfall intensities are obtained from HIRDS under the current climate.

Rainfall intensities for the 2040 climate are estimated by first using table 1, which gives the percentage increase in rainfall intensity per degree Celsius of warming for a range of storm durations and recurrence intervals. For a 20-minute duration we need to interpolate between the rows of table 1 for 10 minutes and 30 minutes. For consistency with the way this table was created (see the Climate Change Effects manual), the interpolation should be logarithmic (although in this case it makes little difference if you use linear interpolation).

If P20 is the percentage increase in rainfall per degree of warming, then the logarithmic interpolation is calculated as:

P20 = P10 [P30/P10](20–10)/(30–10)

For example, with a two-year ARI, P10 = 8 per cent, P30 = 7.2 per cent and so P20 = 7.59 per cent.

The next step in estimating rainfall change is to estimate the amount of warming. From the Climate Change Effects manual (table 2.2, Otago, last column) the average warming (across all emission scenarios) is 0.9 degrees Celsius for 2040. This happens to be the same as the average temperature increase for the middle-of-the-road (A1B) scenario in table 2 above. A check of the Climate Change Effects manual (figure 2.3, top-left map) shows there are no strong within-regional variations in the projected temperature rise for Otago.

The results of this calculation are given in table 8.

Table 8: Calculated rainfall intensities for the Stoney Creek case study
Average recurrence interval (years) Intensity for 20-minute duration – current climate (mm/h) Increase in rainfall for 0.9° C warming (%) Intensity for 20-minute duration – 2040 climate (mm/h)
2 19.9 6.8 21.3
5 24.0 6.9 25.7
10 30.7 7.0 32.9
20 37.7 7.1 40.4
50 51.9 7.2 55.6
100 68.1 7.2 73.0

In table 8 we have assumed an increase of 0.9 degrees, which is the average of the projected increases given in table 2 for the A1B scenario for 2040. The projected increases for Otago range from 0.2 to 1.3 degrees, depending on which global climate model is used (these numbers are given in brackets in table 2, just after the 0.9°C value for Otago).

To assess the range of possible changes in rainfall, the calculations in table 8 should be repeated using other temperature rises from table 2. We suggest taking the mid-point between the mean and the lower bound for A1B (ie, for Otago, [0.9 + 0.2]/2 = 0.55 degrees), as well as the mid-point between the mean and the upper bound (ie, [0.9 + 1.3]/2 = 1.1 degrees). The extreme ends of the ranges are not used because they are slightly less likely than the central values (see the Climate Change Effects manual, section 2.1 and Appendix 3).

6.2.2 Step 2: Convert rainfall to flow

The formula for the Rational Method can be written (see section 4.2.2) as:

Q = C i A /3.6

where:

  • Q is an estimate of the peak design discharge in cubic metres per second
  • C is the run-off coefficient
  • i is the rainfall intensity in millimetres per hour, for a duration equal to the time of concentration of the catchment
  • A is the catchment area in square kilometres.

Because the peak design discharge formula depends linearly on the rainfall rate, this method will predict that any percentage increase in rainfall intensity leads to the same percentage increase in peak design discharge, provided no other variables change. However, in this case the run-off coefficient, C, was assessed as being dependent on the return period of the event (Ministry of Works and Development, 1978), as well as the physiographic features of the catchment. The results of this calculation using 0.9 degrees Celsius as the temperature change for 2040 are given in the first four columns of table 9.

Table 9: Calculated flood flows for the Stoney Creek case study
Average recurrence interval (years) Run-off coefficient (dimensionless) Flood peak for current climate (m 3 /s) Flood peak for 2040 climate (m 3 /s) Water depth for current climate (m) Water depth for 2040 climate
(m)
2 0.45 5.0 5.3 0.46 0.48
5 0.45 6.0 6.4 0.51 0.53
10 0.55 9.4 10.0 0.67 0.70
20 0.55 11.5 12.3 0.76 0.79
50 0.65 18.7 20.1 1.01 1.06
100 0.65 24.6 26.4 1.20 1.25

6.2.3 Step 3: Convert flow to inundation

To assess the impacts of the change in flood peak on inundation in a simple screening study where detailed flood-plain mapping information is not available, the techniques described in section 5.1 are appropriate. In a uniform reach of river, information on channel width and roughness can be used to convert river flows into river levels using Manning’s equation or a similar approach. In this example, the channel is approximately 10 metres wide and has a slope of approximately 0.003, so the estimated Manning roughness coefficient is 0.03 (see, for example, Hicks and Mason, 1998, for methods for estimating roughness). The Manning equation for depth of flow in a wide rectangular channel is:

D = [Q n / (W Sf1/2) ]3/5

where:

  • Q is flow in cubic metres per second
  • W and D are width and depth, respectively, in metres
  • Sf is the friction slope
  • n is the Manning roughness coefficient.

Applying this equation to our case study, we obtain the results shown in the last two columns of table 9. The estimated water depths at the flood peaks have increased by between 0.02 metres and 0.05 metres, depending on flood magnitude. When using a screening technique, such a small change would not be considered significant in most practical settings. The uncertainty of these estimates should be assessed by varying the assumed and estimated input values for slope, width, roughness coefficient and peak flow within likely bounds; for example, the roughness coefficient can vary significantly with flow. If a larger change in depth had been predicted, then its significance for inundation could be investigated through a site visit to gather evidence on historical flood occurrence at the site, and by visually assessing or surveying the area that could be flooded if the predicted water levels were to occur.

6.2.4 Notes

This case study considered only one method for assessing potential climate change impacts, but it is good practice to use several different methods and compare the results where information is limited. Other sources of information would include national or regional (eg, local regional council) studies of flood frequency to supplement the Rational Method, and field inspection/surveying of the stream to supplement local knowledge of flood frequency. To compare the performance of the Rational Method in New Zealand it was checked against two other methods: TM61 and Regional Flood Frequency (see McKerchar and Macky, 2001). Note that estimating flood depths using Manning’s equation with a single roughness coefficient at widely varying flows is a major simplification and would only be appropriate for screening.

6.3 The Hutt River case study

In the early 1990s, concern was expressed that climate change could increase the risk of Lower Hutt being inundated as a result of overtopping of the stopbanks downstream of the Taita Gorge (Leong, et al, 1992). This study used the knowledge available at the time to form a preliminary assessment of the likelihood of this happening. The study is an example of a ‘screening assessment’, since the rainfall scenarios used were simple scaled versions of measured events (similar to the approach outlined in section 3.3.2). No attempt was made to do sophisticated climate modelling, partly because of the state of climate modelling at the time, but also because there was no point doing a lot of work on rainfall estimates if the risk of inundation could be shown to be low.

6.3.1 Step 1: Estimate rainfall

The first step was to assess what rainfall data was available for the catchment. This data had to be at hourly time steps to provide sufficient detail to enable an adequate simulation of river flows. It was impractical to use all the rainfall data, so the values associated with the largest flood each year were extracted. The data also needed to provide a realistic picture of the spatial variation of rainfall over the catchment to enable the known rapid increase in rainfall across the catchment to be taken into account. (The rainfall varies from less than 1400 millimetres per year in the lower reaches to more than 6000 millimetres per year in the headwaters.) A table of the available automatic rainfall recorder records was compiled for the period 1971 to 1989.

In 1981, there was a substantial expansion in the network of automatic recorders, with up to 13 recorders in operation; before 1981, there were only four or fewer gauges operating. The increase in the number of gauges meant that in subsequent years separate spatial patterns could be made for each storm event. However, before 1981, all events had to use the same spatial rainfall pattern based on the network of daily and storage gauges operating in the catchment. This difference affected the reliability of the results obtained.

6.3.2 Step 2: Convert rainfall to flow

To determine the likelihood of inundation of Lower Hutt, a rainfall-to-flow model was built for the catchment to convert rainfall to river flow. The model was spatially distributed and enabled different amounts of rainfall to be used as input to allow spatial variation of rainfall to be taken into account. Thus, the parts of the model representing the headwaters had a lot more rainfall as input than areas near the catchment outlet.

Data for the 1986 storm was used to calibrate the model (by adjusting the coefficients from the run-off model). The calibrated model was then used to generate the floods for the years 1971–1985 and 1987–1989. The model results for each year were compared to the measured river flows, and conclusions were drawn about the quality of the model simulations. For simulations of floods before the expansion of the rainfall network in 1981, the flood peaks were generally underestimated. However, with the expansion of the rainfall network and the use of different spatial rainfall patterns for each event, good check results were obtained. The flood peak flows were overestimated by 1 per cent and their standard error of estimate, a measure of the random variation about the simulated value, was ± 9 per cent.

Every rainfall amount for each annual flood event was then successively increased by 5 per cent, 10 per cent and 15 per cent. The percentage increases in the rainfalls were chosen in order to bracket the increases likely to occur as a result of climate change and to provide a range of potential risk scenarios. The modified rainfalls were then run through the model to form corresponding climate change-affected flows. The peak flows were extracted from the data and used in an ‘extreme value’ analysis 12 to derive the changes in the design flows used for sizing the stopbanks along the Hutt River. The results showed that:

  • a 5 per cent increase in rainfall would lead to an increase of 6.7 per cent in all flood flows
  • a 10 per cent rainfall increase would lead to an increase of 13.4 per cent in flows
  • a 15 per cent rainfall increase would lead to an increase of 20.3 per cent in flows.

The slightly greater proportional increases in flow arise from the fact that the catchment has a limited amount of natural storage, and so as the rainfall increases, proportionally more of it enters the river channel and contributes to the river flow.

6.3.3 Step 3: Convert flow to inundation

The final step in the investigation was to turn the increased flows into water levels and compare the new levels with the heights of the stopbanks. For sites on rivers with established sets of water-level and flow measurements, such as at Taita on the Hutt River, this can be easily done by using the ‘rating curve’. Note that a more complex inundation analysis would probably be necessary if the water was predicted to overtop the stopbanks.

6.3.4 Notes

The Hutt River study, which assumed only that rainfall would increase with climate change, concluded that:

  • the model could accurately and reliably estimate flows in the Hutt River
  • spatial rainfall patterns for each storm give better results than a standard pattern based on storage gauges
  • rainfall increases of 5, 10 and 15 per cent led to flow increases of 6.7, 13.4 and 20.3 per cent, respectively. The 15 per cent increase scenario would represent a mid-range estimate for the rainfall changes expected by 2100
  • the increases in flow will decrease the level of flood protection provided for Lower Hutt. For example, for the 15 per cent increase in rainfall scenario the current 100-year flood will become a 33-year flood in the future
  • it would be unlikely that present stopbanks at Taita would be undermined by the 11 per cent increase in channel width that would result from the increase in flood flows
  • since this work was undertaken, significant effort has been put into understanding the flood risk for the Hutt Valley. This has culminated in the development of the Hutt River flood management plan. 13

The study was limited by the technology and tools available in the 1990s. If the study were repeated today a number of new methods, such as the Virtual Climate Network described by Tait et al (2006), could be used to provide better prediction of rainfall, flood flows and flood levels. The percentage increases in daily rainfall from table 1 could be applied to the historical data of Tait et al to produce rainfall for a future climate scenario. However, this would still need to be reconciled with any projected impacts of climate change on mean and seasonal rainfall.

6.4 The Westport case study

The aim of this 2005 study was to assess the extent, depth and likely locations of areas that could be inundated in the coastal town of Westport as a result of the Buller River flooding. The Buller River has inundated Westport in the past, so this case study is not a case of ‘Will it occur?’ but rather ‘How bad might it be?’ The Westport case is complex because flooding can be exacerbated by sea conditions (ie, the state of the tide and onshore storm surge). Consequently, a simple screening approach, while a necessary preliminary step, was not expected to provide the complete answer.

6.4.1 Preliminary screening

The first step in the study was to carry out a preliminary screening, as follows.

  1. A weather model was used to produce rainfall for current conditions because there were not enough representative rainfall gauges in the catchment. The model used was RAMS (see Pielke et al, 1992) run at a five-kilometre resolution.
  2. The rainfall was used in a rainfall-to-flow model to produce flows in the Buller River, and the flow model was adjusted to obtain a satisfactory comparison with measurements.
  3. Rainfalls were then increased using a percentage increase per degree Celsius, in line with the Ministry for the Environment guidelines at the time, Preparing for Climate Change: A Guide for Local Government in New Zealand (Ministry for the Environment, 2004b, table 7). New flood flows were generated from the new rainfalls.
  4. The flood flows were then fed into a hydraulic model of Westport (with raised sea levels from the then current Ministry for the Environment guidelines) to generate the areas that would be inundated and the likely depths of inundation.

Based on the results from this screening, it was clear that significant areas of Westport could be at risk of inundation if rainfall amounts increased. Accordingly, a more sophisticated analysis was warranted.

6.4.2 Advanced analysis

A new analysis was undertaken, working through the three steps for estimating changes in rainfall, converting rainfall to flood flows, and then converting those flood flows to flood inundation. Details are given in Gray et al, 2005.

Step 1: Estimate rainfall

What changes would occur to the spatial rainfall distribution as a result of climate change? A range of climate change scenarios were considered to allow an assessment of the probable risks associated with the choice of a particular climate change scenario. Three historical events (12 November 1999, 17/18 August 2000 and 8 December 2001) were selected for study. In the detailed analysis, the simple scaling used in the screening process was replaced by modelled rainfalls for three possible future climates, with air temperature increased by 0.5, 1.0 and 2.7 degrees Celsius. The rainfall increased by 3, 5 and 33 per cent on average for the three temperature scenarios, through both an increase in the water holding capacity of the air as well as through changes in the intensity of the storms. The advantage of this approach over the simple scaling was that changes in the locations of areas of intense rainfall could be taken into account in generating the corresponding flood flows.

Step 2: Convert rainfall to flow

Each rainfall scenario was run through the rainfall-to-flow model of the catchment to derive the whole input hydrograph at an hourly time step. The TopNet model was used in this study. Figure 9 shows for the mid-high 2080 scenario there is a significant projected increase in peak flow on the Buller River at Te Kuha. The example shown in figure 9 compares recorded data for a flood in December 2001 (a river flood peak of this size occurs about once every four years at this site) against modelled river flow for the mid-high scenario in 2080. The modelled flow was produced using a TopNet model, which had been verified against recorded data (Gray et al, 2005, figure 8). There is a large change in projected flood peak magnitude for this example.

Figure 9: Comparison of flood hydrographs for the Buller River at Te Kuha for the base case (thick line) and mid-high 2080 scenario (thin line)

p class="description-link">Read a description of this image

Figure 9: Comparison of flood hydrographs for the Buller River at Te Kuha for the base case (thick line) and mid-high 2080 scenario (thin line)
Shows a comparison of flood hydrographs for the Buller River at Te Kuha for the base case (thick line) and mid-high 2080 scenario (thin line). The plots show a time series of river flows over 7 days (2 December to 9 December). Each plot has two peaks, one early in the series, one around 7 December. The base case has a small peak of around 3000 m3/s early on, and a bigger peak of around 5000 m3/s on the 7th. The mid-high 2080 scenario has a early peak of around 5000 m3/s and a large peak of 9600 m3/s on the 7th.

Step 3: Convert flood flow to inundation

The various flood hydrographs were then fed into the hydraulic model, built using the Hydro‑2de system (see Beffa and Connell, 2001), of the Buller River and the low-lying areas on the true right bank, where Westport is situated. Hydro-2de is a two-dimensional flood inundation model (see section 5.3.2). Using the morphology of the river channel and the topography of the land, the model calculated the water levels and velocities on the river flood plain, which includes Westport. The flow of water in the town was analysed with respect to its location and depth, and maps were produced that showed the inundation hazard.

Inundation modelling of Westport is complicated by the fact that inundation can occur as a result of a river flood, a high sea level brought about by a high tide or an onshore wind, or a combination of these factors. For the Westport case an annual exceedance probability (AEP) of inundation was pre-set at the design value of 0.02, corresponding to the 50-year average recurrence interval standard for floor levels from the Building Act 2004. A joint probability distribution was used to determine what this meant in terms of river flood AEPs and sea-level AEPs. This is because the seriousness for inundation of a moderate flood in the river can be exacerbated by a high sea level. To help understand what combination of river flood and sea-level events could lead to an inundation AEP of 0.02, a series of inundation model runs were made with different sets of flood and sea-level data. For each run the area of inundation was extracted. From this data a graph that shows the percentage of Westport that would be inundated for different AEP values could be constructed.

Once the base case and the relevant graphs had been produced for the selected AEP, investigating the effect of climate change impacts on rainfall amounted to supplying the inundation model with the revised flood and sea-level data for each scenario and then calculating the area of the town that would be inundated. The new probability of this occurring was then estimated.

The results of these calculations are summarised in table 10. In this case the climate change scenarios included different target years (2030s and 2080s), as well as an assessment of the effects of uncertainty in climate change projections by looking at both a mid-low and a mid-high scenario. Note that the climate scenarios used here were based on the 2004 Climate Change Effects manual, which has since been updated.

Table 10: Summary of 2 per cent AEP flood inundation results for Westport
Scenario Temperature rise
°C
Rainfall enhancement
%
Flow enhancement
%
Peak flow
(m 3 /s)
Water-level rise
(m)
Peak sea level
(m)
Inundation extent: % Westport
Base case 0 0 0 8,785 0 1.24 4.3
Mid-low 2030s 0.5 3 3.9 9,180 0.065 1.27 12.8
Mid-high 2030s 1 5 9.7 9,692 0.15 1.35 29.6
Mid-low 2080s              
Mid-high 2080s 2.7 33 37.5 12,198 0.49 1.69 79.1

Source: Gray et al, 2005

Table 10 indicates that under the base case (current climate, flood protection works as at 2005), 4.3 per cent of Westport has a 1-in-50 chance of being inundated each year. Under the mid-low 2030s climate scenario, 12.8 per cent of Westport has a 1-in-50 chance of being inundated each year; this assumes that flood protection works and other relevant hydraulic features do not change. Similar interpretations can be made for the other rows in the table. There are many combinations of river flood and sea level that could cause the same flooding: the combination shown in each row of table 10 is the most likely of the many river–sea-level combinations that can cause the same inundation.

The potential impact of the mid-high scenario for the 2080s is graphically illustrated in figure 10.

Figure 10: 1-in-50-year inundation areas in downtown Westport, with the present climate (top) and a mid-high scenario for 2080 (bottom)

1-in-50-year inundation areas in downtown Westport, with the present climate

and a mid-high scenario for 2080

Figure 10: 1-in-50-year inundation areas in downtown Westport, with the present climate (left) and a mid-high scenario for 2080 (right)
The figure shows two aerial maps of Westport. The left hand map is for the present climate and shows that only 4% of Westport would get flooded in a 1 in 50 year event. Levels of inundation in that event are shown to be generally below 0.5 m. The right hand map shows the flooding that might be expected from a 1 in 50 year event in 2080 for a mid-high scenario. The map shows that 80% of Westport is flooded, with levels exceeding 1 metre in many places.

Note: Both scenarios assume flood protection works as at 2005.

6.4.3 Notes

The Westport study required a number of simplifying assumptions to make it tractable.

  • The effect of climate change on the frequency of extreme events was ignored, even though extreme events can be expected to occur more frequently.
  • The relative humidity of the initial conditions has been assumed to be unchanged from the base case. This limits the potential of the atmosphere to hold water and could lead to an underestimation of rainfall.
  • The effects of climate change on land use were ignored.
  • The effects of changes in temperature on freezing levels, and hence snow melt and soil permeability, were ignored.
  • It was assumed that river beds would scour in a predictable manner, and so random factors such as a debris jam on the bridge were ignored.
  • The rise in river-bed level near the coast that would accompany a general sea-level rise (see section 5.1) was not included.

These factors are in addition to the uncertainties inherent in the climate change-induced rainfall scenarios, related to both the model and the input data, and to natural changes in things such as the El Niño / La Niña pattern.

Despite these limitations, the study has shown that climate change can be expected to exacerbate the inundation of parts of Westport. The study provides guidance on how potential future inundation can be reduced by showing where inundation is likely to occur and where it will be greatest. Serious consideration must be given to how best to reduce the effects of the increasing flooding that could occur in future due to the climate-induced changes in river flooding and sea-level rise. The consequences of a mean sea-level rise of at least 0.8 metres (as recommended in the Coastal Hazards and Climate Change manual, Ministry for the Environment, 2008b) have not been investigated.

6.5 Leith Lindsay’s flood protection scheme, North Dunedin

The Leith Stream and its tributary, Lindsay’s Creek, pose a flood hazard in the reaches that flow through the urban area of North Dunedin. Damaging floods occurred in 1877, 1923 and 1929, and since then flood channel enhancement and flood protection works have been undertaken to reduce potential flood damage. The Otago Regional Council has been undertaking a programme of reviewing and enhancing the flood protection works, particularly in the vicinity of the iconic river reaches in the grounds of Otago University.

To allow for climate change effects, a study was undertaken to look at the possible changes in flood risk as a consequence of temperature increases. The approach was to assemble records of storm rainfalls and flood flows for recent events recorded over the Leith catchment. A rainfall-losses/run-off routing model, calibrated for recent storms, was used to assess the expected increase in peak flood flow resulting from expected increases in design storm rainfall intensities. Finally, both laboratory and mathematical hydraulic models were used to assess an expected increase in water levels and therefore the increase in flood hazard.

6.5.1 Estimate rainfall

The study used expected annual mean temperature changes by the 2080s in the range 0.4 to 3.1 degrees, as recommended for the Otago area in the 2004 Climate Change Effects manual, to increase design storm rainfall intensities. These changes suggested, for example, that 12-hour duration, 1-in-100 annual exceedance probability (AEP) rainfall intensities should be expected to increase by between 3 and 21 per cent.

Note that the revision of the 2004 manual (published in 2008) altered the expected temperature change for Otago by 2090 to 2.0 degrees average and a range of 0.8 to 4.6 degrees, based on the six SRES scenarios in table 3. It also suggested the 12-hour duration, 1-in-100 AEP rainfall intensities should be expected to increase by between 6 and 37 per cent, with an average value of 16 per cent.

6.5.2 Convert rainfall to flow

The study used the upper limit percentage increases in storm rainfalls (21 per cent) with a calibrated rainfall-losses/run-off routing model for the Leith Lindsay catchment to estimate that the 1-in-100 AEP flood peak for the Leith Stream above the tidal limits could increase from the present-day value of 171 cubic metres per second to 200 cubic metres per second – a 17 per cent increase. This is in contrast to the Westport example, where smaller percentage increases in rainfall led to larger percentage increases in flow.

6.5.3 Inundation aspects of the study

The design flood estimates were used with laboratory and mathematical models to assess the performance of the proposed works. The results showed the scheme would perform safely during floods of increased magnitude. Accordingly, it was concluded the proposed scheme would perform safely under the extreme and long-range climate change scenarios developed using the Ministry for the Environment’s (2004a) guidelines.

6.5.4 Notes

In the context of the Leith flood protection works, the council has recognised the flood magnitude for a given standard of protection is expected to increase, but also that there is considerable uncertainty about the magnitude of the increase. The works have been designed to allow for enhancement to maintain the protection standard, if that proves necessary in the future. The council’s strategy to allow for climate change was one of a number of concerns about the scheme that were the subject of an unsuccessful appeal to the Environment Court (Gillies and Johnstone v Otago Regional Council 2008). 14 In summing up the reasons for dismissing the appeal, the Court concluded that if the level of protection is considered inadequate in the future, there is the potential for further works to be undertaken.


12 A statistical procedure that is used to take a series of annual peak flood flows and derive the magnitude of events expected to occur on average, say, once every 50 years.

14 Gillies and Johnstone v Otago Regional Council, unreported (23 May 2008) ENVC, Christchurch, C 060/08.