The use of mean capacity estimates does not reflect the short-term water situation associated with below-mean dam storage levels, which elevates the risk of system failure in the short-to-medium term and is arguably the main driver of recent investments in desalination in Australia. Neither does it assess the risk of dam spills that can occur with a sequence of wet years. A system model was developed to assess the risks associated with timing of investment in the second (Binningup) desalination plant proposed for Western Australia under different climate-change assumptions. Inputs to the model are an assumed probability-distribution of catchment yields (representing each climate scenario), projected urban demand and availability of water from groundwater sources (both based on existing institutional arrangements). The model tracks dam storage levels over the next 10 years and measures the frequency of dam spills and of urban-supply restrictions (triggered by low storage levels). Each scenario is analysed from 1000 random sequences of weather patterns, drawn from the probability distribution for that climate scenario. The detail of the simulation model is described in Brennan (2008).

Source: Data provided by Water Corporation (Michael Loh, Personal Communication, October 2006) and Water Corporation (2005). Source of IOCI climate scenarios discussed in text.

In all of the scenarios, it is assumed that the Water Corporation’s proposed new desalination plant is providing water at full capacity in the summer of 2011/12. It is assumed that both desalination plants run at full capacity even when dams are full. While in reality these desalination plants would be switched off if rainfall significantly recovered from recent years, the assumption of full utilisation is useful to illustrate the extent of the potential surplus associated with desalination investment.

Results of the analysis are presented in the Tables 2 and 3. Table 2 shows the (unrestricted) demand forecast over time, and the risk of a sprinkler ban in each year for each climate scenario. As the climate assumptions become more conservative, the risk of a sprinkler ban increases. For all but the driest climate scenario, the risk of a sprinkler ban is already improving in the period prior to 2011. This suggests that the system reliability is in a state of recovery from the low dam reserves of 150GL at the end of the 2007 summer, due to the boost in supply provided by the Kwinana desalination plant. This confirms the average water-balance analysis presented in Table 1 — water in 2011 is not in a net-deficit situation for these climate scenarios. In contrast, for the most conservative climate scenario, the risk of sprinkler bans increases over the period prior 2011. The augmentation of supply reduces the risk of a sprinkler ban down to 3.4 per cent for this scenario in 2010/11. This would not be considered a very high risk in most capital cities for Australia; for example, planning for augmentation is usually done on a sprinkler-ban risk tolerance of one year in 25 (ACTEW 2004). However, in Western Australia planning is based on a one in 200 tolerance to the risk of sprinkler bans.

Also shown in Table 2 is average dam storage level, as measured at the end of winter (October), for each climate scenario. The simulation begins at the start of winter 2007 and thus has had two winters by the end of October 2008. There is already a substantial difference in end-of-winter storage levels for all scenarios. Compared to the actual storage level of 220GL at the end of October 2006, some recovery of dam storage levels is expected even for the most conservative of climate scenarios by winter 2008. However, there is already a large divergence in recovery of dam reserves by 2008 and this divergence increases over time. Expected end-of-winter dam levels for the most conservative climate scenario remain stagnant (and combined with growing demand imply a deteriorating reliability). There is a boost to dam storage levels in all scenarios following supply augmentation in 2011, because the new desalination plant leads to a reduction in the drawdown of dams. For the 1975–2006 climate scenario, dam storage levels approach capacity over the simulation period.

Table 3 presents the analysis of the risk of ‘dams running over top’. First, the risk of winter spills occurring is measured. Dams inevitably spill in high-flow years if they have a high inflow-to-storage ratio, and spills from the two dams in the Perth metropolitan system (Churchman and Stirling) that have high inflow-to-storage ratios were not included in the measures shown here. The risk of winter spills shows the risk that one of the other dams will be full enough to spill in a high-flow winter. In many cases these spills are not very substantial; so, in order to gauge events that involve serious loss of water (and would raise the question as to why the desalination plant is still running), the risk of spills exceeding 10GL was also calculated.

The risk of winter spills is very high for the 1975–2006 climate, and is exacerbated after 2011 with the introduction of the second desalination plant. In 2012, dams are expected to spill in 66 per cent of years. For the 2000–06 climate scenario, the risk of dam spillage in the near future is 6–7 per cent, but it increases to 27 per cent in 2012 following the accumulation of surface water once the second desalination plant is in use. For the IOCI worst-case climate scenario, the risk of dam spillage is substantially lower. For the Water Corporation’s planning scenario (2001–06 climate) there is virtually no risk of dams spilling because system yield is so low.

 

The high risk of dam spills for the 1975–2006 climate also corresponds to a high risk of spills that are of a large volume. If the supply augmentation occurs in 2011, the risk of dams spilling more than 10GL of water in 2002 is 54 per cent. In the case of the 2000–06 climate, the risk of dam spills exceeding 10GL in 2012 is 8 per cent. For the drier climate scenarios the risk is insignificant.

A comparison of the 2000–06 climate scenario with the 2001–06 scenario highlights the current decision-making dilemma regarding climate uncertainty. Adding the relatively wet year of 2000 has a substantial impact on the apparent desirability of supply augmentation. The 2001–06 climate scenario implies that the probability of a sprinkler ban is 16 per cent in 2010, prior to the introduction of the new desalination plant. In contrast, the 2000–06 scenario implies that there is little risk of a sprinkler ban even before the plant is introduced, and the risk of dams spilling more than 10GL per annum is 10 per cent by 2013. Using the ‘ruler and pen’ approach to climate forecasting means that our estimate of whether we will be ‘rooned’ by drought or ‘rooned’ by desalination obsolescence all hinges on whether or not we include 2000 in the time series.