Will Cape Town be the world’s first major city to run out of water?

By Nicky Allsopp, Manager, SAEON Fynbos Node

Residents queue to fill water containers in Cape Town (Picture: CNN)

Rainfall in Western Cape, where Cape Town is located, dwindled dramatically in the three years to 2017.

Capetonians are being asked to use 50 litres of water per person per day from their taps from 1 February 2018 in order to avert "Day Zero", the day the City says the taps will no longer run in residential areas.

Water use dropped to 550 million litres a day (on 5 February) from 600 million litres in the wake of this announcement, but the City wants water use to drop to 450 million litres/day in order to ensure that there is enough water in the dams to continue supplying water.

Day Zero is a moving target but is based on when dams supplying the City’s water reach an average of 13.5% full. With a steady reduction in water use by Capetonians this date has moved back four months from April through to July.

Many factors will determine the date water may stop running from residential taps and when water flow resumes. These include the levels of water use reduction people in the city can achieve and the rate at which supplementary water supplies from aquifers and desalination plants can be brought on line. Plans are for these to start augmenting water supply from May 2018, but Capetonians will remain reliant on rainfall refilling the dams for the majority of their water needs.

Other factors that determine the short- and longer-term water supply are when the winter rains will arrive and whether that rain will be enough to fill the dams to supply water through the winter and the next summer. Can SAEON’s data provide any light on these questions?

Figure 1. The probability that streamflow greater than the average monthly streamflow will start in a particular month from the long-term record (black bars). Blue bars show the probability that high flows begin in any month when the annual rainfall is 20% or more than the average, and the khaki bars show the probability of high flows beginning in a particular month when rainfall for that year was 20% or lower than average. Data from Langrivier, a gauged stream in Jonkershoek.

When will the streams start flowing again?

I looked at the first month in each year in which streamflow exceeded the long-term monthly average for the Langrivier (http://www.ecologi.st/post/2017-04-01-Langrivier/) and took this as the month that the streams filling Cape Town’s dams start delivering substantial water.

Langrivier in the Jonkershoek valley has its catchment in the same mountains as the Berg and Theewaterskloof dams and so can serve as a surrogate for streams filling these dams. To fill our dams we need streamflow to exceed this average in four or more months. In the current drought period, streamflow exceeded this average during one month in 2015, three months in 2016 and three months in 2017.

The usual date for high flows to start falls in May or June, occasionally this is as early as April and sometimes as late as July or August (Figure 1). In exceptionally high rainfall years this occurs quite commonly in April and in drier years substantial streamflow may only start in July or August.

We have no records of the first monthly streamflow exceeding the average in March. Based on probabilities, the later the onset of heavy rain, the more likely we are to get less total rain and hence lower streamflow.

From this data, it seems unlikely that there will be an early onset of rain that will increase dam volume in April irrespective of whether subsequent rainfall is low or high. The South African Weather Service is predicting that rains are likely to start late in the South Western Cape.

Will the dams fill up this winter?

With uncertainty of the amount and spread of rain in the coming season, can we look at the past for some help in understanding what might happen?

We can see the low rainfall reflected in low annual streamflow in Figure 2 and above-average rainfall results in high levels of streamflow. This figure allows us to judge if conditions similar to the present were experienced during the period from 1961.

There have been some fairly persistent dry and wet eras: 1963-1975 was generally drier, 1983-1993 generally wetter. Otherwise the graphs suggest high variability between years, with streamflow generally quite responsive to low or high rainfall in the same year with no clear suggestion that rain in one year influenced streamflow in the subsequent year. Our efforts of prediction are frustrated by the relative shortness of the record, the high variability in rainfall between years, missing data and the prospect that we are entering an era where global change is influencing weather patterns.

Since 2014 we see a decline in streamflow which is unprecedented in this 56-year record. If 2018 receives average or better rainfall, the previous record suggests that streamflow will recover quickly. But during this record the system hasn’t been as dry as the period 2015-17 and we see in Figure 3 that dam storage has dropped each year despite similar rainfall in those years. So dam levels may take a while to recover even if rainfall is high.

Figure 2: Percentage deviation of Langrivier stream flow (pink bars) and Dwarsberg rainfall (blue bars) from their long-term means. Bars below zero reflect rainfall and streamflow lower than average, and values above zero reflect rainfall and streamflow above average. Langrivier in Jonkershoek is the nearest gauged stream to the Dwarsberg weather station (lying 4 km westward), so we can expect that the patterns reflect general trends of water flowing into the Berg and Theewaterskloof dams. Years without bars are where data are missing or unreliable for either streamflow or rainfall.

Typically, the dams serving Cape Town fill up during the winter rainy season and store sufficient water to supply water for residential, industrial and agricultural use through the dry summer months.

The last four years have seen less than average rainfall falling in the catchments and water use has exceeded the capacity of this rain to refill the dams (Figure 3).

Figure 3. Cumulative rainfall at the Dwarsberg automatic weather station (left), situated in the mountains that form the catchment for both the Theewaterskloof and Berg River dams. Dam levels (right) as a percentage of total capacity on 5 February for the years 2013-2018 (dam level data supplied by City of Cape Town). The relatively higher volume of water in the Berg River dam reflects lower rates of abstraction from this dam compared with Theewaterskloof. At full capacity Theewaterskloof dam holds 480 billion litres compared with 130 billion litres in the Berg River dam (on 5 February 2018 there was 60 billion litres in the Theewaterskloof dam compared with 69 billion litres in the Berg River dam). Total dam storage for all the major dams supplying Cape Town is 25% on this date.

Dam volume is a dynamic value and reflects both the amount of water flowing into a dam and that being removed for water use (residential, industrial and agricultural use) and evaporation. Some of the things to notice about these dam levels are that while the dams deliver water in the dry season, they still retained a high volume of water in February of 2013 to 2015. But as the drought became prolonged over several years, dam levels dropped progressively. This is despite peak demand by the city falling from a high of 12 million litres/day in mid-summer 2014/15 to half this at present and very similar rainfall totals for 2015-2017.

While SAEON’s data do not provide a crystal ball for foretelling the future, the data suggest that while in the short-term “Day Zero” may be averted through water savings, unless streams start flowing early in the winter, dam levels will not rise significantly before “Day Zero” in July. In the medium term, dam levels will only rise sufficiently to supply water into next year if rainfall is high, since current augmentation schemes will only supply a portion of the daily needs of Capetonians.