Poor resource management and infrastructure limitations make the effects of drought worse for cities in developing countries. One way to alleviate the impact without large investments is targeted demand management. This has worked well in studies that focused on some of the recent droughts, including Cape Town's ‘Day Zero’ drought of 2016–2018. Many studies have measured demand response to a drought at a coarse time resolution, but few have measured it at an hourly resolution or compared weekday with weekend use. In this study we evaluated households' hourly time-of-use behaviour in response to the Cape Town drought at two prominent inflection points identified by previous studies: the announcement of the Critical Water Shortages Disaster Plan in October 2017 and the introduction of Level 6B restrictions in February 2018. The first major reduction was caused by residents reducing their usage by about a third in the early morning and evening hours on weekdays, and the second, even larger, reduction was achieved in the mid-morning hours on weekdays when home owners were not at home but ensured that domestic workers used water sparingly.

  • Demand side management works better than supply side in a developing country.

  • Cape Town's severe 2016-2018 drought necessitated drastic demand management.

  • We assessed consumer behaviour change with time-of-day and day-of-the-week analysis of household water usage.

  • The first big reduction was due to residents reducing usage in the early morning and evening on weekdays.

  • The second occurred in the mid-morning on weekdays when residents were not at home but ensured their domestic workers reduced their usage.

From 2016 to 2018 the Western Cape province of South Africa experienced what was believed to be the worst drought in a century and was declared a disaster area (Bosman 2017; Botai et al. 2017; Ziervogel 2019; Brühl & Visser 2021). The severity of the drought resulted in the province not only anticipating but planning for the possibility of taps running dry. In an effort to curb demand and avoid what was referred to as ‘Day Zero’, the City of Cape Town introduced water restrictions, enforced by high tariffs, fines, and management devices on offenders' meters. The restrictions, communicated via mass and social media, progressively reduced the permitted daily water usage per person (Ziervogel 2019). Residential water users, especially the affluent, were obliged to reduce their water usage drastically (Brühl & Visser 2021; Cook et al. 2021). Two announcements in particular led to substantial reductions: the Critical Water Shortages Disaster Plan and the Level 6B water use restrictions (Booysen et al. 2019; Parks et al. 2019).

The finest response resolution used in drought response studies has been daily usage. Although studies have been done with sub-minute time resolution to assess usage profiles and identify end uses (Beal et al. 2011; Beal et al. 2013; Stewart et al. 2013), they were not done to assess usage response during a drought. Our study of behavioural change was not research-driven but a natural experiment. We looked at behaviour change as a response to an actual disaster. To the best of our knowledge, no studies have been done of changes in usage through the hour of the day and the day of the week in response to a drought. Both could be instructive in understanding and managing drought response and user behaviour.

Several big cities worldwide have been affected by severe droughts in recent years (AghaKouchak et al. 2014). They have had their drinking water threatened by severe water shortages. Examples are cities in Australia and Spain, and São Paulo, California, and Cape Town (Kenney et al. 2008; Lewis et al. 2011; Syme et al. 2011; AghaKouchak et al. 2014; Saft et al. 2015; Steffen et al. 2018; Brühl & Visser 2021; Pamla et al. 2021). Water shortages are especially devastating for developing countries because of overpopulation, high rates of urbanisation and financial constraints (Turok 2012; Cook et al. 2021; Egerer et al. 2021; Stoler et al. 2021). The possible increase in drought crises in big cities in the future gives urgent relevance to our study of drought management in Cape Town.

Internationally, the topic of change in consumers' water usage profiles during water shortages has been well researched. It has been found that getting the public to understand and carry out government policies during a drought was essential for minimising the severity and duration of the effects (Teo et al. 2013). The information released to the public via mass media is therefore crucial for changing consumer behaviour. Quesnel & Ajami (2017) found that the volume of mass media was important: household water usage changed much more during the San Francisco drought of 2012–2015 than during the 2007–2009 drought in that city, which received significantly less media attention. They found specific inflection points directly linked to certain announcements, such as the declaration of the state of emergency, which indicates the effect that mass media announcements about drought conditions can have on curbing water usage. The influence of mass media on consumers' water usage during a drought is therefore vital. It allows for improved management of the general supply and demand of water, encourages better preparedness for a future crisis, and makes it possible to set new norms for water usage after the drought (Teo et al. 2013; Anderson et al. 2018; Manouseli et al. 2018). Consequently, an in-depth understanding of how consumers adapt during a drought can help policy makers in their efforts to minimise the immediate-term and long-term effects of a drought. There is a need for more studies that investigate changes in residents' water usage profiles as a drought progresses (Anderson et al. 2018).

South Africa has produced a small number of studies of consumer water usage behaviour, particularly of the Western Cape drought, some of them quantifying the effect of nudges and behavioural interventions in reducing water usage (Brick & Visser 2017; Brick et al. 2017; Visser et al. 2019; Brühl & Visser 2021; Cook et al. 2021). These studies reported reductions ranging from 0.6 to 1.3% as a result of behavioural nudges, a 3% natural reduction in households whose initial usage was 50 kL per month, and a 48% reduction as a result of behavioural interventions targeted at affluent water users between 2015 and 2018 using monthly water bills.

Booysen et al. (2019) investigated in particular how water usage was affected by mass media announcements and the public's engagement with these announcements through social media and internet searches. They found that household users were more likely to reduce their water usage significantly if it appeared there was a high probability of their taps running dry and an imminent threat of having to queue for water at distribution points. They found that the most effective interventions to reduce water usage were the three-phase disaster plan announced in October 2017, which caused widespread alarm in the Western Cape, and the most severe restrictions, Level 6B, announced in February 2018. In this paper we assess in more detail, at a higher temporal resolution, the substantial behaviour change that Booysen et al. identified as resulting from these two interventions, and as a result of the mass media and social media frenzy of the time. For more detailed contextual information, see the analyses by Booysen et al. (2019), Parks et al. (2019) and Ziervogel (2019).

We expand on the existing body of research on drought response by exploring how successive interventions during the Western Cape drought affected household water usage. We look specifically at the effects on time-of-day usage. We investigate the effects on consumer behaviour of two key media announcements, the Critical Water Shortages Disaster Plan and the Level 6B water use restrictions, which conveyed to the public the severity and possible consequences of the drought.

Our analysis of water savings at the two inflection points found by Booysen et al. (2019) and Brühl & Visser (2021), indicating significant changes in usage that coincided with the announcement of the Critical Water Shortages Disaster Plan and the Level 6B water restrictions and the peaks in mass and social media activity, could help to explain changes in daily usage observed in other studies.

As the severity of the Western Cape drought increased, the City of Cape Town imposed water restrictions to curb demand. These became progressively stricter as the water crisis worsened. The following describes the timeline of the restrictions and our two influential events, the Critical Shortage Disaster Plan and the Level 6B water restrictions. The main events are shown in Figure 1 and described below.

Figure 1

Timeline showing major events and levels of restrictions (Brühl & Visser 2021) and the two inflection points identified by Booysen et al. (2019) and (Brühl & Visser 2021) that indicate significant behaviour change.

Figure 1

Timeline showing major events and levels of restrictions (Brühl & Visser 2021) and the two inflection points identified by Booysen et al. (2019) and (Brühl & Visser 2021) that indicate significant behaviour change.

The Western Cape Province was declared a disaster area on 22 May 2017, which would normally be the beginning of the rainy season. The province had progressively implemented water restrictions and the city was at Level 3 (Brühl & Visser 2021; Gittins et al. 2021). Thereafter, Level 4 restrictions limited usage to 100 litres per person per day. Level 4B reduced this to 87 litres and household water usage to 10.5 kL/month. Our analysis starts after this. On 17 August 2017, the mayor announced the Resilience Plan, which introduced fines for exceeding the daily water usage limit and the installation of water control devices. Level 5 restrictions maintained the 87 litres per person per day but increased the fines for non-compliance. On 4 October 2017, the Critical Water Shortages Disaster Plan was released to the public, causing much alarm, and press publications criticism of the authorities' management of the crisis. A press release was issued to calm the panic-stricken public. Next followed Level 6, on 1 January 2018, with conditions not much different from those of Level 5. Finally, on 1 February 2018, came Level 6B, which drastically reduced daily water usage limit to 50 litres per person. Our study focused on the announcements of the Critical Water Shortages Disaster Plan on 4 October 2017 and the Level 6B water restrictions on 1 February 2018. We compared residential time-of-day (TOU) water usage before and after each of these dates.

The Critical Water Shortages Disaster Plan, 4 October 2017

The City of Cape Town prepared the Critical Water Shortages Disaster Plan for implementation in the worst case scenario, if the drought became worse before the next rainy season. The plan had three phases, with restrictions becoming increasingly stringent. Phase 1, Preservation Restrictions, was immediately active when the plan was released. Urban water usage was to be reduced by 40% – in effect a daily water usage reduction of 500 million litres. This was to be done by cutting off the water for short periods in specific neighbourhoods. Phase 2, Disaster Restrictions, was to be implemented only if Phase 1 did not achieve sufficient reductions. At this point, water distribution points would be set up and water supply would be cut from a large number of businesses and households. Phase 3, Full-Scale Disaster Implementation, was to be introduced only if the Western Cape's water supply had no surface water available. Distribution points would be used to supply drinking water, which would come from ground water, spring water and aquifers and by delivery from other provinces.

The Level 6B water use restrictions, 1 February 2018

The Level 6B water restrictions introduced severe limitations on water usage. Usage was still restricted to 50 litres per person per day and 10.5 kL per month per household, with fines for exceeding the limit. All commercial water usage was to be reduced by 45%. Some detailed restrictions were emphasised, such as: municipal water was not to be used for swimming pools, portable pools and water features, or to establish new landscaping or sports fields, or to wash cars; borehole water was to be used sparingly, to avoid depleting underground resources; grey water was to be used to flush toilets, and toilets were to be flushed only when necessary. Level 6B was the last set of water usage restrictions before the next expected rainy season from April 2018. At the end of March, the expected Day Zero was extended to July.

We assessed changes in behaviour after the two events identified as particularly influential in changing usage: the announcements of the Critical Water Shortages Disaster Plan (4 October 2017) and the Level 6B water restrictions (1 February 2018). We evaluated the usage for each hour of the day before and the day after each announcement and then assessed the changes in usage profiles before and after the two events evident from the inflection point at the date of each of the announcements. We used four ‘before’ and ‘after’ periods to analyse the effects of the two announcements. These were chosen to assess the longest viable periods without including periods of exceptional use due to public holidays, such as Christmas and New Year, and other externally influenced periods of abnormal usage, such as school holidays, while taking data availability into account. The periods chosen for the first inflection point were six weeks before 14 October 2017 and six weeks after 1 November 2017, and for the second, one week before 22 January 2018 one week following 5 February 2018.

Before the water crisis, two types of metering devices were installed in homes in the Cape Town area for research projects: the ‘Dropula’, to measure cold water usage, in 16 homes, the ‘Geasy’, to measure hot water usage, in 33 homes (Booysen et al. 2019; Ripunda & Booysen 2019). Both devices measure volume at 0.5 litres resolution at a frequency of one sample per minute. Data were collected from October 2016 to February 2018 for the Dropula and from June 2015 to February 2018 for the Geasy. The participants in the study were self-selected volunteers from more affluent, high earning households that were inclined to be water conscious. Notably, their peer group had been classified as high water users (Brick & Visser 2017; Brick et al. 2017). Although the sample set does not represent the Cape Town's population in general, their behavioural response provides interesting possible explanations of the reductions seen by Booysen et al. (2019).

Response to the Critical Water Shortages Disaster Plan: the first inflection point

Announcement of this plan in October 2017 made a visible change in hourly behaviour. Figure 2 shows the time-of-day water usage profile before and after the first inflection point for the median user in each dataset. The change in absolute usage and as a percentage are shown. The results are further disaggregated into weekdays, weekends, and hot and cold water. The aggregated results are given in Table 1 with a coarser disaggregation into four blocks per day.

Table 1

Aggregated usage before and after first inflection point, with disaggregated time segments

Before
After
Reduction
[L/h][L/d][L/h][L/d][L/h][L/d][%]
Cold Weekday  249  179  70 28% 
 (00:00–05:00) 0.2  0.2  0.03  14% 
 (05:00–08:00) 13.3  9.1  4.23  32% 
 (08:00–17:00) 15.8  11.9  3.90  25% 
 (17:00–00:00) 8.7  5.8  2.92  34% 
 Weekend day  157  108  49 31% 
 (00:00–05:00) 0.2  0.3  −0.11  −56% 
 (05:00–08:00) 3.0  2.0  0.94  32% 
 (08:00–17:00) 10.5  7.3  3.18  30% 
 (17:00–00:00) 8.3  5.3  2.93  36% 
Hot Weekday  46  36  10 23% 
 (00:00–05:00) 0.0  0.0  0.00  – 
 (05:00–08:00) 4.2  2.9  1.28  31% 
 (08:00–17:00) 2.6  2.3  0.38  14% 
 (17:00–00:00) 1.0  0.7  0.32  32% 
 Weekend day  25  17  8 34% 
 (00:00–05:00) 0.0  0.0  0.00  – 
 (05:00–08:00) 0.4  0.4  0.00  0% 
 (08:00–17:00) 2.0  1.1  0.88  44% 
 (17:00–00:00) 0.9  0.8  0.08  9% 
Before
After
Reduction
[L/h][L/d][L/h][L/d][L/h][L/d][%]
Cold Weekday  249  179  70 28% 
 (00:00–05:00) 0.2  0.2  0.03  14% 
 (05:00–08:00) 13.3  9.1  4.23  32% 
 (08:00–17:00) 15.8  11.9  3.90  25% 
 (17:00–00:00) 8.7  5.8  2.92  34% 
 Weekend day  157  108  49 31% 
 (00:00–05:00) 0.2  0.3  −0.11  −56% 
 (05:00–08:00) 3.0  2.0  0.94  32% 
 (08:00–17:00) 10.5  7.3  3.18  30% 
 (17:00–00:00) 8.3  5.3  2.93  36% 
Hot Weekday  46  36  10 23% 
 (00:00–05:00) 0.0  0.0  0.00  – 
 (05:00–08:00) 4.2  2.9  1.28  31% 
 (08:00–17:00) 2.6  2.3  0.38  14% 
 (17:00–00:00) 1.0  0.7  0.32  32% 
 Weekend day  25  17  8 34% 
 (00:00–05:00) 0.0  0.0  0.00  – 
 (05:00–08:00) 0.4  0.4  0.00  0% 
 (08:00–17:00) 2.0  1.1  0.88  44% 
 (17:00–00:00) 0.9  0.8  0.08  9% 
Figure 2

Daily cold and hot water usage profiles before and after first inflection point for weekdays and weekend days. Changes in volume used shown as absolute changes and percentage changes.

Figure 2

Daily cold and hot water usage profiles before and after first inflection point for weekdays and weekend days. Changes in volume used shown as absolute changes and percentage changes.

The four weekday median usage profiles in Figure 2(a) all show a peak in the morning between 8am and 10am. The cold water peak before the first announcement is 31.0 L/h between 8am and 9am, but it drops to 20.2 L/h in the period after the announcement – a reduction of 10.8 L or 35%. In fact, the new morning peak is 21.5 L/h and occurs between 9am and 10am, which is when most residents are not be expected to be at home. This probably points to the effect of domestic workers' activities on water usage (most of our users are affluent and would have domestic workers). This is corroborated by the fact that the hot water peak precedes the cold water peak. The before cold water profile gradually reduces until 5pm, while the after profile is generally lower except for a small increase as it approaches the before profile at 12pm (around lunch time) and then increases at 3pm, after which it follows the previous profile until 5pm. The first increase is probably related to lunch activities, and the second one to end-of-day washing up. From 5pm onwards the residents are expected to be back at home. Accordingly, the before profile increases from 5pm, with a lower local peak of 9.9 L/h at 6pm – probably related to children bathing and food preparation – and an evening peak of 11.0 L/h at 9pm, probably related to washing dishes and showering or bathing. As expected, the before profile then rapidly reduces towards midnight. During the same period, the after profile is substantially lower, with the 6pm peak reducing to an inconspicuous 6.8 L/h and the 9pm peak reducing to 7.3 L/h.

The hot water profile follows a similar pattern, with a morning peak for both before and after at 7am, indicative of residents showering or bathing. The before profile peak is 8.5 L/h, and the after 6.5 L/h, a reduction of 2 L/h or 23%. This would point to residents reducing water usage in the mornings by either showering for a shorter period or using less water to bath. For the rest of the day the reduction seems insignificant, except for a reduction at 7pm from 1.8 L/h to 0.5 L/h, which would be linked to reduced water usage for evening showers or baths.

Figure 2(b) and 2(c) shows the reduction from before the announcement to after the announcement, as a volume and a percentage reduction. Figure 2(b) shows that the largest volume reduction occurred between 7am and 2pm, with a peak reduction at 8am (10.8 L/h) and a secondary peak at 10am (9.3 L/h). The cold evening reduction seems fairly flat at approximately 3.5 L/h. The 8am reduction and the 7pm reductions are matched with reductions in hot water, as evidenced by the percentage reduction shown in Figure 2(c).

The aggregated results in Table 1 provide another perspective on these results. The median weekday daily water used before the first announcement is 249 L/d. It drops by 28% (a reduction of 70 L/h) to 179 L/d afterwards. The main reductions in cold water usage are during the period when the residents would be home and awake, 5am to 8am and 5pm to midnight, with respective reductions of 32 and 34%. These timeslots are matched in the hot water used on weekdays, with respective reductions of 31 and 32%. Like the results in Figure 2, the morning reduction is greater in volume. The weekend cold results in Table 1 further confirm that residents caused the reduction, with a fairly flat reduction ranging from 30% to 36% in the waking hours, and a large reduction of 44% in hot water usage during the day at weekends.

It is also informative to consider when savings were less. A noteworthy period is during weekdays between 8am and 5pm, when cold usage reduces by only 25% (3.90 L/h) and hot usage by only 14% (0.38 L/h), despite this period's dominant contribution to initial daily flow. This is the period when residents are mostly not at home but domestic workers are busy in the house. Another noteworthy point is the lower reduction in hot water usage during weekend evenings, which is only 9% (0.08 L/h), showing that residents may initially have been less willing to forego ‘comfort use’ during leisure time – perhaps as a respite from their saving efforts.

Response to the Level 6B restrictions: the second inflection point

The Level 6B restrictions also had a substantial impact on behaviour, as can be seen in Figure 3. Table 2 again captures the aggregated results, which paint a similar picture.

Table 2

Aggregated usage before and after second inflection point, with disaggregated time segments

Before
After
Reduction
[L/h][L/d][L/h][L/d][L/h][L/d][%]
Cold Weekday  127  75  51 41% 
 (00:00–05:00) 0.2  0.0  0.18  100% 
 (05:00–08:00) 6.0  4.7  1.32  22% 
 (08:00–17:00) 8.0  4.2  3.85  48% 
 (17:00–00:00) 4.9  3.1  1.78  36% 
 Weekend day  50  28  22 44% 
 (00:00–05:00) 0.2  0.0  0.23  100% 
 (05:00–08:00) 1.4  0.8  0.57  42% 
 (08:00–17:00) 3.1  1.6  1.49  48% 
 (17:00–00:00) 2.7  1.8  0.89  34% 
Hot Weekday  21  16  5 23% 
 (00:00–05:00) 0.0  0.0  0.00  – 
 (05:00–08:00) 6.0  4.7  1.32  22% 
 (08:00–17:00) 1.2  0.5  0.64  55% 
 (17:00–00:00) 0.3  0.3  0.03  9% 
 Weekend day  11  5  5 51% 
 (00:00–05:00) 0.1  0.0  0.05  100% 
 (05:00–08:00) 1.4  0.8  0.57  42% 
 (08:00–17:00) 0.6  0.3  0.30  46% 
 (17:00–00:00) 0.6  0.3  0.34  58% 
Before
After
Reduction
[L/h][L/d][L/h][L/d][L/h][L/d][%]
Cold Weekday  127  75  51 41% 
 (00:00–05:00) 0.2  0.0  0.18  100% 
 (05:00–08:00) 6.0  4.7  1.32  22% 
 (08:00–17:00) 8.0  4.2  3.85  48% 
 (17:00–00:00) 4.9  3.1  1.78  36% 
 Weekend day  50  28  22 44% 
 (00:00–05:00) 0.2  0.0  0.23  100% 
 (05:00–08:00) 1.4  0.8  0.57  42% 
 (08:00–17:00) 3.1  1.6  1.49  48% 
 (17:00–00:00) 2.7  1.8  0.89  34% 
Hot Weekday  21  16  5 23% 
 (00:00–05:00) 0.0  0.0  0.00  – 
 (05:00–08:00) 6.0  4.7  1.32  22% 
 (08:00–17:00) 1.2  0.5  0.64  55% 
 (17:00–00:00) 0.3  0.3  0.03  9% 
 Weekend day  11  5  5 51% 
 (00:00–05:00) 0.1  0.0  0.05  100% 
 (05:00–08:00) 1.4  0.8  0.57  42% 
 (08:00–17:00) 0.6  0.3  0.30  46% 
 (17:00–00:00) 0.6  0.3  0.34  58% 
Figure 3

Daily cold and hot water usage profiles before and after second inflection point for weekdays and weekend days. Changes in volume used shown as absolute changes and percentage changes.

Figure 3

Daily cold and hot water usage profiles before and after second inflection point for weekdays and weekend days. Changes in volume used shown as absolute changes and percentage changes.

The profiles in Figure 3(a), plotted on the same scale as those in Figure 2, clearly demonstrate the lower base from which behaviour changed. It is important to keep this in mind when considering the values in Table 2, which is also why the percentage metric is an important proxy for the change effort rather than the volumetric impact of the change.

The first thing that is apparent from Figure 3(a) is that the usage in the before profile has already reduced substantially from that of the after profile in Figure 2(a). As Table 2 shows, the daily usage was at a mere 127 L/d at the end of January 2018 (‘Before’ in Table 2), down from the 179 L/d in November 2017 (‘After’ in Table 1).

The four profiles in Figure 3(a) have a clear morning peak, similar to those in Figure 2. The cold water initially peaks at 16.4 L/h at 9am, while hot water peaks at 4.8 L/h at 8am, again indicating early usage for bathing or showering, followed by usage that did not include hot water. After Level 6b restrictions are introduced, the cold water usage at 9am reduces by 54% (8.8 L/h) to a mere 7.6 L/h and the peak shifts to 8am, which reduces by 32% (4.4 L/h) from 13.8 L/h to 9.4 L/h. This distinction is important, as residents are likely to be responsible for the 8am change, and domestic workers for the 9am change. The corresponding reduction in the hot water use is from the 4.8 L/h at 8am to 4.5 L/h, a 6% reduction. This possibly indicates that the potential for reducing the amount of water used for showering and bathing was already exhausted at this point, with the remaining potential for saving limited to end uses that depend on cold water.

The before profile drops sharply from 9am to 12pm, after which it stays mostly flat apart from small localised peaks around 6pm and 10pm. The profile after the inflection point shows a substantial drop until 12pm, after which the reduction is minimal in terms of volume. This is clearly shown in Figure 3(b), with a volume reduction in cold water used of 8.8 L/h at 9am, 7.8 L/h at 10am, 5.7 L at 11am, and no other reductions above 5 L/h for the day. The percentage reduction in Figure 3(c) also paints an interesting picture, albeit noisy due to the low baseline from which change occurred. The change in behaviour, with percentage change acting as a proxy, fluctuates at approximately 30% from 9am to 9pm. Despite the much lower levels of change in the hot water usage, it has the same levels of percentage change during these times.

When we look at the segmented times in Table 2, these changes start to make sense. The water reductions during weekdays are 41% (51 L/d) for cold water and 23% (5 L/d) for hot water. These changes are driven mostly by reductions during the hours from 8am to 5pm. During this weekday timeslot, the cold water reduces from 8.0 L/h by 48% (3.85 L/h) to 4.2 L/h and the hot water from 1.2 L/h by 55% (0.64 L/h) to 0.5 L/h. The only segment that has nearly as large changes is the period from 5pm to midnight, with a 36% reduction.

What makes the daytime change interesting is that most residents would be away during this time. It is therefore plausible to assume that this reduction was due to a behaviour change by the domestic workers, who do cleaning and washing – activities that use a large amount of water. It would seem therefore that usage by domestic workers was not taken into account in the first phase of the Critical Water Shortages Disaster Plan, but this was rectified in the more desperate second phase.

Again the absence of substantial changes is informative – the change in weekday usage after 5pm was only 9%, substantiating the deduction that potential for changes to showering and bathing during weekdays was already exhausted at this point, although substantial savings were still made in cold water usage in the evenings – this could indicate reduced flushing of toilets with municipal water, as instructed by the city.

For the weekends, the profiles of water usage before and after the second inflection point are shown in Figure 3(d). It is clear that hot water usage has already reduced substantially from the usage in November 2017, with cold water usage per weekend day reducing by 53% from 108 L/d to 50 L/d at the end of January 2018 and hot by 36% from 16 L/d to 11 L/d. This change was gradual, as shown by Booysen et al. (2019).

Looking at the weekend changes after the second inflection point, we see the volume changes are made up of reductions at 8am and 10am. Table 2 helps to interpret these results. It shows that the reduction during the waking hours of weekend days are mostly flat in terms of effort, ranging from 34% to 48%. The equivalent savings in hot water usage range from 42% to 58%, with the latter indicating a reduction in ‘comfort use’ during leisure time that was still apparent in November of 2017.

Interestingly, the aggregated percentage reduction after the second inflection point is substantially larger than after the first inflection point, despite the volume changes being lower, indicating a more desperate reduction when ‘Day Zero’ was said to be fast approaching.

A severe drought is a major threat to a city with a large and growing water-using population. Developed countries can more readily afford to respond in the hour of need with improved mitigation and prevention measures, including expensive supply-side infrastructure investments and purchasing power. However, in developing countries, with their faster rate of urbanisation, the response is often constricted on the supply side and largely limited to what can be done on the demand side, with users as the main target. One problem they face in particular, although this is not limited to developing countries, is that of inequality, which has the consequence of large-scale employment of domestic workers. Having multiple parties in a household, not all living there full time, can complicate demand side management of household water usage, and the subsequent data analysis.

In this paper we assessed the response to Cape Town's ‘Day Zero’ drought of 2017–2018. Despite an abundance of work assessing behavioural response to this specific drought and many others, mostly in developed countries such as the US and Australia, ours is the first to look at high-resolution and high-frequency measurements – half a litre sampled per minute and analysed at hourly intervals – of user behaviour in response to and in the throes of a severe drought. The results, admittedly based on a small set of smart cold and hot water meters in different households, demonstrate clear tendencies. Unfortunately such fine temporal and volumetric resolution comes at the cost of perspective. Accordingly, we opted to zoom into two key events during the drought identified by Booysen et al. (2019) and Brühl & Visser (2021), as clearly of significance: the 4 October 2017 Critical Water Shortages Disaster Plan, when the city had an estimated five months of water left at the beginning of the dry season, and the introduction of Level 6B restrictions on 1 February 2018, with dam levels below 15% and ‘Day Zero’ projected to be approximately 10 weeks away.

We observe that the reductions in total hot and cold water usage as a result of the first inflection point were approximately a third. These reductions were mostly driven by residents reducing their usage in the morning and evening hours during weekdays, and during the waking hours of weekend days. The reduction during work hours was substantially less, possibly indicating the exclusion of domestic workers from saving efforts. Although reduction in hot water use was similar after both inflection points, the leisure hours of weekend afternoons showed very little reduction after the first inflection point.

After the second inflection point, which was associated with more urgency, smaller volumetric savings were observed, but they were greater as a percentage of usage before that inflection point. Cold water usage reduced by approximately 40%, and hot by 23% during the week and 51% during weekends. However, the 41% reduction in cold water usage during the weekdays was from a larger base, and the largest reduction – a 48% reduction for the timeslot – was due to a reduction between 8am and 5pm. The volume reduction during this timeslot was more than double that of any other timeslot after the second inflection point. Given the profile of the users in our dataset, this reduction would have been mostly as a result of reduced usage by domestic workers, possibly indicating their inclusion in saving efforts after the Level 6B announcements.

Data cannot be made publicly available; readers should contact the corresponding author for details.

AghaKouchak
A.
,
Cheng
L.
,
Mazdiyasni
O.
&
Farahmand
A.
2014
Global warming and changes in risk of concurrent climate extremes: insights from the 2014 California drought
.
Geophysical Research Letters
41
,
8847
8852
.
https://doi.org/10.1002/2014GL062308
.
Anderson
B.
,
Manouseli
D.
&
Nagarajan
M.
2018
Estimating scenarios for domestic water demand under drought conditions in England and Wales
.
Water Science and Technology: Water Supply
18
(
6
),
2100
2107
.
https://doi.org/10.2166/ws.2018.035
.
Beal
C.
,
Stewart
R.
,
Huang
T. (A. )
&
Rey
E.
2011
SEQ residential end use study
.
Water (Australia)
38
(
1
),
92
96
.
Beal
C.
,
Stewart
R.
,
Gardner
J.
,
Fielding
K. S.
,
Spinks
A.
&
McCrae
R.
2013
MIND OR MACHINE? Examining the drivers of residential water end-use efficiency
.
Water (Australia)
40
(
3
),
1
5
.
Booysen
M. J.
,
Visser
M.
&
Burger
R.
2019
Temporal case study of household behavioural response to Cape Town's ‘Day Zero’ using smart meter data
.
Water Research
149
,
414
420
.
https://doi.org/10.1016/j.watres.2018.11.035
.
Bosman
R.
2017
Critical Water Shortages Disaster Plan – Public Summary October
.
City of Cape Town
.
Botai
C.
,
Botai
J.
,
de Wit
J.
,
Ncongwane
K.
&
Adeola
A.
2017
Drought characteristics over the Western Cape Province, South Africa
.
Water
9
(
11
),
876
.
https://doi.org/10.3390/w9110876
.
Brick
K.
&
Visser
M.
2017
Green Nudges in the DSM Toolkit: Evidence From Drought-Stricken Cape Town
.
https://doi.org/10.13140/RG.2.2.16413.00489
.
Brick
K.
,
De Martino
S.
&
Visser
M.
2017
Behavioural nudges for water conservation: experimental evidence from Cape Town
.
https://doi.org/10.13140/RG.2.2.25430.75848
Brühl
J.
&
Visser
M.
2021
The Cape Town drought: a study of the combined effectiveness of measures implemented to prevent ‘Day Zero’
.
Water Resources and Economics
34
,
100177
.
https://doi.org/10.1016/j.wre.2021.100177
.
Cook
J.
,
Brühl
J.
&
Visser
M.
2021
Distributional statistics of municipal water use during Cape Town's drought: implications for affordability, conservation, and tariffs
.
Water Resources Research
57
,
e2020WR028219
.
https://doi.org/10.1029/2020WR028219
.
Egerer
M.
,
Haase
D.
,
McPhearson
T.
,
Frantzeskaki
N.
,
Andersson
E.
,
Nagendra
H.
&
Ossola
A.
2021
Urban change as an untapped opportunity for climate adaptation
.
npj Urban Sustain
1
(
22
).
https://doi.org/10.1038/s42949-021-00024-y
Gittins
J. R.
,
Hemingway
J. R.
&
Dajka
J. C.
2021
How a water-resources crisis highlights social-ecological disconnects
.
Water Research
194
,
2021
.
https://doi.org/10.1016/j.watres.2021.116937
.
Kenney
D. S.
,
Goemans
C.
,
Klein
R.
,
Lowrey
J.
&
Reidy
K.
2008
Residential water demand management: lessons from Aurora, Colorado 1
.
JAWRA Journal of the American Water Resources Association
44
(
1
),
192
207
.
https://doi.org/10.1111/J.1752-1688.2007.00147.X
.
Lewis
S. L.
,
Brando
P. M.
,
Phillips
O. L.
,
Van Der Heijden
G. M. F.
&
Nepstad
D.
2011
The 2010 Amazon drought
.
Science
331
(
6017
),
554
.
https://doi.org/10.1126/science.1200807
.
Manouseli
D.
,
Anderson
B.
&
Nagarajan
M.
2018
Domestic water demand during droughts in temperate climates: synthesising evidence for an integrated framework
.
Water Resources Management
32
(
2
),
433
447
.
Pamla
A.
,
Thondhlana
G.
&
Ruwanza
S.
2021
Persistent droughts and water scarcity: households’ perceptions and practices in Makhanda, South Africa
.
Land 2021
10
(
6
),
593
.
https://doi.org/10.3390/land10060593
.
Parks
R.
,
McLaren
M.
,
Toumi
R.
&
Rivett
U.
2019
Experiences and Lessons in Managing Water From Cape Town
.
Quesnel
K. J.
&
Ajami
N. K.
2017
Changes in water consumption linked to heavy news media coverage of extreme climatic events
.
Science Advances
3
(
10
),
1
10
.
https://doi.org/10.1126/sciadv.1700784
.
Ripunda
C.
&
Booysen
M. J.
2019
Understanding and Affecting School Water Behaviour Using Technological Interventions
.
https://doi.org/10.31224/osf.io/c7ujh
.
Saft
M.
,
Western
A. W.
,
Zhang
L.
,
Peel
M. C.
&
Potter
N. J.
2015
The influence of multiyear drought on the annual rainfall-runoff relationship: an Australian perspective
.
Water Resources Research
51
(
4
),
2444
2463
.
https://doi.org/10.1002/2014WR015348
.
Steffen
W.
,
Rice
M.
,
Hughes
L.
&
Dean
A.
2018
Climate Change and Drought June 2018 Factsheet
.
Stewart
R. A.
,
Willis
R. M.
,
Panuwatwanich
K.
&
Sahin
O.
2013
Showering behavioural response to alarming visual display monitors: longitudinal mixed method study
.
Behaviour & Information Technology
32
(
7
),
695
711
.
https://doi.org/10.1080/0144929X.2011.577195
.
Stoler
J.
,
Miller
J. D.
,
Brewis
A.
,
Freeman
M. C.
,
Harris
L. M.
,
Jepson
M.
,
Pearson
A. L.
,
Rosinger
A. Y.
,
Shah
S. H.
,
Staddon
C.
,
Workman
C.
,
Wutich
A.
,
Young
S. L.
,
Adams
E.
,
Ahmed
F.
,
Alexander
M.
,
Asiki
G.
,
Balogun
M.
,
Boivin
M. J.
,
Carrillo
G.
,
Chapman
K.
,
Cole
S.
,
Collins
S. M.
,
Eini-Zinab
H.
,
Escobar-Vargas
J.
,
Ghattas
H.
,
Ghorbani
M.
,
Hagaman
A.
,
Hawley
N.
,
Jamaluddine
Z.
,
Krishnakumar
D.
,
Maes
K.
,
Mathad
J.
,
Maupin
J.
,
Owuor
P. M.
,
Melgar-Quiñonez
H.
,
Morales
M. M.
,
Moran
J.
,
Omidvar
N.
,
Rasheed
S.
,
Samayoa-Figueroa
L.
,
Sánchez-Rodriguez
E. C.
,
Santoso
M. V.
,
Schuster
R. C.
,
Sheikhi
M.
,
Srivastava
S.
,
Sullivan
A.
,
Tesfaye
Y.
,
Triviño
N.
,
Trowell
A.
,
Tshala-Katumbay
D.
&
Tutu
R.
2021
Household water insecurity will complicate the ongoing COVID-19 response: evidence from 29 sites in 23 low- and middle-income countries
.
International Journal of Hygiene and Environmental Health
234
,
2021
.
https://doi.org/10.1016/j.ijheh.2021.113715
.
Syme
G. J.
,
Nancarrow
B. E.
&
Seligman
C.
2011
The evaluation of information campaigns to promote voluntary household water conservation
.
Evaluation Review
24
(
6
),
539
578
.
https://doi.org/10.1177/0193841X0002400601
.
Teo
M.
,
Goonetilleke
A.
&
Ziyath
A. M.
2013
An integrated framework for assessing community resilience in disaster management
. In:
Proceedings of the 9th Annual International Conference of the International Institute for Infrastructure Renewal and Reconstruction, Risk-Informed Disaster Management: Planning for Response, Recovery and Resilience
.
Technical report
.
Turok
I.
2012
Urbanisation and development in South Africa: Economic imperatives, spatial distortions and strategic responses
. In:
International Institute for Environment and Development United Nations Population Fund Human Settlements Group
.
Visser
M.
,
Booysen
M.
,
Brühl
J.
&
Berger
K. J.
2019
Saving water at Cape Town Schools by using smart metering and behavioural change
.
https://doi.org/10.31224/osf.io/7kzwv
.
Ziervogel
G.
2019
Unpacking the Cape Town Drought: Lessons Learned
.
Technical report, March 2019
, pp.
1
26
.
https://doi.org/10.13140/RG.2.2.36540.69760
.
This is an Open Access article distributed under the terms of the Creative Commons Attribution Licence (CC BY-NC-ND 4.0), which permits copying and redistribution for non-commercial purposes with no derivatives, provided the original work is properly cited (http://creativecommons.org/licenses/by-nc-nd/4.0/).