Potential of efficient toilets in Hanoi, Vietnam

Residential water demand-side management is important in urban water resource management (Renwick & Green 2000; Tanverakul & Lee 2016). Water use can be managed adequately through both price and non-price means. Although the price of water is one of the most significant factors in decreasing residential water consumption (Grafton et al. 2001; Arbues & Barberan, 2004), there is a limit to pricing as a driver for achieving efficient water use. First, since water is generally regarded as a necessity for human life, price manipulation can be used only in limited cases (Renwick & Green 2000). Second, changes to water use behavior in response to price change take considerable time (Dalhusein et al. 2003; Nauges & Thomas 2003). Third, since the price elasticity of higher-income households is significantly greater than that of lower-income households (Renwick & Archibald 1998), only lower-income households decrease their water consumption in response to an increase in price (Renwick & Green 2000).

The non-price ways to alter water use behavior include public education campaigns, rationing, water use restrictions, water-efficient devices, and so on. Although public education campaigns can result in significant water savings (up to 25%) in short-term or crisis situations, no long-term effect has been observed (Syme et al. 2000). Rationing and water use restrictions are effective measures, and Renwick & Green (2000) reported observed reductions in water consumption of about 19 to 29%. Although it is possible to use such methods during a crisis, it is not realistic to think of them as permanent measures.

The installation of water-saving devices has been studied (Randolph & Troy 2008) and promoted by many water authorities, mainly in developed countries. Although water-saving devices appear to reduce water consumption, there have been controversial discussions about their effectiveness (Olmstead & Stavins 2009). Suero et al. (2012) demonstrated the effectiveness of replacing standard devices with water-saving devices through a study in California, Seattle, and Florida. Retrofitting 96 households with efficient showerheads, toilets, and washing machines reduced water consumption by 10.9%, 13.3%, and 14.5%, respectively. On the other hand, some authors report that only the replacement of toilets with more efficient versions is effective. For example, Grafton et al. (2001) estimated that of all water-saving devices, only the efficient (low-volume/dual-flush) toilet has a statistically significant effect on water consumption in the 10 Organisation for Economic Co-operation and Development (OECD) countries whose data were analyzed. The U.S. General Accounting Office (2000) also reported that changing to an efficient toilet resulted in water savings of 23 to 40 liters per day (L/d) based on data from 12,000 randomly-selected households in 12 study sites in the United States. Lee et al. (2011) conducted a four-year study in Florida and observed that high efficiency toilet rebate programs were effective in reducing water demand. Low et al. (2015), in a study in Melbourne (Australia) during the Millennium Drought, found that toilet replacement was ineffective. During the drought, the government funded water rebate and exchange programs for residential water users. Replacement of showerheads, toilets, and washing machines with more water-efficient models reduced demand by 5.5, 0.44, and 8.67 gigaliters per year (GL/year), respectively. Compared to the efficient showerheads and washing machines, the volumes saved by efficient toilets was relatively small.

In developing countries, rapid increases in water demand are expected alongside economic growth; residential water demand-side management is extremely important for sustainable water supply. The introduction of water-saving devices is supposed to be a useful way to achieve sustainable water use. A few studies have considered the possibility of conservation using water-saving devices in developing countries. There appears to have been only one trial, however, in the city of Florianópolis, southern Brazil, where Proenca et al. (2011) compared the water-saving potential of replacing conventional toilets (e.g., a toilet with wall flushing valve and bowl-and-tank toilet) with a dual-flush toilet, and/or gray- or rainwater use. According to their calculations, for residential water use, toilet replacement has the greatest water-saving potential at 14.7% of potable water. There is still a need, therefore, to verify the water-saving potential in developing countries. This study focused on toilet replacement and calculated potential residential water savings in urban Hanoi, Vietnam, a country presently enjoying remarkable economic growth and increasing water demand.

Conventionally, water-saving potential is calculated using single parameter values such as average toilet tank volume, average number of toilet flushes per person, and average family size. Suero et al. (2012) observed differences in water use before and after retrofits, and compared them to water savings predicted using simple analytical, regression, and hybrid models in Seattle. For toilet water use, both simple analytical and regression models were effective, and the technological and behavioral variables (i.e., toilet tank volumes and number of toilet flushes) had a significant influence on water savings. However, this sort of deterministic approach can prove problematic, because customer populations are heterogeneous and parameter values uncertain, differing between individuals (Whitcomb 1990; Rosenberg 2007; Abdallah & Rosenberg 2014). In order to consider the probability distributions for parameters, Rosenberg (2007) adopted Monte Carlo simulations. In this study, we also used a Monte Carlo simulation to take into account the current distribution of appliances and individual behavior.

This study looks at water use in 134 households in urban Hanoi, sampled by the snowball method. A questionnaire and measurement survey were used to define demographic, technological, and behavioral water use factors for each household. The information concerned 1) family size, 2) household monthly income, and 3) the declared frequency of toilet use at home. Measurements were made of 4) type of toilet (single- or dual- flush), 5) appliance performance (liters per flush), and 6) frequency (toilet flushes per household per day) to know their actual distribution.

The number of toilet flushes was measured using a dedicated sensor in each toilet tank (Otaki et al. 2017), which accumulated counts for several months. In dual-flush toilets, two sensors were installed to enable full and half flushes to be distinguished. The number of toilet flushes per person per day was calculated using the cumulative number of toilet flushes, family size, and the number of measurement days.

• W_i is the water saving of the selected toilet per day because of the improved efficiency (L/d)

• V_Li is the current full-flush volume of the selected toilet (L)

• F_Li is the current full-flush frequency per day of the selected toilet

• V_Si is the current half-flush volume of the selected toilet (L), and

• F_Si is the current half flush frequency per day of the selected toilet.

This calculation was carried out one million times. The flush form and volume of selected toilets followed the current distribution among households surveyed. The working life of a toilet was taken as 20 years (n = 20), although to accelerate toilet replacement using a rebate for efficient toilets, periods of five (n = 5) and three years (n = 3) were used. The water saving, converted into annual average amount per capita per day (L/p/d), was calculated for each n year to evaluate the effects of efficient toilets.

• W_analytical is the annual average water saving per capita per day (L/p/d) due to installation of an efficient toilet

• V_L0 is the full flush volume of the current toilet (L)

• F_L is the number of full flushes per capita per day

• V_S0 is the half-flush volume of the current toilet (L)

• F_S is the number of half flushes per capita per day, and

• n is the replacement interval.

The study population comprised people living within 10 km of central Hanoi. The mean number of people per household was 4.69 (SD = 1.65), with 6% of households having two persons, 16% three, 28% four, 22% five, 17% six, and 10% seven or more. Although the average household size was slightly larger than the 3.4 in the urban Red River Delta (The Vietnamese Population and Housing Census 2009), this is not significant because water consumption per person per toilet flush does not depend on family size.

Vietnamese houses often have more than three floors (Otani et al. 2015) and most houses have more than one toilet. The 134 households surveyed had 248 toilets between them, all of which were Western-style flush toilets.

The study covered 132 single-flush toilets 116 dual-flush units, and water savings could be expected by replacing the single-flush toilets with efficient, low-volume, dual-flush units. Flush type is not related to household income or family size (p > 0.05).

The median flush volume of all toilets surveyed was 6.3 L/flush (for a full flush) and 3.0 L/flush (for a half flush). These values are almost the same as for the popular efficient toilet. However, the large distribution range means that some large-flush toilets remain (Figure 1).

The average number of toilet flushes (both full- and half- flush) measured per person per day was 4.00, broken down into 2.47 full-flushes and 1.53 half-flushes (by converting from the ratio in dual-flush households). The average number of toilet flushes given in answers to the questionnaire was 4.50. As shown in Figure 2, there is no correlation between the measured flush frequency and the answers in the questionnaire survey. The line on the figure indicated where the questionnaire answers were correct. In other words, flush frequencies determined from a questionnaire are inaccurate. Questionnaires have been used in some previous studies to determine flush frequency (O'Toole et al. 2009; Otani et al. 2015; Tanaka et al. 2016), which may have resulted in inaccurate toilet water use estimates. No relationship was found between household income and number of toilet flushes daily.

Figure 3 shows the distribution of per capita toilet water consumption per day. Almost 80% of households used between 10 and 30 L/p/d, and median value was 20.5 L/p/d. According to the statistics (IBNET Database 2018), individual consumption in Hanoi was 117.6 L/p/d in 2014. Therefore, the ratio of toilet to total personal water use was a little less than 20%.

Figure 4 shows the annual average water-saving potential converted into per capita per day (L/p/d), derived from both the Monte Carlo simulation and the analytical model. Even if all toilets were replaced with efficient units within three years, the resultant savings would only be 2.2 L/p/d, so the effect of toilet replacement was limited in this study group. The analytical model indicated water savings 10 to 20% smaller than in the Monte Carlo simulation, no matter how long the replacement period was.

Water saving capacity estimates were based on the replacement of standard with water-efficient toilets, but the saving was quite small (2.2 L/p/d or 1.9% of individual consumption), even though it was assumed that all toilets were replaced with more efficient units within three years. This suggests that rebates to encourage toilet replacement are ineffective, contrary to experience in developed countries, although the value accounted for around 10% saving as far as toilet use was concerned. This may be because this study's focus was on Hanoi, where it is possible that relatively new devices are in use.

The current median flush volume in the households surveyed was about that of popular efficient toilets. As the current distribution of toilet flush volumes is an important factor for estimating water-saving potential, it is better to measure toilet tank volume in a sufficiently large number of households. In fact, the measured flush volumes (and frequencies) were smaller than those used in a previous study based on questionnaire answers. Otani et al. (2015) calculated the number of toilet flushes at home using a questionnaire in Hanoi. The values obtained, 2.1 times per day (full flush) and 4.8 times per day (half flush), were larger than our measured values – 2.47 and 1.53 respectively. There is a clear difference between actual behavior and the answers to the questionnaire, because of which direct measurement of the distribution of toilet flush volume and flush frequency is recommended.

In developing countries, the introduction of flush toilets is relatively recent and it must be expected that people will have newer toilets with smaller flush volumes. This results in less water-saving when changing toilets to more efficient units than is the case in developed countries. Among the homes studied, efficient toilets are already in the majority, although there are still some large-flush volume units in use.

Recently, the evaluation method has shifted from the deterministic approach (i.e. methods using the mean and representative values) to the stochastic one (i.e. methods using the probability distributions for each parameter). In this study, we took account of the probability distributions for toilet flush volume and flush frequency to avoid under-estimation of the water-saving potential.

The installation of water-saving toilets seemed to have limited potential effect on water consumption in urban Hanoi. This is thought to be because relatively new toilets are already in widespread use. In developing countries, new technologies tend to be introduced in a single, initial step. Since the water-saving impact of installing efficient toilets depends on the prevalence of older, less efficient toilets in general, it is important to investigate the prevalent toilet types, flush volumes, and typical flush frequencies, before determining policies.

Monte Carlo simulation can take the distributions of flush volume and frequency into account. It might thus minimize under- or over-estimation of the potential for savings.

This work was supported by Core Research for Evolutional Science and Technology (CREST) of Japan Science and Technology Corporation (JST) and JSPS KAKENHI Grant Number 17H01936.