Rainwater Harvesting during the COVID Outbreak: A Case Study in Brazil †

: This work assessed the potable water savings potential for different scenarios in a ﬂat in Florian ó polis, Brazil. An uncertainty analysis was also performed to understand which parameters most inﬂuenced the results. First, it was necessary to evaluate the water consumption and calculate the water end-uses during a home-ofﬁce period due to the coronavirus pandemic. The water end-uses were obtained by monitoring the users’ consumptions for sixteen days and comparing them with the water meter on a daily basis. The results were very close to those measured using the water meter, with an average absolute error of 5.6%. The base consumption was, on average, 249.2 litres per capita per day. With a home-ofﬁce regime and an uninterrupted occupation, the coronavirus pandemic could be postulated to justify the more intense consumption patterns. Regarding the percentage of non-potable end-uses, an average of 25.8% was obtained. Potable water savings were simulated using the computer program Netuno, version 4. Seventy scenarios were evaluated, including different rainwater catchment areas and water and rainwater demands. The main results were that rainwater harvesting through a reduced area, 17.5% of the roof, obtained signiﬁcant results, compared to the simulation with the whole roof, with a potable water savings potential of 16%.


Introduction
One of the simplest methods to optimise water consumption is to return to the ancient knowledge of rainwater harvesting systems and use rainwater for non-potable purposes in buildings. Rainwater harvesting is a technique that has been widely known and disseminated in society for thousands of years [1]. According to Gnadlinger [1], there is no single reason why rainwater is no longer the focus of water harvesting techniques. However, the author comments that some factors were climate change, with droughts generating local inefficiency of systems, the desire for a centralised water management system, and the focus on large water supply projects, such as dams and wells.
Studies on the potable water savings potential through the implementation of rainwater harvesting systems are abundant in the literature. Examples of residential buildings [2], industries and agriculture [3,4], schools and universities [5,6], and offices [7] are found in the literature. Local rainwater harvesting can decrease the number of distribution pipes and reservoirs, thus decreasing leakage losses. Lower volumes of water are also withdrawn from rivers and aquifers, which benefits the environment. Finally, using rainwater for non-potable purposes decreases the amount of water treatment chemicals. Studies on life cycle analysis (LCA) have also demonstrated the potential to decrease the environmental impacts of water supply through rainwater harvesting systems [8].

Method
The case study consisted of two parts. The first was the water consumption analysis, which measured the use frequencies and flow rates of the appliances while monitoring the water meter of one flat. The measurements were made during the pandemic period, with the social isolation of the users. The second part consisted of simulating the potential for potable water savings through the theoretical implementation of a rainwater harvesting system.

Object of Study and Water Monitoring
A flat in a multi-family residential building in Florianópolis, Brazil, was chosen to be evaluated for the design of a rainwater harvesting system. The definition considered estimating the water end-uses by monitoring the two residents for sixteen days. Figure 1 shows the location and floor plan of the flat. The green area shows the roof area owned solely by the flat owners, and the red area shows the part of the roof shared with the building and its residents. The flat is located at longitude 48 • 30 08 west and latitude 27 • 36 12 south.
The monitoring of water consumption was carried out through questionnaires on the uses of the water appliances. Both residents filled out forms for sixteen days. The specific questionnaires for each room presented items regarding the environment, water appliance, frequency of use in the day, and the flow rate (litres/s, litres/cycle, litres/discharge, or litres), which were counted between 00:00 and 23:59 each day.
The volume of water consumed in each water appliance was measured to calculate the water flow. A pre-established volume was measured for showers, taps, and sinks, and the filling time was recorded. For these appliances, an average of three measurements was taken. Regarding bowl-and-tank toilets, the volume of water in each flushing was measured for half and full flush.
For appliances with cycles, such as the washing machine and dishwasher, the consumptions indicated in the appliances' manuals were used. For the drinking water consumption, we used the consumptions indicated by the users at the end of the day, considering the average number of glasses of water drunk and the glass volumes. For appliances with cycles, such as the washing machine and dishwasher, the consumptions indicated in the appliances' manuals were used. For the drinking water consumption, we used the consumptions indicated by the users at the end of the day, considering the average number of glasses of water drunk and the glass volumes.

Rainwater Harvesting System
The rainwater harvesting system was modelled based on similar works and using the computer program Netuno, version 4. The program, created by Ghisi and Cordova [14], was based on a deterministic water balance similar to the yield-before-spillage and yield-after-spillage approaches. As for the simulation parameters, the program required local pluviometry data, water consumption characteristics, technical definitions (such as first flush volume and the runoff coefficient of the roof), and upper and lower tank volume definitions.
The goal was to analyse the potential for potable water savings under the consumption found during the pandemic times. However, there is much variability within some of the parameters. To include the uncertainty analysis in the simulation result, 70 different scenarios were modelled based on the range of three parameters. Two values were used for the harvesting area, combined with seven water demands and five rainwater demands. Table 1 shows the data used in the simulations. Range between 1000 and 6000 litres (step of 250 litres) * These results were found during the first part of the research and are presented in Section 3.1.

Rainwater Harvesting System
The rainwater harvesting system was modelled based on similar works and using the computer program Netuno, version 4. The program, created by Ghisi and Cordova [14], was based on a deterministic water balance similar to the yield-before-spillage and yieldafter-spillage approaches. As for the simulation parameters, the program required local pluviometry data, water consumption characteristics, technical definitions (such as first flush volume and the runoff coefficient of the roof), and upper and lower tank volume definitions.
The goal was to analyse the potential for potable water savings under the consumption found during the pandemic times. However, there is much variability within some of the parameters. To include the uncertainty analysis in the simulation result, 70 different scenarios were modelled based on the range of three parameters. Two values were used for the harvesting area, combined with seven water demands and five rainwater demands. Table 1 shows the data used in the simulations. The two harvesting areas were modelled to represent the whole roof and the private part of the roof, including the shared and individual parts. This division occurred because one part was owned solely by the flat owners while the other was shared with other building residents. To simplify the results, the private roof is stated as PR and was 22 m 2 . Shared plus the private roof is stated as SPR and was 126 m 2 . Figure 1 shows the shared part in red, while the external boundaries in green show the private area.

Water Consumption and End-Uses
Water consumption was gathered and analysed in comparison to the water metering. The results were similar, with an absolute mean error of 5.6%. This similarity meant that, on average, daily estimates of water consumption varied by ±5.6% compared to the values registered on the water meter. Daily average water consumption via water meters was 249 litres/capita/day, and the average non-potable water use was estimated to be 25.8%. We considered rainwater could be used only for the washing machine and toilets as a non-potable source. Table 2 shows the water flow for each of the water appliances. M (millilitres) and T (seconds) stand for the measurement and timing, according to the method shown in Section 2.1. For the different scenarios of potable water demand (range between −15 and +15%), the minimum and maximum daily water consumption ranged between 250 and 750 litres/day. This range showed how much variability was found within the measurements of daily water demand. Additionally, the daily average water consumption obtained was higher than in previous literature, which states an average figure of 150 litres/capita/day as the Brazilian pattern.
The distributions of the water consumption within rooms and water devices are shown in Figure 2a,b. Figure 2c shows the measured versus metered water consumption. One can see that most of the water consumption occurred in bathroom 1, bathroom 2, and the kitchen, with little demand in the other rooms. Regarding water appliances, consumption was higher for showers, kitchen taps, and toilets. the kitchen, with little demand in the other rooms. Regarding water applianc consumption was higher for showers, kitchen taps, and toilets.
Comparing the results to those of Freitas and Ghisi [16], one can see that this c study flat had a higher water consumption per capita than other studies in the sa region. Additionally, they obtained a non-potable water consumption of 42.2% of the da water demand.

Rainwater Harvesting System
The comparison between the PR and the SPR was the first assessment, resulting i difference of approximately 8% in potable water savings. This assessment was perform with the baseline consumption and the 25% non-potable water end-uses, presented Section 3.1. By harvesting rainwater with the 126 m 2 roof, 24% of potable water savin were obtained, while with the 22 m 2 alternative, 16% savings were obtained. Both resu were obtained with a lower tank of 3000 litres, which was indicated as the optim technical solution.
The second assessment was the uncertainty analysis within the results obtained the water metering. In order to do so, the water demand was ranged, according to Ta 1. The results were then checked for PR and SPR. The main conclusion was that the PR smaller roof area for harvesting, presented more sensitivity to the total water demand this scenario, rainwater was scarcer, and the potable water savings potential dropp when higher water demand was included. For the SPR, almost all demands were met the rainwater harvesting systems.
The third assessment was the range of the parameter "rainwater demand", wh ranged around the figure of 25%. The main result was the opposite of the seco assessment, with less impact on PR and more on SPR. Such a result can be explained the analysis that SPR provided more rainwater. In this scenario, when non-potable wa was needed, rainwater would be available in response to a larger harvesting area. The area, on the contrary, did not present extra rainwater for the system, being less affec by the parameter. Comparing the results to those of Freitas and Ghisi [16], one can see that this case study flat had a higher water consumption per capita than other studies in the same region. Additionally, they obtained a non-potable water consumption of 42.2% of the daily water demand.

Rainwater Harvesting System
The comparison between the PR and the SPR was the first assessment, resulting in a difference of approximately 8% in potable water savings. This assessment was performed with the baseline consumption and the 25% non-potable water end-uses, presented in Section 3.1. By harvesting rainwater with the 126 m 2 roof, 24% of potable water savings were obtained, while with the 22 m 2 alternative, 16% savings were obtained. Both results were obtained with a lower tank of 3000 litres, which was indicated as the optimal technical solution.
The second assessment was the uncertainty analysis within the results obtained in the water metering. In order to do so, the water demand was ranged, according to Table 1. The results were then checked for PR and SPR. The main conclusion was that the PR, a smaller roof area for harvesting, presented more sensitivity to the total water demand. In this scenario, rainwater was scarcer, and the potable water savings potential dropped when higher water demand was included. For the SPR, almost all demands were met by the rainwater harvesting systems.
The third assessment was the range of the parameter "rainwater demand", which ranged around the figure of 25%. The main result was the opposite of the second assessment, with less impact on PR and more on SPR. Such a result can be explained by the analysis that SPR provided more rainwater. In this scenario, when non-potable water was needed, rainwater would be available in response to a larger harvesting area. The PR area, on the contrary, did not present extra rainwater for the system, being less affected by the parameter.
Both assessments were engaging, showing that PR harvesting proved to be a good alternative. However, if more non-potable (more than 25%) water is needed, the SPR alternative would become more attractive, requiring the approval of the remaining building users. Nevertheless, by dividing the potable water savings potential of the PR by the SPR results, a referential percentage of 65% was obtained. This result means that even with only 17.5% of the total roof area, the users might benefit from more than half of the potable water savings potential. It is easier to install and start a sustainable water practice within the flat. Comparing the results to Freitas and Ghisi's [16], one can see that this study had a much lower potable water savings potential, mainly due to the lower non-potable water demand. One can obtain higher figures in houses with gardens and patio cleanings.

Conclusions
The potable water savings potential ranged from 15.80% to 24.43% when considering both roof area possibilities. The results were higher than those obtained in the literature for multi-family buildings and lower than those found for single-family buildings. Additionally, it was found that even the smaller roof area proved to be an exciting approach for the users, starting a sustainable water practice in the flat.
The flat presented higher consumption than the region's average water consumption, and one can postulate that the continuous stay of users due to pandemic isolation might have influenced the results. The non-potable water demand percentage for the flat was similar to previous literature. Further studies can better understand the effects of different user patterns, helping to improve rainwater harvesting dynamics.

Data Availability Statement:
The data presented in this study are available on request from the corresponding author.