Understanding the traditional wisdom of harvesting rainwater in household yards: construction and rainwater usage patterns of settlement water cellars in semi-arid China

ABSTRACT This study focuses on the heritage users and artisans of water cellars – the principal source of potable water in semi-arid China for centuries – to examine their construction and rainwater usage patterns. The main advantages of water cellars are their low construction cost, user-friendly technology, and water usage stabilization features. The concrete impermeable layer was introduced by artisans in 1970, but residents’ and artisans’ preferences for raw material differ. Furthermore, the survey results showed that the daily average water-use per person was 28.5 L, when there was no drought and harvested rainwater was used for bathing. In contrast, about 14 L of water per person is required daily, when usage is kept to a minimum for bathing and laundry, in the event of a drought. The wisdom behind such adjustments is irrefutable, but it is inadequate to alleviate the discomfort of extreme water-saving. This evidence-based awareness of the advantages of water cellars – as well as the limitations of traditional wisdom – will facilitate their sustainable development.


Introduction
Since time immemorial, traditional rainwater harvesting systems have helped sustain the human race (Akpinar Ferrand and Cecunjanin 2014), and the rainwater harvesting technology still aids in solving present-day water shortage and artificial recharge problems (Krishna et al. 2020). Over time, distinctive construction features and usage traditions of these systems have been revealed through research, but most of the attention of academics and the general public has been devoted to large-scale, conspicuous, and culturally significant systems, such as stepwells (Das and Sharma 2002;Priya 2010) and qanats (Megdiche-Kharrat, Ragala, and Moussa 2019; Razzaghi-Asl 2017). The smaller, conventional -and importantly, inconspicuous -systems are ignored until they require maintenance. This study focuses on water cellars constructed via conventional technologies in the traditional villages in semi-arid China. These are small-scale, underground rainwater harvesting reservoirs, in which the runoff catchment is typically the roof and yard. In most traditional villages in semi-arid China (including in Shanxi and Shaanxi) domestic water consumption depends, completely or partially, on these cellars.
However, most studies have focused on modern, concrete water cellars. Since the late 20th century, the Chinese government has been promoting these cellars as relevant water supply projects -the project operated by the China Women's Development Foundation being the most famous -in regions such as Gansu (Zhu 2003), where drinking water supply continues to be a critical issue. Consequently, numerous studies have evaluated the projects in terms of community building (Chen 2003), institutional improvement (Song and Liu 2006), and poverty alleviation effects (Lin 2018). Conversely, the systematic conservation of traditional Chinese villages from 2012 onwards, has created a gap in research related to water cellars, which date back to the Ming and Qing dynasties in China (Hu 2007). Therefore, awareness of water cellar construction techniques is currently limited to modern design (Hou, Hu, and Chen 2011;Niu 2004;Zhang and Chen 1997).
Historians investigated water-use customs related to water cellars built with conventional construction technologies by interpreting ancient inscriptions (Hu 2007). Recently, many studies have examined this unique system from several scientific perspectives. Xu and Lei (2017) revealed the impressive flood-control functions of water cellars -traditional wisdom from the perspective of settlements -using modern methods to calculate their rainwater storage capacity. Another study on the value of traditional water cellars stressed that flood-control was not one of the primary intentions of the builders . Moreover, as most of these remote settlements are devoid of piped water supply, by presupposing piped water intervention, Zhou, Matsumoto, and Sawaki (2021) used a questionnaire to quantify residents' willingness regarding the sustainable use of water cellars and some renovation options of water cellars. Still, research related to their construction and rainwater usage patterns is scanty. As opposed to water cellars, traditional wisdom on larger rainwater harvesting systems has already been quantified. Bandara et al. (2010) pointed to how scientific awareness of traditional village tank cascade systems can provide implications for their plans and technical design. In some pivotal publications regarding the traditional wisdom of qanats (e.g., Ebrahimi et al. 2021;Radaei et al. 2020;Taghavi-Jeloudar et al. 2013), the environment and an ecological approach to macro water management is the main focus. However, unlike these large systems, relatively small water cellars, built and used by private individuals, require evidence-based investigation of the heritage users and artisans to reveal the traditional wisdom behind their construction and use.
Over time, residents and artisans have contributed several beneficial features -using traditional wisdomin the construction of these cellars: indigenous construction formulas, advantageous location, suitable raw materials, and flexible water usage based on rainfall variation. Each of these features is backed by sophisticated economic, environmental, and social trade-offs. Additionally, water cellars are undergoing some modern changes (e.g., use of modern materials and new technologies to build the impermeable layers), but their impact remains unknown. Thus, this study intends to develop an evidence-based awareness of the traditional wisdom of water cellars, in terms of construction and rainwater use patterns, and employ a quantitative and qualitative approach that focuses on the users and artisans of water cellars in traditional Chinese villages, in order to (1) interpret the construction formula, location, and raw materials, as well as the trade-offs behind differences in individual water cellars; (2) reveal how residents adjust the amount and distribution of rainwater usage to deal with variable rainfall; and (3) clarify the potential impact of modernization on the construction of water cellars.

Method
This study analyzes the case of Waling village near Yangquan City, Shanxi, China (Figure 1), where water cellars are the sole domestic drinking water source in each household. According to the community leader, 260 permanent residents and 80 households live all year round in the village, which covers an area of 46.5 ha. Water for agriculture is mostly sourced from a giant waterlogging pool, adjacent to the rice terraces in the southern part of the village. Thus, water storage for domestic and agricultural usages, albeit originating from rainfall, are separated. Eight water wells, dug over time, failed to extract any water ( Figure 1). Considering that the village has not experienced any relocation, nor been connected to piped water, it can be inferred that traditional techniques of water cellar construction have been passed down for generations -according to local, unpublished records of Waling village, these techniques date back to the Ming Dynasty. Additionally, this village could not be connected to piped water supply, because of its isolated location in the mountains, which allowed us to observe the authenticity of water cellar usage. Moreover, the village has been designated as a Chinese National Traditional Village (2013) and a Chinese Historical and Cultural Village (2019), primarily because it is an example of the complementary relationship between spatial layout and nature; it houses many historical buildings, which inhibits the anthropogenic destruction of the historic environment and maintains its socio-cultural sustainability.
From November 15 to 19, 2020, a questionnaire survey was conducted with the residents to understand the construction and rainwater usage patterns of water cellars for prolonged daily use ( Table 1). The questionnaire included demographic questions concerning age, household size, etc. Table 2 shows that currently, the village is home to an aging population (91.1% of the residents are aged 65-84 years). A total of 56 households were investigated, accounting for approximately 70% of the total number of households. The village's conservation and development planning was completed in 2014, which provided essential material to understand its water-use history and culture. Additionally, in-depth interviews with 15 households and two artisans were conducted on 20 November 2020 to reveal the water cellar construction and usage features. Notably, all data in the Tables 2-3 and Figures 4-17 in this study are based on the questionnaire. Using this integrated survey method, we can explain how such a rainwater harvesting system is implemented through a collaborative effort of residents and artisans, as well as highlight the flexible rainwater usage based on climate change.

Construction formula
The construction formula is outlined in Figure 2, and typical water cellars are displayed in Figure 3. Based on interviews with the artisans, the construction procedure entails three sequential steps: the earthwork for building the storage body (approximately 20 days), the impermeable layer (around 15 days), and appendages (around 7 days), including cover, platform, sedimentation tank, and water inlet pipe parts for rainwater collection.
First, the storage body relies entirely on the weightbearing properties of the soil. As soil is naturally dense, cohesive, and has poor permeability, it provides structural stability and is also applied to a well-known local housing structure: cave dwellings. The cellar can be described as a "Thin vase with a large belly and a narrow neck." The depth of the storage body is about 6-8 m, with a diameter of around 3 m. The small opening (approximately 50 cm) of the water cellar prevents debris and people from falling into its interior. The small size also makes the stone cover lighter in weight, which is convenient for daily water retrieval. According to the artisans, this vase shape contributes to the stability of the structure.
Moreover, to make a curvature for maximum structural force, any cross-section of the storage body ought to be circular, and these circles should be concentric to avoid collapse ( Figure 2). The laborious earthwork sequence does not involve any artisan to save labor costs. However, to complete the earthwork, 43 households (76.8%) with their relatives, neighbors, and friends provided labor free of charge ( Figure 4). As a token of appreciation, the owner provided the participants with three free meals. Thus, albeit painstaking, earthwork can be festive and collaborative, which reflects rural interpersonal relationships. Second, the impermeable layer -the part that determines a cellar's longevity -is made from a mixture of lime and local red clay in a 3:7 ratio. The local red clay is poorly permeable and differs from the average soil, which makes it a naturally impermeable raw material. The artisans use this mixture to make large sticks, which are sharp at both ends, and nail them all over the bottom and the wet body of water cellars ( Figure 2). Then, they hammer these sticks tightly against the inner wall of water cellars -a step which is usually repeated three times. The thickness of this impermeable layer is approximately 2 cm. This step is equally laborious -not for the residents, but for the artisans. This layer is hammered until it sets against the inner wall without air bubbles, which is a measure of the artisans' skills. It is only when this layer is damaged that owners build a new, relocated cellar. The damaged one is filled, as the broken parts cannot be identified for repair or patching, nor do seepage hazards ever disappear. The preservation of this layer entails two vital precautions: in order to protect this layer, water level must be kept below the wet body when collecting rainwater (Figure 2), and rainwater inside the cellar must not be exhausted to avoid breakage due to over-drying.
In the 1970s (Figure 5), the concrete water cellarwithout any usage precautions and with a longer lifespan -was developed. However, the traditional method continues to be used because the red clay is believed to possess a natural pollutant adsorbing quality. According to a comparative study by Liu et al. (2012), this effect exists even in the concrete cellar, but to a lesser degree.
Third, the design of the appendages is more flexible, allowing for smooth entrance of rainwater into the storage body. The artisans only need to ensure a certain angle of slope for the inlet pipe. The sedimentation tank is generally a square design with a brick pavement interior and is of a free size. It serves to concentrate rainwater from the yard/road, enabling it to pass through the inlet pipe more easily, and plays the role of settling the debris present in the rainwater. When it is not collecting rainwater, it is usually kept covered. Some water cellars are not designed with sedimentation tanks, which does not affect the rainwater harvesting of the cellar. The platforms are generally built 20 cm above the ground, to prevent rainwater from seeping through water cellar mouths. Occasionally, the platforms will be built close to the ground level of the yard because some owners do not like the raised platforms interfering with their daily activities.

Location attributes and associated trade-offs
There are two contrasting positions for water cellars: yards (82.1%) and public spaces (17.9%) (Figure 6), which are divided into four categories (in the middle of yards: 28.6% vs. close to the door: 23.2% vs. close to the kitchen: 17.9% vs. close to the main house: 12.5%). The public space here mainly refers to vacant space adjacent to the road in the village. Most of the cellars are built in yards to reduce the water intake time and ensure the safety of cellars from sabotage or unintentional damage. Notably, opening the sedimentation tank or removing the covers without the owner's permission is prohibited, which makes public spaces a suitable location. The covers of some cellars in public spaces are locked or weighed down with heavy objects if their owners foresee a prolonged absence.
In terms of the public-space location, the reason, "want free use of the yard," accounted for 100% of the respondents. Additionally, 70% of the respondents thought that public spaces offered greater access to rainwater than yards (Figure 7). Conversely, there were more reasons for choosing yards, because of the strong correlation between water usage and daily activities. First, "in the middle of yards" is highly associated with "visual comfort" (62.5%) and "personal preference" (62.5%). Second, regarding "close to the kitchen," considering that water is most frequently used in the kitchen, "shorten water intake distance" has the highest score (90.0%). Notably, "visual comfort" and "personal preference" played a sizeable part, with 60% score for each. Third, 85.7% of the respondents appreciated "close to the main house," "avoid intersecting with daily action paths," and "visual comfort." Finally, with respect to "close to the door," it is more dictated by "water intake in the road" (61.5%), which can shorten the length of the water inlet pipes. Additionally, as the kitchen is usually located next to the door, 69.2% appreciated "shorten water intake distance".
According to the interviews with artisans and residents, the use of concrete for some impermeable layers did not influence the placement of the cellars. This was authenticated by the crossanalysis of the two kinds of impermeable layers and the location of water cellars, as shown in Figure 8, where there were no significant differences between the locations under the two materials.

Raw material attributes and associated trade-offs of choice
The cover, platform, and sedimentation tank parts of water cellars use diverse raw materials. The raw materials were divided into two categories: traditional materials, including wood, stone, and brick, and modern materials, including iron, concrete, and plastic. Additionally, the overlapping cover and platform had the greatest impact on the landscape. The raw material cross-analysis is illustrated in Figure 9. Only 15.1% of the   cellars used traditional raw materials for cover and platform. The mixture of modern and traditional materials accounted for 51.0%, and cellars using modern materials for both cover and platform accounted for 34.0%.
The statistical results for the types of materials used in building the three parts of the cellars are summarized in Figure 10. First, the cover has the most diverse raw materials, containing modified materials such as iron and plastic; however, the other two have more homogeneous materials. Second, traditional materials were used the most for constructing the sedimentation tank (70.0%), followed by the cover (48.2%) and the platform (34.0%). Figure 11 shows how the material selection for these three parts is characterized by years of construction. Regarding modern materials, the adoption   of plastic covers is more recent -the 2000s. In terms of traditional materials, wood has been consistently used for the cover, and stones have been used since the 2000s; similarly, stones have been consistently used for the platform and sedimentation tank. Additionally, only concrete usage for building the cover experienced a marked increase, from the 1970s to the 2000s. Figure 12 examines whether the modernization of the underground impermeable layer interferes with the visible raw materials above the ground. In terms of the cover, only concrete corresponds to a slightly higher percentage of the concrete impermeable layer than the clay layer. Conversely, for the other two parts, the percentage of the clay impermeable layer is higher than or equal to the percentage of concrete in all materials. The results suggest no correlation between above-and belowground modernization. Figure 13 shows whether the location of a cellar influences the material modernization of the three parts. The platform alone reflects the difference in the yard location, where the use of modern materials is higher than traditional ones. Figure 14 elaborates on the landscape formation of the three parts. Overall, of the six influencing factors, in terms of the platform and sedimentation tank, "imitation of others" and "harmony with surrounding paving" have no apparent correlation with the formation. However, for the cover, "imitation of others" shows a correlation with wood (33.3%) and iron (37.5%); "harmony with surrounding paving" shows a correlation with stone (41.7%).
The remaining four factors were significantly correlated. First, regarding the cover, except for plastic cover (n = 1), iron obtained the highest percentage for "beautification," "better function," and "economical." This is attributed to the novelty of using iron for the water cellar in the settlement, as well as for being fastened to the platform. However, iron scores the lowest in terms of "artisans' advice," which is understandable, because they need to own this material. For the remaining three materials, there is no significant difference in terms of "better function" and "economical." However, the scores for concrete are different from the other two: it scores the lowest in "beautification," despite being the most recommended by artisans. This implies that the concrete cover is not in line     with the aesthetic preferences of the residents. Regarding the platform, except for the rarely used bricks (n = 2), stone -a native material -scores lower than concrete for all the factors, except "beautification" (56.3%) and "harmony with surrounding paving" (18.8%). This is because artisans prefer concrete over stones, which require effort to select, carry, polish, and turn the holes for the cover. Third, for the sedimentation tank, stone is not preferred by the artisans due to the extra effort required to shape it, even though it is the most economical material. Additionally, there was no major difference between the three materials. Finally, focusing on concrete, we observe that the factor of "artisans' advice" has the highest percentage in the three parts, compared to the other materials.

Rainwater usage patterns
Rainwater harvesting, naturally, is limited to the threemonth long rainy season from late-June till early-September, which accounts for nearly 70% of the year's rainfall. Thus, a water cellar should be designed to fulfill the annual water usage requirements of a family. The amount of rainwater usage necessitates separate discussions due to the impact of climate change, which has led to variations in the volume of rainfall.

Rainwater usage in the absence of drought
In order to ascertain the volume of water used, the respondents' water intake containers for the kitchen, water cellar, and flush toilet were measured and the volumes were determined; thereafter, they were invited to use these containers to recreate the water extraction scenarios of various daily water activities and the corresponding values were recorded. The volume of rainwater used for different household activities per person per day is summarized in Table 3. Figure 15 indicates that the frequency of bathing is reduced sequentially from summer to winter. Thus, water used for bathing, as shown in Table 3, refers to usage during summer, because it may vary greatly during winter. Figure 16 shows the water usage rates for household appliances, where washing machines have no impact, whereas flush toilets significantly affect rainwater usage; thus, water usage for toilet is excluded from total rainwater usage. The average water-use was 28.5 L per person per day (standard deviation [SD] = 1.7). The smallest gap between the respondents was for drinking (SD = 0.2), personal hygiene (SD = 0.4), and bathing (SD = 0.7). Assuming a normal parent-child family consisting of three adults, the minimum storage body capacity of the water cellar in this village will Figure 14. Trade-offs behind the raw material choice of cover, platform, and sedimentation tank *p < 0.05 (More than one option can be selected).
be 31.2 m 3 ; similarly, an extended family with six adults should have a minimum storage body capacity of 62.4 m 3 .

Rainwater usage during droughts
According to interviews with residents, in the event of a drought, rainwater in water cellars needs to provide sufficient support for two years (the maximum duration of the collected rainwater's use) rather than one year; the biggest lifestyle change in such a situation is bathing merging with personal hygiene and minimum laundry. The average wateruse is 28.5 L per person per day; assuming the same amount of rainfall is supporting water usage for 2 years, the average amount should be split to 14.3 L per person per day. Meanwhile, if the rainwater consumption for bathing and laundry in Table 3 is removed, the total sum will be approximately 14.2 L per person per day. The similarity between the two computational results confirms the validity of the residents' judgments regarding the regulation of rainwater usage.
At times, prolonged drought conditions could occur, such as during the late-1990s, when they had to resort to other highly exhausting measures, as shown in Figure 17. The most recent severe drought occurred in 2009, when the village committee, together with the township government, financed water trucks to fill household water cellars. With such a local policy, the residents no longer needed to worry about prolonged droughts.

Discussion
(1) Advantages of water cellars include low construction cost (all construction materials are derived from the natural environment), accessible technology for residents (users can determine the location and storage body capacity), and relatively stabilized daily water usage, except in the case of a severe drought. However, there are two labor-intensive construction steps: the earthwork performed by the owner and the impermeable layer construction, which is assigned to the artisans. The Figure 15. Frequency of bathing in all seasons (n = 56). Note: (1) Drinking is sourced from the water cellar, and the residents cannot afford bottled water.
(2) "Personal hygiene," consisting of washing the face and feet and brushing teeth, is a daily activity, and separate from "bathing." (3) The term "bathing" here signifies rinsing the body, not including the face and feet, with a wet cloth rather than bathing in a bucket, which is water-intensive. (4) Cleaning houses and yards are both waterless activities, consisting of only sweeping. (5) The planting rate of vegetable plots in yards is much lower because the main vegetable plots with large areas are joined with the farmland. They do not use rainwater in water cellars but reuse the water used in the kitchen and for personal hygiene to water vegetables in yards.
residents tackle the strain of earthwork through cooperation. Additionally, in 1970, the concrete impermeable layer was introduced by the artisans -it is quicker and easier to build and functions relatively better. However, this has also led to the gradual retirement of the traditional technique of creating clay impermeable layers, which constitute the most distinctive indigenous feature of water cellars. As clay impermeable layers offer little functional advantage over modern concrete ones, encouraging the use of this labor-intensive technique is illogical. In order to prevent the extinction of this traditional technique, it is better to identify and inherit it from the perspective of intangible heritage.
(2) The most common location of the water cellar is in the yards, followed by public spaces. The location differs among yards, which is influenced by multiple trade-offs, such as rainwater stream flow, visual comfort, and compatibility with daily tasks. As determined by these tradeoffs, the location varies from the middle of yards, to close to the door, close to the kitchen, and finally, close to the main house. However, it is not sufficient to examine the trade-offs solely from the perspectives of the residents; further research should involve the influence of physical aspects, such as the layout and size of the yard and the profile of the ground. (3) The trade-offs that determine the raw material used for constructing the parts of the cellar -the cover, platform, and sedimentation tank -are highly interrelated. Despite the proliferation of modern materials, traditional materials have been consistently used for these parts. The combination of cover and platform is mostly made of a mixture of traditional and modern raw materials. Concrete is mostly used for the platform, accounting for 66% of its total usage. Currently, the cover does not use a lot of concrete (35.7%), but this has been consistently changing for the past few years because the artisans prefer concrete -even though it is contrary to the preference of the residents. Residents prefer iron, wood, and stone for the cover and stone for the platform; however, the Figure 16. Ownership rate of water-consuming appliances (n = 55) (More than one option can be selected) According to the survey, the laundry function of washing machines is rarely used; tumble dry function is used relatively more during winter; flush toilets are not installed with automatic flushing but still consume a lot of rainwater. Figure 17. Measures during prolonged drought conditions (n = 56) (More than one option can be selected).
artisans tend to use concrete rather than stone for both parts, as polishing stones is laborintensive. Thus, to avoid ruining the historic environment with higher usage of modern raw materials, the differences in material use tendencies between residents and artisans should be addressed. Residents should assert their aesthetic preferences to the artisans, and a corresponding increase in labor cost ought to be considered. As residents are generally reluctant to bear any additional costs, their greater participation in construction is recommended, such as preparing traditional materials and polishing the raw materials. (4) Material modernization for building the parts below and above the ground emerged simultaneously during the 1970s. However, the modernization of underground impermeable layers has no correlation with the location and raw materials used. Despite these modernizations, a concrete storage body of water cellars was not found.
(5) The regulation of rainwater consumption is efficient. When not encountering a drought, residents change the form of bathing to body rubbing, so that the cellars contain the required volume for year-round supply. With an average water-use of 28.5 L per person per day and severely reduced rainwater availability during a drought, they further change their bathing into personal cleaning and rarely do their laundry, so that the average daily per person water-use is controlled to 14 L -about half the usual amount. Such regulations coincide with their empirical judgment that the average annual rainwater use during droughts should be half the volume used during non-drought periods. Faced with frequent rainfall variations, they must continuously adjust their water consumption patterns to save water. (6) This study implies that the sustainability of water cellars in traditional villages should include two aspects -cultural significance and landscape character. First, the central element of cultural significance is the village's shared vernacular construction techniques and rituals. Second, the landscape character includes the spatial features and individual landscape aspects of the water cellar. However, although this paper has warned of the dangers of modernization regarding both aspects, we are more concerned that the role of the water cellar will be marginalized in the future with the introduction of piped water. Therefore, researchers need to explore ways in which water cellars and piped water can coexist. (7) Finally, this study has shown that water cellars can secure considerable amounts of water at low construction and maintenance costs. Thus, in the context of SDGs Goal 6 seeking to ensure access to water and sanitation for all, indigenous rainwater harvesting techniques such as water cellars can be improved and promoted in waterscarce areas where water solutions are urgently needed.

Conclusion
Using a quantitative and qualitative approach, this paper contributes to the evidence-based awareness of water cellars constructed via conventional technologies. This study highlights the concern for the thoughts and interests of heritage users and artisans. The traditional wisdom behind water cellars has both significant advantages and vulnerabilities. The key advantages of this system include a low construction cost, user-friendly technology, and water usage stabilization features. However, some degree of modernization must be promoted, to compensate for the traditional deficiencies of water cellars without compromising the expression of traditional features. Moreover, the empirically corroborated traditional wisdom behind changing the water consumption patterns cannot free residents from the discomfort of extreme water-saving. Therefore, evidence-based awareness of traditional wisdom is essential to understand the drawbacks of traditional creation and use.