Identification of suitable sites for traditional pokhari water harvesting in mountain rural communities of the Himalaya

For centuries, farmers have adapted to climate variability by implementing various kinds of practices including soil and water conservation techniques and water harvesting (WH) (Al-Adamat et al. 2012; Adhikari 2018). This latter includes a wide range of techniques aimed at storing runoff water in the soil or dedicated structures to use it subsequently for productive purposes (Adhikari et al. 2018; Subedi et al. 2020). WH has been applied in many areas of the world through low-cost and eco-friendly interventions, developed from the indigenous knowledge of local people (Lal & Verma 2008; Chen et al. 2019). In mountainous areas, where water availability is a major concern, WH allows extending the cropping season and increasing food security by providing water while conventional water sources are drying out (Subedi et al. 2020). Moreover, it reduces soil erosion and flood risk and can increase water infiltration in soil and groundwater downstream (Lal & Verma 2008; Dile et al. 2016). WH has been proven to considerably increase crop yields, soil moisture, and farmers’ income; and to reduce soil erosion in various mountainous zones of the world, among which the Himalayan region, including Nepal (Bastakoti et al. 2016; Adhikari 2018; Adhikari et al. 2018; Subedi et al. 2020).

Nepal is a landlocked country in the Himalayan region (UNDP 2020). Moving from North to South, the Nepalese climate ranges from an extremely cold subarctic climate to a monsoon-influenced tropical one, covering the Köppen and Geiger classification codes Dfd, Dwc, Dwa, Cwb, and Cwa. Despite being perceived as a water-rich country, with a mean annual rainfall of 1,344 mm, some areas of Nepal suffer from water stress during the dry season, especially from March to June (Adhikari 2018). The reason is the unbalanced rainfall distribution throughout the year with about 80% of total annual precipitation limited to the monsoon season between June and September (Adhikari et al. 2018). Moreover, climate change is altering rainfall patterns and decreasing total precipitation (Adhikari 2018; Subedi et al. 2020). Climate change consequences can be particularly severe in Nepal, where around 80% of the population depends on subsistence agriculture, especially in hilly areas, and more than 50% of agricultural land is rainfed (Subedi et al. 2020). Nepalese farmers report an increased frequency of droughts, late monsoon onset, rising temperatures, reduction of water availability, and flooding during monsoon as major impacts on their livelihoods (Bastakoti et al. 2016). If monsoon onset is delayed, farmers need to irrigate their crops to avoid severe crop losses (Bastakoti et al. 2016; Pandey 2019).

Recently, pokharis are being abandoned due to the development of irrigation facilities in some areas, while many others are degrading because of abandonment of farmlands (Chen et al. 2017, 2019). Underutilization and poor management of these traditional structures reduces not only water availability in dry months but also ecosystem services provided by pokharis like downhill flood risk reduction and sediment control (Lal & Verma 2008; Bastakoti et al. 2016; Chen et al. 2019).

Various authors (Lal & Verma 2008; Bastakoti et al. 2016; Chen et al. 2017; Adhikari 2018), also reporting farmers’ opinions, point out that restoring existing pokharis and building new ones could foster resilience to climate change through increased water availability. Usually, the construction sites of new pokharis, and ponds in general, are selected ‘arbitrarily’, considering only the proximity to the point where water is needed (Chen et al. 2017).

However, a more comprehensive and systematic site selection can lead to more efficient WH structures, which in turn contribute to improving water availability and agricultural productivity (Adham et al. 2016; Mugo & Odera 2019). Nevertheless, site selection is a very challenging task especially at large spatial scales, owing to the lack of detailed hydrological and soil data (Singhai et al. 2019).

In the last three decades, many studies have attempted to identify the most suitable sites for different types of WH structures, often using geographical information system (GIS) and multi-criteria decision-making (MCDM) (Adham et al. 2016). GIS is widely used because it allows us to analyze large areas by integrating spatial physical and socio-economic information in a timely and cost-effective way (Adham et al. 2016). On the other hand, MCDM allows us to compare and relate different kinds of data producing a single final output displaying various options (Krois & Schulte 2014).

The literature on the topic reports the need to select some criteria for WH structures' best siting, such as environmental and physical characteristics of the area of interest, and its socio-economic features (Adham et al. 2016; Grum et al. 2016). Common physical criteria considered for best-siting analysis are slope, soil type, runoff, and land use (Al-Adamat et al. 2012; Dile et al. 2016). On the other hand, the main socio-economic data are the distance from roads, villages, and fields (de Winnaar et al. 2007). In several cases, these latter criteria and participatory activities to assess stakeholders’ preferences are neglected despite being necessary to ensure the proper implementation of site-specific effective measures (Adham et al. 2016; Grum et al. 2016).

In this work, we use MCDM and GIS to produce, to our knowledge, the first best-siting analysis for pokharis. While considering similar MCDM parameters as other conventional best-siting studies for WH structures, we include innovative parameters, such as upslope contributing area and elevation.

Finally, we complement the geophysical analysis with community-driven criteria, including indigenous knowledge, to integrate and validate our analysis.

The study area, which is around 1,186 km² large, ranges between 28°12′51″N, 84°4′26″E and 28°10′37″N, 84°11′53″E. The elevation ranges from 407 m to 1,443 m a.s.l. South-facing slopes are usually terraced for cultivation, while North-facing slopes of forest and shrublands do not receive enough solar radiation to allow for cropping.

Our analysis focuses on four communities within the study area: Thulakot, Majhkot, Begnas, and Hansapur (hereon ‘locations of interest’), considered representative of the local socio-economical setting. The four locations of interest are also characterized by their position on the upslope parts of the watersheds in the hilly mountainous area, located in the northern surroundings of Pokhara city (Figure 2). These locations (i.e. upslope and south-facing) have been selected since they are the most prone to water scarcity. The total area of the location of interest is 12.11 km².

All the datasets used in this work – i.e. soil type, European Space Agency (ESA) Worldcover Land Use/Land Cover (LULC) map and a Digital Elevation Model (DEM) – were downloaded from websites of different public agencies. Flow accumulation (also said upslope contributing area), slope, and elevation maps were derived from the DEM through QGIS software. The lakes and rivers spatial data retrieved by Open Street Map were provided by the International Rainwater Harvesting Alliance (IRHA), a Swiss NGO carrying out a project in this area with its Nepalese local partner NGO Kanchan Nepal. We downloaded a gridded soil map from the Nepalese National Soil Science Research Center (NARC) website. The soil data were developed with machine learning and predict soil conditions up to 2 m. Kanchan Nepal provided some GPS locations of existing pokharis in the study area as well, while some other locations were gathered during a field visit in the same zone. Data on terraced and private areas, and traditional criteria considered in pokhari building, were collected during the field visit. This work was carried out in the framework of IRHA's project. Table 1 reports the sources from which we accessed the data.

Spatial MCDM procedures, also known as the weighted overlay process (WOP), are used to combine different thematic maps through a few simple steps to obtain a single final suitability map (Krois & Schulte 2014; Adham et al. 2016). A suitability map allows us to visualize multiple options ranked based on preferences expressed through the categorization of various criteria, weighted depending on the influence they have on the overall objective (Al-Adamat et al. 2012). The methodological steps of MCDM can be summarized as follows (Grum et al. 2016):

• 1.

  Selection of relevant suitability criteria for achieving the objective;

• 2.

  Establishment of suitability ranges for each criterion (usually on a scale from 1 to 9 where 9 represents the optimal value) and normalization of each criterion on a 1–9 scale;

• 3.

  Assignment of a weight to each criterion according to the influence they have on the result – potentially according to experts or local people's opinions. Regardless of their distribution, the sum of all weights must be 1.

• 4.

  Combination of suitability values of all weighted criteria to produce the final suitability map.

We reviewed various best-siting works carried out using MCDM and GIS focusing on ponds in different geographic areas and selected the most relevant criteria for the objective of the analysis (de Winnaar et al. 2007; Al-Adamat et al. 2012; Dile et al. 2016; Grum et al. 2016). As summarized in the review paper of Adham et al. (2016), the most common criteria are slope, soil type, and land use. Other common geophysical criteria are rainfall and runoff, while the main socio-economic criteria are the distance from roads, villages, streams, and fields. The slope affects the quantity of rainfall that infiltrates into the soil or runs off, the distribution of runoff, and the quantity of earthwork needed to build WH reservoirs (Ramakrishnan et al. 2009; Dile et al. 2016). Moreover, a small slope is needed to convey water towards storage sites, thus gentle slopes are preferred for WH implementation with respect to flat areas (Ward et al. 2021). The soil type affects water infiltration and storage in the soil (Mugo & Odera 2019), and consequently, the runoff (Singhai et al. 2019). Soil type suitability varies with the type of WH structure. For instance, sandy soils are better suited for recharge structures, while clayey soils are preferred for water storage structures, such as ponds (Al-Adamat et al. 2010; Adham et al. 2016). LULC informs on land surface characteristics (Ward et al. 2021). These latter control the quantity of runoff produced in an area, and water infiltration in soil (de Winnaar et al. 2007). It can be easily extracted by remote sensing data (Adham et al. 2016). Rainfall quantity is usually considered in WH best siting to locate areas where more runoff is generated, as it can be used to recharge the structures (Chen et al. 2017).

Concerning socio-economic criteria, distance from houses and cropped fields are considered because harvested water is mainly required for domestic and agricultural use, thus structures should be as close as possible to where water is needed (de Winnaar et al. 2007; Chen et al. 2017). The same occurs for roads, from which WH structures can be easily accessed. Distance from streams is sometimes considered since part of streams’ discharge can be used to recharge WH structures (Al-Adamat et al. 2010).

Following the literature analysis, the selected parameters considered in this analysis are slope, LULC, soil type, upslope contributing area, elevation, and distance from rivers and lakes. In our case, we decided to introduce an upslope contributing area as a proxy for the presence of minimum water flow (which can feed pokharis) instead of estimating runoff. The upslope contributing area represents the area draining water to a downstream point, and it is measured in km². It is higher in riverbeds and lower in upslope zones (Chen et al. 2017). We did not include a rainfall layer because, due to monsoons, our study area receives enough rainwater, thus prioritizing areas with higher rainfall is unnecessary. Distance from fields and settlements was considered at a later stage of the final analysis, to include the outcomes of the field visit, as these criteria were indicated directly by local people.

Elevation and distance from rivers and lakes were both used to prioritize areas more prone to water scarcity, which are those more uphill and further away from rivers and lakes (which contribute to groundwater recharge and irrigation).

The value ranges/categories of each layer were given a suitability score spanning from 1 (unsuitable values/categories) to 9 (optimal values/categories – suitability scores are referred to as x₁ in Equation (1)). Regarding the suitability score given to each value range of each criterion, we assigned them as follows: According to the literature, optimal values for slope for ponds best siting range from 0–3% to 0–5% (Ramakrishnan et al. 2009; Al-Adamat et al. 2010). On this basis, we assigned the highest suitability score to the 0–3.5% slope range. To other ranges of increasing slopes are assigned decreasing suitability values as summarized in Table 2. Regarding the upslope contributing area, we assigned the highest suitability score (9) to contributing areas of 0.09–0.18 km², considering that the quantity of runoff should be enough to recharge pokharis but not too high to prevent structural damages. This range was selected by adapting to our study area the upslope contributing area values proposed by Márquez et al. (2021). Consequently, upslope contributing areas higher than 0.18 km² were given lower suitability, while contributing areas smaller than 0.09 km² were given an intermediate suitability score. In the case of farm ponds, usually, the highest suitability score for land use is given to agricultural land if water is used for irrigation (Dile et al. 2016), while when ponds are used mainly for herding, also shrubland, bareland, and grassland areas can be marked as optimally suitable (Grum et al. 2016; Singhai et al. 2019). Another reason for which we did not assign the highest suitability score to cultivated land is that in our study area, cropped fields are terraced, thus it is not advisable to build ponds there in order to cultivate all the available surface. Moreover, in our locations of interest shrublands, barelands and grasslands are usually located near cultivations, thus the collected water can be easily accessed for supplementary irrigation as well. Furthermore, the land use map used in this work marked as ‘forest’ all the areas with a tree cover >10%, thus, not having information on the actual presence of open forest (which could have been very suitable for ponds) and closed forest (not suitable as there would not be enough space for a pond), we decided to assign it an intermediate suitability score. According to the literature (Al-Adamat et al. 2012; Adham et al. 2016; Ward et al. 2021), less-draining soils such as clay, silty loam, and loamy soils are ideal for ponds construction. We gave both silt and silty loam soil a suitability value of 9. The highest elevation and distances from water bodies were assigned maximum suitability in order to prioritize more water-scarce areas which cannot benefit from this surface water and groundwater recharge. Hence, zones closer to water bodies were given lower suitability. Table 2 summarizes suitability values assigned to each criterion category and sources used to set the suitability ranges.

Following the MCDM steps, each layer was given a weight based on the relative influence on the resulting map. Weight assignment is a complicated task that might produce misleading results (Al-Adamat et al. 2012). Weights of criteria for MCDM are computed mainly through the Analytical Hierarchy Process (AHP) method which allows us to assign weights by comparing the selected criteria in a pairwise matrix and inspecting the relative importance of each layer (Grum et al. 2016; Ward et al. 2021). The pairs of criteria are set according to literature and experts’ opinions (Krois & Schulte 2014; Ward et al. 2021) or through participatory workshops with stakeholders and NGO consultants (Grum et al. 2016). However, AHP is highly influenced by expertise (Adham et al. 2016). Sometimes, weights are assigned directly on the basis of literature or opinions, without applying the pairwise matrix (Al-Adamat et al. 2010; Krois & Schulte 2014) or equal weights are assigned to all the criteria to avoid biases (de Winnaar et al. 2007; Al-Adamat et al. 2012). Owing to these reasons, we decided to assign equal weights to all the layers except for the distance from water bodies, which received a lower weight. Distance from rivers and lakes serves to prioritize areas that cannot likely count on groundwater recharge or surface water for agriculture. However, stream water cannot always be easily accessed in hilly areas, so this criterion has a lower influence on the final selection of best sites. We considered this weight setting since it is the most conservative and balanced.

We performed a sensitivity analysis to assess the variability of the results at varying weights of the best-siting criteria, as a measure of uncertainty following Dile et al. (2016). We first assigned equal weights to all the criteria (variation 1). We then varied the weights of each criterion using plausible strategies. In variation 2, the highest weight was given to slope, which is usually assigned a higher weight (Al-Adamat et al. 2010; Singhai et al. 2019). Usually, the soil is weighted slightly higher than land use (Mbilinyi et al. 2007; Krois & Schulte 2014), thus we also tested soil type influence on the result in variation 3. The fourth variation was set to give land use the highest weight, while in variation 5, we tested the influence of upslope contributing area on the result by attributing the highest weight. Table 3 summarizes the tested weight distributions.

We conducted a field visit in the study area between the 16th and the 30th of October 2021 with three specific objectives: (a) surveying additional GPS locations of existing traditional ponds beyond the 13 ones provided by Kanchan Nepal to validate the preliminary best-siting results (pokhari GPS survey); (b) collecting information on land ownership and terraces presence; (c) gathering information on traditional criteria for selecting suitable pokharis’ locations.

Concerning the validation of the preliminary results, we checked in which suitability quartile the existing pokharis are located. Those found within the high suitability areas obtained with the preliminary GIS siting were considered successful. Field inspection of suitable areas was also carried out to check for ‘false positives’, i.e. areas with theoretical high suitability, but with local constraints hindering the construction of pokharis (e.g. presence of agricultural terraces or private land ownership).

The features resulting as ‘false positives’ were then removed from high suitability areas in the final map. Finally, we complemented our set of siting criteria with the indigenous knowledge (additional socio-economic criteria) to account for potentially locally relevant factors that are usually neglected in best-siting studies. Traditional criteria to locate new ponds were surveyed through two focus group discussions with two different groups of local pokharis building experts (i.e. four people in Hansapur and seven people in Majhkot communities). They were all elder men who accompanied the authors to visit existing pokharis nearby their villages together with Kanchan Nepal members. These latter had already gained local people's trust during years of collaboration in development projects carried out in the same villages. During traditional ponds inspections, the NGO members also facilitated our interviews by translating from Nepalese language to English. Villagers were asked the following simple questions.

• How do you select where to build new pokharis? What factors do you consider?

• Which are the features characterizing these ponds?

The preliminary best-siting map is shown in Figure 4. Unsurprisingly, higher suitability values lie in areas of higher elevation – where people rely on pokharis the most – which were given the highest priority. On the other hand, the lowest suitability values are reported in pixels corresponding to permanent water bodies followed by urban areas (Figure 4).

In variation 1, where the distance from water bodies weight was higher, one pokhari laid in Q3 due to its proximity to a river, while all the other ponds were located in Q4 areas. The same occurred with another pokhari in the initial weight setting (the reason is addressed in the discussion).

Once we verified the robustness of the initial weight assignment, we proceeded with this preliminary map to obtain the final result. Considering the average suitability score of GPS points corresponding to existing pokharis and the standard deviation of all the different weight assignments (Table 4), the preliminary best-siting map can be considered successfully validated. The single suitability score of each point for each weight condition is reported in the Supplementary material.

Kanchan Nepal members, together with local people, identified on the maps prepared for each location of interest some private lands. In the same maps, we also marked the location of terraces. These features, accounting for 3.18 km² (corresponding to 29% of high suitability areas within the four locations of interest in the preliminary map), were marked as unsuitable. It is important to note that the information on land ownership was not fully available for the whole study area.

Furthermore, we gathered information on both pokharis’ characteristics and local knowledge criteria used to build these traditional ponds.

Local experts in pokharis’ building explained that traditional criteria considered when locating new ponds are the following, ranked by importance:

• Accessibility (e.g. proximity to villages and roads or paths) – usually pokharis are built within 2 km from villages and as close as possible to tracks;

• Small flat areas where water stagnates over 4–5 years during monsoons.

Sometimes pokharis are built nearby small water springs which will feed the ponds themselves. Accessibility to pokharis is the most important criterion, as pond water is used mainly for cattle rearing and irrigation, which are activities carried out in areas surrounding the villages. Still for accessibility reasons, pokharis are located also near roads and tracks used by herders to move their cattle. Pokharis, especially in the past had also social functions, as people walking on mountain paths and carrying out daily activities used to rest near those ponds. Water stagnation during monsoons considered by indigenous knowledge supports our choice to include the upslope contributing area as best-siting criterion. We added these criteria to the preliminary bio-geophysical criteria set shown in Table 2. Creating our final socio-bio-geophysical criteria set. The upslope contributing area was already included in the MCDM analysis, we selected only Q4 suitable areas from the preliminary best-siting map and we removed those >2 km far from villages and those falling on private and terraced lands, which are unusable for community pond building, identified during the field visit.

The main result of our study provides a first guidance for the identification and spread of potential sites for pokharis in Nepal. The sensitivity analysis confirms the robustness of our approach, which is also supported by the in-field validation. In fact, compared with other water harvesting best-siting analysis performed in other regions of the world, our results are among the ones with the highest correspondence between modelled sites and existing ones, with a 100% correspondence. The same percentage was obtained by Ramakrishnan et al. (2009) for ponds, while slightly lower accuracy was reported for other structures. Singhai et al. (2019) reported a suitability map accuracy of 90% with 31 surveyed farm ponds; Grum et al. (2016) validated check dams and percolation pond suitability, for these latter, an accuracy of 63% over 92 infiltration ponds is reported. In these studies, the number of existing structures is higher compared with our work, but the study areas are wider than ours. Moreover, the mentioned studies consider various WH structures and not just ponds as in our case. Our methodology also includes two rather uncommon criteria: upslope contributing area and elevation. This latter was introduced to prioritize higher altitude areas which are those more prone to water scarcity, while upslope contributing area was used to find minimum water accumulation potential, and it is consistent with traditional criteria reported by local experts. Moreover, the upslope contributing area can be easily computed from a DEM on GIS softwares compared with runoff estimation, which is more complicated and data-demanding.

Another key result of this study is the validated selection of siting criteria using participatory methods. We combined the benefits of an accurate validation with existing ponds with traditional knowledge to develop an efficient approach easily acceptable by local people and linked with their indigenous expertise. Other studies took into account traditional knowledge. For instance, Al-Adamat et al. (2012) asked local people to point out potential WH best sites on a map and to select some best-siting criteria like distance from roads, borders, villages, farmlands, and wells. Grum et al. (2016) also included a participatory approach to identify the most relevant criteria for local people in selecting the best WH techniques, and then they processed retrieved criteria with MCDM to sort out preferred water harvesting techniques. Finally, Mbilinyi et al. (2005) reported in their study carried out in Tanzania the traditional criteria used to locate new WH structures.

Concerning our comparison with traditional criteria, we added distance from villages as a socio-economic criterion pointed out by local people and upslope contributing area as a proxy for water stagnation points over years. We did not take into account springs, which was an additional indigenous criterion, as our analysis focused on rainwater storage.

Such a straightforward and validated approach for mapping best location of pokharis can immediately support government, NGOs and private efforts to build new distributed water points across Nepal. Increasing water available for farming, depending on the actual number of new structures and their size, can considerably improve the availability of water in the dry season, thus positively affecting both cattle health and agricultural productivity. Cascading impacts of increase water availability with pokharis might likely include increased food security and income stability, thus potentially reducing outmigration from rural areas, which is a pivotal cause of environmental degradation in Nepal (Pandey 2019). Other ecosystem services from pokharis, such as water retention and flood prevention, reduction of land erosion, cultural and social value would increase as well, improving people's wellbeing.

Our results report good correspondence between existing pokharis’ locations and produced best-siting map. Thus, the presented method offers a useful guidance for WH structure best siting despite some limitations mainly due to data availability and quality. Regarding this latter issue, some considerations were made: the LULC layer was, despite being the best option available, unable to report the presence of terraces. These were reported instead as bareland or grassland (the best categories for pond building) leading to many suitable areas for pond construction. This problem was identified during the field visit and suitable areas were reduced as terraces are unsuitable for pokhari building. Terraced fields are usually private, and these limited lands often represent the main source of subsistence for farmers, thus it is not advisable to use those areas for community-owned ponds. Some techniques for terrace detection from remote sensing do exist (Tarolli & Sofia 2016), but the lack of high-resolution topographic data (1–2 m) hindered its estimation. Interestingly, the LULC reports the biggest pokhari as ‘permanent water body’, thanks to its good spatial resolution. Owing to this reason, the just mentioned pokhari is the one resulting in the third suitability quartile (see Supplementary material), because permanent water bodies were given the lowest suitability as they usually represent rivers or lakes. Actually, the water body in this case is the pokhari itself, thus the lower score is a ‘false negative’, i.e. an area resulting in a lower suitability score which, checking on the field, has instead a high suitability. This explains the 100% correspondence between retrieved best sites and existing structures despite one pond laying in the third quartile. The main limitation of the LULC map – for the scope of this work – lies in the broadness of the categories used in the classification. Particularly, pixels with a wood cover exceeding 10% are classified as tree cover even if ‘other land cover classes can be present below the canopy, even with a density higher than trees’ as stated in the ESA Worldcover Product User manual (Zanaga et al. 2021). The same occurs for grassland and shrubland. This affects our analysis as, for instance, open forests, which are very suitable, are not distinguished from closed forests, which are less suitable for WH structures having less free space. Despite being the best option available, the DEM is not ideal for best-siting analysis of relatively small structures, as in this case. A spatial resolution of 30 m cannot report small flat areas surrounded by steeper areas, which would be actually suitable for pokharis, especially in hilly/mountainous regions like Nepal. Also, data about land ownership were not available from institutions either before or during the field visit. However, local people provided information on some land property, which allowed us to exclude at least some private lands. Lastly, a digital map of housing was not available, so we had to remove areas too far from villages from the final best-siting map by visual inspection. The availability of a digital layer of settlements would have allowed the automatized integration of this traditional criteria into the GIS analysis, making this step faster and more accurate. Despite these limitations related to data availability and resolution, our methodology provides a first reliable guidance to large-scale planning of pokharis, with a good level of accuracy granted by the high validation standard and the integration with participatory methods. Particularly useful to overcome the quantitative data limitation was the integration of qualitative information from field inspection and participatory research. As in other studies (Castelli et al. 2018, 2021; Piemontese et al. 2021), the integration of indigenous knowledge was essential to (a) integrate the scarce data, (b) validate the overall approach and criteria, and (c) identify limitation and potential improvement for future work.

With detailed soil and geological data, areas more prone to landslides (which in Nepal are common) could be estimated to remove suitable sites where water infiltration in soil could increase the risk of land collapse. This analysis could be done with the aforementioned high-resolution topography data if they will become available for Nepal too, as well as automated detection of terraces and small flat areas. Moreover, once new pokharis are built in at least some identified suitable spots, a thorough quantification of the benefits provided by these structures in terms of farm productivity and ecosystem services provision could be carried out. Furthermore, the proposed approach can be a guidance for similar works in other data-scarce areas with similar characteristics, to support decision-makers in sustainable land and water management and foster resilience to climate change.

We introduced a methodology for large-scale siting of traditional WH structures, which is consistent with traditional criteria. To our knowledge, it is the first best-siting analysis for Nepalese traditional ponds and one of the first best-siting works for Nepal. The best-siting analysis identified about 0.4 km² of highly suitable land for pokharis’ construction, corresponding to 3.47% of the total location of interest's area covering around 12 km². As our analysis was successfully validated with a 100% agreement with existing pokharis, it can give decision-makers reliable guidance for water resources planning and management. Compared with the traditional siting methods focused on the local scale and individual structures, the combination of GIS and MCDM allows a fast and cost-effective best-siting analysis on wide areas, providing information that can support large-scale national to regional planning. Moreover, software and datasets used are freely available, so this analysis can be replicated by other institution or for other areas. Before building new structures, we recommend a field check, especially for land ownership and accessibility as this information is usually very difficult to retrieve remotely. Finally, the development of higher-resolution or better-quality datasets could improve the accuracy of our best-siting tool and improve the usefulness of our approach. The proposed methodology favours the prompt implementation of water harvesting techniques that, together with other soil and water conservation measures, can increase farming productivity through increased water availability for agriculture and livestock rearing.

This work was carried out in the framework of the project ‘Rain Communities’ carried out by the Swiss NGO International Rainwater Harvesting Alliance (IRHA) with its Nepalese partner NGO Kanchan Nepal. We would like to acknowledge the institutions funding the project, i.e. République et Canton de Genève, Puplinge and 1% fund. We would like to acknowledge also all the NGO members who supported us during the development of this work, especially concerning the field visit and all the villagers who collaborated with us during the participatory analysis.

The authors are thankful to editor and reviewers for evaluating and improving this manuscript.

All relevant data are included in the paper or its Supplementary Information.

The authors declare there is no conflict.