A study on the effectiveness of percolation ponds as a stormwater harvesting alternative for a semi-urban catchment

With growing urban population and urban water demand, stormwater harvesting (SWH) is becoming one of the preferred approaches for stormwater management in urban areas. Urban SWH refers to the practice of collecting, storing/infiltrating and treating runoff from impervious areas such as roof tops, parking lots, roads, etc., for partially meeting water demands of a community (Larm 2000; Rauch et al. 2012; Nnadi et al. 2015; Bonneau et al. 2017; Niaizi et al. 2017). Conventional urban stormwater systems do not have the provision for harvesting stormwater, rather they were mostly built for safely discharging the excess runoff from impervious areas (Burns et al. 2012, 2015). Apart from the hydrological impacts of urbanization on runoff and infiltration processes, the other notable impact of urbanization is the skewness in water availability and water demand. Urbanization results in an increase in population which inherently results in increased demand for water. To meet the twin objectives of runoff volume reduction and partially meeting water demands locally, countries like the USA, Australia and New Zealand have developed technologies for sustainable management of urban stormwater. Such techniques are variously termed as LID (low impact development), SUDS (sustainable urban drainage system), WSUD (water sensitive urban design), and LIUDD (low impact urban design and development) in various countries (Elliott & Trowsdale 2007; Ray et al. 2010; Sitzenfrei et al. 2013; Fletcher et al. 2015; Eckart et al. 2017). Burns et al. (2012), Her et al. (2017), Holman-Dodds et al. (2003) and Tillinghast et al. (2011) performed a study to relate urban development with receiving watercourse attenuation. The results suggested that the increase in peak flows during a particular rainfall event are mainly influenced by changes in the urban layer which restricts the water being infiltrated into the soil. Qin et al. (2016), Tillinghast et al. (2011) and Zhou et al. (2015) performed studies to provide awareness for improvement in levels of water infiltrated and attaining the natural peak flow values in urban hydrographs. The major target is on improving the percentage of deteriorating groundwater levels and increasing water demand. Green infrastructure, another way to reduce stormwater flow, includes infiltrative rain gardens which have no impact on the environment and this technique requires less maintenance (Yang et al. 2015; Limos et al. 2016; Shaneyfelt et al. 2017; Wadzuk et al. 2017). Potentials to reduce peak runoff, reuse excess water and mimic pre-developed scenarios for an urbanized catchment furthermore includes the resilient design dimensions sustaining mutable storm events (Sorup et al. 2016). For managing urban runoff quantity, Fletcher et al. (2015) suggested that stormwater infiltration and retention-based technologies may be adopted either close to the source or at the end of the catchment (commonly known as decentralized SWH practice). Burns et al. (2012) demonstrated that the typical rainwater tank scenarios can concurrently assist in restoring pre-development flow regimes and reliably augment potable supply. Many studies on the application of permeable pavements and rainwater harvesting tanks in reduced pervious areas report that in the case of area restriction for construction of stormwater control measures (SCM), permeable pavement or rainwater harvesting or their combination is relatively effective in reducing potential urban runoff (Fletcher et al. 2007; Hamdan 2009; Alam et al. 2012; Mahmoud et al. 2014; Mankad et al. 2015). For micro-scale studies, implementation of SWH like infiltration trenches (Ahammed et al. 2013) and bioretention in combination with rain gardens and bioswales (Luell et al. 2011), in a decentralized manner, promises substantial achievement in stormwater goals. From a management perspective, infiltration trenches and bioretention tanks require periodic cleaning and pre-treatment. Measures like green roofs and grass swales (Singh & Kandasamy 2009; Penn et al. 2013) were found to be the best alternative to reduce the peak runoff but due to load considerations, their implementation on old buildings and commercial regions becomes highly questionable. For developing countries like India, Gogate et al. (2017) and Koh et al. (2016) developed a multiple attribute decision-making process for selecting the most suitable stormwater management alternative for a part of Pune city, and they concluded that leaky wells combined with rain gardens was the most suitable SWH alternative.

In summary, it can be stated that the choice of an appropriate SWH method, in general, is controlled by factors like economy, space availability, ease of implementation, and technology available locally. By taking into account the aforementioned factors in selecting an appropriate SWH technology for a locality, research is being presented in this paper to demonstrate the effectiveness of percolation ponds as a viable alternative for SWH for towns/cities in a developing country like India, where the implementation of LIDs like rain gardens, porous pavements, vegetative swales, bio-filters and infiltration trenches may not be feasible both practically and economically. The inspiration for use of percolation ponds as a SWH system was derived from the ancient temple tank system prevalent in South India. In ancient and medieval times, a temple tank served not only a spiritual purpose but also served as a water harvesting structure (Rao & Han 1987; Lei et al. 2008; Arulbalaji & Maya 2019). These tanks were more common in South Indian temples, since all the rivers in South India were rain-fed, and hence were less dependable for their flow during dry months. To overcome a shortage of water during dry months ponds were constructed which stored excess runoff during rainy seasons and percolated the water into the groundwater system. These ponds were well connected with each other such that excess flow from one pond would flow to the next connected pond. Hence, the ponds not only served as percolation units but also served as temporary detention for flood water. Other reasons for selecting percolation ponds is that they are economical, can be constructed with locally available technology, can be easily retrofitted into the existing storm sewer system and, more importantly, can facilitate in increasing groundwater levels locally through infiltration (Central Groundwater Board 2007; Gogate et al. 2017). In this paper, we present a methodology for SWH by emulating the ancient temple tank systems. The SWH system is designed by adopting a decentralized approach. In a decentralized approach a larger catchment is divided into smaller units so that it becomes easy to construct and maintain a SWH system. To check the efficacy of the proposed SWH system, the optimization model was applied to a case study area which is a part of Katpadi Town in the state of Tamil Nadu, India. The size of the study area is about 19 km² and almost 33% of this area is residential, the remaining area consisting of undeveloped land and vegetated land. The results indicated that a decentralized approach to SWH through a system of percolation ponds and lined channels can improve water availability potential locally.

The pre- and post-urbanized runoff resulting from Block 1 is used as input to one of the constraints in the optimization model. Block 2 consists of the optimization model framework, wherein decision variables, objective function and constraints are discussed. The optimization model is solved using the appropriate optimization tool. The output of Block 2 will be the optimal dimensions of the percolation ponds required. Finally, the performance of the designed percolation ponds under different rainfall events are determined using pre-defined efficiency measures through the performance evaluation of the SWH system. A detailed application of the proposed methodology to a case study area is discussed in the next section.

To demonstrate the applicability of the proposed SWH system, the optimization model was applied to a case study region in Vellore, Tamil Nadu. Table 2 shows the topographical characteristics of the case study area. The case study area selected is a part of the larger Katpadi region, which in itself is a part of larger Vellore town. The total geographic area of the study area is about 19.01 km². In this area about 30.7% is urbanized, water bodies occupy about 3.1% of the area, 28% is vegetated land and about 38.2% is barren or undeveloped land.

A schematic figure of the case study area can be seen in Figure 2. The major portion of the urbanized area is residential with a small portion occupied by commercial developments and institutions like schools and local offices. There exists a combined sewer system in the study area, laid along the major roads to carry domestic wastewater and stormwater (https://www.twadboard.tn.gov.in/). The average annual rainfall for the region is about 971 mm and much of the rainfall in the region occurs between the months of July and November. Most of the residences in the area do not practise any water harvesting methods. The average groundwater level in the area measured along the 21 bore and open wells (http://www.fnwm.gov.in/) fluctuates between a maximum of 17.1 m bgl (below ground level) to a minimum of 8.41 m bgl. The only source of groundwater recharge in the area is due to natural infiltration process, and no specific measures are being taken by the community in the area to improve groundwater level (Shanmugasundharam et al. 2017). As per the CCBP report (CCBP 2019), Katpadi town is categorized as a high risk zone for water shortage wherein ward nos. 1, 2, 13, 14 and 15 corresponding to the sub-catchment numbers 1, 4, 6, 7, 13, 14 and 16 as per the catchment division proposed in the study, experience water shortage problems throughout the summer season. The soil type in the study area is sandy loam with an average hydraulic conductivity of 1.1 cm/h (Kanagaraj et al. 2019). In future, as the area becomes more urbanized and the population grows, immense stress on the groundwater source can be expected. It is, therefore, the right time to implement suitable SWH measures which will increase the water level in the groundwater table. To assess the level of decentralization that will be beneficial from a hydrologic and economic perspective, the following two SWH options were considered:

• Option I: SWH through a series of interconnected percolation ponds and with decentralization at sub-catchment unit (SCU) level.

• Option II: SWH through a series of interconnected percolation ponds, with decentralization at sub-catchment (SC) level.

The catchment here is the entire study area; sub-catchment is the unit of division when the catchment is divided into smaller units; and sub-catchment unit is the unit of division resulting from further division of a sub-catchment.

For decentralization of the SWH system at SC level, the entire catchment area was divided into 18 sub-catchments, with each sub-catchment roughly corresponding to a ward as defined by Vellore Municipal Corporation (https://www.tn.gov.in/dtp). The further division of a sub-catchment into SCUs was accomplished by roughly identifying the area enclosed within the major roads (Figure 3).

For this purpose, Google Earth VHR satellite image depositories with a spatial resolution ranging from 15 m to 15 cm were used. The final classification of the study area into SCs and SCUs can be seen in Figure 4. One of the primary assumptions in the proposed optimization model is that each SC or SCU will have only one percolation pond such that all the surface runoff from the impervious area within an SC/SCU will flow through gravity towards the pond. This will be possible if the pond is located at the lowest elevation within a SC/SCU. The lowest elevation point within a SC/SCU was identified using elevation profile mapping of the study area in Google Earth and this lowest point in a SC/SCU is identified as the suitable site for situating a percolation pond (the arrangement of interconnected tanks is shown in Figure 5). It should be noted that the soil type is similar (sandy loam) throughout the study area and the groundwater table is well below the ground level in all the sub-catchments. Figure 5 shows a schematic representation of interconnected ponds within a sub-catchment. Within the catchment, the ponds are arranged in such a fashion that the excess runoff from an upstream pond will flow into the next pond. All the ponds therefore are interconnected and the entire arrangement of percolation ponds at catchment level can be envisaged as a cascade of micro-reservoirs.

As the total system cost will directly depend on the size of the pond, the decision variables for the optimization model will be: length of percolation pond in catchment division i, L_i (in m), width of the percolation pond in catchment division i, B_i (m), and depth of the percolation pond in catchment division i, D_i (m). The lower limit for the pond dimensions were selected arbitrarily as 5 m for length and width of the pond and 0.5 m for the depth of the pond. The upper limit for depth of the pond was taken as 4 m, since the maximum depth of a percolation pond, in general, is about 4 m (CPHEEO 2012). The upper limit for length and width of the pond in a catchment division were defined based on the area constraint for that respective catchment unit.

The optimization model discussed in the earlier section was applied to the case study area of Katpadi, a semi-urbanized catchment. The optimal dimensions of the pond in the catchment unit were determined for either of the options for a design rainfall intensity of five-year return period. The design rainfall intensity for each catchment unit was determined using the IDF equation shown in Table 5. The IDF curve was developed using daily rainfall data from the years 1950 to 2019. When applied to a catchment unit, the duration of rainfall in the IDF equation was taken equivalent to the time of concentration of the catchment unit (a valid assumption made in the application of rational method (Chow et al. 1987).

The optimization model was solved using the fmincon (non-linear constrained optimization algorithm) toolbox in MATLAB. After solving the optimization model in MATLAB, the optimal pond dimensions for Option I and Option II were obtained. The optimal dimensions for the percolation ponds under Option II and a sample result of optimal dimensions of percolation ponds under Option I are shown in Table 6.

Due to a paucity of space, the optimal pond dimensions for remaining sub-catchments under Option I are provided in the Supplementary material (Tables SCU1 to SCU18). From the results, it can be inferred that the size of the pond in a catchment unit is majorly influenced by factors such as size of the catchment unit, location of the pond in the series and space availability within the catchment unit. For instance, SC13 has the largest pond dimensions as it receives runoff from many upstream sub-catchments; also, the space available is sufficiently large. For the catchment units of roughly equivalent size (SC15, SC16 and SC17) and not receiving any runoff from upstream ponds, the size of the pond (Table 6) is influenced by the percentage of imperviousness (45.78%, 52.79% and 11.72% for SC15, SC16 and SC17, respectively) of the catchment unit (Table 3). In Table 7, the percentage area occupied by percolation ponds in a catchment unit with respect to the total impervious area of that catchment unit are shown.

From the results shown in Table 7, it can be inferred that the distribution of percolation ponds in Option II is more commensurate with the space availability constraint and imperviousness of the sub-catchment, whereas in Option I it can be inferred that the size of percolation ponds in a sub-catchment is mostly influenced by the percentage imperviousness in the sub-catchment. This can be observed for sub-catchment SC15 and SC17, which have almost the same area, i.e., 166.5 ha and 166.4 ha, respectively. The percentage of area under percolation ponds is 0.75% for SC15 and 0.39% for SC17, which is almost proportional to the percentage impervious area of SC15 (45.78%) and SC17 (11.72%), respectively. This is in contrast to Option II in which percentage area under percolation ponds is significantly less for SC15 (0.02%), compared to SC17 (0.19%), implying that the size of pond is inversely related to percentage impervious area. In Option I, SWH is more evenly distributed, with eight sub-catchments having more than 0.3% area under percolation ponds; seven sub-catchments having percentage area under percolation ponds between 0.15% and 0.3%; and three sub-catchments having percentage area under percolation pond as less than 0.15%. In Option II, there are six sub-catchments where the percentage area occupied by percolation ponds is less than 0.15%; in five sub-catchments the percentage area under percolation ponds is between 0.1% and 0.2%; in three sub-catchments percentage area under percolation ponds is between 0.2% and 0.5%; and there are four sub-catchments where percentage area under percolation ponds is more than 0.5%. This implies the area under SWH is more skewed under Option II when compared to under Option I.

In Table 8, an economic comparison is shown between Option I and Option II. The capital cost of the SWH system under Option I was obtained as Rs. 555 Million, whereas under Option II it is Rs. 714 Million. The unit cost of SWH system (calculated as a ratio of volume of water infiltrated to the capital cost of the system) under Option I is better when compared to under Option II. Hence, it can be inferred that economically Option I is preferable to Option II. Table 8 also shows the per-capita water demand that can be delivered to the population of the study area (which is about 38,908 as in year 2019) under Option I and Option II, which is 86 lpcd and 52 lpcd, respectively. This indicates a better per-capita water demand can be achieved with Option I when compared to Option II.

In Equation (10), is the volume of runoff retained by the SWH system and is the runoff volume from the contributing catchment unit.

Table 9 provides a summary of the efficiency measures under Option I and Option II. As can be observed from their definitions, E_rmax is defined for a single rainfall event, whereas E_r and E_i are defined for total annual rainfall. To understand the performance of the proposed SWH system in managing peak runoff of different magnitudes, E_rmax was determined for three different return periods, defined as day-to-day domain (0.2-year return period), design domain (5-year return period) and extreme event domain (66-year return period) (Sorup et al. 2016). E_r and E_i were determined for the daily annual rainfall data for the latest year 2019. The efficiency results are tabulated in Table 9. From the results it can be observed that the runoff capture efficiency is significantly higher for Option II for all the rainfall domain events, when compared to Option I. Between the different rainfall domains, E_rmax is higher for both SWH options for day-to-day domain and reduces significantly for extreme event domain. It is, however, interesting to note that the difference in E_rmax value between Option I and Option II for the day-to-day rainfall event is significantly large which implies that Option II is significantly better at capturing runoff resulting from low rainfall events when compared to Option I. Even for extreme events, Option II performs better when compared to Option I. When it comes to E_r, which indicates the ability of a SWH system to exploit rainwater as a resource, there is little significant difference between Option I and Option II. This implies that for daily rainfall events, division of catchment into sub-catchments or sub-catchment units has little bearing on reducing runoff volume. The last column in Table 9, E_i, indicates the capability of the proposed SWH system in infiltrating the stored water. This is again higher under Option I (5.56%) when compared to under Option II (3.37%).

To understand the functioning of the percolation ponds in controlling peak runoff volume and infiltrating the captured runoff within the pond, the designed optimal SWH system was simulated for a daily rainfall data series (23 days) of year 2015 (the highest rainfall year within the past 25 years). The rationale for selecting this set of data was that it was the duration in which a total depth of about 504.7 mm of rainfall occurred, which was about 38% of the total annual rainfall (1,374 mm) in that year.

The rainfall hyetograph for the data used for simulating the SWH system is shown in Figure 6. For the sake of brevity, the hydrographs are shown only for a few sub-catchments, i.e., most impervious (Figure 7(b)), least impervious (Figure 7(a)), smallest (Figure 7(c)) and largest (Figure 7(d)), along with the hydrograph for the entire catchment. The result indicates that implementation of Option I, when compared to Option II, has a significant effect on peak reduction in the largest sub-catchment, whereas for least urbanized sub-catchment there is little significant effect on runoff volume reduction in either of the options. For the largest sub-catchment, the runoff hydrograph resulting from Option I is closer to the pre-urbanized hydrograph, compared to the runoff hydrograph from Option II. When the entire study area is considered, the resulting runoff hydrograph (Figure 8) is very similar in both the SWH options, and in both cases, the resulting runoff hydrograph is lower than the post-urbanized hydrograph.

Finally, Figure 9 shows the potential depth of monthly infiltration that can be achieved after implementing the proposed SWH system. The values are for the monthly rainfall data of the year 2019.

In this paper, a decentralized SWH system using a series of percolation ponds for a semi-urbanized catchment has been proposed. The methodology of the proposed SWH design starts with the division of a larger catchment into smaller sub-catchments and sub-catchment units, determination of suitable location for situating percolation ponds using DEM and Google Earth archive and, finally, determining the optimal dimensions for the percolation ponds against minimization of the total system cost. The proposed methodology was applied to a case study area in Vellore, Tamil Nadu. The efficiency of the proposed SWH system in reducing peak runoff volume was estimated for rainfall intensity pertaining to day-to-day domain, design domain, and extreme domain values. The results indicated that the performance of percolation ponds in reducing the peak runoff is better for Option II when compared to Option I, under all three domains. The inference from this result can be that further division of a sub-catchment into SCUs can have a negative impact on peak runoff volume reducing capacity of the ponds. However, with respect to capital cost for implementation, unit cost of SWH, potential volume of water infiltrated, percentage area under stormwater harvesting system for the entire catchment and the ability of the proposed SWH system in exploiting rainwater as a resource, Option I fare better compared to Option II in all the aforementioned metrics. When the proposed SWH system was evaluated for the highest rainfall month that had occurred within the last 25 years, the resulting runoff hydrographs (Figure 7(a)–7(d)) for the least urbanized, most urbanized, largest and smallest sub-catchments indicated that there was a considerable reduction in peak runoff in both Option I and Option II.

Although, in the present study, the viability of using a series of interconnected percolation ponds for SWH has been discussed, there are however certain limitations in the proposed methodology. The primary limitation is with respect to the use of percolation ponds, which should be adopted only when the groundwater table is well below (preferably 8 m bgl) the ground level and also the underlying soil should be moderately permeable to facilitate faster infiltration of water from the ponds. Also, the methodology discussed will be more suitable for towns/cities where sufficient space is available for the construction of ponds. Existence of a well-laid out storm sewer network for an area can minimize the cost and effort in diverting water from impervious areas to the ponds. The optimization model was solved using a deterministic approach, whereas, in a more realistic setting, there can be many uncertainties involved, for example, uncertainty in rainfall, uncertainty in estimating infiltration rate from the ponds and uncertainty in land use changes. However, the methodology presented in this research paper can be considered as a step towards designing a more complicated SWH system using percolation ponds which can include the aforementioned uncertainties.

To conclude, we feel there are many cities/towns across India, where the proposed SWH system can be implemented beneficially, notwithstanding the limitations mentioned above. There is, however, a lot of scope for further improving the proposed method of SWH using percolation ponds. For example, in the optimization framework, the possibility of supplying the harvested water for satisfying a proportion of daily water demand was not considered. When included in the optimization model, this can affect the sizing of the ponds. Another interesting area for research would be to examine the uncertainties involved in the percolation process through the ponds, and its effect on their recharge potential. It will also be interesting to evaluate the performance of the proposed SWH system under climate change scenario and potential change in urbanization of the area in future.

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