Hydrologic and cost benefit analysis at local scale of streambed recharge structures in Rajasthan (India) and their value for securing irrigation water supplies

The surveys of check dam impoundments allowed calculation of basic metrics for each check dam, as shown in Table 1. Components of the daily water balances for these four check dams in 2016 are plotted in Fig. 3.

Water balance components for the four check dams are summarized for each of 3 years in Table 2. Rainfall in 2016 averaged twice as much as for the two preceding years and generated 72% of the estimated check dam inflows over the 3 years. Although average rainfall totals in 2014 and 2015 were almost identical, there were significant differences in the temporal and spatial distribution of rainfall that led to estimated runoff of 22 and 6% of the 3-year total in those years respectively. Hence estimated inflows relate closely to farmers perceptions that 2014 was an “average” year, 2015 a “dry” year and 2016 a “wet” year. In total across sites and years, 76% of estimated inflow was spilled, 19% was recharged and 5% evaporated. The estimated proportion of inflow that was recharged from all check dams was 37, 70 and 10% for 2014, 2015 and 2016 respectively. Conversely spill accounted for 57, 18 and 87% of inflow and evaporation for 7, 16 and 3% over these years. Total annual recharge from the four check dams in these years was estimated to be 976,000, 510,000 and 850,000 m³ respectively.

The recharge to capacity ratio varied between check dams and years, with a mean of 1.66 and range from 0.17 to 3.40. Among years, the mean recharge to capacity ratio for all four check dams ranged from 1.09 in the “dry” year (2015), 1.81 in the “wet” year (2016), to 2.08 in the “average” year (2014), with a coefficient of variation of 0.31. (These values differ slightly from those reported by Dashora et al. (2018), which was the mean of the individual check dam ratios, rather than the ratio of the totals.) Although the check dams contained water for a longer period in the wet year, the mean infiltration rate was lower than in the other years. The reasons for this include possible hydraulic connection between check dams and aquifers, and high sediment loadings due to intense rainfall events up to 250 mm, and are discussed later in the paper. Among check dams, the ratio of average annual recharge to check dam capacity over the 3 years ranged from 1.01 at Sunderpura (with the highest ratio of detention capacity to catchment area) to 2.48 at Badgaon.

The ratio of recharge to runoff, as expected in the “average” year 2014, was ranked in the same order as the ratio of check dam capacity to catchment area, and approximately so in other years. In general, in the low flow year the proportion of flow recharged (except for Badgaon) increased but in 2016, the “wet” year, the proportion of recharge to inflow declined, as expected for all check dams. Changes in the catchment upstream of Hinta check dam occurred (under Mukhyamantri Jal Swawlamban Yojna) in 2016 which may have contributed to an apparently reduced runoff relative to other catchments.

This shows that there are large season-dependent variations in the hydrologic impacts of check dams. Their impact will be felt greatest in relation to natural recharge in dry years, and at these times, the reduced spill could be conspicuous for downstream communities. The results suggest that longer-term studies are required in order to produce reliable conclusions on long-term or average performance of check dams. In this study, by good fortune, rainfall patterns were favourable, including dry, average and wet years, but this cannot generally be relied upon in design of recharge studies. Furthermore, an economic analysis of the whole catchment is also warranted to incorporate effects of reduced downstream flow including costs (e.g. risk of reduced agricultural production downstream) and benefits (reduced flood risk), such as have been evaluated by Chinnasamy et al. (2018b) for flood water harvesting in the Ramganga Basin, India.

The sensitivity of water balance components to the estimated evaporation rate, the discharge coefficient of the weir and the estimated wet-weather-infiltration rate (WWIR) were calculated. Results for the Hinta check dam (the largest check dam studied) in 2014 (the “average” year) are shown in Table 3. Each parameter in turn was increased from its “base case” value by 20% and the corresponding percentage change in each of the water balance components was calculated. Increasing the rate of evaporation by 20% gives a directly proportional increase in the volume evaporated; however, the reduction in recharge is only 3.5% because the evaporation was only 18% of the recharge for the base case. Increasing the weir coefficient has a directly proportional impact on spill and increases the calculated inflow by 7.5% but has no effect on recharge as the recharge calculation does not make use of inflow or outflow. Increasing the WWIR by 20% increases the calculated recharge by only 2.4% and inflow by 1.2% with a very small decline in calculated evaporation. The main reason for low sensitivity of recharge to WWIR is that 87% of the “base case” recharge occurred in dry weather. The low sensitivity of calculated recharge to these parameters that are extrapolated to remote streambed structures, or estimated on the basis of daily gaugeboard readings by farmers, suggests that the method used is suitably robust for estimating recharge from check dams with this level of data availability.

Dashora et al. (2018) analysed the reliability of farmer measurements of check-dam water levels and found that 96% of readings were within ±1 cm of values from concurrent photographs as read by a university researcher. In summary, while runoff estimates are considered unreliable, the recharge estimates were found to be relatively insensitive to unmeasured parameters and are considered robust.

The check-dam inflow volume was calculated on a daily basis using water level measurements and area-volume elevation curves, to calculate inflow, spill, recharge and evaporation. No information on runoff coefficients (ROC) was available in these check dam catchments, so results were compared with estimates from Strange’s table (source: Subramanya 2008), which is commonly used in India to estimate annual monsoon runoff without the ability to validate predictions. Knowing the low level of accuracy of the spill calculation based on only daily water level readings, this analysis was partitioned to focus on only the years and check dams with no spill, where inflow estimates were considered reliable. Rainfall was recorded only at a single rain gauge in or adjacent to each catchment that ranged up to 1,705 ha. The seasonal runoff coefficient was calculated using the seasonal inflow and dividing by the estimated volume of seasonal rainfall on the catchment. Table 4 shows that for the three cases occurring in the “dry” year, 2015, the best predictor occurred under Strange’s “bad” (least runoff) catchment category, which still overestimated runoff by a factor 1.8–5.1. Runoff in the “average” year 2014 at Sunderpura compared very favourably with Strange’s table predicted runoff for an “average” catchment type. This information suggests that Strange’s table should be used with caution.

There are two main factors that dictate recharge from any check dam, the presence of water in the check dam and its rate of infiltration. The MDWIR, measured over days when water level is lower than the weir crest and falling at least as fast as the evaporation rate, is taken to be the rate of recharge during all other times, when inflows are estimated. The errors, some compensating, associated with this approximation are discussed by Dashora et al. (2018).

During this study, MDWIRs ranged between 10 and 57 mm/day, and the mean was 27 mm/day, more than 5 times the estimated evaporation rate from the dam surface of 5 mm/day (Table 5). Importantly, the highest DWIR was observed at Badgaon in 2015, immediately following manual scraping to remove silt from the floor of the check dam. The two lowest rates were recorded in Dharta and Hinta check dams immediately following mechanical scraping of the check dams, with the lowest rate after two successive years of mechanical desilting. MDWIR in the driest year 2015 is particularly clear where the largest increase from 2014 occurred in the hand-scraped check dam and the only decline was in the mechanically scraped check dam. This appears to be evidence of soil compaction by the heavy machinery used at Dharta check dam, which actually had the opposite of the intended effect of scraping.

From Table 5, it can be seen that MDWIR was highest in 2015, the ‘dry’ year, for all check dams except for mechanically scraped Dharta, and was lowest in 2016 the ‘wet’ year for all check dams. In 2014, the ‘average’ year, MDWIR had intermediate values close to the mean of the 3 years for each check dam. Although this is a small sample and influenced also by desilting, it appears that wetter seasons produce lower infiltration rates. One possible reason for this is that higher flow years produce higher rates of siltation; however, this would also suggest a monotonic decline in MDWIR, which did not occur in 2015, even though in three check dams there was no spill and hence no scour and flushing of sediment. Another factor could be differences in water temperature affecting kinematic viscosity of water, and hence hydraulic conductivity of streambed sediments and is discussed later.

A third consideration was the impact of wetter years resulting in higher natural recharge to the unconfined hard-rock aquifer causing water table to rise and hydraulic connection with the impounded surface water. The process of hydraulic connection has been studied and modelled with a boundary element model having a moving upper surface and found to quickly reduce the recharge rate to about 10–20% of the preceding hydraulically disconnected rate, due to change in hydraulic gradient surrounding the streambed (Dillon and Liggett 1983) and the effect is larger for wider impoundments such as check dams (from electrical resistance analogue models, Bouwer 1978; and using a boundary element model, Dillon 1984). The differences in MDWIR between check dams in any year (except the first before any desilting) were similar to or larger than differences between years, suggesting that the effects of desilting, combined with other variations between sites, were as significant as variations between years (Fig. 4). The year of least standard deviation and coefficient of variation was 2014, before desilting commenced. The check dam with highest standard deviation across the three years, more than twice that of others, was Badgaon, the manually scraped check dam, suggesting that this change was significant.

The changes in MDWIR between 2014 and 2015 would be the changes least influenced by groundwater interactions. Relative to the net increase for unscraped check dams of 13%, the Badgaon check dam that was scraped by hand labour increased a further 70%. For the Dharta check dam that was scraped using heavy machinery, its MDWIR decreased 18% and so relative to the unscraped check dams declined by 31%. Although this is a very small data set, the implications could be very important information for the design of future maintenance programs for check dam desilting. Check dams for desilting could be prioritized on the basis of measured MDWIR. This occurred in 2016 when the two check dams with lowest MDWIR in 2015 were scraped. Unfortunately, this was by mechanical scraping and MDWIR decreased in both cases, but decreases were recorded at all check dams. More data will be needed to help formulate appropriate strategies, accounting for the benefits and costs of desilting, with a greater emphasis on methods that avoid compaction of sediments. Elsewhere, it has been shown that scraping has improved recharge capacity. In a study by Mousavi and Rezai (1999) of three artificial recharge sites in Isfahan Province, central Iran, on average, the average infiltration capacity of the untreated recharge facilities had declined to 20.3% of its initial value, and that 68.3% of initial infiltration capacity was restored after scraping off the top 15 cm of silt.

Variations in DWIR are shown over the 3 years and four sites in Fig. 5. Each site and year follows a pattern of high rates early in each monsoon season, reaching 150 mm/day at each site for short periods. This reduces to low values in the mid-monsoon period and in prolonged monsoons, most notably in 2016, when there are extended periods of very low infiltration rates. Finally, as the season ends, the rate of infiltration increases at most check dams.

Although the air temperature ranges from 15 to 30 °C during the ponding period, there was no correlation found between DWIR and air temperature. Air temperature falls at the end of each monsoon season when infiltration rates increase; hence, daily air temperature was not responsible for the observed increase in infiltration rates. As water level rises within a check dam it had been expected that this would increase the hydraulic gradient for seepage; however, this is not supported by the observations. These show lower DWIRs at higher storage levels, but due to the larger surface area, daily volumes of recharge are less affected.

Groundwater levels were monitored daily at three wells closest to each check dam when they contained water and weekly throughout the rest of the year. At Badgaon the closest well was 68 m upstream of the weir. For other sites, the closest well was >470 m from the check dam. Figure 6 compares daily values for the groundwater in the proximal well and the Badgaon check dam water level. The term ‘reduced level’ (RL) is used as the cease to flow (spillway crest) level of the horizontal broad crested weir is adopted as a temporary bench mark and assigned a reduced level of 100 m (RL 100 m). The lowest elevation of the impoundment is at RL 97.7 m. The DWIR is plotted daily using a secondary axis (on the right).

The bottom of well B9 is at RL 80 m, and each year in the dry season the well ran dry. In the wet seasons of 2014, 2015 and 2016, the peak groundwater level reached RL 97.7 m (the lowest ground surface), 95, and 98.7 m respectively. The DWIR reduces to approximately zero whenever the groundwater elevation at B9 is above RL 97 m, which corresponds with hydraulic connection between the aquifer and the check dam. This is most clearly observed in 2016. At this location, surface-water levels remained above groundwater level so no direct groundwater discharge to the check dam occurred; however, observation of Fig. 3 shows a sustained period of spill between 24 August and 14 September 2016 with only intermittent rainfall, suggesting that groundwater discharge was occurring into the stream upstream of the check dam.

Similar data are shown in Fig. 7 for Sunderpura check dam, where the closest observation well (SP26) is 656 m downstream of the check dam, and the local temporary benchmark for the crest of the check dam is taken to be 100 m and the lowest ground surface of the impoundment is 97.7 m. In the wet seasons of 2014, 2015 and 2016 the peak groundwater level reached 97.4, 96.4, and more than 100.5 m respectively. In years 2014 and 2015 there was no spill and marginal evidence of hydraulic connection in 2014; however, in 2016, the wet year, the groundwater level in SP26 not only demonstrated hydraulic connection but rose to above the level of surface water in the check dam due to significant natural recharge. This suggests that groundwater discharge to the check dam occurred between 24 August and 7 September as also indicated by spill in Fig. 3. When groundwater levels in well SP26 dropped below about 96.5 m the DWIR again increased, suggesting that the check dam and aquifer had again become hydraulically disconnected.

At Dharta and Hinta check dams, the monitored wells were not sufficiently close to the check dams to conclusively determine whether hydraulic connection occurred in 2014. Hydraulic connection was unlikely in 2015 when levels peaked 5–10 m below the lowest levels of the impoundments, but was evident from a period of low DWIR, and from evidence of continuous flow from 24 August to 4 Sept 2016 for Dharta and from 24 August to 10 Sept 2016 for Hinta. Thus, each check dam, depending on its position in the catchment, the nature of the aquifer and ambient groundwater depth, had different propensities for hydraulic connection, which suggests that for long-term studies there would be a great advantage in having monitoring wells very close to check dams to interpret recharge behaviour.

The cost of each check dam was taken from government department records and varied from 150,000 to 500,000 Indian Rupees (INR) in their year of construction, which ranged from 1995 to 2005 (Table 6). The benefit cost evaluation here is done in Indian Rupees (INR) in the year 2014. In that year the average exchange rate with the US dollar was 60 INR per US$ (max 63.69, min 58.43).

The maintenance costs of each check dam (primarily scraping for silt removal and occasional repairs to concrete structures) were estimated for manual and mechanical scraping (Table 7). No maintenance costs were recorded for the four check dams from their establishment until after 2014, and for Sunderpura check dam there has been no maintenance for 22 years. The check dams had quite different mean annual maintenance costs, which discounted to year 2014 INR, ranged from 0 to 7.3%/year of the capital costs with a mean of 2.9%/year.

Table 8 summarizes the annualized costs expressed in terms of the present value in 2014 INR, for capital, and applied to all check dams with the assumed maintenance cost, taken as the mean from Table 7. It also shows the average annual recharge over the 3 years of record, 2014–2016. Using the approximation that the mean annual recharge in these 3 years represents the mean annual recharge over a 30-year life of these check dams, given this level of maintenance, the average unit cost of annual recharge (CR) is obtained (Eq. 5).

Benefits were calculated using data provided by the Irrigation Depatment, Rajasthan Department of Agriculture (2017) and ICAR (2009) and Eqs. (6)–(9). Major rabi season crops are wheat and mustard having on average 44 and 36% of the total cropped area, respectively. Water uses for these two dominant crops were measured in farmers’ fields by installing a water meter near the well head. For the remaining mix of crops, the water use was referred from sources as listed in the footnote of Table 9. The area fraction of opium (0.7%) is very small compared to major crops in Dharta watershed, but to eliminate any possible impression that this crop may be responsible for elevating the determined benefit cost ratios, the net profit and water use of wheat were substituted for opium. This marginally reduced the area-weighted profit per unit volume of water from 2.42 to 2.36 INR/m³ (in INR 2014) for Dharta watershed using groundwater. This lower value was used in the benefit:cost analysis.

This paper has addressed only one aspect of the benefits of recharge enhancement, being the main motivation to secure and expand rabi season irrigation supplies. Beernaerts (2006) proposed an evaluation framework for groundwater recharge enhancement systems for improved irrigated agriculture with economic, social and environmental indicators. In addition to the economic evaluation of the current study, Beernaerts (2006) proposed evaluating: economic impacts on improved soil fertility using silt removed from check dams; the potential for increase or decrease in fluoride or arsenic concentration in the aquifer and for pollution by faecal pathogens, taking account of water quality improvements (e.g., as reported by Patel 2002 with localized lowering of salinity); the rise in water table with consequent reduction in energy required for pumping, improved conditions for survival of riparian trees and control of saline intrusion in coastal areas; the current externalities faced by downstream communities of reduced water availability for irrigation or city water supply dams and reduced incidences and severity of flooding and erosion. Beernaerts (2006) also considered a range of financial indicators for evaluating economic impacts of recharge enhancement and suggested that those measures that brought supply side and demand-side (such as water use efficiency) into perspective would have advantage to communities for decision making on water management.

Gale et al. (2006) reported the results of a project to determine impacts of recharge structures on livelihoods in village communities. Their study showed the reasons why this is difficult. They found that a lack of baseline data on the ‘before’ recharge situation hampered longitudinal comparisons of ‘before’ and ‘after’ income. MAR typically forms one of a number of watershed activities and agricultural reforms aimed at improving resource productivity, generating employment and supporting livelihoods. Hence it would generally not be straight forward to isolate the effects of recharge structures from the other effects. Gale et al. (2006) proposed that comparisons between ‘with’ and ‘without’ recharge structures also require a control group with a similar environmental and socio-economic profile to the ‘with’ group. This would enable confounding effects of rainfall variations between years and other factors affecting productivity to be removed. Finally, changes in economic conditions, access to infrastructure and other external factors may also impact on agricultural production. Gale et al. (2006) concluded that “attributing changes in livelihood strategies and outcomes to watershed development, and to MAR specifically, is therefore difficult to do with confidence”. However, surveys of farmers for their records and perceptions did reveal an association between increased livelihoods and streambed recharge structures. The approach used in this current study using recharge volumes from check dams determined from daily water balances, assuming a proportion of this (here 100%) contributes to crop production, and using crop area weighted mix of net profit per cubic meter to value the benefit, is a more rational approach than the ‘before and after’ method, but is not without limitations, as discussed earlier.

A comparable analysis for agricultural value of water use using surface-water irrigation was undertaken by Garg et al. (2012) in the Upper Bhima sub-basin of Maharashtra. The gross income was calculated using crop yields and support prices of crops for 2003–2004 for a known volume of dam water released and cropped area. Then deducting the cost of cultivation gave net income and water productivity was calculated to range between US$0.02/m³ and US$0.03/m³. Other researchers, Hussain et al. (2007), compiled more recent estimates of the average value of agricultural water in several countries that varied among countries. For Indian irrigated areas, the average value of agricultural water use was estimated as US$0.09/m³. After accounting for inflation, the higher value of water benefits obtained for the current study (Table 9) is attributed to higher efficiency in the use of groundwater over surface water, as has been observed by Burke and Moench (2000).

Benefit cost analysis is presented for the four studied check dams of Dharta watershed in Table 10. The unit cost of recharge, CR (Eq. 5), varies from 0.36 to 1.73 INR/m³ (from Table 8). The unit benefit of recharge, BR (Eq. 8), was found to be 2.36 INR/m³ (from Table 9) and is assumed constant for all check dams in this area. These unit values attributable to benefits and costs were used to calculate benefit:cost ratio (BCR) which ranged from 1.3 at Sunderpura check dam to 6.4 at Hinta check dam with a mean of 4.1 (Table 10). The Hinta check dam had the highest recharge (about half of the aggregate recharge of the four check dams), the lowest unit costs of recharge, and contributed 57% of aggregate net benefits from all four check dams. This suggests that the siting and design features were well matched for this check dam.

The earliest BCR found for Indian streambed recharge structures was for Baramati Taluka of Pune District, Maharashtra, which was declared in 1963 to be an area of “precarious scarcity” with total or near total crop failure occurring in 1 year in 3, due to drought. In response, a combined effort of government and NGOs constructed the first percolation tank in 1968, and by the end of 1978 more than 149 had been built with a combined capacity of 15 Mm³ (Dillon 1983). Percolation tanks are similar to check dams except that they have an earthen embankment and a separate concrete spillway, rather than just a weir on a stream. At 1980 the feasibility criterion for construction of percolation tanks was for cost to be less than 1.9 INR/m³ detention capacity and the average value was 0.92INR/m³ capacity (Dillon 1983). (In the current study this was 7.9 INR/m³ in 2014 Indian Rupees, approximately, half the cost from 1978 inflated at 8% pa rate to 14.7 INR/m³.) There were no measurements to enable recharge to be estimated, although Dillon (1983) recommended installation of gaugeboards and monitoring water levels in streambed structures and wells. The current study demonstrates the value of having such data.

Based on Maharashtra Irrigation Department records, the mean capacity of these first 149 percolation tanks was 98,700 m³, and the average area benefitted by each tank was estimated to be 36.2 ha. The crop mix in this area was jowar (millet) (45%), wheat (30%), sugar cane (10%), and vegetables, onions and gram (each 5%). The Irrigation Department’s estimated average annual increase in income per tank was 56,000 INR and net costs of production were not reported. The capacity of a tank divided by the estimated benefitted area is 272 mm. This is not inconsistent with water use results for the current study considering crop types and the likelihood of annual recharge to exceed detention capacity. The resulting benefit:cost ratio (BCR) derived by Dillon (1983) based on a discount rate of 10% was >6. Using a 30-year life and 8% discount rate, as per the current study, BCR would have been 7.0; however, for comparison with current data from Rajasthan (where profit is 28% of the increase in income), and allowing 3% capital costs for annual maintenance, and a 30-year life and 8% discount rate, a BCR of 1.5 is achieved. In the Baramati area the Irrigation Department also estimated a silt loading to percolation tanks of 1.7 m³/year/ha of catchment area. It would be very informative to record the current status, maintenance history and also the DWIR on these percolation tanks 40–50 years since construction.

The CGWB (2000) Guide on Artificial Recharge to Ground Water gives a brief summary of the procedure for undertaking an analysis to determine the benefit:cost ratio (BCR) of a recharge project, and gives the computed BCR for a number of recharge structures in Amravati District, Maharashtra. For three percolation tanks, BCR is 1.3–2.0, increasing with impoundment size (from 49,000 to 132,000 m³), and with an aggregate mean BCR of 1.76. Cement plugs are smaller, with a typical detention volume of 4,000–10,000m³, and for ten such structures the estimated BCR ranged from 0.88 to 5.80 with a mean of 1.88. The method of recharge benefit calculation assumed that the difference in recharge, and hence area that can be irrigated resulting in increased crop income, is attributable entirely to the check dam. However, in the current study there are significant variations in recharge and crop production between years due to the amount and pattern of rainfall within the monsoon. This and other agronomic factors would confound estimates of benefits attributable to check dams. A series of years before and after check dam construction would need to be compared to give confidence to these benefit estimates. Another advantage of the method used in this paper is that it can be applied even in the absence of information before the check dam was constructed. As presented in Table 9 for the area-weighted crop mix, the benefits can be calculated from per unit volume of recharged water even in a study period of 3 years. The benefit calculation depends on average crop mix in an area; thus, it can be applied for any check dam within the catchment.

Machiwal et al. (2004) used the contrast between expected production before and after check dam construction for a hypothetical example consistent with the authors’ perceptions, to estimate benefits of small-water-harvesting structures suitable for catchments smaller than 30 ha in southern Rajasthan. Their analysis used a lifetime of 10 years and a 10% discount rate; the analysis took into account capital costs based on design of the structure and presumed an annual maintenance cost of 10% of capital costs for masonry structures and 25%/year for earthen structures. The net profits per ha of wheat and mustard crops were slightly higher than those used in this current study after taking into account inflation. Importantly, they also considered benefits to occur during the kharif season, with one third of the benefits attributed to extra cropping in the kharif in their lower rainfall (541 mm) study area. In the current study, it was not possible to quantify benefits attributable to check dams of “life-saving” irrigation during dry spells in the monsoon. These benefits were ignored on the conservative presumption that the residual groundwater available for rabi irrigation may be reduced by irrigation in the kharif. Hypothetical benefit:cost ratios calculated by Machiwal et al. (2004) ranged from 5 for earthen structures to 3.5 and 1.9 for dry stone masonry of simple and more complex design, respectively. They concluded that concrete weirs, which cost considerably more, were not cost-effective in catchments of less than 30 ha.

By 2007 the Central Ground Water Board had revised and expanded its Manual on Artificial Recharge of Ground Water (CGWB 2007). In this, a chapter is devoted to economic evaluation of recharge schemes. The volume of recharge is estimated from the product of seepage rate, average area of water spread and the duration of storage. This implies monitoring of water levels in the check dam; however, the degree of adherence to this is unknown. Benefits can include income from additional agricultural production, and savings in pumping energy, as well as savings from constructing new tube wells and installing new pumps that would otherwise be done. The CGWB (2007) template for cost benefit analysis adopts the income from increased production rather than the net income after costs (28% of income in the current study), and the default value suggests crop irrigation requirement is 670–1,000 mm/ year, which is somewhat larger than the 413 mm/year for the cropping mix in the current study. Hence, on balance, the default values in CGWB (2007) likely overestimate the benefits in comparison with the current study. However conservative default values for annual costs are likely to at least partially compensate, by including an annual interest expenditure of 10% of the capital cost, maintenance and repairs charges of 2.5%, depreciation of civil works 5% and miscellaneous expenditure of 1%. An example of percolation tanks in Baramati Taluka of Pune District, Maharashtra, is given that used a life time of 15 years and very high discount rate of 15% at an undeclared date, with an interest rate on loans of 11.5%. The capital cost of 9 INR/m³ capacity, pre-2007, would considerably exceed the average capital costs for the current study at 7.9INR (in 2014) /m³ capacity. The size of the tank evaluated was 130,000m³, very similar to that of the Dharta check dam. For that example, the present value of costs exceeded the present value of benefits unless a 75% subsidy was applied to the capital cost. It is unclear whether capital costs as well as interest on capital are accounted for in this analysis, but it seems at odds with the BCR of 1.5 re-evaluated for the average of 149 check dams in the Baramati area, as shown in the preceding text.

Malik et al. (2014) followed a different approach, by interviewing 120 farmers with and without small private rainwater harvesting structures on their land in western Madhya Pradesh and recording the changes in reported crops and livestock. There was negligible change in kharif cropped area and livestock numbers, but a very substantial increase in cropped area in the rabi season followed construction of structures. In this area, of ~800-mm annual rainfall, irrigation water was drawn directly from farm water storages rather than by recharge to the aquifer and then extracting. Benefit cost ratios of 1.9 and 1.5 were found for alluvial and hard-rock areas respectively without a government subsidy. In hard-rock areas, the depth of excavation and hence impoundment depth was constrained. These structures occupied between 6 and 10% of a farmer’s land area, and so it is not surprising that with evaporation losses and low use of water for irrigation in the kharif season, the available irrigation water would be less than detention capacity, so the benefit:cost ratio would be expected to be lower than that found for groundwater recharge structures in wadis found in the current study, where average recharge was 1.66 times detention capacity.