Rainwater harvesting in India: some critical issues for basin planning and research

Transcription

1 1 Rainwater harvesting in India: some critica issues for basin panning and research M. Dinesh Kumar 1, Shantanu Ghosh 1, Ankit Pate 1, O. P. Singh 1 and R. Ravindranath 2 1 Internationa Water Management Institute, Anand Programme Office, Eecon Campus, Anand-Sojitra Road, VV Nagar, Anand, Gujarat, India PIN Foundation for Ecoogica Security, Near IRMA Gate, Anand, Gujarat, India PIN Abstract Often, as a frantic response to probems of water scarcity and consequent hardships faced by both urban and rura communities, India has invested heaviy in rainwater harvesting. Unike investment in arge water resource systems, these efforts, by and arge, ack hydroogica panning and sound economic anaysis: research on the impact of oca water harvesting/groundwater recharge activities in India is very sparse. This paper identifies six critica issues in rainwater harvesting efforts in water-scarce regions of India. First: there is no emphasis on potentia oca suppies and the demand they have to cater for: oca suppy potentia is ow in most water scarce regions, a fact compounded by poor reiabiity, and demand far exceeds the suppy potentia. Second: there are compexities in the economic evauation of RWH, due to ack of scientific data on infows, runoff coection and storage efficiency, beneficiaries, vaue of the incrementa benefits generated and scae considerations. With higher degrees of basin deveopment, the margina benefit from water harvesting at the basin eve reduces, whie margina cost increases. Third: in many basins, there is a strong trade-off between maximising hydroogica benefits and improving cost effectiveness. Fourth: many water-scarce basins are characterised by wide disparity in demand between upper catchments and ower catchments, so that there is a trade-off in maximising benefits of upstream water harvesting with optimising basin-wide benefits. Fifth: in many water-scarce basins, oca water harvesting merey divides the hydroogica benefits rather than augmenting them. Finay, poor integration between surface water and groundwater systems, and ack of incusion of natura recharge, utimatey eads to reduction in potentia for artificia recharge in hard rock areas. Introduction India has a ong tradition of water harvesting. Many of the traditiona water harvesting systems have either faen into disuse due to a variety of physica, socia, economic, cutura and poitica factors which have caused their deterioration, decine of institutions which have nurtured them (Agarwa and Narain, 1997) or have ost their reevance in the modern day context due to their inabiity to meet the desires of the communities. Whie the first dimension of the decine in water harvesting tradition has been we researched and documented, the second dimension is much ess understood and appreciated. The fact that different periods in history are marked by the genesis, rise and fa of some new water harvesting tradition (Pandey et a., 2003), is aso not appreciated. In this paper, the term rainwater harvesting structures refer to a sma structura interventions to impound excess runoff (

2 2 from agricutura fieds and runoff from sma natura catchments ( km 2 ), either for direct use by humans and catte or for aquifer recharging. They do not incude roof water harvesting systems that capture water from artificia catchments, or arge dams and diversion head-works that harness water from arge catchments. Water harversting in India has boomed during the past two decades in two markedy different ways from the traditiona techniques. They can now use recent advances in soi, geosciences and hydro-sciences, pus modern-day techniques and technoogies in survey and investigation, earth moving and construction. Whie the traditiona techniques represented the best engineering feats of the times, in terms of water technoogy used for water harnessing and distribution (Agarwa and Narain, 1997) and the voume of water handed, modern water harvesting systems are at best miniature versions of the arge water resource systems that use advances in civi engineering and hydroogy. The imited Indian research on rainwater harvesting (RWH)/artificia recharge so far has focused on the engineering performance of individua structures (Muraidharan and Athawae, 1998). Whie much anecdota evidence on the socia and economic gains exist, there is itte understanding based on empirica work of: (1) the impacts of water harvesting activities on oca hydroogica regime in terms of net water gain; (2)basin eve impacts on overa basin water baance; and (3)economic imperatives from a ong term perspective. Of ate, researchers have raised questions of the possibe unintended impacts of water harvesting (Bacheor et a., 2002), and its economics (Kumar, 2004). One of the reasons for the paucity of empirica research on the hydroogica and economic aspects of water harvesting systems is the inabiity to generate accurate scientific data on hydrauic, hydroogica and meteoroogica parameters that govern the performance and impact of water harvesting. Anaysis of water harvesting systems aso misses the infuence of the scae factor. Here, we do not try to anayse the physica performance or hydroogica impacts of water harvesting by taking a different approach. We begin with the basic premise that scae considerations are important in anaysing the impact of water harvesting, i.e. one has to move from anaysis at the oca eve to the river basin eve and that basin eve impacts are not aways aggregates of oca impacts. The paper therefore has the foowing objectives: (1) to discuss the critica issues in rainwater harvesting not ony from a micro perspective but aso a macro perspective; and (2) to present the issues for future research in water harvesting and recharging. Critica issues in rainwater harvesting One of the most important underying vaues in rainwater harvesting is that it is a benign technoogy (Bacheor et a., 2002) and cannot create undesirabe consequences. Water harvesting initiatives are driven by firm beiefs and assumptions, some of which are: (1) there is a huge amount of monsoon fow which remains uncaptured and eventuay ends up in the natura sinks, especiay seas and oceans, supported by the nationa eve aggregates of macro hydroogy; (2) oca water needs are so sma that exogenous water is not needed; (3) oca water harvesting systems are aways sma and are therefore cost effective; (4) since the economic, socia and environmenta vaues of water are very high in regions hit by water shortages, water harvesting interventions are viabe, supported by the assumption that cost effective aternatives that can bring in the same amount of water do not exist; (5) incrementa structures ead to incrementa benefits; and (6) being sma, with ow water storage and diversion capacities, they do not pose negative consequences for downstream uses. Lack of emphasis on oca water demand and potentia puppies Rainwater harvesting ignores a few critica parameters that govern the potentia of RWHS in meeting oca water demand. First is the hydroogica regime of the region/ ocaity. Second is the reiabiity of the suppies, governed by the reiabiity of rainfa. Third is the constraint imposed by oca geoogica and geo-hydroogica settings on recharge potentia. Fourth is the aggregate demand for water from various sectors within the oca area. Some basic hydroogica phenomena, which make these parameters very critica in deciding the scope of rainwater harvesting and groundwater recharging, are: For runoff harvesting, the rainfa has to exceed a threshod to generate runoff, a threshod which varies according to the nature of sois and and cover. The estimated runoff based on the regression equation derived from observed fows in the Hathmati sub-basin of Sabarmati basin (R= *X ) in western India (source: GOG, 1994) shows that for the runoff to exceed 100 mm, the minimum rainfa required is 682 mm, whie in the Kabani sub-basin of Cauvery, runoff starts when the rainfa exceeds 366 mm 1. However, actua runoff rates depend on the strength of the correation between rainfa and runoff in a given basin, and this reationship weakens if year-to-year changes in rainfa intensity and pattern are major. Regions with ower mean annua rainfa experience higher variabiity, and vice versa (Pisharoty, 1990). Hence, in regions with ower mean annua rainfa, rainwater harvesting as a dependabe source of water is ikey to be ow. Generay, a arger magnitude of annua rainfa means a greater number of rain days and a smaer magnitude of annua rainfa means ess rainy days spread over the rainy season (Pisharoty, 1990); exampes from Gujarat iustrate this further (see Kumar, 2002b; Kumar, 2004). Fewer rainy days aso mean onger dry spes and thus greater osses from evaporation for the same region. High intensity rainfas are common in semi-arid and arid regions of India (Garg, 1987 as cited in Figure 24; Athawae, 2003). Higher intensity rainfa can ead to high intensity runoff occurring in short durations, imiting the effective storage capacity of rainwater harvesting 1 The regression equation for Kabani estimated by Nationa Water Deveopment Agency based on observed fows was R=0.6363N where N is the rainfa (mm) and R the runoff (mm). (

3 3 systems to amost equa actua storage size. High evaporation during the rainy season means osses from surface storage structures. It aso means a faster rate of soi moisture depetion through both evaporation from barren sois and evapotranspiration,decreasing the rate and quantity of soi infitration. This reduces runoff generation potentia. Among the seven ocations in Gujarat for which ET 0 (reference evapotranspiration) data are avaiabe, ET 0 during monsoon (June to September) varies from a ow of 543 mm in Vadodara to 714mm in Rajkot. As a percentage of annua ET 0, it varies from 33% in semi-humid Surat to 37.3% in Bhuj, Kachchh (source: authors anaysis based on data from IMD, Ahmedabad). In the case of Rajasthan, ET 0 during monsoon ranges from 433 mm in the hi station of Mt. Abu to mm in Jaisamer in the Thar Desert. In percentage terms, it varies from 32% of the tota annua ET 0 in Sawaimadhupur to 49.3% in Anupgarh (GOR, 1992). Among the ten ocations seected aong the Narmada basin in Madhya Pradesh, the vaues range from 429 mm to 600 mm, as a percentage of tota ET 0 ranging from 31.3% in Betu to 35% in Manda (source: GOMP, 1972). Soi infitration capacity can be a imiting factor for recharge. In sandy and sandy oam sois, the infitration capacity of the recharge area can be sustained through continuous remova of sois. But cayey sois have inherent imitation. Resuts obtained from short-term infitration tests carried out in dug wes in Andhra Pradesh in two different soi conditions show that the infitration rate becomes negigibe (< 0.60 mm hr 1 ) within 10 minutes of starting the test in the case of sity cay, whereas infitration stabiises at a rate of mm hr 1 within the first 25 minutes in the case of sandy oam (NGRI, 2000). If the infitration rate approaches to zero fast, it wi negativey affect the recharge efficiency of percoation ponds. As thin soi cover has ow infitration potentia (Muraidharan and Athawae, 1998), the extent of the probem woud be arger in hard rock areas (idea for percoation ponds) with thin soi cover. Based on severa infitration studies, Dickenson (1994) shows that the rate of infitration decines to a minimum vaue within 4-5 days of ponding. This wi aso have adverse effects on the performance of structures buit in areas experiencing fash foods and high evaporation rates, soutions for which woud be wetting or drying of pond beds through reguation of infows. For artificia recharge, the storage potentia of the aquifer is extremey important. The storage potentia of an aquifer vis-à-vis the additiona recharge is determined by the geoogica formation characteristics, and the ikey depth of the dewatered zone. In hiy catchments, the area avaiabe for cutivation is generay very ow, keeping agricutura water demand ow. At the same time, the surface water potentia avaiabe for harvesting is generay high due to high rainfa and runoff coefficients. On the contrary, towards the vaeys and pains, the area avaiabe for cutivation increases, raising agricutura water demand, yet the surface water potentia for harnessing is generay ow due to the ower rainfa, and ow runoff coefficients owing to gente sopes, high PET and deeper soi profies. Limitations imposed by hydroogica regimes Loca water management interventions are often based on very itte understanding of the oca hydroogica regimes, which govern the potentia suppies of water for harvesting. There is a deep-rooted beief that the greater the size of the water impounding structure, the higher wi be the hydroogica benefit in terms of water storage and recharge. Part of the reason for this misunderstanding is the ack of data on streamfows for sma rainwater catchments. Whie runoff harvesting is most suited to areas with high runoff catchment area to run on area ratio (Lajee and Facknath, 1999), this is aso ignored. The higher the aridity, the arger woud be the required catchment area to the cropped area for the same water yied (Prinz, 2002). Often, encroachment of catchments of water harvesting systems for crop cutivation is very rampant, reducing the runoff prospects. The states which have taken up rainwater harvesting and groundwater recharge programmes on a arge scae are Gujarat, Rajasthan, Maharashtra, Tami Nadu, Karnataka, Andhra Pradesh (AP), Madhya Pradesh (MP), Orissa and Chhattisgarh. A major part of these regions is covered by six water-scarce river basin systems, namey: Sabarmati, rivers of Kachchh and Saurashtra, Pennar, Cauvery, eastfowing rivers between Mahanadi and Godavari, eastfowing rivers between Pennar and Kanyakumari, which have ess than 1000 m 3 of renewabe water per annum (Gupta, 2000: pp 116). Now et us ook at the hydroogica regime existing in these states. The percentage area of each state faing under different rainfa regimes (<300 mm, mm, mm, mm, mm and >2500 mm); and different PE regimes (<1500 mm, mm, mm and >3500 mm) has been anaysed (data from Pisharoty, 1990). It is assumed that regions with reativey ow rainfa wi have higher potentia evapotranspiration due to reativey ow humidity and a higher number of sunny days (Pisharoty, 1990). Lower rainfa, couped with higher PE, reduces the runoff potentia and high evaporation from the impounded runoff, thereby increasing the dryness (Hurd et a., 1999). The anaysis shows that Gujarat and Rajasthan have 11% and 42% area, respectivey, faing under extremey ow rainfa (< 300 mm); and 39% and 32%, respectivey under ow rainfa ( mm). In the case of Maharashtra, MP, AP, Karnataka and Tami Nadu, a ion s share (85% and above) fas in the medium rainfa regime (see Figure 1). As regards PE, most of Gujarat and Rajasthan have high evaporation ( mm), as does neary 35 56% of the geographica area of the other states (except Orissa and Chhattisgarh) with 38 65% faing in the medium evaporation regime ( mm). Orissa and Chhattisgarh fa entirey in the medium evaporation regime. Overa, a arge section of the area (of the nine states considered) has medium rainfa, and medium to high evaporation (see Figure 2). We then anaysed the proportion of the geographica area from each of these states faing under different rainfa variabiity casses such as >25%, 25 30%, 30 40%, 40 50% and >50%, and the different percentage of PE occurring during monsoon. Anayses showed that more than 50% of (

4 4 Figure 1. Average mean annua rainfa: prepared on GIS patform based on data from Pisharoty (1990) Figure 2. Average annua evaporation: prepared on GIS patform based on data from Pisharoty (1990) (

5 5 Figure 3. Average coefficient of varaition of rainfa: prepared on GIS patform based on data from Pisharoty (1990) the tota geographica area of a the states put together experience medium variabiity, neary 25% experience high to very high variabiity and neary 20% experience ow variabiity in rainfa (see Figure 3). They coincide with medium rainfa medium to high evaporation, ow rainfa very high evaporation and high rainfa medium evaporation regimes, respectivey. Synthesising readings from Figures 1, 2 and 3, regions with high rainfa variabiity coincide with those with ow magnitudes of rainfa and high PE, which aso have a high dryness ratio. In such areas, a sight variation in precipitation or PE can substantiay change the water stress on bioogica systems as compared with humid regions (Hurd et a., 1999). The higher the variabiity in rainfa, the ower woud be the dependabiity of oca water harvesting/ recharge systems. This is because the chances of occurrence of ow rainfa and extremey ow runoff woud be higher under such circumstances, whie the demand for water woud be high due to environmenta stress caused by poor soi moisture storage, ow runoff and high temperature. In the third step, we anaysed the average number of rain days and its variabiity across regions. We attempted to find the percentage of geographica area of each region faing under different rain day regimes (say <20, 20 30, 30 40, 40 50, and >75 days). We aso anaysed the impications for the magnitude of rainfa in each rainfa event and the maximum and minimum daiy rainfa under different rainfa regimes. The anaysis shows that neary 21% of Gujarat and 45% of Rajasthan state receive ess than 20 days, neary 51% of Gujarat and 70% of Rajasthan fa in areas which experience ess than 30 days and neary one-third of both states receive days of rain a year. As regards the other five states, the area which receives rain days ranges from 9 to 27%; days of rain ranges from 29 39%; days of rain ranges from 27 58%. The Western Ghat in Maharashtra and Karnataka receive heavy rains spread over many days (>75). Both Orissa and Chattisgarh receive days of rain in a year. Synthesising the resuts of the spatia anaysis of rainfa, PE, rainfa variabiity and number of rain days that are provided in Figures 1-4, the foowing trends can be estabished: the inter-annua variabiity in rainfa increases with reducing rainfa; the number of wet spes reduces with owering magnitude of rainfa; the PE increases with owering magnitude of rainfa. The impications of this trend on the potentia for water harvesting in ow rainfa regions woud be: (1) the runoff potentia by and arge woud be ow due to a high dryness ratio; (2) evaporation from surface storage woud be high due to high PE; and (3) the probabiity of occurrence woud be ow. Limitations imposed by high water demands Water harvesting arguments totay miss out on the water demand avaiabiity perspective at the micro eve. Ideay, RWHS woud work if an area which has uncommitted fows to harness has an un-met demand or vice versa. This is unike arge water resource systems where provisions exist for transfer of water from areas of surpus to deficit areas. The water demand of an area is determined by the agrocimate and existing socioeconomic system which, in fact, gets adjusted by the natura resource environment of the viage, the technoogies avaiabe for accessing them and (

6 6 Figure 4. Average rain days the institutiona and poicy environments over a period of time. Regions which were heaviy into irrigated agricuture in the past, supported by good water endowments, institutiona support and favourabe poicies, might continue to demand arge quantities of water for irrigation even when they run out of water. This is because communities take some time to devise coping and adaptive strategies to manage conditions of water deficits. Studies in a viage in Mandvi tauka of Kachchh, which is one of the most arid districts in India, showed that the annua water withdrawa from aquifers for irrigating crops is MCM. The entire water requirements in the viage were being met by groundwater, which was experiencing severe over-draft conditions (Kumar, 1997). Tota rainwater faing over the viage is neary MCM (source: based on data provided in Kumar, 1997, on geographica area and the mean annua rainfa of Kachchh). With a surface water potentia of MCM km 2 (IRMA/UNICEF, 2001), the voume of runoff that woud be avaiabe for repenishment through natura and artificia recharge from within the viage is ony 0.40 MCM. The runoff is, therefore, a sma fraction of the tota consumptive use. This means that the viage has to depend on exogenous sources of water to keep water use sustainabe. This is representative of amost the entire region of peninsuar India, excuding Keraa, centra India and western India. In the viage of Manund, in the Patan district of north Gujarat, which has seen widespread pond de-siting, the tota groundwater abstraction for agricuture aone was estimated to be 3.78 MCM (or 275 mm), with 35 deep tube wes pumping water at a rate of neary gaons per hour for neary 1500 hours a year (Kumar, 2000b). The groundwater condition of the viage is typica of the north Gujarat region. Against this, the tota amount of rainfa over the viage is ony 7.56 MCM, with a mean annua rainfa of 550 mm over an area of 1374 ha. The runoff which this amount of rainfa can generate is 63.8 mm as per the rainfa runoff reationship, with tota runoff being MCM. But, in practice, it is unikey to achieve this as farmers harness for crop production a significant chunk of the runoff generated in situ within the catchment, unike arge basins where a good part is under virgin catchments. Kumar (2000b) estimated the groundwater over-draft in the viage as being neary mm by considering the recharge as 5% of the annua rainfa. Hence, even if the entire runoff generated is harnessed for recharge, it woud amount to ony 25.7% of the over-draft. On the other hand, there are many regions in India where the economic demand for water is far beow what the natura endowment can provide, such as the entire Ganga- Brahmaputra basin. This region has an enormous amount of static groundwater, estimated to be BCM, apart from having high rainfa and a cod, sub-humid cimate that generates sufficient surface fows. Cheaper access to water might increase the demand for irrigation water sighty but the cod and humid cimate and very ow per capita arabe and impose imits (Shah 2001; Kumar, 2003). Irrigation intensities are aready high in Uttar Pradesh and Haryana. Though irrigation intensity in Bihar is ow, the sub-humid and cod cimate reduces the irrigation requirement significanty. In most of this region, the issue is not one of the physica avaiabiity of water, but the abiity of communities to access it for irrigation (Kumar, 2003; Shah, 2001); water harvesting does not offer any economic (

7 7 soution here for the poorer communities to access water. Issues in evauating costs and economics In the panning of arge water resource systems, cost and economics are important considerations in evauating different options. But unfortunatey, the same does not seem to be appicabe in the case of sma systems, though concerns about economics of recharge systems in certain situations were raised by authors such as Phadtare (1988) and Kumar (2004). Part of the reason for ack of emphasis on cost is the ack of scientific understanding of the hydroogica aspects of sma scae interventions, such as the voume of streamfow avaiabe at the point of impoundment, its pattern, the amount that coud be impounded or recharged and the infuence area of the recharge system. Even though simuation modes are avaiabe for anaysing catchment hydroogy, there are great difficuties in generating the vita data at the micro eve on daiy rainfa, soi infitration rates, catchment sopes, and cover and PET which determine the potentia infows; and evaporation rates that determine the potentia outfows. Further, for sma water harvesting projects, impemented by oca agencies and NGOs with sma budgets, the cost of hydroogica investigations and panning is hard to justify. Often, provision for such items is not made in sma water harvesting projects. That said, the amount of runoff which a water harvesting structure coud capture depends not ony on tota runoff voume but aso on how it occurs. A tota annua runoff of 20 cm occurring over a catchment of one km 2 can generate a surface fow of 0.20 MCM. But the amount that coud be captured depends on the pattern. As Garg (1987) points out, in arid and semi-arid regions in India, high intensity rainfas of short duration are quite common (source: Garg, 1987 as cited in Athawae, 2003: Figure 24). These runoffs generate fash foods 2. If the entire runoff occurs during a major rainfa event, the runoff coection efficiency woud fa with the reducing capacity of the structures buit. However, if structures are buit arge enough to capture the high intensity runoff, thereby increasing the runoff coection efficiency, that woud mean infating the cost per unit voume of water captured. In fact, authors such as Oweis et a. (1999) have argued that runoff harvesting shoud be encouraged in arid areas ony if the harvested water is directy diverted to the crops for use. Even given data on infows and runoff coection efficiencies, predicting the impacts on the oca hydroogica regime is aso extremey compex, requiring accurate data on geoogica and geo-hydroogica profies, and variabes. In ieu of these difficuties in assessing the effective storage, unit costs are worked out on the basis of the design storage capacity of the structures and rue of thumb guestimates on the number of fiings. The Shri Vivekananda Research and Training Institute, Mandvi, Kachchh, which has done pioneering work on artificia groundwater recharge in India, often resorts to this strategy to evauate the cost effectiveness of recharge structures buit in Kachchh (e.g. Raju, 1995). The recent book by Dr. R.N. Athawae on rainwater harvesting in India, athough it covers a range of technica aspects of water harvesting in different regions of India, does not dea with economic issues (Athawae, 2003). Scae considerations are extremey important in evauating the cost and economics of water harvesting/ groundwater recharge structures because of the hydroogica integration of catchments at the eve of watershed and river basins. The cost and economics of water harvesting systems cannot be performed for individua systems in isoation when the amount of surpus water avaiabe in a basin is imited. This is because incrementa structures do not resut in proportiona increase in the hydroogica benefits (Kumar, 2000a), as interventions in the upper catchments reduce the potentia hydroogica benefits from the ower systems. What is important is the incrementa hydroogica benefits due to the new structure. A system in itsef may be costeffective and economicay viabe if evauated independenty but, if evauated as a part of a arge-scae water-harvesting intervention at the eve of river basins, the system may not be justifiabe on costs aone when compared with the additiona benefit it brings in. In any basin, the margina benefit from a new water harvesting structure woud be smaer at higher degrees of basin deveopment, whie the margina cost is higher (see Figure 5). The reasons for this are: (1) the higher the degree of basin deveopment, the ower the chances of getting sociay and economicay viabe sites for buiding water impounding structures, and (2) with a higher degree of deveopment, the socia and environmenta costs of harvesting every unit of water increases (Frederick, 1993), reducing the net economic vaue of benefits. Therefore, the cost and economic evauation shoud move from watershed to basin eve. As Figure 5 indicates, the eve at which basin deveopment can be carried out depends on whether we consider the fows in a wet year, a dry year or a norma year. Nevertheess, there is a stage of deveopment (marked by O in the chart) beyond which the negative socia, economic and environmenta benefits start accruing, reducing the overa benefits. Here, O is the optimum eve of water resource deveopment. However, it is important to keep in mind that the negative socia and environmenta effects of over- Margina Cost/ benefit Margina benefit (Socia, Environmenta and Economic) Dry Year Wet Year O O Wet Year 2 Many parts of Kachchh, which records one of the owest mean annua rainfas (350 mm), experienced foods during 1992 and 2003 with many WH structures overfowing. Fash foods occur even in some of the semiarid and water scarce basins such as Sabarmati and Banas (Kumar, 2002b). Figure 5. Degree of Water Deveopment Margina costs and benefits of water harvesting with different degrees of basin deveopment (

8 8 appropriation of a basin s water resources may be borne by a community iving in one part of the basin whie the benefits are accrued to a community iving in another part. Ideay, water deveopment projects in a basin shoud meet the needs and interests of different stakehoders iving in different parts. Therefore, the optimum eve of water deveopment shoud not aim at maximising the net basin eve benefits, but rather at optimising the net hydroogica and socio-economic benefits for different stakehoders and communities across the basin. That said, in certain situations, the oca economic benefits from RWH against the economic costs themseves may be questionabe. But such interventions coud be justified if there are potentia socia benefits in changing patterns of water avaiabiity and use, in terms of increasing water avaiabiity to poorer farmers with ow capabiity and hodings: such decisions shoud be based on evauation of aternative strategies to meet the oca water needs of the poor. The abiity to derive economic benefit from recharge depends on where the recharged water ends up. In regions underain by hard rock geoogy, the groundwater fow patterns are quite compex. Often, the benefits of recharge structures extend up to a few kiometres downstream or upstream, depending on the pattern of occurrence of geoogica structures such as ineaments, fractures and dykes (source: based on Muraidharan and Athawae, 1998). Tracing the recharge water in such situations woud require sophisticated studies invoving isotopes. This is a common probem in the hard rock areas of Saurashtra, Kachchh, north Karnataka and Tami Nadu where arge-scae water harvesting/groundwater recharge interventions are taken up through check dams, ponds and percoation tanks. Often the communities, for whom investments in the recharge system are made, do not get the benefit (Moench and Kumar, 1993). In certain other situations, the recharge water coud end up in saine aquifers. The economics of RWH wi aso be a function of the incrementa vaue of benefits accrued from the use of newy-added water. Apart from the recharge voume, the vaue of the use to which the additiona water is put is extremey important in determining the incrementa benefits, an issue often ignored in the project panning. Often, the benefits of RWHS are not ceary identified or understood. Whie the cost of water harvesting is significant, it is critica to divert the new water to high-vaued uses. Phadtare (1988) pointed out that recharge projects woud be economicay viabe in auvia north Gujarat if the water is diverted for irrigation, as structures are expensive. Yied osses due to moisture stress are extremey high in arid and semi-arid regions and that providing a few protective irrigations coud enhance yied and water productivity of rain-fed crops remarkaby, especiay during drought years (Rockström, 2002). The avaiabe extra water harvested from monsoon rains shoud therefore be diverted to suppementary irrigation in drought years. There are regions where potabe water for peope and ivestock becomes a high priority demand. North-western Rajasthan, which is arid and dominated by pastora communities, named Gujjars, is one such exampe. The socia and economic vaue reaised from the use of water for human drinking and ivestock use, respectivey, woud be much more than the economic vaue reaised from its use in irrigating crops. In such situations, water shoud be diverted for such uses where the opportunity costs are ow and net vaue products are high. But proper water use panning to reaise maximum vaue from the added water is argey missing in water harvesting efforts. Lack of integrated approach In many river basins, surface water and groundwater systems are often inter-connected. Any aterations made in either one coud change the avaiabiity of water in the other (Sohiquio, 1985; Lamas, 2000). In many hiy areas, especiay in the Western Ghats, the water eves rise steepy after monsoon, and groundwater contributes significanty to the streamfows downstream during ean seasons due to the steep groundwater fow gradients. In that case, any water harvesting intervention to store water underground may not make much sense as it woud get rejected and appear as surface fows (Mayya, 2005). On the other hand, in regions with deep water tabe conditions ike in north Gujarat, the runoff moves directy into the groundwater systems of the pains through the sandy river bed as dewatering of the upper aquifers increases the rate and cumuative percoation (Kumar, 2002b). With two-thirds of the country s geographica area underain by hard rock formations, the storage capacity of aquifers poses a major chaenge for artificia recharge. Most parts of the water-scarce states, namey Gujarat, Madhya Pradesh, Maharashtra, Karnataka, Andhra Pradesh, Orissa, Chhattisgarh and Tami Nadu, are underain by hard rocks ranging through basat, crystaine granite hi aquifers and sandstone. A sma area in Gujarat has extensive auvium, i.e. the Narmada vaey and (Cambay basin) (see Figure 6). The hard rock aquifers have no primary porosity and ony secondary porosity. The constraints imposed by hard rock geoogy in recharge efforts through percoation tanks are: high depth to water tabe beow and around the recharge structure due to the occurrence of the recharge pume and shaow bed rocks, which prevent percoation of water (Muraidharan, 1990 as cited in Muraidharan and Athawae, 1998) and the ow infitration capacity of the thin sois overying the hard rock formations. Due to ow specific yied ( ), a sharp rise in water eves is observed in aquifers during monsoon, eaving itte space for infitration from structures. Whie harnessing water for recharge is extremey important during norma and wet years, the natura recharge in hard rock formation is high during such years as it is a function of seasona rainfa (based on regression equations shown in Figure 7 in Athawae, 2003), further reducing the scope for artificia recharge. In Saurashtra, in spite of the poor potentia offered by ow rainfas, high variabiity and high evaporation rates (see Figures 1 3), significant recharge efforts were made. Nevertheess, the biggest constraint in storing water underground during high rainfa years is the poor storage capacity or specific yied of the basat formations. During good rainfa years, the aquifers become saturated with natura recharge immediatey after the rains, eaving no space for entry of water from the recharge systems (Kumar, 2000a). An estimated check dams buit in the region to capture the rainwater and recharge the aquifers are abe to store ony a sma fraction of the surpus runoff. In such situations, proper water use programming is required to use the surpus water effectivey, whereby water from aquifers is pumped out and used during the rainy season itsef thus creating storage space for the incoming fows (Muraidharan and Athawae, 1990; Shah, 2002). (

9 9 Figure 6. Aquifer system in India Trade-off between oca vs. basin impacts in cosed basins Lack of integration between pans for oca water harvesting and basin water resource deveopment, means that RWH often eads to over-appropriation of surface water in river basins. Whie panning of conventiona water deveopment projects is based on dependabe yieds from the catchments, the subsequent pans for WH do not take into account the committed fows for downstream reservoir/water diversion systems. Aso, there is an increasing tendency to beieve that because these structures are so sma they are benign (Batcheor et a., 2002), even though they are present in arge numbers in most cases. The primary reason for this is that the agencies concerned with sma water harvesting in the upper catchment and those concerned with major headworks are different and they do not coordinate their actions at the basin eve. Buiding of tanks and check dams is often the responsibiity of minor parts of an irrigation department or district arms of the rura deveopment departments. This ad hoc approach to panning often eads to over-appropriation of the basin water, with negative consequences for arge schemes downstream (Kumar et a., 2000). Many arge and important river basins in India, which are aso facing water scarcity, are now cosed or do not have uncommitted fows that are utiisabe through conventiona engineering interventions. Exampes of these are Pennar, Cauvery and Vaigai in the south (based on GOI 1999: pp ), and Sabarmati, Banas in the west, which are cosed. In addition to these, a the west-fowing rivers in Saurashtra and Kachchh in Gujarat are aso cosed (Kumar, 2002). Whie Krishna basin is on the verge of cosure, some of the basins which are sti open are Godhavari and Mahanadi in the east (based on GOI, 1999: pp ). Sabarmati basin, for instance, having a drainage area of km 2, has a utiisabe surface fow of MCM aocated to Gujarat (Kumar and Singh, 2001), whereas the tota ive storage capacity of irrigation schemes buit in the basin, estimated to be 1470 MCM (GOI, 1999) is sti sighty beow this. But the basin has many water diversion structures, incuding weirs and a barrage. Indeed, the dependabe runoff upstream of the reservoirs/diversion structures in the basin is far beow the panned water utiisation (estimated to be 1560 MCM as per Kumar and Singh, 2001) eaving no spi-over. At the aggregate eve, the basin is over-appropriated. At the sub-basin eve, the scenario is different. Two of the sub-basins, Dharoi and Hathmati, are heaviy over-appropriated (Kumar et a., 2000) whie one of the sub-basins, Watrak, has uncommitted fows (Kumar and Singh, 2001), which eventuay end up in the Guf of Cambay. It is hard to judge whether a basin is cosed or open on the basis of the storage capacity of reservoirs and the dependabe fows, as many reservoirs aso divert a ot of water during the monsoon season, making the effective water utiisation more than the ive storage capacity. Take, for instance, the Narmada basin. The tota ive storage voume of the termina dam, Sardar Sarovar, is 5800 MCM, whereas the tota water utiisation from this reservoir is MCM. A the 30 arge and 135 medium reservoirs together woud divert a tota of MCM of water for irrigation and various other purposes (NWDA, 2004) but the tota ive storage of these reservoirs woud be much ess, (

10 10 i.e MCM (GOI, 1999: pp 36). This is because a significant amount of water woud be diverted from these reservoirs for kharif irrigation within and outside the basin, particuary from the Sardar Sarovar reservoir. Again, the estimates do not take into account the reservoirs having ive storage capacity of ess than 10 MCM. Trade-off between economics and hydroogica opportunity Regions with semi-arid and arid cimate experience extreme hydroogica events (Hurd et a., 1999). As we have seen, high inter-annua variabiity in rainfa is a common phenomenon in most parts of these water-scarce regions. Rainfa variabiity induces more variabiity in runoff, even in high rainfa as we as ow rainfa regions. We take as an exampe the upper catchment area of Cauvery basin in peninsuar India and one of the catchments of the Sabarmati River basin in north Gujarat of western India. In the Paanpur area of Banaskantha district in north Gujarat, which has semi-arid to arid cimatic conditions, the rainfa records vary from a ow of 56 mm in 1987 to 1584 mm recorded in The runoff estimated on the basis of a regression equation deveoped for Hathmati, a sub-basin of the Sabarmati in north Gujarat, physiographicay quite simiar to the Paanpur area of Banaskantha, shows that the runoff can vary from 0.6 mm to 541 mm. But actua runoff coud be different from this, depending on how other variabes that are not considered in the regression, such as the intensity and pattern (over space and time) of rainfa, infuence the runoff intensity. Thus the owest runoff is cose to 1/1000 th of the highest runoff. Even though what can occur at the sub-basin eve may not be representative of that in sma upper catchments, the difference cannot be drastic. Even for a humid, high rainfa region such as the Wayanad district in Keraa, the runoff estimated for the sma catchment of Karappuzha, on the basis of the rainfa runoff reationship deveoped for the Kabani sub-basin (catchment area of 7040 km 2 ) within the Cauvery river basin, and the observed rainfa of the area, ranges from 528 mm in the owest rainfa year (2002) to 1458 mm in the highest rainfa year (1994) in a 31-year period from Maximizing oca benefits vs. optimum benefits for basin communities Generay, in any river basin, the upper catchments are rich in terms of their abiity to contribute to the basin yieds. This is mainy because of the unique physiographica features, and party because of the cimatic conditions such as steep sopes, high rainfa in the mountains and high humidity which provide a favourabe environment for runoff generation. The upper catchments aso provide a good source of base fows due to forest cover which causes favourabe conditions for water storage and infitration. On the demand side, these regions generay are ess we endowed in terms of avaiabiity of arabe and and consequenty the demand rates for irrigation are generay ow. On the other hand, the ower catchments are generay characterised by ower rainfas and higher eves of aridity (rainfa deficit to meet ET demands) and the better access to arabe and increases the aggregate demand for irrigation. There are numerous exampes for this. A few to cite are: the upper catchment of Cauvery basin in the south, the Narmada basins in centra India, the Sabarmati basin in western India, tributaries of the Indus in the north-western India, the Krishna basin in centra India and the Mahanadi basin in eastern India. Some parts of the Kabani sub-basin of the Cauvery river basin have a cod and semi-humid cimate, and parts of this sub-basin receive the second highest rainfa in India after Chirapunji, with mean annua rainfa over 4000 mm. We have defined the agricutura water demand as a function of per capita net sown area and the ratio of ET 0 and rainfa; and water avaiabiity as a function of rainfa. It is assumed that the higher the ET 0 /R ratio, the higher woud be the irrigation requirement for a unit of and; the higher the per capita (rura popuation) net sown area, the higher woud be the aggregate demand for irrigation per capita. Tabe 1 shows the estimated vaues of two agricutura water demand variabes, ET 0 /R and per capita arabe and; and one water avaiabiity variabe, i.e. rainfa. Major water resource/irrigation projects undertaken in the past tap streamfows generated from the upper catchments, but cater to either the ower parts of these Tabe 1. basins Comparison of agricutura water demand variabes in upper and ower catchment districts of seected Indian river Name of Name of Name of Mean annua Mean annua ET 0 /R Per capita Basin UCD LCD rainfa (mm) PET (mm) net sown area (ha) UCD LCD UCD LCD UCD LCD UCD LCD Sabarmati Dungarpur Ahmedabad Indus Shima Ludhiana Narmada Shahdo Jhabua Cauvery Wayanad Nagapattianan Krishna Raigarh Guntur NA NA NA Mahanadi Raipur Puri UCD: Upper catchment district LCD: Lower Catchment District Source: authors own estimates based on Agricutura Statistics of India and FAO data on precipitation (R) and reference evapotranspiration (ET 0 ) (

11 11 basins or other ess water endowed regions outside these basins (Verghese, 2001 and 2002). Bakhra reservoir and Nanga diversion projects ocated in the high rainfa Shivaik his of Himacha Pradesh essentiay cater to the ravenous ow rainfa and drought prone regions of Punjab and sparse rainfa regions of Rajasthan (Verghese, 2002); the Sardar Sarovar dam harnesses water from ampe rainfa areas in Narmada vaey and takes it to the drought-prone areas of north Gujarat and Saurashtra which are characterised by ow and erratic rainfa (Verghese, 2001). Simiary, the arge reservoir projects in Cauvery transfer water to the drought-prone regions in Tami Nadu and Karnataka. As such, the water demand for irrigation is extremey ow in the upper catchments. Moreover, as irrigation water use efficiency and water productivity are ikey to be high in areas with variabiity in rainfa and high drought-proneness (Rockström, 2002), with transfer of water from the we-endowed regions to the poory-endowed regions, the economic vaue of water in agricuture increases. The recent research carried out by IWMI in water-scarce and and-rich western Punjab and water-rich and and-scarce eastern Uttar Pradesh showed that the vaue of water reaised from irrigation is much higher in Punjab (Rs /m 3) than in eastern UP (Rs. 11/ m 3 ). Because of water scarcity, the farmers in Punjab make better use of water by choosing cropping systems that are economicay more efficient and using agronomic practices to obtain higher yieds, higher physica productivity and greater economic efficiency (Kumar, Maa and Tripathy, 2006). Often water harvesting initiatives, especiay those by NGOs, are driven by considerations other than economic efficiency, the most important of which are socia equity and environmenta justice. For exampe, impounding water in the upper catchments might serve socia objectives of meeting drinking water requirements. As is evident from the above iustrations, there is a cear trade-off between meeting economic efficiency objectives and these deveopmenta goas. Therefore, any water resource intervention in the upper catchment areas which reduces the downstream uses shoud be done with due consideration of the net change in gross vaue product of water in the basin. The gross vaue product can be defined as the sum tota of the incrementa vaue product from the economic uses, environmenta services and socia uses the basin s water resources meet. The amount of water to be captured upstream through RWH interventions shoud aso be optimised to derive maximum regiona socia equity, environmenta vaue and overa output from the economic uses of water. In basins where the avaiabe water resources are aready committed (cosed basins), the chaenge is greater since maximising the gross vaue product might mean reaocating some water from one ow vaued use to a high vaued use. Critica issues for research on RWH in India Issues emerging from past research There have been severa research studies which attempted to anayse the oca and regiona impacts of oca water harvesting/recharge schemes. The research papers deat with the foowing key questions: the physica performance of recharge structures; how much do the recharge structures actuay contribute to groundwater avaiabiity in a region? what is the socio-economic impact of increased water avaiabiity? does water harvesting hep ater the water baance from a river basin perspective? what are the second generation issues in water harvesting? is water harvesting an eixir for farmers, or are they engaged in a process of ecoogica destruction? what are the unintended impacts of rainwater harvesting, particuary on the water use hydroogy of sma catchments? Pate (2002) evauated the various hydroogica and hydrochemica aspects of recharge systems such as percoation tanks, check dams and dug we recharge in three different geoogica settings, namey, miioite imestone, gaj imestone and weathered basat rock. The different hydroogica aspects of artificia recharge system studies were deveopment and decay of the recharge pume, recharge rates and radius of infuence of recharge structures. The hydrochemica aspects were changes in tota dissoved soids and fuoride content of groundwater. The study invoved actua measurements of some parameters governing the physica impacts of recharge structures. The study found that the rate of deveopment and decay of recharge varied according to variations in geoogica settings. It aso estabished that the recharge rates were far higher in the case of percoation tanks and check dams which were periodicay de-sited than those not de-sited. The recharge rate estimated for a norma percoation tank was 7.87 mm day 1 whie that for a de-sited percoation tank was 20.4 mm day 1. Accordingy, the rechargeevaporation ratio was found to be much higher for the desited percoation tank (4.2) against 1.83 for the norma percoation tank. Further, the radius of infuence of percoation tank was found to vary across geoogica formations. Though the study is about the oca hydroogica impacts of recharge schemes, it supports the argument made in this paper that it may not be appropriate to use thumb rues to assess the size of benefits from recharge structures. Paanisamy and others evauated the economic impact of ten percoation tanks from Coimbatore and Avinashi districts of Tami Nadu. It was found that ony 14% of the wes in the vicinity were benefited by the tanks, with a tota area of 14.4 ha and average additiona income at tank catchment ranging from Rs.1323 ha 1 to Rs.2736 ha 1. The anaysis did not invove the cost of the tank structures and was based on one year of data. The study attributed the poor economic performance of the tanks to inadequate rainfa in that particuar year and improper tank ocation (Paanisamy and Kandaswamy (1990) as cited in Muraidharan and Athawae, 1998). These findings corroborate to a great extent the arguments made earier in this paper, with regard to imited physica impact of RWH structures and uncertain benefits. Badiger et a. (2002) made a quick evauation of the variety of physica and socioeconomic impacts of the pa systems buit by PRADAN in northeastern Rajasthan on the basis of studies carried out in four micro catchments. These catchments fe within the arge basin of Mewan in Mewat region of Awar district. The pa project of PRADAN used (