Solar Pumps and Water-Energy Nexus in Gujarat, India: First Assessment of the World’s Largest Pilot on Grid-connected Solar Irrigation Pumps

The epicentre of Gujarat’s perverse nexus between electricity subsidy and groundwater depletion lies in its legacy of 485,000 unmetered tubewell owners who have ercely resisted metering for 20 years. These comprise 40 percent of Gujarat’s irrigation connections but account for 49 percent of agricultural load, 71 percent of energy use in groundwater abstraction and 90 percent of farm power subsidies. Suryashakti Kisan Yojana (SKY), a large, state-wide pilot project which solarised and net-metered 4215 tubewells and began paying farmers for evacuating surplus solar energy, has been enthusiastically embraced by unmetered and metered farmers alike. SKY promises politically acceptable resolution of a wicked energy-water conundrum that aicts much of India and west Asia. Mainstreaming SKY can signicantly reduce groundwater overdraft, GHG emissions and subsidy burden. It will increase farm incomes while radically improving energy-water accounting and management. Gujarat government should invest in compulsory, free-of-cost solarisation of tubewells. We show that doing so is protable for government and benecial for farmers, climate and environment.

larger than other known projects anywhere of grid-connected solar pumps. Similar pilots in India --SuryaRaitha in Karnataka, Vinziyanagaram feeder in Andhra Pradesh and a pilot feeder in Jaipur, Rajasthan--are all single feeder pilots with less than 350 pumps each (Shirisath et al 2020). This paper presents the rst ever assessment of SKY using granular data provided by real time, online SKY monitoring system. Section 2 situates SKY in the history of electricity-groundwater nexus in Gujarat. Section 3 tests key hypotheses about SKY impacts and shows how the SKY-model, if scaled out, can radically improve nances of DISCOMs and socio-ecology of groundwater irrigation. Section 4 shows why Government of India's PMKUSUM-C scheme, styled after SKY, will fail to take SKY to scale. Section 5 explores a way out in compulsory, free-of-cost solarisation of all tubewells. Section 6 concludes the paper.

The Hard Problem of Tubewell Metering
SKY's prime goal was to test whether farmers would willingly embrace a positive-sum model of tubewell solarisation that would bene t farmers and also liberate Gujarat from a perverse subsidy that has drained its sc and strained its aquifers (Shah et al 2004). Wells have played a seminal role in Gujarat's irrigation history (Shah 2009). Informal groundwater markets boomed in Gujarat villages during early 1900's when enterprising farmers invested in oil engines and buried cement pipelines more to sell irrigation service than water own crops (Shah 1993;Hardiman 1998). Spread of electricity and tubewell technology after 1950's gave a llip to water markets. Electric pumps were then metered and electricity tariff drove water price. But metering technology was primitive, costly, laborious and prone to corruption by meter readers against whom farmers frequently agitated. Utilities too loathed the high cost and logistical di culty of metering geographically dispersed tubewells (Shah 1993). During 1980's, many states removed meters and switched to horse-power (HP) linked at tariff. Pressure mounted to do so in Gujarat too which until 1988 recovered US $ 9/MWh[7] from metered tubewell owners. Studies in Gujarat also showed that at tariff would transform Gujarat's 142,000 private electric tubewell owners (Joshi & Acharya 2005) from 'water lords' into owner-managers of quasi-public utilities offering to the poor affordable irrigation that government's 3500 public tubewells had failed to do (Shah 1989;Shah 1993). In 1988, Gujarat also abandoned meters and switched to at tariff. True to the studies, surveys showed that tubewells now worked longer hours and water buyers were paying 30-50 percent less for irrigation than before (Shah 1993). Planners had assumed periodic upward revision in at tariff to recover the cost of rising average power consumption/HP. But this was to prove easier said than done.
Flat tariffs soon proved a Faustian bargain for Gujarat as for much of India. Throughout 1990's, tubewell numbers experienced runaway increase as did power consumption/tubewell. Water sellers made hay; buyers too enjoyed lower prices and better service until groundwater began running out. Without proactive regulation of tubewell numbers and pumping hours, Gujarat got sucked into a perverse nexus in which power subsidies drove groundwater depletion which, in turn, hardened demand for more subsidy, as described in gure 1. North Gujarat, Saurashtra and Kutch experienced secular decline in groundwater levels, even as subsidies drove electricity utility towards bankruptcy. For a while,Jyotigram [8] scheme of feeder separation and farm power rationing capped subsidies as well as groundwater draft. But a million new tubewell connections issued after 2000 mostly to the poor nulli ed the gains (Shah and Chawdhury 2017).
Early 2000's witnessed a growing national resolve to meter farm power supply, even if subsidized. This was formalised in the National Water Policy (MoWR, GoI 2002), legislated in Electricity Act 2003 and speci cally mandated by GERC [9] standing order in 2004. Unmetered supply was thought to undermine energy accounting and auditing, letting DISCOMs hide ine ciency and theft as agricultural consumption (PRAYAS 2018;Gulati andPahwa 2015, Dubash 2007). But farmers ercely opposed metering. While most Indian states failed to even make a start, Gujarat banned issuing new unmetered connections while subsidising metered supply. Kumar et al (2021) have argued that farmer opposition to metering in India is a red herring but even 90 percent subsidy on metered supply in Gujarat (US $ 8.3/MWh for normal and US $ 11.1 /MWh for on-demand connections) failed to lure unmetered farmers to meter tubewells. A global survey by Molle and Closas (2021) shows that metering groundwater wells is resisted and frustrated by farmers not only in India but also in US, Iran, Australia, Greece, France, Italy, Syria, Jordan, Mexico, Chile and other countries for much the same reasons as in India. The survey recommends even today, as Rajadhyaksha Committee had done back in 1985 (Shah 1993), that "metering [of irrigation wells] should be avoided in the absence of adequate resources." (p.11).
Summing up years of failed metering campaign, India's Comptroller and Auditor General lamented in a 2016 report: "Despite explicit order by the regulator to meter all unmetered connections, there was absolutely no progress in this matter… due to stiff resistance offered by farmers with unmetered connections" (CAG 2016: p 101-102). As a solid vote-bank, unmetered farmers not only successfully opposed metering but also any revision of at tariff. Against GERC-mandated tariff of US $ 33.3/HP/year, these pay even today just US $ 9.2/HP/year, less than US $ 10.4/HP they paid in 1988. To lure them to metering, metered tariff too is kept at US $ 8.33/MWh, less than US $ 9.03/MWh in 1988, when rupee was worth 10.4 times more than today![10] The de cit is made good by GoG to GUVNL [11], the holding company of DISCOMs (rows 7,8,9,10 in table 1). In 2014-15, unmetered farmers accounted for 41 percent of Gujarat's tubewells, but 49 percent of tubewell load, 71.3 percent of farm power consumption and 90 percent of farm power subsidy. An average unmetered farmer is alleged to use 3.5 times the electricity and enjoy 13 times the subsidy claimed by average metered tubewell owner, who himself pays only 10 percent of the cost to serve. In the 1980's, Gujarat had on average 8 electric tubewells per village. Flat tariffs leveraged these to expand irrigation through pro-poor water markets while posing little threat to groundwater sustainability (Shah 1993 methods are likely to fail [24]." Tubewell metering is anathema throughout India and indeed elsewhere. Based on the notion that cost of unmetered supply far outweighs its value to farmers, agencies have proposed schemes[13] of incentivizing farmers to willingly embrace metered power supply (Gulati and Pahwa 2015;USAID, 2011;World Bank 2019). Jury is out on how successful these are in onboarding tubewell owners and changing their pumping behaviour. SKY is a far more ambitious pilot with a component of sophisticated, tamperproof remote metering of tubewells. Did SKY succeed in onboarding unmetered farmers? Is metering changing pumping behaviour? Did farmers gain from onboarding SKY? Did the 'GoG-DISCOM combine' (henceforth, the Utility) gain nancially and otherwise? These are critical questions that the SKY monitoring system is uniquely suited to explore. Feeders selected for SKY had metered and unmetered connections earlier. After SKY, all were brought under tamper-proof, remote metering, generating real time data on four categories of agricultural connections: MetSol: previously metered tubewells whose owners onboarded SKY and are now subject to SKY terms under FFSBS FlatSol: previously unmetered at tariff paying farmers who onboarded SKY and are now subject to FFSBS MetNonsol: previously metered farmers on SKY feeders who did not join SKY and continue to pay subsidized grid tariff on metered consumption FlatNonsol: previously unmetered tubewells on SKY feeders who did not join SKY; these are now metered but continue to pay HP-linked at tariff regardless of metered consumption Our analysis in the remainder of this paper is based on this dataset.

SKY Adoption Rates
For metered farmers, bene ts of SKY are obvious. They are metered anyways; and getting Five Fold Sky Bene t Suit (FFSBS) against contributing just 5 percent of capital cost is a lucrative deal for them. But for unmetered farmers, decision is not so clear cut given that it is irreversible. Is joining SKY worth permanently giving up average subsidy of US $ 2468/year (CAG 2016) for a lifetime? We therefore expected lukewarm response to SKY from unmetered farmers compared to metered ones. But SKY results are surprising ( gure 2) since the onboarding ratio is nearly the same at 66-69 percent for both metered and unmetered tubewell owners.
Did hydrogeology have any in uence on farmer decision to join SKY? To explore this, we need to disaggregate onboarding rates across regions. Figure 3 shows service areas of Gujarat's four DISCOMs, each with distinct hydrogeological pro le. DGVCL [14] serves humid-alluvial south Gujarat amply recharged by rainfall and canals. MGVCL serves semi-arid central and eastern Gujarat with alluvial and hard-rock aquifers recharged by canals and modest rainfall. UGVCL serves arid North-Gujarat with stressed alluvial aquifers, and limited recharge sources while PGVCL serves semi-arid Saurashtra with karst aquifers and arid Kutch's alluvial aquifers. Gujarat's energy-water nexus plays out fully in service areas of PGVCL and UGVCL, which account for 90 percent of its unmetered load (Viswanathan 2014:24) as well as of agricultural power consumption (CAG 2016) (table 2). SKY onboarding ratio matters far more in PGVCL and UGVCL than in DGVCL and MGVCL.   3.2 Impact on power consumption/HP An important but unsettled question around unmetered tubewell owners is: do they really consume signi cantly more power/HP than metered consumers or is their alleged excess consumption a g-leaf to conceal distribution losses, theft and ine ciency (Dubash 2007;PRAYAS 2018 (2016) is metered consumption of metered farmers. This too supports the 'excess consumption' thesis. It predicts mean consumption/HP for metered tubewells to be 471 kWh across DISCOMs; but for unmetered tubewells, it predicts 533 kWh/HP in DGVCL, 939 kWh/HP in MGVCL, 1108 kWh/HP in UGVCL and 1150 kWh/HP in PGVCL. Figure 4 plots log values of consumption/HP against connected load. FlatSol and FlatNonsol tend to hoover towards the top suggesting excess consumption to be the central behavioural tendency of unmetered connections.
One unexpected nding is that average metered consumption/HP for FlatNonsol and MetNonsol on SKY feeders is less than half of CAG (2016)  We expected that under SKY power consumption/HP by FlatSol will move away from FlatNonsol and closer to that of MetSol because FlatSol are now metered and can sell surplus energy to the grid at remunerative price. In actuality, however, FFSBS puts con icting pressure on consumption/HP by SKY tubewell owners: [a] moving from 8 hours of day-and-night power earlier to 12 hours of daytime power every day is expected to exert a strong 'supply push', making possible increased consumption than before; [b] on the other hand, saving power and selling it at US $ 48.6/MWh is expected to create a 'demand pull' for reducing consumption. Given that hours of available power is critical during peak irrigation season, it is noteworthy that SKY adopters are using much less of potentially available power/HP compared to non-adopters (last column). It is also likely that FlatSol are unable to shake off abruptly the overhang of a decades-old water selling tradition and desert an established, loyal clientele, reneging on their commitments. There was also much uncertainty about getting paid for energy until they got the rst check from the Utility at year-end when the 'demand pull' to save energy would kick in. Clearly, there is an adjustment process in motion that will work itself out gradually. In a dozen feeders commissioned in 2018, SKY data available for 2019 and 2020 show (in gure 5) that while consumption/HP of MetSol owners increased by 5 percent, that for FlatSol decreased 28.7 percent.
9 out of 12 feeders showed a decline in consumption for MetSol as well as FlatSol. The 3 feeders where it rose were all in PGVCL which experienced a drought in 2020. We also have, for peak irrigation months of February to April, data for 2019,2020 and 2021 for these early 12 feeders ( gure 6). During 2020, 'supply push' of 12 hours/day of daytime power led all tubewell owners to increase consumption. But after the rst annual payment for energy sales was made in May 2020, MetSol and FlatSol steeply reduced consumption by 26.1 and 34.1 percent respectively. This suggests that FlatSol consumption/HP is tending towards that of MetSol at a pace tempered by an adjustment parameter.  Table 6 suggests that our core SKY hypothesis-that consumption/HP of unmetered tubewells will decline under SKY as they become FlatSol--has stronger support in PGVCL and UGVCL where it matters most and weak support in DGVCL and MGVCL where it matters the least. The decline in consumption/HP of FlatSol farmers is higher in PGVCL (18 percent) and UGVCL (33 percent) which are the locus of Gujarat's energy-water nexus. The decline is marginal in DGVCL (4 percent) and MGVCL (6 percent) whose share in Gujarat's agriculture load is minuscule. If SKY is to be mainstreamed, PGVCL and UGVCL should take the lead.

Farmers' Gain from onboarding SKY
SKY farmers can use the same amount of power/HP they used earlier for irrigation and water selling without paying grid power tariff; and they also have the new option to sell energy saved which they could not do earlier. Farmers' incremental gain from onboarding SKY then is the value of surplus power exported plus grid power tariff saved less payment for net energy import (if any). Of 2997 SKY farmers' clean records we found to estimate incremental gains in the rst year, 2608 (87 percent) exported energy while 389, mostly FlatSol owners in DGVCL, were net importers. These had to pay for energy import but could have earned more from water selling. Figure 7 shows farmers' incremental gains/HP from joining SKY. Overall, MetSol gained more per HP because they pumped less and exported more compared to FlatSol; but some FlatSol might have earned more by selling water which gure 7 fails to capture. Earnings/tubewell from energy sales, however, are much higher for FlatSol simply because they had larger connected load ( gure 8). In both gures, DGVCL is an outlier with both FlatSol and MetSol owners ending up net energy importers but DGVCL is insigni cant player in Gujarat's energy-irrigation nexus.
The incremental gains to farmers are large enough to pay back their 5 percent capital contribution [16] (Molle and Closas 2021). SKY farmers not only embraced metering but got their skin in the metering process. SKY feeders show evidence of farmers' concern, never heard of before, to ensure that meters function well, and with minimum downtime. On a separate note, in promoting distributed solar generation, SKY underscores distinct superiority of farm-top solar over roof-top solar promoted aggressively under SURYA-Gujarat scheme that allows premium consumers to escape peak slab in the telescopic tariff structure. SKY bene ts consumers as well as Utility; SURYA-Gujarat bene ts only the consumer, hurtling the DISCOMs towards the dreaded 'utility death syndrome' (Lacy, 2014;Sythoff 2019).
All these make a strong case for mainstreaming and scaling out SKY to solarize Gujarat's 10.3 GW (CAG 2016) of grid-connected tubewell load and for adapting its lessons to other states.  (table 7). They offer only 8 hours/day of daytime power, a FiT of US $ 39.3/MWh, ask farmers to nd their own funds to contribute 40 percent of the capital cost. These together will make it impossible to onboard 70 percent tubewells on each feeder, making PMKUSUM-C a non-starter. For farmers to embrace PMKUSUM-C en masse, its offering must not only seem better than choosing status quo but also have the power to nudge farmers out of the lethargy that made 1/3 rd farmers on SKY feeders forgo the highly attractive FFSBS. The barriers that produced lethargy were diverse. Many farmers were not sure if they would actually get paid for solar power. Psychological barrier of loss aversion--"bird in the hand is worth two in the bush" (Kahneman & Tversky, 1979)-drove many to choose status quo, especially because the decision to join SKY was irreversible. Absentee farm-owners had no interest and their tenants had no power to join SKY. Many small and marginal farmers lacked ready capital or credit-worthiness to make even 5 percent upfront capital contribution. These and other such barriers will undermine PMKUSUM-C, too. To succeed, PMKUSUM-C needs to nudge tubewell owners to overcome these barriers, brush aside lethargy and solarise en masse.

Compulsory, Free-of-cost Solarisation of Tubewells
One way to meet this challenge is by revisiting panel oversizing ratio of 1.75/kW load that requires farmers to contribute signi cant capital upfront. Oversizing panels was necessary in Dhundi and Mujkuva solar pilots to run pumps in low light of mornings and evenings. In Andhra Pradesh solar pilot, highly e cient brushless DC (BLDC) pumps achieved the same result without oversized panels (Sirisath et al 2020). However, in Gujarat's conception of PMKUSUM-C, as in SKY, pumps are to run on grid power; here oversizing panels is not a technical compulsion but an economic incentive. Farmer can claim subsidy of US $ 436.7/HP on panels by investing US $ 290.8/HP and, over 25 years' economic life, generate attractive net present value (NPV) of US $ 410/HP [23] with a bene t-cost ratio (BCR) of 2.41. This is very pro table but may not lure heavy users of unmetered grid power for whom the BCR of status quo is in nity and NPV, around US $ 916.7/HP. [24] Moreover, substantial upfront contribution required to join PMKUSUM-C is sure to repel capital-short or credit unworthy farmers, unwilling absentee tubewell owners, and risk-averse farmers suffering 'loss aversion'.
Recent pilots on grid-connected solar pumps suggest that capital contribution drives farmers' onboarding rate and FiT drives their energy evacuation rate. In the Andhra Pradesh pilot as well as Karnataka's Surya Raitha pilot, farmers enjoyed unmetered, free grid power; yet DISCOMs onboarded 100 percent farmers by assuring: [a] zero farmer contribution to capital; and [b] free power supply as before or better. In SKY too, informal interviews indicated that close to 100 percent farmers would have joined had the Energy Minister abolished upfront farmer contribution altogether as farmers demanded instead of reducing it from 10 to 5 percent. Given this experience, there is a case for PMKUSUM-C is to launch a time-bound campaign for compulsory, free-of-cost solarisation of all grid-connected tubewells. Like compulsory free primary education, compulsory solarisation will be accepted easily only if it is offered free than if farmers are asked to share capital cost. In speci c terms, GoG should: [A] present existing guidelines (column 1 in table 8) as the rst option (PMKUSUM-COI) to metered and unmetered farmers; [B] offer a second option, PMKUSUM-COII (col. 2 in table 8), which will provide, free-of-cost, solar panels of 0.75 kWp/HP of connected load (complete with net metering installation) to all tubewell owners, consistent with recent GoG policy of providing 3.75 kWp subsidised solar panels for 5 HP off-grid pumps. All other terms will remain the same as in PMKUSUM COI. [1] PMKUSUM-CO I [C] to farmers choosing PMKUSUM-COII, provide 60 percent subsidy on additional panels up to the limit imposed by PMKUSUM-COI.
[D] resolve to compulsorily meter all remaining unmetered tubewells at the end of a speci ed transition period (by say December 31, 2023) beyond which all farm power supply in Gujarat will be metered and charged at the solar FiT of US $ 39.3/MWh or current metered tariff of US $ 8.3/MWh, as determined by GERC.
If SKY data are any guide, consumption/HP will decline sharply in response to a FiT of Rs 2.83/kWh. At solar generation of 1500 kWh/kWp/year, 0.75 kWp/HP of solar panel will offer 1150 kWh/HP/year (equivalent to 1438 kWh/HP of grid supply) and exceed consumption/HP of most tubewell owners. Figure 12 shows that those using less than 1000 kWp/HP stand to gain signi cantly from PMKUSUM-COII; those using more than 1250 will lose if they have to pay US $ 39.3/MWh for net import. But if GERC mandates, as it did with SKY, that net imports must be charged at prevailing metered tariff of US $ 8.3/MWh, then onboarding PM-KUSUM-COII becomes pro table for all farmers bar very few who are currently sent out more than 2800 kWh/HP of grid power. The option to buy more panels at 40 percent subsidy will enhance acceptability of PMKUSUM-COII among these as will the high transaction costs of selling water versus selling electricity. The fact that an overwhelming majority will gain from PMKUSUM-COII will also mute opposition to compulsory solarisation from fringe cases.  it also leaves all metered and many unmetered tubewell owners better off than now; [e] PMKUSUM-COI is more bene cial than PMKUSUM-COII for the Utility, but less so for farmers who will prefer the latter since it entails no upfront investment; [e] only unmetered tubewell owners with high current grid power consumption are better off in status quo than migrating to PMKUSUM-C; however, even these will be better off with PMKUSUM-COII (A3) than with subsidized but metered grid power supply (A4).
During recent years, GoG has provided 95 percent subsidy at a rate of 0.75 kWp/HP on some 3500 offgrid solar pumps [26] provided in lieu of metered grid connections. The objective was to avoid future subsidy pay-outs on grid power (B7). In comparison, 100 percent panel subsidy on 0.75 kWp/HP as proposed here under PMKUSUM-COII has even stronger justi cation since future subsidy pay-outs saved on legacy unmetered tubewells are 2.5 times larger per HP and 4-5 times larger per connection. Moreover, it is easier to impose compulsory solarisation when farmers pay nothing than when they are asked to share the cost of solarisation.

Conclusion and Way Forward
The heart of the problem discussed in this paper is the political economy of farm power subsidies in Gujarat. Metering all tubewells can, all at once, snap open this grid lock, drastically reducing subsidy burden, groundwater draft and energy use in irrigation. However, given farmer opposition to metering, this obvious solution has remained on paper all these years and will remain so in foreseeable future in Gujarat. This is even more true of other Indian states where all tubewells are unmetered. The proposal has other bene ts. In 2020, GoG also announced Suryodaya, a new pro-farmer scheme of providing farmers daytime grid power year-round. This will not only increase peak-load but also transmission and distribution losses. Compulsory, free-of-cost tubewell solarisation achieves the goal of Suryodaya without distorting load structure or wheeling power long distance. As net sellers of day-time power to the grid, tubewell owners will be part of the load management solution for the Utility.
The proposal discussed will create a metering regime in which farmers work with the Utility as partners rather than adversaries. All these years, farmers have loathed meters as DISCOMs' technology of control.
SKY has turned electricity meters into cash meters farmers will guard avidly. SKY farmers keenly track solar generation, consumption and export through SKY mobile application and chase Utility staff to x damaged meters. Possibilities also arise of feeder-level farmer organizations aiming to enhance earnings from energy sales by controlling line losses and conserving energy and water.
PMKUSUM-COII will also augur well for carbon footprint of irrigation, eliminating annual use of estimated 17.3 billion kWh (CAG 2016) of mostly thermal energy in pumping groundwater, saving 14.1 million MT of CO2 emissions/year (CEA 2018). Moreover, tubewell solarisation with power-buyback is expected, ceteris paribus, to reduce electricity consumption in pumping groundwater by over 40 percent, reducing signi cantly, though not proportionately, consumptive use of groundwater in irrigation.
PMKUSUM-COII is a bigger, more pro table approach to ramping up distributed solar generation compared to residential or institutional rooftop solar capacity GoG is aggressively promoting. Rooftop solar robs DISCOMs of their premium customers while 'farm-top' solar (for tubewells) liberates them from subsidy burden.
Vala. R.J. SKY Business Model for Solarisation of Agriculture Pumpsets, presentation made in World Bank Seminar on, "Grow Solar, Save Water, Double Farm Income' in New Delhi (2020).

Viswanathan, P.K. Trends and Determinants of Agricultural Power Consumption and Distribution Losses:
An analysis of agricultural feeders across power distribution companies in Gujarat", Gandhinagar, Gujarat Institute of Developent Research, report for GERC (2014). Figure 1 Nexus between electricity subsidies and groundwater Depletion in Gujarat Trend in power consumption/HP during peak irrigation months Gains/HP and Connection from SKY: DISCOM and GoG Perspective Figure 12 Breakeven level of consumption/HP beyond which adopter loses from PMKUSUM COII