Intervention analysis of an irrigation system using a structured system concept

R. Sakthivadivel¹, S. Thiruvengadachari² and Upali A. Amarasinghe¹

1 IWMI - 2 IWMI Consultant (former NRSA)

Abstract

Modernization with a structured system concept is increasingly adopted in irrigation improvement programmes in the Indian subcontinent and elsewhere to alleviate the burden of operating the system. This paper evaluates the performance of the Bhadra reservoir project in India - before, during and after the introduction of modernization with such a concept. The performance analysis, supported by data generated by remote-sensing techniques, office records, field visits and farmer surveys, focused on water management, agricultural productivity and farmer participation and perceptions. The analysis of IRS IC satellite wide field sensor data on 20 overpass dates during the rice growth cycle of the most recent rabi season proved particularly valuable to evaluate water distribution between the distributary commands.

The concept of a structured system and equitable supply-based technology, central to the National Water Management Project intervention in the Bhadra project, did not succeed for various reasons analysed in this paper. The analysis indicates among other factors that water use since then has increased as a result of the cropping pattern being dominated by water-intensive crops such as rice during the rabi season. Preferential allocation to head end of command continues and inequity sets in within the distributary commands. The tail-end water supply deprivation is partially offset by farmers practising deficit irrigation. The kharif and rabi seasons did not come early, so there was water saving. Farmers' organization and participation in decision-making at scheme level and water distribution at distributary level and below are very low. However, agricultural productivity has not registered a significant decline since before the intervention of the National Water Management Project.

The absence of continuing support mechanisms by way of institutional arrangements and effective farmer participation and involvement in implementing the operation plan has been the major cause of the decline in water management, which has slid back to quasi demand-based supply, and in productivity per unit of water. Strong farmer involvement thus holds the key to sustainable performance.

Introduction

The structured system concept in irrigation projects is about maintaining control of the water flow through the use of proportional devices without human intervention, intermittent water supply in the distributary canals and a systematic operation plan. The concept advocated by the World Bank (Shanan, 1992) is being increasingly adopted in the Indian subcontinent in the rehabilitation and modernization of irrigation projects, some of which have been under operation for more than five years after withdrawal of external aid. A holistic performance evaluation of such systems, especially of how well they have been sustained after closure of external credit, is of great significance not only to India but also to other developing countries, given dwindling financial allocations for irrigation operations, maintenance and management, increasing competition for water and interference by vested interests.

The structured system concept is central to the World Bank-funded National Water Management (NWM) project, which covered some 80 schemes in 11 states in India during 1988-1995. Subsequent to the NWM project, the concept was extended to the ongoing Water Resources Consolidation project in the states of Tamil Nadu, Orissa and others. In view of its widespread adoption there is a pressing need to evaluate the adoption of the structured system concept in improving irrigated agricultural performance. This paper attempts to evaluate the Bhadra reservoir project in the state of Karnataka, which was one of the earliest to adopt the structured system concept under the NWM project. The main objective of this paper is to carry out a comparative analysis of the project before, during and after the NWM project. While the main focus will be on water distribution and agricultural productivity, the analysis will also briefly cover other relevant issues such as farmer participation, monitoring and evaluation, and training. The lessons learned could be of use for future implementation of such projects.

The paper builds upon the 1995 evaluation by the International Irrigation Management Institute (IIMI), which focused on water distribution, agricultural productivity, training and farmer participation up to the 1993-94 rabi season. That study found that farmer participation had shown no significant progress and planned farmer organizations had not been created, water distribution along the distributaries had improved, and significant improvements in agricultural output and in the value of that output occurred after NWMP was introduced. IIMI has now revisited the project to evaluate the lingering impact of NWMP two years after external aid stopped in 1995.

The analysis will address the following questions:

• Has there been a decline in irrigated area, change in cropping pattern or in land and water productivity of rice, particularly in the tail-end areas in recent years since external inputs stopped?
• Has water management deteriorated in terms of more water being used, head-to-tail preferential allotment or use, irrigation seasons and crop calendar changed, and the canal operation schedule being different?
• What is the current status of water user associations?

A noteworthy feature of the comparative analysis is the application of satellite remote-sensing techniques to generate objective and de-aggregated information on agricultural productivity during the rabi season, particularly on rice productivity per unit of land. Multiple-date data from the recently launched IRS IC satellite during the 1997 rabi season also provided indications of equity and reliability of the water distribution between the distributary commands. Ground data were obtained from office records, field visits and farmer surveys.

The comparative study argues that the successful operation of a structured system, particularly in a water-rich environment, is complex and requires realistic planning, rigorous implementation (involving the wholehearted participation of farmers) and a mechanism for sustaining the benefits of intervention after cessation of external aid. The conclusions are supported by a critical analysis of proposed NWM interventions for improved water management and analysis of actual water distribution and agricultural productivity across the command area and through the years during the implementation of the NWM project and since the cessation of aid.

The Bhadra reservoir project

The Bhadra reservoir project (Fig 1) was built between 1946 and 1966. The reservoir has a gross storage capacity of 2 025 Mm³, a live storage of 1 608 Mm³ and a surface area of 11 200 ha. An annual withdrawal of 1 747 Mm³ (61.7 TMC), including reservoir evaporation losses, was awarded to the project by the Krishna River Tribunal, with 1 400 Mm³ specific to the right bank canal, whose command is covered under the NWM project. The NWM project was implemented during 1987-89 and most of the physical rehabilitation work was completed by that time. The estimated cost of the project is Rs5 500 per ha (US$160/ha) at 1994 prices. However, external funding for operation, maintenance and management was continued up to 1995. Three periods - pre-intervention up to 1988, intervention from 1989 to 1995 and post-intervention thereafter - are considered for analysis of the performance.

Figure 1. Location of the Bhadra project command area

The right bank canal, with a full supply discharge of 75 cumecs, commands an area of 92 360 ha and has 29 direct distributaries, which irrigate 17 050 ha. The Anvery Branch canal, watering 6 320 ha, branches off at Km76.4 of the right bank canal. Further along, at Km103, the right bank canal bifurcates into the Malebennur Branch, benefiting 23 710 ha, and the Davangere Branch, which waters 45 280 ha. Administratively the whole right bank canal command area is divided into three divisions: Bhadravathi, with three subdivisions; Malebennur, with four subdivisions; and Davangere, with four subdivisions.

The annual rainfall in the command area is about 800 mm and decreases toward the tail end. Most of the rainfall is concentrated during the south-west monsoon period of June to September. The mean monthly evapotranspiration is equal to or below rainfall during July to October, but is substantially above it during the rest of the year. Even during the south-west monsoon period, breaks in the monsoon are frequent, necessitating supplemental irrigation. The command area gently rolls with moderate slopes. Red soil covers 60 percent of the command area. The majority of the holdings are less than 1 ha and are generally split into 2 or 3 segments.

The National Water Management project

The main objective of the National Water Management project was to increase productivity and farm income in the existing irrigation scheme through a more reliable, predictable and equitable irrigation service. This would be achieved by developing the institutional capacity to plan, implement and monitor improved operation and maintenance practices and providing low-cost infrastructure improvements to support the proposed operational plan. The improved water management would come from more systematic canal operation, best use of available rainfall, reduction of irrigation wastage and optimization of the cropping season particularly during the rabi season, and reduction of crops with high irrigation demands by manipulating the frequency of irrigation.

Before the introduction of the NWM project, rice was grown wherever possible. Water availability was good in the distributaries directly branching off the right bank canal and in the head reaches of the branch canals. Irrigation supply was very unreliable in tail reach areas, leading to cultivation of mainly semidry crops. The cropping seasons were kharif from July to November and rabi from January to May. The NWM project aimed at a better use of water resources by providing irrigation supply starting from May, which would enable a better use of rainfall by the kharif crops and would lower the irrigation requirement for the rabi crop, which it would be possible to sow in November. This plan required a carryover storage of 250 Mm³ in the reservoir for irrigation supply in May and part of June. The seasonal calendar proposed under the NWM project was kharif extending from 15 May to 15 October (153 days) and rabi between 15 November and 15 April (121 days).

The assumed cropping plan of the right bank canal command envisaged 100 percent cropping intensity in the kharif season. In the rabi season, rice was to be excluded by releasing water to the canal intermittently during the first two months. But even with 90 percent of the area under semi-dry crops, peak irrigation requirements in the dry season were so high that within the capacity limitations of the canal, only 75-percent intensity could be achieved. The distribution policy, as incorporated in the staff appraisal report, envisaged that the command would be divided into four zones and every year one zone, selected in rotation, would be left out of rabi irrigation.

Under the NWM concept, the available water resources were allocated proportionally within the command area. In principle, each farmer received his share based on the acreage he either owned or rented, irrespective of the crops grown and irrespective of farm location.

In order to prevent the upper parts of the command from monopolizing the water withdrawal to the disadvantage of the lower command, distributaries and lower level channels were to be so sized with proportional outlets that they would not draw more than an equitable share of the capacity of the main canal. The NWM project proposed the establishment of a structured distribution system below the turnout from the right bank canal and branch canals in the distributaries and minors and installation of a network of flow measurement structures at strategic locations. In Bhadra, the level of structuring was restricted to distributary, i.e. branch canal outlets leading to distributaries were operated as non-structured (manual) systems. A scheme-level committee, with the participation of the Irrigation Department, Command Area Development authorities, the Agriculture Department and farmers, would meet before each irrigation season to approve the operational plan and define the rules for implementation, and again meet after each season to evaluate the operation.

Systematic linkages were to be developed between the operation-and-maintenance agency, other government agencies and farmers. Operation and maintenance were to be strengthened by setting up a main canal operating unit with computer facilities.

The operational plan at the scheme level would be the basis for monitoring system operations. Regular reports would be submitted to the project committee on the seasonal and annual operations (water delivery, irrigation efficiency, cropped area, farmer satisfaction) in order to evaluate performance. The subsequent seasonal operation plan would be modified using this feedback. Thus the scheme level monitoring and evaluation would be an essential element of NWMP intervention.

Training would cover orientation training for senior officers, training in planning for the staff of operation and maintenance units, training in design for engineers, and implementation and operation training for scheme-level officers and field workers. Farmers would be trained to understand NWM concepts to lead to their acceptance and create awareness of their own responsibilities. The NWM project considered farmer participation very important to the success of a scheme operation. Effective participation was to be achieved by holding meetings with farmers to discuss NWM plans and to modify the plans to suit specific situations, organizing a farmer committee at each outlet and higher level committees for each minor and distributary canal, and through farmer training.

Basic data

Historic and current-season data on reservoir release and distribution in branch and distributary canals has been collected from the Bhadra Reservoir Project offices. Agricultural productivity data in terms of total irrigated area, area under rice and other crops and rice productivity per unit of land were obtained through the digital analysis of satellite data of the 1986-87 and subsequent rabi seasons. A satellite inventory for the 1994-95 and 1996-97 rabi seasons which was not available at the time of the 1995 IIMI evaluation has been used in the study. This data available at pixel level can be aggregated for any desired aerial unit such as distributary command or at branch canal level. Since crop-cutting experiment data for the 1996-97 rabi season for sample rice plots was not available during the study, a satellite-based rice yield model was developed by normalizing the rice yield model of the 1992-93 rabi season to the current season (Appendix 1). An additional feature of the satellite inventory of the 1996-97 season is the analysis of 20 IRS IC satellite wide-field sensor data sets (with a repeatability of five days compared to 22 days for IRS-2A and IB) between February and June 1997. The average wide-field sensor-based normalized difference vegetation index for each distributary command plotted against time represents the rice growth profile during the rabi season and enables comparative analysis of rice conditions through the season and between distributary commands.

A farmer survey covering 105 farmers in seven distributaries (one each in the head and tail end of the Right Bank Canal direct command, one each in the head and tail end of the Malebennur Branch and one each in the head, middle and tail end of the Davangere Branch) was conducted in April 1997 (Figure 1). Typically, in each distributary three villages were selected, one each in the head, middle and tail reaches; the selection of five farmers in each village was based on village records and distributed between small and large farmers. A structured questionnaire was given to the farmers to fill in, and the data organized in a data base for further analysis. Interaction with the water user associations has made it possible to evaluate their current status and effectiveness.

Application of satellite remote-sensing techniques

The earliest satellite remote-sensing application for de-aggregated irrigation inventory in India was in the Bhadra project (Thiruvengadachari and Sakthivadivel, 1997). The methodology, developed and refined since 1992, has been applied in the Bhadra project to generate de-aggregated information on the cropping pattern and rice productivity per unit of land right up to the reach of a distributary for the rabi seasons from 1987 to 1997. An important application is the development of a rice yield model linking the satellite-derived normalized difference vegetation index to the harvested rice yield in sample plots. A further development is the extension of such spectral yield relationship from one year to another through radiometric normalization of satellite data (appended). In addition to area and yield data, the spatial variability in the rice transplantation calendar could be captured through the analysis of multiple-date satellite data through the season. Satellite remote-sensing data supported by ancillary information have been applied in a geographic information system environment to analyse system performance and to conduct diagnostic analysis of the Bhadra irrigation project.

An important recent development is the improved capability for through-the-season analysis with wide-field sensor data of 180 m spatial resolution and five-day revisit period since December 1995 from IRS IC and since September 1997 from IRS ID satellites. During the 1996-97 rabi season, 19 rounds of cloud-free coverage were obtained and the normalized difference vegetation index profiles were generated and analysed for spatial homogeneity in rice growth and health status.

The NWMP planning process: a critical analysis

The project report ventures that optimum water use would be achieved through more systematic canal operation making the best possible use of the available rainfall and reducing irrigation wastage, optimization of the cropping season, particularly rabi, and reduction of the crop area with high irrigation demand by manipulating irrigation frequency. This has not been completely or adequately addressed under NWMP proposals for intervention in the Bhadra project. Significant deficiencies including unrealistic assumptions and incomplete attention to details have vitiated the planning process and led to unsatisfactory impact during and after implementation. While equity is ensured when the water supply is proportional to the area irrespective of the crops grown, the assumed cropping pattern helped in the determination of expected water requirements and sizing of canal capacities. The catch here is that the assumed cropping pattern was unrealistic.

The high rainfall in the upper command (an annual rainfall of 950 mm is normal in Bhadravathi compared to 650 mm in tail-end Davangere) makes rice the preferred crop but it leads to spatial variability in cropping pattern during the rabi season. The assumption of excluding rabi rice is doomed to fail without strict adherence to the operational schedule of this traditional rice-growing area. Thus the more realistic assumption of a spatially varying cropping pattern and an alternate definition of equity by changing priorities for assured water supply in the two seasons would have had better implementation potential. A careful analysis of past cropping responses to irrigation supplies and rainfall pattern, and consultation with farmer groups, would have resulted in a more realistic cropping pattern. The steps for achieving the farmer consensus and participation necessary for effective supply management have not been sufficiently detailed, nor have specific funds been provided.

In order to use rainfall effectively, the canal supply needs to be variable and capable of adjustment. The NWM project, in contrast, proposed that the branch canals be designed with a capacity equal to the sum of the capacities of off-taking distributaries (with allowance for losses), and be operated at constant maximum flow. This would mean that if a distributary under a branch canal is closed due to lesser demand, the excess flow would just get redistributed among other distributaries or run downstream. In either case no saving would result.

When the main canal is operated at varying flows, closing a distributary can result in saving only if there is a corresponding change in the release at the right bank canal head regulator or by opening a closed distributary of similar capacity. This requires a tightly operated system with excellent communication facilities and an effective decision-making system. The NWM project did not propose to develop such support mechanisms. Thus the assumption of water saving under the project was defeated at the planning stage itself.

Effective use of rainfall and consequent reduced reservoir withdrawals were also to be achieved by advancing the kharif sowing to mid-May by providing irrigation from storage carried over from the previous water year. This was corroborated by a simulation study showing that the carryover storage of 250 Mm³, including storage required to meet all the evaporation losses from the reservoir, was available every year during the 1974-1984 period. The rabi sowing correspondingly would be advanced to November, reducing crop water requirements in that season.

The operational plan proposed running the right bank canal continuously during the advanced kharif season. This would result in the withdrawal of 985 Mm³ from the reservoir. The withdrawal in the rabi season with three initial cycles of on-off operations is computed at 674 Mm³. The annual withdrawal would thus be 1 659 Mm³, and with 175 Mm³ allocated to the canal and reservoir evaporation of 172 Mm³, the total withdrawal works out at 2 006 Mm³ against an allocation of 1 747 Mm³ by the Krishna River Tribunal. Any saving would be only by operating the main canal with a varying flow and, as mentioned earlier, this was not an effective proposition.

Comparative analysis of performance: before, during and after intervention

This section will attempt to compare the performance during and after implementation against what prevailed before 1988. The main focus will be on water management, agricultural productivity and farmer perceptions, with lesser attention to other NWMP elements. The collapse of the guide wall at the entrance of the right bank canal in September 1991 marks a change from the prevailing 75 percent zoning to the irrigation of the whole area in the rabi season. Therefore the performance during the period from 1992 to 1995 is taken to represent the impact during implementation, unless otherwise specified. The post-implementation period is considered to be after 1995.

Water management

The start of the canal operation was always delayed during the pre-intervention period; it was in time as planned during the intervention period and up to the 1995 rabi season and advanced beyond kharif 1995 (Table 1). The actual number of days the canal was operated - as against the actual number of days planned - also varies, especially during the post-intervention period. Substantial deviations from NWM postulates against the actual planned and operated water distribution schedule are seen for all the three periods investigated. The advancing of the kharif season to 15 May, and consequently of the rabi season to 15 November, as contemplated in the original proposal, was not achieved, and the seasonal calendar highly varied from year to year, more so in the kharif season. The start of the kharif season has been delayed particularly since 1994-95 because of monsoon vagaries. The gap between the planned and actual withdrawal from the reservoir is shown in Figure 2. Water use during the rabi season became more critical when the 25 percent zoning policy under the NWM project was abandoned in 1992-93, and the whole right bank canal command was provided irrigation supplies. The excess water use in the rabi season thus increased, against the computed rabi allocation of 674 Mm³, from 19 percent in 1992-93 to more than 30 percent after 1994-95, compared to 35 percent in the pre-NWM years (Table 2). Although withdrawals during kharif were less than those stipulated in the NWM project proposals, the annual withdrawals steadily increased between 1992-93 and 1996-97 (0 to 5 percent). The marginal decrease in excess withdrawal in 1996-97 was due to poor rainfall in the Bhadra catchment area (the area received 1 245 mm in 1996 compared to the 1 569 mm normal rainfall), which resulted in lower reservoir storage levels in that year. Thus the NWM objective of restricting water use within the tribunal allocation was not achieved, as water use increased as of 1992-93. The rather significant increase in water use since 1994-95 is perhaps the result of the physical deterioration of the system due to inadequate maintenance, with more water needed to push supplies to the whole command.

Table 1. Irrigation schedule : planned and actual

^{NWMP planned Kharif: 153 days - Rabi: 121 days}

Table 2. Excess withdrawals from reservoir

^{With reference to 1 659 Mm3 planned under NWMP for annual withdrawal, and 674 Mm3 for rabi withdrawal}

Table 3. Water shares among branch canals and ratio of actual-to-nominalshare

The NWM project also postulated equity, with water delivered proportionally to the area irrigated, irrespective of the cropping pattern. Deviation from equity is seen, with preferential deliveries to the head reaches at the expense of the tail-end portions (Table 3). The direct commands under the right bank canal and under the Anvery Branch are drawing more than their share, with marked decrease in water receipts in Malebennur and Davangere branch canals. The latter seems to be the main sufferer, with only a 41.9 percent allocation against 46.5 percent postulated under the NWM project and less even than in the pre-project 1986-87 rabi season. The inequity evident even during the kharif season indicates that the higher rainfall contribution in the head reach is not being appropriately used. The inequity between the head and tail reaches has also increased in recent years. The similarity in inequity in kharif and rabi seasons seems to be due to the fact that the extent of rice cultivation in both seasons is similar.

In addition to preferential allocation to head-end branch canals, inequity also exists between distributaries under each branch canal, though to different extents. The vegetation index profiles derived from IRS IC satellite wide-field sensor data of 20 overpasses throughout the 1997 rabi season enabled the analysis of spatial and temporal homogeneity in regard to equity, adequacy and timeliness of water distribution (Figure 3). The vegetation index profiles of the right bank canal suggest a different cropping pattern from other bank canals as they started initially with a high variability (20-100) whereas other bank canals are within a limited range (20-40). The dispersion of the profiles within each branch canal and the highest vegetation index value reached are also different. The tight bunching and similar magnitude and shape of profiles in the Davangere and Malebennur branch canal commands indicate better equity, similar adequacy and timeliness, and less so in the Anvery Branch and right bank canal commands. This indicates, as shown by the high yields in Table 4, better management in the middle and water-short tail end of the command compared to the upper part with excess water withdrawal. The data gaps in the vegetation index profiles are due to cloud cover in wide-field sensor data coverage on the overpass dates.

It has not been possible to operate all the distributaries simultaneously at specified discharges in the kharif season because the distributaries in the upper reaches drew higher than specified discharges and also due to the inadequate capacity of the right bank canal. The rabi operational plan calls for a strict calendar for on-off operation of all the distributaries simultaneously with the right bank canal also following the calendar. Rabi zoning has been abandoned since 1992-93, the canal has been operated continuously since 1994, and on/off operations of distributaries do not follow a set calendar either, implying tacit acceptance of field realities.

A major departure in operation and maintenance is that farmers' pressures could not be withstood and this led to a changeover from supply to demand management. There has also been reintroduction of gates at the heads of some minors in some tail-end distributaries of the Davangere branch canal, making a departure from the structured system concept of the NWM project, indicating a tendency to correct deviations in water distribution in the upper command by ad hoc solutions at the tail end. In spite of well-constructed measuring structures at distributary off-takes and along the right bank canal, the monitoring of system operations has remained weak. Only three distributaries, the same every year, were covered under intensive monitoring. The monitoring report does not highlight the difference between targeted and actual deliveries or the relative performance along the sample distributaries or the implications of the current cropping pattern on the operation plan. The scheme-level report is also inadequate as it does not emphasize the pattern of deliveries within the system or extent of equity. Also, the monitoring and evaluation exercise does not cover groundwater status. An indication of the deteriorating efficiency in monitoring and evaluation was the great difficulty the authors faced in obtaining canal discharge data from the divisional offices. While in the early years of NWM intervention, the data were compiled and documented at the division and circle offices, data for 1996-97 had not been compiled at these offices even after the rabi season was over.

Figure 2. Comparison of actual and planned seasonal delivering through the years

Figure 3. Seasonal vegetation index profiles under different branch canals

Agricultural productivity through the years

The spatial variability in irrigation intensity, percentage of rice area, and rice yield across the canal divisions in the rabi season are shown in Table 4. The overall irrigation intensity has increased (statistically significant at 5 percent level) from 75 percent in 1986-87 to 91 percent in 1992-93. The irrigation intensity in the subsequent rabi seasons has declined since 1992-93 and is not significantly different from the level before the implementation of the NWM project. The area under rice has increased (significant at 5 percent level) from 56 percent in 1986-87 to 69 percent in 1992-93. The rice area in the subsequent rabi seasons is similar to that of 1992-93, but significantly different from the 1986-87 rabi season area. The average rice productivity per unit of land, both during and after the intervention years, is significantly different from that of the 1986-87 rabi season. However, the average rice yield in 1996-97, i.e. the first rabi season after the intervention period, is significantly lower than those recorded during the intervention period.

The agriculture performance in all three divisions, i.e. Bhadravathi in the head end, Malebennur division in the middle and Davanagere in the tail end, shows more or less the same trends as that of the total command area (Table 4). For example, in the tail end Davangere canal division, the irrigation intensity rose from 75 percent in 1986-87 to 87 percent in 1992-93 (significant only at 10 percent level). The irrigation intensity in the subsequent time periods, for example 84 percent in 1996-97, is not significantly different (at 5 percent level) from the irrigation intensity in the 1986-87 rabi season. The rice intensity in the tail-end section has increased from 51 percent in 1986-87 to 66 percent in 1992-93 (significant at 5 percent level) and stabilized since then. The rice yield has significantly increased (at 5 percent level) from 3.8 tons per ha in 1986-87 to 4.4 tons per ha in 1992-93. The yields in the other time periods during and after intervention are also significantly higher than those of the 1986-87 rabi season. The irrigation deliveries are available only at the divisional level.

The rice productivity per unit of water diverted at the reservoir has been steadily declining through the years, from a high of 0.45 kg/m³ in the 1992-93 rabi to a low of 0.36 kg/m³ in the 1996-97 rabi season against 0.25 kg/m³ in the 1986-87 rabi (Table 5).

Table 4. Satellite-derived crop pattern in the Bhadra command area Irrigation intensity, percentage rice area and rice yield

Table 5. Rice productivity through the rabi seasons

• Computed as rice area + 0.6 x non-rice area (non-rice area water consumption is approximately 0.6 times the consumed water per unit rice area).
• Computed as equivalent rice area x rice productivity per unit of land.
• Computed as equivalent rice production divided by water volume diverted at the reservoir during the rabi season.

The decrease in irrigated area and rice intensity and yield during the 1996-97 rabi season is attributable partly to low reservoir inflows and partly to greater conveyance losses in recent years due to poor maintenance of, and damage to, irrigation structures. Field visits by the authors as well as observations during farmer surveys, revealed broken irrigation structures in many places in the tail end. However, the decline in rice productivity per unit of land from previous years is owing less to this than to farmers irrigating near-normal areas but at significantly lower depths of application due to conveyance losses and to lesser reservoir withdrawal. These findings are well reflected in the wide-field sensor vegetation index profiles of different branch canals (Figure 3).

Farmers' participation and perceptions

Farmers' participation in the NWM project was extremely weak. Of the five activities proposed for farmer participation - meeting with farmers, creating a farmer organization below each outlet, organizing a water use committee for each minor or distributary, farmer training and creation of a scheme-level committee - none materialized with specific input from the project. However, the Bhadra Command Area Development Authority created nine village-based farmer co-operatives between 1992 and 1994. Among these, the tail-area farmer co-operatives were relatively stronger than head-end co-operatives in solving their water problems during the implementation of the NWM project. One such farmer co-operative, started in 1992 at Tiruchughatta at the tail end of the eighth distributary of the Davangere Branch was doing some good work during the implementation of the NWM project. When we visited in 1997, it had become defunct, although the organization remained in name. The situation was the same for two other organizations we visited that year. The functions of the scheme-level committee suggested under NWMP are looked after by the traditional district-level irrigation consultation committee, which is less effective. Of the 39 members in the committee, only four are farmers; they too are nominated as political representatives and not elected directly by the farmers of the Bhadra reservoir project. Most important, the irrigation consulting committee does not evaluate the system performance and does not oversee NWMP activities.

The analysis of farmers' perceptions is based on the data generated using a farmer survey questionnaire. Three levels of aggregation are considered for the analysis. The first level is reached by considering the total command area as a unit. The second level considers grouping distributaries at head, middle and tail end respectively. The third level assembles the parcels at head, middle and tail reaches along distributaries as separate units. The results of the analysis for the first two levels are presented in figures 4 to 7 and those for the third level analysis in figures 8 to 11.

The results presented in figures 4 to 7 depict the present condition (as of 1997) of water distribution as well as the change of performance and system conditions before and after rehabilitation. Figure 4 presents the irrigation adequacy of present supply and change since rehabilitation for the kharif and rabi seasons. The horizontal bars show the percentage of sampled farmers who reported "never adequate" water supplies at present and "better in the past", i.e. before rehabilitation.

For the command as a whole, only 20 percent of the farmers reported that they never received adequate water in both irrigation seasons. However, there is head-tail disparity in current water adequacy along the main canal. Almost 45 percent of the farmers at the tail end reported that water was never adequate, compared to 9 percent in the middle and nearly zero percent in the head-end distributaries. Farmer perceptions of current adequacy varied significantly depending on the location of the distributaries (Pearsons' Chi square equals 35.6 for kharif, and 21.7 for rabi, df = 4). In the command area, 40 percent of the farmers reported that adequacy was better before rehabilitation. The percentage of farmers expressing such an opinion varied from head (25 percent) to tail (70 percent). As in the case of current adequacy, farmer perceptions of change in adequacy since rehabilitation varied significantly depending on the location of the distributaries (Chi square equals 18.2 for kharif and 11.8 for rabi, df = 4). This type of response in the context of water diversion into the canal being stepped up may be due to the increase in irrigated area after rehabilitation with a high percentage of rice cultivation requiring more water.

Figure 5 presents irrigation equity and change in equity before and after rehabilitation for both kharif and rabi seasons. The horizontal bars represent the percentage of farmers who reported "not fair" at present and "better in the past" before rehabilitation. The figure indicates that equity is not a major problem in the head and middle reaches while it is one in the tail-end distributaries, where 23 percent of the farmers in kharif and 61 percent in rabi reported "not fair" at present. Again, farmer perceptions of current equity varied significantly depending on the location of the distributaries (Chi square equals 19.2 for kharif and 23.8 for rabi, df = 4) : the majority of the farmers in the head end reported either "mostly" or "sometimes fair", whereas the majority of the farmers in the tail end reported either "sometimes" or "rarely fair". A quarter of the farmers in the command area felt that equity was better before rehabilitation. However, this percentage was higher (57 percent for kharif and 73 percent for rabi) for the tail-end farmers. The latter perceived both adequacy and equity as problems mainly because of change to larger areas under rice coupled with main system canal constraints in pushing sufficient water to the tail-end distributaries.

Figure 6 presents farmer perceptions of the functional condition (current and also the change since rehabilitation) of the canals within the unit, and also at the main system level. Uniformly from the head end to the tail end, about 60 percent of the farmers mentioned the poor condition of canals within the unit at present, and about 40 percent of the farmers in each reach pointed to the poor functionality at present of the main system. However, with regard to change in functionality of the system, both within the unit and in the main system, about 30 percent of the farmers in the head and middle and about 70 percent of those at the tail end reported that functionality was better for the whole system before rehabilitation. Yet again, farmer perceptions of change since rehabilitation varied significantly depending on location of the distributaries (Chi square equals 10.1 for within unit and 19.15 for the main system, df = 4). The main reasons for this kind of perception appear to be the mismatch between the water requirement of the irrigated area and high-intensity water-loving crops, and the main system constraint coupled with the operational policy adopted by the irrigation agency.

Figure 7 is a representation of the frequency of conflicts at present and since rehabilitation. When water becomes inequitably shared and inadequate, especially at the tail end, conflicts increase. About 25 percent of the farmers in the whole command reported frequent conflicts at present. However, the proportion in the tail-reach distributaries was 48 percent. Also, 33 percent of the farmers in the command area reported that conflicts became more frequent after rehabilitation, while 57 percent of the farmers at the tail end reported more frequent conflicts since rehabilitation. Farmer perceptions of conflicts at present and also since rehabilitation varied significantly depending on the location of the distributaries (Chi square equals 20.7 for the present and 12.4 for change since rehabilitation, df = 4).

From figures 8 and 9, it can be seen that while the response of farmers concerning adequacy and equity in kharif is not very different along the head, middle and tail reaches of tail-end distributaries, there is a distinct difference in responses during rabi. Both in the middle and tail reaches more farmers reported inadequacy and inequity than did the head-reach farmers. The inadequate and inequitable distribution of water has also given rise to conflicts among farmers, especially within the middle and tail reaches. Though the number of farmers in each category is not adequate for a statistical analysis, the main conclusions that can be drawn from these four figures are that there is inequity and inadequacy of water supply among field plots within the tail distributaries and that the plots within the middle and tail reaches suffer more than those at the head reaches.

Impact of the closure of World Bank credit

The World Bank credit closed in March 1995, after incurring an expenditure of about Rs389 million against the estimated Rs196 million. The balance works estimated at Rs7.8 million have not yet been completed. The impact of the closure of credit in regard to both financial and supervisory inputs is a slackening in the system performance. With the closing of additional operation and maintenance funds coming from external credit, the maintenance has become poor and structures which are broken in many places have not been rebuilt. The NWMP objective of more equitable, reliable and predictable irrigation supplies, which could not be achieved even during the NWMP period, has shown a widening gap with actual performance in recent years. The agricultural calendars stipulated in the NWMP proposal have not been followed; actual and planned operations were different, and the advancing of the kharif season could not be implemented. Water use, particularly in the rabi season, is increasing, and agricultural performance (total irrigated area, area under rice, and rice yield) has registered a decline. Tail-end deprivation, due to preferential water allocation to head-reach areas, is stressing agricultural performance in this command. However, through deficit irrigation, tail-end farmers have maintained irrigation intensity and rice intensity. Farmers' participation, and particularly the institution of water user associations, has declined. No new such associations have been formed and even those existing earlier have become non-functional. Farm surveys indicate that a large majority feels that system performance was better in the past than now.

Summary of the analysis

Many interventions as contemplated in the staff appraisal report of NWMP could not be made fully operational for various reasons analysed in this paper. In spite of these deficiencies, the agricultural performance (irrigation efficiency, rice irrigation intensity, land and water productivity) of the project significantly improved during and after the intervention period compared to the pre-intervention period. For example, in the tail-end Davangere canal division, the irrigation intensity rose from 75 percent in 1986-87 to 87 percent in 1992-93 and slightly declined to 84 percent in 1996-97. The rice intensity in the tail-end section increased from 51 percent in 1986-87 to 66 percent in 1992-93 and has stabilized since then. The rice yield significantly increased, from 3.8 tons per ha in 1986-87 to 4.4 tons per ha in 1992-93 and slightly declined to 4.3 tons per ha in 1996-97. The productivity per unit of water delivered at the reservoir increased from .25 kg/m ³ in 1986-87 to .45 kg/ m ³ in 1992-93 and slightly declined to .36 kg/m ³ in 1996-97. Since the closure of the World Bank loan in 1995, there has been a declining trend in all agricultural results except rice irrigation intensity. Data is not sufficient to pinpoint whether this decline is due to the deteriorating condition of physical management systems or to a low reservoir inflow in 1996-97 or to both.

Three main factors appear to constrain the potential performance of the system. They are:

1. switching to a cropping pattern dominated by water-intensive crops such as rice during the rabi season;
2. abandoning area zoning (as envisaged by the NWM project) and irrigation of the full extent of the command during the rabi season; and
3. lack of proper institutional mechanisms such as farmer organizations and their committed involvement in operation, maintenance and management of the system.

The management problems in a structured system are directly related to the level of structuring. Supply management, in the context of a heterogeneous cropping pattern of different irrigation requirements, cannot succeed unless farmers organize themselves to co-ordinate water distribution below the distributary or minor or outlet. The structured system shifts the requirement for organization and management from the operating agency to the farmers. Thus in traditional demand-based systems such as the Bhadra project, the existence of farmer organizations at scheme level and at the appropriate sub-unit level is necessary to the success of a structured system. There are none in the Bhadra project, where the emphasis continues to be on system rehabilitation rather than on farmer training and organization. Farmer participation both in decision-making at scheme level and in operation and maintenance at the distributary level needs to be given the pre-eminent role, and detailed guidelines and mechanisms need to be put in position in the early stages of implementation itself.

The design of the structured system in the Bhadra project was constrained by limited funds provided for institutional strengthening within the system and by the need to work within existing hydraulic system limitations. The operating procedures of the structured system did not explicitly take into account the heavy rainfall and its spatial variability across the command area. The proposed cropping pattern was unrealistic, particularly in the context of the complete absence of farmer acceptance and adherence to rules of operation. The involvement of farmers in planning, design and implementation becomes more necessary and they need to be committed and motivated.

The successful operation of a structured system, in addition to social change, also requires a concomitant institutional change. In the Bhadra project, for instance, the Karnataka Irrigation Act legalizing the localization (demand-based) system has created confusion between supply and demand management. This confusion continued and exerted pressure on the scheme-operating agency and its operation.

Farmers' involvement and acceptance of the operation plan is essential in the recourse to supply-based technology. In the absence of a scheme-level committee and functional farmer associations, the NWM experiment in Bhadra was bound to fail. The conflicts among the farmers continued and resulted in undue pressures on the scheme operators, who tinkered with the operation in an ad-hoc fashion. The slide back into quasi demand-based supply and ad-hoc functioning prevented the rigorous implementation of operational plans, resulting in less equity, reliability and predictability.

The maintenance of the system failed in view of insufficient grants and more so since it had not been transferred to the water user associations. The farmers continued to feel that maintenance was the responsibility of the Irrigation Department, which did not have adequate funds anyway. Damage to structures by farmers continued and no repairs were undertaken. The creation of water user associations and transfer of operation and management responsibilities of the structured system are essential to successful operation and management activities. A strong monitoring and evaluation function would support the review of operation plans and the necessary mechanism needs to be introduced.

In the absence of a continuing support mechanism by way of institutional arrangements and farmer participation, the Bhadra project is showing signs of decline, with water management sliding back to quasi demand-based supply. Strong farmer involvement thus holds the key to sustainable performance.

Acknowledgements

The co-operation extended by the Bhadra project authorities in providing data and toward field visits is acknowledged. Analysis of satellite data was conducted by the National Remote Sensing Agency in Hyderabad, India.

References

IIMI 1995. Evaluation of schemes under NWMP-II: the Bhadra scheme in Karnataka and the Sathanur scheme in Tamil Nadu, International Irrigation Management Institute, Sri Lanka

Murthy C.S., S. Jonna, P.V. Raju, S. Thiruvengadachari & K.A. Hakeem, 1995. Paddy yield prediction in the Bhadra project command area using remote sensing data. Asia-Pacific Remote Sensing Journal, Vol8 (No1) 79-84

Shanan L. 1992. Planning and management of irrigation systems in developing countries, Agricultural Water Management, Vol23 (No1&2) 234 pp

Thiruvengadachari S. & R. Sakthivadivel. 1997. Satellite remote sensing for assessment of irrigation system performance: a case study in India, Research Report 9, International Irrigation Management Institute, Sri Lanka

Appendix: Satellite-based rice yield estimation in the 1996-97 rabi season

Since crop cutting experiment data (in which rice yields are estimated on the ground from statistically sampled plots) for 1996-97 were not available during the study period, the satellite-based rice estimation model developed for the 1992-93 rabi season was updated for the 1996-97 rabi season through satellite data normalization procedures.

The existing rice yield model, which is based on IRS satellite LISS-I sensor data of the 1992-93 rabi season, was normalized for LISS-I data of the 1996-97 rabi season by comparing the two sets of data for stationary targets whose reflectance remains unchanged. The difference in normalized difference vegetation index value was found to be only 3 to 4 digital counts, which is not significant; hence the 1992-93 model can be used unchanged for the LISS-I data of 1996-97. The LISS-I-based model was then normalized for LISS-III data of the same season by regressing the vegetation index of concurrent LISS-I and LISS-II data of 7 April and 10 April 1997. The rice yield model is now normalized for the LISS-III data of 7 April 1997. However, the model based on a single date (7 April 1997) needs to be updated for the vegetation index data of the milking stage of the rice crop, which would occur in different calendar periods across the command area due to staggered rice transplantation.

Time composite wide-field sensor data was generated by picking the maximum vegetation index value (corresponding to the milking stage of rice) from a set of co-registered 12 wide-field sensor data sets from mid-March to May 1997. The ratio of wide-field sensor-derived vegetation index data of 10 April 1997 to time composite normalized difference vegetation index data is computed and applied to LISS-III-derived vegetation index data, to compute the expected peak (time composite) LISS-III vegetation index value for use in the model.

The assumption that LISS-III and wide-field sensors are equally sensitive to vegetation response was tested. The ratios of LISS-III vegetation index data of 10 April and 4 May were computed, as were the ratios of wide-field sensor data on the same dates, and both sets were regressed with each other. The slope of regression line is 0.96, indicating equal sensitivity of LISS-III and wide-field sensors to vegetation response.

Since crop cutting experiment data for the 1996-97 rabi season was not available, the updated rice yield model was validated through farmer enquiries and enquiries with the Bhadra Command Area Development Authority and its irrigation engineer. In an earlier study, a similar category of rice yield model of the 1992-93 to the 1993-94 rabi seasons has however been validated against actual crop cutting experiment data of the latter season, with deviations from predicted to actual being less than 10 percent (Murthy et al, 1995).