USE OF A SIMULATION MODEL TO IMPROVE THE MANUAL OPERATION PROCEDURES OF AN IRRIGATION CANAL (KIRINDI OYA - SRI LANKA)

USE OF A SIMULATION MODEL TO IMPROVE THE MANUAL OPERATION PROCEDURES OF AN IRRIGATION CANAL (KIRINDI OYA - SRI LANKA) Jacques REY 1 Pierre-Olivier MALATERRE 2 - Jean-Pierre BAUME 3 SUMMARY Following a presentation of the operational challenges faced by the canal managers, the use of a simulation model to improve the manual operation procedures of an irrigation canal is detailed. The pilot field site for this IIMI-CEMAGREF collaboration is the Kirindi Oya Right Bank Main Canal (Sri Lanka). The simulation model RBMC is a fully interfaced hydrodynamic model solving complete Saint-Venant equations. Its first applications provided detailed information on canal behaviour and a diagnosis on the present manual operation procedures. Its integration in a broader Management Information System provided canal managers with a decision-support package helping them to improve operational procedures. These improvements contributed to increase the water use efficiency. INTRODUCTION Various types of mathematical models have been developed during the past 20 years to improve the water management of irrigation systems. Most of them were used by specialists to design irrigation canals, compute crop requirements or test operational procedures. Nevertheless, except for simple water allocation models, few of them were successfully integrated in the real operational management process of irrigation agencies at the field level. High data requirements, and lack of 1 : Associate Irrigation specialist - International Irrigation Management Institute (IIMI, Sri Lanka) 2 : Research Engineer - Irrigation Division, Centre National du Machinisme Agricole du Génie Rural des Eaux et des Forêts (CEMAGREF, Montpellier, France) 3 : Research Engineer - Irrigation Division, CEMAGREF (Montpellier, France) 1

dedicated user interfaces were the main weaknesses that undermined many attempts for transferring these computer tools to the end users. In 1987, a collaborative activity was initiated between IIMI 1 and CEMAGREF 2, to prospect the possibility of providing irrigation canal managers with a decision support tool likely to help them to better understand the hydraulic functioning of their systems. The Kirindi Oya Right Bank Main Canal in southern Sri Lanka was selected as the pilot field site. This paper presents the main phases of this project. The first phase consisted mainly in the development of a mathematical model called RBMC 4 that integrates hydrodynamic simulation models. During the second phase, simulations done with RBMC provided information on canal behaviour and helped in diagnosing weaknesses in current operational procedures. Additional modules were then developed to better integrate RBMC as a support tool in the decision-making process of the canal managers. 1 OPERATIONAL CHALLENGES The Kirindi Oya Right Bank Main Canal is a 30 km long earthen canal. Its design head discharge is about 10 m 3 s -1, its bed width varies from 10 m to 3 m and its average slope is 35 cm per km. The water levels are maintained all along the canal by means of 18 gated cross regulators. Water is delivered to the secondary network through 42 gated offtakes structures. The command area of the canal amounts to 3,300 hectares. It is divided into 5 subdivisions called tracts, each varying from 500 to 1,000 hectares. The water storage at the head is ensured by the Lunugamvehera reservoir, whose capacity is 200 MCM 5. This reservoir supplies water to another main canal (left bank) increasing the total command area of the irrigation system to approximately 10,000 hectares. Due to a severe hydrological deficit in the area, the irrigation agency in charge of the scheme finds it difficult to ensure a cropping intensity higher than 100 %. Even if, during the 4 Right Bank Main Canal 5 Million cubic meters. 2

wet season or Maha (from October to March), most of the command area is under cultivation, it is difficult to ensure adequate water deliveries during the dry season or Yala (from April to August). As a response to the water scarce conditions, diversified cropping patterns were promoted in order to progressively replace paddy cultivation with crops less demanding in water, on the well drained soils of the command area. The consequent water delivery patterns require more sophisticated operational plans (intermittent irrigation). Thus, reinforcing water conservation practices at the main canal level while experiencing new operational scenarios became a challenging concern for the Kirindi Oya canal managers. 2 USE OF SIMULATION MODELLING FOR DIAGNOSIS PURPOSES 2.1 Description of RBMC Two main parts of the terms of reference of the RBMC simulation model were the detailed description of the physical system (canal geometry, offtake and cross regulators, etc.) and the computation of hydraulic variables (discharge, water elevation, flow velocity, etc.) under both steady and unsteady flow conditions. Three CEMAGREF programs (TALWEG, FLUVIA and SIRENE) available on mainframe computers were used to design RBMC. The latter is therefore divided into three units: - the Topography unit checks the inputted topographic data of the canal and generates the data files used by the other two units; - the Steady Flow unit calculates the water surface profile as well as the offtake and cross regulator gate openings for any set of water demands; - the Unsteady Flow unit simulates the flow conditions in the canal after modifications of internal and external boundary conditions: gate settings, water 3

inflows and outflows. Thus, before implementing a new operational procedure in the field, the user can evaluate its likely consequences on the system. A major component of the project was the development of a user-friendly interface for non-specialised personnel from IIMI and ID 6. On-line help is available and data input is fully interactive. Through data editors, users give all the required information for the hydraulic computation. Graphic and tabulated outputs are designed to provide operational assistance to the canal managers. In particular four indicators are computed at each offtake in order to assess the distribution efficiency and timeliness. RBMC's features and outputs meet the needs of different types of users with various interests in using the model, and different levels of practical experience and theoretical knowledge in hydraulics. The user interface handles all the features of the simulation model, managing spatial and temporal information. Modules linked to the unsteady flow unit allow the simulation of any operational rules (either manual or automatic) at moving structures (main sluice, cross regulators, offtakes), as described in the next chapter. 2.2 Simulation of manual operation After calibration of RBMC, IIMI and CEMAGREF used it to study design and operational issues, together with the canal managers. Maximum capacity, overtopping locations (Sally, 1988-89) and influence on water deliveries of water fluctuations in the main canal (Malaterre, 1989) were identified. RBMC was also used to simulate and analyse manual operations, and compare theoretical and real procedures (Malaterre, 1989 and Rey, 1990). The first step was to observe the gate keepers and to understand the cross regulators and offtakes operation. Previous reports and studies were analysed, and data were collected by IIMI through data loggers in 1988. The number and frequency of operations as well as the magnitude of oscillations of the water surface in the main canal were determined through simple monitoring 6 Irrigation Department of Sri Lanka 4

performed by IIMI during the Yala (or dry) season in 1988. This information was completed by field visits in subsequent seasons. Data collection focused especially on the method used to evaluate the magnitude of the operation at the cross regulators in order to be able to simulate it on RBMC. This information was gathered through discussions with the operators and direct measurements at the regulators. Finally, for each cross regulator, the following parameters were determined: - frequency of operation, - duration of an operation, - the time at which work begins, - duration of a working day, - water elevation margin around the target, - maximum and minimum opening and closing during one operation, These parameters and additional rules were introduced in a regulation module linked to the Unsteady Flow unit. This module was tested on several scenarios identified as problems faced by the canal managers. One of them is the transition between two steady states (for example an increase of the main sluice discharge). Simulations proved that the time required for stabilisation was around 2 days, which was confirmed by field observations. Such a long duration was shown to be due mainly to mutual interactions between adjacent cross regulators (1 cross regulator every 1.5 km). Lack of information between gate keepers disturbed operations at regulators and final gate positions were attained only after a long trial and error procedure. New gate positions and optimum time for adjustment can be computed with RBMC. With this improved procedure, the new steady state can be reached much faster (less than one day) and with less operations than with the present practices. These first applications of the simulation model increased the knowledge on the operational procedures performed on the canal. RBMC highlighted problems and difficulties encountered by the canal managers and provided a diagnosis on the canal management (operation but also maintenance and design). Nevertheless, to become operational, RBMC itself is not sufficient. New data, communication flows and processes have to be introduced in the decision-making process. 5

3. A DECISION SUPPORT PACKAGE 3.1 Which management intervention? Based on the operational challenges faced by the system, and on the problems of manageability created by uncoordinated operations, an intervention aimed at rationalising the decision-making process at the main canal level was envisaged in Kirindi Oya. As mentioned previously, the driving motivation for improvement was clearly identified as the need for efficient management of an extremely scarce water resource. Under such conditions, key areas of potential improvements concern both the planning stage (choice of crops and cultivation calendar) and the operational stage (water application methods at the field level, water conveyance and distribution management). Among those, main canal management was identified as a critical factor in performance deficiencies and became the focus of the joint intervention by the ID, IIMI and CEMAGREF. 3.2 Identification of performance drivers To study the decision-making process at the main canal level, the quality of the communication system between the field staff and the canal managers, the indicators used as management signals, and the decision-making rules and procedures had to be assessed. This diagnosis was thus articulated around three main questions: Are management decisions based on reality? The actual deliveries at the head of the secondary canals were insufficiently controlled by the manager, due to the lack of measurements and communication along the canal and water wastage occurred by high reaction time lag. Thus, providing him with a timely and reliable "picture" of his system was identified as a prerequisite condition of management improvement. Are management decisions based on performance analysis? Under water scarce conditions, a key indicator lies simply in the computation at regular intervals of the volume delivered to the different secondary canals. The 6

maintenance of a small database permitting easy access to this indicator (and potentially others) appeared as an interactive means to promote an optimisation attitude at the decision-making stage. Is the logic of management decisions technically sound? The hydraulic behaviour of open channel systems with cross regulators is sometimes far from intuitive and requires a fairly long experience to be completely mastered. In order to be performed efficiently, critical phases of operational management (implying changes of the main discharge) required additional procedures. Operational plans, created by simple decision support modules, were tested with RBMC. This diagnosis phase did not address the issue of suitability of the hardware (modernisation and rehabilitation) and deliberately concentrated on potential improvements by a better management of the existing system. 3.3 Suggested Management Information System (MIS) Following the previous recommendations, a frequent data collection system and communication network were created and are currently being used. Essentially, basic data on water levels and gate openings are collected at all structures operated along the main canal, including gated regulators and offtakes. These data are recorded twice a day by the gate keepers and summarised in a standard format which is handed over on the same day to the canal manager. These new data are converted into discharges, displayed in the manager's office on a board and stored in a computerised database. The complete set of messages presently in use in Kirindi Oya is presented in Figure 1. The boards display the discharges at the offtakes (in cusecs) stimulating timely control of deliveries. The charts summarise the cumulative volumes (in feet) issued at the offtakes, per tract and for the whole canal, focusing the attention of the manager on the key strategic objective of water use efficiency. 7

Field Books Field Books Main Sluice Issues Water Levels HYDRAULICS M INSTRUCTIONS Gate Operations Main Sluice Issues Offtakes - Discharges - Instructions Board (Canal Manager Office) Boards (Field Offices) I Regulators - Guidelines Cultivated Area SCHEDULING S EVALUATION Tract Level Charts (Field Offices) Water Allocation Main Canal Level Charts (Canal Manager Office) Figure 1 - Management Information System Messages This database reflects the information conveyed by the messages and comprises data on canal structures and topography, hydraulic trajectories at the structures (levels, openings, discharges) and some contextual information (areas cultivated, rainfall). Parallel to this first set of activities, the interface between the previous database, hydraulic modules (spreadsheets) and RBMC were developed to allow further improvements in the decision-making process. These modules provide the manager with guidelines for better understanding the hydraulic functioning of the canal, in order to prepare coordinated operational plans at the cross regulators when needed (rotations among tracts, rainfall, start of issues) and test these plans prior to implementation. 4. PRELIMINARY RESULTS IN KIRINDI OYA 4.1 Strengthening of monitoring activities The first part of the management intervention recommendation (introduction of a communication system and of a database) was implemented during the 1991/92 wet season in Kirindi Oya. As a result of new monitoring activities, water distribution at the offtakes was roughly looked into by the manager on a daily basis and control 8

actions were performed to conserve water whenever possible. Though difficult to link exclusively to the management intervention, a significant reduction of water consumption was observed during this season (0.2 m on average). At the same time, interviews at the field level revealed an increased accountability of field staff and easier operating tasks due to the greater responsiveness of the manager to the measured field conditions. In addition to the main objective of providing the manager with a daily picture of water deliveries and regularly computing water consumption, the data collection system and associated database permitted a dramatic increase in the available knowledge on the canal hydraulic functioning, as illustrated below. For example, a systematic study of all steady flow periods occurring in the different reaches of the canal during the season was performed. A reach is considered under steady flow conditions when no significant changes in its inflow and outflow discharges, and upstream and downstream water levels occur during hydraulically representative period of time. Table 1 shows that the usual canal flow conditions were unsteady throughout the season: Table 1 - Steady flow periods REACHES (from Dam to Tail-end) NUMBER OF STEADY FLOW PERIODS (in 19 weeks) TOTAL DURATION OF STEADY FLOW PERIODS (Hours) % TIME OF STEADY FLOW PERIODS 1 12 357 12 2 28 801 27 3 26 559 19 4 6 136 12 5 24 441 15 7 18 297 10 11 33 835 29 13 24 392 14 15 2 39 3 17 10 155 5 A second field of investigation was the hydraulic functioning of the different structures (cross regulators and offtakes) and their related achievement with regard to water delivery targets. For this purpose, a set of 6 indicators was systematically computed for all offtakes along the canal: 9

Three performance indicators: P1, P2 and P3. These indicators (Molden and Gates, 1990) explicitly take into account the targets formulated by the manager in terms of water deliveries. P1: measure of adequacy (indicative of the notion "enough supply"), P2: measure of efficiency (indicative of the notion "no waste"), P3: dependability (indicative of the notion "timeless of the deliveries"). Three descriptive indicators: D1, D2 and D3. These indicators aim at describing hydraulic behaviour at the structures without explicit reference to managers' targets. D1: total number of gate adjustments performed at a given structure, D2: average submergence at the structure (ratio of the downstream level over the upstream level), D3: average level fluctuation upstream of the structure (standard deviation of the upstream level). The set of performance indicators was first used to detect weaknesses in the water delivery process. The quality of water deliveries at one structure is highly dependent on its location along the canal and on the capabilities of field staff in charge of operating it. Discriminating amongst the different structures according to their achievement with regard to common objectives such as adequacy, efficiency and reliability of water deliveries is of great importance to ensure proper management control. For each of the three performance indicators, structures were separated in three groups: values below, equal to, or above the average, and received respectively the flag number -1, 0, and +1 (Cf. Table 2, e.g. for 5 structures). For further comparison, three main categories were finally created according to the value of the sum of the flags. Structures presenting a good water delivery profile could then be separated from those where problems occurred in meeting their targets during the season. Following the same approach, structures were classified according to their values of the descriptive indicators (Cf. Table 3, e.g. for 6 structures). In the process of diagnosing the potential difficulties faced by field staff at the implementation stage, simple information concerning the hydraulic conditions at the structures is useful. A 10

structure is considered as operating well if few operations are done, the degree of submergence is low, and few fluctuations of upstream water level occur. Table 2 - Flags according to performance indicators Tract Structure 7 P1 (Adequacy) P2 (Efficiency) P3 (Dependability) Summary 3 FC55 1-1 -1-1 3 DC1A 1-1 0 0 3 DC1B -1 1 1 1 4 DC1 0 1 1 2 1 FC6 1 1 1 3 Tract Structure D1 (N Operations) Table 3 - Flags according to descriptive indicators D2 (Submergence) D3 (Level fluctuations) Summary 4 FC37-1 -1-1 -3 4 DC1-1 -1 0-2 1 F56 0-1 0-1 5 DC1 1 0-1 0 2 DC9 1-1 1 1 3 FC55 1 0 1 2 This set of flags helps the manager in focusing his attention to critical structures of his system and eases his diagnosis. There is no general correlation between the delivery performance and hydraulic functioning of the offtakes as defined by our flags. The full set of 6 indicators is necessary to formulate an accurate diagnosis at one particular structure. 7 FC: Field channel - DC: Distributory canal 11

4.2 Gradual test of decision support modules Once the functioning of the canal system is known accurately and in a timely fashion, coordinated operational decisions can be studied. The objective presently pursued in Kirindi Oya, in the second part of the management intervention, is to test and gradually introduce optimisation procedures supported by the RBMC simulation model. This approach is currently tested during the 1992/93 wet season. Simple hydraulic information concerning the water volumes needed to fill certain reaches of the canal, the approximate time lags and diffusion along the canal are provided through spreadsheets to the canal manager who can utilise them to interactively design draft operational plans. Previous records of gate operations stored in the database can also be retrieved to support his decisions with past experiences. Once finalised, the plan consisting of a sequence of operations at the structures can be tested and evaluated through the RBMC simulation model. If considered satisfactory, the plan forms the basis for the implementation guidelines transmitted to the field staff; if not, the decision maker can revise it according to the weaknesses detected in the simulation results and re-test it as many times as needed. This optimisation process is only recommended for typical difficult phases of operational management. This approach is expected to gradually improve the managers' knowledge and confidence about the manageability of his system and dramatically speed up the learning process in case of a new engineer taking over the canal supervision. CONCLUSIONS The problems faced in operational water management are relatively well-known and various tools can be designed to assist managers of irrigation systems. Technical improvements have to be introduced with a full awareness of the existing management context. This involves a careful diagnosis of the field situation and usually leads to the design of a management intervention package instead of the introduction of a single, isolated tool. A real problem remains the gap which usually exists between the communication and measurement facilities of a system and the data requirements of many decision support tools. 12

In Kirindi Oya, a gradual approach, involving the upgrading of the data collection and communication system, the introduction of a simple computerised database and a flow simulation model, has been recommended to improve water use efficiency at the main canal level. The research program is still on-going but the first results indicate that this approach is most likely to be validated. REFERENCES BAUME J.P., CERTAIN F., MALATERRE P.O., REY J., SALLY H. "Utilisation d'un modèle informatique hydraulique pour améliorer la gestion des canaux d'irrigation. Application au canal de Kirindi Oya (Sri Lanka)", CIID Budapest - June 1992 CEMAGREF "RBMC Theoretical concepts and User's guide"- Internal report, Irrigation Division, CEMAGREF Montpellier, 1990 IIMI "Hydraulic modelling of a main canal system: The next best thing to buying a canal", IIMI Review, Vol. 2 n 2 - August 1988 MALATERRE P.O. "A study of the Kirindi Oya RBMC operations through an application of a mathematical flow simulation model: prospect of improving performance", CEMAGREF / ENGREF / IIMI - ENGREF internal report, October 1989 MOLDEN, GATES. Journal of Irrigation and drainage engineering - Vol. 116, N 6 Nov./Dec. 1990. REY J. "Field application of the Kirindi Oya RBMC simulation model", Internal Report, ENGREF Paris - August 1990 BAUME J.P., SALLY H., MALATERRE P.O., REY J., "Development and fieldinstallation of a mathematical simulation model in support of irrigation canal management", Research paper IIMI/CEMAGREF 1993. 13