Optimal reservoir operation for irrigation

Abstract

Irrigation is an important purpose for which most reservoirs are operated in developing countries. There is an increasing awareness among irrigation planners and engineers to plan and operate reservoir systems at maximum efficiency and derive maximum benefits from them. As a result, considerable work has been done on reservoir operation for known total irrigation demand on the one hand, and on optimal allocation of available water at the farm level on the other. However, very few studies have attempted to derive optimal reservoir operating policies by integrating reservoir operation with on farm water utilisation for various crops. The present study develops a model for reservoir operation toward this goal and demonstrates its applicability through a case study of an existing reservoir system in Karnataka State, India. A three state variable Stochastic Dynamic Programming (SDP) model has been developed to obtain a steady state optimal operating policy, which specifies the optimal release in any intraseason period for known values of: • initial storage • inflow • initial average soil moisture in the command area for a multiple crop context. The random nature of inflows is incorporated through a one step Markov process. Decision intervals are chosen based on crop growth stages and typical irrigation decision timing in the field. Optimal reservoir releases are based on detailed information about: • individual crops • soil moisture storage characteristics Thus, crop yield response to water deficit and soil moisture dynamics are explicitly accounted for in determining water allocations and reservoir releases. Crop water requirements also incorporate time varying root growth. When the reservoir release (after accounting for losses) is insufficient to meet the irrigation requirements of all crops during a period, competition for water arises. The model allocates water among crops such that the adverse effect of water deficit on yields is minimised, while simultaneously determining the optimal reservoir release. The SDP model has been applied to the Malaprabha Reservoir Project in Karnataka, using a 35 year inflow record to derive the optimal steady state operating policy. Major crops considered include: • Kharif: Maize, Pulses, Sorghum • Rabi: Sorghum, Pulses, Wheat, Safflower • Two season crop: Cotton The implications of the optimal policy were evaluated using a simulation model, which determines: • actual reservoir releases based on the SDP policy • actual allocations to individual crops per decision interval • relative yields of crops at the end of each year Three indicators of performance-considered important for irrigation managers-were used: 1. Reliability (of irrigation water supply) 2. Resiliency 3. Productivity index (based on crop yield) Four different policies were compared through these indicators. The Thomas-Fiering model was used for generating synthetic streamflows required for simulation. For real time implementation of the optimal operating policy, a simple method was adopted to estimate the current period inflow class interval based on its expected value. The models developed in this study are general and may be applied to obtain optimal operating policies and evaluate their implications for any irrigation reservoir system. The thesis consists of five chapters: • Chapter I: Motivation and objectives of the study • Chapter II: Review of literature on reservoir operation and irrigation allocation models • Chapter III: Development of the optimal operating policy model using Stochastic Dynamic Programming • Chapter IV: Development of the simulation model, case study application, and policy implication analysis • Chapter V: Conclusions from the study