A procedure was developed to estimate optimum irrigation requirements for lowland rice cultivation in Southeast Asia. The procedure uses a water balance equation of semi-stochastic nature to maintain minimum desired water depths in paddy fields at the end of each irrigation period. The procedure estimates weekly pan evaporation (EV) and rainfall (RF) at different probability levels, which is then used to determine weekly irrigation requirements at different probability levels.

To illustrate the use of the method, the Kalawewa irrigation scheme in Sri Lanka was selected for demonstration purposes. Different transformations were applied to Rf and EV data in an attempt to normalize these variates and to obtain a unique distribution to describe their variations. Statistical analysis of weekly EV arid RF showed that the power transformation was best able to transform the weekly RF and EV data to normality.

Comparison of the use of the model and current system practices showed that a significant amount of water could be saved even when the system was operated at high probability levels (90% reliability). The irrigation water required when the system was operated at the 72% probability level was about 21% less than the amount required when the system was operated at 90% probability level during some weeks.

The EXTRAN flow routing model was used to simulate water flow in the upper reaches of the main canal system for varying discharges at the head gate each day. The simulated water depths were used to determine the gate settings required at the turnout structures to divert the desired amount of irrigation water into the turnout areas.

The flow simulation for the demonstration area, showed that it was not possible to regulate irrigation water from the main reservoir to meet daily demands at all the turnouts. This was due to the large distances between the regulating reservoir and turnouts that caused appreciable time lag for the flow to reach the turnouts farthest from the regulating reservoir.