An economic model of adaptive irrigation system management : subsistence farming in Sri Lanka

Abstract

The major purpose of this study is to develop an adaptive irrigation management policy for the Uda Walawa project in Sri Lanka. The policy is formulated by applying a linear programming algorithm that utilizes forecast information on a 20-week period of the dry season. The policy is updated at every decision stage using the most current actual and forecast information about reservoir inflows. Eleven forecasting models are developed in this study to generate expected information about reservoir inflows. Four of these are used for the development of an Adaptive Decision Rule (ADR). The four models are (1) a vector autoregression model with log transformed weekly data, (2) an historical average model based on the past seventeen years' inflow records, (3) a composite forecast model (VAR and averages) with Bayesian priors, and (4) a model based on one standard deviation reduction in the composite forecast. Three years-- the maximum reservoir inflow year during that 20-week period, the minimum reservoir inflow year during the same period, and the year 1985-- are used to examine the performance of the ADR. The composite forecast model, adjusted by subtracting one standard deviation from its mean forecast in each period performed very well relative to the other methods for minimum inflow year; while it offered quite reasonable performance in the other two years. The other objective of this study was to examine the present institutional structure of water management activities in the Uda Walawa project. A field survey was conducted and the Generalized Least Square method was used to analyze the survey data. The results showed that fertilizer, labor, agro-chemicals, and an interaction term between delivery and field canals were positively related to the paddy yield. Land size, off-farm income, experience in farming activities, and the distance from the field and delivery canals were negatively related to the paddy yield. Further, the regression results suggest that, under static conditions, reallocation of water may increase the total paddy yield for the region.