Spatial Clustering of Irrigation Networks Using K-Means Method (Case Study of Ghazvin Irrigation Network)

Improving the performance of water conveyance networks is one of the key issues in saving limited water resources. The first step for this improvement is performance evaluation and then presenting the solutions. One of the practical and efficient approches for performance improvement is to extract the homogenous area out of the irrigation network based on the physical and technical features. The main idea behind this research is to present a quantitative benchmark for exploring homogenous areas with similar physical attributes and present the abilites of this method for a real case study. K-Means clustering algorithm, is applied to spatial clustering of irrigation networks based on physical attributes. Data was arranged based on the “objects” and the “features” in the matrix language. Ghazvin irrigation network data was used to form the input matrix. This matrix consisted of 162 rows and 5 columns. Using Davies and Bouldin (DB) index as the cluster validity index, it has been shown that the optimum number of clusters is 10. Each cluster represented a homogenous area in the irrigation network district. Clustering reduces the dimension of assessments from a large extended irrigation district to a limited number of homogeneous regions and provide a context for better and easier decision making, performance evaluation, and allocation of facilities and budget to different regions.