Abstract:
The concepts of seasonal groups and neural networks and their characteristics are the focus of this paper in estimating missing values in monthly streamflows. The group approach recognizes the utility of associative and distributive properties of data points at local and global levels across the data series. At the local level, the associative properties are identified and used in the formation of groups; whereas, at the global level the distributive properties across the data series are recognized and used in the formation of group-clusters. The formation of groups and group-clusters enhances the extraction and utilization of information content of the data set and, thus enhances the development of effective data infilling methods and techniques. Efficacy of the approach for data infilling in monthly streamflow time series has been demonstrated with reasonable degree of success through applications to five rivers across Canada.
