Generalized models for the periodic components of monthly flows

Either observed values of periodic parameters or their fitted functions of time are essentially used at the deseasonalization stage of the mathematical modeling of periodic stochastic hydrologic processes. Despite the fact that these parameters have great importance in simulation and synthetic data generation studies, no attempt on regionalization of these parameters has been met in the literature. In the scope of this study, five alternative approaches for the regionalization of within the year periodicity in the basic parameters of monthly streamflows in a given homogeneous hydrologic subregion have been treated in two broad categories, namely as nonfunctional (or nonparametric) and functional (parametric) approaches. All of the approaches are simply based on the regionalization of the standardized (or normalized) values of monthly periodic parameters or on the regionalization of their dimensionless fitted harmonic functions. Hydrologically similar (homogeneous) sub regions were identified by applying the hierarchical clustering procedure with average linkage based on the Euclidean metric distances in terms of standardized values of the periodic parameters. At-site and regional relative performances of regionalization approaches were assessed by means of the dimensionless average relative bias and the root mean square error of each regionalization procedure. The numerical results of the application on 30 gauges, in two major river basins (Seyhan and Ceyhan) of South Turkey revealed that both type of regionalization procedure may produce reliable regional estimates for monthly flow time series especially at locations where monthly records are not available or they are too short. Taking into account the fact that non functional approach needs much more parameters, the functional approaches seem much more plausible. Keywords: monthly flows, generalized models, periodic parameters, regional homogeneity