RUDIN HOXHA
Abstract
The primary objective of this study is the application of advanced statistical methods, based on empirical data, to design optimal histograms for the water inflows of the three hydropower plants of the Drin Cascade. The focus of this paper is not on identifying a universally superior method, but on selecting the most appropriate technique depending on the specific characteristics of the data and the physical phenomena of each basin. For Fierza and Koman, Knuth’s method proved to be efficient, producing optimal histograms that are statistically accurate and easily interpretable. In contrast, for Vau i Dejës, where the data distribution presents high complexity, the adaptive Bayesian Blocks method emerged as the appropriate solution to capture details that standard methods fail to highlight. Based on the optimal histograms, Probability Density Functions (PDFs) and Cumulative Distribution Functions (CDFs) were constructed for each basin using the Piecewise Linear approach. For seasonal trend analysis, Nadaraya-Watson Kernel Regression with periodic wrapping was applied, enabling the construction of annual hydrographs that clearly identify the nival and pluvial regimes of the cascade.
Key words : Hydrology, Drin Cascade, histogram optimization, Bayesian blocks, kernel regression.