Stochastic assessment of Phien generalized reservoir storage-yield-probability models using global runoff data records

Adebayo Adeloye, Soundharajan Bankaru Swamy, Chuthamat Chiamsathit

Research output: Contribution to journalArticle

50 Downloads (Pure)

Abstract

This study has carried out an assessment of Phien generalised storage-yield-probability (S-Y-P) models using recorded runoff data of six global rivers that were carefully selected such that they satisfy the criteria specified for the models. Using stochastic hydrology, 2000 replicates of the historic records were generated and used to drive the sequent peak algorithm (SPA) for estimating capacity of hypothetical reservoirs at the respective sites. The resulting ensembles of reservoir capacity estimates were then analysed to determine the mean, standard deviation and quantiles, which were then compared with corresponding estimates produced by the Phien models. The results showed that Phien models produced a mix of significant under- and over-predictions of the mean and standard deviation of capacity, with the under-prediction situations occurring as the level of development reduces. On the other hand, consistent over-prediction was obtained for full regulation for all the rivers analysed. The biases in the reservoir capacity quantiles were equally high, implying that the limitations of the Phien models affect the entire distribution function of reservoir capacity. Due to very high values of these errors, it is recommended that the Phien relationships should be avoided for reservoir planning.

Original languageEnglish
Pages (from-to)1433-1441
Number of pages9
JournalJournal of Hydrology
Volume529
Issue numberPart 3
Early online date24 Aug 2015
DOIs
Publication statusPublished - Oct 2015

Keywords

  • Generalised storage-yield function
  • Phien models
  • Reservoir capacity
  • SPA

ASJC Scopus subject areas

  • Water Science and Technology

Fingerprint Dive into the research topics of 'Stochastic assessment of Phien generalized reservoir storage-yield-probability models using global runoff data records'. Together they form a unique fingerprint.

  • Cite this