Informations et ressources scientifiques
sur le développement des zones arides et semi-arides

Accueil du site → Doctorat → Australie → 1993 → Predicting daily runoff using catchment water balance models with stochastic data

University of Wollongong (1993)

Predicting daily runoff using catchment water balance models with stochastic data

Bin Baki Aminuddin

Titre : Predicting daily runoff using catchment water balance models with stochastic data

Auteur : Bin Baki Aminuddin

Université de soutenance : University of Wollongong

Grade : Doctor of Philosophy (PhD) 1993

In water resources design, a long record of runoff is desirable in order that many possible runoff variations occur, including flooding and extreme droughts. However, available runoff records are usually short in length. Records of daily rainfall are much longer compared to runoff records. If long runoff records are required, several methods can be used including : stochastic runoff data generation ; and stochastic rainfall data generation followed by the use of this rainfall data in rainfall-runoff models to generate runoff data. This latter approach was adopted in this research because more variations of rainfall and runoff conditions could be included due to the longer rainfall record. The stochastic runoff data generation was not adopted because it only uses runoff variations available in the short runoff record.
In order to generate synthetic runoff data using statistics of rainfall data, the stochastic behaviour of rainfall data and the relationship between rainfall and runoff must be analysed.
The first phase of this research project involved a study of rainfall and streamflow (runoff) in natural catchments. Catchments used in this study were selected from the Illawarra Region, which include : Kangaroo River at Hampden Bridge, Macquarie Rivulet at Albion Park, Shoalhaven River at Kadoona, Mongarlowe River at Mongarlowe, Corang River at Hockeys, Endrick River at Nowra Road and Bungonia Creek at Bungonia.
Rainfall and runoff form part of the water balance in the catchment’s hydrologic cycle. Several rainfall and runoff models were used ranging from simple to more realistic and complicated ones. The models were : Diskin’s Linear Model, the SFB Model, the original and modified Boughton Models, the original and two modified versions of the Soil Dryness Index (SDI) Model and the Antecedent Precipitation Index (API) Model. From the results of rainfall-runoff modelling, Kuczera’s SDI Model was found to be the most satisfactory model for all catchments analysed.
The second phase of this study involved a study of stochastic daily rainfall generation. Rainfall is the basic weather variable in this study, being the main input data of all the models mentioned above. Stochastic rainfall data was produced using the statistics of the recorded rainfall (stochastic data generation). Several models were used for this stochastic rainfall generation : Lag-one Markov Chain Model, Two-Step Markov Chain with Gamma Distribution Model and the Transition Probability Matrices Model. From the results obtained, the Transition Probability Matrices Model was found to be the most satisfactory daily rainfall stochastic data generation model.
The ability of the stochastic rainfall generation model to generate data for the future was tested by dividing the available rainfall data into two samples : the Earlier Period and the Later Period (during which, concurrent rainfall and runoff data are available). Daily rainfall data were generated for the Later Period, using statistics of data from the Earlier Period (i.e. the Split Sample method). This technique was generally satisfactory using the adopted data generation model.
In the third and final phase of this study, the adopted rainfall and runoff model was then used to generate synthetic runoff data using the stochastic rainfall data (generated using the Later Period’s statistics). From the results of runoff data generation using this combination of rainfall-runoff and stochastic models, successful data generation was achieved.
In this research, the application of a combination of the two approaches (stochastic generation of rainfall and deterministic rainfall-runoff modelling) to generate synthetic runoff data for the future, was tested. Using the Split Sample technique for daily rainfall data generation, stochastic daily rainfall data was generated for the future period, and this data was then used in the rainfall-runoff model to generate synthetic runoff data. Tests of this runoff data indicated that on most occasions, this technique was satisfactory. Therefore, for natural catchments in the Illawarra Region, combinations of a stochastic daily rainfall generation model and deterministic rainfall-runoff model can be used to generate synthetic runoff data for water resources design.


Lien vers la version intégrale (15,9 Mb)

Page publiée le 25 novembre 2014, mise à jour le 3 mai 2019