Sediment yield prediction based on analytical methods and mathematical modelling

Msadala, V. P. (2009-12)
 

Thesis (MScEng (Civil Engineering)--University of Stellenbosch, 2009.

Thesis

ENGLISH ABSTRACT: A study of the state of reservoir sedimentation in South Africa based on reservoir sediment deposit data, has shown that a considerable number of reservoirs have serious sedimentation problems. The analysis of the reservoir sediment deposit data showed that almost 25% of the total number of reservoirs have lost between 10 to 30% of their original storage capacity. The average storage loss due to sedimentation in South African reservoirs is approximately 0.3% per year while the average annual storage loss for all the reservoirs in the world is 0.8%. The aim of this research was to develop sediment yield prediction methods based on analytical approaches and mathematical modelling. The sediment yield prediction methods can be used in planning and management of water resources particularly in reservoir sedimentation control. The catchment erosion and sediment yield modelling methods can be applied in temporal and spatial analysis of sediment yields which results are essential for detailed design of water resources, particularly in the identification of critical erosion areas, sediment sources and formulation of catchment management strategies. Current analytical methods for the prediction of sediment yield have been reviewed. Nine sediment yield regions have been demarcated based on the observed sediment yields and catchment characteristics. Empirical and probabilistic approaches were investigated. The probabilistic approach is based on analysis of the observed sediment yields that were calculated from reservoir sediment deposit, river suspended sediment sampling data and soil erodibility data. The empirical equations have been derived from regression analysis of the variables that were envisaged to have a significant effect on erosion and sediment yields in South Africa. Empirical equations have been developed and shown to have accurate and reliable predictive capability in six of the nine regions. The probabilistic approach has been recommended for the prediction of sediment yields in the remaining three regions where reliable regression equations could not be derived. The predictive accuracy of both the probabilistic and empirical approaches was checked and verified using the discrepancy ratio and graphs of the observed and calculated data. While the analytical methods are needed to predict the sediment yield for the whole catchment, mathematical modelling to predict sediment yields is applied for more detailed analysis of sediment yield within the catchment. An evaluation of available catchment sediment yield mathematical modelling systems was carried out. The main criteria for the choice of a numerical model to be adopted for detailed evaluation was based on the following considerations: the model’s capabilities, user requirements and its application. The SHETRAN model (Ewen et al., 2000) was therefore specifically chosen because of its ability to simulate relatively larger catchment areas (it can handle catchment scales from less than 1km2 to 2500km2), its ability to simulate erosion in channels, gullies and landslides, its applicability to a wide range of land-use types and ability to simulate land use changes. Another model, ACRU (Smithers et al., 2002) was also reviewed. The aim of the model evaluation was to provide a conceptual understanding of catchment sediment yield modelling processes comprising model set up, calibration, validation and simulation. The detailed evaluation of the SHETRAN model was done through a case study of Glenmaggie Dam in Australia. The flow was calibrated and validated using data from 1975 to 1984, and 1996 to 2006 respectively. The results for both the calibration and validation were reasonable and reliable. The sediment load was validated against turbidity derived sediment load data from 1996 to 2006. The model was used to identify sources of sediment and areas of higher sediment yield. The land use of a selected sub-catchment was altered to analyse the impact of land use and vegetative cover on the sediment yield. Based on the results, the SHETRAN model was confirmed to be a reliable model for catchment sediment yield modelling including simulation of different land uses.

AFRIKAANSE OPSOMMING: ‘n Studie van die stand van damtoeslikking in Suid-Afrika toon dat daar ernstige toeslikkingsprobleme by baie reservoirs bestaan. ’n Ontleding van die toeslikkingsyfers gegrond op damkomopmetings toon dat omtrent 25% van die totale getal reservoirs tussen 10 en 30% van hulle oorspronklike opgaarvermoë verloor het. Die gemiddelde tempo van damtoeslikking in Suid-Afrika is 0.3%/jaar, wat laer is as die wêreld gemiddeld van 0.8%/jaar. Die oogmerk met hierdie navorsing was om sedimentlewering voorspellingsmetodes te ontwikkel deur gebruik te maak van analitiese metodes en wiskundige modellering. Die sedimentlewering voorspellingsmetodes kan gebruik word vir die beplanning en bestuur van waterbronne en veral vir damtoeslikking beheer. Die opvangsgebied erosie en die sedimentlewering modelleringsmetodes kan toegepas word in tydveranderlike en ruimtelike ontleding van sedimentlewering. Hierdie inligting word benodig vir die detail ontwerp van waterhulpbronne en veral vir die identifisering van kritiese erosiegebiede, bronne van sediment en die formulering van opvangsgebied-bestuur strategië. ‘n Literatuuroorsig oor die huidige metodes vir die voorspelling van erosie en sedimentlewering is gedoen. Nege sedimentasie streke is afgebaken in Suid-Afrika, gegrond op waargenome damtoeslikkingsdata en opvangsgebied-eienskappe. Proefondervindelike en waarskynlikheidsbenaderinge is ondersoek. Die waarskynlikheidsbenadering is gegrond op die ontleding van waargenome damtoeslikking wat bereken is uit reservoir opmeting data en rivier gesuspendeerde sediment data, asook data oor gronderosie. Die proefondervindelike metode se vergelykings is afgelei vanuit regressie ontleding van die veranderlikes wat ‘n belangrike invloed het op die erosie en sedimentlewering in Suid-Afrika. Daar is bevestig dat die ontwikkelde proefondervindelike (empiriese) vergelykings ‘n akkurate en betroubare voorspellingsvermoë in ses van die nege streke het. Die waarskynlikheidsbenadering is aanbeveel vir die voorspelling van sedimentlewering in die ander drie streke, waar betroubare regressie vergelykings nie afgelei kon word nie. Die voorspellingsakkuraatheid van albei metodes is nagegaan en bevestig deur gebruik te maak van die teenstrydigheidsverhouding en grafieke van die waargenome en berekende data. Analitiese metodes van sedimentleweringsvoorspelling is nodig vir ‘n volle opvangsgebied, terwyl wiskundige modellering om sedimentlewerings te voorspel gebruik kan word om ‘n meer in diepte ontleding van die sedimentlewering binne ‘n opvanggebied te doen. ‘n Evaluasie van beskikbare wiskundige modelle wat opvangsgebied sedimentlewering kan voorspel, is gedoen. Die hoofkriteria vir die keuse van ‘n model vir gebruik by gedetailleerde ontleding is gegrond op die volgende: die vermoëns van die model, wat verbruikers benodig en die aanwending van die model. Die SHETRAN model (Ewen et al., 2000) is spesifiek gekies weens sy vermoë om relatief groter opvangsgebiede te simuleer (dit kan opvangsgebiede van 1km2 tot 2500km2 wees) asook om erosie in kanale, dongas en grondverskuiwing simuleer. Dit kan toegepas word op ‘n wye reeks grondtipes en kan ook die gevolge simuleer as die gebruik van die grond verander. ‘n Ander model, ACRU (Smithers et al., 2002) is ook ondersoek. Die doel van die modelevaluering was om ‘n konseptuele begrip te kry van sedimentlewering modelleringsprosesse wat die opstelling, kalibrasie, toetsing en simulasies insluit. Die volledige evaluasie van SHETRAN is gedoen deur middel van ‘n gevalle-studie van die Glenmaggiedam in Australia. Die riviervloei is gekalibreer en getoets deur gebruik te maak van data wat strek van 1975 tot 1984, en van 1996 tot 2006 onderskeidelik. Die resultate van beide die kalibrasie en die toetswas redelik en betroubaar. Die sedimentlading is gekalibreer teen velddata van 1996 tot 2006. Die model is gebruik om bronne van sediment te identifiseer, asook gebiede met ‘n hoër sedimentlewering. Die gebruik van die grond op ‘n gekose sub-opvangsgebied is verander om die impak van grondgebruik en plantbedekking op sedimentlewering te ontleed. Die resultate bewys dat die SHETRAN model ‘n betroubare model is vir groot opvangsgebied sedimentlewering modellering, asook vir die simulasie van verskillende grondgebruike.

Please refer to this item in SUNScholar by using the following persistent URL: http://hdl.handle.net/10019.1/2863
This item appears in the following collections: