Full Text Available
Note: Clicking the button above will open the full text document at the original institutional repository in a new window.
A comparative analysis and improvement of the empirical design flood estimation approaches
Thesis (MEng)--Stellenbosch University, 2025.
Saved in:
| Main Author: | |
|---|---|
| Other Authors: | |
| Format: | Thesis |
| Language: | English |
| Published: |
Stellenbosch : Stellenbosch University
2025
|
| Subjects: | |
| Tags: |
No Tags, Be the first to tag this record!
|
| _version_ | 1867613741483294720 |
|---|---|
| access_status_str | Open Access |
| author | De Klerk, Barend Jakobus |
| author2 | Du Plessis, Kobus |
| author_browse | De Klerk, Barend Jakobus Du Plessis, Kobus |
| author_facet | Du Plessis, Kobus De Klerk, Barend Jakobus |
| author_sort | De Klerk, Barend Jakobus |
| collection | Thesis |
| dc_rights_str_mv | Stellenbosch University |
| description | Thesis (MEng)--Stellenbosch University, 2025. |
| format | Thesis |
| id | oai:scholar.sun.ac.za:10019.1/134588 |
| institution | Stellenbosch University (South Africa) |
| language | English |
| last_indexed | 2026-06-10T12:40:58.109Z |
| license_str | Other — see source repository |
| provenance_str_mv | Harvested via OAI-PMH from SUNScholar — Stellenbosch University Repository |
| publishDate | 2025 |
| publishDateRange | 2025 |
| publishDateSort | 2025 |
| publisher | Stellenbosch : Stellenbosch University |
| publisherStr | Stellenbosch : Stellenbosch University |
| record_format | dspace |
| source_str | SUNScholar — Stellenbosch University Repository |
| spelling | oai:scholar.sun.ac.za:10019.1/134588 A comparative analysis and improvement of the empirical design flood estimation approaches De Klerk, Barend Jakobus Du Plessis, Kobus Stellenbosch University. Faculty of Engineering. Dept. of Civil Engineering. Flood forecasting -- Mathematical models Hydrologic models Runoff -- Mathematical models Thesis (MEng)--Stellenbosch University, 2025. De Klerk, B. J. 2025. A Comparative Analysis and Improvement of the Empirical Design Flood Estimation Approaches. Unpublished masters thesis. Stellenbosch: Stellenbosch University [online]. Available: https://scholar.sun.ac.za/items/097d4026-7edb-4762-abbc-321e47a22f3f ENGLISH ABSTRACT: The research focused on evaluating the performance of the empirical and standard design flood (SDF) design flood estimation (DFE) methods, as well as enhancing the Midgley and Pitman (MIPI) DFE. To this end, performance measurement methodology involved processing and analysing both probabilistic and traditional DFE peak values at 411 gauging stations and their associated catchments. Data was acquired and prioritised through various sources. The analysis utilized Statflo software, which was modified to include the generalized Pareto (GPA) distribution and various goodness-of-fit (GoF) criteria, such as the adjusted R², Kolmogorov-Smirnov test, Akaike information criterion (AIC), Schwarz Bayesian information criterion (SBC), and root mean square error (RMSE). A ranking approach was employed to identify the best-performing distribution, with the improved probability model, South Africa (IPZA) distribution selected for further analysis. The design flood estimation tool (DFET) was tailored for mass data processing, and alternative areal reduction factors (ARFs) were applied to the SDF method (SDFA). The peak runoff values for each catchment were then calculated using the SDF, SDFA, MIPI and catchment parameter (CAPA) DFE methods. The MIPI equation parameters were analysed to identify correlations affecting the estimated/observed (E/O) flood peak relationship, revealing hydraulic length (L) as the strongest predictor. Consequently, flood peak ratios (E/O) were plotted against hydraulic length to define adjustment ratios (R_VT) based on a selected hydraulic length threshold (L'). These E/O ratios varied according to veld type regions, exhibiting no consistent general pattern. The initial threshold was chosen arbitrarily and then iteratively refined to minimize both average and median deviations, as well as to reduce significant underestimation errors (defined as underestimates exceeding 50%). A specific hydraulic length threshold (L') was ultimately selected based on optimal performance across these criteria, with minor deviations from this value having a negligible impact on overall flood peak estimation accuracy. R_VT values for segments above and below L' were optimized by minimizing deviations from observed probabilistic peaks, resulting in a significant improvement in the accuracy of flood peak estimates (Q_VT). The peak runoff values for each catchment were compared with at-site probabilistic peak values. The performance of each method was measured using various GoF criteria, including standard error of estimate (SE), mean absolute relative error (MARE), root mean square error (RMSE), coefficient of determination (R²), and Nash-Sutcliffe efficiency (NSE). Additional analytical techniques included scatter charts, relative and absolute error assessments, box-and-whisker plots and E/O estimates. Visual regional performance plots were provided to show the performance of the DFE methods on a regional scale. CAPA demonstrated the best performance for the return periods for return periods of 1:2-, 1:5- and 1:200-years, yielding the best scores. SDF, SDFA and the traditional MIPI produced variable results across the return periods. Q_VT performed the best overall for the 1:10 to 1:100-year return periods, ranking highest for all return periods, followed in order by CAPA, MIPI, SDF and SDFA. This confirmed that the Q_VT showed a clear improvement on the traditional MIPI. The performance of traditional DFE methods varies, with the SDF showing the weakest correlation (R² = 0.562) and a tendency to underestimate at lower return periods and overestimate at higher return periods (average slope = 1.294). SDFA improves moderately (R² = 0.564) and overestimates with a slope of 1.624, while the traditional MIPI offers a better fit and overestimating (R² = 0.591, slope = 1.352) flood peaks. CAPA slightly outperforms MIPI (R² = 0.593), but substantially overestimates with a slope of 1.758. Q_VT achieves the best fit (R² = 0.606), though its weaker slope (0.779) suggests tendency to underestimate calculated flood peaks. The analysis of various DFE methods using box-and-whisker plots revealed key insights into their performance across return periods. SDF showed variability, with high overestimates (266%) and a risk of underestimation (down to -100%) at longer return periods, particularly for the 50- and 100-year periods. SDFA offers comparable estimates. MIPI provides stable, reliable runoff values, with minimal risk of underestimation, making it a dependable choice. CAPA shows an upward trend in estimates at longer return periods, balanced by a risk of underestimation at shorter periods. Q_VT similarly to CAPA, with moderate potential for overestimates and some underestimation risk, offering a balanced approach to runoff estimation. Q_VT and CAPA demonstrated the best overall performance when applying the E/O estimates metrics by minimizing extreme deviations and maximizing the proportion of estimates within the acceptable range. While SDF and MIPI yielded reasonable results, SDFA requires further calibration to reduce the frequency of gross errors. The performance of Q_VT was confirmed through independent verification and indicates that Q_VT substantially improved upon the performance of the traditional MIPI flood estimation method. AFRIKAANSE OPSOMMING: Die navorsing het gefokus op die evaluering van die doeltreffendheid van die empiriese en standard ontwerpvloed (SDF) ontwerpvloedskatting (DFE) metodes, sowel as die verbetering van die MIPI DFE. Vir hierdie doel het die doeltreffendheidmetingsmetodologie die verwerking en ontleding van beide waarskynlikheids- en tradisionele DFE piekwaardes by 411 meetstasies en hul geassosieerde opvanggebiede behels. Data is verkry en geprioritiseer deur verskeie bronne. Die ontleding het Statflo sagteware gebruik, wat aangepas is om die veralgemeende Pareto-verdeling (GPA) en verskeie beste passing (GoF) kriteria in te sluit, soos die aangepaste R², Kolmogorov-Smirnov toets, Akaike-inligtingskriterium (AIC), Schwarz Bayesiese inligtingskriterium (SBC), en wortel-gemiddelde-kwadraat fout (RMSE). 'n Rangordebenaderingis gebruik om die mees doeltreffende verdeling te identifiseer, met die verbeterde waarskynlikheidsmodel, Suid Afrika verdeling (IPZA) gekies vir verdere ontleding. Die ontwerpvloedskatting sigblad (DFET) is aangepas vir massa dataverwerking, en alternatiewe area reduksie faktore (ARFs) is op die SDF-metode toegepas (SDFA). Die piek-afloopwaardes vir elke opvanggebied is bereken deur gebruik te maak van die SDF, SDFA, MIPI en CAPA metodes. Die MIPI-vergelykingsparameters is ontleed om korrelasies te identifiseer wat die verwagte/waargenome (E/O) vloedpiekverhouding beïnvloed. Die hidrouliese lengte (L) het die sterkste korrelasie getoon. Gevolglik is E/O geplot teenoor hidrouliese lengte om aanpassingsverhoudings (RVT) te definieer gebaseer op ’n geselekteerde grenswaarde vir hidrouliese lengte (L’). Hierdie E/Overhoudings het volgens veld tipe streke gevarieer sonder om ’n konsekwente algemene patroon te toon. Die aanvanklike hidrouliese grens is arbitrêr gekies en daarna iteratief verfyn om beide gemiddelde en mediaanafwykings te minimaliseer, asook om onaanvaarbare onderskattingsfoute (gedefinieer as onderskattings wat 50% oorskry) te beperk. ’n Spesifieke grenswaarde vir hidrouliese lengte (L’) is uiteindelik geselekteer op grond van optimale prestasie oor hierdie kriteria, met geringe afwykings van hierdie waarde wat ’n minimale impak op die uiteindelike akkuraatheid van vloedpiekskatting gehad het. RVT-waardes vir segmente bo en onder L’ is geoptimaliseer deur afwykings van waargenome waarskynlikheidspieke te minimaliseer, wat gelei het tot ’n beduidende verbetering in die akkuraatheid van vloedpiekskattings (QVT). Die berekende piekvloei waardes was vergelyk met elke stasie se waarskynlikheids-piekwaardes deur gebruik te maak van verskeie GoF-kriteria, insluitend standaardfout van skatting (SE), gemiddelde absolute relatiewe fout (MARE), RMSE, bepaalheidskoëffisiënt (R²), en Nash-Sutcliffe doeltreffendheid (NSE). Bykomende ontledingstegnieke het verspreidingsgrafieke, relatiewe foutbeoordelings, boks-enhttps snorbaard-grafieke, verwagte/waargenome skattings, en geografiese doeltreffendheids-evaluerings ingesluit. CAPA het die beste presteer vir die 1:2-, 1:5- en 1:200-jaar herhaalperiodes en het die beste punte behaal. SDF, SDFA en die tradisionele MIPI het wisselvallige resultate oor die herhaalperiodes gelewer. QVT het die beste prestasie algeheel vir die 1:10 tot 1:100-jaar herhaalperiodes getoon, en het die hoogste rang vir alle herhaalperiodes behaal, gevolg in volgorde (beste na slegste) deur CAPA, MIPI, SDF en SDFA. Dit het bevestig dat die QVT 'n duidelike verbetering op die tradisionele MIPI toon. Die resultate is verder onderworpe aan verspreidingsgrafieke, relatiewe foutontledings, boks-ensnorbaard-grafieke en verwagte/waargenome skattings. Die prestasie van tradisionele DFE-metodes wissel, met die SDF-metode wat die swakste korrelasie toon (R² = 0. 562) en 'n neiging tot onderskatting by laer herhaalperiodes en oorskatting by hoër herhaalperiodes (helling = 1.294). SDFA verbeter matig (R² = 0.564, helling = 1.624), terwyl die tradisionele MIPI 'n beter passing bied (R² = 0.591, helling = 1.352). Die CAPA-metode presteer effens beter as MIPI (R² = 0.593, helling = 1.758), wat 'n sterker verhouding tussen die veranderlikes aandui. Die QVT-metode behaal die beste passing (R² = 0.606), hoewel sy platter helling (0.779) 'n onderskatting n die berekende vloed pieke aantoon. ie ontleding van verskeie DFE-metodes deur middel van boks-en-snorbaard-grafieke het belangrike insigte in hul prestasie oor verskillende herhaalperiodes aan die lig gebring. SDF toon variasie, met uiterste oorberamings (266%) by kort herhaalperiodes en volgehoue risiko's van onderskatting by langer herhaalperiodes (so laag as -100%), veral by die 1:50- en 1:100-jaar herhaalperiodes. SDFA bied soortgelyke resultate. MIPI bied stabiele, betroubare vloed pieke, met minimale risiko van onderskatting, wat dit 'n betroubare keuse maak. CAPA toon 'n opwaartse neiging in skattings by langer herhaalperiodes, gebalanseer deur 'n risiko van onderskatting by korter periodes. QVT presteer soortgelyk aan CAPA, met 'n matige potensiaal vir oorberamings en 'n paar risiko's van onderskatting, wat 'n gebalanseerde benadering tot vloed pieke bied. QVT en CAPA het die beste algehele prestasie getoon toe die E/O skattings-kriteria toegepas is, deur uiterste afwykings te minimaliseer en die aantal skattings binne die aanvaarbare waardes te maksimeer. SDF en MIPI het redelike resultate opgelewer, terwyl SDFA verdere kalibrasie vereis om die frekwensie van groot foute te verminder. Die resultate van QVT is deur onafhanklike verifiëring bevestig en dui daarop dat QVT die prestasie van die tradisionele MIPI-metode wesenlik verbeter het. Masters 2025-12-15T12:21:34Z 2025-12-15T12:21:34Z 2025-12 Thesis https://scholar.sun.ac.za/handle/10019.1/134588 en Stellenbosch University xviii, 274 pages : illustrations application/pdf Stellenbosch : Stellenbosch University |
| spellingShingle | Flood forecasting -- Mathematical models Hydrologic models Runoff -- Mathematical models De Klerk, Barend Jakobus A comparative analysis and improvement of the empirical design flood estimation approaches |
| title | A comparative analysis and improvement of the empirical design flood estimation approaches |
| title_full | A comparative analysis and improvement of the empirical design flood estimation approaches |
| title_fullStr | A comparative analysis and improvement of the empirical design flood estimation approaches |
| title_full_unstemmed | A comparative analysis and improvement of the empirical design flood estimation approaches |
| title_short | A comparative analysis and improvement of the empirical design flood estimation approaches |
| title_sort | comparative analysis and improvement of the empirical design flood estimation approaches |
| topic | Flood forecasting -- Mathematical models Hydrologic models Runoff -- Mathematical models |
| url | https://scholar.sun.ac.za/handle/10019.1/134588 |
| work_keys_str_mv | AT deklerkbarendjakobus acomparativeanalysisandimprovementoftheempiricaldesignfloodestimationapproaches AT deklerkbarendjakobus comparativeanalysisandimprovementoftheempiricaldesignfloodestimationapproaches |