Repository logo
 

Detecting and assessing the impacts of outlier events and data availability on design rainfall and flood estimation in South Africa.

dc.contributor.advisorSmithers, Jeffrey Colin.
dc.contributor.advisorJohnson, Katelyn Ann.
dc.contributor.authorSingh, Keanu Reeve.
dc.date.accessioned2021-06-25T13:08:24Z
dc.date.available2021-06-25T13:08:24Z
dc.date.created2021
dc.date.issued2021
dc.descriptionMasters Degree. University of KwaZulu-Natal, Pietermaritzburg.en_US
dc.description.abstractAccurate Design Rainfall Estimation (DRE) and Design Flood Estimation (DFE) require long periods of quality-controlled data for the planning, design, operation, and improved flood risk assessment of hydraulic structures. However, observed hydrological data frequently include outlier events and there is a decline of hydrological monitoring in South Africa which may impact DRE and DFE. It is therefore necessary to assess the impact of outlier events and reduced data availability on DRE and DFE. The aims of this study were to: (a) assess the impact of outlier events on DRE and DFE in South Africa, (b) assess the performance of outlier detection methods under South African conditions, and (c) assess the impact of reduced data availability on DRE and DFE in South Africa. The impact of synthetic Low Outlier (LO) and High Outlier (HO) events on DRE and DFE from observed and synthetically generated data series were assessed. The performance of the BoxPlot, Modified Z-Score (MSZ) and Multiple Grubbs-Beck Test (MGBT) outlier detection methods were assessed. Record length and network density were reduced to assess the impact of reduced data availability on DRE and DFE. Results from the analysis of observed data show that design rainfall is impacted by up to 22% and design floods by up to 45% in the presence of LOs. Design rainfall is impacted by up to 16% and design floods by up to 46% in the presence of HOs. For synthetically generated data series, design rainfall and floods are impacted by up to 2% and 1% respectively in the presence of LOs and by up to 13% in the presence of HOs. At best, LOs in observed rainfall and streamflow data are under-detected by up to 6% and 30% respectively by the MGBT method, whereas HOs are over-detected up to 50% and 150% respectively by the MZS method. Design rainfall and flood events are impacted by up to 4% and 24% respectively by reduced record lengths, and by up to 4.5% and 60% respectively from a reduced gauged network. This study indicates that outlier detection be adopted as regular practice in South Africa and that additional national resources must be directed towards maintaining and improving the hydrological monitoring networks in South Africa.en_US
dc.description.notesAuthor's Keywords: Design floods, Design rainfall, Network Density, Outliers, Record length.en_US
dc.identifier.urihttps://researchspace.ukzn.ac.za/handle/10413/19532
dc.language.isoenen_US
dc.subject.otherFloods.en_US
dc.subject.otherHydrological monitoring.en_US
dc.subject.otherFlood forecasting.en_US
dc.subject.otherRainfall management.en_US
dc.subject.otherStreamflow.en_US
dc.subject.otherRain gauge.en_US
dc.titleDetecting and assessing the impacts of outlier events and data availability on design rainfall and flood estimation in South Africa.en_US
dc.typeThesisen_US

Files

Original bundle

Now showing 1 - 1 of 1
Loading...
Thumbnail Image
Name:
Singh_Keanu_Reeve_2021.pdf
Size:
8.2 MB
Format:
Adobe Portable Document Format
Description:

License bundle

Now showing 1 - 1 of 1
No Thumbnail Available
Name:
license.txt
Size:
1.64 KB
Format:
Item-specific license agreed upon to submission
Description: