EGU24-12436, updated on 09 Mar 2024
https://doi.org/10.5194/egusphere-egu24-12436
EGU General Assembly 2024
© Author(s) 2024. This work is distributed under
the Creative Commons Attribution 4.0 License.

Detecting Trends In Hydrological Extremes And Non-Stationary Extreme Value Analysis Of Flood Data In Kwazulu-Natal, South Africa

Demian Vusimusi Mukansi1,2, Jeff Smithers1,2,3, Katelyn Johnson1,2, Thomas Kjeldsen1,4, and Macdex Mutema5
Demian Vusimusi Mukansi et al.
  • 1School of Engineering, University of KwaZulu-Natal, Pietermaritzburg, South Africa
  • 2Centre for Water Resources Research, University of KwaZulu-Natal, Durban, South Africa
  • 3Director of Centre for Water Resources Research, University of KwaZulu-Natal, Durban, South Africa
  • 4Department of Architecture and Civil Engineering, University of Bath, United Kingdom
  • 5Department of Agricultural Engineering, Agricultural Research Council, Pretoria, South Africa

In this study, the annual maximum streamflow from 14 stations in KwaZulu-Natal, along the East Coast of South Africa, were analysed. Trends were investigated using the non-parametric Mann-Kendall test and the Sen Slope tests, and the results indicate that the annual maximum streamflow has been decreasing in magnitude at 78 % of stations. Extreme value analysis was performed using both stationary and non-stationary models using time and rainfall as covariates. The results show that the stationary models are superior to non-stationary models at most stations with time as a covariate. Where possible, streamflow stations were linked with rainfall stations to determine the impact of rainfall on annual maximum streamflow. The results indicate that the non-stationary model incorporating observed rainfall as a covariate performed better than the stationary and non-stationary models with only time as a covariate. Therefore, incorporating rainfall in design flood estimation should be considered to account for non-stationary trends and to mitigate the risk of failure of hydraulic structures. Regional magnification factors to account for non-stationarity were not investigated further in this study as the majority of the stations showed a negative trend, which means the application of a regional magnification factor will result in a reduction of the magnitude of the estimated design floods.

How to cite: Mukansi, D. V., Smithers, J., Johnson, K., Kjeldsen, T., and Mutema, M.: Detecting Trends In Hydrological Extremes And Non-Stationary Extreme Value Analysis Of Flood Data In Kwazulu-Natal, South Africa, EGU General Assembly 2024, Vienna, Austria, 14–19 Apr 2024, EGU24-12436, https://doi.org/10.5194/egusphere-egu24-12436, 2024.