Dissertation - Kenechukwu Okoli

  • Date: –14:00
  • Location: Geocentrum Hambergsalen, Geocentrum, Villavägen 16, Uppsala
  • Doctoral student: Kenechukwu Okoli
  • About the dissertation
  • Organiser: Institutionen för geovetenskaper
  • Contact person: Kenechukwu Okoli
  • Disputation

Design flood estimation under uncertainty

A common task in hydrology is the estimation of the design flood, i.e. a value of river discharge corresponding to a given exceedance probability that is often expressed as a return period in years. Flood risk assessment, floodplain mapping and the design of hydraulic structures are a few examples of applications where estimates of design floods are required. Common approaches for estimating a design flood are based on: (i) hydrological methods such as continuous simulations with rainfall–runoff models, or (ii) statistical methods, such as the fitting of a probability distribution function to a record of annual maximum values. In this thesis, these alternative approaches are compared in view of the various sources of uncertainties affecting the estimation of the design flood. Since design floods are typically not known a priori, a series of virtual experiments was developed and implemented for both estimation methods, hence the magnitudes and frequencies of the design floods are known ab initio, and the quality of estimates (i.e., in terms of their accuracy and precision) were analysed. These virtual experiments are defined as ‘numerical experiments with a model considered as the truth and best understanding of the modelled processes’. This thesis looked at the influence of method of estimation, model structure uncertainty, errors in the flow data, and sampling on design flood estimation. The results show that design floods estimated by using a simple rainfall-runoff model have small uncertainties (i.e., variance of the errors) even for high return periods compared to statistical methods. Statistical methods performed better than the simple rainfall-runoff model in terms of median errors for high return periods, but their uncertainty (i.e. variance of the error) is larger. The thesis concludes that given the sources of uncertainty of statistical and hydrological methods, they both should be applied as complementary.

Download the dissertation on this link.