Abstract
The present work aims to provide reliable estimates of extreme discharge
flows and their probability of occurrence. Such estimates are important
for the assessment of the associated hydrological risk of hydraulic
infrastructures, such as bridges and dams, in the design process as well
as during their operations. The hydrological modeling herein developed
was applied to estimate the design floods approaching the new Hintze
Ribeiro bridge, in the north of Portugal. It proposes a statistical
analysis of the maximum annual streamflow data by using a flood
frequency analysis technique. The data series were subject to a
reliability analysis and the specific modeling assumptions, required for
the study, were appropriately given and tested. An extrapolation
technique of the missing instantaneous discharge data was herein
derived. Such technique was validated by two distinct methods. The
estimations are accurate with a mean deviation of 7.2% relative to the
observed data. A set of probabilistic models were considered and the
models’ performance verified by the goodness-of-fit tests and Q-Q plots.
The model and the parameter uncertainties were taken into account. Model
uncertainties were addressed by comparing the estimated design floods
through selecting the best fitting probability model (MS) with the
approach that considered the distribution functions which fit well the
data (MM). On the other hand, the computed flow rates were estimated
with 95% of confidence to reduce the inherent parameter uncertainties.
An additional accuracy assessment of the parametric approaches was
performed through a comparative analysis of such design floods with the
ones retrieved by application of the non-parametric Kernel density
estimate (KDE). The MM approach showed a lower discrepancy (18.5%) to
KDE estimates, when compared with the MS results. A sensitivity analysis
of the associated hydrological risks was also undertaken.