Assistant Research Professor Florida State University
To improve current research on flood frequency analyses, detecting a close-to-real time flood events by studying multiple flood characteristics (peak, volume and duration; not only peak), and choosing proper probability distribution functions for multivariate frequency analyses are crucial. Daily discharge data are widely used in flood frequency analyses due to the lack of adequate historical data at smaller timescales. However, compared to the daily timescale, using instantaneous discharge, or at least hourly data, can better explain flood dynamics and lead to more accurate estimations of the dependency between the flood characteristics. This dependency is the main factor that drives the goodness-of-fit of multivariate cumulative distribution functions like copulas for flood events and estimation of the conditional return periods. Here, annual maximum and peak over threshold methods with two timescales–daily and hourly–records were applied to evaluate the dependency among river flood characteristics across selected watersheds in the USA. The preliminary results showed that the quantile of hourly data in the same return period is generally higher than that for the daily timescale. Gumbel copula was found as the best joint bivariate distribution function, with inconsistent parameters at daily and hourly timescales. Joint bivariate return period in both "and" and "or" cases indicated that multivariate frequency analyses based on "or" case lead to greater high quantiles compared to "and" case. This research has implications for the design of civil infrastructure considering plausible flood events they experience over their lifespa
