Articles | Volume 16, issue 1
Hydrol. Earth Syst. Sci., 16, 241–254, 2012
Hydrol. Earth Syst. Sci., 16, 241–254, 2012

Research article 25 Jan 2012

Research article | 25 Jan 2012

Nonstationarities in the occurrence rates of flood events in Portuguese watersheds

A. T. Silva1, M. M. Portela1, and M. Naghettini2 A. T. Silva et al.
  • 1Instituto Superior Técnico, Technical University of Lisbon, Lisbon, Portugal
  • 2Federal University of Minas Gerais, Belo Horizonte, Brazil

Abstract. An exploratory analysis on the variability of flood occurrence rates in 10 Portuguese watersheds is made, to ascertain if that variability is concurrent with the principle of stationarity. A peaks-over-threshold (POT) sampling technique is applied to 10 long series of mean daily streamflows and to 4 long series of daily rainfall in order to sample the times of occurrence (POT time data) of the peak values of those series. The kernel occurrence rate estimator, coupled with a bootstrap approach, was applied to the POT time data to obtain the time dependent estimated occurrence rate curves, λˆ(t), of floods and extreme rainfall events. The results of the analysis show that the occurrence of those events constitutes an inhomogeneous Poisson process, hence the occurrence rates are nonstationary. An attempt was made to assess whether the North Atlantic Oscillation (NAO) casted any influence on the occurrence rate of floods in the study area. Although further research is warranted, it was found that years with a less-than-average occurrence of floods tend to occur when the winter NAO is in the positive phase, and years with a higher occurrence of floods (more than twice the average) tend to occur when the winter NAO is in the negative phase. Although the number of analyzed watersheds and their uneven spatial distribution hinders the generalization of the findings to the country scale, the authors conclude that the mathematical formulation of the flood frequency models relying on stationarity commonly employed in Portugal should be revised in order to account for possible nonstationarities in the occurrence rates of such events.