Extreme value analysis of precipitation series in the south of Iberian Peninsula

- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner
Monday, 18 January 2010
Exhibit Hall B2 (GWCC)
Jose Manuel Hidalgo-Muñoz, Universidad de Granada, Granada, Granada, Spain; and D. Argüeso, D. Calandria-Hernandez, S. R. Gamiz-Fortis, M. J. Esteban-Parra, and Y. Castro-Diez

Handout (3.0 MB)

The analysis of extreme precipitation events on regional scales has attracted special interest in recent years due to the importance of these studies for applications in risk and impact analysis. Therefore, the study of the occurrence probability of the extreme events in a regional scale is of great interest. In this study, we analyse daily precipitations records of the Southern Iberian Peninsula, an area of special interest due to the unfavourable predictions (increase of heavy precipitation and droughts events) suggested by the models under climate change conditions. The original dataset comprises about 2000 stations, but only stations with less than 10% of missing values in the period 1955-2006 were selected. The selected stations were subjected to a quality control study in order to ensure the homogeneity of the series. Finally, 90 stations were considered in this study. Using daily precipitation records, we have constructed two different extreme precipitation series for each location: the annual maximun (AM) series and the partial duration (PD) series, computed using the exceedances of given appropriated threshold value.

Some statistical distributions of extremes precipitation series commonly used, as Generalized Pareto or Gumbel, have been considered to calculate the best fit to AM and PD series. The L moments method has been used for determining the best fitted distribution and the parameters of this distribution. Results indicate that, in general, Generalized Extreme Value and Gumbel distributions fit better to AM series and Generalized Pareto and Exponential fit better to PD series. Additionally, the values associated with different return periods have been calculated, suggesting differences in western and eastern part of the region of study, from the AM and PD series.