Estimation of Future Return Levels for Heavy Rainfall in the Iberian Peninsula: Comparison of Methodologies

Abstract:

F. J. Acero¹, S. Parey², T.T.H. Hoang², D. Dacunha-Castelle³

¹Dpto. Física, Universidad de Extremadura, Avda. de Elvas s/n, 06006, Badajoz

²EDF/R&D, 6 quai Watier, 78401 Chatou Cedex, France

³Laboratoire de Mathématiques, Université Paris 11, Orsay, France

Trends can already be detected in daily rainfall amount in the Iberian Peninsula (IP), and this will have an impact on the extreme levels. In this study, we compare different ways to estimate future return levels for heavy rainfall, based on the statistical extreme value theory. Both Peaks over Threshold (POT) and block maxima with the Generalized Extreme Value (GEV) distribution will be used and their results compared when linear trends are assumed in the parameters: threshold and scale parameter for POT and location and scale parameter for GEV. But rainfall over the IP is a special variable in that a large number of the values are 0. Thus, the impact of taking this into account is discussed too. Another approach is then tested, based on the evolutions of the mean and variance obtained from the time series of rainy days only, and of the number of rainy days. A statistical test, similar to that designed for temperature in Parey et al. 2013, is used to assess if the trends in extremes can be considered as mostly due to these evolutions when considering only rainy days. The results show that it is mainly the case: the extremes of the residuals, after removing the trends in mean and standard deviation, cannot be differentiated from those of a stationary process. Thus, the future return levels can be estimated from the stationary return level of these residuals and an estimation of the future mean and standard deviation. Moreover, an estimation of the future number of rainy days is used to retrieve the return levels for all days. All of these comparisons are made for an ensemble of high quality rainfall time series observed in the Iberian Peninsula over the period 1961-2010, from which we want to estimate a 20-year return level expected in 2020. The evolutions and the impact of the different approaches will be discussed for 3 seasons: fall, spring and winter.

Parey S., Hoang T.T.H., Dacunha-Castelle D.: The importance of mean and variance in predicting changes in temperature extremes, Journal of Geophysical Research: Atmospheres, Vol. 118, 1–12, 2013.