Estimation of 100-year-return-period temperatures in France in a non-stationary climate: Results from observations and IPCC scenarios

Reference daily series of maximum temperature observed by Météo-France are used to extract the extreme values (maximum per block of time, or values above a threshold) and test the significance of a temporal polynomial trend. When the trend is statistically significant, the Peak Over Threshold method, used for very large or extreme values of stationary stochastic processes, is extended to the non-stationary case to define a new return-level. Extrapolated observed temperatures in a non-stationary climate are then compared with temperatures simulated in A2 and B2 IPCC scenarios, with Météo-France and IPSL atmospheric models in the framework of a French project called IMFREX (from April 2003 to April 2005) coordinated by Météo-France. Most of the 55 reference daily series studied indicate a quadratic evolution of the Poisson process intensity and a stationary Pareto scale parameter from the 1960s or 1970s to 2002, indicating more frequent exceedances at the end of the 20th century. Extrapolation of the significant observed trends allow to reach the low level of centennial temperatures simulated by the atmospheric models for the end of the 21st century.