• extreme precipitation patterns;
  • GCM-downscaled precipitation conditioned on atmospheric circulation;
  • nonlinear dynamics;
  • recurrence quantification analysis;
  • chaotic time series


A statistical framework based on nonlinear dynamics theory and recurrence quantification analysis of dynamical systems is proposed to quantitatively identify the temporal characteristics of extreme (maximum) daily precipitation series. The methodology focuses on both observed and general circulation model (GCM) generated climates for present (1961–2000) and future (2061–2100) periods which correspond to 1xCO2 and 2xCO2 simulations. The daily precipitation has been modelled as a stochastic process coupled with atmospheric circulation. An automated and objective classification of daily circulation patterns (CPs) based on optimized fuzzy rules was used to classify both observed CPs and ECHAM4 GCM-generated CPs for 1xCO2 and 2xCO2 climate simulations (scenarios). The coupled model ‘CP-precipitation’ was suitable for precipitation downscaling. The overall methodology was applied to the medium-sized mountainous Mesochora catchment in Central-Western Greece. Results reveal substantial differences between the observed maximum daily precipitation statistical patterns and those produced by the two climate scenarios. A variable nonlinear deterministic behaviour characterizes all climate scenarios examined. Transitions’ patterns differ in terms of duration and intensity. The 2xCO2 scenario contains the strongest transitions highlighting an unusual shift between floods and droughts. The implications of the results to the predictability of the phenomenon are also discussed. Copyright © 2013 John Wiley & Sons, Ltd.