A novel approach to statistical downscaling considering nonstationarities: application to daily precipitation in the Mediterranean area

摘要: [1] In the present study, nonstationarities in predictor–predictand relationships within framework of statistical downscaling are investigated. In this context, a novel validation approach is introduced which explicitly taken into account. The method based on results from running calibration periods. (non)overlaps bootstrap confidence interval mean model performance (derived by averaging performances all calibration/verification periods) and intervals individual errors used to identify (non)stationary performance. specified procedure demonstrated for daily precipitation Mediterranean area using bias assess skill. A combined circulation-based transfer function–based employed as technique. large-scale seasonal atmospheric regimes, synoptic-scale circulation patterns, their within-type characteristics, related station-based precipitation. Results show that due varying predictors–precipitation specific configurations. regard, frequency changes patterns can damp or increase effects nonstationary relationships. Within scope assessing future under increased greenhouse warming conditions, identification analysis leads substantiated selection models assessments. Using RCP4.5 scenario assumptions, strong increases become apparent over large parts western northern regions winter. spring, summer, autumn, decreases until end 21st century clearly dominate entire area.