Inter-Research > CR > Prepress Abstract

CR prepress abstract   -  DOI:

Projection of future changes in summer thermal stress index in Romania using statistical downscaling and associated uncertainties

Aristita Busuioc*, Alexandru Dumitrescu, Amalia Iriza-Burca, Zenaida Chitu, Rodica Dumitrache, Andreea Dima

*Corresponding author:

ABSTRACT: A statistical downscaling model based on the canonical correlation analysis (CCA-SDM) is used to project future changes in heat thermal stress (THI) at high spatial resolution across Romania and frequency of extremely high THI values (FRTHI) over southern regions. The uncertainty associated to the statistically downscaled changes and validity of the CCA-SDM under future changed climate is assessed. The 2021-2050 and 2071-2100 time horizons are considered. Large-scale predictors (temperature at 850 hPa, specific humidity at 700 hPa) simulated by six CMIP5 (RCP4.5 and RCP8.5 scenarios) and 17 CMIP6 (SSP2-4.5 and SSP5-4.5 scenarios) GCMs are considered to estimate the GCM uncertainty. The uncertainty related to the downscaling technique is estimated by comparing the THI changes (derived directly through dynamical downscaling of three EURO-CORDEX RCMs) with those obtained through the CCA-SDM (driven by the same GCM). The CCA-SDM is skilful in capturing the observed THI magnitude and temporal variability over periods with different climate regimes and less skilful in capturing the FRTHI magnitude. Future statistically downscaled THI changes exhibit an increase over the entire country, with magnitudes depending on the GCM driver and time horizon, the highest values being revealed over mountainous regions. Increases are also revealed for FRTHI changes. THI changes derived from RCMs show similar behaviors to those obtained through CCA-SDM driven by the same GCM. The two climate signals are closest for those RCMs. that are more skilful in capturing the observed characteristics of THI variability, showing an increased confidence of the CCA-SDM transferability in future changed climate.