| |
 |
UNIVERSITY OF BUCHAREST
FACULTY OF PHYSICS Guest
2024-11-23 18:25 |
 |
|
|
|
Conference: Bucharest University Faculty of Physics 2014 Meeting
Section: Atmosphere and Earth Science; Environment Protection
Title:
Statistical downscaling model for thermal stress indices in Romania
Authors:
Andreea DOBRINESCU (1,2), Aristita BUSUIOC (1)
Affiliation:
1) National Meteorological Administration, 97 Sos. Bucuresti-Ploiesti, Sect. 1, Bucharest, Romania
2) University of Bucharest, Faculty of Physics, P.O.BOX MG-11, Magurele, Bucharest, Romania
E-mail
andreea.dobrinescu@meteoromania.ro
Keywords:
thermal stress indices, downscaling, Romania
Abstract:
The future climate change scenarios based on global climate models project significant increase in temperature extremes in the next decades. There are some complex indices quantifying the direct discomfort felt by human body and they were less analysed so far. Two of such indices are known: heat index corroborating the air temperature with the relative humidity (ITU) specific for the summer season; for winter, a cooling index (IR) is dependent on air temperature and wind speed. The IR was calculated at 61 stations while ITU was calculated at 87 stations, uniformly distributed over the Romanian territory. The information supplied by the global climate models (GCMs) are too coarse (about 100 km) to calculate these indices for practical needs. There are two techniques to derive information on higher spatial resolution, usual known as downscaling: dynamical downscaling given by the regional climate models (RCMs) that are nested in the global climate models and statistical downscaling (SD) techniques that are based on statistical relationships between local variables of interest and large-scale variables (predictors). In this paper a statistical downscaling model is develop to estimate the two stress indices (predictands) from the large-scale predictors that are usually well simulated by the GCMs. The SD models are based on the canonical correlation analysis (CCA). Before the CCA, the predictors and predictands are transformed to anomalies (by subtracting the long term mean) and then are projected on their EOF (empirical orthogonal functions). The most important EOFs (explaining the most part of the observed variance) are retained for the CCA model. The air temperature at 850 hPa (T850), sea level pressure (SLP) and specific humidity at 700 hPa (SH700) have been considered as predictors for the two predictands. It was found that the T850 alone is a very good predictor for both indices, the model performance (represented by the correlation between estimated and observed values, as well as by the explained variance of estimated values from the total observed variance) being very high for all stations. The combination between T850 and SLP (for IR) or SH700 (for ITU) increases a little the model performance only for ITU, while for IR the model performance is increased only for a few stations These results show that the statistical model presented in this paper is skillful and can be used in the future to project the future ITU/IR changes on station scale in Romania using the climate change GCM simulations.
|
|
|
|