engleski

RegCM performance over the Adriatic region with respect to EURO-CORDEX and MED-CORDEX regional climate model simulations: evaluation of the near-surface wind

Over the past few decades efforts were put into creation of coordinated sets of regional (downscaled) climate simulations (RCM) with typical horizontal resolutions ranging from 50 km to 12.5 km. This enables to confront the RCM added value issues in comparison with global reanalysis, global climate simulations or 50 km resolution RCMs. The CORDEX initiative provides a possibility to evaluate an ensemble of models at 12.5 km and 50 km resolution. Of our particular interest are two RegCM simulations: ICTP MED- CORDEX simulation using RegCM4.3 and DHMZ EURO- CORDEX simulation using RegCM4.2. Our study explores simulated and observed near-surface wind speed over the Adriatic region. Due to the diverse topography, the Adriatic area represents a significant challenge for RCMs to realistically simulate local wind flows. Observations include the land-based station data from the 1996-2008 period and QuikSCAT data from 2000-2008 period. Observed data have been compared against the daily output obtained from a suite of ten CORDEX RCM simulations (seven from EURO-CORDEX: CLMcom- CCLM4-8-17, DMI-HIRHAM5, IPSL-INERIS-WRF331F, KNMI-RACMO22E, SMHI-RCA4, CNRM-ALADIN5.3 DHMZ- RegCM4.2, and three from MED-CORDEX: CNRM- ALADIN5.2, ICTP-RegCM4.3, UCLM-PROMES), where all simulations are forced by the ECMWF ERA-Interim reanalysis. Furthermore, for the proper comparison between 12.5 km and 50 km resolution simulations, the upscaling of the 12.5 km model resolution to 50 km grid is performed. Also, several methodological aspects related to the interpolation techniques when comparing RCMs and observations (or RCMs at two different resolutions) are examined. To provide an answer if the added value is attained in the 12.5 km simulations when compared to 50 km simulations and ERA-Interim, we discuss a measures-oriented approach (e.g. Brier skill score) and a distributions-oriented approach (e.g. Perkins skill score), both disclosing season and location dependence. In addition, the DHMZ-RegCM4.2 power spectrum is analysed. The results indicate that all RCMs taken into account rarely outperform the ERA-Interim reanalysis, while the upscaled 12.5 km resolution offers better agreement with observations than 50 km resolution simulations. For example, the RegCM4.2 maxima and minima in the annual cycle align better to the observations for the 12.5 km (and 12.5 km upscaled) resolution in comparison to 50 km resolution. Power spectrum of the DHMZ-RegCM4.2 shows two large peaks at 12 h and 24 h period, also present in measurements. When RCMs are explored in terms of the skill in reproducing specific wind regimes over the Adriatic region (e.g. the bora wind), large spread in the RCM ensemble is found. In particular, when looking at the wind rose or results of the principal component analysis, DHMZ-RegCM4.2 model frequently gives the correct angle of the bora flow, but it overestimates the wind magnitude.

added value ; near-surface wind ; RegCM

nije evidentirano