Performance of RegCM4 simulations over Croatia and adjacent climate regions

Abstract

We investigate the impact of horizontal resolution, convection parameterization and lateral boundary conditions on the ability of a regional climate model ‐ RegCM4 to simulate the observed temperature and precipitation climatology over Croatia and surrounding regions. An ensemble of RegCM4 simulations over Europe has been generated including two resolutions (50 and 12.5 km), two different deep convection schemes and different boundary conditions (a reanalysis and four global climate models). Using two observational data sets (with and without the undercatch correction) shows that the magnitude of RegCM4 errors in winter precipitation also depends on observational uncertainty. While the mean precipitation amounts in high‐resolution simulations are generally overestimated in the cold part of the year, low‐resolution RegCM4 simulations are comparable to most of other EURO‐CORDEX models. When forced by ERA‐Interim, RegCM4 performs relatively poorly in reproducing temporal correlations against observations, but performs well in terms of spatial correlation. Whereas precipitation amounts from high‐resolution simulations need further improvements, the RegCM4 skill in simulating near‐surface temperature is high and well suited for impact studies.

Bivariate bias correction of the regional climate model ensemble over the Adriatic region

“…At this point, it cannot be concluded whether systematic error in the ensemble or E‐OBS data creation smooths the elevation dependence. In addition, the warm bias over the Pannonian plain was shown to be sensitive to the implementation of the deep convection parameterization in one of the RCMs used in this study (DHMZ‐RegCM4 model from Güttler et al ., 2020). However, during winter, the weather is under the influence of a high‐pressure system (D'Arrigo et al ., 2005), and the temperature in the lower parts of the domain depends on the duration of the fog, which is a process that cannot be a well‐resolved event in RCMs at an ~10 km spatial resolution.…”

“…Considering that any gridded dataset has uncertainties, it is worth emphasizing the basic assumption that systematic errors in the RCMs are usually larger than the associated uncertainty of the gridded datasets such as the E-OBS (Güttler et al, 2020).…”

“…To our knowledge, in this subdomain bias, correction of climate model output has only been performed by using simple univariate methods and on a coarser spatial grid (Gajić‐Čapka et al ., 2017). Considering that any gridded dataset has uncertainties, it is worth emphasizing the basic assumption that systematic errors in the RCMs are usually larger than the associated uncertainty of the gridded datasets such as the E‐OBS (Güttler et al ., 2020).…”

Očekivano toplinsko opterećenje Dubrovnika, Osijeka, Rijeke, Zadra i Zagreba prema projekcijama regionalnih klimatskih modela

“…This adjustment was made on a monthly basis, which means that the observed and modelled daily data from a certain period were taken separately for each month, and the monthly means of modelled data corrected by this method match the monthly means of the measured values. The analysis of regional climate model systematic errors is provided in other studies (e.g., Güttler et al, 2020 for the DHMZ-RegCM4 simulations). No additional adjustments were performed before the quantile mapping step.…”

“…Currently, RegCM4 (Giorgi et al, 2012) is one of the latest versions of a regional climate model that can be used for any part of the Earth at a resolution of up to approximately 10 km (hydrostatic limit). The implementation of the RegCM4 model by the Croatian Meteorological and Hydrological Service (DHMZ) is evaluated in Güttler et al (2020) and will be denoted DHMZ-RegCM4 in this paper. Similar to the RegCM4 model, the SMHI-RCA4 regional climate model was developed by the Swedish Meteorological and Hydrological Institute (SMHI; Wang et al, 2015).…”