The ENEA-REG system (v1.0), a multi-component regional earth system model. Sensitivity to different atmospheric component over Med-CORDEX region

Abstract. In this study, a new regional Earth system model is developed and applied to the Med-CORDEX region. The ENEA-REG system is made up of two interchangeable regional climate models as atmospheric components (RegCM and WRF), a river model (HD), and an ocean model (MITgcm); processes taking place at the land surface are represented within the atmospheric models with the possibility to use several land surface schemes of different complexity. The coupling between these components is performed through the RegESM driver. Here, we present and describe our regional Earth system model and evaluate its components using a multidecadal hindcast simulation over the period 1980–2013 driven by ERA-INTERIM reanalysis. We show how the atmospheric components are able to correctly reproduce both large-scale and local features of the Euro-Mediterranean climate, although some remarkable biases are relevant for some variables. In particular, WRF has a significant cold bias during winter over North-Eastern bound of the domain, while RegCM systematically overestimates the wind speed over the Mediterranean Sea. This latter bias has severe consequences on the ocean component: we show that when WRF is used as the atmospheric component of the Earth system, the performances of the ocean model are remarkably better compared with the RegCM version. Our regional Earth system model allows studying the Euro-Mediterranean climate system and can be applied to both hindcast and scenario simulations.



32
The Mediterranean basin is a complex region, characterized by the presence of pronounced 2015; Turuncoglu and Sannino, 2017). Given the relatively fine spatial scales at which these 38 processes take place, the Mediterranean basin provides a good opportunity to study regional 39 climate, with a special focus on the air-sea coupling (Sevault et al., 2014;Turuncoglu and 40 Sannino, 2017). For these reasons, regional coupled models have been developed and used to  Table 1. 147 The other supported atmospheric component of the regional Earth system model is RegCM 148 (version 4.5) a hydrostatic, compressible, sigma-p vertical coordinate model initially developed 149 by Giorgi (1990) and Giorgi et al. (1993aGiorgi et al. ( , 1993b and then modified as discussed by Giorgi et al.  The model domain (Figure 2b) is projected on a Lambert conformal grid with a horizontal 163 resolution of 20 km and with 23 vertical levels extending from land surface up to 50 hPa. 164 Similarly to WRF, we used ERA-Interim data to force RegCM and 6 grid-points in each side are 165 selected as relaxation zone with an exponentially decreasing relaxation coefficient (Giorgi et al. 166 1993) ( Table 1). 167 A few modifications have been made both in WRF and RegCM to receive the oceanic surface The river discharge is a key variable in the Earth system modeling as it closes the water cycle 216 between the atmosphere and ocean. The ENEA Sannino 2017) to retrieve surface runoff and drainage from the atmospheric components of the 223 regional coupled model and to provide the river discharge to the ocean component (Figure 1). are filled by interpolation (Reynolds et al., 2007).

241
Salinity data for the Mediterranean Sea are obtained from DIVA (data-interpolating variational 242 analysis); this tool allows to interpolate in situ observations to obtain gridded climatologies 243 (Brasseur et al., 1996).
For the mixed layer depth, we use a global climatology computed from more than one million 245 Argo profiles collected from 2000 to present (Holte et al., 2017); this climatology provides 246 estimates of monthly mixed layer depth on a global 1° gridded map.

247
As reference dataset to evaluate the performances of the atmospheric components of the ENEA- In general, the spatial performances of the ENEA-REG system are better when WRF is used as

322
Looking at precipitation, WRF shows a systematic dry bias over sea with respect to ERA5, while  configurations, a similar monthly distribution with ERA5 dataset with a peak in the late summer 369 caused by sparse precipitation and high evaporation. The total E-P estimated simulated using 370 WRF is 890±43 mm/yr while with RegCM we obtain a mean annual estimate of 909±45 mm/yr; 371 in contrast, ERA5 data has a lower E-P of 729±56 mm/yr.

372
In addition to freshwater flux, wind speed is also a key variable for ocean models as it controls 373 the evaporation over the sea surface and affects the ocean circulation through the drag stress. addition, it is responsible for the large evaporative flux described in Figure 5. 384 It should also be noted that the large bias found over mountainous regions is clearly an artifact 385 due to the spatial resolution differences, with ERA5 reanalysis reproducing lower wind speed 386 than both WRF and RegCM because of its coarser resolution. In general, the two atmospheric underestimated by the interactive river routing model (Figure 12); this underestimation is more 492 evident in RegCM as a consequence of the larger drier precipitation bias found over land ( Figure   493 4), resulting in a lower river baseline with respect to WRF (Figure 12). 494 Looking at the monthly SSS anomalies (Figure 13a) we found a similar temporal variability Ocean. In general, the two-way exchange at the strait is constituted by an upper inflow of 510 Atlantic water and a lower outflow of relatively colder and saltier Mediterranean water.

511
However, the semidiurnal tidal effect is strong enough to reverse the direction of the flows 512 during part of the tidal cycle. As this exchange represents the main driver of the circulation in the 513 basin, the estimation of its value has been faced for decades.

514
The inflow transport derived from the two coupled simulations is about 1 Sv ( Table 2); 515 similarly, the models predict a net transport of 0.06 Sv. Unfortunately, the estimate of the 516 transport obtained from the direct measurements of velocities is affected by the limited number 517 of moorings used that cannot resolve the structure of the entire section. Therefore, some 518 numerical models have also been used to reproduce and quantify the two way-exchange .  Table 3; here we present estimates from the DIVA 568 data, while for the two simulations we show the anomalies with respect to the reference data. The  (1983,1987, and 1993 paragraph 4.2.4). 600 The mean annual salinity averaged over the whole column (Table 3)  documented also in observations (Lascaratos et al. 1999;Malanotte et al., 1999;Roether et al., 649 2007). These events (1983,1987 and 1989), corresponding to intense atmospheric fluxes, have  (Figure 17b).

660
In addition to the temporal evolution of MLD, in Figure 18 we compare the mean spatial pattern the discovery of the EMT. As described by many authors, there is observational evidence that 669 during the '90s the main source of EMDW migrated to the Aegean Sea (Lascaratos et al., 1993;670 Malanotte et al., 1999;Wu et al., 2000;Roether et al., 2007;Beuvier et al., 2010). The common 671 understanding is that the EMT has been the effect of many concurrent causes that make this

709
We presented a newly designed regional Earth system model used to study the climate variability interior of the regional model domain (Waldron et al.1996;Heikkilä et al., 2011); this is achieved 735 by relaxing the model state towards the driving large-scale fields by adding a non-physical term 736 to the model equation (Omrani et al., 2015). Clearly, the spectral nudging allows a stronger 737 control by the driving forcing and thus a greater consistency between the regional model and 738 large-scale climate coming from the driving model. Nowadays, there is still some controversy on the use of indiscriminate nudging in regional climate models (e.g. Omrani et al., 2015). Some 740 studies agree that nudging does not allow the regional model to deviate much from the driving 741 fields limiting the internal physics of the regional climate model (

764
This analysis reveals that spectral nudging helps to keep the large scale circulation of the  (Figure 3).
Notwithstanding the better performances, nudging has also to be used with caution: strong 770 inconsistencies between regional model and driving large-scale fields may lead to unrealistic 771 compensations within the model, for example, anomalous heat fluxes compensating for 772 temperature biases (Brune and Baehr, 2020). 773 We conclude that in the context of coupled atmosphere-ocean models, the correct representation 774 of surface winds is crucial to simulate ocean-atmosphere interactions correctly.       Note that in the bias panels ERA5 data are interpolated into the atmospheric model grid. Mind 1209 also the differences in colour scales between DJF and JJA climatologies.  Note that in the bias panels ERA5 data are interpolated into the atmospheric model grid.