Hydrol. Earth Syst. Sci., 25, 3471–3492, 2021 https://doi.org/10.5194/hess-25-3471-2021 © Author(s) 2021. This work is distributed under the Creative Commons Attribution 4.0 License. Changes in the simulation of atmospheric instability over the Iberian Peninsula due to the use of 3DVAR data assimilation Santos J. González-Rojí1,2, Sheila Carreno-Madinabeitia3,4, Jon Sáenz5,6, and Gabriel Ibarra-Berastegi7,6 1Oeschger Centre for Climate Change Research, University of Bern, Bern, Switzerland 2Climate and Environmental Physics, University of Bern, Bern, Switzerland 3Department of Mathematics, University of the Basque Country (UPV/EHU), Vitoria-Gasteiz, Spain 4TECNALIA, Basque Research and Technology Alliance (BRTA), Parque Tecnológico de Álava, Vitoria-Gasteiz, Spain 5Department of Physics, University of the Basque Country (UPV/EHU), Leioa, Spain 6Plentziako Itsas Estazioa (BEGIK), University of the Basque Country (UPV/EHU), Plentzia, Spain 7Department of Energy Engineering, University of the Basque Country (UPV/EHU), Bilbao, Spain Correspondence: Santos J. González-Rojí ([email protected]) Received: 3 February 2020 – Discussion started: 11 March 2020 Revised: 6 May 2021 – Accepted: 26 May 2021 – Published: 18 June 2021 Abstract. The ability of two downscaling experiments underestimate (depending on the parameter) the variability of to correctly simulate thermodynamic conditions over the the reference values of the parameters, but D is able to cap- Iberian Peninsula (IP) is compared in this paper. To do so, ture it in most of the seasons. In general, D is able to produce three parameters used to evaluate the unstable conditions in more reliable results due to the more realistic values of dew the atmosphere are evaluated: the total totals index (TT), point temperature and virtual temperature profiles over the convective available potential energy (CAPE), and convec- IP. The heterogeneity of the studied variables is highlighted tive inhibition (CIN). The Weather and Research Forecast- in the mean maps over the IP. According to those for D, the ing (WRF) model is used for the simulations. The N exper- unstable air masses are found along the entire Atlantic coast iment is driven by ERA-Interim’s initial and boundary con- during winter, but in summer they are located particularly ditions. The D experiment has the same configuration as N, over the Mediterranean coast. The convective inhibition is but the 3DVAR data assimilation step is additionally run at more extended towards inland at 00:00 UTC in those areas. 00:00, 06:00, 12:00, and 18:00 UTC. Eight radiosondes are However, high values are also observed near the southeast- available over the IP, and the vertical temperature and mois- ern corner of the IP (near Murcia) at 12:00 UTC. Finally, no ture profiles from the radiosondes provided by the University linear relationship between TT, CAPE, or CIN was found, of Wyoming and the Integrated Global Radiosonde Archive and consequently, CAPE and CIN should be preferred for (IGRA) were used to calculate three parameters commonly the study of the instability of the atmosphere as more atmo- used to represent atmospheric instability by our own method- spheric layers are employed during their calculation than for ology using the R package aiRthermo. According to the val- the TT index. idation, the correlation, standard deviation (SD), and root mean squared error (RMSE) obtained by the D experiment for all the variables at most of the stations are better than those for N. The different methods produce small discrepan- 1 Introduction cies between the values for TT, but these are larger for CAPE and CIN due to the dependency of these quantities on the Precipitation is one of the most important variables involved initial conditions assumed for the calculation of a lifted air in the water balance, and its variability determines the wa- parcel. Similar results arise from the seasonal analysis con- ter resources of the planet. Following the definitions of re- cerning both WRF experiments: N tends to overestimate or gional models, precipitation can be separated into two cat- egories: large-scale and convective precipitation. In general, Published by Copernicus Publications on behalf of the European Geosciences Union. 3472 S. J. González-Rojí et al.: Changes in the simulation of atmospheric instability over the IP convective precipitation is frequently associated with precip- Kingdom (Holley et al., 2014) suggests that the reduction of itation extreme events due to high intensity over a short dura- CAPE overnight is over 500 J/kg. tion. However, the simulation of these events is a well-known On the global scale, CAPE follows the spatial pattern of problem in the modelling community (Sillmann et al., 2013) surface specific humidity and air temperature, which means due to restrictions in the resolution, poor representation of that it increases from pole to Equator (Riemann-Campe et al., complex topography, insufficient assimilated observations, 2009). The minimums are obtained in arid regions and over forecast errors, or deficiencies in the microphysics schemes areas with cold water upwelling. Focusing on Europe, con- in the numerical models. In order to avoid these problems, as vective storms develop for lower values than the United previously done in the literature (Viceto et al., 2017), this pa- States (Graf et al., 2011), and several studies have tried to per focuses on the evaluation of the atmospheric conditions determine the most active regions. Amongst them, Romero favourable for the development of convective precipitation et al.(2007) found that the region with highest instability is rather than the validation of the simulation of extreme events. located along a zonal belt over south-central Europe, partic- The evaluation of the atmospheric conditions is typically ularly over the western Mediterranean Sea and the surround- based on the calculation of some instability indices such as ing areas. This agrees with Brooks et al.(2003), who found the lifted index (LI) (Galway, 1956), the K index (George, that the favourable environment for thunderstorms is devel- 1960), the total totals index (TT) (Miller, 1972), or the oped in southern Europe and that the highest number of days Showalter index (S) (Showalter, 1953). These conditions can in such a regime is located over the Iberian Peninsula (here- be also evaluated by convective available potential energy after, IP), south of the Alps, and the northern Balkans. How- (CAPE) (Moncrieff, 1981) or convective inhibition (CIN) ever, van Delden(2001) found that southwestern France and (Moncrieff, 1981). All of these variables are commonly used the Basque Country seem to be a preferred region for the for- in the literature for these kind of studies (e.g. Ye et al., mation of severe storms that drift towards the northeast. More 1998; DeRubertis, 2006; Viceto et al., 2017). CAPE and CIN recent studies based on lightning data (Enno et al., 2020) and are based on the adiabatic lifting of a parcel, while most regional climate models using higher resolution (Mohr et al., of the others are based on differences in the values of sev- 2015; Rädler et al., 2018) highlighted the same areas with eral variables at different pressure levels. The deep convec- favourable environments for thunderstorms in Europe, which tion is caused by three ingredients: high levels of moisture are located in particular over northern Italy (Po Valley), east in the planetary boundary layer (PBL), potential instability, of the Adriatic Sea (Albania, Bosnia, and Serbia), and in the and forced lifting (Johns and Doswell, 1992; McNulty, 1995; northeastern IP and southern France (near the Gulf of Lyon). Holley et al., 2014; Gascón et al., 2015). CAPE and CIN pro- Over the IP, the seasonality of precipitation is determined vide information about the first two ingredients (Holley et al., by different sources of moisture due to seasonal variations of 2014), and both can give details about the genesis and inten- the global atmospheric circulation and contrasting climatic sity of the atmospheric convection (Riemann-Campe et al., regions (influenced by the strong topography). Northern and 2009). However, previous studies (Angus et al., 1988; López western IP are mainly affected by stratiform precipitation et al., 2001) suggest that CAPE should not be used alone but during winter, while eastern and southern IP receive great should be combined with other indices. The final ingredient, amounts of precipitation during autumn due to convective which is the forced lifting, is usually caused by the orogra- activity (Rodríguez-Puebla et al., 1998; Esteban-Parra et al., phy (Doswell et al., 1998; Siedlecki, 2009), the convergence 1998; Romero et al., 1999; Iturrioz et al., 2007). Maximum of horizontal moisture fluxes (McNulty, 1995), or the breezes precipitation amounts over central IP are measured in early in coastal regions (van Delden, 2001). Thus, the high spatial spring (Tullot, 2000). and temporal resolution is important for these kind of stud- Previous studies over the IP (Viceto et al., 2017) sug- ies focusing on the atmospheric convection, and that is why gest that CAPE shows a high spatio-temporal variability: the regional simulations are needed (Siedlecki, 2009). values in winter and spring over land are small due to the The probability of occurrence of convective precipitation reduced surface temperature, and the differences between is not the same through the day, and previous studies sup- Atlantic and Mediterranean regions are remarkable