Interaction of a sea breeze and a moist convection over the northeastern Adriatic coast: an analysis of the sensitivity experiments using the high‐resolution mesoscale model Gabrijela Poljak 1 , Maja Telišman Prtenjak 1 , Marko Kvakić 2 , Kristina Šariri 3 and Željko Večenaj 1 1 Andrija Mohorovičić Geophysical Institute, Department of Geophysics, Faculty of Science, University of Zagreb, Zagreb, Croatia 2 ILRI, Nairobi, Kenya 3 Croatian Metrology Institute, Zagreb, Croatia email: [email protected] & [email protected] References Babić K, Mikuš P, Prtenjak MT (2012): The relationship between shallow thermal circulation regimes and cumulonimbus clouds along northeastern Adriatic coast. Geofizika, 29,103-120. Mikuš P, Prtenjak MT, Strelec Mahović N (2012): Analysis of the convective activity and its synoptic background over Croatia. Atmos. Res., 104/105,139-153. Pielke RA (2002): Mesoscale meteorological modeling, 2nd Edition, Academic Press, USA. Poljak G, Prtenjak, MT, Kvakić M, Strelec Mahović N, Babić K (2014): Wind patterns associated with the development of daytime thunderstorms over Istria, Ann. Geophys., 32, 401–420. Sović I, Šariri K, Živčić M (2013): High frequency microseismic noise as possible earthquake precursor. Research in Geophysics, 3, 8-12. Fig. 1: The satellite SST and ground temperature distribution in the area of interest (the northern Adriatic at 11 CET on 9 July 2006) and measuring sites as white circles (Udine, San Pietro Capofiume (SPC), Portorož-Airport, Pazin, Poreč, Pula-airport). Motivation & Aim ¾ The northeastern (NE) Adriatic, in particular, the Istrian peninsula, represents the most convective area in Croatia especially in the warm part of the year (Mikuš et al., 2012). ¾ In at least 51% of daytime Cb occurrences, sea breezes (SB) develops along the coast. (e.g. Babić et al., 2012; Poljak et al., 2014). ¾ The SB characteristics highly depend on location and local topographic characteristics. ¾ The influence of the size and temperature of the water surface and topography on the SB- Cb interplay at particular Adriatic area is still insufficiently known. Model WRF-ARW, V3 ¾ initial and boundary conditions (every 6 h) ECMWF analysis; ¾ vegetation and land-use data: USGS 24; ¾ 3 domains (dx=13.5 km, 4.5 km, 1.5 km) & 2- way nesting on a Lambert conformal projection; ¾ top of the atmosphere = 50 hPa & 80 sigma levels; Several physical options for all domains: Constant: ¾ RRTM for the longwave radiation; ¾ Dudhia scheme for the shortwave radiation; ¾ a five-layer thermal diffusion scheme for the soil temp.; ¾ the Betts-Miller-Janjić cumulus parameterization in 2 outer domain Varying: ¾ PBL schemes: YSU, MYJ and BL ¾ Microphysics schemes: Kessler, Lin, WSM6 Sensitivity numerical tests: (i) a varying SST field provided by the MSG SEVIRI geostationary satellite data (at 5 km) every hour (SST L3C hourly data, Fig. 1); (ii) a modified topography. Results -> Sensitivity tests Impact of the SST variability (9/July/2006 at 15 CET) ECMWF SST satellite SST Fig. 6: Distribution of 10-m wind, (wind vectors and speed in m/s) in (a,b), sensible heat flux, W/m 2 (c,d) and 10-m wind with the maximum simulated radar reflectivity factor, dBZ (e,f). More realistic SST distribution caused a larger sea-land temperature difference, stronger SB with larger vertical speeds and stronger deep convection. Summary ¾ The application of the three methods in the evaluation of the model and the determination of a model setup did not give a specific conclusion. However a choice of MYJ and Lin for PBL and microphysics options gave satisfactory result in the most cases. ¾ Sensistivity tests have shown the impact of the SB-Cb interaction. It primarily takes place in the boundary layer due to SB modification and thus affects the convergence zone and the position and the lifespan of convective cells. Data ¾ Three chosen cases from Poljak et al. (2004): (i) 8/June/2003 (ii) 9/July/2006 (iii) 8/August/2006 ¾ Radiosounding (Udine & SPC) and near surface data (Fig. 1). Fig. 2: Modifying of the topography was done by a simple cosine weight function which is zero at the boundary and one in the center, defined over a 100x100 point square area. Results -> Evaluation of the model In-situ statistical evaluation: Taylor diagram for Pula-Airport Fig. 3: Taylor diagram for temperature (°C) and mixing ratio (kg/kg) for three selected cases. The acceptable mesoscale model skill could be done if Stdev_WRF ~ Stdev_M and RMSD < Stdev_M (Pielke, 2002). This is mostly not valid for Kessler microphysics option. Fig. 4: Experiments biases for all experiments for speed (m/s) in Pula Airport. Biases computed as observed minus each experiment forecast. Regardless the model setup, the WRF model always underestimated wind speed. In-situ statistical evaluation: spectral analysis ¸ Fig. 5: Observed versus modelled time series (top) and power spectra (bottom) for Portorož-Airport site for 8/June/2003; (a,f) the horizontal zonal wind (m/s), (b,g) zonal wind component (m/s), (c,h) merdional wind component (m/s), (d,i) temperature (°C), (e,j) mixing ratio (kg/kg). The observed time series and spectrum is presented by black line. The image moment analysis, IMA (e.g. Sović et al., 2013) 8/Jun/2003 Temperature (°C) 8/Aug/2006 9/Jul/2006 Mixing ratio (kg/kg) 8/Jun/2003 9/Jul/2006 8/Aug/2006 Experiments biases {obs-fcst} for speed (m/s) PBL 8/Jun/2004 9/Jul/2007 8/Aug/2006 YSU -0.87 -0.99 -0.85 -0.82 -1.12 -0.72 -0.90 -0.93 -1.08 -1.09 -1.06 -1.34 MYJ -0.86 -1.05 -1.09 -1.01 -0.97 -0.38 -1.32 -1.10 -1.11 -0.83 -0.77 -1.18 BL -0.80 -0.88 -0.81 -0.85 -1.09 -0.83 -0.92 -1.08 -1.13 -0.97 -0.99 -1.00 O Kess Lin WSM6 O Kess Lin WSM6 O Kess Lin WSM6 Microphysics <-0.8 -0.8->-0.9 -0.9->-1.0 -1.0->-1.1 -1.0->-1.1 >-1.2 1 12 24 36 0 1 2 3 4 5 6 7 hour V H [m s -1 ] sim01 sim02 sim08 sim11 sim12 sim18 sim21 sim22 sim28 sim61 sim62 sim68 obsv 1 12 24 36 -4 -2 0 2 4 6 8 hour u [m s -1 ] 1 12 24 36 -4 -2 0 2 4 hour v [m s -1 ] 1 12 24 36 18 20 22 24 26 28 30 32 hour t [°C] 1 12 24 36 0.008 0.01 0.012 0.014 0.016 0.018 0.02 hour q [kg kg -1 ] 2 3 5 10 20 30 0 0.2 0.4 0.6 0.8 1 1.2 T [h] fS V H (f) [m 2 s -2 ] 2 3 5 10 20 30 0 1 2 3 4 5 T [h] fS u (f) [m 2 s -2 ] 2 3 5 10 20 30 0 0.2 0.4 0.6 0.8 1 1.2 T [h] fS v (f) [m 2 s -2 ] 2 3 5 10 20 30 0 5 10 15 20 T [h] fS t (f) [°C 2 ] 2 3 5 10 20 30 0 1 2 3 4 5 x 10 -6 T [h] fS q (f) [kg 2 kg -2 ] (a) (g) (b) (f) (c) (d) (e) (j) (i) (h) Fig. 5: The IMA method is used for the analysis of convection comparing the maximum simulated radar reflectivity factor (dBZ) with observed ones (source: ARSO-www.meteo.si). The model is more successful when the Euclidean distance (ED) is smaller, which means that the model is closer to reality. Impact of the modified topography (MT): 9 July 2006 Fig. 7: Diurnal evolution of the modeled 10-m wind (in m/s) associated with the maximum simulated equivalent radar reflectivity factor (dBZ). The field comparison shows: (i) deeper intrusion of the flows with eastern directions over the peninsula; (ii) later development, smaller humidity advection and slower inland penetration of the western SB (eliminated slope winds); (iii) the mountain ridge control the onset and accelerated the convection. Acknowledgements This work has been supported by the CATURBO (HRZZ, No. 09/151) project.