th 5 European Conference on Severe Storms 12 - 16 October 2009 - Landshut - GERMANY

ECSS 2009 Abstracts by session

ECSS 2009 - 5th European Conference on Severe Storms 12-16 October 2009 - Landshut – GERMANY List of the abstract accepted for presentation at the conference: O – Oral presentation P – Poster presentation

Session 06: Numerical studies, e.g. of convective initiation, storm life cycles and phenomena

Page Type Abstract Title Author(s)

Explicit forecasting of supercells with the operational A. Seifert, M. Baldauf, C. O COSMO-DE Gebhardt, S. Theis Maritime convective initiation of the severe thunderstorm of J.-B. Cohuet, R. Romero, V. 161 O 4 October 2007 in Mallorca: numerical experiments Homar, V. Ducroq, C. Ramis Multi-decadal dynamical downscaling of tropical cyclones 163 O in East Asia using spectrally nudged regional climate F. Feser, H. von Storch models D. Mastrangelo, K. Horvath, M. Observational and numerical analysis of a heavy 165 O M. Miglietta, A. Moscatello, A. precipitation event over southern Italy Riccio An Analysis of numerically simulated mesovortices and 167 O A. D. Schenkman, M. Xue tornado-like vorticies in mesoscale convective system EnKF Analysis of the 29 May 2004 Oklahoma City T. E. Thompson, L. J. Wicker, D. 169 O Supercell using Rapid-Scan Phased Array Radar Data E. Forsyth, M. I. Biggerstaff Sensitivity of quantitative precipitation forecast to soil K. van Weverberg, N. P. M. van 171 O moisture initialization, microphysics parameterization Lipzig, L. Delobbe, D. Lauwaet and horizontal resolution A comparison of transient impinging jet and cooling source B. C. Vermeire, L. G. Orf, E. 173 P downburst models Savory B. R. S. B. Basnayake, S. Das, M. Composite characteristics of severe thunderstorms over 175 P K. Das, M. Rahman, M. A. Bangladesh simulated by WRF-ARW Model Sarker, M.A.R. Akand The influence of boundary layer conditions on storm life 177 P M. Curic, D. Janc, N. Kovacevic cycles Numerical simulations of supercells over idealized 179 P P. Markowski, N. Dotzek orography Long-lasting deep convective systems in the Mediterranean M. Pasqui, S. Melani, B. Gozzini, 181 P basin: a model study F. Pasi Numerical simulation of tornado-scale vortices occurred in 183 P K. Tsuboki, A. Sakakibara a winter cold-air outbreak over the Sea of Japan An investigation of a severe multicellular storm in the 185 P U. Wissmeier, R. Goler tropics Forecasting skill study of different non-hydrostatic 187 P meteorological model configurations in severe A. Bertozzi, P. Randi convective events simulation

159 Page Type Abstract Title Author(s)

Parameterization and development of statistical model for P N. Huseynov, B. Malikov meteorological elements of convective instability Characteristics of convective processes in inland Northeast F. Espejo, E. Alvarez, F. Cortes, 189 P Spain C. Lafragüeta High-Resolution Simulations of Convective Cold Pools over P. Knippertz, J. Trentmann, A. P the Northwestern Sahara Seifert A right flank supercell in Romagna; Splitting storm system 191 P P. Randi, A. Bertozzi case study Validation of deep convection in the convective-scale NWP P K. Wapler, A. Seifert, B. Ritter model COSMO-DE Generation of Windstorm in the Eastern Mountainous 193 P H. Choi, D. S. Choi Coast of Korea

160 5h European Conference on Severe Storms 12 - 16 October 2009 - Landshut - GERMANY

MARITIME CONVECTIVE INITIATION OF THE SEVERE THUNDERSTORM OF 4 OCTOBER 2007 IN MALLORCA: NUMERICAL EXPERIMENTS

J.B. COHUET1, R. ROMERO2, V. HOMAR2, V. DUCROCQ1, C. RAMIS2

1Météo France, 42 av. G. Coriolis 31057 Toulouse, France, 2 Universitat de les Illes Balears, Carrer de Valldemosa 07122 Palma de Mallorca, Spain,

(15 September 2009)

I. INTRODUCTION III. RESULTS AND CONCLUSIONS

The Mediterranean basin is regularly affected by Méso-NH simulations have been performed with severe weather associated with deep convection. Even if several initial conditions from ECMWF, ARPEGE and they are usually tied to coastal orography, some severe ALADIN, and confirm the crucial effects of initial state on thunderstorms develop and mature over the sea. A good and the numerical modeling. Failure of experiments initialized recent example is the severe thunderstorm that crossed the with ALADIN data is clearly due to an incorrect island of Mallorca in the afternoon of 4th October 2007. representation of the environment over the Mediterranean Generated early in the morning offshore of Murcia Sea. Experiments fed with ECMWF and ARPEGE present (southeastern Spain), this storm organized progressively into more accurate low-level flows and achieve simulations with a squall line structure with a northeastwards movement. more realistic evolutions. However, differences in the low- Arriving in Palma city, this squall line was accompanied by level temperature fields lead to a best convective triggering severe gusts, heavy rain and several F2-F3 tornadoes, with ARPEGE experiments, in which the front is associated leading to significant damages in the southwestern part of with a stronger convergence of winds. Moreover, differences the island and eventually to one fatality. pointed out in experiments performed with Méso-NH, WRF and MM5 with the same initial conditions indicate the An observational study of weather patterns favourable benefits of a multi-model approach, since each one is to such squall line development is presented in Ramis et al, characterized by its own physical and dynamical (2009). The triggering and evolution of convection in these parameterizations. kind of events depend on both synoptic and mesoscale features. Representing such interaction is a real challenge for The analysis of the ensemble of experiments leads to numerical models used in weather forecasting, all the more the conclusion that the low-level convergence appears to be since the squall line was triggered over the sea. in this case the key element to explain the triggering of deep convection. The position of a low-pressure area to the south of the Balearic archipelago in the lower troposphere is an important factor that improves the triggering location of the II. PRESENTATION OF RESEARCH simulated storm. The experiment performed with WRF forced by ECMWF analysis succeeds in reproducing the In order to better identify mechanisms responsible for squall line evolution with the best spatiotemporal the initiation and evolution of this severe thunderstorm, representation. In the morning, this simulation triggers a three mesoscale numerical models have been used (Méso- convective cell to the south of Murcia, over an area with NH, WRF and MM5) combined with several initial and significant wind convergence and characterized by a boundary conditions (from ECMWF, ARPEGE and pronounced convective instability. Later on, this ALADIN). All the experiments were achieved with a fine thunderstorm organizes in a linear structure and moves over resolution mesh (between 2 and 2.4 km) centred over the sea along a thermal boundary to reach Mallorca around Mallorca and no deep convection parameterization was 16 UTC (15:30 UTC in reality). The high sheared activated. Analyses rather than forecasts were used to environment resulting from an easterly low-level flow provide the most realistic initial and boundary conditions overlain by strong southerly winds appears to be necessary available. to sustain the squall line.

The best simulations are expected to provide Of particular importance is that the MM5 model is not interesting comprehensive elements of squall lines dynamics able to trigger a convective cell offshore of Murcia, as and structure. This study also allows assessing initial and opposed to the other models. In order to attempt to remedy boundary conditions influences as well as model this failure to correctly simulate the event, two different parameterization impacts on the modeling framework. approaches have been considered. Since in the MM5 Another objective is to test simulation improvements by experiment, the lack of low-levels convergence seems to assimilating synthetically generated observations avoid an efficient convective triggering, the first corresponding to mesoscale features, in order to remedy the improvement was to assimilate pseudo-observations in order lack of data over the sea. The idea is to assimilate pseudo- to synthetically generate a convergence line. Such observations of plausible structures that are believed to be observations are inspired from an experiment with Méso-NH necessary for convective initiation. forced by ARPEGE analyses in which initial stages of the

161 5h European Conference on Severe Storms 12 - 16 October 2009 - Landshut - GERMANY

convective cell are well captured. Only wind pseudo- observations are provided in the morning, at several levels in the planetary boundary layer, in order to initiate convection. Data assimilation of these pseudo-observations by the nudging method results in a well located convective triggering necessary to the further realistic evolution of the squall line. Although the simulated squall line is slightly off to the west of Mallorca,the signature is very similar to observed radar reflectivities.

The second approach was to explore the sensitivity of the MM5 reference simulation to the planetary boundary layer scheme, since on the whole domain, the convection develops later than with other models. The substitution of MRF scheme (Hong and Pan, 1996) -daily used in the

Universitat de les Illes Balears forecasts- by ETA scheme (Janjic, 1994) leads to significant improvements in the FIG. 2: Vertical vorticity and horizontal wind at 200 m on domain simulation. With this later scheme, using a 1.5 order of indicated on FIG.1. closure and a prognostic equation to compute Turbulent Kinetic Energy, convection is favoured and the triggering of the squall line takes place with rather accurate To conclude, the success of a numerical simulation to spatiotemporal features. The lower diffusion of vertical capture a severe convective event with a maritime initiation velocity patterns and a better representation of the moisture closely depends on a conjunction of factors from distinct vertical profile in the lowest levels are crucial factors to this scales. On the 4th October 2007 event, in a favourable success. synoptic environment which has been identified, a mesoscale circulation was necessary to lead to the squall line Given the successful results of the Méso-NH triggering. Finally, the ability of mesoscale numerical experiment fed with ECMWF analysis -which simulates the models to reproduce such maritime convective event is squall line in remarkable agreement with remote sensing highly dependent on initial condition accuracy at small observations, despite a time lag-, further experiments have scales, an ever-present challenge over the data-void been carried out. An analysis of the internal structure and Mediterranean Sea. dynamics of convection confirms the squall line organization of the simulated convective system. In addition, a very high resolution simulation with a grid size of 600 meters, performed on a domain surrounding the squall line already formed, points out a bow-echo in reflectivities IV. AKNOWLEDGMENTS located in the southern sector of the squall line arriving to Mallorca (FIG. 1). The authors would like to thank Juan Escobar from the Laboratoire d’Aérologie whose help was essential to succeed in running Méso-NH at the UIB. Thanks are also due to Regina Guilabert and Jose Arrecio for their software assistance. At least, UIB authors acknowledge support from MEDICANES/CGL2008-01271/CLI project, from the Spanish Ministerio de Ciencia e Innovacion.

V. REFERENCES

Atkins N., St Laurent N., 2009: Bow echo mesovorticies, Part II: Their genesis. Mon. Wea. Rev., 137 1497-1513. Hong S., and Pan H., 1996: Nonlocal boundary layer vertical diffusion in a Medium-Range Forecast model. Mon. Wea. FIG. 1: Very high resolution simulation with Méso-NH and a 600 m grid size at 12:30 UTC: Reflectivities and indication of FIG.2 Rev., 124 2322-2339 domain. Janjic Z., 1994: The step-mountain Eta Coordinate model: further developments of convection, viscous sublayer and Near the bow apex and in other undulations of the turbulence closure schemes. Mon. Wea. Rev., 122 927-945 squall line leading edge, where the rear inflow jet is severe, Ramis C., Romero R., Homar V., 2009: The severe the highly detailed simulation allows highlighting individual thunderstorm of 4th October 2007 in Mallorca (Spain): an vertical mesovortices. Their respective location, with the observational study. Nat. Haz. and Earth. Syst. Sci., 9 positive vortex to the north, indicates that they are enhanced 1237-1245. by updrafts ahead of the gust front (FIG. 2). The theory of such vortices genesis by updrafts is extremely recent and is described in Atkins and St Laurent, (2009). The presence of this couplet of mesovortices demonstrates the favourable environment to the tornadogenesis.

162 5h European Conference on Severe Storms 12 - 16 October 2009 - Landshut - GERMANY

MULTI-DECADAL DYNAMICAL DOWNSCALING OF TROPICAL CYCLONES IN EAST ASIA USING SPECTRALLY NUDGED REGIONAL CLIMATE MODELS Frauke Feser1 and Hans von Storch2

1GKSS Research Institute, Max-Planck-Str. 1, 21502 Geesthacht, [email protected] 2GKSS Research Institute, Max-Planck-Str. 1, 21502 Geesthacht, [email protected] (Dated: 15 September 2009)

I. INTRODUCTION NCEP deviates a bit. One simulation without spectral Tropical cyclones (TCs) can cause flooding and nudging is also close to the best track data, but the other extreme weather conditions affecting the coastal population. unconstrained RCM simulations deviate to the north of the In order to estimate changes of TCs in the past in location, typhoon track. For the tracking a simple minimum pressure intensity or frequency, a modelling approach was chosen. search was performed for the sea level pressure field. Global reanalyses data are available only at a coarse In a second step, an ensemble of modelled typhoon resolution, for higher resolutions downscaling is necessary. events was examined for the typhoon season 2004 for SE In this study the feasibility to reconstruct the weather Asia and the western Pacific (Feser and von Storch, 2008b) . (including typhoons) of SE Asia for the last decades using 12 typhoons between May and October of the western North an atmospheric regional climate model was analyzed. Global Pacific typhoon season 2004 were analyzed. The selected National Centers for Environmental Prediction - National typhoons were chosen according to the typhoons presented Center for Atmospheric Research (NCEP-NCAR) reanalysis by Zhang et al. (2007). The comparison to Best Track data data were dynamically downscaled to 50 km and in a revealed improved SLP and near-surface wind values by the double-nesting approach to 18 km grid distance using a RCM compared to the reanalyses for most cases. The state-of-the-art regional climate model, the COSMO model reanalyses thereby showed smaller great circle distances to in Climate Mode (CCLM). the best track data than the regional model.

II. PRESENTATION OF RESEARCH Typhoon Model RMSE RMSE FF First, an ensemble of one simulated typhoon case SLP [hPa] [kt] will be presented. The regional simulations were not only Dianmu NCEP 53.83 46.43 driven by lateral boundary conditions, but also with a CCLM 0.5o 25.49 14.34 spectral nudging technique (von Storch et al., 2000) that CCLM0.165o 23.67 14.35 enforced the observed large-scale state inside of the model Songda NCEP 58.27 50.14 domain. Tropical storms which are coarsely described by the CCLM 0.5o 48.02 39.38 reanalyses were correctly identified and tracked; CCLM 0.165o 46.61 40.54 considerably deeper core pressure and higher wind speeds TABLE I: RMSE of SLP and 10m wind speed between JMA best were simulated compared to the driving reanalyses. Several track data and NCEP, CCLM 0.5°, and CCLM 0.165°. sensitivity experiments with varied grid distances, different initial starting dates of the simulations and changed spectral nudging parameters were computed (Feser and von Storch, 2008a).

FIG. 1: Effect of spectral nudging and different initial states.

The track of Typhoon Winnie is shown in Fig. 1 for the best-track data, NCEP reanalysis, and for four simulations with spectrally nudged CCLM and four unconstrained CCLM simulations. The NCEP track deviates only in the earlier part of the development from the best- track data; the spectrally nudged RCM simulations all follow FIG. 2: Brier Skill Score between JMA best track data and NCEP, the best-track data closely, even during the phase when CCLM 0.5°, and CCLM 0.165°.

163 5h European Conference on Severe Storms 12 - 16 October 2009 - Landshut - GERMANY

IV. AKNOWLEDGMENTS Table 1 shows the root mean square error (RMS) The authors would like to thank B. Gardeike who helped for sea level pressure (SLP) and wind at a height of 10 with the preparation of the Figures. The climate version of meters for two sample typhoons. The RMS is for both the ‘Lokal-Modell’ is the community model of the German typhoons and both wind speed and SLP larger for the global climate research. The authors also thank the German reanalyses than for the regional model at both 50 and 18 km Climate Computing Center (DKRZ) which provided the resolution. The regional model runs show only slightly computer hardware for the LAM simulations in the project deviating results. ‘Regional Atmospheric Modelling’. The NCEP-NCAR Figure 2 shows the Brier Skill Score between best reanalyses data was provided by the National Center for track data of the Japan Meteorological Agency (JMA) and Atmospheric Research (NCAR). This study is part of the o o NCEP, CCLM 0.5 , and CCLM 0.165 , for SLP (upper informal project ‘Comparative Study on the retrospective panel) and 10 m wind speed (lower panel) for the analysed simulation of typhoon seasons in SE Asia’ typhoons. For values larger than 0 CCLM is closer to the www.coastdat.eu/typhoon). best track than NCEP, 0 means CCLM is equally close to the best track as NCEP, for values smaller than 0 NCEP is closer to the best track than CCLM. In 7 out of the 10 cases, the high-resolution (0.165o) CCLM performs better than the V. REFERENCES coarse grid (0.5o) CCLM; in 3 cases the improved resolution Feser, F., von Storch, H., 2008b: Regional modelling of the does not lead to better results in terms of SLP. The result is western Pacific typhoon season 2004. less good for wind speed. In terms of this variable, CCLM is Meteorolog. Z., 17, 519-528 not performing better than NCEP in 2 out of 10 cases; Usage Feser, F., von Storch, H., 2008a: A dynamical downscaling of 0.165o grid-sizes leads to results closer to the best track case study for typhoons in SE Asia using a regional numbers than 0.5o grid-sizes only in 3 out of 10 cases. In 7 climate model. of the 10 cases, the 0.5o CCLM is performing better than the Mon. Wea. Rev., 136, 1806-1815 0.165o CCLM. von Storch, H., Langenberg, H., Feser, F., 2000: A Spectral Nudging Technique for Dynamical Downscaling Purposes. Mon. Wea. Rev. 128, 3664–3673 III. RESULTS AND CONCLUSIONS Zhang, X., Li, T., Weng, F., Wu, C.-C., Xu, L., 2007: Some typhoons can be described at least partly Reanalysis of western Pacific typhoons in 2004 with realistically by dynamical downscaling of coarse-grid NCEP multisatellite observations. reanalyses with regional atmospheric models. The presented Meteor. Atmos. Phys. 97, 3–18, Doi: 10.1007/s00703-006- experiments indicate that lateral boundary control is not 0240-5. always sufficient to reconstruct the numbers and tracks of typhoons. Of course, this depends on the size of the domain, for smaller domains the lateral control is more efficient. The inclusion of large-scale constraints, as for instance spectral nudging, is helpful in forming the storms along the right track. Using a double-nesting approach has an effect on the simulation, which is believed to be a slight improvement compared to the 50-km simulation in terms of core pressure development and near-surface wind speed. But the typhoon tracks were very similar. Precipitation patterns and rainfall amounts were simulated more realistically by the RCM using the higher resolution compared to the 50 km run. The location of the typhoon tracks is already well represented in the reanalyses. A comparison of typhoon core pressure along the track revealed RCM values which were closer to the JMA best track data than the global reanalyses for 9 out of 10 cases. Some SLP developments in the RCM followed closely the JMA best track values, but some showed a time lag as e.g. typhoon Songda. In 7 out of 10 cases the highest resolution simulation was closest to the best track SLP. For near-surface wind speed CCLM was closer to the best track wind speed in 8 out of 10 cases compared to NCEP. Still, for all analysed typhoons the RCM underestimated the low pressure values and the strength of the maximum winds. It is concluded that regional models can improve simulated typhoon developments from global forcing reanalyses data not by altering the track, but by giving lower core pressure and higher wind speeds and more realistic precipitation patterns even though these values still do not reach observed values.

164 5th European Conference on Severe Storms 12 - 16 October 2009 - Landshut - GERMANY

OBSERVATIONAL AND NUMERICAL ANALSYIS OF A HEAVY PRECIPITATION EVENT OVER SOUTHERN ITALY D. Mastrangelo1, K. Horvath2, M.M. Miglietta3,4, A. Moscatello4, A. Riccio5

1DISAM, Parthenope University of Naples, Naples, Italy, [email protected] 2Meteorological and Hydrological Service, Zagreb, Croatia, [email protected] 3ISAC-CNR, Padua, Italy, [email protected] 4ISAC-CNR, Lecce, Italy, [email protected] 5DSA, Parthenope University of Naples, Naples, Italy, [email protected] (Dated: 15 September 2009)

I. INTRODUCTION The LLJ advected a potentially unstable airstream with high The Ionian Italian region is one of the Mediterranean moisture content very close to the surface. At about 00 UTC areas prone to heavy precipitation (Federico et al., 2002; 13 November, in the middle of the event, a weak short-wave Miglietta and Regano, 2008). The nearly arc-shaped trough (SWT), developed on the forward flank of a potential topography with steep mountains on the western boundary vorticity (PV) anomaly, approached southern Italy moving may favour the southerly unstable flow coming from the from southwest to northeast. The SWT was associated with Mediterranean Sea to release instability. dry air and, crossing the Apennines, contributed to deepen a In the present study, a heavy precipitation event weak surface low over the Adriatic Sea. The associated affecting Basilicata and Apulia during 12 and 13 November circulation favored the shift of the convective line towards 2004 is analyzed through observations and numerical Salento. During the afternoon of 13 November, following simulations performed with the Weather Research and the whole synoptic system shift, the LLJ and the convective Forecasting (WRF) Model. Rain gauges recorded two line moved definitively northeastward leaving southern distinct precipitation maxima in different phases of the Italy. event: on 12 November, a maximum of about 250 mm in less than 12 hours occurred over the northernmost coastal zone of the Gulf of Taranto; on 13 November, a maximum amount of about 200 mm was recorded over the central part of Salento peninsula. Such amounts are about one third of the mean annual precipitation of the affected zones. The first stages of this event were analysed in Horvath et al. (2006) that focused their attention mainly on the generation of a deep cyclone in the lee of the Atlas Mountains and on severe windstorms over Croatia. Here, we focus on the precipitation features over Italy to identify the main synoptic and mesoscale factors responsible for the generation of convection, and to highlight the main features of the event.

II. THE EVENT The heavy precipitation event lasted more than 24 h starting in the afternoon of 12 November 2004. As two distinct rainfall maxima can be identified in the observations, the event can be roughly subdivided into two phases. During the first phase, corresponding to the second FIG. 1: Observed (top panels) and simulated (bottom panels) precipitation (mm) accumulated during the first phase of the event half of 12 November, the largest precipitation affected (a, c), from 12 UTC 12 November to 00 UTC 13 November, and the localized Ionian coastal areas of southern Basilicata and second phase (b, d), from 00 UTC to 16 UTC 13 November. adjacent Apulia territory (Fig. 1). The two most intense rain peaks were observed at 16 UTC and 22 UTC, the latter producing the largest rain rate of the event, 103 mm in a 2-h III. NUMERICAL OUTPUTS period. Convection organized along a line that, in the The WRF model has been implemented on two following hours, shifted and persisted over a narrow area of domains nested with a two-way technique. The outer domain Salento for about 12 h leading to the second phase of the covers the central and western Mediterranean basin and event. During this phase, a maximum amount of 145 mm north-western Africa with grid spacing of 16 km. The inner was recorded from 00 UTC to 16 UTC 13 November, domain is centred over southern Italy and Ionian Sea; it is whereas 202 mm were recorded on 13 November in the made up of 128 × 148 grid points with grid spacing of 4 km. same narrow area. No convection parameterization is used on the inner domain. The event was associated with a wide Atlantic trough The simulation is initialized at 00 UTC 12 November 2004 that, at the surface, induced a cyclonic circulation leading to and integrated for 72 h. Initial and boundary conditions are a LLJ that extended from northern Africa to southern Italy. provided by ECMWF forecast with grid spacing of 0.5° ×

165 5th European Conference on Severe Storms 12 - 16 October 2009 - Landshut - GERMANY

0.5°. The boundary conditions, including sea surface were directly involved in triggering convection during the temperature field, are updated every 6 h. first phase of the event. Indeed, as diagnosed through the The model realistically predicts the observed nondimensional Froude number, during the afternoon of the precipitation, although a delay of some hours affects the 12 November, the southerly flow impinging over the synoptic evolution and the shift of the linear convective mountain range of Calabria passed from a “flow around” to system from the first to the second maximum area (FIG. 1). a “flow over” regime as the LLJ reached the area. Convection developed over Calabria reliefs was subsequently advected by the mid-tropospheric currents, and reached the first maximum rainfall area where the LLJ continued to feed it. The upper-level SWT associated with the passage of a PV anomaly over southern Italy was responsible for the shift of convection from the area of the first to the second maximum. Moving over the Apennines, the PV anomaly deepened a shallow vortex previously over southern Tyrrhenian. The resulting low- and mid-tropospheric currents veered from south to southwest favouring the shift of the convective line to the east. Moreover, the dryer air associated with the SWT strongly reduced CAPE and increased the LFC over the areas not affected by the LLJ. Consequently, convection was inhibited beneath the SWT and downstream of the convective line where, the convergence between the impinging south-easterly LLJ and the convective outflow was still sufficient to lift the LLJ to the LFC and maintain convection. Beneath the convective line, a cold downdraft outflow developed at about 07 UTC 13 November (FIG. 2). Its propagation caused new convection upstream that caused the convective line to further shift to the east, reaching the area of maximum FIG. 2: 925-hPa equivalent potential temperature (shaded, every 2 observed precipitation amount (FIG. 3). K) and wind vectors, and precipitation > 5 mm (blue contour) simulated at 07 UTC 13 November 2004. The red line indicates the location of the vertical cross section shown in FIG. 3. IV. ACKNOWLEDGMENTS The authors are greatful to SMA S.p.A. and ARPA Basilicata for having provided observational data, and to Aeronautica Militare for satellite and lightning data.

V. REFERENCES Federico, S., Bellecci, C., and Colacino, M.: Numerical simulation of Crotone flood: storm evolution, Nuov. Cim. C, 26C, 357–371, 2003 Horvath, K., L. Fita, R. Romero, and B. Ivančan-Picek, 2006: A numerical study of the first phase of a deep Mediterranean cyclone: Cyclogenesis in the lee of the Atlas Mountains. Meteorol. Zeitschrift 15, 2, 133-146. Miglietta, M. M. and Regano, A.: An observational and numerical study of a flash-flood event over south-eastern Italy, Nat. Hazards Earth Syst. Sci., 8, 1417-1430, 2008

FIG. 3: Vertical cross section of equivalent potential temperature (shaded, every 2 K), potential temperature (contoured, every 2 K), and circulation vectors along the line marked in FIG. 2, at 07 UTC 13 November 2004.

The analysis of numerical outputs of the inner domain reveals that the LLJ had a major role in the event since it supplied moisture to convection throughout the event. Moreover, a strong gradient in moisture content made the warm airstream advected by the LLJ potentially unstable and lowered the level of free convection. As a result, convection could easily develop even without a strong lifting mechanism such as the orography. However, the Apennines

166 5th European Conference on Severe Storms 12 - 16 October 2009 - Landshut - GERMANY

An Analysis of Numerically Simulated Mesovortices and Tornado-like Vortices in a Mesoscale Convective System

Alexander D. Schenkman ([email protected] ) and Ming Xue ([email protected]) School of Meteorology and Center for Analysis and Prediction of Storms University of Oklahoma, USA

15 September 2009

I. INTRODUCTION near the surface and are only present in the lowest kilometer of the atmosphere. A tornado climatology presented in Trapp et al. For brevity we present here only fields showing the (2005) found that up to 20 percent of tornadoes in the United thermodynamic and kinematic structures of the dominant States are spawned by quasi-linear convective systems mesovortex and associated TLVs at 0332 UTC. Figure 1a (QLCS) and bow-echoes. Fujita (1978) presented a shows that a strong TLV is present near the surface. The -1 conceptual model that summarized the lifecycle of a bow maximum vertical vorticity for this TLV is around 0.4 s . echo from a single ‘tall’ echo to the comma-echo stage The TLV is embedded within a well-defined mesovortex. during which the bow-echo contains a single cyclonic line- The mesovortex is most apparent around 1.6 km AGL where end vortex (hereafter, LEV) at its northern point. Fujita’s the wind field from the TLV no longer obscures the wind conceptual model also proposed that tornadoes and field of the mesovortex (Fig. 1b). A current of high e air damaging winds were most likely just north of the apex of from east of the mesovortex circulation is wrapping around the bow-echo as well as at the southeastern tip of the the mesovortex at this height. Near the surface this high e comma-echo. More recent studies (e.g., Atkins et al. 2004, is present mainly on the south and west sides of the Wakimoto et al 2006b, Atkins and St. Laurent 2009a,b) have mesovortex. This suggests that a downdraft is likely present suggested that tornadoes and damaging winds in these to the south and west of the TLV and mesovortex. locations of bow echo are spawned by the development of sub-storm scale mesovortices. Mesovortices are typically 03:32Z Wed 9 May 2007 e,vort, wind the parent circulation for the tornadoes spawned by bow- Z = 0.15 KM AGL (a) 340. echoes (Trapp and Weisman 2003, Weisman and Trapp 339. 36.8 2003, Atkins et al. 2004). 337. In this study, we use the Advanced Regional 335. Prediction System (ARPS, Xue et al. 2003) model to 35.2 333. simulate one such case where several EF-0 and EF-1 331. tornadoes were spawned by a mesoscale convective system 33.6 329.

(MCS) that occurred on 8-9 May 2007 in Central Oklahoma. (km) 327.

Observations of the MCS show that the tornadoes were 32.0 325. associated with mesovortices that developed and rotated 323.

321. around a larger cyclonic LEV. A successful simulation of 30.4 the both the cyclonic LEV and associated mesovortices was 319. recently obtained by assimilating data from operational 317. 28.8 WSR-88D Doppler radars and a network of experimental X- 315. band radars of CASA (Center for Collaborative Adaptive Z=1.6 KM AGL Sensing of the Atmosphere) at 2 km and 400 m resolutions (b) 340. (Schenkman et al. (2009a,b), to be submitted to Mon. Wea. 339. 36.8 Rev.). In the present study, we further refine the grid 337. resolution, by placing a nested 100 m grid within the 400 m 335. grid, starting from the interpolated 400 m initial condition at 35.2 333. 0300 UTC. The increased resolution allows us to more 331. closely examine mesovortices as well as the genesis of 33.6 329. tornado-like vortices (hereafter, TLVs) in the model. (km) 327. 32.0 325. 323.

II. Results 321. 30.4 319.

Results from the 100 m simulations show that the 317. 28.8 model is able to resolve two distinct scales of vortices: 315. mesovortices and smaller scale TLVs. A mesovortex 22.4 24.0 25.6 27.2 28.8 30.4 32.0 develops in the northern portion of the leading edge of a 5.0 5.0 (km) FIG. 1: Horizontal wind vectors, vorticity (contoured in 2x10-3 surging rear-inflow jet around 0315 UTC. The mesovortex -3 intervals starting at 5x10 ) and e (shaded, K) at 0332 UTC at (a) has a core-diameter of about 5 km and extends from the Z=150 m AGL and (b) Z=1.6 km AGL. surface up to around 2.5 km AGL. TLVs are embedded within the mesovortex circulation. The TLVs are strongest

167 5th European Conference on Severe Storms 12 - 16 October 2009 - Landshut - GERMANY

To establish the cause of this downdraft as well as to understanding of the relationship between mesovortices and examine vorticity sources, backward trajectories are TLVs. calculated for parcels that end up near the surface south of the TLV. These trajectories show that a weak downdraft is present south of the mesovortex (Fig. 2). This downdraft appears to be forced by a downward directed pressure IV. AKNOWLEDGMENTS This work was primarily supported by NSF grant EEC- gradient force (PGF) caused by low-level low pressure 0313747, as part of ERC CASA. Partial support was also associated with the TLV. provided by ATM-0530814 and ATM-0802888.

Height of Parcel in Warm Downdraft 2200 V. REFERENCES

2000 Atkins N. T., Arnott J. M., Przybylinski, R. W.,Wolf R.A., 1800 Ketcham K. D., 2004: Vortex Structure and Evolution 1600 within Bow Echoes. Part I: Single-Doppler and Damage

1400 Analysis of the 29 June 1998 Derecho. Mon. Wea. Rev., 132 2224– 2242 1200 Atkins N. T., St. Laurent M., 2009: Bow Echo Mesovortices. Height above MSL (m) 1000 Part I: Processes that Influence Their Damaging Potential.

800 Mon. Wea. Rev., 137 1497-1513 Atkins N. T., St. Laurent M., 2009: Bow Echo Mesovortices. 600 Part I: Their Genesis. Mon. Wea. Rev., 137 1514-1532 400 0300 0303 0306 0310 0313 0317 0320 0323 0326 0329 0332 Fujita T.,1978: Manual of downburst identification for Time (UTC) project Nimrod., 104 pp. Perturbation Pressure Vs. Time Š100 Schenkman A. D., Xue M. Shapiro A., Brewster K., Gao, J, 2009: Impact of Radar Data Assimilation on The Analysis

Š150 and Prediction of the 8-9 May Oklahoma Tornadic Mesoscale Convective System. Part I: Mesoscale Features Š200 on a 2 km Grid. Mon. Wea. Rev., To Be Submitted Schenkman A. D., Xue M. Shapiro A., Brewster K., Gao J, Š250 2009: Impact of Radar Data Assimilation on The Analysis and Prediction of the 8-9 May Oklahoma Tornadic Pressure Pert. (Pa) Š300 Mesoscale Convective System. Part II: Sub-Storm Scale Mesovortices on a 400 m Grid. Mon. Wea. Rev., To Be Š350 Submitted Trapp R. J., Weisman M. L., 2003: Low-Level Mesovortices Š400 0300 0303 0306 0310 0313 0317 0320 0323 0326 0330 0333 within Squall Lines and Bow Echoes. Part II: Their Time (UTC) Genesis and Implications. Mon. Wea. Rev., 131 2804- FIG. 1: Time evolution of (top) parcel height and (bottom) 2823 perturbation pressure in a weak warm downdraft south of a TLV Trapp R. J., Tessendorf S. A., Godfrey E. S., Brook H. E., 2005: Tornadoes from Squall Lines and Bow Echoes. Part A backward trajectory is also calculated for a parcel I: Climatological Distribution. Wea. and forecasting, 20 that terminates within the TLV. Vorticity generation terms 23-34 via tilting and stretching are calculated for this trajectory Wakimoto R. M., Murphey H. V., Davis C. A., Atkins N. T., (not shown). These calculations show stretching is an order 2006: High Winds Generated by Bow Echoes. Part II: The of magnitude larger than tilting suggesting most of the Relationship between the Mesovortices and Damaging vertical vorticity came from the came from stretching of Straight-Line Winds. Mon. Wea. Rev., 134 2813-2829 existing vertical vorticity near the surface. Weisman M. L. Trapp R. J, 2003: Low-Level Mesovortices within Squall Lines and Bow Echoes. Part I: Overview and Dependence on Environmental Shear. Mon. Wea. III. SUMMARY AND CONCLUSIONS Rev., 131 2779-2803 In this extended abstract, results from a 100 m Xue M., Wang D. H., Gao J., Brewster K., Droegemeier K. resolution simulation of mesovortices and tornado-like K., 2003: The Advanced Regional Prediction System vortices initialized with real data are briefly discussed. (ARPS) Storm-Scale Numerical Weather Prediction These results show that TLVs are embedded within the Model. Meteor. Atmos. Physics, 82 139-170 circulation of a mesovortex. A warm downdraft is generated south of the mesovortex as high e air is forced to descend by a downward directed vertical PGF. Trajectory calculations also show that vorticity is predominantly amplified within TLVs via stretching. While the preliminary results presented herein have determined the cause of warm downdrafts as well as the dominant vorticity amplification terms, further work is needed to understand the temporal evolution of the mesovortices and TLVs. Additionally, the initial source of vorticity for the TLVs must be determined through more in depth analysis. We seek to obtain a more complete

168 5th European Conference on Severe Storms 12 - 16 October 2009 - Landshut - GERMANY

ENKF ANALYSIS OF THE 29 MAY 2004 OKLAHOMA CITY SUPERCELL USING RAPID-SCAN PHASED ARRAY RADAR DATA Therese E. Thompson1, Louis J. Wicker2, Douglas E. Forsyth2, Michael I. Biggerstaff1

1University of Oklahoma, 120 David L. Boren Blvd. Suite 5900 Norman, OK. 73072, USA, [email protected] 2National Severe Storms Laboratory, 120 David L. Boren Blvd. 4366 Norman, OK. 73072, USA, [email protected] (Dated: 15 September 2009)

I. INTRODUCTION system is the NSSL Collaborative Model for Multiscale The Phased Array Radar (PAR) is a new and unique Atmospheric Simulation (NCOMMAS). The Lin-Farley- Doppler radar (10 cm) that rapidly scans precipitating clouds Orville (LFO) microphysics scheme is used. The model via electronic beam steering. It is a part of the National domain is 120 km in the horizontal and 20 km in the Weather Radar Testbed (NWRT) in Norman, Oklahoma. vertical. The model grid moves to match storm motion, -1 The PAR is the first phased array radar system adapted to which was 6 m s to the east during the integration period. observe weather. On 29 May 2004 the PAR observed a The grid spacing is 2 km in the horizontal and 400 m in the tornadic supercell as it traversed central Oklahoma causing vertical. Forty ensemble members are used. The 2236 UTC considerable damage beginning near the town of Geary, OK Weatherford, OK sounding is used to initialize the ensemble and continuing eastward through northern sections of the members at 0030 UTC. Variations among the ensemble Oklahoma City metropolitan area. This data set represents members are created by randomly adding warm 5.0 K the first tornadic supercell observed using a rapid-scan 10- bubbles in the localized area where the observed storm cm radar, covering a 90 degree sector using seven elevation reflectivity is greater then 15 dBZ during the next twenty angles every 20-30 seconds during a three-hour period when minutes as in DW09. Each member is integrated for 20 the storm was within 150 km of the radar. Rapid-scan radar minutes before PAR data is assimilated. The ensemble data is believed to be a critical tool needed to produce real- spread is maintained using additive noise following DW09. time high-resolution storm-scale analyses and forecasts for The additive noise is applied every time a radar volume is the explicit prediction of severe weather such as hail and assimilated and the perturbation values were chosen to tornadoes. Accordingly, this study investigates whether the provide sufficient ensemble spread. Objectively analyzed use of rapid-scan data positively impacts storm-scale PAR radial velocity and reflectivity as well as no analyses and forecasts relative to less-frequent radar data precipitation observations are assimilated into the model volumes now available from the U.S. Weather Surveillance beginning at 0050 UTC. The reflectivity observations are Radars, 1988, Doppler (WSR-88D) radar system. Storm- not used to update potential temperature or water vapor scale EnKF techniques developed by Dowell and Wicker mixing ratio, as several studies have found this helps (2009, hereafter known as DW09) are used to assimilate the mitigate excessive model error growth in the boundary layer. PAR Doppler and reflectivity data from the 29 May 2004 One experiment assimilates a single PAR volume tornadic storm, producing storm-scale analyses and forecasts every five minutes that is similar to the data frequency of the that are compared to those produced by using less frequent operational WSR-88D network currently running in the data. United States. This experiment is therefore designated as “five-minute simultaneous”, hereafter denoted as 5min. A data assimilation experiment to assess the potential impact II. EXPERIMENT DESIGN of rapid-scan data uses a three-dimensional PAR volume The PAR data were collected over a 90° sector, with every minute. This experiment is called one-minute seven elevation angles, 0.75°, 2.27°, 3.78°, 5.28°, 6.78°, simultaneous (or rapid-scan), hereafter denoted as 1min. 8.28°, and 9.78°. As soon as a volume scan was completed a new scan started, making the times of the scans irregular III. RESULTS AND CONCLUSIONS but approximately every 20 seconds. This study uses data The results show that assimilating the one-minute from 0050 UTC to 0140 UTC because the tornadic storm data decreases the spin-up time (the time required for the was located within 100 km from the radar beginning at 0050 analysis to develop deep convection and capture the basic UTC. The gate spacing for the radar beam is 0.24 km. The storm structure) compared to the conventional five-minute PAR beamwidth is a function of the azimuthal angle from data. Results are evaluated by examining the analysis fields the flat-plated radar dish. At the center of the sector scan, for the characteristic structures of tornadic supercells. For the beam width is 1.5°, which then broadens to 2.1° at both example, after ten minutes of data assimilation (not shown), edges sector. The raw PAR data required hand editing to the 1min analysis has a well-developed storm having remove velocity and range aliasing as well as ground clutter supercell reflectivity structures, including a hook echo, near the radar. The radar data are then objectively analyzed strong precipitation core, forward flank, and mesocyclonic to a horizontal 2 km Cartesian grid on the individual PPI circulation in the wind field. However, the 5min experiment scan using a Cressman scheme. This analysis technique is shows a poorly organized and still-developing storm. used to thin the data horizontally to decorrelate observation Reduction of spin-up time is important toward being able to errors while not introducing error via vertical interpolations. initialize storm-scale forecasts quickly and accurately. The cloud-scale model used in the data assimilation After twenty minutes of data assimilation the 1min

169 5th European Conference on Severe Storms 12 - 16 October 2009 - Landshut - GERMANY z = 1 km reflectivity (Fig. 1) analyses depict a distinct hook appears to be most similar to the observed storm. The 1min echo and heavy precipitation core. The 5min reflectivity storm analyses contains typical supercell features after only analyses only depict a developing inflow notch. The hook 10-15 minutes of assimilation, while in the 5min experiment echo is believed to strongly reflect the presence of a deep these features are broader, weaker, and less mature. Rapid- mesocyclonic circulation, which is confirmed by examining scan radar data therefore “spins-up” the supercell storm the flow and vertical velocity. The analyzed wind field much more quickly in a storm-scale EnKF analysis system indicates a strong mesocyclone with nearly a closed relative to analyses generated via the current operational circulation, while the 5min analysis wind field only has data frequency. Rapid-scan data also helps reduce the weak cyclonic flow around the updraft. The 1min analysis errors relative to those seen in the conventional data experiment’s vertical velocity analysis has a stronger updraft assimilation, even after the conventional analyses have near the cloud base of approximately 14 m s-1 and what “caught up” and contain a mature convective storm. could be a rear-flank downdraft (RFD). The 5min analysis -1 has a weaker updraft near the cloud base (~8 m s ) and a IV. AKNOWLEDGMENTS less intense downdraft that is outside the mesocyclone The authors would like to thank National Severe Storms region of the storm. In the inflow region of the 1min Laboratory and the National Science Foundation for funding analyses at 0111 UTC (Fig. 1) there is a maximum in this work. Funding was provided by NOAA/Office of positive vertical vorticity nearly collocated with the Oceanic and Atmospheric Research under the NOAA- maximum in vertical velocity. The 5min analysis has University of Oklahoma Cooperative Agreement weaker updrafts and smaller values of vertical vorticity #NA17RJ1227, U.S. Department of Commerce as well as generally collocated in the inflow region. The motion and the NSF grant ATM-0802717, "Multiscale Analyses of vorticity are weaker and more spatial diffuse compared to Tornadic Storms using Multiparameter Mobile Radars". the 1min analyses. Therefore the 1min experiment has developed a more vigorous supercell compared to the 5min assimilation experiment. West to east vertical cross-sections V. REFERENCES slicing through the updraft of each 1min and 5min storm Dowell, D. C., and L. J. Wicker, 2009: Additive noise for (Fig. 1) also show significant differences in supercell storm-scale ensemble forecasting and data assimilation. structure and intensity. A Bounded Weak Echo Region J. Atmos. Ocea. Tech., 26, 911-927. (BWER) is present in the both cases. However, the 5min BWER is shallower with the largest reflectivity considerably downshear of the main storm updraft. This suggests a less mature storm. The 1min experiment has a strong updraft (maximum vertical velocity ~39 m s-1) collocated with the BWER location. Vertical vorticity values in the 1min experiment are indicative of a significant mesocyclone. Near the surface (the lowest model layer which is located at 0.2 km AGL, not shown) the 1min potential temperature analyses depict a more mature cold pool structure at this time. The 5min experiment’s cold pool is weaker and more disorganized. The 1min analysis has cold air under the forward flank and a region of warm air that could be descending air from the RFD on the upshear region of the storm. As the data assimilation continues, however, the temperature deficits within the cold pool grow too large. These deficits are believed to be much smaller in most tornadic supercells. Our experiments suggest the intensity of the cold air results from the accumulation of model error associated with the microphysical parameterization. The dynamics associated with an overly strong cold pool subsequently affect the whole storm system, and lead to a premature “gusting out” and demise of the analyzed supercell storm. Verifying convective storm analyses is difficult due to a lack of direct observations from within the convective system. The 29 May 2004 storm is unique in that a number of operational and research radars were able to observe the storm. However, using other radars to compare model results presents challenges due to the differences between the various radar systems and difficulties involved with pre- processing the data. Preliminary statistics indicate that the 1min experiment radial velocities inside the storm have lower root-mean-square innovations than the 5min experiment. Further results from these comparisons will be FIG. 1: 0111 UTC EnSRF analyses of vertical cross-sections of shown at the conference. reflectivity, vertical velocity, and the vertical vorticity for the (left) To summarize, the EnSRF experiment using the one- 1min assimilation experiment and (right) 5min assimilation minute PAR volumes generates a vigorous supercell that experiment.

170 5h European Conference on Severe Storms 12 - 16 October 2009 - Landshut - GERMANY

6(16,7,9,7<2)48$17,7$7,9(35(&,3,7$7,21)25(&$677262,/02,6785( ,1,7,$/,=$7,210,&523+<6,&63$5$0(7(5,=$7,21$1'+25,=217$/ 5(62/87,21 Kwinten Van Weverberg1, Nicole P.M. van Lizpig1, Laurent Delobbe2 , Dirk Lauwaet1

1'HSDUWPHQWRI(DUWKDQG(QYLURQPHQWDO6FLHQFHV.8/HXYHQ&HOHVWLMQHQODDQ(%/HXYHQ%HOJLXP NZLQWHQYDQZHYHUEHUJ#HHVNXOHXYHQEH 22EVHUYDWLRQV'HSDUWPHQW5R\DO0HWHRURORJLFDO,QVWLWXWHRI%HOJLXP$Y&LUFXODLUH%8FFOH%HOJLXP (Dated: 15 September 2009)

,,1752'8&7,21 realistic size distribution assumptions for the precipitating A major debate in the atmospheric mesoscale hydrometeors. All precipitating hydrometeors in the LFO83 modelling community is associated with the strong scheme are represented by exponential size distributions of

deficiencies in several aspects of the quantitative the form:

¢ ¡ ¢ ¡

, (1) precipitation forecast (QPF), despite a continuously = 0 −λ )exp()( improving representation of the physical processes and where N is the number of particles per unit volume per unit improving numerical techniques. The scope of the present size range, D is the maximum dimension of a particle and research is to gain insight in how operationally feasible N0x and λx are the intercept and slope of the exponential size modifications to aspects of the QPF interrelate and if current distribution respectively. While the intercept parameter of all insights for mainly idealized experiments could be hydrometeors is assumed constant, slope parameters, confirmed for real case simulations. We performed a number assuming all hydrometeors to be constant density spheres,

of sensitivity experiments, including modified soil moisture are determined by

£ £ content, more realistic microphysical size distribution ¨ 25.0 assumptions and enhanced spatial resolution. Two cases of £  πρ 0  , (2) λ =  £ 

 § 

¥ ¦ extreme convection were selected, one having strong vertical  ρ ¤  wind shear conditions and modest thermodynamic instability where ρ is the hydrometeor density, q the hydrometeor and a second having almost no vertical shear and strong x x mixing ratio and ρ is the air density. From many thermodynamic instability. observational data it is clear that intercept parameters are not

constant, but vary over many orders of magnitude. Therefore  we diagnosed the intercept parameter of rain using a mixing ,,02'(/6(783 ratio dependent relation (Zhang et al. 2008) and the snow In order to assess the impact of the sensitivity intercept parameter of snow using a temperature dependent experiments we employed the Advanced Regional relation (following Houze et al. 1979). Further, the constant Prediction System (ARPS - Xue et al. 2000 and Xue et al. density sphere relation for snow is found to be problematic 2001). The model was applied using one-way grid nesting in a number of studies to mainly cold season stratiform with two levels. Analysis data on a 0.25° horizontal precipitation. Hence we calculated the slope parameter and resolution from the global operational model operated by the terminal fall velocity of snow using V-D and m-D relations European Centre for Medium-Range Weather Forecasts found by Locatelli and Hobbs (1974) for graupellike snow

(ECMWF) were used as initial conditions and as 6-hourly as proposed by Woods et al. (2005)

 1

 ©

lateral boundary conditions for the model integration with a  ()+1



 0 () +Γ 1  , (3)  9-km grid spacing and a domain size of 1620 km × 1620 km.

λ =  

 Within this domain, a smaller domain, centred over Belgium  ρ  and covering 540 km × 540 km with a 3-km resolution was Last, Gilmore et al. found a strong sensitivity of the surface nested. Cloud microphysics was parameterized following precipitation to the way the hail/ graupel variable was Lin et al. (1983 further referred to as LFO83), including five parameterized in idealized cases of extreme convection. The water species (cloud water, cloud ice, rain water, snow and two microphysical size distribution experiments we hail). conducted had all modifications previously mentioned, but one experiment having the original formulations of large hail, (MIRSH) while in a second experiment we replaced ,,,(;3(5,0(17$/'(6,*1 A series of five sensitivity experiments has been these formulations with those typical for small graupel. We designed to compare the impact of modifications to soil increased the constant intercept parameter and used the m-D moisture content, size distribution assumptions and and V-D relations found by Locatelli and Hobbs (1974) for enhanced spatial resolution. A first set of two experiments lump graupel in the calculation of the slope parameter and consisted of a decrease of the initial soil moisture to the the terminal fall velocity (MIRSG). lowest recorded value by Nachtergaele and Poesen (2002) A last experiment was conducted to investigate the over one year in a Belgian loam area (DRYSOIL) and an influence of enhanced horizontal spatial grid spacing increase to saturated volumetric soil moisture content (GR1500) by decreasing the grid spacing to 1500 m as (WETSOIL). compared to the 3000 m of the CONTROL experiment. In a second set of experiments we applied more

171 5h European Conference on Severe Storms 12 - 16 October 2009 - Landshut - GERMANY

increased mean surface precipitation. Due to enhanced ,,,5(68/76$1'&21&/86,216 collection of snow by graupel as compared to CONTROL graupel volume and maximum graupel amount grow to Table 1 summarizes the surface precipitation much larger values. This leads to much broader precipitation statistics and a number of thermodynamic and heat balance swaths and larger accumulated surface precipitation. statistics for all experiments and both cases. During the Increasing the horizontal resolution in these cases clearly shear-driven case surface precipitation is mostly improved in improves the surface precipitation as convection was forced the DRYSOIL and the MIRSH experiments. The main on more realistic scales. reason for the improvements in the DRYSOIL experiment is the decreased buoyancy leading to lower precipitation  values. In the MIRSH experiment increased depositional ,9$.12:/('*0(176 growth of snow leads to more snow at the expense of cloud This research was carried out in the framework of the water and hail and hence slower precipitation fallout despite QUEST-B project, funded by the Flemish Fund for stronger updrafts. Replacing large hail by small graupel Scientific Research (FWO-Flanders). Further we would like (MIRSG) did not decrease the precipitation further, to acknowledge the Center for Analysis and Prediction of contradicting the results of Gilmore et al. (2004). Increasing Storms (CAPS) of Oklahoma University for providing the the horizontal resolution makes the precipitation fields very ARPS source code online and Jerry Straka and Ming Xue for noisy with widespread precipitation due to grid-scale storms. the fruitful discussions leading to the experiments carried  out. 6KHDU &2175 0,56+ 0,56* 3.1 2.6 2.7  0UU 95()(5(1&(6 ;UU 47.8 46.7 49.9 Gilmore, M.S., J.M. Straka, E.N. Rasmussen, 2004: &$3( 387.3 403.2 400.7 Precipitation uncertainty due to variations in /&/ 505.8 516.4 518.3 precipitation particle parameters within a simple

;Z 10.5 11.0 12.4 microphysics scheme.  !#"$#%& , 132, 2610-2627. ;FS 6.9 8.1 6.6 Houze, R.A., P.V., Hobbs, P.H. Herzegh, D.B. Parsons,

 *5 '5<62, :(762, 1979: Size distributions of precipitation particles in ,&-#&.0/21 frontal clouds. '!)(+* ., 36, 156-162. 0UU 5.3 2.7 2.9 Lin, Y.-L., R.D. Farley, H.D. Orville, 1983: Bulk ;UU 53.3 40.9 54.4

parameterization of the snow field in a cloud model. '! 4

343.1 312.1 413.4 354

,+ 0* (7686  92* &:) &$3( 1 , 22, 1065-1092. /&/ 480.8 592.9 480.9 Locatelli, J.D., P.V. Hobbs, 1974: Fall speeds and masses of

;Z 11.8 9.3 11.8 solid precipitation particles 8'!;<&=65>@?!-#"5 -  , 79, 2185- ;FS 5.7 7.3 6.7 2197. %XR\DQF\ &2175 0,56+ 0,56* Nachtergaele, J., J. Poesen, 2002: Spatial and temporal

variations in resistance of loess-derived soils to 4

0UU 11.1 11.1 13.7 4

 DE.!01

;UU 198.7 179.8 130.9 ephemeral gully erosion. A

 *5 '5<62, :(762, bulk microphysical scheme with snow habit prediction. ,&-#&.0/21 0UU 10.0 9.6 11.1 '!)(+* ., 64, 3927-3948. ;UU 139.7 190.2 206.0 Xue, M., K. K. Droegemeier, V. Wong, 2000: TheAdvanced &$3( 765.7 785.5 866.9 Regional Prediction System (ARPS) – A multi-scale /&/ 448.2 459.0 396.5 nonhydrostatic atmospheric simulation and prediction model. Part I: Model dynamics and verification.

;Z 16.0 16.9 4

#8:G @(* ,2-  H$>G?!-  5.2 4.5 F2* , 75, 161-193. ;FS Xue, M., K. K. Droegemeier, V. Wong, A. Shapiro, K. TABLE I: Precipitation and (thermo)dynamic statistics of all model experiments (Mean surface precipitation (Mrr), Maximum surface Brewster, F. Carr, D. Weber, Y. Liu, D. Wang, 2001: precipitation (Xrr), Convective Available Potential Energy (CAPE), The Advanced Regional Prediction System (ARPS) – A Lifted Condensation Level (LCL), Maximum vertical velocity (Xw) multi-scale nonhydrostatic atmospheric simulation and

and Maximum cold pool temperature deviation from domain prediction tool. Part II: Model physics and applications.

4

#8:G @(* ,2-  H$>G?!- average temperature (Xcp)). F2* ., 76, 143-165. Zhang, G., M. Xue, Q. Cao, D. Dawson, 2008: Diagnosing In the buoyancy-driven case, the DRYSOIL the intercept parameter for exponential raindrop size

experiment yields similar improvements of the moist distribution based on video disdrometer observations:

4 354 4

7F&* &:)J 8K 1 ,L 0*  &M ? processes and the surface precipitation, but the MIRSH model development. '!!(I686 , experiment does not yield clear improvements. There is a 47, 2983-2992. clearly increased vertical velocity as buoyancy increases associated with the increased depositional growth of snow. CAPE is indeed the only driver of the convection in this case and hence the increased buoyancy compensates for slower precipitation fallout. The MISRG experiment in this case leads to a decreased peak precipitation but a strongly

172 5th European Conference on Severe Storms 12 - 16 October 2009 - Landshut - GERMANY

A COMPARISON OF IMPINGING JET AND COOLING SOURCE DOWNBURST MODELS Brian C. Vermeire1, Leigh G. Orf2, Eric Savory3

1University of Western Ontario, 1151 Richmond Street, London, ON, Canada, [email protected] 2Central Michigan University, Mount Pleasant, MI, USA 3University of Western Ontario, 1151 Richmond Street, London, ON, Canada

15 September 2009

I. INTRODUCTION based on the stability affected log law. Nondimensional A thunderstorm downburst is defined as an intense wind shear is specified according to Hogstrom (1996). downdraft of air that induces an outburst of damaging winds The forcing function for the cooling source models on or near the surface of the earth. The structure, evolution has an ellipsoidal shape identical to that in Lin et al. (2007). and atmospheric conditions conducive to downburst events The centre experiences the greatest cooling, decreasing in have been documented (Fujita 1985), showing that magnitude towards the outer edge. The intensity of the downbursts are caused by liquid water loading and cooling function increases from 0 K/s to a peak of -0.08 K/s thermodynamic cooling. This produces a dense, cool during the simulation. The ellipsoid has horizontal and downdraft of air that descends from the base of a vertical half-widths of 1200 m and 1800 m respectively and thunderstorm cloud. A leading edge roll vortex is formed is centred at a height of 2000 m. Surface roughness levels of due to shear instability with the surrounding air. The 0.1 m, 0.03m, 0.01 m, 0.003 m and 0.001 m are investigated. downdraft then impinges on the ground and spreads radially The impinging jet simulations are performed at Re = 105 at high velocities, propelling the roll vortex in front of the with an Hj/Dj ratio of 2.0. These are chosen according to the outflow. availability of experimental data and the practical limitations Previous studies on downbursts include both a of future experiments. Surface roughness levels of 10-4, 10-5 -6 cooling source approach, as taken in Lin et al. (2007), and an and 10 Dj are considered. impinging jet model after Kim and Hangan (2007). The The scaling approach used in this study is based on cooling source model uses a specified cooling function in a a Galilean-invariant frame of reference, fixed to the centre of dry adiabatic atmospheric model. This cooling function is the roll vortex. This allows for a direct comparison of the specified to mimic the effects of melting and sublimation in simulated outflows, independent of the inlet conditions. This an actual downburst event, resulting in a physically realistic avoids scaling issues as a downburst, unlike an impinging simulation. The impinging jet model takes advantage of jet, does not have a universal length or velocity scale. The observed similarities in the outflows of downbursts and length scales used to compare the two outflows include the laboratory scale jets. The advantage of the impinging jet is horizontal and vertical vortex diameters (DV and HV), the that it can be simulated in a laboratory setting; however it is height of the vortex centre (CY), and the horizontal and limited by physically unrealistic forcing. The objective of vertical locations of the peak velocity (xUp and yUp) as shown the present work is to outline the similarities of these two in Figure 1. The dimensions xUp and yUp are measured from modelling methods. This is performed by using a novel the surface and the left hand side of the primary vortex vortex scaling approach, solving problems associated with respectively. the scaling of thunderstorm downbursts.

II. PRESENTATION OF RESEARCH Both the models are solved using large eddy simulation (LES) with the Bryan Cloud Model (CM1) (Bryan 2002). This model is completely dry, utilizing a Klemp-Wilhelmson time splitting scheme for integration of acoustic waves. A 5th order scheme is used for horizontal and vertical advection. Subgrid turbulence is handled by the FIG. 1: Length scales used for comparing vortex structures. k-ε turbulence model. The computational domain for the cooling source simulations extends to 3.5 km in both This yields four separate nondimensional groups that can be horizontal directions and 4.0 km vertically. Horizontal grid compared, specifically xUp/D, yUp/D, CY/D and the aspect spacing is constant 10 m, while vertical spacing stretches ratio of the vortex DV/HV. The location of the peak velocity from 1 m at the surface to 50 m at the top of the domain. For in the domain is readily available, but the determination of the impinging jet model, a constant horizontal grid spacing the vortex parameters DV, HV and CY require an objective and frame independent vortex identification method. The of 0.01 jet diameters (Dj) is used, extending to 3.5 Dj. The method employed is based on the q-surface after Hunt et al. vertical grid stretches from 0.001 Dj at the surface to 0.07 Dj at the top of the domain. The forcing functions are centred (1988), allowing the extents of the vortex to be defined as on one vertical edge of the domain for both models and the any interconnected region with positive q, where: two mating faces are symmetry boundary conditions. This allows for the simulation of one quarter of a downburst event, as both models are axisymmetric. The other two faces and top of the domain are specified as outflow boundary conditions. The surface is modelled by a drag coefficient, This allows the location of peak velocity to be scaled by the

173 5th European Conference on Severe Storms 12 - 16 October 2009 - Landshut - GERMANY jet diameter as shown in Figure 2, where each point represents a different time in the simulation.

FIG. 4: Scaled vortex height both sets of simulations during times * of high outflow velocity (dt = tVj/Dj for the impinging jet).

For the cooling source simulations it is apparent that the FIG. 2: Scaled locations of peak velocity for impinging jet (IJ) and vortex centre starts at CY/D = 0.4 at the beginning of the cooling source (CS) simulations. outflow period. It then varies depending on the level of surface roughness. It is apparent that the impinging jet It is observed that UP for the impinging jet vortex height is relatively insensitive to the amount of simulations occurs in one grouped region A, and in two surface roughness. For all cases, the impinging jet vortex distinct grouped regions, B and C, for the cooling source centre is higher than that of the cooling source model. simulations. For the impinging jet simulations the horizontal Therefore, it cannot be concluded that they produce similar location remains between 20-40 percent of Dj. By outflow features. decreasing the surface roughness, the point of peak velocity tends to move closer to the surface. For the two roughest III. RESULTS AND CONCLUSIONS cooling source models, the point of peak velocity tends to Both cooling source and impinging jet downburst occur towards the front of the vortex and high above the simulation have been performed. The outflow features from surface in group B. For the smooth cooling source models, these simulations were scaled based on the extents of the the point of peak velocity is extremely close to the surface, primary roll vortex, resulting in a direct comparison of the under 5 percent of Dj. It is evident that there is no overlap outflows. It has been shown that there is no similarity between regions A and B or regions A and C, suggesting between the location of peak velocity, height of the roll that the model outflows are not similar. vortex above the surface and vortex aspect ratio. Based on The second comparison that can be made is these results it can be concluded that the outflow from an between the roll vortex aspect ratios for both modelling 5 impinging jet of Re 10 and Hj/Dj = 2 is not similar to that of methods. These are shown in Figure 3 for both sets of a realistic downburst event. This suggests that an simulations. investigation of the Reynolds number effects on transient impinging jet outflows is necessary.

V. REFERENCES Fujita T., 1985: The downburst microburst and macroburst. University of Chicago, Department of Geophysical Sciences, 128 pp

Lin W., Orf L., Savory E., Novacco C., 2008: Proposed large-scale modelling of the transient features of a downburst. Wind and Structures, 10 315– 346

Kim J., Hangan H., 2007: Numerical simulations of impinging jets with application to thunderstorm downbursts. Journal of Wind Engineering and Industrial FIG. 3: Aspect ratio both sets of simulations during time of high * Aerodynamics, 95 279– 298 outflow velocity (dt = tVj/Dj for the impinging jet). Bryan G., 2002: An investigation of the convective region of The cooling source aspect ratio is highly sensitive to the numerically simulated squall lines. PhD Thesis. The level of surface roughness. With the exception of the 0.003 Pennsylvania State University, 181 pp m roughness case, all aspect ratios lie between 1 and 2. The impinging jet simulations are insensitive to the level of Hogstrom U., 1996: Review of some basic characteristics of surface roughness. All aspect ratios lie between 0.8 and 1, the atmospheric surface layer. Boundary Layer significantly different from the cooling source results. Meteorology, 78 215– 246 The final nondimensional group to be compared is the relative height of the roll vortex from the surface. These Hunt J., Wray A., Moin P., 1988: Eddies, stream and are shown in Figure 4 for both simulation cases. convergence zones in turbulent flows. Center for Turbulence Research Report, CTR-S88 p. 193

174 5h European Conference on Severe Storms 12 - 16 October 2009 - Landshut - GERMANY

Composite characteristics of severe thunderstorms over Bangladesh simulated by WRF-ARW Model B.R.S.B. Basnayake, Someshwar Das, M.K. Das, M. Rahman, M.A. Sarker and M.A.R. Akand

SAARC Meteorological Research Centre (SMRC), Dhaka 1207, Bangladesh, [email protected] (Dated: 15 September 2009)

I. INTRODUCTION III. RESULTS AND CONCLUSIONS

Severe thunderstorms known as Nor’westers are About 15 Nor’wester events observed over among the most common natural phenomena in Bangladesh Bangladesh and it’s surrounding, during pre-monsoon and adjoining northeastern parts of India, especially during season in 2008, were selected for the study. Rainfall rates pre-monsoon season. These systems are embedded within (mm/h) of selected events, derived from Dhaka radar, are squall lines, which travel several hundred kilometres and shown in Fig. 2. Figure 3 shows vertical structure of these cause high number of loss of life and property each year. events in terms of rainfall rate (mm/h) derived from 2A25 Several hundreds people died and few ferries were capsized TRMM data. Selected Nor’wester events were simulated by due to lightning and gusty winds triggered by the WRF-ARW model to investigate the structure of the Nor’westers during the pre-monsoon season of 2008. These systems. systems develop mainly due to merging of cold dry northwesterly winds and southerly low level warm moist winds from the Bay of Bengal. As the lifetime of these systems is only few hours, prediction of these systems is challenging using any conventional forecasting technique. These systems develop mainly due to merging of cold dry northwesterly winds and southerly low level warm moist winds from the Bay of Bengal. As the lifetime of these systems is only few hours, prediction of these systems is challenging using any conventional forecasting technique.

II. DATA AND METHODOLOGY

Advanced Research WRF (ARW) model, developed by the National Centre for Atmospheric Research (NCAR) of USA is utilized to simulate Nor’wester events in and around Bangladesh. NCEP-FNL data is utilized as initial and lateral boundary conditions (LBCs) at six hourly intervals. Domain is selected to cover the whole Bangladesh at 9 km FIG. 2: Rainfall (mm/h) observed by Dhaka Radar during horizontal resolution and 27 vertical sigma levels. Kain- Nor’wester events. Fritch cumulus parameterization scheme is used for simulating all the events. Similarly, surface layer is treated using Monin-Obukhov with Carslon-Bolan viscous sub- layer option and boundary layer is treated with Yonsei University scheme.

FIG. 3: Vertical structure of rainfall rate (mm/h) associated with Nor’westers derived from 2A25 TRMM data.

Model simulated Nor’wester events are shown in figure 4. WRF model is able to simulate the rainfall FIG. 1: Map showing meteorological stations in Bangladesh. associated with the events with some spatial and temporal

175 5h European Conference on Severe Storms 12 - 16 October 2009 - Landshut - GERMANY shifting. Figure 4(a) shows model simulated rainfall pattern Composite characteristics of 15 selected events are at 1500 GMT on 05th May 2008. It turns out that model calculated (Table 1). It is interesting to see that the model underestimates the intensity of this event as it unable to simulated direction of movement, speed of movement and simulates double-squall line shape system. Model simulated precipitation rate at the surface of the system is well Nor’wester event on 30th May 2008 is shown in figure 4(b). comparable with the observations. However there are some Intensity of model simulation is somewhat comparable with differences of the characteristics simulated by the model the radar observation (Fig. 3). However it simulated with a with the observations. temporal shifting of about 4 hours. Characteristics Observation Model Cloud Top Altitude (km) 13.1 15.2 Altitude of Core precipitation 3.6 7.3 (km) Intensity of Core precipitation 21.3 107.8 (mm/h) Precipitation rate at surface 29.7 26.5 (mm/h) Direction of movement (°) 293 263 Speed of movement (km/h) 47.8 48.7 Maximum wind speed at surface 20 7.1 (m/s) Length of Squall line (km) 186.3 271.4 Updraft speed (max) (m/s) NA 2.33 Downdraft speed (max) (m/s) NA 0.36 Liquid water content (mg/m3) NA 610

TABLE I: Composite characteristic of selected Nor’wester events.

IV. AKNOWLEDGMENTS

The authors would like to thank National Centre or Atmospheric Research (NCAR) for providing WRF model, NASA-JAXA for the TRMM data, Bangladesh Meteorological Department (BMD) for providing Radar observations. They also wish to thank Director of SAARC (a) Meteorological Research Centre (SMRC) for providing facilities to conduct this research at the Centre.

V. REFERENCES

Das Someshwar, Basnayake B.R.S.B., Das M.K., Akand M.A.R., Rahman M.M., Sarker M.M.A., 2009: Composite Characteristics of Nor’westers observed by TRMM and Simulated by WRF Model. SMRC Report No. 25 44pp. Das Someshwar, Ashrit, R., Moncrieff, M.W., 2006a: Simulation of a Himalayan Cloudburst Event. Journal of Earth System Science, 115(3) 299-313. Islam M.N., Terao T., Uyeda H., Hayashi T., Kikushi K., 2005: Spatial and temporal variations of precipitation in and around Bangladesh. J. Meteo. Soc. Japan., 83 21-39. Islam M.N., Uyeda H., 2008: Vertical variation of rain intensity in different rainy periods in and around Bangladesh derived from TRMM observations. Int. J. Climatology, 28 273-279. Karmakar S., 2001: Climatology of thunderstorm days over Bangladesh during the pre-monsoon season. Bangladesh J Sci. and Tech., 3(1) 103-112. Karmakar S., Alam M.M., 2007: Troposphere moisture and its relationship with rainfall due to Nor’westers in Bangladesh. Mausam, 58, 153-160. Prasad K., 2006: Environment and synoptic conditions associated with Nor’westers and tornadoes in Bangladesh (b) – An appraisal on Numerical Weather Prediction (NWP) FIG. 4: Model simulated rainfall (mm/h) of Nor’wester guidance products. SMRC Report No. 14, 75pp. events (a) 05th May 2008 (b) 30th May 2008.

176 5h European Conference on Severe Storms 12 - 16 October 2009 - Landshut - GERMANY

THE INFLUENCE OF BOUNDARY LAYER CONDITIONS ON STORM LIFE CYCLES Mladjen Ćurić 1 , Dejan Janc, Nemanja Kovačević

1 Institute of meteorology, University of Belgrade, Dobračina 16, [email protected].

I. INTRODUCTION wind components, perturbation potential temperature and pressure, turbulent kinetic energy, mass concentration of Convective storms are considered to be amongst the water vapour, cloud water, cloud ice, rain, hail and snow and most devastating weather phenomena that cause great seeding agent as well as the number concentrations of rain, demage to crops and property, because they are associated cloud ice, hail and snow. by hail events and flash floods in small areas. Cloud- For the simulation presented in this paper, the model resolving mesoscale models (CRM) can be used to reliably was configured with the domain 64km× 64km× 18km with forecast such events. The reliability of the model products the 600 m grid spacing in horizontal and 300 m in vertical. strongly depends on the boundary layer characteristics taken The simulations were terminated at 80 min. Long and short for the initialization of the model. Single-soundings were time steps are 3s and 0.5s respectively. The wave-radiating sufficient to enable successful forecasting of maximum condition is applied for lateral boundaries. The upper cloud tops by using one-dimensional convective cloud boundary with the Rayleigh spongy layer is used, while the models (e.g. Ćurić and Janc, 1993). Meanwhile many lower boundary is free slip. disadvantages of single soundings become apparent by using Model bulk microphysics treats two categories of two and three-dimensional convective cloud models as well non-precipitating (cloud water and cloud ice) and three as the CRM models in recent time. This is due to the fact categories of precipitating elements (rain, hail and snow). that such initial state is not fully consistent with real case Rain, hail and snow are each represented by an exponential that characterizes non-homogeneity in horizontal. On the size spectrum. Cloud water and cloud ice spectra are other hand convective storms are frequently initialized in supposed to be monodisperse. Two-moment bulk interval between two subsequent routine soundings. This microphysical scheme is used following Murakami (1990). requires the adjustment of routine soundings data in The turbulence is treated by 1.5-order turbulent kinetic boundary layer regarding both time and location. energy formulation. The Coriolis force is neglected in our There are at least two ways in which these problems simulations. can be solved. The first one is using several soundings over The reference state is homogeneous in the horizontal small area ( ∼ 10,000km2 ) performed two and more times a using a single sounding giving the values of temperature, day. Only few countries are capable to provide the dense humidity, pressure, wind velocity and direction. The model network of soundings sites due to high cost. The other one is cloud is initiated by introducing an ellipsoidal warm bubble data interpolation from larger-scale models in order to with 1.5 K amplitude in its centre having a horizontal radius provide more real data distribution in time and space. of 10 km and a vertical radius of 1.5 km. The coordinates of However such models mainly cannot provide CRM models the warm bubble centre are (x, y) = (16, 40) km in the with data that are necessary for successful storm horizontal and 1.5 km in the vertical. The midnight Belgrade initialization. In praxis the CRM models therefore use single soundings of 13 July 1982 is used. Temperature profile in soundings that are close in time with occurence of adjusted regarding to storm initialization time by complex convective storm (Swan, 1998) or idealized soundings with radiation-low equations adopted for clear sky conditions that associated hodographs for different sensitive studies (Ćurić are favourable for convective cloud occurence. Location of et al., 2003; Gilmore et al., 2004; van den Heever and storm initialization is over Zlatibor plateau with expressed Cotton, 2004). Our investigation is targeted to improvent of low-level winds from the Western Morava valley to initial conditions given by a single sounding. This would be mountain slopes in time of storm occurence. This initial performed by the adjustment of temperature and wind conditions are favourable for isolated storm formation at this Ć ć profile in the boundary layer taking into account both place part of Serbia ( uri , 1982). Low-level winds have opposite and time of a storm initialization. We select for our direction compared to winds in original sounding. The upper investigation the Western Serbia region as well-known level wind is mainly from NW direction, while its speed source region for individual convective storms. varies from 5 m/s near the ground to about 18 m/s at 10 km height. At lowest 3 km the water vapour mixing ratio II. DESCRIPTION OF THE MODEL reaches its maximum value of 14.5 g/kg at p=880 mb.

The model used is developed by Ćurić et al (2003; III. RESULTS AND CONCLUSIONS 2006). This model numerically integrates the time- dependent, nonhydrostatic and fully compressible equations. For purpose of this study we shall present two The model uses the generalized terrain-following coordinate model cases with the CRM model performed by Ćurić et al. in the vertical, while the horizontal coordinates are the same (2003). The first one refers to original sounding (OS case), as in the Cartesian system. while the other one refers to modified sounding (MS case) The model’s basic prognostic variables are: Cartesian regarding temperature and wind profiles in the boundary layer. The storm occurs by radar at 10.15 GMT over Zlatibor

177 5h European Conference on Severe Storms 12 - 16 October 2009 - Landshut - GERMANY slopes and it moves towards the Western Morava valley. speed as well as the accumulated precipitation at the surface Some results of model storm simulation are presented in crucially depend on initial conditions in the boundary layer. Figs. 1-4. Such results would not be essential if they cannot match the observations. Meanwhile, the good agreement between radar and model storm characteristics is occurred for analyzed storm. Also, the accumulated precipitation data taken from the dense rain-gauge network agree well with their model counterparts. Investigations presented here give the idea how the problem of initial boundary-layer conditions for storm initialization by the CRM model may be solved in successful way.

FIG. 1: Visual appearances of reflectivity greater or equal to 20 dBZ as viewed from the west at t=54 min of simulated time for theOS case.

FIG. 2: As in Fig. 1 but for the MS case.

Fig. 1 and 2 show clearly that the simulated storm radar reflectivity fields are quite different for the OS and MS cases. In the OS case only single-cell storm is simulated moving fastly in NW-SE direction with mid-tropospheric wind. In the MS case the storm is the multi-cell one that FIG.4: As in Fig. 3 but for the MS case. propagates more slowly towards the Western Morava valley. With regard to accumulated precipitation at the ground, the IV. AKNOWLEDGMENTS OS case shows only one cell with maximum around 8 mm at t=80 min as it is presented in Fig. 3. In contrast the This research was supported with the Ministry of accumulated precipitation encircles much wider area having Science of Serbia. We gratefully acknowledge Mr. D. two distinct cells with local maxima for the MS case (Fig. Bulatović for technical preparing of pictures. 4). This is attributed to the storm splitting in sharp-sheared environment (van den Heever and Cotton, 2004). V. REFERENCES

Ćurić, M.,1982: The development of the cumulonimbus clouds which move along a valley. In: Cloud Dynamics (eds. Agee,E.M., Asai,T.). Dordrecht: D. Reidel, 259-272. Ćurić, M., Janc, D., 1993: Predictive capabilities of a one- dimensional convective cloud model with forced lifting and a new entrainment formulation. J. Climate Appl. Meteor., 32, 1733-1740. Ć ć ć č ć uri ,M., Janc,D.,Vujovi ,D.,Vu kovi ,V., 2003: The effects of a river valley on an isolated cumulonimbus cloud development. Atmos. Res., 66 123-139. Gilmore, M.S., Straka J.M., E.N. Rasmussen,E.N., 2004: Precipitation uncertainty due to variations in precipitation particle parameters within a simple microphysics scheme. Mon. Wea. Rev., 132 2610-2627. Murakami, M., 1990: Numerical modeling of dynamical and

microphysical evolution of an isolated convective cloud- The 19 July 1981 CCOPE cloud-. J. Meteor. Soc. Japan , FIG. 3: Distribution of cumulative precipitation at the surface for 68 107-128. the RS case at t=80 min with contour intervals of 2 mm starting at Swan, H, 1998: Sensitivity to the representation of 2mm. precipitating ice in CRM simulations of deep convection. Atmos. Res., 47-48 415-435. Given results clearly show that the storm van den Heever, S.C., Cotton, W.R., 2004: The impact of development, cell-organization, cloud life cycle, propagation hail size on simulated supercell storms. J. Atmos. Sci., 61 1596-1609.

178 5th European Conference on Severe Storms 12 - 16 October 2009 - Landshut - GERMANY

Numerical simulations of supercells over idealized orography

Paul Markowski1 and Nikolai Dotzek2

1Penn State University, University Park, Pennsylvania, USA, [email protected] and 2Deutsches Zentrum fur¨ Luft- und Raumfahrt (DLR), Institut fur¨ Physik der Atmosphare,¨ Oberpfaffenhofen, Germany, [email protected]

I. INTRODUCTION The model was initialized with a sounding similar to that used by Weisman and Klemp (1982). The environmental wind Despite decades of observing and simulating deep moist profile is defined by the quarter-circle hodograph used by Ro- convection, we know little about how the underlying orogra- tunno and Klemp (1982). The hodograph was shifted in three different ways in order to vary the terrain-relative winds. One phy influences convective storms. At the 5th ECSS we report −1 on our recent investigation of the effects of idealized orogra- wind profile has 4 m s surface easterlies (in the far field, away from the hill), one has calm surface winds, and the other phy on supercell storms. −1 Although many investigators, at least anecdotally, express has 4 m s surface westerlies. The problem we are studying little doubt that terrain can have an appreciable effect on con- is not Galilean invariant, which adds extra dimensions to the vective storms, there are few formal papers on the influence parameter space (this is generally the case when one wishes of terrain on convective storms. The primary difficulty with to include the effects of the lower boundary, whether surface observational studies (e.g., Hannesen et al. 2000; LaPenta et fluxes and/or sloping terrain are included). al. 2005; Bosart et al. 2006; these and other references are Storms are initiated with a warm bubble, and an eastward- available upon request) is that it is never possible to know moving supercell results in every case. The ground-relative how the storms would have evolved in the absence of terrain. motion of the supercell, as well as the upslope and downslope Thus, observational work tends to remain fairly speculative side of the terrain, varies depending on the location of the about the impact of terrain on the observed structure and evo- hodograph trace relative to the origin of the hodograph. A lution of convection. A numerical modeling approach ought north-south oriented hill having a height of 500 m and zonal to be better suited for this line of work, for models allow the half-width of 10 km was centered at different longitudes user to compare a simulation with terrain against a simulation (ranging from x = 85–145 km), depending on the storm mo- without terrain (e.g., Frame et al. 2006; Curi´ c´ et al. 2007). tion, so that the storm would cross the ridge at approximately the same time in each of the simulations. Herein we present The present study on the influence of terrain on supercells the results for the experiments in which the supercells crossed uses idealized terrain rather than actual terrain. It is much eas- the hills at approximately t = 2 h. Other hill locations/hill ier for us to develop a dynamical understanding of the cause- passage times were tested—we report only the results that and-effect relationship of the storm-terrain interactions if the are robust and do not depend on the exact timing of the hill- terrain configuration is kept simple. Below we briefly sum- crossing. marize the simulations with a two-dimensional hill parallel to the y axis. At the ECSS we will present the results of addi- tional simulations with two-dimensional terrain [e.g., we also III. RESULTS investigated the effects of a storm passing over an idealized valley, similar to the case observed by Bosart et al. (2006)], as well as simulations with three-dimensional terrain (e.g., an In general, the simulated supercells weaken (in terms of isolated hill, channeled flow through a mountain gap). In the both low-level and midlevel updraft strength and vertical vor- case of three-dimensional terrain, there is also the possibility ticity) on the lee slopes of the hills (Figs. 1 and 2), where the of the storms interacting with preexisting terrain-induced vor- lee and windward slopes are defined relative to the direction of tices (Smolarkiewicz and Rotunno 1989; Epifanio and Durran the surface wind, not the storm motion [e.g., in the case of the 2002). hodograph trace that is shifted to the left of the origin, there is easterly low-level flow; thus, the eastern (western) slope of the hill is the windward (lee) slope]. The results turn out to be fairly intuitive in that the supercells simply appear to be II. METHODOLOGY responding to changes in environmental convective inhibition (CIN) and relative humidity that are induced by the airflow The simulations were performed using the Bryan Cloud over the terrain. For example, in the lee of a hill, isentropic Model 1 (Bryan and Fritsch 2002) with a terrain-following surfaces are depressed (a hydraulic jump may be observed if vertical coordinate. The horizontal grid spacing is 500 m; the the flow is very strong, depending on the ambient stratifica- vertical grid spacing varies from 100 m in the lowest 1 km to tion) and relative humidity is anomalously low. Both effects 500 m at the top of the domain. The lower and upper bound- contribute to anomalously large CIN on the lee slope. In some aries are free-slip (sensitivity tests were performed with sur- of the cases, some modest intensification of the storms oc- face drag at the lower boundary); a Rayleigh sponge occupies curs on the windward slope (e.g., Fig. 1; notice the lifting of the uppermost 4 km of the model domain. the isentropes on the windward slope—the lifting was associ-

179 2

easterly ground-relative surface winds westerly ground-relative surface winds a) a)

1:40 2:40 2:00 2:20 2:20 2:40 1:20 1:40 3:00 w 1:00 0:40 1:00 1:20 1 km w1 km 2:00 10 7.5 5 m s-1 intensification weakening 10 7.5 5 m s-1 weakening

b) θ b) θ

c) RH c) RH

FIG. 1: Model output from a simulation with westerly low-level, FIG. 2: As in Fig. 1, but for a simulation with easterly low-level, ground-relative winds and a hill centered at x = 145 km (the hodo- ground-relative winds and a hill centered at x = 85 km. graph is shown schematically at the top). (a) Rainwater fields at 1 km at 20-min intervals (color shading), with vertical velocity at 1 km overlaid (contours). The light (dark) brown shading indicates the re- the airflow over it is itself a very difficult problem outside of gion where the terrain elevation is ≥90% (≥50%) of the height of the a limited number of idealized situations, as many have de- hilltop. (b) Vertical cross-section of steady-state potential tempera- voted a significant fraction of their careers to studying this ture in a simulation with the same terrain configuration but without a problem alone (e.g., Smith 1979, 1989; Durran 2003). (Most storm. (c) As in (b), but relative humidity (%) is displayed. studies looking at terrain-induced waves only consider rela- tively simple upstream wind and temperature profiles; super- cell hodographs not only have shear but may have large varia- ated with humidification and CIN reduction). In the case of tions of shear direction and magnitude with height, and sound- calm ground-relative winds at the surface (not shown here)— ings often have large vertical variations in static stability.) But the case in which the hill perturbs the isentropic surfaces the if one knows how a hill affects the isentropic surfaces, then least—the hill has the smallest effect on the supercell. it seems fairly straightforward to determine the effects on en- vironmental CIN and relative humidity, and ultimately the ef- fects of the hill on the overlying storm. IV. CONCLUSIONS

Regardless of the set-up of the numerical simulation, changes in storm evolution relative to the control can be at- V. ACKNOWLEDGMENTS tributed to changes in the environment that are associated with airflow over hill, with the environment on the lee slope be- The lead author is grateful for the support of the Deutsches ing more hostile to the storms than the far-field environment Zentrum fur¨ Luft- und Raumfahrt (DLR), Institut fur¨ Physik and windward-slope environment in terms of CIN and rela- der Atmosphare,¨ where he has been a Visiting Scientist in the tive humidity. Predicting the details of how a hill influences summer and fall of 2009.

180 5th European Conference on Severe Storms 12 - 16 October 2009 - Landshut - GERMANY

LONG-LASTING DEEP CONVECTIVE SYSTEMS IN THE MEDITERRANEAN BASIN: A MODEL STUDY Pasqui M., S. Melani, B. Gozzini and F. Pasi

Institute of Biometeorology, Via G. Caproni, 8, 50145 Florence, Italy, [email protected] (Dated: 15 September 2009)

I. INTRODUCTION parameterization short and long wave scheme and the Land Ecosystem Atmosphere Feedback scheme (LEAF-3) for soil Long lasting deep convective storms are of particular – vegetation – atmosphere energy and moisture exchanges, interest because of their potential damaging power. Many described in Walko et al. (2000). studies, analyzing the dynamical characteristics of these Using the RAMS model, simulations for these severe precipitating systems, highlighted their stationary behaviour convective events were performed at very high horizontal as one of the most important dynamical features. It is and vertical resolution. Grids specifications following: for responsible of large rainfall amount and can produce each event a nested grid approach with three grids at casualties and hazards over a large area. Such specific different horizontal resolution: 32km, 8km and 2km in order dynamical feature is the consequence of interaction between to guarantee the proper description of large to local scale large and local scale atmospheric circulation. Thus it is dynamics features. The 2km inner grid has a stretched quite difficult to reproduce with numerical models as both vertical spacing from 22m, near surface, to 800m, in the dynamical and thermo – dynamical description must be well troposphere. The atmospheric forcing was provided by the simulated (Krichak and Levin, 2000). NCEP – NCAR Reanalysis 6 – hours dataset for the coarse 4 different storms (see Fig.1 and Fig.2), three over grid, while the one – way nesting from coarse to medium, sea: 4th Dec 2004, 4th Nov 2008 and 8th Jan 2009; and one and from medium to fine grid is set every 1 hour. Observed over land (18th Sep 2007), were simulated by means of a non sea surface temperature (SST) from MODIS was used in – hydrostatic, high resolution numerical model (the Regional order to provide a good description of vapour exchanges Atmospheric Modelling System – RAMS) and analysed between sea and atmosphere boundary layer. RAMS model using remote sensed data (see Melani et al., 2009, this provides a full comprehensive simulation system that is able Conference). to represent atmospheric evolution of such a severe Among these events, the December 4th, 2004 storm is convective systems. of particular interest. It developed over the Tyrrhenian Sea, between Corsica Island and Italy as a stationary long live severe convective storm. It lasted over 8 hours, producing heavy rains even far from the convective downdraft areas and intense wind speed. Over the anvil cloud top a so-called V – shape ice plumes was observed. This unusual cloud top feature is composed of very small ice particles with a very high 3.9µm reflectivity values, as observed and modelled by Melani et al. (2003a, b). Fig.1: 4th December 2004, 12:41 UTC. NOAA – 16 AVHRR channel 4 image of the cloud top temperature (Event A).

II. PRESENTATION OF RESEARCH $" The Regional Atmospheric Modeling System, RAMS, has been used operationally at Consortium LAMMA (http://www.lamma.rete.toscana.it) and National Research Council (http://www.ibimet.cnr.it) since 1999. The latest !" #" RAMS version, 6.0, has been used for this study as Fig. 2: Thermal infrared satellite images form MSG (Met 9, Ch 10): modelling component (Meneguzzo et al. (2004), Soderman 18th Sep 2007, 13:30 UTC (event B); 4th Nov 2008, 5:30 UTC et al. (2003), Pasqui et al. (2005)). A general description of (event C); 8th Jan 2009, 17:00 UTC (event D). the model can be found in Pielke et al. (1992), while a technical description can be found on the ATMET web site (http://www.atmet.com). III. RESULTS AND CONCLUSIONS Today RAMS represents one of the state-of-the-art models in atmospheric science and it is continuously In RAMS model, providing a full comprehensive improved on the basis of a multi-disciplinary work. The simulation system, was able to represent atmospheric physical package of the model describes a number of evolution of such a severe convective systems. Using the atmospheric effects: a two-way interactive nested grid detailed parameterisation scheme for cloud microphysics structure, an atmospheric turbulent diffusion processes dynamics, it reproduces quite accurately the main feature of according with the Mellor-Yamada scheme, a cloud convective storms (see Fig.3). In particular location, cloud microphysics parameterization (Walko et al., 1995, Meyers top properties (skin temperature and shape) and rainfall et al., 1997), modified Kain-Fritsch type cumulus amount of active cells were represented. In general the parameterization, the Harrington radiative transfer stationary phase was reproduce, even if duration was in 5th European Conference on Severe Storms 12 - 16 October 2009 - Landshut - GERMANY general shorter than the observed one. Topography plays a Hydrology, 288, 37-56, 2004. key role for the studied events since the triggering Pasqui M., Baldi M, Giuliani G., Gozzini B., Maracchi G. mechanism is due to the instabilities generated by the and Montagnani S., 2005. Operational Numerical Weather interaction between large scale flow and mountain chain Prediction systems based on Linux cluster architectures, respectively over Corsica, Sardinia, Appennini and Etna Nuovo Cimento, 28 C, N.2, DOI 10.1393/ncc/i2005- mountain. The stationary phase was always guaranteed by 10183-4. the evaporation from the positive sea surface temperature Pielke, R. A. & Coauthors, 1992: A comprehensive anomaly for cases A, C and D, and from the positive soil meteorological modelling system-RAMS. Meteor. Atmos. moisture anomaly present for case B. Phys., 49, 69– 91. Walko, R. L., W. R. Cotton, M. P. Meyers, and J.Y. Harrington, 1995: New RAMS cloud microphysics parameterization. Part I: The single-moment scheme. Atmos. Res., 38, 29–62. Walko, R.L., L.E. Band, J. Baron, T.G.F. Kittel, R. Lammers, T.J. Lee, D. Ojima, R.A. Pielke, C. Taylor, C. Tague, C.J. Tremback, and P.L. Vidale, 2000: Coupled atmosphere-biophysics-hydrology models for environmental modeling. J. Appl. Meteor., 39, 931-944.

Fig.3: Event A modelled by RAMS. Total condensate concentration: in white it is shown the isosurface at 0.1 g/kg along with horizontal wind streamlines at 14km above ground level.

IV. AKNOWLEDGMENTS

The authors would like to thank Francesca Guarnieri, Graziano Giuliani for their support and the LAMMA Consortium for availability of satellite data used in such research.

V. REFERENCES

Krichak, S.O. Levin, Z., 2000: Mesoscale Simulation of Life Cycle of Cloud Microphysics During Hazardous Weather Conditions in the Southeastern Mediterranean Atmos. Res., 53, 63-89. Meyers, M. P., R. L. Walko, J. Y. Harrington, W. R. Cotton, 1997: New RAMS cloud microphysics parameterization. Part II: The two-moment scheme. Atmos. Res., 45, 3-39. Melani, S., E. Cattani, V. Levizzani, M. Cervino, F. Torricella, and M.J. Costa, 2003a: Radiative effects of simulated cirrus clouds on top of a deep convective storm in METEOSAT Second Generation SEVIRI channels. Meteor. Atmos. Phys., 83, 109-122. Melani, S., E. Cattani, F. Torricella, and V. Levizzani, 2003b: Characterization of plumes on top of deep convective storm using AVHRR imagery and radiative transfer simulations. Atmos. Res., 67-68, 485-499. Melani, S, M. Pasqui, A. Antonini, A. Ortolani and V. Levizzani: The synergy of GEO-LEO satellite observations in analysing enhanced-V features on top of severe storms. ECSS, 2009 Meneguzzo F., M. Pasqui, G. Menduni, G. Messeri, B. Gozzini, D. Grifoni, M. Rossi and G. Maracchi, “Sensitivity of meteorological high-resolution numerical simulations of the biggest floods occurred over the arno river basin, Italy, in the 20th century.”, Journal of 5th European Conference on Severe Storms 12 - 16 October 2009 - Landshut - GERMANY

NUMERICAL SIMULATION OF TORNADO-SCALE VORTICES OCCURRED IN A WINTER COLD-AIR OUTBREAK OVER THE SEA OF JAPAN Kazuhisa Tsuboki1, and Atsushi Sakakibara1

1Hydrospheric Atmospheric Research Center (HyARC), Nagoya University, Furo-cho, Chikusa-ku, Nagoya, 464-8601 Japan, [email protected] (Dated: 15 September 2009)

I. INTRODUCTION following coordinate is used for the vertical coordinate. Tornado-scale intense vortices are called “tatsumaki” Time integration is performed by the mode-splitting in Japanese. Both tornado and waterspout are called as technique. Cloud microphysics is a bulk method of cold rain. tatsumaki in Japan. It occurs below intense convective One-dimensional model is used for the computation of the clouds. Niino et al. (1997) showed that a majority of surface temperature. tornadoes occurs in coastal regions of Japan and that about In the present study, one-way triple-nesting 12 % of tornadoes are associated with winter monsoon. technique was used for the high-resolution simulation of When a cold-air outbreak occurs over the Sea of Japan, intense vortices of tatsumaki with 3 different horizontal shallow convective clouds develop over the sea. They are resolution of 2000 m, 250 m and 50 m. The initial and usually ordinary cells and have short life time. Two major boundary conditions were provided by the forecast model types of tornado-genesis are known; one is a supercell type output data of JMA (the Japan Meteorological Agency). The and the other is non-supercell type along a meso-front. The initial time is 00 UTC, 25 December 2005. The domains are mechanism of tatsumaki is unknown. Since a tatsumaki in shown in Fig.1. the winter monsoon occurs over the sea and its life time is short with their small scale, there are almost no reports about tatsumaki in the coastal regions of the Sea of Japan except Kobayashi et al (2007). They observed a tatsumaki in the coastal region using a Doppler radar and photograph and showed its horizontal scale is 150 m at a height of the cloud base. Although a tatsumaki occasionally causes severe disasters along the coastal region, its characteristics and structure have been unknown for long time. On 25 December 2005, a train-accident occurred in the coastal region owing to a strong gust wind and 5 people were killed. Since the gust wind associated with snowstorm has not been studied well in the region, its mechanism was not identified. A tatsumaki is most possible phenomenon for the gust wind. Even though the tatsumaki frequently occurs when the cold-air outbreak of the winter monsoon occurs over the sea, it is still a mistily in a snowstorm. Observation FIG. 1: Computational domains for the triple-nested simulation is very difficult because of its small scale and short life time. experiments. The outermost, middle and innermost regions are A high-resolution numerical simulation is a possible those of 2000 m, 250m and 50 m resolution simulations. approach to clarify characteristics and structure of the tatsumaki in snowstorm. Since the horizontal scale is an order of 100 m, a very high resolution of a several 10 m is III. RESULTS AND CONCLUSIONS necessary and the computation is very large. The huge parallel computer named the Earth Simulator makes it The simulation of 50 m resolution was performed for possible such huge simulation of tatsumaki. 2.5 hours. The cold-air outbreak occurred over the Sea of For simulations of high-impact weather systems, we Japan. The front of the cold air pushed the warm and moist have been developing a cloud-resolving model named southwesterly toward the southeast. Shallow convective CReSS (the Cloud Resolving Storm Simulator) (Tsuboki, clouds developed in the cold air of the winter monsoon. A 2004; Tsuboki and Sakakibara, 2001, 2002). The purpose of large number of intense vortices were simulated in the the present study is to clarify characteristics and structure of domain of the 50 m simulation. We found both positive the tatsumaki associated with the snowstorm of winter and negative vortices of tatsumaki in the simulation. Typical monsoon in the coastal region using the cloud-resolving vortices of positive and negative tatsumakis are shown in model. Figs. 2 and 3, respectively. Both tatsumakis have absolute value of vorticity of about 0.3 s-1 and maximum wind speed reaches about 30 m s-1. Their horizontal scale is about 200 m. II. NUMERICAL MODEL AND This is almost the same scale that the observed winter EXPERIMENTAL DESIGN tornado observed by Kobayashi et al. (2007) The surface data were output every 20 seconds and The basic equations of the CReSS model are the used for traces of each vortex. The minimum absolute value non-hydrostatic and compressible equations. Terrain of vorticity is 0.2 s-1 for a tatsumaki vortex and 10 km

183 5th European Conference on Severe Storms 12 - 16 October 2009 - Landshut - GERMANY regions (200 grids) from the inflow boundary were excluded The velocity field of both positive and negative from the analysis. The total number of vortices is 418. tatsumakis is balanced with the pressure field (the Positive and negative vortices are 58 and 42 %, respectively. cyclostrophic balance). Both positive and negative vortices Histogram of vorticity (Fig. 4) shows that most tatsumakis of tatsumaki are accompanied with negative pressure have vorticity between 0.2 and 0.3. The number decreases deviation. Figure 5 shows a scatter diagram of vorticity and significantly with increase of vorticity. A small number of pressure deviation. The increase of vorticity roughly tatsumaki were found with vorticity larger than 0.3 s-1. corresponds to the increase of pressure deviation. Its maximum deviation reaches about 4 hPa. Both positive and negative tatsumakis accompany intense wind of 30-35 m s-1. The maximum wind speed reaches 40 m s-1. The relationship between vorticity and wind speed is not clear (figure is not shown).

FIG. 2: An example of typical positive tatsumaki. Thick and thin contours are vorticity and wind speed at the surface, respectively. The arrows are horizontal wind vectors at the surface.

FIG. 5: Scatter diagram of vorticity of tatsumaki vs. pressure deviation at the surface obtained in the 50 m simulation. Data of vorticity between -0.2 and 0.2 are excluded.

IV. AKNOWLEDGMENTS

The present study is a part of the research project led by Dr. W. Ofuchi, JAMSTEC. The simulations of this work were performed by the Earth Simulator.

V. REFERENCES

Kobayashi, F, Sugimoto, Y., Suzuki, T., Maesaka, T., and Moteki, Q., 2007: Doppler radar observation of a tornado FIG. 3: The same as Fig. 2 but for a negative tatsumaki. generated over the Japan Sea coast during a cold air

outbreak. FamilyName N., FamilyName N., Year: Title. J. Meteor. Soc. Japan, 85, 321–334. Niino, H., Fujitani, T., and Watanabe, N., 1997: A statistical study of tornadoes and waterspouts in Japan from 1961 to 1993. J. Climate, 10, 1730-1752. Tsuboki, K., 2004: High resolution modeling of multi-scale cloud and precipitation systems using a cloud-resolving model. Annual report of the Earth Simulator Center, April 2003.March 2004, 21-26. Tsuboki, K. and Sakakibara, A., 2001: CReSS User's Guide 2nd Edition, 210p. Tsuboki, K. and A. Sakakibara, 2002: Largescale parallel computing of Cloud Resolving Storm Simulator. High Performance Computing, Springer, H. P. Zima et al. Eds, 243-259.

FIG. 4: Histogram of vorticity of tatsumaki obtained in the 50 m simulation. Vorticity between -0.2 and 0.2 s-1 are not counted.

184 5h European Conference on Severe Storms 12 - 16 October 2009 - Landshut - GERMANY

AN INVESTIGATION OF A SEVERE MULTICELLULAR STORM IN THE TROPICS Ulrike Wissmeier1, Robert Goler2

1Meteorological Institute, Theresienstr. 37, 80333 Munich, Germany, [email protected] 2Meteorological Institute Munich, [email protected] (Dated: 15 September 2009)

I. INTRODUCTION stretched from 120 m at the bottom of the domain to 1 km at the top. Convection is triggered by an axis-symmetric Studies of tropical thunderstorms have shown that vertical thermal perturbation of horizontal radius 4 km and vertical wind shear has a large influence on storm evolution (e.g., extent 500 m. A temperature excess of 2 K is specified at the Wissmeier and Goler 2009) and that convergence caused by centre of the thermal and decreases gradually to 0 K at its sea breezes can trigger thunderstorms. In their study of edge. convection in the Darwin area in the “Top End” of northern A number of experiments is performed with a Australia, Keenan and Carbone (1992) found that initial combination of a northerly (Nsb), westerly (Wsb), and/or a convective features tend to evolve towards a line of northwesterly sea breeze (NWsb). Each sea breeze is thunderstorms, which is oriented perpendicular to the low- initialised using a box of cold air in the north, west, and level shear and shows forward-moving squall-like northwest of the domain at the beginning of the simulation, characteristics. A particular phenomenon occurring in the respectively. The depth and temperature excess of the sea Darwin region is the so-called “Northeaster”, which is a breezes are zsb = 2 km and θsb = -2 K, respectively, and these multicellular storm complex or squall line approaching the values are based on observations of sea breezes in the coastal city from the northeast. A good example of a Darwin region (Todd Smith, Darwin Regional Forecasting Northeaster passed over Darwin during the afternoon of 14 Centre, personal communication; May et al. 2002). Each November 2005. The automatic weather station at the airport experiment is run for 180 minutes, thereby allowing the was hit by lightning and stopped recording during the storm. initial updraught and the subsequent storm system enough The storm produced wind gusts of up to 93 km h-1 which time to develop. uprooted or snapped trees along a 1 km stretch of a highway adjacent to the airport, and the outward bound section of the III. RESULTS AND CONCLUSIONS highway was blocked. Power supplies to many residents were disrupted for up to an hour. It is the observations The basic experiment is initialised with a Nsb, a NWsb, and relating to this storm that motivate the formulation of the a thermal perturbation which is placed so that convection idealised numerical calculations described herein. develops at the Nsb front. Horizontal cross-sections through The aim of this work is to perform idealised the initial updraught with its cold pool and the subsequent numerical simulations relevant to the Northeaster that storm system at mid-levels are shown in Figures 1a and b at occurred on 14 November 2005, and to investigate the times t = 50 and 110 min after model initialisation, influence of the environmental vertical wind shear on the respectively. The model output compares well with reality in evolution of the model storm system that develops. A the following respects: particular focus is to examine how the additional lifting and low-level vertical wind shear provided by the sea breeze(s) • Initial cell (see Fig. 1a) progresses to the west; lead to the formation of a severe multicell complex. In the • New updraughts develop on the gust front of the course of studying the formation of new updraughts, the initial cell, but behind the sea breeze front applicability of the Rotunno-Klemp-Weisman-criterion (formation of a multicell complex, see Fig. 1b); (Rotunno et al. 1988) is tested. The observations of the 14 November case serve as a reality check on the numerical • New cell development along the southern flank of solutions. the multicell complex; • Propagation speed and direction of the multicell complex (40 km h-1, west-northwest); II. MODEL CONFIGURATION • Orientation of the multicell complex (north- northeast/south-southwest); and To study multicell thunderstorms, especially the Northeaster of 14 November 2005, and the circumstances which leads to • Length of the line of convection (25 to 30 km). their formation the three-dimensional, non-hydrostatic cloud-scale model of Bryan and Fritsch (2002) and Bryan The role of the two sea breezes in the evolution of the (2002) is used where the ice microphysics scheme is multicell complex on 14 November 2005 is studied by included. modifying the basic experiment. A total of seven sensitivity Convection is initialised in an environment with experiments is considered here. Four experiments the wind, vertical temperature, and moisture profiles taken investigate the importance of the sea breezes: no sea breezes from the 14 November 2005, 0000 UTC Darwin sounding. (EXP1); only a Nsb (EXP2); only a NWsb (EXP3); and the To represent the mid-afternoon conditions when the NWsb replaced by a Wsb (EXP4). Two experiments use the Northeaster developed, the lowest 1 km of the sounding is basic experiment with the position of the initial convection modified to give a convectively-mixed boundary layer. The changed: convection is triggered ahead of the Nsb (EXP5); calculated CAPE based on these values is 4129 J kg-1. and convection is triggered behind the Nsb (EXP6). The The model domain size is (90 x 60 x 28) km3, with basic experiment, initialised with different idealised wind a horizontal grid spacing of 1 km, and a vertical grid, profiles, provides the basis for EXP7.

185 5h European Conference on Severe Storms 12 - 16 October 2009 - Landshut - GERMANY

regions where new cell development at the gust a) front of the initial updraught is most likely. • The low-level convergence at the edge of the cold pool needs to be strong and persist for a sufficient time so that new cells can develop. • Large horizontal convergence at the gust front is achieved if the strength of the cold pool is comparable to that of the opposing environmental flow. Whether the vertical shear generated by the environmental wind is approximately balanced by the gust front shear can be determined via an equation similar to the Rotunno-Klemp- Weisman-criterion (Rotunno et al. 1988). • The gust front is strong if the initial updraught is tilted significantly, so that the downdraught does not fall into the buoyant air, but supplies a specific region of the cold pool continuously with cool air. • Updraught tilting is caused by strong environmental wind shear and by the vorticity generated by the sea breeze(s). • Even though the convergence at the gust front is strong, convection can be suppressed by subsidence from pre-existing neighbouring cells. b) The aim of this work has been to provide a deeper understanding of the overall evolution of the Northeasters and of the factors which lead to the development of thunderstorm complexes with squall-like characteristics. Even though the experiments were motivated specifically by thunderstorms in the Darwin area, the basic principles and relations itemised above are likely to apply also to other regions in the tropics and mid-latitudes.

IV. ACKNOWLEDGEMENTS

We wish to express our gratitude to Roger K. Smith for his thoughtful comments, and to George Bryan for kindly making his model available. The first author is grateful to the German Research Foundation (Deutsche Forschungs- gemeinschaft) for providing financial support for this study, and to the European Meteorological Society for endowing a travel support fund (Young Scientist Travel Award).

V. REFERENCES

FIG. 1: Mid-level storm structure depicted at a) t = 50 and b) 110 Bryan G. H., 2002: An investigation of the convective region min. The sea breeze fronts and the gust front are denoted by the of numerically simulated squall lines. Ph.D. thesis, The dotted (red) and solid (black) lines, respectively, and represent the Pennsylvania State University, 181 pp. -0.5 K temperature perturbation contour. Vectors represent Bryan G. H., Fritsch J. M., 2002: A benchmark simulation horizontal flow at z = 4.6 km, and the total precipitation mixing for moist nonhydrostatic numerical models. Mon. Wea. -1 ratio qn=qr+qs+qg is contoured in grey at 2 g kg intervals, with the Rev., 130, 2917-2928. zero contour omitted. Regions of updraught velocities at z = 4.6 km -1 Keenan T. D., Carbone R. E., 1992: A preliminary larger than 5 m s are shaded. morphology of precipitation systems in tropical northern Australia. Quart. J. Roy. Meteor. Soc., 118, 283-326. The sensitivity experiments, EXP1 to EXP7, show the May P. T., Jameson A. R., Keenan T. D., Johnston P. E., following: Lucas C., 2002: Combined wind profiler/polarimetric radar studies of the vertical motion and microphysical • A sea breeze can supply lifting to the initial cell characteristics of tropical sea-breeze thunderstorms. Mon what leads to a stronger updraught, to more Wea. Rev., 130, 2228-2239. precipitation loading within the thunderstorm, and Rotunno R., Klemp J. B., Weisman M. L., 1988: A theory for to a stronger downdraught and gust front than if strong, long-lived squall lines. J. Atmos. Sci., 45, 463-481. there had been no sea breeze or a sea breeze which Wissmeier U., Goler R., 2009: A comparison of tropical and is located far away from the initial cell. mid-latitude thunderstorm evolution in response to wind • A strong updraught and downdraught are not shear. J. Atmos. Sci., 66, 2385-2401. indicators as to whether a (large) multicell complex will develop. Large low-level horizontal convergence is the primary factor determining the

186 5th European Conference on Severe Storms 12 - 16 October 2009 - Landshut - GERMANY

FORECASTING SKILL STUDY OF DIFFERENT NON-HYDROSTATIC METEOROLOGICAL MODEL CONFIGURATIONS IN SEVERE CONVECTIVE EVENTS SIMULATION A.Bertozzi1, P.Randi1

(1) MeteoCenter - Epson Meteo Competence Center , via A. Einstein 8 c/o APC-CNR Faenza (RA), ITALY, [email protected]

(Dated: 15 September 2009)

I. INTRODUCTION general, the non-explicit models were calculated by means This study is part of a direct evaluation of the best of a fixed extension grid measuring 1200x1200 km and 50 possible configuration for a mid to high resolution model to vertical levels hyperbolic tangent vertically distributed. be adapted as an RUC for the central Mediterranean and The explicit models are nested into non-explicit parent another high resolution model for explicit convection domain (1200x1200 km big) with 2 or 3 nesting levels. intended for use in making forecasts over Italy. The part of The explicit domains were sized at 600x400 km in order to the work presented in this poster is intended to test two types attain an integral representation of the Po River Valley and of model simulations on a case study of very severe Northern Adriatic PBLs. convective events that occurred over the south and south- The number of vertical levels is 65, always with a eastern Po River Valley, often with ruinous results. These hyperbolic tangent distribution. events were all either underestimated or completely ignored The simulations were extended for 48 hours with the by mid to mid-high resolution global models or LAM. convective event centred between +18 and +36 hours in Actually, these are all phenomena (50-60% of the overall order to remove all potential spin-ups. cases of severe convective events over the south and south- The set of severe convective events taken into consideration eastern Po River Valley) that cannot be attributed to direct in the area being studied numbered about 20 outbreaks synoptic frontal action, but only by dry-lines or between 1999 and 2008, all of which manifested similar mesofrontogenesis locally generated or prefrontal instability. characteristics, like various intensities of hail storms, downbursts up to 100 km/h wind gusts, precipitation with rain-rates generally in excess of 100 mm/h, and a few II. PARAMETRIZED AND EXPLICIT vortical phenomena (land spouts, waterspouts, and a pair of CONVECTION MODELS RESPONSE : suspected mesocyclonic tornadoes). However, in nearly all 20 OUTBREAKS REFORECASTING cases, none were correctly simulated by operational models (1999-2008, 45 TOTAL CONVECTIVE (GCM above all). EVENTS) Series of test simulations were conducted to evaluate the responsiveness of a set of mid resolution models run setted up with a series of parameterization schemes combinations, The particular mesoscale conditions that occur under the (PBL, Surface physics, low level control convective circumstances being studied in the south and south-eastern schemes, and deep layer convective schemes, surface layer) , part of the Po River Valley are particularly difficult for for the purpose of profiling the model performances almost all convection parametrization schemes whether low themselves for use as RUC for the central basin of the level or deep layer control. Mediterranean, The second part of the research, still That environmental condition are frequently characterised ongoing, is directed in Data assimilation systems by deep dry intrusion, low relative hygrometry in the cloud In addition to the study on the configuration between 20 and layer and the presence of very energetic PBL, difficul 8 km, an in-depth analyses with re-forecasting of the events parametrization of entrainment-detrainment due to complex on finer grid domains (from 6 to 1 km) were conducted to orographic profile and heterogeneous environmental evaluate the impact of different types of microphysics, PBL, conditions produced by rapidly changing heat flux dynamics and modified parameters of Landuse for the purpose of from nearby sea (responsible to complex PBL energy improving the skill of the model in managing convection, budget) and convective triggering made by mountain-plain which generally appears to be overestimated. solenoidal circulation. -Modifications in Landuse- NON-EXPLICIT MODELS (20- Contemporaneously, in nearly all the conditions analysed, 8 km) there are upper level southern or south-western currents that 3 different configurations of LANDUSE were adopted: the give rise to more uniform and rich hygrometric profiles first (cited as LU0) is the standard one; the second (cited as north of the Po, where parameterization schemes often tend LU1) was provided with a 3% to 9% increase of the to overestimate phenomena and concentrate widespread parameters of surface roughness (see table 1). rainfall maximums. In the third (cited as LU2), there were further significant Such events are very frequently underestimated, wrongly corrections concerning the same parameters (see table 1). identified, or completely ignored by mid resolution global -Modifications in land use - EXPLICIT MODELS (6-1 km) models and LAM with parameterized convection. 3 different configurations of LANDUSE were adopted: The model used derives from the WRF framework, using The first (cited as LU0) is the standard one; ARW (EM) and NMM core dynamics with variational the second (cited as LU1) was provided with a -4% and initialization and with or without the nested domain. In -8,25% in the parameters of surface roughness (see table 2)

187 5th European Conference on Severe Storms 12 - 16 October 2009 - Landshut - GERMANY

In the third (cited as LU2) there were further significant maximum sometimes of several hundreds kilometers. The corrections concerning the same parameters (see tables) scheme skill deteriorates reducing grid space The modifications were made by evaluating: - KF put from 80 to 95% of rainfalls on Alps where RH 1-) overestimates as high as 20-30% and higher of the profile in cloud layer tends to be more rich due to surface wind during the advection phase for models with predominant sotherly or south westerly mid tropospheric domains between 15 and 8 km winds. 2 ) increasing undrestimation below 5 km resolution relative - GR scheme behaviour is more heterogeneous; it produces to tests conducted for the sensitivity of the grid/bias less rainfall on Alps and a little bit more on Apennines when resolution on the wind. These tests were conducted for wind coupled to MYJ PBL, a lot of rainfalls (strongly power use overstimated) on large rainfalls maximums in central Po Valley or northerly Po Delta when copuled to BouLac PBL Land use Parametrized convection and partially with Pleim PBL or frequently similar to KF with Pleim PBL but with less rainfalls on Alps roughnesss Soil moisture - BMJ has adeguate skill to locate correctly rainfall LU1 LU2 LU1 LU2 maximums, expecially with BouLac PBL, but always produces strong understimation of 48h cumulated rain Urban and Built-Up Land 5,20% 11,40% - - Explicit convection models Dryland Cropland and Pasture 7,10% 14,20% - - - a 3 km grid domain seems to be the best solution here for Deciduous Broadleaf Forest 3,50% 7,00% 4,00% 8,00% depiction of severe storm in southern part on Po Valley , Evergreen Needleleaf Forest 4,50% 9,00% 4,00% 8,00% coarse model is less able to resolve correctly mesoscale convective features Industrial or Commercial 2,50% 5,00% 4,00% 8,00% - the convection localization skill increase costantly until 2 High Intensity Residential 6,20% 10,00% 4,00% 8,00% km grid, after that the improvement is small or negligible - the convective rainfall overstimation dramatically explode Low Intensity Residential 7,00% 12,00% 4,00% 8,00% below 3 km grid in southern Po Valley, below 5 km in the Barren or Sparsely Vegetated 5,00% 10,00% 2,00% 4,00% rest of domain, the Ferrier microphysics extremize this behaviour Mixed Forest 5,00% 10,00% 4,00% 8,00% - Microphysics Lin and WSM 6-class have a better skill in Cropland/Woodland Mosaic 6,50% 12,00% 2,00% 4,00% contanining rainfalls overstimation below 3 km and have Cropland/Grassland Mosaic 7,00% 13,00% 2,00% 4,00% better skill in localizing convective overtures - Variational initialization give a better skill in cell Irrigated Cropland and Pasture 7,50% 13,00% 2,00% 4,00% localization and trigger timing but it can't contain rainfall TABLE I: landuse modifications in 20km -8 km grid models (non overstimation explicit) - Landuse parameter fine tuning reduces convective rainfall overstimation from 20 to 35%; the reduction of landuse Land use Explicit convection roughess below -15% con drive too large negative bias, roughnesss Soil moisture otherwise, +15% of rougness lenght drive to large negative temperature bias to (and positive wind bias) , due to domain LU1 LU2 LU1 LU2 convection and cold pool effects “explosions” and increased Urban and Built-Up Land -3,00% -5,00% - - downdrafts number. A change in summer landuse roughess and soil moisture contents parameter respectively from -8 Dryland Cropl. Pasture -2,00% -4,00% - - and -12% and -4% seems to be the best for overstimation Deciduous Broadleaf Forest -7,50% -13,00% -2,00% -4,00% bias Evergreen Needleleaf Forest -7,50% -13,00% -2,00% -4,00% - spatial correlation get worse below 2 km only due to very large rainfall overstimation despite better localization skill Industrial or Commercial -2,50% -5,00% -2,00% -4,00% High Intensity Residential -6,20% -12,00% -2,00% -4,00% V. REFERENCES Low Intensity Residential -7,00% -12,00% -2,00% -4,00% Kain and Fritsch: - one dimensionale entraining deteraining plume model nad -5,00% -10,00% -2,00% -4,00% Barren or Sparsely Vegetated application in convective parametrization (j.Atmos.Sci.) Mixed Forest -5,00% -10,00% -2,00% -4,00% - the role of convective trigger functions in numerical foreecast of convective systems (1992) M.A.phys Cropland/Woodland Mosaic -6,50% -12,00% -2,00% -4,00% - Convective parametrizations for mesoscale models (1993) Cropland/Grassland Mosaic -7,00% -13,00% -2,00% -4,00% – AMS Irrigated Cropl. Pasture -7,50% -13,00% -2,00% -4,00% - Kain Kain Fritsch convective parametrization, an update (2004) TABLE 2 landuse modifications in 20km -8 km grid models (non explicit) Gallus: - Eta simulation and three extreme convrctive events, Sensitivity to resolution and convective parametrizations III.RESULTS AND CONCLUSIONS (1999) - j.Atmos.Sci. Parametrized convection models Gallus and Johnson: - PBL scheme play a foundamental rule in convective - Heat and moisture budget iof an intensive midlatitude parametrization skill due to latent heat and sensible heat squall line (1991) - j.Atmos.Sci. flux from Adriatic and heat and energy budget inside Po Stensrud: Valley Parametrization schemes, Cambridge press (2007) - KF scheme sistematically fails to correctly locate rainfall

188 5th European Conference on Severe Storms 12 - 16 October 2009 - Landshut - GERMANY

CHARACTERISTICS OF CONVECTIVE PROCESSES IN INLAND NORTHEAST SPAIN Francisco Espejo1, Evelio Álvarez2, Francisco Cortés3, Cristina Lafragüeta3

1AEMET-State Meteorological Agency, Aragon Regional Office, Spain, [email protected] 2AEMET-State Meteorological Agency, La Rioja Regional Office, Spain, [email protected] 3SODEMASA, Aragon Regional Government, Spain, [email protected], [email protected] (Dated: 15 September 2009)

I. INTRODUCTION considered. FIG. 2 shows the mean number of strokes per The interior of the Northeastern Iberian Peninsula Km2 and warm season (a) and the mean storm days per presents some special characteristics regarding convective warm season (b), whereas FIG. 3 shows the seasonal processes. Due to both the proximity to the Atlantic Ocean distribution of precipitation maxima (a) and the percentage and the Mediterranean Sea, the effect of high mountain of summer rainfall to the annual total (b). ranges (the Pyrenees and the Iberian Range), and the As seen in FIG. 2a, the Eastern Iberian Range presence of a high portion of elevated plateau-like terrain in presents the maximum density of strokes, whereas as for the latter, convection in this area presents records of mean storm days FIG. 2b shows two maxima, a relative one frequency and severity in Spain. Especially during the warm coinciding with the maximum of stroke density and an season (April-September), storm severity features such as absolute one located in the Pyrenees. the number of lightning strokes, high maxima of rainfall, severe hail and even F3 tornadoes can be observed in the area.

FIG. 2: Climatology of lightning stroke density (a) and days of storm (b) per warm season (April-September). FIG. 1: Area of study. On the other hand, FIG. 3 shows that in spite of its II. PRESENTATION OF RESEARCH proximity to the Mediterranean, it is in the summer when the In this work we follow the line of research presented Eastern Iberian Range has more rainfall (a), which seems to in ECSS 2007, extending the study period to warm seasons be linked to the higher frequency of storms during the warm (April-September) between 2002 and 2008. The aim is to season (FIG. 2). characterize the convective processes in the area and to establish their main driving factors. The first part of the study presents a classification of 26 convection-prone situation types, shown in TABLE I, selected through wind speed and direction in points A, B, C (FIG. 1) at the 500 hPa level.

FIG. 3: Seasonal maximum percentage of precipitation (a) and TABLE I: Classification of convective-prone situation types. summer rainfall percentage to the annual total (b).

Besides, a climatological revision of the study area is carried out using three different parameters for the period

189 5th European Conference on Severe Storms 12 - 16 October 2009 - Landshut - GERMANY

The second part is a systematization of the possible surface level, instability and differential thickness advection. causes of local convection. This has been made using the From the statistical analysis, we can conclude that instability data from AEMET-HIRLAM 0.5º model, through various is the most important parameter regarding prediction of parameters involving different factors and levels. A convective phenomena in the area, followed by humidity statistical study of the correlation of factors and effects convergence and wind convergence, the two latter probably (detected lightning strokes), using logistical regression being caused by the confrontation of humid Mediterranean techniques, has been carried out for the warm season (April- flows with drier and cooler winds in the presence of a September) between 2002 and 2008 in the region of interest. ground configuration and a geographic position prone to The parameters considered are: humidity favour convection. convergence and wind convergence at low levels, pressure at Both methodologies show similar results in surface level, instability, geostrofic vorticity advection, modellization as well as in validation, with a skill score differential thickness advection and Q vector divergence around 85%, as shown in TABLES II and III. (Hoskins). Before regression, variables were tested for colineality and correlation, not detecting any trace of these. Consequently, all variables were included in the analysis. Regression analysis was carried out in two different ways, considering the 2002-2007 period and validating the results with data of 2008 and, alternatively, considering a TABLE II: Classification of training sample (2002-2007) and validation sample (2008). random sample of 60% of all possible data (2002-2008) and validating them with the remaining 40%, trying to avoid a possible seasonality in the results.

III. RESULTS AND CONCLUSIONS In the study period and area considered 847 days of TABLE III: Classification of training sample (60%) and validation storm were counted which, according to the classification sample (40%). established in TABLE I, the most numerous were SAT-SAT Finally, it is noteworthy that the results obtained in (South Atlantic) situations, followed by Air Mass storms and this study corroborate and complete the ones obtained in SAT-W situations (FIG. 4). previous studies (Espejo and Álvarez, 2007).

IV. REFERENCES

Álvarez, E., 2001: Climatología de descargas eléctricas. V Simposio Nacional de Predicción, INM.

Buisán, S., Espejo, F., Sanz G. et al., 2009: Characterization of the convective activity in the eastern Iberian Range, Spain. International Conference on Alpine Meteorology.

Buisán, S., Espejo, F., Sanz G. et al., 2009: Convective FIG.4: Classification of days of storm according to situation type. activity and severe weather in Teruel province, Spain. April-September 2002-2008. Description, episodes and possible future trends. I Spain- China Symposium on Geophysical & Geochemical Besides, we analyzed the mean maximum rainfall Geosystems. associated to each storm type (FIG. 5), being COL (Cutoff Low) and AFR (Northern African) the situations registering Doswell, C.A., Brooks, H.E., Maddox, R.A., 1996: Flash more rainfall. Flood Forecasting: An Ingredients-Based Methodology. Weather and Forecasting, 11 560-581.

Espejo, F., Álvarez, E., 2007: Characterisation of the evolution of convective processes in the Ebro Basin (NE Spain). 4th European Conference on Severe Storms. ECSS 2007.

Espejo, F., Sanz, R., 2001: El tornado del 28 de agosto de 1999 en Teruel. V Simposio Nacional de Predicción, INM.

Pérez, C., 2005: Técnicas estadísticas con SPSS 12. Aplicaciones al análisis de datos. Pearson Prentice Hall, 13 513-542. FIG.5: Mean maximum daily rainfall per storm type. April- September. 2002-2008. Vilar del Hoyo, L., 2006: Empleo de regresión logística para The regression models using the two methodologies la obtención de riesgo humano de incendios forestales. presented take into account 5 variables: humidity CDRom XII Congreso Nacional de Tecnologías de la convergence and wind convergence at low levels, pressure at Información Geográfica.

190 5th European Conference on Severe Storms 12 - 16 October 2009 - Landshut - GERMANY

A RIGHT FLANK SUPERCELL IN ROMAGNA; SPLITTING STORM SYSTEM CASE STUDY P.Randi1, A.Bertozzi1

(1) MeteoCenter - Epson Meteo Competence Center , via A. Einstein 8 c/o APC-CNR Faenza (RA), ITALY, [email protected]

(Dated: 15 September 2009)

I. INTRODUCTION • In mid-low troposphere (isobaric surface 700 hPa), two dry During the afternoon of July 21, 2008, most of the intrusions entered from the western quadrants, one of Alpine foothills, flatlands, and coast of the region of Romagna were orographic origins over the Po valley and evolving ESE, and subjected to particularly intense severe weather conditions a second entering over Tuscany and the northern Apennine with at least four heavy hailstorms of which at least two arc across the Ligurian Gulf with a WSW-ENE trajectory. were of the highest level with large hailstones (more than 5 These dry intrusions favoured the onset of dry-lines in the centimeters in diameter). Apennines and south of the Po River Valley that tended to Furthermore, two supercell systems developed - one lift the existing unstable and more humid air, especially on over the Ravenna area during the early afternoon and a the lower isobaric levels, over the Romagna region; making second one over the Rimini area a few hours later - giving the situation a typical environmental indicator of strong rise to one ofthemost significant severe weather outbreaks to storm episodes. have occurred over the past 5 years. The objective of this case-study was to seek the main triggering factors of such atmospheric violence through the-analysis of high- resolution, non-hydrostatic models with explicit calculation. More specifically, there was the onset of a “splitting storm” from unidirectional environmental windshear during the early afternoon over the Ravenna area, with the development of a right flank supercell system over the eastern plains and coastline (where the maximum damage was recorded). The mesoscale interactions between mid- high troposphere atlantic trough, dry katabatic flows of orographic origins (Apennine and Alpine), very hot and, above all, humid maritime currents entering the Adriatic basin (Adriatic LLJ), and very hot inert air present in the Po River Valley, gave rise, in a pre-frontal or frontal setting, to several episodes of notable violence, very often triggered in the context of an orographic dry-line, as punctually occurred under these circumstances.

II. ANALYSIS OF ENVIRONMENTAL CONDITIONS AND OPERATIONAL MODEL OUTPUT FIG. 1: Mteteocenter NMM 4km (explicit convection); 700 hpa We consider the sum of the dynamics to now in the dry intrusion high, mid, and low tropospheres with many model output, its seems clear that the situation was heading towards the • The entry of unstable cold currents from the NE, first over intense production of storm activity, particularly over the Triveneto region and then over eastern Emilia and eastern Emilia, northern Romagna and Marche, and that it Romagna; such currents eventually converged with warmer could be summarised as follows: south-western flows arriving from Tuscany and the • Entry of the polar jet-effect wind in correspondence to the Apennine ridge in proximity to coastal areas and the Alpine arc with high wind magnitude values in the high Romagna inland, and initially with weak very humid troposphere and a strong effective divergence on the currents rising from ESE of the mid- Adriatic to establish a ascending branch of the jet itself, aimed at identifying the somewhat favourable status of directional windshear intense vertical motions of a dynamic nature. Although the between the low and mid troposphere (clockwise jet core does not directly involve the region, the velocity still hodograph). reached between 60 and 70 knots. • Frontogenesis near the ground with ongoing undulations • Distinctly cyclonic winds from WSW leaving a trough triggered by the closure of the baric low over the Ligurian over France and heading ESE in mid troposphere, with a Gulf downwind of the Alpine chain and as a consequence of strong horizontal thermal gradient forming between the the airshed of cold polar maritime air across central Europe Alpine arc and regions of central Italy causing rapid cold and France. The undulation included the region of Romagna advection and the transport of positive vorticity, especially in the warm system with the cold front advancing from the over the Triveneto and Romagna regions. W over the Ligurian Gulf, Corsica, and upper Tuscany, and

191 5th European Conference on Severe Storms 12 - 16 October 2009 - Landshut - GERMANY a warm active branch N of the course of the Po River. III.RESULTS AND CONCLUSIONS In addition to the conditions already verified previously, other thermodynamic and convective parameters capable of The analysis by means of numerical hi-res models over a justifying the escalation of events that will characterise the limited area (Explicit NMM model 2 and 4 km), especially afternoon of the 21st were reviewed. by exploiting explicit resolution of the dynamics inherent in PBL and convection, has confirmed highly reliable skills in First of all, high CAPE (Convective Available Potential relation to the event. Even the mesocale magnitude and Energy) values induced by the presence of very humid air evolution of convective parameters (stability indexes), of and unstable air were found at low altitudes over the environmental windshear and low level moisture Romagna sector. Somewhat significant values, well over convergence, which are strongly influenced by the 2000 J/kg, were recorded as soon as the early afternoon over interaction between reliefs, the Adriatic Sea and the Po much of the Romagna region, particularly over the belt of River Valley according to prevailing currents in the low the inland plain, the Ravenna and Rimini coasts, and the tropospheric layers, have provided highly interesting surrounding plains as well. indications under these circumstances. During the late afternoon, the situation did not change much, and only towards sunset, with the advance of the dry intrusion even in the lower levels, did the CAPE high IV. AKNOWLEDGMENTS migrate to the confines between southern Romagna and The authors would like to thank Arpa Veneto and Arpa SMR Marche where, not by chance, other intense convective for radar imaging and Niccolò Ubalducci for photo sets systems were developing, including a second supercell. Other convective parameters or indexes of stability contributed to painting an overall picture that could lead to severe events in the area in question; in fact, even the correlated analyses of K values (Whiting index) and TT (Total Totals index) indicated highly favourable conditions. Even the Lifted index and SWEAT (Severe Weather Threat) indicated conditions favourable to severe storm activity, with the LI reaching values of -5°C and the SWEAT in the 500s. In particular, the latter index denotes environmental windshear conditions highly inclined to triggering convective systems of a supercell nature and highly indicative of vortical phenomena. Windshear values favourable to the genesis of supercell storms can also be deduced through the use of hodographic tracking according to the radiosonde planned to be carried out at 14z on the 21st via explicit model with domain at a resolution of 2 km (NNM MeteoCenter), as reported below (clockwise hodograph).

FIG. 2: Mteteocenter NMM 2km (explicit convection); derived hodograph 14z 21/7/2008

192 5h European Conference on Severe Storms 12 - 16 October 2009 - Landshut - GERMANY

GENERATION OF WINDSTORM IN THE EASTERN MOUNTAINOUS COAST OF KOREA Hyo Choi1, Doo Sun Choi1

1Dept. of Atmospheric Environmental Sciences, Kangneung-Wonju National University, Gangneung 210-702, Korea, [email protected] (Dated: 15 September 2009)

I. INTRODUCTION 3). On the other hand, at 09 LST, October 28, negative The formation of severe windstorm in the mountainous geopotential tendency line (-160m/day) lies on the study area, coast has been faced by open questions of several authors where the atmospheric depth of 500 hPa level should be (Bosart and Seimon, 1988; Eom, 1975; Kanak., et al., 2007), much shrunken and positive vorticity induces downward because its occurrence is due to lee cyclogenesis associated motion of air, resulting in high wind speed more than 9m/s. with mesoscale convection and microphysical processes. Lee According to Bernoulli theory, velocity is inversely cyclogenesis accompanying windstorm in the coast is proportional to vertical cross sectional area. Thus shrunken generally detected when downslope motion of air is atmosphere in the mountainous coast induces a strong organized into jetlike motion or the downslope motion exists channel flow, resulting in a strong wind storm. over a steep terrain. In this study, windstorm generation is From night until the next day morning, October 28, explained through numerical simulation by WRF Model-2.2. synoptic southerly wind changed into downslope westerly wind passing over mountain terrain with steep dropoff in elevation (orography) and combined with land breeze, II. NUMERICAL METHOD AND INPUT DATA becoming a strong westerly windstorm in the mountainous For the generation of windstorm around Kangnung city coast (Fig. 4). This downslope windstrom showed a (K; 37045N, 128054E) in the eastern mountainous coast of hydraulic jump in the lee side of the mountain and Froude Korea, the 3D-WRF 2.2 model was adopted for 48 hours numbers were 1.0 near the mountain top and 1.6 in the numerical simulation from Ocrober 27 until 29, 2003 and one downwind side, Kangnung coastal city. way, triple nesting was performed using a horizontal grid of

27 km covering a 91 x 91 grid square in the coarse mesh domain and a 9 km interval in the second domain and a 3 km in the third domain with the same grid square. NCEP/NCAR of FNL (1.00 x 1.00) meteorological data was used to the model. WSM 6 scheme was used for microphysical processes and YSU PBL scheme for the planetary boundary layer. Kain-Fritsch (new Eta) for cumulus parameterization, the five thermal diffusion model for land surface, RRTM long wave radiation scheme and dudhia short wave radiation schemes were also used.

III. RESULTS AND CONCLUSIONS At 21 LST, October 27, before windstorm event in the study area, a low pressure with central pressure 1005 hPa was located in northeastern China and another center of 1011 hPa in the further northeastern area (Fig. 1). Whole Korean peninsula is underneath low pressure system. Before the passage of cold front through Gangneung city at 21 LST, wind was southerly or southwesterly of less than 5m/s. However, as the low pressure system moved eastward at 09 LST, October 28, three hours before windstorm occurrenc in the both mountain and coast, especially intruding into the east sea, the low pressure was intensified under mososcale convection over the sea surface like cyclogenesis, appearing in the narrow displaced isobaric contours on a surface map and producing a strong wind in the mountainous coastal region. After cold front passage trough the study area, the wind strengthened to more than 9m/s at the city and 13m/s near Mt. Taegualyang (T) in the west of the city. The previous southerly and southwesterly changed to northwesterly and westerly wind. FIG. 1: Surface weather maps at (a) 21 LST (UTC=LST - 9 At 21 LST, October 27, before windstorm event, positive hrs), October 27, 2003 and (b) 09 LST, October 28 (two geopotential height tendency line at 500 hPa level (20m/day) hours before windstorm occurrence at Kangnung city; K). lies on the study area, where atmospheric depth of 500 hPa Red line denotes cold front. level should be expanded and negative vorticity induces upward motion of air, producing low wind speed (Figs. 2 and

193 5h European Conference on Severe Storms 12 - 16 October 2009 - Landshut - GERMANY

FIG. 2: 500hPa height change for 24 hours (m/day) called Geopotential tendency at (a) 21 LST, October 27, 2003 in the expansion (20m/day) and (b) 09 LST, October 28 in the shrunken (-160m/day). Circle denotes Kangnung city.

FIG. 4: (a) Surface wind (m/s; full bar-5m/s) and (b) vertical profile of horizontal wind in a box on a horizontal line in (a) at 21 LST, October 27. (c) and (d) at 11 LST, October 28. T and K denote Mt. Taegulyang (865m) and Kangnung city.

IV. AKNOWLEDGMENTS This work was funded by the Korea Meteorological Administration for Research and development Program under Grant “CATER 2006-2308” for 2006~2009”.

V. REFERENCES Bosart L. F., Seimon A., 1988: A case study of an unusually

intense atmopsheric gravity wave. Mon. Wea. Rev., 116 FIG. 3: (a) relative vorticity (10-5/sec) at 500 hPa at 21 LST, 1857-1886. October 27, 2003 and (b) 09 LST, October 28. White area Eom J., 1975: Analysis of the internal gravity wave denotes negative vorticity (upward motion) and red area, occurrences of April 19, 1970 in the Midwest. Mon. Wea. vice verse. Rev., 103 217-226.

Kanak J., Benko M., Simon A., Sokol A., 2007: Case study

of the 9 May 2003 windstorm in southwestern Slovakia. Atmos. Res., 83 162-175.

194