Content uploaded by Mario Marcello Miglietta
Author content
All content in this area was uploaded by Mario Marcello Miglietta on Sep 26, 2016
Content may be subject to copyright.
Characteristics and Predictability of a Supercell during 1"
HyMeX SOP1 2"
Mario Marcello Miglietta1 3"
ISAC-CNR, Lecce, Italy 4"
5"
Agostino Manzato 6"
OSMER, ARPA Friuli Venezia Giulia, Visco, Italy 7"
8"
Richard Rotunno 9"
NCAR*, Boulder, USA 10"
11"
12"
*The National Center for Atmospheric Research is sponsored by 13"
the National Science Foundation. 14"
15"
"""""""""""""""""""""""""""""""""""""""""""""""""""""""""""""
1Corresponding+ author+ address:+ Mario" Marcello" Miglietta," CNR-ISAC," strada" provinciale" Lecce-Monteroni," 73100"
Lecce,"Italy."E-mail:"m.miglietta@isac.cnr.it""
Abstract 16"
An analysis is presented here of intense convection affecting the Friuli-Venezia Giulia region 17"
(FVG, northeastern Italy) during the Intensive Observation Period 2b (IOP2b) in the first Special 18"
Observation Period (SOP1) of the HyMeX (HYdrological cycle in Mediterranean EXperiment). The 19"
present study focuses on the first of three severe-convection episodes that affected FVG on the 20"
morning of 12 September 2012. In the first episode, a supercell, which produced hail and severe 21"
damage to trees and buildings, was observed on the plain of FVG. The available observations are 22"
analyzed together with a high-resolution mesoscale model, in order to identify the relevant 23"
mechanisms for the formation and development of the cell. Six different simulations were 24"
performed starting at three different initial times, using respectively two different analysis/forecasts 25"
as initial/boundary conditions. A large spread in forecast precipitation is found among the six 26"
simulations. Only a few of the simulations were able to reproduce intense rainfall on the plain of 27"
FVG during the morning, although with significant differences in the rainfall distribution among 28"
them. One of the six simulations well reproduces the observed elongated distribution of the intense 29"
rainfall maximum; the characteristics of the cell responsible for this distribution are consistent with 30"
those expected for a supercell and its simulated evolution near the Adriatic coast agrees well with 31"
the other observations. Some instability parameters over the FVG plain and offshore (over the 32"
northern Adriatic Sea) are analyzed which showed that during this event the potential instability 33"
varies significantly over small space and time intervals and among the simulations. The best 34"
simulations have the best match to the observed potential instability calculated using the mean 35"
characteristics of the lowest 500 m layer. 36"
37"
Keywords: 38"
HyMeX; Supercells; convection; predictability; instability indices; limited area model 39"
1. Introduction 40"
Recent events have made it clear that severe convection is not rare in the Mediterranean region and 41"
may sometimes produce significant damage and casualties. For example, on 8 July 2015, an EF4 42"
tornado struck the area west of Venice, causing one death and 72 injured, while a multi-vortex EF3 43"
tornado affected southeastern Italy on 28 November 2012, causing one casualty and estimated 44"
damage of 60 M€ to the largest steel plant in Europe (Miglietta and Rotunno, 2016). The 45"
monitoring and prediction of such severe localized convective events requires a deeper 46"
understanding of the relevant mechanisms necessary for their development. The present paper is a 47"
contribution towards this goal. 48"
Routine short-term (0-36 h) numerical weather forecasts of deep convection have existed for about 49"
a decade (Weisman et al., 2008). These forecasts strongly rely on mesoscale [O(100-1000 km)] 50"
features in the initial condition to predict the location and timing of areas of convection, as well as 51"
the type of convection (supercells, squall lines, etc.). The state-of-the-art practice is to use 52"
ensembles of such forecasts generated by diverse initial conditions in order to estimate the forecast 53"
uncertainty (Schwartz et al., 2015). As the predictability limit for convective-scale elements is at 54"
most a few hours (Lilly, 1990), the precise location, timing and type of convection within the 55"
mesoscale-model-predicted area is of course not possible. The possibility of making short-term (0-56"
60 min) forecasts using cloud-scale models of severe convection is described in Stensrud et al. 57"
(2009). 58"
As the studies quoted above are in the context of the physical geography of the US, the experience 59"
gained from them cannot be simply applied to the Mediterranean basin, where the different physical 60"
geography (e.g., the presence of complex orography, coastlines, land-sea gradients of temperature 61"
and surface drag) may affect the conditions for the initiation and development of convection. 62"
The Friuli-Venezia Giulia region (FVG, northeastern Italy) has a high incidence of deep convection 63"
and thus is a natural laboratory for the analysis of such events. Together with the peak in the 64"
average yearly rainfall in the Alpine region (Frei and Schär, 1998, Isotta et al., 2014), a high 65"
frequency of thunderstorms (Feudale and Manzato, 2014), hailstorms (Manzato, 2012), tornadoes 66"
and waterspouts (Giaiotti et al., 2007) have been identified in the climatology of FVG; other high 67"
impact events, such as bow-echoes (Pucillo and Manzato, 2010) and heavy rain episodes (Davolio 68"
et al., 2016) have also been observed. The frequent occurrence of these events is due to the interplay 69"
between frontal systems, orography, and the relatively warm and shallow-water bathymetry 70"
Adriatic Sea that borders FVG. 71"
For these reasons, FVG was included as a target area both in the Mesoscale Alpine Programme 72"
(Bougeault et al., 2001) and in the Hydrological cycle in the Mediterranean Experiment (HyMeX, 73"
http://www.hymex.org), first Special Observation Period (SOP1; Ducrocq et al., 2014). In the latter 74"
campaign, three severe-convection episodes affected FVG during Intensive Observation Period 2b 75"
(IOP2b), on the morning of 12 September 2012. In the first of these episodes, a supercell formed on 76"
the plain of FVG and produced hail and severe damage to trees and buildings near the coast. This 77"
case is investigated here with a high-resolution numerical weather prediction model to explore the 78"
characteristics of the event and the sensitivity of precipitation and of the supercell features to 79"
different initial and boundary conditions. Although the physics parameterizations can also influence 80"
the simulations, here the schemes are kept fixed for simulations (although some preliminary 81"
sensitivity tests have been performed; see Section 3). 82"
The paper is organized as follows. Section 2 provides a synoptic and mesoscale overview of the 83"
event. Section 3 is focused on numerical simulations, including a comparison with the available 84"
data (surface-station measurements, satellite-derived winds and radiosounding profiles) and a 85"
discussion of the modeling results from a predictability perspective. Section 4 presents the features 86"
of the simulated storm, identifying the characteristics typical of supercells. Conclusions are 87"
summarized in Section 5. 88"
2. Synoptic and mesoscale conditions 89"
A detailed description of the synoptic conditions during IOP2b of HyMeX SOP1 is provided in 90"
Manzato et al. (2015; M15); here only a brief summary of the most relevant features is reported. In 91"
Fig. 1a, the 500-hPa geopotential height map shows a diffluent trough, associated with a cold front, 92"
moving across western France, from the north Atlantic southeastward on the morning of 12 Sept. 93"
2012. The trough and cold front reach NE Italy in the late afternoon of the same day. 94"
A closer look at the MSLP field in Fig. 1a shows two small-scale cyclones which help guide warm, 95"
moist air to FVG: an orographic lee cyclone in the Gulf of Genoa and another cyclone over the Po 96"
Valley. The latter pressure minimum is associated with a low-level cyclonic circulation that 97"
straddles the Adriatic coast of FVG. This latter circulation has intense southeasterly wind, which 98"
moves warm and moist air northward along the east side of the Adriatic and a southwesterly wind 99"
that flows downslope across the Apennines. This flow configuration produces an elongated tongue 100"
that brings moist air from the sea inland where it is available to feed convection. Further 101"
verification of this mesoscale flow pattern is found in the present case by the satellite-retrieved 102"
surface wind from ASCAT at 0839 UTC over the Adriatic Sea (Fig. 1b). 103"
Figure 2 shows the 6-h accumulated rainfall from 0600 to 1200 UTC, Sept. 12. The data are 104"
obtained from the 6-h accumulated rain estimated by the Fossalon di Grado radar using the 105"
Marshall-Palmer (1948) equation and corrected with raingauge measurements. Intense and 106"
widespread precipitation affects the region, with a peak of more than 150 mm in 6 h in the western 107"
part of the FVG Prealps, and another band of intense precipitation (estimated in about 75 mm) 108"
generated by the supercell, extending from the Veneto region eastward along the coastal regions 109"
(near Palazzolo in Fig. 2). 110"
The evolution of the atmospheric vertical structure at Udine (46.03°N, 13.18°E), which is near the 111"
center of the FVG plain, is shown in Fig. 3. Several features conducive to intense convection 112"
occurring on the morning of Sept. 12 are identified. First, a southeasterly low-level jet of moist and 113"
warm air between 400 and 1400 m amsl at 0600 UTC, which is responsible for the high values of 114"
equivalent potential temperature (θe > 330 K) in the lower troposphere. The latter feature, combined 115"
with cold-air advection at middle levels (see the slope of the black isotherms between 4 and 7 km 116"
amsl), produce conditions of potential instability, increasing the value of CAPE to approximately 117"
2000 J kg-1. At the same time, weak convective inhibition (CIN) makes the environment favorable 118"
to the triggering of convection. An increase of the vertical wind shear is also apparent, due to the 119"
intensification of the upper-level wind and the rotation of the low-level wind from westerly to 120"
easterly. Such unstable conditions are then quickly eliminated by the entrance of much colder air 121"
associated with a frontal system in the afternoon of Sept. 12. 122"
M15’s Fig. 4 shows images from the EUMETSAT (European Organisation for the Exploitation of 123"
Meteorological Satellites) experimental 2.5 min rapid scans from the High Resolution Visible of the 124"
Meteosat Second Generation, which helps identify the transit of cloud systems across the region and 125"
the triggering of convection. At around 0630 UTC, 12 Sept., M15’s Fig. 4 shows the first 126"
convective cell is triggered in the Alpine foothills, in the western part of the FVG region, where the 127"
6-h accumulated rainfall maximum of 150 mm is observed. The radar data presented here in Fig. 4 128"
shows that in the next 1.5 h, the “northern storm” (as referred to in M15) intensifies approximately 129"
in the same location. In the following hour, a new cell, the “southern storm”, develops and grows 130"
rapidly. The evolution of the two cells is clearly identified in detail in Fig. 4. 131"
The radar reflectivity maps of Fig. 4 indicate the triggering of the southern cell near the Adriatic 132"
Sea and its subsequent northward movement (between 0740 and 0820 UTC). Since the 133"
environmental wind is westerly and this cell moves to the left of the flow, there is a strong 134"
suggestion that it is a left mover of a previous supercell splitting (Weisman and Klemp, 1982). 135"
From the analysis of M15 (their Fig. 13), the northern storm is also very likely a supercell. When 136"
the two (probable) supercells get closer to each other at about 0830 UTC, the southern storm 137"
strengthens, probably because it is able to intercept the moist low-level air (Fig. 1b) that had been 138"
feeding the northern storm, so that the latter suddenly dissipates (Fig. 4). Finally, the remaining cell 139"
deviates eastward along the coast, where it exhibits supercellular features, such as the hail reported 140"
in Latisana and the intense downdraft/outflow in Palazzolo at about 0900 UTC (Fig. 4). Later on, 141"
the system moves along the coast evolving into a bow-echo pattern, as shown in M15’s Fig. 5d. 142"
Fig. 5 shows θe observed at four surface stations, whose locations are shown in Fig. 2. While Udine, 143"
which is north of the area of main convective activity, shows small variations of θe, the other three 144"
stations exhibit very sharp drops corresponding to the storm passage. Such large drops in θe suggest 145"
a strong efficiency of the storm in converting the environmental thermodynamic energy into 146"
precipitation and kinetic energy. 147"
3. Numerical simulations 148"
3.1 Simulations with WRF model 149"
The supercell described above is analyzed here by means of numerical simulations performed with 150"
the Weather Research and Forecasting (WRF) model, version 3.5.1 (see www.wrf-model.org; 151"
Skamarock et al., 2008). WRF is a state-of-the-art weather prediction system that solves the fully 152"
compressible, nonhydrostatic equations of atmospheric motion. Forty terrain-following hydrostatic-153"
pressure vertical levels are used in the present simulations with more closely spaced levels in the 154"
boundary layer. In order to analyze the detailed evolution of the meteorological parameters, model 155"
output is saved every 5 min. 156"
The model is implemented using three different one-way-nested grids, with horizontal spacings 157"
respectively of 9, 3, and 1 km, extending for 190 (in the east–west direction) x 150 (in the north–158"
south direction) grid points in the outer grid, 271 x 163 in the middle grid and 181 x 181 in the 159"
inner grid (Fig. 6). The area of interest is in the center of the inner domain and on the eastern side of 160"
the two outer domains, in order to properly represent the large-scale evolution of the trough, which 161"
propagates from the west. 162"
Since the predictability of the event is the main subject of the present paper, different 163"
initial/boundary conditions are considered to force the simulations. In particular, the European 164"
Centre for Medium-range Weather Forecasts (ECMWF) and the Global Forecasting System (GFS) 165"
analysis are used as initial conditions. The boundary conditions are updated every 3 h with the 166"
ECMWF Integrated Forecasting System (IFS) and GFS forecasts, thus the simulations are 167"
performed in an operational-like configuration. Also, different starting times have been tested to 168"
initialize the model simulations, respectively 0000 and 1200 UTC, 11 Sept. 2012, and 0000 UTC, 169"
12 Sept. 2012. Simulations are named according to the initial time, for example the run forced with 170"
GFS data starting at 11 UTC, 12 Sept., is named as GFS1112. As will be shown, GFS1112 is the 171"
simulation that best reproduces the observed evolution of the supercell and is considered the 172"
“control run” hereafter. 173"
Preliminary experiments were undertaken to identify an optimal set of parameterizations, able to 174"
better reproduce the cell evolution. Following the outcome of these experiments, the model is 175"
implemented with the following schemes: Thompson et al. (2008) microphysics; Rapid Radiative 176"
Transfer Model (RRTM) for longwave radiation based on Mlawer et al. (1997); Dudhia (1989) for 177"
shortwave radiation; the unified Noah land-surface model (Niu et al., 2011); Mellor-Yamada-Janjic 178"
planetary boundary layer (Janjic, 2001). Thus, the schemes are the same as those employed for the 179"
simulations with the WRF model in M15. Since the large-scale forcing is the same, the better result 180"
of the GFS1112 run in reproducing the supercell evolution compared to the simulation in M15 181"
indicates the importance of domain sizes (which are larger in the present study) and, mainly, of the 182"
fine horizontal resolution required for a realistic simulation of meso-γ scale features. 183"
There is no consensus on the use of convection parameterization for grid spacing slightly smaller 184"
than 10 km, as in the outer domain used here. Done et al. (2006) reported on cases of fairly intense 185"
convection with mesoscale organization, showing there is no advantage of using a cumulus 186"
parameterization even for grid spacing slightly larger than 10 km. In the present study, no cumulus 187"
scheme is used in any domain. In order to see how this choice affects the simulation, an additional 188"
run was performed, using the same configuration as GFS1112, but switching on the Kain (2004) 189"
cumulus convection scheme in the outer grid and leaving the explicit treatment of convection only 190"
in the two inner domains. 191"
Additionally, different schemes for the planetary boundary layer (PBL) have been tested: we 192"
believe that the PBL parameterization plays a key role in the present case study by modifying the 193"
characteristics of the atmosphere in the low levels, thus the instability properties and the flow 194"
dynamics. In both cases, differences of the simulations with the GFS1112 run are relatively minor 195"
compared to those emerging among experiments with different initial times and/or large-scale 196"
forcing: they show the same solution characteristics, with only slight modifications in the rainfall 197"
amount and distribution. As a consequence, hereafter the predictability analysis will focus on the 198"
sensitivity to different initial/boundary conditions without considering the role of parameterization 199"
schemes. 200"
3.2 Mesoscale and precipitation patterns 201"
In M15, both models (MOLOCH and WRF) used for the simulation of the event were able to 202"
reproduce the triggering of convection over the foothills of the Alps. Also, they simulated some 203"
mesoscale features that possibly played a key role during this phase, such as the tongue of warm air 204"
advected by an intense low-level jet. However, both models missed the exact timing, location and 205"
movement of the cells. Thus, although both models captured the mesoscale environment fairly well, 206"
they were far from an accurate simulation of the convective system. 207"
In order to better explore this point, six experiments were performed with the WRF model using 208"
different initial and boundary conditions, as discussed in Subsection 3.1. The characteristics of the 209"
low-level inflow of warm, moist air at 0600 UTC are shown in Fig. 7 for all runs. The simulations 210"
show important mesoscale differences near the mountain and over the plain are already present 211"
before the triggering of convection near the foothills. Thus, the discrepancies growing from initial 212"
small-scale differences spreading upscale as a consequence of moist convection (Zhang et al., 2002, 213"
2003) probably have a minor effect here. 214"
The ECMWF runs (Fig. 7, right) all show a similar pattern at 0600 UTC: the warm-air tongue 215"
penetrates inland, although with a different northward extent and intensity (the earlier the starting 216"
time, the cooler the air and the narrower the jet), while the cold air always remains confined near 217"
the mountains and the foothills. The differences among the GFS-forced runs are more apparent (Fig. 218"
7, left). The GFS1200 run shows a peculiar configuration, since it is characterized by a wide area of 219"
very cold, low-level air, extending from the mountains to the sea, while the warm tongue is 220"
confined to a very narrow region close to the coast. 221"
Compared with the ECMWF simulations, in the GFS1100 and GFS1112 simulations the cold air 222"
shows a similar southward extent near the foothills, but is somewhat cooler. Also the inland 223"
penetration of the warm air is much deeper, with values close to 340 K simulated even near the 224"
foothills (in particular in GFS1112), and shifted farther to the west than in the other runs. The 225"
westward deflection of the warm air in its northern tip can be probably attributed to the cyclonic 226"
circulation around the pressure meso-low in the Po valley (see Fig. 1), centered near Venice, which 227"
appears slightly deeper in GFS1112 run than in the other runs. This feature plays an important role 228"
in favoring supercell development since it prevents the warm tongue from moving eastward, forcing 229"
it to remain in the vicinity of the foothills. Between the two experiments, GFS1112 has a much 230"
deeper penetration inland, and a stronger and more extensive low-level jet than GFS1100. 231"
As a consequence of the variety in the simulated mesoscale patterns, there is a large variation in the 232"
way the flow interacts with the orography and in the simulated precipitation. To summarize the 233"
differences among the simulations, Fig. 8 shows the 6-h accumulated precipitation from 0600 to 234"
1200 UTC. Only the GFS runs and the ECM1200 run are able to reproduce intense rainfall in the 235"
FVG plain and the coastal area, although with significant differences in its distribution. Similarly, 236"
the variation of maximum updraft vertical velocity with time (not shown) is characterized by a 237"
strong variability. 238"
Two experiments shown in Fig. 8, ECM1100 and ECM1112, produce precipitation mainly in the 239"
northeastern part of the region, near the border with Slovenia, in an area where the observed 240"
precipitation is much smaller (cf. Fig. 8 with Fig. 2), while no precipitation is simulated near the 241"
coast. In these two cases, the warm tongue is advected too far north at later times, while the cold air 242"
remains confined very close to the mountains (not shown); thus the direct orographic uplift is 243"
mainly responsible for rainfall in these runs. 244"
The other experiment forced with ECMWF data (Fig. 8f) reproduces an intense rainfall peak in the 245"
area affected by the supercell, but the simulated precipitation is about 200 mm, about twice that 246"
observed (and about the sum of the rainfall maximum in the northern and southern storms). The 247"
coastal rainfall is produced –as in the GFS experiments- at the northern terminus of the low-level 248"
warm inflow, which remains quasi-stationary near the coast for several hours (not shown). The peak 249"
in vertical velocity is above 20 m s-1 for a few minutes, but the cell does not show the rotation 250"
typical of supercells. Two minor rainfall peaks are also simulated, one in the northeastern region, 251"
and another one (corresponding to the observed maximum) near the foothills at the border with 252"
Veneto region, but they are significantly underestimated and shifted northward. 253"
The simulated precipitation in the GFS1200 experiment shown in Fig. 8 is also very different from 254"
the observations. Some orographic precipitation is again shown near Slovenia, while a rainband is 255"
elongated from the west side of the region to the east, following the eastward movement of the 256"
northern tip of the warm tongue in the morning of Sept. 12. However, the timing is incorrect, the 257"
precipitation is shifted to the north (cf. Fig. 8 with Fig. 2) and the intensity is significantly 258"
underestimated. 259"
The other two GFS-forced experiments shown in Fig. 8 better simulate the observed precipitation. 260"
Both experiments, in particular GFS1112, show a persistent vertical velocity larger than 20 m s-1 261"
lasting for more than one hour. Also, the two simulations are the only ones that produce some 262"
rainfall in the northern part of the region near the Alps, in agreement with the observations. The 263"
GFS1100 run reproduces fairly well the rainfall amount associated with the supercell near the coast, 264"
although the affected region has a shorter east-west extent than that observed, due to the earlier 265"
weakening of the supercell. Finally, the GFS1112 run reproduces well the observed elongation of 266"
the intense rainfall maximum toward Slovenia, and the precipitation amount is close to the 267"
observed. The presence of the mountains in Slovenia seems to prevent a longer duration of the 268"
supercell, which lasts for about 1.5-2 hours (in agreement with the analysis in M15’s Section 5). 269"
The observed maximum in the foothills is well captured in the simulation, although separated into 270"
two distinct and weaker maxima. 271"
Apparently, the successful simulation required a deep inland penetration and a cyclonic rotation of 272"
very warm, moist air on the west side of the FVG plain, which is accomplished only in the 273"
GFS1100 and GFS1112 experiments (Fig. 7). Small-scale variations in θe are known to affect the 274"
potential instability of parcels in the layer where convection originates (Done et al., 2011). The 275"
presence of high-θe values in the low levels is required to make the atmosphere more unstable and 276"
allow for the triggering of convection near the foothills, which, as shown in Section 4, leads to the 277"
cold-air outflow that plays an important role in the later evolution of the storms. (Note that the 278"
strong gradient of θe corresponds to an area of low-level convergence between the southerly inflow 279"
and a northeasterly barrier wind from the Alps, which affects the pre-Alpine region and the northern 280"
part of the Po valley as discussed in M15 and Davolio et al., 2016 and favors convective triggering.) 281"
Figure 9 shows that intense precipitation (larger than 40 mm h-1) in the GFS1112 run is initially 282"
triggered near the rainfall maximum observed in the foothills (Fig. 2). This is an area where CIN is 283"
low (below 20 J kg-1) and bordered on its southern side by a band of high CAPE (above 1500 J kg-
284"
1), which was advected northward on the morning of Sept. 12. The simulated soundings near the 285"
foothills show that the advection of low-level moisture dramatically increases the instability of the 286"
environment in that period. For example, at the point (46.0°N, 12.75°E), the CAPE is about 350 J 287"
kg-1 and CIN is -25 J kg-1 at 0300 UTC; after 3 hours, an increase in the mixing ratio of about 2 g 288"
kg-1 at the level of the most unstable parcel (950 hPa), produces an increase in θe from 329 K to 336 289"
K and reduces the inhibition while the CAPE increases up to about 1400 J kg-1. Changes in the 290"
upper-level profiles, due to the incoming trough, appear as relatively minor in this stage. 291"
3.3 Simulated and observed vertical profile and instability indices 292"
Figure 10 shows the time evolution of the atmospheric profiles simulated at the grid point closest to 293"
Udine in the GFS1112 run. Since the time resolution of the model output shown in Fig. 10 is 1 h, 294"
the rapid changes in the meteorological fields can be detected with greater detail than in the 295"
observed time evolution (Fig. 3). 296"
Figure 10 shows that the vertical structure of the equivalent potential temperature at Udine, 297"
simulated by the model, appears consistent with the observed evolution. In particular, the GFS1112 298"
run correctly represents the arrival of cold, dry air at midlevels, the presence of low-level high-θe 299"
air, the large instability (high CAPE) in the early morning of Sept. 12, the rotation of the wind 300"
vector in the low levels along with the intensification of the wind speed in the upper levels and the 301"
transit of the cold front in the evening. However, Fig. 10 shows that the evolution in the simulation 302"
occurs a few hours earlier than in the observations. The more-frequent model output in Fig. 10 303"
shows the presence of high-θe air extending from the ground to the upper troposphere 304"
corresponding to the development of deep convection, simulated both in the morning and afternoon 305"
of Sept. 12. 306"
One of the main aspects that influence the predictability of this event is the model ability to 307"
simulate the vertical structure of the temperature, humidity and wind, which determine potential 308"
instability and type of convection (e.g. supercells, multicells, etc.). In order to compare the potential 309"
instability and wind profiles among the simulations as well as with the observed values, sounding-310"
derived and model-derived indices are calculated (Manzato, 2008), before and during the 311"
convection. Four instability indices are shown in Fig. 11 for all six runs, to represent the evolution 312"
on the morning of Sept. 12. On the right side of Fig. 11 both observed values (cross marks) derived 313"
from the Udine radiosoundings (46.03°N, 13.18 °E) and simulated values every 5 min are shown. 314"
On the left side of Fig. 11, the simulated indices are shown at a grid point located offshore over the 315"
Adriatic Sea (45.4°N, 13.0 °E). All these indices are computed with the “Tv method” described in 316"
Manzato and Morgan (2003), which uses also the “virtual correction” suggested by Doswell and 317"
Rasmussen (1994). Moreover, a centered moving average of 3 points (10 min of time interval) has 318"
been applied to smooth the fast varying indices: even so, the time evolution is very fast, with the 319"
indices over land having much sharper fluctuations than those offshore. 320"
Figure 11a shows the classical Lifted Index (LI; Galway, 1956), which uses as the initial parcel the 321"
mean air properties in the lowest 500 m. Between 0500 and 0800 UTC the potential instability is 322"
much larger offshore (LI in the range -5 to -8 °C) than over land (LI from 0 to -4 °C). In the GFS 323"
runs, the model-derived LI is closer to the observed values in Udine, in particular at 0535 UTC. 324"
However, the Lifted Index using the most unstable parcel method (MUP, not shown), in which the 325"
initial parcel corresponds to the maximum θe in the lowest 250 hPa, shows that ECMWF runs have 326"
a better estimate of LI using MUP, due to its better estimation of the maximum θe in the lowest 250 327"
hPa (Figure 11b), which is located above 500 m. In conclusion, it seems that potential instability 328"
based on the lowest 500 m is better described by GFS, while that based on the most unstable parcel 329"
(located at higher levels) is better described by ECMWF. The fact that GFS1112 better reproduces 330"
the observed dynamics and the precipitation indicates that the potential instability based on the very 331"
low levels (lowest 500 m) is the most important to be well predicted in this case. 332"
Figure 11b also shows that the sounding-derived maximum θe at 0535 UTC is higher than the 333"
highest simulated θe, while the observed maximum θe at 1105 UTC is slightly lower than the lowest 334"
simulated θe. This means that the drop of almost 20 K in air mass between 0535 and 1105 UTC is 335"
underestimated by all six models. Consistent with the LI, Fig. 11b shows the simulated maxima of θe 336"
have much higher values offshore than inland. This very strong north-south gradient of θe across the 337"
coast is very significant considering that the two locations are only 70 km apart. The feature of the 338"
simulations corresponds well with the surface observations, as the value of θe in Udine is on 339"
average 10 K lower than it is in Lignano before storm passage (Fig. 5). This observation means that 340"
the very warm, moist air remains mainly near the coast in the west part of FVG. 341"
The fast variations in the simulated Udine indices can be attributed to sudden changes in the 342"
northern extent of the low-level jet, probably associated with the movement of the convective cells 343"
along the region, which may temporarily block the southerly inflow of warm and moist air. Figure 344"
11c shows the meridional component of the mean wind in the lowest 500 m (LLWv, with positive 345"
values indicating southerly flow). We see that, while a weak southerly component (from about 0 to 346"
4 m s-1) is simulated for most of the time in Udine, for a short period the wind becomes northerly in 347"
some runs (GFS1100, GFS1200, ECM1200), due to the outflow associated with the northern storm 348"
(Fig. 4). 349"
Fig. 11c shows that LLWv in the point offshore is southerly for all time and all runs, apart from 350"
ECMWF1100, and shows a progressive intensification of its magnitude until about 9 m s-1, 351"
followed by a sudden drop (occurring between 0730 or 0830 UTC, depending on the model), as to 352"
track the passage of a low-level jet. This is in agreement with the presence of a southerly low-level 353"
jet over the Adriatic Sea, observed at 0839 UTC in Fig. 1b. The LLWv drop is less pronounced in 354"
the three GFS runs than in the remaining two ECMWF simulations, hence GFS runs are more 355"
efficient in pushing the high-θe air of the lowest 500 m from the sea inland. 356"
Lastly, Figure 11d shows that the storm-relative helicity, which is calculated considering the 357"
simulated eastward storm speed of 7 m s-1, is in good agreement with observations, in particular for 358"
GFS1112 and GFS1200 (note also that the values for the GFS1112 run are the highest in the point 359"
offshore). The parameter shows intense fluctuations inland; offshore the peak occurs a couple of 360"
hours earlier (between 0700 and 0800 UTC, depending on the model run). In contrast with most of 361"
the other indices, the peaks of this parameter are much higher inland than offshore, probably due to 362"
the much stronger wind shear brought about by larger drag over land, in particular in the presence 363"
of complex orography. Simulated values of helicity larger than 100 m2 s-2 in Udine occur in all 364"
models, denoting a larger potential for cyclonic updraft rotation inland. The increase in this index 365"
reflects the larger vorticity advection associated with the cold front and the upper-level trough 366"
approaching from the northwest. 367"
In conclusion, from this analysis we have learned that, at least in this case, a more realistic 368"
simulation of the lowest 500 m (both in terms of θe and wind structure) seems to have a strong 369"
influence on the better predictions of some forecasts with respect to others. 370"
4. Supercell features 371"
In the present Section, the characteristics of the cell generated by the merging of the northern and 372"
southern storm are analyzed and compared with the classical supercell conceptual model. Thus, the 373"
3D structure of the flow around the simulated supercell is investigated more deeply. In the 374"
following, only the GFS1112 run, which reproduces better the observed 6-h accumulated rainfall, is 375"
considered. 376"
Following Rotunno and Klemp (1985), two distinctive features are recognized as hallmarks of 377"
supercell thunderstorms: the propagation to the right of the mean tropospheric wind shear (apart 378"
from the left movers, as in Fig. 4) and a significant degree of organized rotation around the updraft 379"
that persists for tens of minutes (Doswell and Burgess, 1993; Thompson, 1998). For the former, the 380"
simulated cell movement is from west-northwest to east-southeast, approximately coincident with 381"
the rainband elongation from Veneto toward Slovenia (Fig. 2), and agrees well with the 382"
observations (cf. with M15’s Fig. 5); thus, it is rightward of the average tropospheric wind shear, 383"
which in the morning of Sept. 12 is approximately west-south-westerly (Figs. 3 and 10). M15’s Fig. 384"
13 confirms that the northern storm and the merged cell, as simulated in that paper, have the 385"
rightward propagation typical of supercells. For the latter feature, Fig. 12a shows the vertical 386"
velocity and the vertical component of vorticity at 5500 m at 0905 UTC, Sept. 12. The near 387"
superposition of the maxima of the two fields is a typical feature of supercellular systems. Also, on 388"
the northern side of the cell, a small area of anticyclonic circulation is present, which can be 389"
associated with the splitting of the original cell (generated on the Alpine foothills at about 0600 390"
UTC) induced by the downdraft, according to the conceptual model discussed in Rotunno (1981, his 391"
Fig. 3) and Rotunno and Klemp (1982). 392"
Figure 12b shows the structure of the cell at lower levels and suggests that the intense rainfall 393"
induces strong evaporation contributing to the formation of a cold pool on the rear flank, which is 394"
crucial for the baroclinic generation of vorticity along the cold-air boundary and the low-level 395"
rotation of the system (Rotunno and Klemp, 1985). Also, the inflow of moist, potentially unstable 396"
low-level air from the Adriatic Sea that feeds the updraft is necessary to continuously trigger 397"
convection above the surface gust front. Such features are consistent with the conceptual model in 398"
Markowski and Richardson (2010)’s Fig. 8.20. However, in the present case the supercell moves in 399"
synchronicity with the high-θe tongue; thus the evolution of the synoptic and mesoscale features 400"
appears to control its displacement. (In particular, its eastward movement follows the mesocyclone, 401"
while the southward movement corresponds to the intrinsic dynamics of a supercell.) While sharing 402"
a similarity with supercells forming on and moving with a dryline (Bluestein et al., 2015), in the 403"
present case there is also the low-level jet coming from the Adriatic Sea. As discussed in Feudale 404"
and Manzato (2014), this jet can be considered as occurring in a “channel” formed by the Dinaric 405"
Alps and the Apennines; the jet is influenced by the Alpine barrier on its northern side, causing 406"
local convergence and inhomogeneity in the high-θe tongue, and thereby influencing the evolution 407"
of the supercell in way that appears qualitatively different from supercells over the U.S. Great 408"
Plains. 409"
Figure 13 shows a 3D view during the mature stage of the cell. The intense updraft (of more than 30 410"
m s-1) is generated where the cold pool in the rear of the cell and the warm air inflow meet, a few 411"
km from the coast. The outflow associated with the downdraft is apparent, as well as the low-level 412"
rotation below the updraft. The high rainwater content assumes a bow-echo pattern, induced by the 413"
downward movement of the potentially cold mid-level air, which is further cooled down by the 414"
rainfall evaporation and moves underneath the low level inflow. 415"
The low-level rotation is more apparent in Fig. 14, where two specific trajectories are shown. In 416"
agreement with the Browning (1964)’s supercell model and with Klemp et al. (1982), the rear-flank 417"
updraft is created by ambient air entering the storm along the right (south) flank and then wrapped 418"
around the rear flank by the strengthening mesocyclone, while the forward-flank downdraft is 419"
wrapped around the north side of the updraft. 420"
5. Conclusions 421"
The present paper focuses on a severe-convection episode occurring in Friuli Venezia Giulia (FVG, 422"
northeastern Italy) on the morning of 12 September 2012, during the Intensive Observation Period 423"
2b (IOP2b) in the first Special Observation Period (SOP1) of the HyMeX campaign. One supercell, 424"
which produced hail and severe damage to trees and buildings, developed on the plain of FVG and 425"
was generated by the interaction between two previously existing cells. 426"
Observations are analyzed together with high-resolution WRF model simulations to identify the 427"
mechanisms responsible for the formation and development of the cell. Among six runs, starting 428"
from different large scale forcing at different times, the simulation initialized at 1200 UTC, 11 429"
September 2012 forced with GFS analysis/forecasts (GFS1112) is the one that best reproduces the 430"
observations which included a supercell thunderstorm. 431"
The mesoscale features responsible for the event are well identified. Warm, moist air, mainly 432"
confined near the coast, is advected by a low-level jet toward the Prealps (Fig. 2), producing large 433"
instability (although smaller than offshore) by increasing the local value of water-vapor mixing 434"
ratio, θe and CAPE; at the same time, an area of low-level convergence between the southerly 435"
inflow and a northeasterly barrier wind from the Alps favors convective triggering in the foothills; 436"
finally, cold air advection in the middle levels enhances potential instability. 437"
A strong sensitivity to the initial and boundary conditions is demonstrated: only a two of six 438"
simulations (GFS112 and GFS1100) were able to reproduce a persistent updraft rotation and the 439"
rightward movement typical of supercells; in two runs (ECMWF1100 and ECMWF1112), no 440"
precipitation is simulated along the coast, with rainfall generated by direct orographic uplift only 441"
near the Alps; in the other two runs (GFS1200 and ECMWF1200), the timing and the intensity of 442"
rainfall is far from that observed (the triggering of the cells occurs early in the simulations, and may 443"
be possibly affected by the model spin up). Considering that all above experiments are undertaken 444"
in an operational-like mode, this result clearly shows that an ensemble approach (even a “poor man” 445"
ensemble, as the one shown here) appears absolutely necessary to provide some indication on the 446"
risk of localized severe convective weather. From the preliminary experiments performed to 447"
identify an “optimal” setup, the sensitivity to physics appears minor compared to that due to 448"
different larger-scale forcing and initial starting times, at least for this case. 449"
Apparently, the successful simulation requires a deeper low in the Po Valley giving a farther inland 450"
penetration of deep (~500m) low-level high-θe air, especially on the west side of the FVG plain, a 451"
weak (but nonzero) convective inhibition, necessary to confine the release of convection near the 452"
foothills, where the cold-air outflow plays an important role in the later evolution of the storm and, 453"
finally, low-level cold air confined near the mountains and the foothills. The latter point is very 454"
tricky, since cold air generally remains confined mainly in the narrow Alpine valleys, which are 455"
well below the resolution of a large-scale model. Thus, the presence of a cold-air damming may 456"
easily be missed or misrepresented in the initial and boundary conditions. The significant 457"
climatological underestimation of the rainfall simulated by ECMWF forecasts in the FVG plain and 458"
coastal area during summer (Manzato et al., 2016) is probably also a consequence of such a 459"
limitation. 460"
The analysis of some instability parameters over the FVG plain and offshore (over the northern 461"
Adriatic Sea) before and during the event reveals significant small-scale variations in space and 462"
time, mainly as a consequence of the variations in the low-level θe. In particular, the sudden 463"
variations simulated in Udine are probably associated with the movement of the convective cells, 464"
which may temporary limit the tongue of warm air more to the south. 465"
Lastly, supercell features emerging from the best simulation are consistent with the classical 466"
supercell model developed in Rotunno and Klemp (1985) mainly for US Plain supercells. However, 467"
while in the latter case the pattern of θe is generally homogeneous and stationary, in the present case 468"
the synchronous movement of the high θe tongue with the cell is a distinctive feature, which 469"
appears to be controlled mainly by mesoscale features. Also, the interaction of the moist and warm 470"
low-level jet with the Alps causes local convergence and inhomogeneity in the high-θe tongue (as 471"
shown also in Figure 1b), influencing the evolution of the supercell. The generality of these results 472"
should be tested extending a similar analysis to other Mediterranean events. 473"
For future work, we plan to simulate the environment conducive to the present supercell in idealized 474"
conditions. In this way, we can systematically analyze the sensitivity of the solution to a range of 475"
values or to small perturbations added to the relevant parameters, in order to better understand the 476"
mechanisms that may have affected the triggering and development of the supercell, making 477"
apparent the differences with respect to the US supercell environment. Also, since the 478"
characteristics of the present supercell appear to have survived only for a short period (a few tens of 479"
minutes), the reason for such a short lifetime need to be analyzed and discussed, possibly 480"
considering the role of the orography. 481"
ACKNOWLEDGEMENTS 482"
This work is a contribution to the HyMeX program. Arturo Pucillo is gratefully acknowledged for 483"
Fig. 1a, Stefano Zecchetto for Fig. 1b. Figures 7, 8, 12-15 were produced with VAPOR 484"
(www.vapor.ucar.edu). Figures 3, 4 and 5 are made with the NCAR Zebra software (Corbet et al. 485"
1994). The stay of R. Rotunno in Lecce was supported by CNR short-term mobility program; the 486"
stay of M. Miglietta in Boulder was supported by NCAR. 487"
BIBLIOGRAPHY 488"
Bluestein HB, Snyder JC, Hauser JB. 2015. A multiscale overview of the El Reno, Oklahoma, 489"
tornadic supercell of 31 May 2013. Weather and Forecasting 30: 525-552. 490"
Browning KA. 1964. Airflow and precipitation trajectories within severe local storms which travel 491"
to the right of the winds. J. Atmos. Sci. 21: 634-639. 492"
Bougeault P, Binder P, Buzzi A, Dirks R, Houze R, Kuettner J, Smith RB, Steinacker R, Volkert H. 493"
2001. The MAP Special Observing Period. Bull. Am. Meteorol. Soc. 82: 433-462. 494"
Corbet JM, Mueller C, Burghart C, Gould K, Granger G. 1994. ZEB: software for integration, 495"
display and management of diverse environmental datasets. Bull. Am. Meteorol. Soc. 75: 783-792. 496"
Davolio S, Volontè A, Manzato A, Pucillo A, Cicogna A, Ferrario ME. 2016. Mechanisms 497"
producing different precipitation patterns over North-Eastern Italy: insights from HyMeX-SOP1 498"
and previous events. Q. J. R. Meteorol. Soc., doi: 10.1002/qj.2731. 499"
Done JM, Craig GC, Gray SL, Clark PA, Gray MEB. 2006. Mesoscale simulations of organised 500"
convection: Importance of convective equilibrium. Q. J. R. Meteorol. Soc. 132: 737–756. 501"
Done JM, Craig GC, Gray SL, Clark PA. 2012. Case-to-case variability of predictability of deep 502"
convection in a mesoscale model. Q. J. R. Meteorol. Soc. 138: 638–648. DOI:10.1002/qj.943. 503"
Doswell CA III, Burgess DW. 1993. Tornadoes and tornadic storms. A review of conceptual 504"
models. The Tornado: Its Structure, Dynamics, Hazards, and Prediction, Geophys. Monograph 79: 505"
161–172. 506"
Doswell CA III, Rasmussen EN. 1994. The effect of neglecting the virtual temperature correction 507"
on CAPE calculations. Weather and Forecasting 9: 625-629. 508"
Ducrocq V, Braud I, Davolio S, Ferretti R, Flamant C, Jansa A, Kalthoff N, Richard E, Taupier-509"
Letage I, Ayral P-A, Belamari S, Berne A, Borga M, Boudevillain B, Bock O, Boichard J-L, Bouin 510"
M-N, Bousquet O, Bouvier C, Chiggiato J, Cimini D, Corsmeier U, Coppola L, Cocquerez P, Defer 511"
E, Delanoë J, Di Girolamo P, Doerenbecher A, Drobinski P, Dufournet Y, Fourrié N, Gourley JJ, 512"
Labatut L, Lambert D, Le Coz J, Marzano FS, Molinié G, Montani A, Nord G, Nuret M, Ramage 513"
K, Rison W, Roussot O, Said F, Schwarzenboeck A, Testor P, Van Baelen J, Vincendon B, Aran 514"
M, Tamayo J. 2014. HyMeX-SOP1: The Field Campaign Dedicated to Heavy Precipitation and 515"
Flash Flooding in the Northwestern Mediterranean. Bull. Amer. Meteor. Soc. 95: 1083–1100. 516"
Dudhia J. 1989. Numerical study of convection observed during the Winter Monsoon Experiment 517"
using a mesoscale two-dimensional model. J. Atmos. Sci. 46: 3077–3107. 518"
Feudale L, Manzato A. 2014. Cloud–to–Ground Lightning Distribution and its Relationship with 519"
Orography and anthropogenic emissions in the Po Valley, J. Appl. Meteorol. Climatol. 53: 2651-520"
2670. 521"
Frei C, Schär C. 1998. A precipitation climatology of the Alps from high-resolution rain-gauge 522"
observations. Int. J. Climatol. 18: 873-900. 523"
Galway JG. 1956. The lifted index as a predictor of latent instability. Bull. Am. Meteorol. Soc. 37: 524"
528–529. 525"
Giaiotti DB, Giovannoni M, Pucillo A, Stel F. 2007. The climatology of tornadoes and waterspouts 526"
in Italy. Atmos. Res. 83: 534–541. 527"
Isotta FA, Frei C, Weilguni V, Perčec Tadić M, Lassègues P, Rudolf B, Pavan V, Cacciamani C, 528"
Antolini G, Ratto SM, Maraldo L, Micheletti S, Bonati V, Lussana C, Ronchi C, Panettieri E, 529"
Marigo G, Vertačnik G. 2014. The climate of daily precipitation in the Alps: Development and 530"
analysis of a high resolution grid dataset from pan-Alpine rain-gauge data. Int. J. Climatol. 34: 531"
1657–1675, doi:10.1002/joc.3794. 532"
Janjic ZI. 2001. ‘Nonsingular implementation of the Mellor–Yamada Level 2.5 Scheme in the NCEP 533"
Meso model,’ NCEP Tech. Rep. 437, 61 pp. 534"
Kain JS. 2004. The Kain–Fritsch convective parameterization: an update. J. Appl. Meteorol. 43: 535"
170–181. 536"
Klemp JB, Wilhelmson RB, Ray PS. 1981. Observed and numerically simulated structure of a 537"
mature supercell thunderstorm. J. Atmos. Sci. 38: 1558–1580. 538"
Lilly DK. 1990. Numerical prediction of thunderstorm - Has its time come? Q. J. R. Meteorol. Soc. 539"
116: 779–798. 540"
Manzato A, Morgan GM. 2003. Evaluating the sounding instability with the Lifted Parcel Theory. 541"
Atmos. Res. 67–68: 455–473. 542"
Manzato A. 2008. A Verification of Numerical Model Forecasts for Sounding-Derived Indices 543"
above Udine, NE Italy. Weather and Forecasting 23: 477-495. 544"
Manzato A. 2012. Hail in NE Italy: Climatology and bivariate analysis with the sounding--derived 545"
indices. J. Appl. Meteorol. Climatol. 51: 449-467. 546"
Manzato A, Davolio S, Miglietta MM, Pucillo A, Setvak M. 2015. 12 September 2012: A supercell 547"
outbreak in NE Italy? Atmos. Res. 153: 98-118. 548"
Manzato A, Cicogna A, Pucillo A. 2016. 6-hour maximum rain in Friuli Venezia Giulia: 549"
Climatology and ECMWF-based forecasts. Atmos. Res. 169 B: 465-484. 550"
Markowski P, Richardson Y. 2010. Mesoscale Meteorology in Midlatitudes, Wiley, 430 pp. 551"
Marshall JS, Palmer WMcK. 1948. The distribution of raindrops with size. J. Meteorol. 5: 165–166. 552"
Miglietta MM, Rotunno R. 2016. An EF3 multi-vortex tornado over the Ionian region: is it time for 553"
a dedicated warning system over Italy? Bull. Am. Meteorol. Soc. 97: 23-30. 554"
Mlawer EJ, Taubman SJ, Brown PD, Iacono MJ, Clough SA. 1997. Radiative transfer for 555"
inhomogeneous atmosphere: RRTM, a validated correlated-k model for the longwave. J. Geophys. 556"
Res. 102: 16663–16682. 557"
Niu G-Y, Yang Z-L, Mitchell KE, Chen F, Ek MB, Barlage M, Longuevergne L, Kumar A, 558"
Manning K, Niyogi D, Rosero E, Tewari M, Xia Y. 2011. The community Noah land surface model 559"
with multiparameterization options (Noah-MP): 1. Model description and evaluation with local-560"
scale measurements. J. Geophys. Res. 116: D12109. http://dx.doi.org/10.1029/2010JD015139. 561"
Pucillo A, Manzato A. 2010. ‘08/08/08: the “Olympic storm” event and its implication about severe 562"
weather’ In IV Hymex Workshop, Bologna 8–10 June 2010 (page 71 in Book of Abstracts of IV 563"
Hymex Workshop (available online at: 564"
www.hymex.org/public/workshops/4/BOOK_OF_ABSTRACT_HYMEX_2010.pdf?1407740830). 565"
Rotunno R. 1981. On the evolution of thunderstorm rotation. Mon. Weather Rev. 109: 577-586. 566"
Rotunno R, Klemp JB. 1982. The influence of the shear-induced pressure gradient on thunderstorm 567"
motion. Mon. Weather Rev. 110: 136–151. 568"
Rotunno R, Klemp JB. 1985. On the rotation and propagation of simulated supercell thunderstorms. 569"
J. Atmos. Sci. 42: 271-292. 570"
Schwartz CS, Romine GS, Sobash RA, Fossell KR, Weisman ML. 2015. NCAR’s Experimental 571"
Real-Time Convection-Allowing Ensemble Prediction System. Weather and Forecasting 30: 1645–572"
1654. 573"
Skamarock WC, Klemp JB, Dudhia J, Gill DO, Barker DM, Duda M, Huang X-Y, Wang W, 574"
Powers JG. 2008. ‘A description of the advanced research WRF Version 3,’ NCAR Technical Note 575"
NCAR/TN–475+STR, 113 pp. NCAR: Boulder, Colorado, USA. 576"
Stensrud DJ, Xue M, Wicker LJ, Kelleher KE, Foster MP, Schaefer JT, Schneider RS, Benjamin 577"
SG, Weygandt SS, Ferree JT, Tuell JP. 2009. Convective-scale warn-on forecast: A vision for 2020. 578"
Bull. Am. Meteorol. Soc. 90: 1487–1499, doi:10.1175/ 2009BAMS2795.1. 579"
Thompson RL. 1998. Eta model storm-relative winds associated with tornadic and nontornadic 580"
supercells. Weather and Forecasting 13: 125–137. 581"
Thompson G, Field PR, Rasmussen RM, Hall WD. 2008. Explicit Forecasts of winter precipitation 582"
using an improved bulk microphysics scheme. Part II: Implementation of a new snow 583"
parameterization. Mon. Weather Rev. 136: 5095–5115. 584"
Weisman ML, Klemp JB. 1982. The dependence of numerically simulated convective storms on 585"
vertical wind shear and buoyancy. Mon. Weather Rev. 110: 504-520. 586"
Weisman ML, Davis C, Wang W, Manning KW, Klemp JB. 2008. Experiences with 0–36-h 587"
explicit convective forecasts with the WRF-ARW model. Weather and Forecasting 23: 407–437. 588"
Zhang F, Snyder C, Rotunno R. 2002. Mesoscale predictability of the “surprise” snowstorm of 24–589"
25 January 2000. Mon. Weather Rev. 130: 1617–1632. 590"
Zhang F, Snyder C, Rotunno R. 2003. Effects of moist convection on mesoscale predictability. J. 591"
Atmos. Sci. 60: 1173–1185. 592"
593"
594"
595"
FIGURE CAPTIONS: 596"
Figure 1: a) 500 hPa geopotential height (violet lines) and mean sea level pressure (black lines) at 597"
0600 UTC, 12 Sept. 2012, from the ECMWF forecast (initial time 0000, 12 Sept 2012) (Courtesy of 598"
Arturo Pucillo, OSMER – ARPA FVG); b) wind data from MetOP-ASCATa scatterometer (12.5 599"
km horizontal resolution) at 0839 UTC, 12 September (Courtesy of Stefano Zecchetto, ISAC-600"
CNR). The location of the places mentioned in the text is shown in a) and the box in a) corresponds 601"
to the domain shown in b). 602"
Figure 2: FVG 6-h accumulated rainfall in the period 06-12 UTC, 12 Sept. 2012 (6-level radar data 603"
corrected with raingauge observations). 604"
Figure 3: Vertical-time series in the layer 1000 – 400 hPa (left scale in hPa) of soundings at the 605"
location of Udine sounding station (46.03°N, 13.18°E) from 1200 UTC, 11 Sept. to 0000 UTC, 13 606"
Sept., with horizontal winds (red barbs) and θe (colorbar scale). Superimposed are estimates of 607"
CAPE (red line) and CIN (blue line), LFC (Level of Free Convection) (“+” signs), and temperature 608"
(black contours). Observed soundings are reported every 6 h between 0000 UTC of Sept. 12 and 609"
0000 UTC of Sept. 13, due to the request of two additional soundings at 0600 and 1800 UTC (note 610"
that the wind at 1800 UTC is not shown because it was not recorded correctly). 611"
Figure 4: Vertical Maximum Intensity (VMI) of the reflectivity measured by the Fossalon di Grado 612"
radar at 0810, 0820, 0830, 0840, 0850 and 0900 UTC, 12 Sept., with equivalent potential 613"
temperature and 10m wind observed by surface stations 5 minutes later and CESI (Centro 614"
Elettrotecnico Sperimentale Italiano) cloud-to-ground lightning ±6 minutes around the nominal 615"
time. 616"
Figure 5: Equivalent potential temperature (θe) every 5 min observed in San Vito (cyan), Palazzolo 617"
(red), Lignano (blue) and Udine (green) between 0300 and 1500 UTC, 12 Sept. 2012. 618"
Figure 6: Model grids and topography. 619"
Figure 7: Wind vectors at 350 m height (white arrows), θe at 300 m (shaded, no data between 332 620"
and 335 K) at 0600 UTC, 12 Sept. 2012, from GFS runs initialized at a) 0000 UTC, 11 Sept., b) 621"
1200 UTC, 11 Sept., c) 0000 UTC, 12 Sept and from ECMWF runs initialized at d) 0000 UTC, 11 622"
Sept., e) 1200 UTC, 11 Sept., f) 0000 UTC, 12 Sept. 623"
Figure 8: 6 h rainfall simulated (WRF inner grid) from 0600 UTC to 1200 UTC, 12 Sept. 2012, 624"
from GFS runs initialized at a) 0000 UTC, 11 Sept., b) 1200 UTC, 11 Sept., c) 0000 UTC, 12 Sept 625"
and from ECMWF runs initialized at d) 0000 UTC, 11 Sept., e) 1200 UTC, 11 Sept., f) 0000 UTC, 626"
12 Sept. 627"
Figure 9: CAPE (green shading) and CIN (red contour = -20 J kg-1) at 0600 UTC, 12 Sept.; hourly 628"
rainfall (blue-red shading) at 0700 UTC. The orography is in grey tones. 629"
Figure 10: As Fig. 3, but for GFS1112 simulated soundings at the location of Udine sounding 630"
station (46.03°N, 13.18°E). Temperature and θe are shown every hour, while horizontal winds every 631"
6 hours. 632"
Figure 11. From top to bottom: 10-min moving average of a) Lifted Index, b) θe of the most 633"
unstable parcel, c) Low level v-wind component and d) Storm Relative helicity, offshore over the 634"
Adriatic sea (45.4°N, 13.0 °E) (left) and in the grid point closer to Udine (46.03°N, 13.18 °E) 635"
(right) between 0500 UTC and 1200 UTC of Sept. 12. Note that the 0600 (1200) UTC Udine 636"
(WMO16044) sounding has been launched at 0526 (1059) UTC and has reached 500 hPa at 0545 637"
(1116) UTC, so that the corresponding indices are plotted at 0535 (1105) UTC. 638"
Figure 12: a) Vertical velocity (shading; value between –1 and 1 m s-1 are not shown), vertical 639"
component of vorticity (c.i. = 0.005 s-1; black for negative values, white for positive, 0 not shown), 640"
and wind vectors at 5500 m, b) Maximum reflectivity (shading; values below 30 dbZ are not 641"
shown), θe (c.i. = 3 K; yellow contours) and wind vectors at 100 m, simulated by the WRF model 642"
inner grid (GFS1112 run) at 0905 UTC, Sept. 12. 643"
Figure 13: as Figure 7, but for GFS1112 run at 0915 UTC, Sept. 12. The isosurface of w = 12 m s-1 644"
(orange) and rainwater content at 1000 m height (c.i. = 1 g kg-1) are also shown. 645"
Figure 14: Two different trajectories wrapping around due to supercell rotation. The colors along 646"
the trajectories represent θe (in K). 647"
648"
649"
650"
651"
652"
653"
Figure 1: a) 500 hPa geopotential height (violet lines) and mean sea level pressure (black lines) at 654"
0600 UTC, 12 Sept. 2012, from the ECMWF forecast (initial time 0000, 12 Sept 2012) (Courtesy of 655"
Arturo Pucillo, OSMER – ARPA FVG); b) wind data from MetOP-ASCATa scatterometer (12.5 656"
km horizontal resolution) at 0839 UTC, 12 September (Courtesy of Stefano Zecchetto, ISAC-657"
CNR). The location of the places mentioned in the text is shown in a) and the box in a) corresponds 658"
to the domain shown in b). 659"
660"
Figure 2: FVG 6-h accumulated rainfall in the period 06-12 UTC, 12 Sept. 2012 (6-level radar data 661"
corrected with raingauge observations). 662"
663"
Figure 3: Vertical-time series in the layer 1000 – 400 hPa (left scale in hPa) of soundings at the 664"
location of Udine sounding station (46.03°N, 13.18°E) from 1200 UTC, 11 Sept. to 0000 UTC, 13 665"
Sept., with horizontal winds (red barbs) and θe (colorbar scale). Superimposed are estimates of 666"
CAPE (red line) and CIN (blue line), LFC (Level of Free Convection) (“+” signs), and temperature 667"
(black contours). Observed soundings are reported every 6 h between 0000 UTC of Sept. 12 and 668"
0000 UTC of Sept. 13, due to the request of two additional soundings at 0600 and 1800 UTC (note 669"
that the wind at 1800 UTC is not shown because it was not recorded correctly). 670"
671"
Figure 4: Vertical Maximum Intensity (VMI) of the reflectivity measured by the Fossalon di Grado 672"
radar at 0810, 0820, 0830, 0840, 0850 and 0900 UTC, 12 Sept., with equivalent potential 673"
temperature and 10m wind observed by surface stations 5 minutes later and CESI (Centro 674"
Elettrotecnico Sperimentale Italiano) cloud-to-ground lightning ±6 minutes around the nominal 675"
time. 676"
677"
Figure 5: Equivalent potential temperature (θe) every 5 min observed in San Vito (cyan), Palazzolo 678"
(red), Lignano (blue) and Udine (green) between 0300 and 1500 UTC, 12 Sept. 2012. 679"
680"
Figure 6: Model grids and topography. 681"
682"
683"
Figure 7: Wind vectors at 350 m height (white arrows), θe at 300 m (shaded, no data between 332 684"
and 335 K) at 0600 UTC, 12 Sept. 2012, from GFS runs initialized at a) 0000 UTC, 11 Sept., b) 685"
1200 UTC, 11 Sept., c) 0000 UTC, 12 Sept and from ECMWF runs initialized at d) 0000 UTC, 11 686"
Sept., e) 1200 UTC, 11 Sept., f) 0000 UTC, 12 Sept. 687"
688"
Figure 8: 6 h rainfall simulated (WRF inner grid) from 0600 UTC to 1200 UTC, 12 Sept. 2012, 689"
from GFS runs initialized at a) 0000 UTC, 11 Sept., b) 1200 UTC, 11 Sept., c) 0000 UTC, 12 Sept 690"
and from ECMWF runs initialized at d) 0000 UTC, 11 Sept., e) 1200 UTC, 11 Sept., f) 0000 UTC, 691"
12 Sept. 692"
693"
Figure 9: CAPE (green shading) and CIN (red contour = -20 J kg-1) at 0600 UTC, 12 Sept.; hourly 694"
rainfall (blue-red shading) at 0700 UTC. The orography is in grey tones. 695"
696"
Figure 10: As Fig. 3, but for GFS1112 simulated soundings at the location of Udine sounding 697"
station (46.03°N, 13.18°E). Temperature and θe are shown every hour, while horizontal winds every 698"
6 hours. 699"
700"
701"
Figure 11. From top to bottom: 10-min moving average of a) Lifted Index, b) θe of the most 702"
unstable parcel, c) Low level v-wind component and d) Storm Relative helicity, offshore over the 703"
Adriatic sea (45.4°N, 13.0 °E) (left) and in the grid point closer to Udine (46.03°N, 13.18 °E) 704"
(right) between 0500 UTC and 1200 UTC of Sept. 12. Note that the 0600 (1200) UTC Udine 705"
(WMO16044) sounding has been launched at 0526 (1059) UTC and has reached 500 hPa at 0545 706"
(1116) UTC, so that the corresponding indices are plotted at 0535 (1105) UTC. 707"
708"
Figure 12: a) Vertical velocity (shading; value between –1 and 1 m s-1 are not shown), vertical 709"
component of vorticity (c.i. = 0.005 s-1; black for negative values, white for positive, 0 not shown), 710"
and wind vectors at 5500 m, b) Maximum reflectivity (shading; values below 30 dbZ are not 711"
shown), θe (c.i. = 3 K; yellow contours) and wind vectors at 100 m, simulated by the WRF model 712"
inner grid (GFS1112 run) at 0905 UTC, Sept. 12. 713"
714"
Figure 13: as Figure 7, but for GFS1112 run at 0915 UTC, Sept. 12. The isosurface of w = 12 m s-1 715"
(orange) and rainwater content at 1000 m height (c.i. = 1 g kg-1) are also shown. 716"
717"
Figure 14: Two different trajectories wrapping around due to supercell rotation. The colors along 718"
the trajectories represent θe (in K). 719"