Вы находитесь на странице: 1из 28

1 JUNE 2018 TASZAREK ET AL.

4281

Climatological Aspects of Convective Parameters over Europe:


A Comparison of ERA-Interim and Sounding Data

MATEUSZ TASZAREK
Department of Climatology, Institute of Physical Geography and Environmental Planning, Adam Mickiewicz
University, Pozna n, and Skywarn Poland, Warsaw, Poland

HAROLD E. BROOKS
NOAA/National Severe Storms Laboratory, and School of Meteorology, University of Oklahoma, Norman, Oklahoma

BARTOSZ CZERNECKI
Department of Climatology, Institute of Physical Geography and Environmental Planning,
Adam Mickiewicz University, Pozna n, Poland

PIOTR SZUSTER
Department of Computer Science, Cracow University of Technology, Krakow, and Skywarn Poland, Warsaw, Poland

KRZYSZTOF FORTUNIAK
Department of Meteorology and Climatology, Faculty of Geographical Sciences, University of Łód
z, Łód
z, Poland

(Manuscript received 5 September 2017, in final form 20 January 2018)

ABSTRACT

We compare over 1 million sounding measurements with ERA-Interim reanalysis for the 38-yr period
from 1979 to 2016. The large dataset allows us to provide an improved insight into the spatial and tem-
poral distributions of the prerequisites of deep moist convection across Europe. In addition, ERA-
Interim is also evaluated. ERA-Interim estimates parameters describing boundary layer moisture and
midtropospheric lapse rates well, with correlation coefficients of 0.94. Mixed-layer CAPE is, on average,
underestimated by the reanalysis while the most unstable CAPE is overestimated. Vertical shear pa-
rameters in the reanalysis are better correlated with observations than CAPE, but are underestimated by
approximately 1–2 m s21. The underestimation decreases as the depth of the shear layer increases.
Compared to radiosonde observations, instability in ERA-Interim is overestimated in southern Europe
and underestimated over eastern Europe. High values of instability are observed from May to September,
out of phase with the climatological pattern of wind shear, which peaks in winter. From September to
April, favorable conditions for thunderstorms occur mainly over southern and western Europe with the
peak location and higher frequency shifting to central and eastern Europe from May to August. For
southern Europe, the annual cycle peaks in September with high values of inhibition suppressing thun-
derstorm activity in July and August. The area with the highest annual number of days with environ-
mental conditions favorable for thunderstorms extends from Italy and Austria eastward through the
Carpathians and Balkans.

1. Introduction
Denotes content that is immediately available upon publication
as open access. Around 9000 severe thunderstorm incidents causing
100 fatalities and 500 injuries are reported each year in
Corresponding author: Mateusz Taszarek, mateusz.taszarek@ Europe according to the European Severe Weather
amu.edu.pl. Database (ESWD; Dotzek et al. 2009). Unfortunately,

DOI: 10.1175/JCLI-D-17-0596.1
Ó 2018 American Meteorological Society. For information regarding reuse of this content and general copyright information, consult the AMS Copyright
Policy (www.ametsoc.org/PUBSReuseLicenses).
4282 JOURNAL OF CLIMATE VOLUME 31

TABLE 1. Number of 1200 UTC soundings used in the study given months and domains for years 1979–2016.

Western Central Europe Eastern Southern


Domain Europe and Balkans Europe Europe Other Total
Stations 20 24 23 20 32 119
Jan 15 249 17 865 13 747 13 004 22 292 82 157
Feb 13 975 16 478 13 035 12 049 20 805 76 342
Mar 15 449 18 060 14 553 13 319 23 177 84 558
Apr 15 536 18 124 15 262 13 146 24 055 86 123
May 16 207 18 561 16 172 13 603 25 332 89 875
Jun 15 587 17 953 15 452 12 968 24 895 86 855
Jul 16 124 18 223 15 780 12 978 25 282 88 387
Aug 16 589 18 597 16 074 13 078 25 848 90 186
Sep 15 599 17 812 15 283 12 494 24 340 85 528
Oct 16 048 18 346 15 393 12 995 24 812 87 594
Nov 15 494 17 870 14 693 12 819 23 422 84 298
Dec 15 778 18 318 14 935 13 274 23 351 85 656
Total 187 635 216 207 180 379 155 727 287 611 1 027 559

geographical bias toward densely populated areas and ingredients; Doswell et al. 1996) that reflect environ-
errors in the databases make it difficult to determine the mental conditions favorable for thunderstorms. The
true coverage of severe thunderstorms (Groenemeijer relationship between these parameters and thunder-
et al. 2017). Lightning is detected more objectively (Betz storm occurrence allows them to be applied as a proxy
et al. 2009; Pohjola and Mäkelä 2013; Anderson and for the probability of different weather events occur-
Klugmann 2014; Poelman et al. 2016), but the networks ring (Allen and Karoly 2014). Understanding the cli-
still suffer from spatial inhomogeneities and short re- matological aspects of convective parameters provides
cord lengths. To overcome these issues, many re- an estimate of where and when the corresponding events
searchers have addressed the idea of using covariates are the most likely to occur, subject to the quality of the
in the form of convective parameters (thunderstorm relationship between the covariates and the events.

FIG. 1. ERA-Interim model domain (individual grid points are represented by very small
gray pluses), ERA-Interim orography in meters above sea level (shaded color scale), location
of sounding stations (circles), and subdomains (black polygons). Filled red circles indicate
stations for which individual scatterplots will be shown later.
1 JUNE 2018 TASZAREK ET AL. 4283

TABLE 2. Description of model grid for ERA-Interim. dataset, Brooks et al. (2003) found that environments
with high instability along with high vertical wind shear
Resolution 0.758 3 0.758
Latitude extension 31.58–758N (59 grid points) were associated with thunderstorms producing severe
Longitude extension 278W–53.258E (108 grid points) weather. The notion that the probability of convective
Vertical levels 1000–50 hPa 1 surface data (30 levels) hazards is a function of thermodynamic instability and
Total grid points 6372 deep layer shear (DLS) was also confirmed in many
Timeframe 1979–2016 (1200 UTC data)
other studies (Rasmussen and Blanchard 1998; Craven
and Brooks 2004; Brooks 2009, 2013; Allen et al. 2011;
Numerous studies using proximity soundings (radio- Allen and Karoly 2014; Púcik et al. 2015; Taszarek et al.
sonde or numerical model profiles taken close to par- 2017). The height of the lifting condensation level
ticular phenomena in space and time) assessed which (LCL), along with low-level shear (LLS) and storm-
parameters can be regarded as valuable thunderstorm relative helicity (SRH; Davies and Johns 1993), were
predictors (covariates). The occurrence of thunder- found to be good predictors for tornadic supercells and
storms is strongly dependent on convective available for estimating tornado intensity (e.g., Thompson et al.
potential energy (CAPE), with values of roughly 100– 2003; Groenemeijer and van Delden 2007; Grams
200 J kg21 discriminating between lightning and non- et al. 2012; Taszarek and Kolendowicz 2013). In ad-
lightning events (Craven and Brooks 2004; Kaltenböck dition, Grünwald and Brooks (2011) indicated that tor-
et al. 2009; Kolendowicz et al. 2017; Taszarek et al. nadoes in Europe usually form with lower LCL and CAPE
2017). Basing on European records, Westermayer et al. than those occurring in the United States.
(2017) suggested that the probability for lightning is The climatological distribution of convective pa-
highest when CAPE exceeds 400 J kg21 and convective rameters may not be useful in making a forecast on a
inhibition (CIN) is no lower than 250 J kg21. They particular day, but it can help in understanding the
highlighted that relative humidity in the low to mid- baseline probability of a particular event at different
troposphere has a major influence on storm occur- locations (Brooks et al. 2007). The annual cycles of
rence with low relative humidity strongly suppressing convective parameters at a variety of locations in Eu-
thunderstorms. rope indicate that conditions supportive of severe
The presence of sufficient instability ensures that thunderstorms are the most frequent during spring and
convective updrafts, once initiated, can become strong. summer (Brooks et al. 2007). The highest mean CAPE
Weisman and Klemp (1982) noted that vertical wind is observed in the southern and southeastern parts of
shear promotes organization and longevity of these Europe during summer, and the western part during
updrafts. The establishment of dynamically forced winter (Riemann-Campe et al. 2009). Brooks et al.
vertical pressure gradients can significantly enhance (2003) used data from a global reanalysis dataset to
the strength of the updraft, especially in supercell develop covariate discriminants that identify an in-
thunderstorms. Using soundings from a global reanalysis creased probability of an environment to produce

TABLE 3. Parameters used in the study.

Parameter Abbreviation Units


0–500-m AGL mixed layer convective available potential energy ML CAPE J kg21
Surface-based convective available potential energy SB CAPE J kg21
Most-unstable convective available potential energy MU CAPE J kg21
0–500-m AGL mixed layer convective inhibition ML CIN J kg21
Theoretical maximum parcel updraft speed (square root of 2 3 ML CAPE) ML WMAX m s21
Theoretical maximum parcel updraft speed (square root of 2 3 MU CAPE) MU WMAX m s21
Theoretical maximum speed to overcome CIN (square root of 2 3 ML CIN) ML WINIT m s21
0–500-m AGL mixed layer lifted condensation level ML LCL m AGL
0–500-m AGL mixed layer mixing ratio MIXR g kg21
Precipitable water PW mm
800–500-hPa temperature lapse rate LR85 K km21
0–6-km AGL bulk shear (deep layer shear) DLS m s21
0–3-km AGL bulk shear (midlevel shear) MLS m s21
0–1-km AGL bulk shear (low-level shear) LLS m s21
0–3-km AGL storm relative helicity SRH03 m2 s22
DLS 3 ML WMAX ML WMAXSHEAR m2 s22
4284 JOURNAL OF CLIMATE VOLUME 31

TABLE 4. Mean errors and mean percentage errors from ERA-Interim for all analyzed parameters given various domains.

Mean error (mean percentage error)


Central Europe
Western Europe and Balkans Eastern Europe Southern Europe All
21
MIXR (g kg ) 0.17 (2.9) 0.14 (2.5) 0.11 (2.3) 0.24 (3.4) 0.15 (2.7)
LR85 (K km21) 0.13 (2.2) 0.14 (2.3) 0.03 (0.5) 0.12 (1.8) 0.11 (1.9)
MU CAPE (J kg21) 27.28 (19.5) 77.06 (35.2) 230.13 (214.1) 239.11 (46.9) 55.22 (24.6)
SB CAPE (J kg21) 30.02 (21.4) 85.95 (40.6) 23.89 (22.1) 259.99 (53.5) 69.73 (33.1)
ML CAPE (J kg21) 2.12 (5.5) 24.05 (25.4) 232.03 (236.9) 22.89 (22.5) 29.28 (213.6)
DLS (m s21) 21.28 (27.8) 21.12 (27.2) 21.30 (28.5) 20.81 (25.5) 21.09 (27.0)
MLS (m s21) 21.59 (215.7) 21.40 (214.4) 21.63 (216.6) 21.37 (214.3) 21.46 (215.0)
LLS (m s21) 21.83 (228.4) 21.51 (225.9) 21.60 (224.9) 21.69 (233.7) 21.61 (226.9)
SRH03 (m2 s22) 214.30 (211.9) 213.79 (212.1) 215.15 (213.2) 214.36 (212.9) 213.81 (212.1)
ML LCL (m AGL) 266.28 (27.2) 273.55 (26.4) 280.55 (26.8) 2186.57 (212.3) 291.98 (27.9)
PW (mm) 0.62 (4.0) 0.44 (2.8) 0.69 (4.7) 0.17 (0.9) 0.47 (3.1)
ML CIN (J kg21) 4.18 (38.6) 5.36 (33.9) 2.81 (24.0) 11.08 (19.1) 5.27 (25.7)
ML WMAXSHEAR (m2 s22) 26.21 (29.0) 211.67 (215.5) 222.59 (232.5) 28.76 (28.8) 210.41 (215.0)

severe thunderstorms and tornadoes. They estimated The majority of the climatologies discussed above
the frequency of such environments for the whole were based on reanalyses. Reanalysis products are
world based on relationships developed from data in generated by the assimilation of observational data
the United States. For Europe, they concluded that the (e.g., surface observations, satellite information, and
southern part has the greatest frequency of significant radiosondes) over a given period of time. The goal of
severe thunderstorm environments. The preferred areas reanalysis is to provide a climatological snapshot of
for severe thunderstorms in the ERA-40 reanalysis conditions that are as close to reality as possible
(Uppala et al. 2005) were also found along a zonal belt (Thorne and Vose 2010). Data assimilation entails the
over southern and central Europe by Romero et al. (2007). incorporation of observations into a background field
In recent years, a growing number of studies have to produce an initial condition (Mooney et al. 2011).
analyzed long-term trends in convective parameters to Given the inherent uncertainties in the forecast
define how changing climate affects the occurrence of model, input data, and data assimilation, it is impor-
severe thunderstorms. An increase of low-level mois- tant to assess the quality of these reanalyses (Hodges
ture over most parts of Europe from 1979 to 2012 was et al. 2011). This was done for basic atmospheric pa-
seen in the reanalysis (Pistotnik et al. 2016). Steep rameters (e.g., temperature and dewpoint) in the nu-
vertical temperature gradients became less frequent in merous studies (e.g., Tian et al. 2010; Mooney et al.
northwestern Europe and more frequent in south- 2011; Szczypta et al. 2011; Bao and Zhang 2013; de
eastern Europe where CAPE increased as well. Future Leeuw et al. 2015; Guo et al. 2016; Duruisseau et al.
changes of severe thunderstorm environments in Eu- 2017), but only little attention has been paid to con-
rope based on climate model simulations have been vective parameters (e.g., thermodynamic instability).
addressed by Marsh et al. (2009). A small increase in Although Gensini et al. (2014) performed such an analysis
favorable severe environments was observed for most for United States and Allen and Karoly (2014) for Aus-
locations resulting from an increase in the joint oc- tralia, no efforts were made for Europe. Because convec-
currence of high CAPE and high DLS situations. Púcik tive parameters include information from basic quantities,
et al. (2017) found that the simultaneous occurrence of such as temperature, moisture, and winds, at a variety of
latent instability, strong DLS, and model precipitation levels, they are a good test of the quality of a reanalysis.
is simulated to increase by up to 100% in representa- In this paper, we compare over 1 million sounding
tive concentration pathway 8.5 (RCP8.5)1 and by 30%– measurements with the ERA-Interim reanalysis for the
50% in the RCP4.5 scenario by the end of the century 38-yr period from 1979 to 2016. The main focus is on
in central and eastern Europe. An increase in the ingredients (predictors) supporting development of
number of severe thunderstorm environments in the thunderstorms and influencing their intensity. We look
RCP8.5 scenario was also noted by Viceto et al. (2017) at the spatial and temporal distribution of these pa-
for the Iberian Peninsula. rameters for both sounding and reanalysis data in order
to gain a better understanding of the evolution of the
atmosphere supporting thunderstorms. In addition,
1
For further details see van Vuuren et al. (2011). ERA-Interim is evaluated.
1 JUNE 2018 TASZAREK ET AL. 4285

FIG. 2. Comparison of sounding observations and ERA-Interim proximity soundings (1 027 559 cases) for (top
left)–(bottom right) selected parameters. The gray line denotes a one-to-one ratio. The red line denotes the best fit
line. Value in the top-left corner denotes correlation.

2. Dataset and methodology For the years 1979–2016 all available measurements
for 1200 UTC (1100 LT in western Europe and 1500 LT
a. Radiosonde data
in far eastern Europe) were downloaded from 132 sta-
The radiosonde measurements were derived from tions over Europe (;1.3 million soundings). We focused
the sounding database of the University of Wyoming. only on 1200 UTC because this time best represents the
4286 JOURNAL OF CLIMATE VOLUME 31

FIG. 3. Box-and-whisker plots representing monthly variability of (left) MIXR and (right) LR85 for sounding
observations (blue) and ERA-Interim proximity soundings (red): (top)–(bottom) western Europe, central Europe
and Balkans, eastern Europe, and southern Europe. The median is denoted as a horizontal line inside the box, the
edges of the box represent the 25th and 75th percentiles, and whiskers represent the 10th and 90th percentiles.

typical preconvective storm environment, as indicated by or severe wind gusts is also estimated to peak in Europe
peak cloud-to-ground lightning activity between 1400 and between 1500 and 1800 UTC (Groenemeijer and Kühne
1600 UTC (Antonescu and Burcea 2010; Wapler 2013; 2014; Taszarek and Brooks 2015; Punge and Kunz 2016,
Virts et al. 2013; Mäkelä et al. 2014; Taszarek et al. 2015; Celi
nski-Mysław and Palarz 2017).
Poelman et al. 2016). The highest frequency of severe After quality control (discussed later), the final
thunderstorm phenomena such as tornadoes, large hail, number of stations dropped to 119 with over 1 million
1 JUNE 2018 TASZAREK ET AL. 4287

FIG. 4. (top) Mean annual values of MIXR for ERA-Interim and (bottom) sounding locations for (left) the cold season
(October–March) and (right) the warm season (April–September). A delicate smoothing was applied for better clarity.

soundings evenly distributed throughout the year extends from 31.58 to 758N (59 grid points) and from
(Table 1). For statistical purposes, we clustered the 278W to 53.258E (108 grid points), giving 6372 grid
stations into four subdomains representing different points (Fig. 1). For each sounding site, we picked the
regions of Europe, each with distinctive climate con- nearest (by geographical distance) grid point from the
ditions. Those regions are western Europe (WE), reanalysis, and produced a pseudo-sounding from the re-
which is under the influence of the Atlantic Ocean and analysis to compare with the observational sounding. A
humid oceanic climate, central Europe and the Balkans similar method was used previously by Thompson et al.
(CE&B), located in the transitional zone between an (2003), Allen et al. (2011), and Gensini et al. (2014).
oceanic and continental climate, eastern Europe (EE),
c. Convective parameters
representing a continental climate, and southern Eu-
rope (SE), consisting of stations under the influence of Convective parameters for both sounding and re-
the Mediterranean Sea (Fig. 1). analysis data were processed using a sounding analysis
program developed by Szuster (2016). For each profile,
b. Reanalysis data
pressure, height, temperature, dewpoint, and U and V
The European Centre for Medium-Range Weather winds were interpolated in the vertical and the param-
Forecasts (ECMWF) interim reanalysis (ERA-Interim; eters that have shown value in predicting thunderstorms
Dee et al. 2011) was used for the period 1979–2016. The and their intensities were calculated (Table 3). For
dataset used has 0.758 horizontal grid spacing with 29 mixed-layer (ML) calculations, we used the layer from
pressure levels from 1000 to 50 hPa. An additional 30th 0 to 500 m AGL. In addition, a virtual temperature
level is the surface with temperature and dewpoint for correction was included (Doswell and Rasmussen 1994).
2 m above ground level (AGL) and U and V wind vec- To compute SRH and estimate storm motion, we ap-
tors for 10 m AGL (Table 2). The research domain plied the ID method (Bunkers et al. 2000).
4288 JOURNAL OF CLIMATE VOLUME 31

FIG. 5. As in Fig. 4, but for LR85.

We defined cold (October–March) and warm (April– were checked, and we discarded soundings with obvi-
September) seasons. This division was based on light- ous errors or unrealistic values, which we took as mid-
ning activity over Europe, which is the highest from tropospheric lapse rates . 9 K km21, low-tropospheric lapse
April to September (Anderson and Klugmann 2014). rates . 11 K km21, DLS . 70 m s21, MLS . 45 m s21,
With our interest in convection, this seemed an appro- LLS . 35 m s21, MIXR . 20 g kg21, PW . 65 mm, MU
priate choice. For the annual cycle of parameters asso- CAPE . 8000 J kg21, and ML CAPE . 6000 J kg21.
ciated with CAPE or CIN we included only nonzero Since one of our motivations was to investigate the
CAPE cases. This approach was also applied in the pre- annual cycle, we excluded all stations with fewer than
vious studies (Rasmussen and Blanchard 1998; Brooks 100 soundings measurements available in any month.
et al. 2003; Craven and Brooks 2004; Brooks 2009; We also excluded stations if the difference between
Taszarek et al. 2017) and was intended to focus mainly on the month with the fewest and most soundings was
unstable environments. more than 3% of available data for the station to avoid
biases toward particular months when calculating
d. Quality control assumptions
climatological statistics. The number of soundings
Given the initial database of 1.3 million soundings, a per station that passed quality control assumptions is
considerable number contained errors of various kinds, listed in the appendix.
including incomplete profiles and invalid or unrealistic
e. Limitations of the study
values. To deal with these, as a first step, we excluded
levels with incomplete measurements (e.g., temperature An important limitation to this study is the changing
was available but dewpoints were missing). Then, all quantity and quality of radiosonde measurements over
soundings with measurements not reaching 6 km AGL time (especially humidity), which impact derived vari-
or containing fewer than 10 pressure levels were dis- ables such as CAPE and CIN (Guichard et al. 2000;
carded. Next, temperature and wind vertical gradients Cady-Pereira et al. 2008), as well as the availability of
1 JUNE 2018 TASZAREK ET AL. 4289

FIG. 6. As in Fig. 3, but for ML WMAX and MU WMAX. Values in parentheses at top left (right) denote ML
CAPE (MU CAPE). Only cases with nonzero ML WMAX and MU WMAX are used.

measurements, which in our database ranges from 4000 sharp boundaries such as mountains or coastal areas
to 10 000 soundings per site. Although we are aware of may be not well represented in the model data. In some
these issues, we believe that quality control assumptions cases, it is possible that the methodology of choosing
and the large sample size should limit their negative influ- the closest grid point to the station location may not
ence when focusing on large-scale climatological patterns. be representative of the observations. Variations in the
Another limitation is related to the pseudosoundings depiction of the boundary layer (Brooks et al. 2003;
generated from ERA-Interim. Given the resolution of Thompson et al. 2003; Allen et al. 2011) and limited
the reanalysis (Fig. 1, Table 2), soundings located over vertical resolution of the reanalysis (30 levels) compared
4290 JOURNAL OF CLIMATE VOLUME 31

FIG. 7. As in Fig. 4, but for ML CAPE.

to an average of 47 levels per observed sounding may we are aware that observations also contain errors).
also result in differences between model data and ob- The mean errors from ERA-Interim are within
servations. On the other hand, the reanalysis can be 0.15 g kg21 for MIXR, 0.11 K km21 for LR85, and
more representative than station data with fewer verti- 0.47 mm for PW, all close to the ranges for radiosonde
cal levels on average, as is often seen in EE and SE accuracy (Table 4). These parameters are also quite
stations (see the appendix). well correlated with correlation coefficients greater
It is also possible that the model formulation may in- than 0.9 (Fig. 2). Larger differences are observed for
troduce problems. As an example, given the horizontal CAPE parameters. Among different parcel types,
resolution of ERA-Interim, a convective parameterization ERA-Interim overestimates MU CAPE on average by
is necessary. This may lead to errors in the vertical 24.6%, and SB CAPE by 33.1% (Table 4). It is notable
profile of temperature and moisture as a result of the that the highest overestimation is observed over
approximations associated with the parameterization, SE (46.9% and 53.5%, respectively), while an un-
and subsequent errors in CAPE. Similarly, quantities derestimation (214.1% and 22.1%, respectively) is
associated with low levels may be impacted by the pres- seen in EE. Conversely, ML CAPE is underestimated
ence of boundary layer parameterization schemes. It is in ERA-Interim on average by 13.6%. The highest
beyond the scope of this paper to diagnose the source of ratio of underestimation is observed again over EE and
the errors, but it provides a caveat on the results. exceeds 36%. Scatterplots show a large spread, most
likely related to the high spatial variability of CAPE.
CAPE can change significantly from a few hundred to
3. Results even few thousand joules per kilogram over the dis-
tance of several kilometers, so that a ‘‘point to point’’
a. Correlation and mean errors
sounding and reanalysis comparison can show large
We will use the sounding observations as a baseline errors as a result of the gradients. Although the re-
for ‘‘truth’’ to compare the reanalysis data (although analysis does not provide exactly the same values in the
1 JUNE 2018 TASZAREK ET AL. 4291

FIG. 8. As in Fig. 3, but for DLS and LLS.

same locations as the observations, the correlation (Table 4). The highest ratio of underestimation is with
coefficients for MU and SB CAPE are approximately LLS (27%), decreasing with increasing depth (15% for
0.7 (Fig. 2). A lower correlation (0.55) is observed with MLS and 7% for DLS). Scatterplots support the im-
ML CAPE, which indicates that the boundary may not proved estimation with height, with the correlation co-
be depicted well in ERA-Interim, especially over EE. efficient for LLS, MLS, and DLS being 0.74, 0.83, and
The boundary layer parameterization is a likely can- 0.92 (Fig. 2). SE shows the largest underestimation of
didate for such issues. LLS (33.7%) while it has the smallest underestimation
A comparison of shear values indicates that ERA- of DLS (5.5%). Again, the boundary layer is a likely cul-
Interim underestimates them by approximately 1–2 m s21 prit. SRH is, on average, underestimated in ERA-Interim
4292 JOURNAL OF CLIMATE VOLUME 31

FIG. 9. As in Fig. 4, but for DLS.

by 12% and no significant regional differences are seen. (19.1%). The LCL computed with the use of ML
The SRH in ERA-Interim is in better agreement with parcel is underestimated by ERA-Interim on average
observations when values are higher than 200–250 m2 s22 by 8% (around 90 m). The largest errors (12.3%) are
(Fig. 2). observed over SE while the lowest are over CE&B
ML WMAXSHEAR (a composite parameter of in- (6.4%). The correlation coefficient between re-
stability and shear; Table 3), which is regarded as a good analysis and observation is 0.85 (Fig. 2).
severe thunderstorm predictor, is also underestimated in
b. Low-level moisture and lapse rates
the reanalysis data on average by 15%. This result is
heavily dependent on the choice of parcel type. We use Boundary layer moisture and temperature lapse rates
ML version for consistency with previous studies are often regarded as basic ingredients for thermody-
(Brooks et al. 2007; Marsh et al. 2009; Brooks 2013; namic instability. MIXR usually peaks during summer
Allen and Karoly 2014; Taszarek et al. 2017). Regional under strong diurnal heating and evapotranspiration.
analysis of this parameter indicates that it is under- The peak is earliest in EE (July) and latest in SE (August)
estimated most in EE (32.5%) and CE&B (15.5%) in (Fig. 3). During the cold season, the lowest values are
ERA-Interim. The correlation coefficient of 0.6 shows a observed over EE while the highest are over SE and WE.
slightly better relationship than ML CAPE. High values are observed all year in the Mediterranean
ML CIN was overestimated on average by 25.7%, basin (Fig. 4). Areas with elevated terrain (Spain, Turkey,
which means that convective inhibition was weaker in and the Alps) are characterized by lower MIXR on av-
ERA-Interim than in reality. The use of a limited erage. Results regarding the spatial distribution and an-
number of pressure levels in the reanalysis influences nual cycle of PW (not shown) are similar.
the computation of CIN and leads to weaker convec- LR85 shows a less pronounced annual cycle with peak
tive inhibition in general. The largest differences were values in the spring in all regions except the SE, which
observed over WE (38.6%) with the smallest over SE has high lapse rates through August (Fig. 3). The highest
1 JUNE 2018 TASZAREK ET AL. 4293

FIG. 10. As in Fig. 4, but for LLS.

values are observed over the area surrounding the describe the potential instability of the atmosphere and
Mediterranean Sea (Fig. 5). The largest amplitude of the has some advantages in presentation in comparison to
annual cycle of LR85 is observed over EE, which during CAPE (Brooks 2013). Physically, there is no difference
the cold season has modest lapse rates. Conversely, between looking at WMAX and CAPE, but the display
there are weak annual cycles over WE and SE. The naturally compresses the effect of extremely large values
highest values of LR85 are usually observed in the early of CAPE and allows us to see more details in the range
summer when strong diurnal heating has begun in the of low values. The warm season is characterized by large
boundary layer but cool air masses still remain in the values of instability with the timing of the peak slightly
midlevels. This is less pronounced (but still visible) over later in SE than other regions (Fig. 6). The SE has the
SE and WE, which are heated all year by the warm highest peak of instability and largest amplitude of the
waters of the Mediterranean Sea and Atlantic Ocean. annual cycle, with WE having the smallest peak and
The influence of the Gulf Stream during wintertime amplitude of the cycle. ML instability is underestimated
results in enhanced lapse rates in the corridor from in EE, particularly in the cool season. MU instability is
Iceland up to the Norwegian Sea (Fig. 5). Although overestimated by the reanalysis in the SE, particularly in
LR85 derived from ERA-Interim seems to be quite well the autumn, making the annual cycle more asymmetric.
correlated with observations, a slight overestimation is It is worth to highlight that except for EE, ML CAPE in
apparent, especially during the cool season. The ex- ERA-Interim has a good agreement with observations
ception is EE, which has an almost perfect fit between for the summer season, but tends to be lower in the
reanalysis and observational data. winter when shallower less well-mixed boundary layers
are common. A spatial maximum of mean ML CAPE is
c. Instability
observed over south-central Mediterranean Sea during
The parcel theory estimate of vertical velocity (WMAX; the cold season, and central Italy during the warm sea-
Table 3) associated with a value of CAPE has been used to son (Fig. 7). Observational data indicate also high mean
4294 JOURNAL OF CLIMATE VOLUME 31

FIG. 11. As in Fig. 3, but for ML WMAXSHEAR and ML WINIT. Values in parentheses denote ML CIN. Only
cases with nonzero ML WINIT and ML WMAXSHEAR are used.

values over northeastern Spain, the Carpathians, and features such as the jet stream. Jets are strongest during
southwestern Russia. the wintertime, and this is when the highest values of wind
shear are observed (Fig. 8). The general climatological
d. Vertical wind shear
pattern is anticorrelated with instability. Both DLS and
Vertical wind shear is an important parameter that in- LLS peak in winter throughout Europe and the annual
fluences storm organization and the potential to produce minimum is during the warm season. Low values extend
severe weather. It is directly related to synoptic-scale somewhat later into September in the SE, where the
1 JUNE 2018 TASZAREK ET AL. 4295

FIG. 12. As in Fig. 4, but for ML WMAXSHEAR.

annual cycle is weaker than in other regions. The highest magnitude for a given wind aloft. The highest clima-
annual amplitude is in EE. tological observed values of LLS are found from
Shear values are usually underestimated in the re- northern Germany to the British Isles throughout the
analysis (Fig. 8), consistent with previous work (Zwiers year. The reanalysis underestimates both DLS and
and Kharin 1998; Brooks et al. 2003; Allen et al. 2011; LLS over much of Europe but there is qualitative
Duruisseau et al. 2017). A larger underestimation is agreement in the spatial and temporal patterns with
found in LLS than deeper layers, especially evident observations.
during summertime when the 25th percentile of obser-
e. WMAXSHEAR and convective inhibition
vational data is at the median of the reanalysis. Despite
the underestimate, the pattern of the annual cycle is WMAXSHEAR as the product of WMAX and DLS
similar in the observations and reanalysis. can be calculated for any thermodynamic parcel. In
In the spatial analysis, the highest values of DLS in comparison to WMAX or DLS alone, the combination
ERA-Interim during the cold season are observed leads to smaller differences between cold and warm
over the British Isles, Norway, and the Alps while the seasons. This is evident over WE, which has a weak
minimum falls on the western Mediterranean Sea annual cycle in ML WMAXSHEAR (Fig. 11). Severe
(Fig. 9). During the warm season, DLS is considerably thunderstorm potential over WE is roughly the same
lower. A regional minimum is observed over the Uk- throughout the year, based on this parameter, but
raine. The spatial distribution of LLS indicates higher during wintertime it is mostly driven by high shear,
values over northern Europe and lower values over SE whereas during summertime it is driven more by in-
(Fig. 10). It is also notable that shear is higher over the stability. Larger annual cycles are observed over EE
land surface than water. This is probably due to higher and CE&B. As with CAPE, peak values are observed
friction over land, which contributes to slowing down in the summer. Over SE, relatively high values are
the surface winds, and thus enhancing overall shear observed throughout the year. The peak in annual cycle
4296 JOURNAL OF CLIMATE VOLUME 31

FIG. 13. As in Fig. 4, but for ML CIN.

is 1–2 months later than over CE&B. Both reanalysis over WE, SE, and CE&B (Fig. 11). This suggests that
and sounding data indicate that the highest warm sea- the influence of ML CIN on the storm activity in these
son means of ML WMAXSHEAR are found from Italy regions is not very strong. Conversely, large negative
through the western Mediterranean Sea, with a second values of ML CIN are observed over SE, with sum-
corridor from the western Ukraine to Bulgaria (Fig. 12). mertime values frequently below 250 J kg21. The most
During the cold season, ML WMAXSHEAR peaks negative values are in the western and central Medi-
over the central Mediterranean Sea and the waters terranean Sea (Fig. 13), indicating that although high
west of Norway, driven by the high shear, and relatively ML WMAXSHEAR is frequently seen there (Fig. 12),
warm water. much of the potential convective activity is suppressed
Regardless of high ML WMAXSHEAR indicating by the CIN. This so-called severe weather efficiency
the potential for severe weather, no storm will form if (Brooks 2009) has also been noted by Groenmeijer
convective initiation does not take place. For this et al. (2017), who found that the fraction of severe
purpose, we analyze the convective inhibition (ML thunderstorms compared to days with potential for a
CIN) and the theoretical parcel vertical velocity to severe thunderstorm (CAPE 3 DLS . 10 000 m3 s23) is
overcome CIN (ML WINIT; Westermayer et al. 2017) much higher over continental Europe (mostly between
for initiation. The annual cycle of ML CIN is similar to 0.6 and 0.8) than over the central Mediterranean Sea
ML CAPE, with the most negative values of ML CIN (below 0.2).
during the summer, although the reanalysis has more
f. Lifted condensation level
positive values than observations (Fig. 11). As sug-
gested by Westermayer et al. (2017), the probability There are differences in ML LCL among various
for a thunderstorm sharply decreases when CIN drops regions. The annual cycle of ML LCL has the lowest
below 250 J kg21. Such values are in general between amplitudes over WE and the highest over EE (Fig. 14).
the 75th and 90th percentile of summertime distributions There are consistently high values in the warm season
1 JUNE 2018 TASZAREK ET AL. 4297

in all regions, with WE having the lowest ML LCL on


average. The annual cycle for SE shows a nearly
monotonic increase from January to July, followed by a
decrease for the rest of the year. The highest values
during the warm season are observed over the regions
just north of the Mediterranean Sea and over the
eastern Ukraine and southwestern Russia (Fig. 15).
The smallest values are over northwestern Europe. The
reanalysis represents ML LCL well during the warm
season over all regions except SE, but tends to un-
derestimate it during the cold months. As with in-
stability, the largest underestimation of ML LCL is
seen over EE. Over SE, ML LCL is underestimated
throughout the year.
g. Annual cycles at individual stations
Apart from the general climatological overview, it is
worthwhile to look at individual locations to provide
more details on the annual cycle of shear and instability.
To do so, we have chosen six stations for each region
approximately evenly distributed in space and contain-
ing roughly 10 000 observations each (Fig. 1; also, see the
appendix). Following Brooks et al. (2007), we computed a
60-day running mean of ML WMAX and DLS for each
location.
Stations located on the western coast under the
strong influence of Atlantic Ocean (La Coruna, Brest,
and Stavanger) show a weak annual cycle of insta-
bility but large changes in vertical wind shear (Fig. 16).
Stations located farther to the east (Bordeaux, Trappes,
and De Bilt) have a similar shear pattern to western
stations, but have higher instability during summer-
time. The reanalysis represents observations relatively
well in Bordeaux and De Bilt but tends to un-
derestimate instability over Brest and Stavanger. Shear
is most notably underestimated over Brest, Trappes,
and La Coruna.
The annual cycles in CE&B resemble the shape of a
bow with substantial changes of both shear and in-
stability (Fig. 17). Although all the stations show sim-
ilar patterns, higher values of ML WMAX are observed
in the southern part of domain (Udine, Budapest, and
Bucharest). Shear decreases and instability increases
from January to July, with the pattern reversing in the rest
of the year. This is most obvious over the northern part of
the domain (Stuttgart, Prague, and Legionowo), whereas
in the southern part of the domain shear during the warm
season is relatively constant with changes only in the in-
FIG. 14. As in Fig. 3, but for ML LCL. stability. It is also worth noting that considerable values
of ML CIN (which contribute to convective suppres-
sion) are observed in July over Udine and Bucharest
(Fig. 17). Comparing reanalysis and observations, it can
be noted that shear is underestimated at all locations.
4298 JOURNAL OF CLIMATE VOLUME 31

FIG. 15. As in Fig. 4, but for ML LCL.

The accuracy of the reanalysis instability varies with annual cycle than the rest of stations in SE. Small
the location, which may be a result of a complex Eu- changes in instability and large changes in shear are
ropean orography not being well represented in the more similar to WE than to Murcia located roughly on
reanalysis. the same latitude.
A similar pattern is observed in EE. The main dif-
h. Frequency of thunderstorm and severe
ference is that the amplitudes of the annual cycles of
thunderstorm environments
both shear and instability are larger (Fig. 18). As in the
previous domains, shear in the reanalysis is under- Since our main interest in this study is ingredients for
estimated in each location, as is instability. Over some deep moist convection, we estimate the number of days
locations (Moscow, Kiev, Rostov, and Saratov), the per year with conditions favoring severe thunder-
underestimation of ML WMAX in summer is consid- storms by using instability and shear parameters. A
erable. This is of particular interest because all these similar approach was used in many previous studies
stations are located in the plains without complex (Brooks et al. 2003; Trapp et al. 2007, 2009; Marsh et al.
orography. 2007, 2009; Diffenbaugh et al. 2013; Allen et al. 2014;
Unlike the other regions, most of the stations in SE Allen and Karoly 2014; Seeley and Romps 2015; Púcik
have a weak annual cycle of shear and large changes in et al. 2017). We define a potential thunderstorm day
instability (Fig. 19). In the summer and autumn, ML if ML CAPE . 100 J kg21 and ML CIN . 250 J kg21
CIN is considerable and limits the development of (Table 5), based on estimates of thunderstorm events
convection. In contrast to other locations, the reanalysis in Europe with their accompanying atmospheric con-
tends to overestimate instability, particularly in the ditions (Kaltenböck et al. 2009; Westermayer et al.
western parts of the domain (Murcia, Ajaccio, and Ca- 2017; Kolendowicz et al. 2017; Taszarek et al. 2017).
gliari). These locations are in coastal regions or sur- The ML CIN term is used to exclude cases with inhi-
rounded by complex terrain. Athens has a very different bition strong enough to suppress deep moist convection
1 JUNE 2018 TASZAREK ET AL. 4299

FIG. 16. Scatterplot of DLS and ML WMAX for chosen sounding sites in the WE domain. Lines and points represent 60-day
moving average for ERA-Interim (red) and sounding observations (green). Darker points represent the 15th day of month
numbered respectively for January, April, July, and October. Black points represent days in which mean 60-day moving average
ML CIN , 250 J kg21.

(Sander 2011; Gensini and Ashley 2011; Diffenbaugh observations, peak values occur over the Balkan Pen-
et al. 2013). A day with a potential severe thunder- insula and EE.
storm adds WMAXSHEAR . 300 m2 s22 (Taszarek Potential severe thunderstorms follow a similar pat-
et el. 2017). However, we are aware that no discrimi- tern to potential thunderstorm but with one-third to
nator will perfectly distinguish between severe and one-half the frequency. In the reanalysis, more than
nonsevere thunderstorm environments (Doswell and 15 days per warm season occur in the Balkan Peninsula
Schultz 2006), and thus any such analysis will always and a corridor from Italy to Germany (Fig. 21). In the
be subject to a probabilistic interpretation (Allen and observations, the Carpathians and most of EE have the
Karoly 2014). highest frequency. During the cold season, severe
During the cold season, conditions conducive to thunderstorm environments are observed most often
thunderstorm development are mainly over the Med- over the central Mediterranean Sea.
iterranean area, especially its central part, with a sec- From September to April the most frequent condi-
ondary maximum along the western coast of Europe tions supportive of both thunderstorms and severe
(Fig. 20). During the warm season, convective activity thunderstorms are mainly over SE and WE (Fig. 22).
moves inland. In the reanalysis, the most frequent From May to August, the main storm activity moves to
thunderstorm environments occur over Italy, Austria, EE and CE&B. For SE, the peak frequency is in Sep-
and the Balkan Peninsula, with a second region of high tember whereas in July and August thunderstorm ac-
frequency from Germany to western Russia. In the tivity is suppressed due to considerable CIN. It is again
4300 JOURNAL OF CLIMATE VOLUME 31

FIG. 17. As in Fig. 16, but for the CE&B domain.

evident that the reanalysis underestimates thunderstorm Allen and Karoly 2014; Gensini et al. 2014). Zwiers and
environments over EE (Fig. 22). Kharin (1998) pointed out that low-level winds in
the NCEP–NCAR reanalysis are underestimated. A
4. Comparison with previous studies systematic underestimation of the wind speed in ERA-
Interim was also indicated by Duruisseau et al. (2017).
Caution must be taken when looking at fields in- Both Brooks et al. (2003) and Allen et al. (2011)
volving strong vertical gradients, which the reanalysis pointed out that LLS is underestimated in the re-
has difficulties with (Brooks 2009). Allen and Karoly analysis and suggested that this may be due to its poor
(2014) suggested that sharp horizontal boundaries such vertical resolution being not able to capture important
as coastal regions or topographic boundaries may ex- gradients.
pose poor spatial resolution of the reanalysis. We sus- Our findings concerning the climatological aspects
pect that some of the errors we found may be associated of CAPE and CIN are in agreement with Siedlecki
with strongly diversified orography and the coastline (2009), who analyzed individual sounding stations,
of Europe. and Riemann-Campe et al. (2009), who used ERA-40
The performance of the reanalysis in reproducing reanalysis. The highest means of CAPE over the Med-
observations has been analyzed in many previous iterranean Sea, the Balkan Peninsula, and southwestern
studies. Our work supports the result that kinematic Spain were also noted by Romero et al. (2007) and
parameters are better represented in contrast to ther- Pistotnik et al. (2016).
modynamic indices in a variety of reanalyses (Lee 2002; The notion of high (low) instability being associated
Niall and Walsh 2005; Thompson et al. 2003, 2007; with low (high) shear was also pointed out by Brooks
1 JUNE 2018 TASZAREK ET AL. 4301

FIG. 18. As in Fig. 16, but for the EE domain.

et al. (2007). In general, our study agree with their re- in northern Italy. Contrary to the ERA-Interim results
sults that there is more variability in convective pa- presented here, Púcik et al. (2017) found local maxima
rameters in a cross section from north to south in central of severe thunderstorm environments over southern France
Europe compared to one from east to west, mostly be- and northeastern Spain using European Coordinated
cause of the increase in moisture from north to south. In Regional Downscaling Experiment (EURO-CORDEX)
addition, in comparison to the United States, the lapse data (Jacob et al. 2014).
rates and mixing ratios in Europe are lower, reflecting
the absence of source areas comparable to the Rocky
5. Summary and concluding remarks
Mountains (lapse rates) or Gulf of Mexico (low-level
moisture). As a result, high values of CAPE are much In this study, over 1 million sounding measurements
less likely in Europe than North America, as also seen in was compared with ERA-Interim reanalysis for the
Brooks et al. (2003). 38-yr period from 1979 to 2016. The large dataset allowed
The European lightning climatology has a peak us to provide an improved insight into the spatial and
lightning density over northern Italy and the Balkan temporal distributions of the prerequisites of deep moist
Peninsula (Anderson and Klugmann 2014), which our convection across Europe. In addition, ERA-Interim
estimates of days with potential thunderstorms supports. was evaluated. Our findings are consistent with re-
We also find a similar patterns in the annual cycle of sults from previous studies of thunderstorm environ-
thunderstorms. Our results regarding the spatial distribu- ments in Europe (Brooks et al. 2003; Brooks et al. 2007;
tion of severe thunderstorm environments agree with Romero et al. 2007; Siedlecki 2009; Brooks 2009;
Groenemeijer et al. (2017), indicating a peak frequency Pistotnik et al. 2016; Kolendowicz et al. 2017; Púcik
4302 JOURNAL OF CLIMATE VOLUME 31

FIG. 19. As in Fig. 16, but for the SE domain.

et al. 2017; Taszarek et al. 2017; Groenemeijer et al. is overestimated in southern Europe and under-
2017) but expand on that work because of the large estimated over eastern Europe. Vertical shear pa-
sample size. rameters are better correlated than CAPE but are
ERA-Interim estimates basic parameters describing underestimated by approximately 1–2 m s21. The rel-
moisture (MIXR) and lapse rates (LR85) well with ative amount of underestimation decreases from LLS
correlation coefficients of 0.94. Bulk instability param- to deeper shear layers.
eters, such as CAPE, are not represented as well. ML Enhanced values of instability are observed from May
CAPE is, on average, underestimated while MU CAPE to September, out of phase with the climatological pat-
is overestimated, indicating possible problems with the tern of wind shear, which peaks in winter. From Sep-
boundary layer representation. Instability in ERA-Interim tember to April, favorable conditions for thunderstorms
occur mainly over southern and western Europe, with
the peak location and higher frequency shifting to cen-
TABLE 5. Definition of potential thunderstorm and potential tral and eastern Europe from May to August. For
severe thunderstorm day. southern Europe, the annual cycle peaks in September
with high values of inhibition suppressing thunderstorm
Potential thunderstorm ML CAPE . 100 J kg21
ML CIN . 250 J kg21 activity in July and August. The area with the highest
Potential severe ML CAPE . 100 J kg21 annual number of days with environmental conditions
thunderstorm favorable for thunderstorms in the reanalysis extends
ML CIN . 250 J kg21 from Italy and Austria eastward through the Carpathians
ML WMAXSHEAR . 300 m2 s22
and Balkans.
1 JUNE 2018 TASZAREK ET AL. 4303

FIG. 20. As in Fig. 4, but for the mean annual number of days with potential thunderstorm (ML CAPE . 100 J kg21
and ML CIN . 250 J kg21).

Although most of our results are not qualitatively shear being poorly represented in the reanalysis
surprising, we have developed quantitative assess- (Púcik et al. 2017). However, as shown here, many
ments, leading to more precise descriptions of cli- errors in ERA-Interim may not affect the qualitative
matological aspects of ingredients for deep moist interpretation of results, but could lead to quantita-
convection in Europe. Our results may be used as a tive differences if reanalysis-based thresholds are
background for future studies on thunderstorm en- applied directly to observed soundings. Severe thun-
vironments in Europe, and also can be a valuable derstorm discriminators (Brooks et al. 2003; Trapp
source of information for various groups such as et al. 2009; Allen et al. 2011) or composite parameters
weather forecasters and insurance companies. Dif- (Thompson et al. 2003; Sherburn and Parker 2014)
ferences that were found between ERA-Interim and designed with certain models and domains should
observational data indicate that researchers doing be applied to other datasets and geographical loca-
work based solely on the reanalysis data should be tions with caution because they may not perform
cautious when drawing conclusions regarding in- quantitatively in the same way. Researchers using
stability and low-level shear parameters. Issues with reanalysis datasets to analyze convective variables
these near-ground parameters may be indicative of should consider examining the errors in parameters
the challenges in representing the boundary and before applying the results to other models or ob-
gradients within it. A similar caveat relates to areas servations. Future studies will continue to address
with strong horizontal gradients in topography, which this problem and more comparisons between model
may lead to orographically induced mesoscale circu- and observational data will be available to assess how
lations that modify the environment around the well various numerical models represent the observed
mountain ranges and modulate latent instability and environment.
4304 JOURNAL OF CLIMATE VOLUME 31

FIG. 21. As in Fig. 4, but for the mean annual number of days with potential severe thunderstorm (ML CAPE .
100 J kg21, ML CIN . 250 J kg21 and ML WMAXSHEAR . 300 m2 s22).

Acknowledgments. We appreciate valuable com- National Science Centre (Project No. 2014/13/N/ST10/
ments of Pieter Groenemeijer, John Allen, and anon- 01708). The lead author was supported by the doctoral
ymous reviewers who helped to improve this study. scholarship of the National Science Centre (Project No.
This research was supported by the grant of the Polish 2015/16/T/ST10/00373) and the Foundation for Polish

FIG. 22. (left) Mean annual number of days with potential thunderstorm (ML CAPE . 100 J kg21 and ML CIN . 250 J kg21) per
station given a particular domain: WE (green), CE&B (red), EE (blue), and SE (orange). (right) As at (left), but for potential severe
thunderstorm (ML CAPE . 100 J kg21, ML CIN . 250 J kg21, and ML WMAXSHEAR . 300 m2 s22). Solid (dashed) line indicate
estimates based on sounding observations (ERA-Interim).
TABLE A1. List of sounding stations (by WMO ID) including number of 1200 UTC sounding observations used in the study, average number of pressure levels per measurement, and
years of operation.

Avg. No. of Avg. No. of Avg. No. of


1 JUNE 2018

WMO levels per levels per levels per


ID No. sounding Years of operation WMO ID No. sounding Years of operation WMO ID No. sounding Years of operation
01001 11 525 44.4 1979–2016 10200 7412 56.0 1982–2011 17220 9240 42.3 1979–2016
01152 8935 37.6 1979–2014 10238 9745 56.7 1979–2016 17240 7820 42.6 1979–2016
01241 10 123 40.3 1979–2016 10393 8775 74.2 1991–2016 17281 3925 71.5 2005–16
01400 5416 45.6 1994–2016 10410 12 984 65.0 1979–2016 17351 6517 52.6 1994–2016
01415 11 054 41.2 1979–2016 10437 2336 22.4 1982–2004 22217 9578 42.5 1979–2016
02365 11 644 47.9 1979–2016 10548 8709 67.9 1991–2016 22271 9200 37.4 1979–2016
02527 8377 34.5 1979–2016 10618 9937 59.1 1979–2016 22522 8510 34.6 1979–2016
02591 10 132 47.5 1979–2016 10739 12 935 64.7 1979–2016 22820 6289 37.1 1979–2016
02836 12 747 50.8 1979–2016 10771 9615 59.1 1979–2016 22845 7417 34.3 1979–2016
02963 13 254 50.8 1979–2016 10868 12 954 56.7 1979–2016 23205 9237 33.8 1979–2016
03005 12 597 101.8 1979–2016 11035 13 215 45.5 1979–2016 26038 5559 22.4 1979–2013
03238 3225 126.6 2002–16 11520 13 343 53.6 1979–2016 26063 11 859 39.8 1979–2016
03354 4811 126.3 1997–2016 11747 4730 66.5 2003–16 26298 8256 38.2 1979–2016
03808 12 884 95.4 1979–2016 11952 12 410 53.2 1979–2016 26477 6708 42.0 1979–2016
03882 6284 100.4 1980–2016 12120 10 540 50.8 1979–2016 26629 5013 22.1 1979–2010
03953 12 491 39.6 1979–2016 12374 11 910 52.6 1979–2016 26702 5663 32.5 1979–2016
04018 11 331 45.5 1979–2016 12425 8060 61.8 1979–2016 26781 9750 36.8 1979–2016
06011 11 484 41.6 1979–2016 12843 12 286 34.7 1979–2016 27199 6196 48.4 1994–2016
06260 11 799 48.4 1979–2014 12982 5584 30.1 1979–2016 27459 6192 48.1 1994–2016
06610 13 134 48.2 1979–2016 13275 6270 30.2 1979–2016 27595 8764 34.7 1979–2016
07110 11 946 21.2 1979–2016 14240 6151 51.6 1996–2016 27612 10 426 38.2 1979–2016
07145 12 169 20.9 1979–2016 14430 4342 63.9 2002–16 27707 8744 35.7 1979–2016
TASZAREK ET AL.

07180 9234 17.5 1979–2011 15120 3580 19.6 1979–2010 27730 5700 55.1 1994–2016
07481 8803 18.2 1979–2009 15420 9337 46.0 1979–2016 27962 6241 46.0 1979–2016
07510 12 292 27.6 1979–2016 15614 11 125 29.7 1979–2016 27995 5144 54.1 1994–2016
07645 11 854 21.2 1979–2016 16044 12 697 57.8 1979–2016 33041 5993 35.3 1979–2016
07761 12 118 21.9 1979–2016 16080 12 466 60.2 1979–2016 33345 9699 29.4 1979–2016
08001 10 752 53.8 1979–2016 16113 3064 66.9 1999–2016 33393 3968 19.6 1979–2008
08023 9069 51.6 1986–2016 16144 5714 49.6 1987–2016 33658 4006 19.4 1979–1997
08160 6407 51.9 1992–2015 16245 9712 69.2 1987–2016 33791 4584 18.0 1979–2016
08190 2714 68.7 2007–16 16320 12 481 51.7 1979–2016 33837 4034 22.1 1979–1998
08221 11 425 60.6 1979–2016 16429 11 699 60.4 1979–2016 34009 10 492 36.2 1979–2016
08302 6846 69.3 1979–2016 16546 1750 103.7 2012–16 34122 8467 34.5 1979–2016
08430 10 162 60.3 1984–2016 16560 9737 55.7 1979–2012 34172 8614 36.6 1979–2016
08495 11 510 75.1 1979–2014 16622 5429 24.0 1980–2014 34247 10 340 37.9 1979–2016
08579 9762 33.5 1980–2016 16716 10 087 32.7 1979–2016 34300 4395 23.7 1979–1999
10035 13 015 52.7 1979–2016 16754 3964 29.6 1979–2014 34731 10 109 33.8 1979–2016
10046 2344 21.9 1980–1994 17030 9031 43.6 1979–2016 34858 9241 41.4 1979–2016
10113 1939 100.9 2011–16 17095 2644 85.1 2007–16 60390 10 587 26.9 1979–2016
10184 8703 69.8 1991–2016 17130 10 011 38.7 1979–2016 TOTAL 1 027 559 47.7 —
4305
4306 JOURNAL OF CLIMATE VOLUME 31

Science (FNP). The reanalysis and sounding computa- Cady-Pereira, K. E., M. W. Shephard, D. D. Turner, E. J. Mlawer,
tions were performed in Poznan Supercomputing and S. A. Clough, and T. J. Wagner, 2008: Improved daytime
column-integrated precipitable water vapor from Vaisala ra-
Networking Center (Grant 331).
diosonde humidity sensors. J. Atmos. Oceanic Technol., 25,
873–883, https://doi.org/10.1175/2007JTECHA1027.1.
Celi
nski-Mysław, D., and A. Palarz, 2017: The occurrence of
APPENDIX convective systems with a bow echo in warm season in
Poland. Atmos. Res., 193, 26–35, https://doi.org/10.1016/
j.atmosres.2017.04.015.
List of Sounding Stations Craven, J. P., and H. E. Brooks, 2004: Baseline climatology of
sounding derived parameters associated with deep moist
The observations used, average number of pressure convection. Natl. Wea. Dig., 28, 13–24, http://www.nssl.noaa.
levels per measurement, and years of operation are gov/users/brooks/public_html/papers/cravenbrooksnwa.pdf.
shown in Table A1. Davies, J. M., and R. H. Johns, 1993: Some wind and instability
parameters associated with strong and violent tornadoes: 1.
Wind shear and helicity. The Tornado: Its Structure, Dynam-
REFERENCES ics, Prediction, and Hazards, Geophys. Monogr., Vol. 79,
Allen, J. T., and D. J. Karoly, 2014: A climatology of Australian Amer. Geophys. Union, 573–582.
severe thunderstorm environments 1979–2011: Inter-annual Dee, D. P., and Coauthors, 2011: The ERA-Interim reanalysis:
variability and ENSO influence. Int. J. Climatol., 34, 81–97, Configuration and performance of the data assimilation sys-
https://doi.org/10.1002/joc.3667. tem. Quart. J. Roy. Meteor. Soc., 137, 553–597, https://doi.org/
——, ——, and G. A. Mills, 2011: A severe thunderstorm clima- 10.1002/qj.828.
tology for Australia and associated thunderstorm environ- de Leeuw, J., J. Methven, and M. Blackburn, 2015: Evaluation of
ments. Austr. Meteor. Ocean., 61, 143–158, https://doi.org/ ERA-Interim reanalysis precipitation products using England
10.22499/2.6103.001. and Wales observations. Quart. J. Roy. Meteor. Soc., 141, 798–
——, ——, and K. J. Walsh, 2014: Future Australian severe thun- 806, https://doi.org/10.1002/qj.2395.
derstorm environments. Part II: The influence of a strongly Diffenbaugh, N. S., M. Scherer, and R. J. Trapp, 2013: Robust in-
warming climate on convective environments. J. Climate, 27, creases in severe thunderstorm environments in response to
3848–3868, https://doi.org/10.1175/JCLI-D-13-00426.1. greenhouse forcing. Proc. Natl. Acad. Sci. USA, 110, 16 361–
Anderson, G., and D. Klugmann, 2014: A European lightning 16 366, https://doi.org/10.1073/pnas.1307758110.
density analysis using 5 years of ATDnet data. Nat. Hazard. Doswell, C. A., III, and E. N. Rasmussen, 1994: The effect of ne-
Earth. Syst. Sci., 14, 815–829, https://doi.org/10.5194/nhess-14- glecting the virtual temperature correction on CAPE calcu-
815-2014. lations. Wea. Forecasting, 9, 625–629, https://doi.org/10.1175/
Antonescu, B., and S. Burcea, 2010: A cloud-to-ground lightning 1520-0434(1994)009,0625:TEONTV.2.0.CO;2.
climatology for Romania. Mon. Wea. Rev., 138, 579–591, ——, and D. M. Schultz, 2006: On the use of indices and parameters
https://doi.org/10.1175/2009MWR2975.1. in forecasting severe storms. Electronic J. Severe Storms Me-
Bao, X., and F. Zhang, 2013: Evaluation of NCEP–CFSR, NCEP– teor., 1 (3), http://www.ejssm.org/ojs/index.php/ejssm/article/
NCAR, ERA-Interim, and ERA-40 reanalysis datasets viewArticle/11/12.
against independent sounding observations over the Tibetan ——, H. E. Brooks, and R. A. Maddox, 1996: Flash flood
Plateau. J. Climate, 26, 206–214, https://doi.org/10.1175/JCLI- forecasting: An ingredients-based methodology. Wea.
D-12-00056.1. Forecasting, 11, 560–581, https://doi.org/10.1175/1520-
Betz, H. D., K. Schmidt, P. Laroche, P. Blanchet, W. P. Oettinger, 0434(1996)011,0560:FFFAIB.2.0.CO;2.
E. Defer, Z. Dziewit, and J. Konarski, 2009: LINET—An in- Dotzek, N., P. Groenemeijer, B. Feuerstein, and A. M. Holzer, 2009:
ternational lightning detection network in Europe. Atmos. Res., Overview of ESSL’s severe convective storms research using
91, 564–573, https://doi.org/10.1016/j.atmosres.2008.06.012. the European Severe Weather Database ESWD. Atmos. Res.,
Brooks, H. E., 2009: Proximity soundings for severe convection for 93, 575–586, https://doi.org/10.1016/j.atmosres.2008.10.020.
Europe and the United States from reanalysis data. Atmos. Res., Duruisseau, F., N. Huret, A. Andral, and C. Camy-Peyret, 2017:
93, 546–553, https://doi.org/10.1016/j.atmosres.2008.10.005. Assessment of the ERA-Interim reanalysis winds using high-
——, 2013: Severe thunderstorms and climate change. Atmos. Res., altitude stratospheric balloons. J. Atmos. Sci., 74, 2065–2080,
123, 129–138, https://doi.org/10.1016/j.atmosres.2012.04.002. https://doi.org/10.1175/JAS-D-16-0137.1.
——, J. W. Lee, and J. P. Craven, 2003: The spatial distribution of Gensini, V. A., and W. S. Ashley, 2011: Climatology of potentially
severe thunderstorms and tornado environments from global severe convective environments from North American regional
reanalysis data. Atmos. Res., 67–68, 73–94, https://doi.org/ reanalysis. Electronic J. Severe Storms Meteor, 6 (8), http://
10.1016/S0169-8095(03)00045-0. www.ejssm.org/ojs/index.php/ejssm/article/viewArticle/85.
——, A. R. Anderson, K. Riemann, I. Ebbers, and H. Flachs, 2007: ——, T. L. Mote, and H. E. Brooks, 2014: Severe-thunderstorm
Climatological aspects of convective parameters from the reanalysis environments and collocated radiosonde observa-
NCAR/NCEP reanalysis. Atmos. Res., 83, 294–305, https://doi.org/ tions. J. Appl. Meteor. Climatol., 53, 742–751, https://doi.org/
10.1016/j.atmosres.2005.08.005. 10.1175/JAMC-D-13-0263.1.
Bunkers, M. J., B. A. Klimowski, J. W. Zeitler, R. L. Thompson, Grams, J. S., R. L. Thompson, D. V. Snively, J. A. Prentice,
and M. L. Weisman, 2000: Predicting supercell motion G. M. Hodges, and L. J. Reames, 2012: A climatology and
using a new hodograph technique. Wea. Forecasting, 15, comparison of parameters for significant tornado events in
61–79, https://doi.org/10.1175/1520-0434(2000)015,0061: the United States. Wea. Forecasting, 27, 106–123, https://
PSMUAN.2.0.CO;2. doi.org/10.1175/WAF-D-11-00008.1.
1 JUNE 2018 TASZAREK ET AL. 4307

Groenemeijer, P., and A. van Delden, 2007: Sounding-derived Pistotnik, G., P. Groenemeijer, and R. Sausen, 2016: Validation of
parameters associated with large hail and tornadoes in the convective parameters in MPI-ESM decadal hindcasts (1971–
Netherlands. Atmos. Res., 83, 473–487, https://doi.org/10.1016/ 2012) against ERA-Interim reanalyses. Meteor. Z., 25, 753–
j.atmosres.2005.08.006. 766, https://doi.org/10.1127/metz/2016/0649.
——, and T. Kühne, 2014: A climatology of tornadoes in Europe: Poelman, D. R., W. Schulz, G. Diendorfer, and M. Bernardi, 2016:
Results from the European Severe Weather Database. Mon. The European lightning location system EUCLID—Part 2:
Wea. Rev., 142, 4775–4790, https://doi.org/10.1175/MWR-D- Observations. Nat. Hazard. Earth Syst. Sci., 16, 607–616,
14-00107.1. https://doi.org/10.5194/nhess-16-607-2016.
——, and Coauthors, 2017: Severe convective storms in Europe: Pohjola, H., and A. Mäkelä, 2013: The comparison of GLD360 and
Ten years of research and education at the European Severe EUCLID lightning location systems in Europe. Atmos. Res.,
Storms Laboratory. Bull. Amer. Meteor. Soc., 98, 2641–2651, 123, 117–128, https://doi.org/10.1016/j.atmosres.2012.10.019.
https://doi.org/10.1175/BAMS-D-16-0067.1. Púcik, T., P. Groenemeijer, D. Rýva, and M. Kolár, 2015: Prox-
Grünwald, S., and H. E. Brooks, 2011: Relationship between imity soundings of severe and nonsevere thunderstorms in
sounding derived parameters and the strength of tornadoes in central Europe. Mon. Wea. Rev., 143, 4805–4821, https://doi.
Europe and the USA from reanalysis data. Atmos. Res., 100, org/10.1175/MWR-D-15-0104.1.
479–488, https://doi.org/10.1016/j.atmosres.2010.11.011. ——, and Coauthors, 2017: Future changes in European severe
Guichard, F., D. Parsons, and E. Miller, 2000: Thermodynamic and convection environments in a regional climate model ensem-
radiative impact of the correction of sounding humidity bias in ble. J. Climate, 30, 6771–6794, https://doi.org/10.1175/JCLI-D-
the tropics. J. Climate, 13, 3611–3624, https://doi.org/10.1175/ 16-0777.1.
1520-0442(2000)013,3611:TARIOT.2.0.CO;2. Punge, H. J., and M. Kunz, 2016: Hail observations and hailstorm
Guo, Y., S. Zhang, J. Yan, Z. Chen, and X. Ruan, 2016: A com- characteristics in Europe: A review. Atmos. Res., 176, 159–184,
parison of atmospheric temperature over China between https://doi.org/10.1016/j.atmosres.2016.02.012.
radiosonde observations and multiple reanalysis datasets. Rasmussen, E. N., and D. O. Blanchard, 1998: A baseline clima-
J. Meteor. Res., 30, 242–257, https://doi.org/10.1007/s13351- tology of sounding-derived supercell and tornado forecast
016-5169-0. parameters. Wea. Forecasting, 13, 1148–1164, https://doi.org/
Hodges, K. I., R. W. Lee, and L. Bengtsson, 2011: A comparison of 10.1175/1520-0434(1998)013,1148:ABCOSD.2.0.CO;2.
extratropical cyclones in recent reanalyses ERA-Interim, Riemann-Campe, K., K. Fraedrich, and F. Lunkeit, 2009: Global
NASA MERRA, NCEP CFSR, and JRA-25. J. Climate, 24, climatology of convective available potential energy (CAPE)
4888–4906, https://doi.org/10.1175/2011JCLI4097.1. and convective inhibition (CIN) in ERA-40 reanalysis. Atmos.
Jacob, D., and Coauthors, 2014: EURO-CORDEX: New high- Res., 93, 534–545, https://doi.org/10.1016/j.atmosres.2008.09.037.
resolution climate change projections for European impact Romero, R., M. Gayà, and C. A. Doswell, 2007: European clima-
research. Reg. Environ. Change, 14, 563–578, https://doi.org/ tology of severe convective storm environmental parameters:
10.1007/s10113-013-0499-2. A test for significant tornado events. Atmos. Res., 83, 389–404,
Kaltenböck, R., G. Diendorfer, and N. Dotzek, 2009: Evaluation of https://doi.org/10.1016/j.atmosres.2005.06.011.
thunderstorm indices from ECMWF analyses, lightning data Sander, J., 2011: Extremwetterereignisse im Klimawandel: Be-
and severe storm reports. Atmos. Res., 93, 381–396, https:// wertung der derzeitigen und zukünftigen Gefährdung. Ph.D.
doi.org/10.1016/j.atmosres.2008.11.005. thesis, University of Munich, 125 pp.
Kolendowicz, L., M. Taszarek, and B. Czernecki, 2017: Atmospheric Seeley, J. T., and D. M. Romps, 2015: The effect of global warming
circulation and sounding-derived parameters associated with on severe thunderstorms in the United States. J. Climate, 28,
thunderstorm occurrence in central Europe. Atmos. Res., 191, 2443–2458, https://doi.org/10.1175/JCLI-D-14-00382.1.
101–114, https://doi.org/10.1016/j.atmosres.2017.03.009. Sherburn, K. D., and M. D. Parker, 2014: Climatology and in-
Lee, J. W., 2002: Tornado proximity soundings from the NCEP/ gredients of significant severe convection in high-shear, low-
NCAR reanalysis data. M.S. thesis, Dept. of Meteorology, CAPE environments. Wea. Forecasting, 29, 854–877, https://
University of Oklahoma, 61 pp. doi.org/10.1175/WAF-D-13-00041.1.
Mäkelä, A., S. E. Enno, and J. Haapalainen, 2014: Nordic Light- Siedlecki, M., 2009: Selected instability indices in Europe. Theor.
ning Information System: Thunderstorm climate of northern Appl. Climatol., 96, 85–94, https://doi.org/10.1007/s00704-008-
Europe for the period 2002–2011. Atmos. Res., 139, 46–61, 0034-4.
https://doi.org/10.1016/j.atmosres.2014.01.008. Szczypta, C., J. C. Calvet, C. Albergel, G. Balsamo, S. Boussetta,
Marsh, P. T., H. E. Brooks, and D. J. Karoly, 2007: Assessment of D. Carrer, S. Lafont, and C. Meurey, 2011: Verification of
the severe weather environment in North America simulated the new ECMWF ERA-Interim reanalysis over France.
by a global climate model. Atmos. Sci. Lett., 8, 100–106, https:// Hydrol. Earth Syst. Sci., 15, 647–666, https://doi.org/
doi.org/10.1002/asl.159. 10.5194/hess-15-647-2011.
——, ——, and ——, 2009: Preliminary investigation into the se- Szuster, P., 2016: The implementation of computing system for
vere thunderstorm environment of Europe simulated by the determining the temperature change for trial air particles,
Community Climate System Model 3. Atmos. Res., 93, 607– M.S. thesis, Dept. of Computer Science, Tadeusz Kosciuszko
618, https://doi.org/10.1016/j.atmosres.2008.09.014. Cracow University of Technology, 42 pp.
Mooney, P. A., F. J. Mulligan, and R. Fealy, 2011: Comparison Taszarek, M., and L. Kolendowicz, 2013: Sounding-derived pa-
of ERA-40, ERA-Interim and NCEP/NCAR reanalysis rameters associated with tornado occurrence in Poland and
data with observed surface air temperatures over Ireland. Universal Tornadic Index. Atmos. Res., 134, 186–197, https://
Int. J. Climatol., 31, 545–557, https://doi.org/10.1002/joc.2098. doi.org/10.1016/j.atmosres.2013.07.016.
Niall, S., and K. Walsh, 2005: The impact of climate change on ——, and H. E. Brooks, 2015: Tornado climatology of Poland.
hailstorms in southeastern Australia. Int. J. Climatol., 25, Mon. Wea. Rev., 143, 702–717, https://doi.org/10.1175/MWR-
1933–1952, https://doi.org/10.1002/joc.1233. D-14-00185.1.
4308 JOURNAL OF CLIMATE VOLUME 31

——, B. Czernecki, and A. Kozioł, 2015: A cloud-to-ground Geophys. Res. Lett., 36, L01703, https://doi.org/10.1029/
lightning climatology for Poland. Mon. Wea. Rev., 143, 2008GL036203.
4285–4304, https://doi.org/10.1175/MWR-D-15-0206.1. Uppala, S. M., and Coauthors, 2005: The ERA-40 Re-Analysis.
——, H. E. Brooks, and B. Czernecki, 2017: Sounding-derived pa- Quart. J. Roy. Meteor. Soc., 131, 2961–3012, https://doi.org/
rameters associated with convective hazards in Europe. Mon. Wea. 10.1256/qj.04.176.
Rev., 145, 1511–1528, https://doi.org/10.1175/MWR-D-16-0384.1. van Vuuren, D. P., J. A. Edmonds, M. Kainuma, K. Riahi, and
Thompson, R. L., R. Edwards, J. A. Hart, K. L. Elmore, and J. Weyant, 2011: A special issue on the RCPs. Climatic
P. Markowski, 2003: Close proximity soundings within su- Change, 109, 1–4, https://doi.org/10.1007/s10584-011-0157-y.
percell environments obtained from the Rapid Update Cycle. Viceto, C., M. Marta-Almeida, and A. Rocha, 2017: Future climate
Wea. Forecasting, 18, 1243–1261, https://doi.org/10.1175/1520- change of stability indices for the Iberian Peninsula. Int.
0434(2003)018,1243:CPSWSE.2.0.CO;2. J. Climatol., 37, 4390–4408, https://doi.org/10.1002/joc.5094.
——, C. Mead, and R. Edwards, 2007: Effective storm-relative helicity Virts, K. S., J. M. Wallace, M. L. Hutchins, and R. H. Holzworth,
and bulk shear in supercell thunderstorm environments. Wea. 2013: Highlights of a new ground-based, hourly global light-
Forecasting, 22, 102–115, https://doi.org/10.1175/WAF969.1. ning climatology. Bull. Amer. Meteor. Soc., 94, 1381–1391,
Thorne, P. W., and R. S. Vose, 2010: Reanalyses suitable for https://doi.org/10.1175/BAMS-D-12-00082.1.
characterizing long-term trends. Bull. Amer. Meteor. Soc., 91, Wapler, K., 2013: High-resolution climatology of lightning char-
353–361, https://doi.org/10.1175/2009BAMS2858.1. acteristics within central Europe. Meteor. Atmos. Phys., 122,
Tian, B., D. E. Waliser, E. J. Fetzer, and Y. L. Yung, 2010: Vertical 175–184, https://doi.org/10.1007/s00703-013-0285-1.
moist thermodynamic structure of the Madden–Julian oscilla- Weisman, M. L., and J. B. Klemp, 1982: The dependence of nu-
tion in atmospheric infrared sounder retrievals: An update merically simulated convective storms on vertical wind shear
and a comparison to ECMWF Interim Re-Analysis. Mon. Wea. and buoyancy. Mon. Wea. Rev., 110, 504–520, https://doi.org/
Rev., 138, 4576–4582, https://doi.org/10.1175/2010MWR3486.1. 10.1175/1520-0493(1982)110,0504:TDONSC.2.0.CO;2.
Trapp, R. J., N. S. Diffenbaugh, H. E. Brooks, M. E. Baldwin, Westermayer, A. T., P. Groenemeijer, G. Pistotnik, R. Sausen, and
E. D. Robinson, and J. S. Pal, 2007: Changes in severe E. Faust, 2017: Identification of favorable environments for
thunderstorm environment frequency during the 21st century thunderstorms in reanalysis data. Meteor. Z., 26, 59–70, https://
caused by anthropogenically enhanced global radiative doi.org/10.1127/metz/2016/0754.
forcing. Proc. Natl. Acad. Sci. USA, 104, 19 719–19 723, Zwiers, F., and V. Kharin, 1998: Changes in the extremes of the
https://doi.org/10.1073/pnas.0705494104. climate simulated by CCC GCM2 under CO2 doubling.
——, ——, and A. Gluhovsky, 2009: Transient response of severe J. Climate, 11, 2200–2222, https://doi.org/10.1175/1520-
thunderstorm forcing to elevated greenhouse gas concentrations. 0442(1998)011,2200:CITEOT.2.0.CO;2.

Вам также может понравиться