Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
Forecaster
C ontents
5
Introduction
10
15
19
22
32
Simulation Training
A tool for forecasters and staff
34
38
43
48
Cover:
Medicane on 7/11/14,
Crdit Meteo-France
Printed by Mto-France
Editor Will Lang - Met Office
Layout Kirsi Hindstrom - Basic weather Services
Published by Mto-France
COM/CGN/PPN - Trappes
I ntroduction
Dear Colleagues,
Welcome to the 20th edition of The European Forecaster, a newsletter which has always sought to offer
something a little different to academic journals or trade publications, with its focus firmly on the forecasting profession itself. Once more the WGCEF kindly thanks Meteo France for their support in publishing this
edition. We hope it will find a home in forecast offices across Europe, and that our forecasting colleagues
find plenty to interest them within its pages in some of the all-too-brief quieter moments during their operational duties!
As we all know, the modern operational meteorologist is multi-skilled not just a scientist but an expert
decision-maker and communicator who can offer excellent service to our customers. We are involved with
the latest developments in technology and making use of new sources of information. And we understand
that our participation in training, strategic and business matters, and research can be as beneficial as our
traditional operational roles. I am very pleased, therefore, to see that the articles in this edition are equally
wide-ranging in scope. As ever, we have some excellent case-studies, here concerning a freezing rain event,
the Saint Nicolas storm, and the catastrophic Balkan floods of May 2014. We have a summary of current
thinking around bow echo convective phenomena, and presentation of a sophisticated new verification
scheme. There are also articles outlining simulation training for forecasters, and our increasing reliance on
social media.
Looking back on the last twelve months, it is encouraging to see the WGCEF gaining greater influence and
recognition within the European meteorological circles, largely through our involvement in EUMETNET, but
also through the dedication of our members. Increasingly our advice and experience is being sought, and
the cross-border networks we have formed have allowed us to share knowledge, data and discussions
around warnings to the mutual benefit of our organisations and nations, and for Europe as a whole. I am
sure that in the coming years, the modest efforts and resources of the WGCEF will continue to clearly
demonstrate the merits of collaboration in operational meteorology and in continued development of our
forecasting capabilities and skills.
Will Lang
Chair, WGCEF
May 2015
List of Participants
Members: Will Lang (UK, Chair),
Andre-Charles Letestu (Switzerland, Host),
Klaus Baehnke (Germany), Dick Blaauboer
(EUMETNET FP), Christian Csekits (Austria),
Jos Diepeveen (Netherlands), Karen-Helen
Doublet (Norway), Tessy Eiffener
(Luxembourg), Alessandro Fuccello (Italy),
Bruno Gillet-Chaulet (France), Tim Hewson
(ECWMF), Cecilia Karlsson (Finland), Piotr
Manczak (Poland), Janez Markosek (Slovenia), Jean Nemeghaire (Belgium), Lola
Olmeda (Spain), Taimi Paljak (Estonia) Antii Pelkonnen (Finland), Chryssoula Petrou
(Greece), Vida Raliene (Lithuania), Natasa Strelec-Mahovic (Croatia)
Observers and Guests: Daniel Cattani (Switzerland), Pierre Eckert (Switzerland),
Knut-Helge Midtboe (Norway), Marcel van Schaik (Netherlands),
Apologies
Evelyn Cusack (Ireland, Vice-Chair), Bernard Roulet (France), Knut-Jacob Simonsen
(Denmark)
Actions
All: Promote, and increase circulation of, The
European Forecaster.
All: Send presentations to Andre-Charles for the
website
All: Send articles for next newsletter to Will between
January and March 2015
All: Discuss possible topics for next meeting via
email
Will, Evelyn and Jean: Organise next years meeting
Will, Bernard: Produce next issue of The European
Forecaster
Christian: Report back on Storm-Naming task at next
meeting.
Annex A: Summary
of NMS Developments
Andre-Charles (Switzerland)
A new radar has been installed in Vallee
A new web platform has been designed to warn
the public of dangerous conditions, including pollen
forecasts.
Klaus (Germany)
DWD is installing a new Cray HPC
The DWD global model will have 13km resolution
and 90 vertical levels. Within this is nested a
European model, and a 2.2km resolution model
covering Germany.
The European Forecaster
Bruno (France)
ARPEGE is moving to 7km resolution with 105
layers, and a new DA system.
The new PCMT convection scheme has been
delayed.
The domain of AROME is being widened, incorporating an hourly DA cycle.
Staff reductions continue.
New Exec Director is seeking a reorganisation of
forecasting, with a focus on higher quality prediction
processes which require less human input.
Lola (Spain)
The ongoing reorganisation of the service results in
reduction of personnel and greater use of automated
products, based on the Global Forecast Editor (GFE)
approach.
Closer links with media are being developed.
Work is underway to improve marine forecasts.
A renewal of the HPC is expected early in 2015.
Chryssoula (Greece)
The service has financial difficulties, and when forecasters retire they are not being replaced.
Observing stations are being automated, especially
at night, with only the international airports retaining manual observations.
A colour-coded warning system has been developed,
aligned with MeteoAlarm.
The COSMO model is used. Performance can fluctuate, particularly for wind forecasts between the
islands.
Karen (Norway)
They are cooperating with Sweden to run local
AROME model.
There is increased cooperation between forecasters
and R&D, with forecasters now expected to take part
in project work.
There is emphasis on the development of marine
services, especially for the Arctic.
Vida (Lithuania)
No significant staff reductions this year.
A new workstation for aviation has been developed, and general forecasting is moving onto IBL
workstations.
They are moving into a new, better equipped, forecast room.
There has been a decision to join MeteoAlarm.
Janez (Slovenia)
No staff changes in the last year.
There is further automation of the monitoring
network.
A new radar is operational in western Slovenia
Forecasters use 3 hrly ALADIN at 5km along with
EC models.
The Met Service has a developing Twitter presence.
UK (Will)
Meteorologists are being trained in a 5 Facet
Model of modern forecasting which includes
emphasis on customer needs, soft skills and
involvement in development.
UKMO makes extensive and two-way use of social
media.
A 24/7 Space Weather centre has recently become
operational
Antii (Finland)
FMI celebrated 90 years of aviation weather
services this year.
A reorganisation is underway, which should allow
greater cooperation between operations and
research.
Piotr (Poland)
There will be further automation of the observing
network, and this process is becoming faster, with
staff reductions likely.
A new provider for aviation weather services is to
be selected later this year.
During the last storm season, a special team
supported the regional centres.
Jos (Netherlands)
The KNMI reorganisation continues, with a new
financial model and a flatter management structure.
There should be better integration between
research and development.
Staff are used flexibly, depending on the weather
situation.
Tessy (Luxembourg)
There are 22 staff at MeteoLux, but 3 new forecasters will increase that number to 25.
Its funding model has changed, with half provided
by the airport and now half by the state.
A new warnings system has been developed with
stronger emphasis on collaboration with civil protection and on publication of impact-based alerts
which push information towards the media and
public.
Flood warnings are now available on Meteoalarm,
in partnership with the hydrological agency.
Jean (Belgium)
There is a new dual polarisation radar at
Maastricht, and 3 new lidars for aerosol measurement.
There is discussion of longer range forecasts, eg up
to 2 weeks.
Introduction
Forecast errors are inherent elements of the forecasters job. Some are inevitable. Some are unimportant. But some are fraught with consequences
and make a deep impression. Nowadays the synoptic scale is widely under control. Numerical weather
prediction models have become remarkably powerful, for the short range at least. Synoptic elements
then often take all our attention; the mesoscale is
sometimes put aside and remains a challenge.
During the last few years, as a forecaster, I have
been witness to several situations where violent
small scale phenomena had a national impact. In
some cases, these events were not anticipated and
called into question the role of the forecasters. In
this short article, a recent situation is described and
issues raised by the events are discussed.
Figure 1: Thursday 3 July, 12 UTC, Geopotential height and temperature at 500 hPa, ARPEGE analysis.
Scenario:
The situation took place in the summer of 2014. On
Thursday, 3d of July, around midday, a cut-off low
was situated over Iberia (Figure 1). At lower levels, a
barometric marsh was spreading over a large part
of Western Europe with warm air (Figure 2). It was
mostly dry over France at that time except over the
southern regions. Instability developed over Spain in
response to diurnal heating and the presence of cold
air in the mid and upper levels. Showers, thunderstorms and lightning were observed. Because of
southerly flow, clouds and rain were crossing over
the Pyrenees and were reaching the south western
part of France in the evening (Figure 3). This was not
expected - a sunny day had been forecast, an error
that sometimes happens, fortunately in this case
without damaging repercussions!
For the next day, models gave the following
scenario: The cut-off low was predicted to turn to a
dynamic short wave trough, moving quickly northeastwards to the Alpine regions (Figure 4). At lower
levels, pressure was forecast to decrease, flow
speeding up with continual warm advection
Figure 3: Thursday 3 July, 12 UTC, MSG image and lightning strikes.
10 The European Forecaster
11
Right Decision:
This was the right decision to make. Over a wide
part of eastern and south-eastern regions of
France, thunderstorms were numerous and sometimes violent (Figures 10, 11). Various risk criteria
were met: wind gusts more than 100 km/h, hail-
Figure 12: Time series of level of the Nive near Cambo-lesBains with previous records (blue straight lines).
However:
Late Signal:
At this stage, it must be said with strength and
honesty that this episode couldnt have been
predicted the day before. Looking carefully at the
accumulated precipitation charts from different
models available at that time (Figures 14, 15), no
signal emerged to bring to light an exceptional
rainy event on the South-West of France. Rain was
predicted, in the form of showers or thunderstorms, but the amount of precipitation remained
under severe thresholds over this area.
Figure 17: Vigilance watch map base time Friday 4 July, 7 a.m.
13
Conclusion:
To be able to detect when and where models move
away from reality and to be then in a position to
correct the forecast is the primary task of the forecaster. Nevertheless, a diagnostic sometimes is not
obvious. Furthermore, these unexpected situations
correspond to those the forecaster is the least
prepared for, by definition! Luckily, these conditions seem to remain rare. As far as I am
concerned, in the past 15 years, I can remember 4
or 5 circumstances similar to the one described in
this article. And no doubt that the improvements of
numerical weather prediction will make them less
and less frequent in the future. However, they are
important because they can have detrimental
consequences for the forecast team by creating
conflicts between forecasters themselves and
between forecasters and management.
These situations point up some psychological
aspects:
Attention is often focused on a specific area where
troubles are expected. Our mind is less attentive to
what happens elsewhere.
These cases can occur most often at the worst
times, such as at the end of a night shift, when
tiredness is maximal. The ability to assimilate fresh
information, to react in a proper way, and the motivation to trigger a new warning production cycle
are minimal. The paradox is that a forecaster who is
not aware of the situation (open-minded) could be
more reactive!
Introduction
An active low pressure system moved east across the
northern part of the North Sea during 5 and 6 December
2013. Because of this system, a weather alarm was
issued for very strong gusts along our coast and further
inland. This storm also caused water levels to rise along
our coast because of the strong northwesterly winds at
the western flank of the aforementioned low.
In this article I would like to give an overview of the
development of the low pressure area, the model forecasts and the consequences that this storm had for
our country. This will be done from the perspective of a
marine forecaster and therefore water level forecasts
will be mentioned as well. This is also the reason that
this article will start with a short description of the
vulnerability of The Netherlands to high water levels.
Vulnerability
a)
b)
c)
Figure 4 abc
Figures 1 and 2
15
Figure 8
Model Output
Figure 5
Figures
6 and 7
16 The European Forecaster
Figures 9 and 10
Figure 11 and 12
17
Water levels
As mentioned above, water levels started to rise
after the passage of the cold front. The synoptic
charts show a northwesterly flow over a large part of
the North Sea. This caused water levels to rise along
the Dutch coast. ECMWF did calculate the storm
depression in an early stage. That is why water level
calculations showed a peak around the 5 th of
December as well. An example of this water level
calculation can be seen in figure 13. Water levels
are calculated by a water level model (WAQUA) and
all ECMWF members are used as input for this water
level model. As a consequence 52 water level
members are represented. The colored dashed lines
in this graph represent several important water level
thresholds. A video capture of a 3D animation
(figure 14) gives a representation of water levels
along the Dutch coast.
Figure 15
Figure 13
Figure 14
Introduction
Social media complement and increasingly determine the public discourse. By March 2010, 30
million Germans were already members of internet
communities. Maintaining contact with friends and
families, the exchange of common interests and
experiences, and the opportunity to find new
friends are the main reasons for the membership of
users in these communities. Seven social media
sites were among the twenty most visited websites
in Germany (YouTube, Facebook, Wikipedia,
LinkedIn, Twitter, Blogger, Xing). The VZ network
alone had almost 17 millon users, and Facebook
had over 11 million in Germany. Although Twitter
has so far only 350,000 German-speaking active
accounts, it continues to receive high media attention. Among Twitter users an above average number
of opinion leaders were present.
All this, along with the pressure to respond in crisis
situations, urged the German Weather Service (DWD)
to enter the field of modern social media in 2011.
Facebook
With the appearance of modern means of communication such as smartphones, the advent of new
media phenomena such as "shitstorms" as well as
severe weather events, the German Weather Service
decided to play a bigger part in the social media
sector.
Thus DWD started its Facebook page (Fig. 2) on the
web in March 2011.
19
Twitter
Another important social media tool is Twitter.
Twitter is an online social networking service that
enables users to send and read short 140-character messages. The German Weather Service has
participated in this service since 2011. The aim is
to position the public weather service in social
YouTube
YouTube is a very important social platform.
YouTube is a video portal of Google. Users can view
and upload free of charge video clips on this portal.
With the installation of a professional television
studio in the headquarters of the German Weather
Service in Offenbach (Fig. 5) it was possible to
Figure 5: TV-studio
at DWD-Headquarters
Figure 3: Twitter_presse
Figure 5:
DWD on YouTube
Figure 4: Twitter_agar
20 The European Forecaster
Flickr
Flickr is a commercial web-service portal that allows
users to upload digital and digitized pictures with
comments and notes to the site and make it available to other users.
DWDs Flickr-site (Fig. 7) is supplied with interesting
weather pictures by the PR department and has
about 69 000 followers. The aim is also to position
the DWD in social media.
References
Leitfaden Social Media, BITKOM 2010,
Bundesverband Informationswirtschaft,
Telekommunikation und neue Medien
e. V., Albrechtstrae 10 A, 10117 Berlin-Mitte
www.facebook.com/DeutscherWetterdienst
www.twitter.com/dwd_presse
https://twitter.com/dwd_agrar
www.youtube.com/DWDderWetterdienst
www.flickr.com/deutscherwetterdienst
Summary
The DWD is widely represented in the field of social
media. This had already started with a daily
newsletter with interesting, current meteorological
topics in 1994, but later reinforced with Facebook,
The European Forecaster
21
Bow Echoes :
Conceptual Schemes and European Relevance
Laurent Goulet, Meteo France
Introduction
Convective systems are traditionally classified
according to three categories: single cells, multicells and supercells (Weisman and Klemp, 1982).
These features can be discriminated in various
ways, such as by their complexity, the environment
in which they develop, or their consequences. For
example, amongst the three categories, supercells
are probably the most complex.
However, convective reality is more complex than
prescribed by this classification. For instance, the
multicell class encompasses convective systems
whose effects can be very different. For example,
squall lines, which belong to this category, are very
mobile systems providing potentially strong gusts.
On the other hand, multicells can be also stationary, as regularly observed in Mediterranean regions,
instead giving significant rain accumulation.
Figure 1 : COMET illustration of typical bow echo life cycle, as first described
by Fujita (1978). One can notice three main steps :
the Echo (A), Bow Echo (B and and C), and Comma Echo (D and E) phases.
The first topic concerns the optimal environment favouring such a phenomenon
(Johns and Hirt, 1987; Evans and Doswell,
2001). The second one is about mechanisms which are responsible for the
strongest winds. In this respect, great attention has been paid to mesoscale vortices
related to bow echoes life cycle (Weisman,
1993; Atkins et al, 2005; Wakimoto et al,
2006).
Figure 2 :
Radar
reflectivities
from the French
radar network
ARAMIS, at
2245 utc (a),
0015 utc (b),
0100 utc (c).
This sequence
shows typical
life cycle of a
bow echo from
a linear squall
line (a) to a
comma shape
echo (c). The
bow echo
phase seems to
be consecutive
to merging of
the squall line
with the
convective
system number
2, off the
Pyrnes
Orientales.
23
(a) RIJ : the pressure gradient paradigm (from Lafore and Moncrieff, 1989)
Mechanisms
a. The RKW Paradigm
The cold pool is very prominent inside bow echoes.
Its boundaries are a site of strong baroclinic horizontal vorticity. In the RKW theory (Rotunno,
Klemp and Weisman, 1988), this baroclinic vorticity
has to be more or less balanced by the environmental one (associated to wind shear). According to this
condition, ascending motions are more upright and
stronger. More severe and more durable storms are
favoured.
b. Mesovortices Genesis
One important bow echo attribute is its mesovortices. These have two types: 1) the midlevel bookend
vortices, which appear at the ends of the convective
line; 2) the low level vortices which develop along
the leading edge of the storm system.
Mechanisms proposed in various studies are generally based on the tilting of horizontal crosswise
vorticity (Figure 4). Horizontal vorticity is most of the
time baroclinically generated along the cold pool
boundary. However, some of the genesis of the
midlevel bookend vortices, could alternatively, at
the beginning, be the result of environmental wind
shear.
25
European Relevance
Bow echoes were first identified in the USA (Nolan,
1959; Fujita, 1978). In Europe, the first studies
concerning bow echoes date back to the mid-nineties
(eg, Ramis et al, 1997). If interest remains less than
in USA, various studies show that no European country is immune to this phenomenon (Figure 7). Indeed,
bow echoes have been observed in Scandinavia
(Punkka et al, 2006), in Central Europe (Tuschy,
2009; Gatzen, 2004; Schmid et al, 2000), in Great
Britain (Clark, 2007; Clark et al, 2014), and in
Southern Europe (Ramis et al, 1997). Moreover, one
can identify typical American features - mesovortices, tornadoes and damaging winds and the RIJ.
Consequently, European bow echoes seem similar to
their American counterparts.
Figure 6: Meso high surge. One intriguing companion phenomenom of bowing segment apparition (Adams-Seilin et al, 2010). This
consists in a sharp pressure increase ahead of the linear convective system, just before a bowing step. Meso high surge implies winds rotation.
This one could help bow echo development.
26 The European Forecaster
Meteorological
Ingredients (from
Johns and Hirt, 1987;
Evans and Doswell,
2001; Burke and
Schultze, 2004; Cohen
et al, 2007)
Bow echoes can emerge all year
round. Of course, they are more
frequently observed during the
warm season from May to
August. However, this already
shows that bow echoes can
develop in various environments.
(e) Finland
(f) France
-1
-1
-1
-1
27
Parameter
Warm period
Cold season
CAPE
MUST BE HIGHT
LOW
DCAPE
HIGHTER
LOW
Mean flow
WEAKER
MUST BE STRONG
(0-6 km)
(20/30 kt)
(45/55 kt)
Deep shear
WEAKER
MUST BE STRONG
(0-6 km)
(20/30 kt)
(45/55 kt)
CAN BE WEAK
MUST BE STRONG
(2500 J/kg)
Parameter
DERECHO
Discriminator (Strong Bow Echoes)
Warm Season
CAPE
NO
HIGH
DCAPE
NO
HIGH
YES
-7.3C/km
20/30 KT
YES
YES
YES
VERY WEAK
MCS SPEED
YES
YES
YES
YES
Concluding Remarks
and Bow Echo Monitoring
Figure 9:
Simulated
refelectivity
produced by (a)
AROME, on July
19th, 2011 over
Southern France;
(b) WRF-ARW on
May 8th, 2009
over Missouri
(USA).
29
References
(c) Bow Echo
case :
10 December
2000, temporal
evolution
of 2 m
Temperature,
Mean Sea
Level Pressure
and 10 m Gust.
31
Simulation Training
A tool for forecasters and staff
Jos Diepeveen, Heleen ter Pelkwijk, Frans Debie, John Kambeel, KNMI
Introduction
Simulation is extensively used for educational
purposes, most frequently by way of adaptive hypermedia. It is often used in the training of civilian and
military personnel. This usually occurs when it is
prohibitively expensive or simply too dangerous to
allow trainees to use the real equipment in the real
world. In such situations they spend time learning
valuable lessons in a "safe" virtual environment yet
living a lifelike experience (or at least this is the
goal). Often the advantage is to permit mistakes
during training for a safety-critical system.
Simulation training can also be useful in the case of
forecasting the weather.
Set up
In the schematic below the set up of the simulator
is simply shown. To maintain pressure during the
training, time runs twice as fast as in real life .
Higher speeds are also possible but have not
The Concept
The learning process can be divided
into three phases:
1) The Education phase: gaining
knowledge
2) The Training phase: transforming
knowledge into skills
3) The Practice phase: applying and
reinforcing skills in near real-time
situations under a lot of pressure
Obviously, the third stage is a
phase in which simulation training
could be helpful. At KNMI a wish
had risen to examine all three
stages of the learning process.
Previously, education had tended
to get stuck in phase 1 and 2. This
article will briefly discuss the third
phase.
With use of a simulator, the goal is
to examine two aspects of forecasters work, namely the meteorological
Technical details
The Meteo-Simulator at KNMI is an update of
Cloudy Camel (Pelkwijk, Higgings, Mills)
Initially it was designed specifically for KNMI, but
now it is also usable for other institutes. It is based
on php/jquery&javascript and adaption to php
requires a webserver (local or company webserver),
some of which can be taken freely from the internet.
Experiences
At KNMI we have developed experience in simulation training over the last few years. In 2014 the
training was evaluated with the use of the newly
designed Meteo Simulator.
The chosen meteorological situation was not a trivial case, the main issues were:
- Model analysis of the position of a low was wrong
(about 30-40 nm, image 3)
- Winds and precipitation prediction were very
uncertain. (image 3)
- A chemical accident had occurred in an industrial
area with questions about the dispersion with an
uncertain wind-forecast.
Sources/references:
The subjects for training were shift leaders and
senior forecasters:
http://en.wikipedia.org/wiki/Simulation#Simulation_
in_education_and_training
33
Introduction
Since 1985, MeteoSwiss has used a global score for
systematically assessing the basic weather forecasts issued by the regional forecasting centres.
This assessment is done for two main reasons.
Firstly, it is used for administrative purposes as the
weather centres are expected to communicate in a
simple way to the general public and to the government the evolution of the quality of their forecasts.
On the other hand, the forecasters need to know
the performance of their predictions in order to
improve them. In 2013, we developed a new verification scheme, called COMFORT (for COntinuous
MeteoSwiss FORecast qualiTy), which also accounts
for benefits from the evolution of the forecasting
system as well as of the present automated observation networks.
COMFORT is a global measure of accuracy which
verifies deterministic forecasts of quantities representing sensible weather in Switzerland, namely:
precipitation (without distinction of its type),
sunshine duration, minimum and maximum
temperatures, and wind speed. Specifically,
COMFORT assesses the generic forecasts which are edited numerically by
the forecasters. These forecasts serve
then as a basis for generating a variety
of products, ranging from web apps to
agriculture or support for TV broadcasts.
A requirement that COMFORT had
ideally to fulfil was to encode in a
single value the general forecast quality, together with the capability to
provide intelligible explanation for a
high/low global score, typically
computed over a long period and over
a large territory, to people that are
neither experts in verification, nor forecasters. A way of conciliating these
conflicting requirements was to make
it possible to focus on specific periods
and/or geographical areas in order to
34 The European Forecaster
detect and analyse forecasts whose accuracy deviates from the average. Also, forecasts for all timeranges are verified using the same spatial and
temporal resolutions, which allows comparison
across different lead times. In parallel, COMFORT
can be applied to NWP forecasts, typically the
First Guess predictions which initialize the forecast editing tool used by MeteoSwiss bench forecasters, making it possible to measure forecasters
added value with respect to NWP.
This approach explained above can be applied independently to each verified quantity, which in our
case are:
Verification Principles
As mentioned in the introduction, we consider deterministic forecasts only. For any verified quantity, the
forecasts accuracy is split into three qualifications :
correct, useful and useless. These categories are
defined by two thresholds that should be seen as
tunable parameters which depend on the verification
context. Both thresholds are defined once for all
when setting up the verification framework. The first
threshold defines a tolerance interval around the
forecast value. This threshold should be seen as an
estimation of the maximum error below which a forecast is assumed as completely correct. The second
threshold is the maximum error beyond which the
forecast is considered too erroneous to be of any
value, and defines the utility interval around the
forecast value. Deciding whether a forecast is correct
or useless remains largely subjective and depends
on the verification context. For instance, the thresh-
Figure 2: Behaviour
of the partial score for
precipitation with respect
to the observation, for
three different values
of the forecast;
3, 8 and 20 mm.
35
Tests
A significant piece of the work related
to the development of the COMFORT
score was devoted to simulations
with the aim of testing with real data
different properties of the score such
as its spatial and temporal variability,
its sensitivity to perturbations of
different kinds, its ability to reflect
theoretical enhancements to the forecasts, and its robustness against
hedging. Each quantity involved in
the verification was considered sepa Figure 3: Example of daily analysis with the scores for precipitation, sunshine
rately. The tested forecasts were
duration, temperature minimum and maximum and wind for the forecaster (VAL)
predictions edited by forecasters,
and first guess (FG). In the centre is the score for each parameter and underneath
First Guess forecasts obtained from
the global score.
different numerical outputs, and in
addition different reference forecasts:
The tuning of the parameters and in each partial persistence for temperatures and various poorscore can be made following different approaches, man predictions for relative sunshine and precipidepending on the verification context. For instance, tation.
in a customer-oriented system, thresholds might be
imposed by each specific client according to their The robustness of the score against hedging was
requirements. The thresholds that we have fixed for tested by considering different no-skill or no
our verification purposes are mostly empirical and risk forecasts in order to check that there is no
try to represent, for each verified parameter, reason- obvious systematic way of obtaining better longable estimations of what a correct, useful or useless term results, at least for short-range predictions, by
forecast for the general public is. Also, we have forecasting some predefined scheme rather than
made the choice of setting the same thresholds for best-judgement.
all regions in Switzerland as this allows easier explanation and comparison of the forecast accuracy from Simulations were also made with the aim of estimating COMFORTs sensitivity to different theoretione region to another.
cal forecast enhancements. On one hand, this was
The weights i, which should always sum to 1, repre- useful to answer a question asked by the
sent the relative importance of each verified quantity MeteoSwiss leadership when fixing quantitative
in the global score and can also be adjusted according to the verification context. We give a similar
weight to all verified parameters except for wind: 0.3
for precipitation, 0.3 for sunshine, 0.15 for Tmin as
well as for Tmax, and 0.1 for wind. The main reason
for setting such a smaller weight for wind is the difficulty of having representative observations especially in mountainous regions which prevail in the
country. We thus have made the choice of verifying
wind speed only at selected stations which capture
Figure 4: Quarterly score communicated to the government. Solid
the dominant winds blowing in Switzerland. For
lines represents the global score performed by the forecaster within
countries with larger flatlands or coastlines, where the 3 weather centres in Switzerland, and dashed line represent the
measures might be more representative of the global to achieve for 3 forecast ranges.
36 The European Forecaster
Communication
Every quarter, the global score COMFORT obtained
by three administrative regions for the elapsed period is communicated to the leadership and to the
government, allowing them to monitor the overall
evolution of the forecast quality. Fig 4 shows the
quarterly evolution of the score for day +1, day +3
and day +5; the dashed lines represent the goal
fixed by the government.
In parallel, regular feedback in the form of daily
bulletins verifying in greater detail individual forecasts is automatically delivered to forecasters (see
Fig 3). These bulletins show the partial scores of the
forecasters predictions (VAL) as well as scores
achieved by the NWP First Guess forecast (FG) for
all forecast regions and for a given day. In particular, this allows forecasters to see what value they
added to NWP on a concrete occasion. For each verified quantity, mean values for the whole responsibility region are provided, as well as a global score
for that day.
Reference
Sideris, I. V., Gabella, M., Erdin, R., and Germann, U.
(2011). Real-time radar-raingauge merging using
spatiotemporal co-kriging with external drift in the
alpine terrain of Switzerland. Quarterly Journal of
the Royal Meteorological Society, 00:122.
Letestu A.-C. The New Production Process at
MeteoSwiss. The European forecaster, 15, may
2010.
Cattani, D., Faes A., Giroud Gaillard M., Matter M.,
Global Forecast Quality Score for Administrative
Purposes. To appear in MAUSAM (July 2015)
37
Introduction
Synoptic Situation
Figure 1: 500hPa geopotential and MSL pressure field over Europe and North Atlantic (left) and 850hPa temperature field (right) on 30 and
31 January 2014. Source: www.wetterzentrale.de
38 The European Forecaster
39
Issued Warnings
Due to heavy precipitation predicted by different NWP
models (Figure 7 all showing the 72 hours accumulation in the time period 31 January 00 UTC to 3
February 00 UTC), the NMS issued many severe
weather warnings starting with the first warning on
the morning of 30 January. In the first warning only
heavy snowfall for some regions was mentioned, but
on the evening of the same day, severe freezing rain
was added into the warning text. In the following
days severe weather warnings were issued twice a
day and, in accordance to the text, the appropriate
colour on the Meteoalarm map was used. The most
severe situation was predicted for Sunday, 2 February
2014, when for all five regions in Slovenia the highest
level of weather warnings were declared (Figure 8).
Figure 11: Damage in forests and on power and railway infrastructure, Feb. 2014.
Figure 10: Freezing rain in Postojna 2 Feb. 2014.
The railway infrastructure in particular the connection between Ljubljana and the coast (Koper) - was
severely damaged. Pylons of power lines were
broken and for more than a year only diesel locomotives maintained rail transport from and to the Port
of Koper.
The European Forecaster
41
The estimated total damage caused by this long lasting freezing rain was around 400 million EUR. As can
be seen in Figure 12, the major part of Slovenia was
damaged by freezing rain/snow and a small region
in the South by flooding.
Conclusion
References:
1. Meteorological archive of EARS
2. Daily bulletins of Civil protection authorities
3. http://www.wetterzentrale.de
4. Tim Hewson: Current activities at ECMWF related
to severe weather prediction (presentation at WGCEF
meeting in Geneve 2014)
Figure 12: Territory with material damage due to freezing rain or snow (purple) and due to flooding (blue cross-hatching).
Source: Daily Bulletin of Civil Protection Authorities.
Introduction
43
Synoptic Situation
In the period between 12th and 15th May an upperlevel pressure trough was moving from the west of
Europe towards the southeast. The axis of the
trough, and with it also the surface front, moved
across central Europe and the Balkans during the
night of 11th and the morning of 12th May. After the
front passed over central Europe and the Balkan
Peninsula, the cold air from the north of the continent arrived (Figure 6).
Figure 10: Surface air pressure analysis and the position of frontal
systems on 15th May 2014, 0600 UTC (Source: DWD)
45
Figure 13:
a) Aladin forecast
of 24-hour
precipitation for
the period from
06 UTC 15th till 06
UTC 16th May;
b) interpolated
measured
precipitation for
the period from
06 UTC 15th till
06 UTC 16th May
46 The European Forecaster
Conclusion
The flood that occurred in the lower Sava river basin
was a consequence of increased precipitation
between the 11th and 17th May 2014, especially in
the areas of east Croatia, north-west Bosnia and
central Serbia. The cause of the increased precipitation was a strong and stationary cyclone, with a
centre located over the southeast Europe. The flood
was also preceded by a very wet period, so that the
soil was saturated with water in the area of the Sava
river basin. The event was correctly forecast and all
forecasting services issued alerts for high precipitation. However, due to abnormal intensity and
stationarity of the cyclone that brought extreme rainfall (in some places exceeding the average monthly
precipitation amounts by several times), the largescale flood was inevitable.
47