mise en page 1 › newsletter17 › full.pdfminutesofthe2011annualmeeting,bergen, norway,7th october...

The EuropeanForecaster

Newsletter of the WGCEF N° 17 - June 2012

Cover: "Explosion maritime"

Dangerous waves breaking

on the French Riviera

Introduction

Minutes of the 2011 Annual Meeting, Bergen, Norway, 7th October

The Messina Flood of 22nd November 2011

Flood Forecasting for England and Wales

Mesoscale Modelling in the Netherlands

Services of DWD (Deutscher Wetterdienst) for Civil Protection Authoritiesand Media

The Forecasters' Contribution to the Impact Assessment in Weather Warningsat the Royal Meteorological Institute of Belgium (RMIB)

Communications between the NMHS ans Media and Civil Protection in Croatia

What Forecasters Need to Know about Training for Trainers

Experimental daily forecasts of Northern Atlantic Weather Regimes and HeavyPrecipitating Events (HPEs) over Southern France with the Meteo France globalensemble system PEARP

Representatives of the WGCEF

Printed by Météo-FranceEditor Will Lang - Met OfficeLafout Kirsi Hindstrom - Basic weather Services

Published by Météo-FranceDirection commerciale et de la communicationD2C/IMP - TrappesCr

édit:DamienBouic,http://www.db-prods.net/

5

6

10

12

16

22

27

31

34

37

43

C ontents

Dear readers,

Trying to put together a mental picture of striking severe weather events since our 16th edition (May 2011)makes me feel a little confused! The only thing that comes to mind immediately is the 2011/2012 winter inEurope. From late January until early February 2012 we were really in the grip of extreme conditions. FromThe Ukraine, via the Balkan region and Italy, then towards the United Kingdom and the Low countries,extreme low temperatures and heavy snowfall ended a winter that had been relatively mild until then. Manypeople died or were injured in the cold spell. Beyond this, I cannot recall many extremes. Have I forgottenthe weather events of 2011 entirely? That could very well be due to the ‘meteorological memory overflow’from which I necessarily suffer in order survive the vast amount of data that come to forecasters daily ? Or itmight be the case that we really did encounter fewer big weather events during this period. And yes lookingat the internet I feel relieved, with only some relatively small heat waves during the 2011 summer and somemajor rain and river flooding events in France, Italy and Ireland from late October to early November.

Writing this introduction for the 17th edition of the European Forecaster I am well aware that this will be mylast one. During our upcoming meeting a new Chairperson will take on this post and I realise that this willend a long period for me. I have had more than 10 years of direct involvement with the Working Group, andto be honest not much has changed in terms of the ‘spirit’ of the group. Of course within our forecastingbusiness huge progress has been made by the use of ensemble techniques and also from the new non-hydrostatic very high resolution models. Meteoalarm has made us change and improve our warningsystems on the national level. And the way we communicate with our colleagues from Civil Protectionduring severe weather events and the way we make use of New Media is bringing a lot of change. But thespirit within our WGCEF seems to be very stable, devoted and directed towards improving our working abili-ties by sharing best practice examples and by having discussions on topics of common interest.

We consider ourselves as the professional representatives from the community of European weather fore-casts, and Eumetnet, the European network of co-operating weather services, has got an eye on us becauseof that. This can be seen as a major compliment on all the work that has been done and will continue in theWGCEF. For this we will discuss our future relation with Eumetnet during our upcoming meeting in October.

And now it is reading time. This nice edition of the European Forecasters’ newsletter again reflects all thetopics discussed in Bergen during our 2011 meeting. I call on all of you to recommend this edition to yourcolleagues. I also call on all readers to send in new contributions for the next (18th) edition. All articles werereviewed by Will Lang (from UKMO). Bernard Roulet and Météo France made it again possible to present thehigh quality edited and printed edition. Thanks to André-Charles Letestu (Météo Suisse) our websitewww.euroforecaster.org is updated with actual information on the Working Group. Our web archive showsthe previous editions of the Newsletter.

I wish you inspiring reading hours and hope to see you during our next WgCEF meeting in Vilnius, Lithuania,

on Friday 5th October 2012.

The European Forecaster 5

I ntroduction

Frank KroonenbergChairperson of WGCEF

Minutes of the 2011 Annual Meeting, Bergen,Norway, 7th October

Introductions

Karen-Helen opened the meeting and welcomed thegroup to Bergen. Noting the heavy showers outside,she stressed that she could not guarantee theweather, but also that it was very changeable so atleast some fine conditions could be expected, ifonly briefly.

Frank thanked Karen-Helen on behalf of the groupfor her hospitality, and for her and her colleagueVibekke Maeland for organising the meeting. Hethen extended a warm welcome to new (or longabsent) members of the group: Lola, Christian andAlessandro.

The Chairman then recalled his last visit to Bergen,30 years previously, on board the Dutch weathership 'Cumulus'. The ship was usually stationed atposition 'Mike' at 66N, 02E. Bergen had proved tobe a welcome and beautiful sight, and skies hadremained blue during his stay.

Frank passed on the apologies from group membersunable to attend the meeting:

Manfred Kurz, Merike Merilan, Teresa Abrantes,Stefano Micheletti, Tessy Eiffener, ChryssoulaPetrou and Panagiotis Giannopoulos.

The agenda was agreed.

Actions from Last Meeting

These actions – to organise the current meeting, toupdate the website, and to produce the newsletter –had all been completed.

Chairman's Report

Frank noted that the year since the 2010 meeting inDublin had seen some very significant weather-related events. In particular the earthquake andtsunami in Japan had been a concern for manyEuropean National Met Services (NMSs). There hadbeen much collaboration, particularly in the field ofdispersion modelling and forecasting.

The economic situation continues to cause uncer-tainty within many NMSs. At KNMI in the Nether-lands, there is ongoing discussion around further

Present:

Frank Kroonenberg (Netherlands, Chair)Will Lang (UK, Vice Chair)Klaus Baehnke (Germany)Christian Csekits (Austria)Evelyn Cusack (Ireland)Karen-Helen Doublet (Norway)Alessandro Fucello (Italy)Andre-Charles Letestu (Switzerland)Lola Olmeda (Spain)Janez Markosek (Slovenia)Jean Nemeghaire (Belgium)Antti Pelkonen (Finland)Bernard Roulet (France)Knud-Jacob Simonsen (Denmark)Zoran Vakula (Croatia)

The European Forecaster6

privatisation. Commercial activities at KNMI ceasedin 2000. Current civil protection customers remainvery happy with the service they receive, but therecession brings increased pressure to cut costswithin government organisations.

Newsletter and Website

Frank encouraged all members and their colleaguesto contribute to the WGCEF Newsletter. Bernardconfirmed that Meteo France were happy to contin-ue to publish the journal, and it was agreed that thegroup should send a formal message of thanks toMeteo France directors for their support for thegroup over the years. Alessandro noted the impor-tance of maintaining a 'hard copy' version of thenewsletter. It was felt that its appeal to forecasterswould be diminished if it were to become 'on-lineonly' like many other journals.

Andre-Charles presented several options for arevamp of the www.euroforecaster.org. These werewell-received, and a preferred option was chosen.The updated site will be based on website formatscommonly used for blogging, and the group agreedthat this approach would encourage more regularuse and interaction between group members.Evelyn and Knud-Jacob volunteered to assist Andre-Charles in further exploring and implementingoptions for the updated site. Other suggestedimprovements to be investigated by this WGCEFsubgroup would be new logo, and the ability toaccess the site from smartphones.

Andre-Charles, Evelyn and Knud-Jacob agreed toform a WGCEF Website Group and report onprogress at the next meeting.

Round Table Discussion on NHMSDevelopments

Group members were each asked to briefly describesignificant changes to their organisations and theirroles as forecasters.

Will (UK)• The Met Office remains a 'government agency',but has transferred from the Ministry of Defence tothe Department of Business, Innovation and Skills.

• This change will entail a renegotiation of our rela-tionship with military customers.

• There is a current pay and recruitment freeze at

the Met Office as a whole, but despite this we havebeen able to recruit new forecasters.

• The 24/7 operational and computing facilities atthe Met Office have made it a key component ofseveral initiatives to build partnerships betweengovernment and scientific organisations. As anexample, the Natural Hazards Partnership seeks tocombine expertise in many different fields to beable to warn for phenomena as diverse as volcanicash, floods and enhanced solar activity.

Jean (Belgium)• New procedures for forecasting thunderstormshave been developed.

• INCA is being used experimentally.

• Forecasts for Fukushima were produced usingUKMO and Meteo France dispersion models.

Alessandro (Italy)• The Italian Air Force is being reduced in size, sosome organisational change is possible due to areview of civil and defence responsibilities.

Evelyn (Ireland)• 300 people visited Valentia Observatory to cele-brate 150 years of observations.

• Staff numbers have reduced to 190.

• The general and aviation forecasting offices havebeen amalgamated.

• 4 senior forecasters have been lost, so 3 new fore-casters have been recruited with a modelling back-ground.

• Government agencies including Met Eireann arejoining together to form a National EmergencyCentre. A strong forecasting component is required.

Bernard (France)• A new Vigilance procedure has been introducedfor storm surge.

• Aviation forecasting is now centralised inToulouse.

Andre-Charles (Switzerland)• Meteo-Suisse's status is changing.

• Regional offices are being reorganised.

• There is a need to exploit commercial potential, soa new product management department has beencreated.

• 20 jobs are going.



• It has been agreed that data is to be made freelyavailable, though it is still not clear what this meansin practice.

• Meteo-Suisse is the 'Single Official Voice' for themost severe ('Level 4 and 5') weather warnings,which it issues on behalf of the government.

Christian (Austria)• 50% of ZAMG comes from government, but therest must come from commercial activities.

• 2 forecasters have been lost in the past year, withthe number of forecasters in the Vienna officereduced to 12.

• Some scientists have been used for operationalforecasting work, though they may not be availablefor duties at nights and weekends.

Zoran (Croatia)• No major changes so far this year, though theupcoming election may well result in less money forthe met service.

• Staff continue to be lost through retirement, andthey are unable to recruit new staff.

• There is a shortage of scientists and mathemati-cians available to the service.

Antti (Finland)• A new radar (doppler + dual-polarisation) hasbeen installed.

• There are new warnings for water level and waveheight.

• The Finnish government has required that FMI doall aviation observations.

Janoz (Slovenia)• Elections are expected soon, so it is uncertainwhat will happen next year.

• No new meteorologists have been employed.

• A new model is now operational (as described atlast year's meeting).

• There is increased use of first-guess model inputto products, allowing more time for forecasting.

Klaus (Germany)• Staff numbers at DWD are now reduced to 2400.

• 3 regional offices are no longer open at night.

• The AUTOWARN system is due to become fullyoperational in 2015.

• There will be more automated observing.

Knud-Jacob (Denmark)• Danish and Swedish airspace join together from1st January 2012, so the aviation sectors of the twomet services are joining forces.

• DMI has been appointed to provide tsunami warn-ings to Denmark, the Faroes and Greenland.

Karen-Helen (Norway)• The yr.no website continues to be a big success,with 4 million users per week. It has become themain channel for forecasts in Norway.

• There is an ongoing process to ensure that the siteuses the best data, and makes best use of forecast-er intervention where required.

Lola (Spain)• There is a new organisation of forecasting intoforecasting centres.

• New systems, using digital forecasting tech-niques, are enabling increased automation.

Frank (Netherlands)• Only one manual observation remains in theNetherlands, at Schipol Airport.

• A decision on further privatisation is due in 2012.

• 17.5% cuts to the KNMI budget are expected overthe next 4 years.

Presentations

Members then gave presentations on the followingtopics:

“Communication from NHMSs towards Media andCivil Protection”.

“New Developments on Short-Term High ResolutionModel Ensembling and Operational Use ofProbabilistic Values in Short-Range Forecasts”.

Chairperson Nomination for NextMeeting

Having been Chair/Vice-Chair for the group for fouryears, Frank announced his intention to step downbefore the next meeting. Will Lang, current Vice-Chair indicated that he would be glad to serve asChair from next year – an offer which was agreed bythe rest of the group.

The position of Vice-Chair would then becomevacant at the next meeting, so members were


asked to consider proposing themselves for thisrole.

Date and Place of Next Meeting

Karen-Helen and Frank stated that Vida Raliene(Lithuanian) had confirmed her offer to host nextyear's meeting, and the group provisionally agreedto meet in Vilnius on 5th October 2012.

Topics for Next Meeting

The group agreed the following topics for presenta-tions at the 2012 meeting in Vilnius :

• Severe weather case studies.

• Examples of how we can improve our serviceseven given current financial and staffing restraintsin many of our organisations.

Any Other Business

Will thanked the members of the group who hadcontributed to his work on the WMO PWS ServiceDelivery Strategy earlier in the year. Membersprovided examples of service delivery metrics usedin their organisations, several of which were quotedin the final version of the strategy.

Close of Meeting

The Chair and Vice-Chair again thanked the membersfor their participation, wished them well for thecoming year, and formally closed the meeting.

Will Lang18th October 2011



Vorticity and the wind at 300 hPa on November 21st

showed a large anomaly in the dynamic tropopauseover the Strait of Sicily 24 hours later, with ajetstreak further south between Tunisia and Libya(see Fig. 2).

The precipitation forecast by the ECMWF model andCOSMO ME gave a clear signal for very high valuesover small spatial scales equivalent to a maximumof 60 mm (ECMWF) and 220 mm in 12 hours(COSMO ME; see Fig. 3) These values were extreme-ly localized on the Sicilian Ionic coastal area anddid not give similarly strong signals for Calabria.

The evolution of weather conditions on the eveningof Monday 21st and the morning of Tuesday 22nd

confirmed the risk of strong convective activity inSardinia, and the localised nature of precipitationforecast for the island by the models compares wellwith the actual precipitation distribution.

On the 22nd November 2011, intense and persistentrain affected the province of Messina in Sicily -especially the areas along the Tyrrhenian coast -and Calabria in the south of the Italian mainland.The municipalities most affected were Saponara,Villafranca Tirrena, Rometta and Barcellona Pozzo diGotto (Fig.1), with 4 dead and about 700 displacedpersons (data from Protection Civile). Strongconvective precipitation fell in these areas, withhourly precipitation amounts of the order of 60-100mm, and total cumulative values around 150 mm(data from the observational network and the CivilProtection rainfall network).

The event was associated with a so-called V-shapedstorm (named after the shape of the convectivesystem) within a baroclinic area, a deep cycloniccirculation at high altitude, combined with a signifi-cant warm anomaly in the lower layer, activated bystrong sirocco winds. The 500 hPa Potential

The Messina Flood of 22nd November 2011Alessandro Fuccello, Italian Air Force Weather Service

� Figure 1


� Figure 2

On Sicily, however, the shift of the anomaly in thealtitude over the intense southern moist, flowallowed the development of a supercell (V-shaped

storm), which had its southern peak in Messina andspread to Calabria, as is clear from the compositesatellite/radar/lightning image in Fig.4.

� Figure 3

� Figure 4



Summary

In the UK, the organisations responsible for floodprediction and warning have joined forces to deliverjoined-up flood forecasting services to governmentand to emergency responders. In the FloodForecasting Centre, a team of experienced andmulti-skilled operational hydrometeorologists fromthe Met Office and the Environment Agency hasbeen created to aid the design of and to use newmodels, tools and techniques in flood forecastingfor England and Wales.

Background

The summer of 2007 saw unprecedented rainfallacross the UK. 13 people died in the resulting flood-ing, which also caused billions of pounds worth ofdamage to communities, homes and critical infra-structure. The subsequent independent review ofthese events – the Pitt Review - highlighted theneed for a much more integrated response to flood-ing across both national and local government andthe other agencies respon-sible for warning for andresponding to flood risk. Inparticular, the reviewrecognised the importanceof creating a joined-upoperational flood forecast-ing service between theMet Office – responsiblefor weather warnings forthe UK – and the Environment Agency (EA) – respon-sible for flood warning inEngland and Wales1. ThePitt Review also recom-

mended that probabilistic approaches to flood fore-casting should be employed, using warnings withlonger lead times to allow a proportionate and risk-based response.

At the same time, the Met Office was restructuringits Public Weather Service to ensure its alignmentwith the requirements of the UK Civil Contingenciescommunity at both national and local levels. Therewas a move towards use of impact-based warningsand advice, instead of just pure meteorologicalguidance.

The Flood Forecasting Centre (FFC) was created in2009. Now located at the Met Office headquartersin Exeter, the Centre is run and operated by bothMet Office and EA staff and provides a focus for thehydrometeorological forecasting and developmentwithin both organisations. A 24/7 operational rosteris staffed by EA hydrologists who have been trainedas meteorologists, and by Met Office forecasterswho have undertaken flood forecasting training; allare now regarded as hydrometeorologists, who areskilled in both disciplined. They form a highly effec-

Flood Forecasting for England and WalesWill Lang, Flood Forecasting Centre, Met Office, UK

1 - Separate arrangements exist for Scotland and for Northern Ireland. For example, the Met Office works with SEPA in Scotland to deliv-er the Scottish Flood Forecasting Service.

� Figure 1

The Flood Forecasting Centre,based within the Met OfficeOperations Centre


tive communication and decision-making link in theflood forecasting process.

The Role of the Flood ForecastingCentre

Until recently, ultimate responsibility for flood fore-casting as a whole was unclear in the UK. The EAissues Flood Warnings and Alerts for sections ofrivers and coasts, but there was no national-scalemonitoring, modelling or forecasting. The Met Officeissued heavy rainfall warnings, though these wereissued based on accumulation thresholdexceedance rather than the expectation of floodingfrom surface water. In Summer 2007, both sourcesof flooding were significant, and it was unclear whowas responsible for a consistent message.

So a principle strategic aim of the FFC is to provide acentre of expertise and authoritative advice for allsources of flooding; from the rivers, the sea,groundwater, and from 'flash flooding'; associatedwith intense rainfall.

This is an ambitious objective, which has becomereliant on building a very strong partnershipbetween the two organisations involved. And nowthe strength and effectiveness of this partnership is

being used as a model for further and morecomplex collaboration across government andscience bodies, notably the UK’s Natural HazardsPartnership (NHP).

In addition to operational activities, the FFC'sServices team seeks to work across both organisa-tions to co-ordinate development and implementa-tion of new flood forecasting science and systems.Their work has proved invaluable in acting as an'expert customer' which can steer science and tech-nical strategy in the wider EA and Met Office.

Services for EmergencyResponders

Given the diversity of the emergency respondercustomer base, FFC flood risk products are designedto be as simple and useful as possible. The FloodGuidance Statement (fig 2) consists of 'traffic-light'coloured maps of England and Wales for the nextfive days, accompanied by explanatory text and(when necessary) graphics. Each local authority orcounty is assigned a flood risk of either GREEN,YELLOW, AMBER or RED for each day, consistent

� Figure 2

An example Flood Guidance Statement



with NSWWS warnings issued by the Met Office. TheFGS is issued routinely at 1030 local time each day,but its content is under constant review and can beupdated and reissued when ever flood risk isdeemed to have changed significantly. The productis issued in PDF format, either by email directly to acustomer, or to be viewed via 'Hazard Manager', theMet Office's web portal available to emergencyresponders.

Flood risk is calculated using a matrix of likelihoodversus impact (fig 3). Note that there are times inwhich the FGS colour state will not in itself distin-guish between 'low likelihood, high impact' and'high likelihood, low impact' events. In these casesthere is need for careful clarification and communi-cation of the nature of the risk.

Hydrometeorological Forecastingand Impact Assessment

FFC Hydrometeorologists ('Hydromets') are trainedin both operational meteorology and hydrology, andtherefore are able to contribute their expertisethroughout the flood forecasting process. They areoften able to advise the Chief Forecaster on quanti-tative precipitation forecasting, particularly in situa-tions which can give rise to intense or prolongedrainfall. They are also familiar with the mechanismsby which coastal flooding occurs, and are trained toidentify potential storm surge scenarios many daysin advance.

Having influenced the meteorological aspects of theforecasting process, the Hydromets then take

responsibility for determining the likelihood offlooding. A range of tools are available to guidetheir assessment.

For river flooding, the FFC makes use of the Grid-to-Grid (G2G) model developed at the UK's Centre forEcology and Hydrology. G2G uses gridded NWP andobservational data in a grid-based runoff and rout-ing model, allowing 1km resolution river flow fore-casting anywhere in England and Wales, even forungauged catchments. High resolution NWP dataare now available to 5 days ahead, and high resolu-tion ensemble data blended with radar output nowoffers the prospect of reliable probabilistic rainfallforecasts for use in hydrological models.

In the case of coastal flooding, the FFC uses surgeand wave models in conjunction with astronomicaltide predictions to derive changing sea levels at anumber of reference ports. These forecasts are thenused by regional flood forecasting teams to assessthe potential for flooding and the need to activateflood defences.

For of surface water flooding, or for 'flash flood-ing' from rapidly responding small catchments,the FFC looks to combine access to the bestobservational and forecast data with conceptualmodels and with tools calibrated using pastevents. The ability to identify a location particu-larly at risk from very heavy rainfall, and to distin-guish between mere 'heavy rain' situations andthose conducive to intense rain and flooding isparamount – especially when this guidance canbe communicated as soon as possible to localresponders.

� Figure 3


Knowledge the likelihood of extreme rainfall or ofincreased river or sea levels only represents parts ofthe FFCs forecasting process. As with NSWWS, mostFFC forecasts and warnings rely on an assessmentof impact. Prior to an event, there will be extensiveconsultation with regional EA flood forecastingteams and their Met Office counterparts to discussand agree the effects of flooding at a local level.This dialogue is informed by reference to standard-ised definitions of impacts related to factors suchas number of properties affected or by areal extentof disruption.

During a flood event, it can be difficult to make anobjective assessment of impact, so FFC staff under-take real-time monitoring of news reports and socialmedia to accurately gauge the scale of events asthey unfold. This information is stored along withforecast and observational data to aid the post-event review process.

Future Developments

The FFC Development Programme has oversight andinfluence over new model and observational capa-

bilities and aims to exploit them as soon as possi-ble. A challenge in the coming year will be to makebest use of high-resolution ensemble NWP modelsin both river and surface water flood forecasting. Anew wave ensemble capability will soon enhanceour coastal forecasts too.

We aim to further empower our Hydromet team,increasing their expertise and influence so they arerecognised as leaders in the field of hydrometeoro-logy, both in the UK and internationally.

A greater understanding of impacts remains a keygoal for the Centre, with the view to developing amore rigorous and objective method of impactassessment.

Looking further ahead, the FFC is contributing to aJoint Forecasting Strategy between the EA and MetOffice, which looks to further integrate national-scale models and services, and to extend thisjoined-up modelling capability to include multiplenatural hazards.



Mesoscale Modelling in the NetherlandsSander Tijm, KNMI

The installation of a new supercomputer hasenabled KNMI to finally start running the mesoscalemodel HARMONIE on a regular basis. In this articlewe will describe the setup of the system, our firstexperiences with the model in severe weatherconditions and some of the postprocessingdesigned to make optimal use of the possibilities ofthe mesoscale model.

The HARMONIE/AROME MesoscaleModel

The HARMONIE (HIRLAM ALADIN Research onMesoscale Operational NWP In Euromed) modelenvironment is a system which encompasses(amongst others) the AROME model and a scriptenvironment that enables the easy running of themodel of choice. The mesoscale model is run on aLambert grid of 800x800 points and has a resolu-tion of 2.5 km with a timestep of 60 seconds. Themodel is currently driven by the regional modelHIRLAM at the boundaries and uses 3D-Var for data-assimilation. Every three hours a forecast is made to24 hours ahead. In the near future we will startmaking forecasts out to 48 hours, as the users ofthe model like to work with a single model for thewhole forecasting period of 36 to 48 hours.

The main reason for running a high-resolutionmesoscale model like HARMONIE is the ability ofmesoscale models to represent severe convectiveweather well. As the most severe cases with deepconvection over the Netherlands usually developover France and Belgium, these countries have tobe included in the model domain. If the modeldomain is too small, then the convection canalready have some organization when the modeldomain is entered. In the model driving themesoscale model the convection will be parameter-ized and the dynamic state of the convection willnot be present. The convection and the organizationof the convection will then start to build from theboundary, lagging behind the real development andorganization by hours. This has a significant impacton the speed of advance of the convective complexand may cause the modelled convective complex to

lag behind the real complex by hours, as an orga-nized convective system can advance much quickerthan the initial disorganized convection.

One example of such a case is the convectivesystem that crossed the Netherlands on 14 July2010. Very strong gusts associated with thisconvective system caused caravans to be blownover, causing two fatalities on a camping site in thesmall village of Vethuizen. When re-running thiscase on a small domain of 300x300 points themodel showed very strong deep convection (seefigure 1). The comparison with the radar, however,shows that the modelled convection lags behindthe observed one by 3-5 hours. This is caused bythe fact that this convective system already devel-oped over France, outside of the model domain.Therefore the convective system was already insome state of development and organization, whilethe air mass containing the convection entered thedomain through the boundary. But in the model theconvection only started to develop when the unsta-ble air entered the model domain, lagging behindthe real development and organization. And it isthis organization that makes the convective systemdevelop and advance faster than individual cells,which develop initially in the model close to theboundary.

A second experiment, with the southern boundarymoved south by about 500 km, provides a muchbetter fit to the observed precipitation. Figure 2shows the rain, cloud water and cloud ice distribu-tion at the same time as figure 1. It shows that therain in the HARMONIE experiment is still laggingbehind the observed rain, but the distance betweenthe observed and modelled rain is much smallerthan with the small domain, and also the organiza-tion of the precipitation, with a bow echo on themost northeastern part of the convective system, ispresent in the run on the large domain.

The difference is large, especially at the back end ofthe convective system. In figure 1, the build-up ofthe convection starting close to the boundary isclearly visible with convective towers rising slowlyfrom the southern boundary to the North. Figure 2


shows a full-grown and organized system already atthe place where the boundary is in the first experi-ment. It is this organization, - present in the experi-ment on the large domain whereas it clearly is noton the small domain - that shows that you cannot

make the model domain too small, else you maymiss or erroneously represent cases where theconvection is already present in reality at theboundary of the mesoscale model domain.

� Figure 1

Observed radar precipitation and forecast cloud ice (grey), cloud water (yellow) and rain (pink) at 17 UTC on 14 July 2010 for the300x300 points experiment. The plot comes from a 3D visualization tool. The southern boundary is situated on the bottom leftcorner of the image

� Figure 2

Observed radar precipitation and forecast cloud ice (grey), cloud water (yellow) and rain (pink) at 17 UTC on 14 July 2010 for the500x500 points experiment.



First OperationalExperiencesWith HARMONIE

From 7 December 2011 HARMONIE has beenrunning on a regular schedule with 24-hour fore-casts every three hours. During this time a few inter-esting situations have arisen, showing the potentialof the model and the additional value for the fore-casts of KNMI. Three of these situations involvedthe passage of a cold front with line convection,something that usually happens once or twice everyyear, but in this case occurred three times in onemonth.

Line convection is very interesting as it producesa very narrow band with intense precipitation thatcan be accompanied by strong wind gusts. In thisarticle we describe one of these cases, 3 January2012. On this day a deep cyclone moved overScotland and the northern part of the North Sea tothe East (see figure 3). The cold front associatedwith this system passed the Netherlands between13 and 17 UTC. A narrow band with stronglyforced convection, the line convection, was clear-ly visible in the radar images over the UK. It wasforecast by HARMONIE to break up when itapproached the Netherlands, but it was still intactat 13 UTC.

However, between 13 and 14 UTC it did start tobreak up and figure 4 shows the situation around14 UTC. The line convection, or what is left of it,is si tuated over the Nor thwest of theNetherlands. Some extreme wind gusts wereobserved on this system with maximum gusts of94 knots at the Dutch West coast (IJmuiden) and80 knots observed at one of the Wadden Isles(Vlieland).

HARMONIE forecast the breakup of the line convec-tion quite well in a qualitative way, and the modelalso forecast that the strongest gusts would beassociated with the line convection that was break-ing up. But the maximum gusts were forecast to beclose to 70 knots, well below the maximumobserved gusts of 80 and 94 knots. These extremegusts were probably very localised, as these gustsare among the strongest ever reported in theNetherlands and widespread significant damagewas not reported.

One of the issues with the output of mesoscalemodels is the very small-scale features that arepresent in these runs. These are so small that it issometimes hard to distinguish in the plots that areused in the forecasting office, even when zoomed inon a small country like the Netherlands. Thereforewe now also have another way of plotting the windgust forecasts. For this we divide the country into

three areas, water(sea), the coastal arearanging from the coast-line to approximately50 km inland, and theinland area. Then weplot the maximuminstantaneous gust ateach output time stepfor these areas. Tocompare with theobservations we alsoplot all the observa-tions, color coded inthe same way as in thethree areas.

Figure 5 shows theforecast of the maxi-mum instantaneous

� Figure 3

HARMONIE rain rate on 3January 2012 at 12 UTC


wind gust based on the forecast from 03 UTC on 3January 2012. It shows that the maximum gust isforecast over the sea (close to the coast, not shownhere) around 13 UTC (03 UTC + 10 hours). In theobservations it is shown that the maximum gustsare observed on 13.30 and 14.00 UTC, so aboutone hour later than was forecast.

Forecasting Lightning IntensityWith HARMONIEThe hydrometeors in HARMONIE, prognostic rain,snow and graupel, which are not present in HIRLAM,have the advantage that new forecasting tools canbe developed based on these parameters. Until

� Figure 4

HARMONIE +14h forecast of precipitation intensity (left) valid at 14 UTC and the observed radar image (right) at 13.55 UTC (14.55 localtime) on 3 January 2012

� Figure 5

Maximum wind gust forecast by HARMONIE in the 03 UTC run on 3 January 2012 for the sea points(blue), coastal points (red) and the points over land (green) in and close to the Netherlands. Alsoshown are the warning criteria (yellow, orange, which is the weather alert criterion inland, and red,the weather alert criterion at the coast)



now the forecasting of, for example, the chance oflightning or the chance of reaching a certain light-ning intensity threshold was based on statisticalmethods that needed a significant number of casesto get the correct probabilities. Also, these methodshad the drawback that they included the modelerrors (differences in timing and placement ofconvection), meaning that the maximum thresholdsfor which the forecasts could be made was less thanwhat was needed. In the Netherlands the weatheralert criterion for lightning is 500 discharges in 5minutes over an area of 50x50 km2, or smaller, butthe statistical methods can only be derived for light-ning thresholds of 200 discharges in an area of60x90 km2.

The addition of graupel to the prognostic modelparameters enables us to use this parameter, whichis one of the most important factors responsible forcharge separation in thunderstorms, to make fore-casts of lightning intensity. Eight cases with thun-derstorms were used to find a relation between thelightning intensity and the graupel. Four of thesecases were with very intense thunderstorms andfour with limited lightning intensity. For all thesecases a whole day was used in deriving the relation,so periods with no or only a small amount of grau-pel are included in the derivation of the relation.

By first trying to find a relation between the totalgraupel in the model over the lightning observationarea, and then trying to make it applicable on thesmaller areas that are used in the issue of weatheralert, we were able to find a relation between theforecast graupel and the lightning intensity forareas of 25x25 km2. In this derivation we haveexcluded the cases where there was a clear differ-ence between the model and the observations andtried to account for changes in phase. By doing thisa very clear relation can be found between the verti-cal integrated graupel and the lightning intensity.

Figure 7 shows the result of this relation for thecase of 30 April 2012. On this day warm air waspresent over the Netherlands, Belgium andGermany, causing the destabilization of the atmos-phere. Thunderstorms, developing on a cold frontover France and Germany, moved into Belgium andthe first very intense thunderstorms of the convec-tive season 2012 started to form. The model fore-cast these thunderstorms in more or less the rightplace and time, indicating that the lightning intensi-ty could be as high as 150 discharges per 5 minutesin a 25x25 km2 area.

Note the method that is used here is a ‘perfect prog’method. This means that it does not take into

� Figure 6

Observed gusts at all Dutch stations from 00 to 24 UTC on 3 January 2012 for the sea stations (blue,black dots for the stations on the Northern North Sea), coastal stations (red) and inland stations(green)


account the cases where the model convection fore-cast is bad (where showers develop in reality butnot in the model, or the other way around). So whenthe forecast of convection is bad, this method willgive a bad lightning forecast, whereas a statisticalmethod will still give some signal that should point

� Figure 7

Lightning intensity forecast from 00 UTC on 30 April 2012, valid time 19 UTC, based on the vertically integrated graupel inHARMONIE

in the right direction. One other issue that has to beraised is the fact that the relation is only valid forHARMONIE. The method can be used for any model(it was first used on WRF) but our experiencesuggests that the graupel is very differently repre-sented in the different mesoscale models.



Introduction

While the Deutscher Wetter-dienst (DWD) had over 3000members of staff just after unifi-cation with the weather serviceof the German DemocraticRepublic in 1993, it continuedto contribute to the ongoing,nation-wide process to reducecosts of the civil service whilestill achieving a high level ofefficiency. The main challengesin this context were: concentra-tion on core tasks, greater useof the opportunities provided by information tech-nology as well as process optimisation and qualitymanagement. At the same time competition withprivate weather companies increased more andmore. This lead to a reorganisation of DWD’s weath-er forecasting services, especially in collaborationwith civil protection authorities and the media afterthe year 2000.

Organisation of the ForecastingService at DWD

Currently weather forecasting at DWD is done at theCentral Forecasting Office in Offenbach and at 6Regional Forecasting Centres (Fig 1) throughoutGermany.

The forecasting process is guidedand co-ordinated by meteorologists

at the Central Forecasting Office in Offenbach. Basedon numerical products and breaking the forecastingprocess in to three time stages - medium range, shortrange and nowcasting - the the Offenbach meteorolo-gists work with the Regional Offices to achieve a‘Single Voice Policy’ for the forecasts. This appliesespecially to early warnings (lead time 7 – 2 days),prewarnings (48 – 12 hrs) and actual weather warn-ings (Fig 2).

During the reorganization of the forecasting serviceafter 2004 the main objectives were newly rede-fined, while DWD concentrated on its core tasks.This meant the forecast and warning business areawas restricted to its main target customers. Theywere redefined into 7 groups (Fig 3). Thesecustomers should be supplied with forecasts andwarnings by a state-owned authority in a reliableway based on the principle of protection of life andproperty.

Service for Civil ProtectionAuthorities

Weather warnings are issued at regional forecastingoffices guided and co-ordinated by the Central

Services of DWD (Deutscher Wetterdienst)

for Civil Protection Authorities and MediaKlaus Bähnke, Deutscher Wetterdienst/Offenbach

� Figure 1

� Figure 2


� Figure 3

� Figure 4

� Figure 5



Forecasting Office for the main target customers likefire brigades, technical relief organization (THW) et.al. In case of severe weather warnings, forecasts aredisseminated to the Operational Centre of the feder-al government as well as to the joint informationheadquarters (GMLZ) (Fig 4).

DWD has two main tools for the dissemination ofweather warnings: unrestricted information bywww.DWD.www especially for the general public,and FeWIS (FeuerwehrInformatinsSystem) for firebrigades (Fig 5).

FeWIS supplies a closed circuit group with onlineweather information. Users get information about thepresent weather status (Fig 6) over the whole ofGermany. Fire brigades in bigger cities can be provid-ed with their area warning status (Fig 7) broken downinto smaller districts.

In cases of severe weather warnings caused byheavy convective thunder and showers, customershave access to DWD’s radartracking-system KonRaD(fig 8).

Disaster management authorities are assisted by aspecial dispersion model in case of dangerouschemical accidents (Fig 9).

During droughts and when there is a danger offorest fires, authorities are provided with the resultsof a forest fire model (Fig 10).

DWD arranges trainings and exercises for customersand authorities. User conferences and workshopsas well as feedback meetings are performed follow-ing very extreme weather incidents like the “Kyrill”-storm.

Currently FeWIS is the most important severe weath-er information tool for disaster management inGermany. In recent years the number of users hasrisen to over 1450.

Service for Media

With restriction to its core tasks in 2004, DeutscherWetterdienst decided to issue only standard fore-casts for the media. This meant the production onlyof basic numerical forecasts, basic weather chartsas well as occasional radio forecasts and interviews(Fig 11).

Since then DWD only delivers those media forecaststhat usually can be expected by a national weather

service. This means products like standard nationalweather forecasts, a daily “Top Story of the Day”, anewsletter, special media information in case ofsevere weather and a statement on Facebook (Fig12). These information services are freely disseminat-ed via the internet on www.DWD.www. Oral state-ments are only provided on special request but notregularly. Forecasters support DWD’s public relationand press department on request (Fig 13).

Summary

• Deutscher Wetterdienst provides weather forecastproducts for its key accounts, i.e. civil protectionauthorities, fire brigades et. al. The main dissemina-tion tools are the website www.DWD.www and theclosed use circuit online tool FeWIS.

• Media services are issued on a limited basic level.

• Value added forecasts are not produced regularly.


� Figure 7

� Figure 8

� Figure 9

� Figure 6



� Figure 10

� Figure 11

� Figure 12

� Figure 13


The Forecasters' Contribution to the ImpactAssessment in Weather Warnings at the RoyalMeteorological Institute of Belgium (RMIB)Jean Neméghaire, assistant and forecaster at the RMIB, Belgiumwith the cooperation of Karim Hamid, forecaster

Foreword

At the last WGCEF meeting in Bergen, three issueswere been presented concerning Belgium. Firstly,outputs from the short-term high-resolution ensem-ble GLAMEPS (0) were verified against the ECMWFand Aladin model outputs. Also, it was reportedhow efforts are made to contribute to severe weath-er management communications with media andCivil Protection through a federal Crisis Centre.

In this newsletter we discuss the third issuepresented : the short range weather warningsissued by our forecasters.

Introduction

The severity level of weather warnings relies on apotential impact matrix which is illustrated here forsome frequent weather events. The warning levelsare determined by forecasters and produced using acolour and associated symbol(s). These warningsare issued in alignment with Meteoalarm objectivesand criteria, using four colours to identify anddescribe the severity of the potential impact ofweather events.

This potential impact estimated by forecastersdepends both on specified quantitative and/or quali-tative thresholds and on an evaluation of the risk toreach or exceed these thresholds. Some potentialimpact matrices built on thresholds (columns) andrisk assessment (rows) are presented below.

For winds (gusts), snow/ice, freezing rain, andrainfall amounts associated to severe weatherevents, quantitative thresholds based on meanreturn periods have been defined.

For thunderstorm warnings the thresholds are most-ly qualitative and the risk is evaluated using differ-ent tools such the probability scores derived usingan interactive checklist focused on the potential fordevelopment of convective weather events. Moredetails on the interactive checklist currently based

only on one regional model (ETA) can be providedby the author on request. Furthermore other dataare used like the analysis and forecasts of meteoro-logical fields issued from numerical models. SAFNowcasting products (cloud, precipitation andrapidly developing thunderstorm products), satel-lite images derived from RGB compositions (likethe severe convection product), radar images andthe lightning detection system (SAFIR) are mostlyinterpreted for nowcasting (very short range fore-casts up to 1 or 3 hours ahead).

For cold spells and heatwaves, threshold values forsurface temperatures (minima and maxima) and thenumber of consecutive days of an event are calcu-lated to assess the potential impact. For heatwaves,the weather criteria along with an ozone concentra-tion above a critical value are required to issue ared level warning in co-operation with the Ministryof Public Health.

All weather warnings issued by forecasters andreported here have a validity period of up to 48hours ahead. They are forecast for the whole coun-try and/or on a regional basis in accordance withnine Belgian provinces and a coastal area. For theBelgian maritime area only gale or storm warningsare currently issued.

Below we focus on the assessment of the potentialimpact for a selection of weather events forecast atshort ranges. The methodology takes advantage ofthe meteorological expertise of forecasters usingtheir analysis and their assessment of the mostprobable scenario. Furthermore a fine-tuning of thewarning level(s) tends to take human and economicfactors into account.

Example 1. WIND WARNINGS

The potential impact matrix below is based on gustforecasts. Columns show gust thresholds and rowsthe estimated risk levels. The icons with symbolsand colours represent the potential impact levelassessed by forecasters.



The risk level is evaluated as a complex function ofseveral factors:

• Meteorological factors – the particular weathersituation, the scale and location of the weathersystems involved, and the effects of these factorson gusts. The gust forecasts are derived fromseveral NWP deterministic models (a poor man'sensemble).

• Surface and soil factors - including surface rough-ness, soil moisture and vegetation type.

• Human and economic factors – including popula-tion density, economic sensitivity, time of day,week or year, traffic, or leisure activities.

The summer season is defined as when a largeamount of trees are still in leaf. The thresholds are


adjusted and there are five columns for the summerseason. Otherwise the risk level is evaluated as inwinter.

DECISION to forecast the potentialIMPACT level

The potential impact level depends both on theforecast gusts reaching or exceeding the thresholdvalues AND on the evaluation of the risk level.Different threshold values are provided in the twoprevious tables, respectively for the winter andsummer seasons.

The impact level is then tuned by forecasters:

A/ the level is increased when

• the timing of the expected maximum gusts corre-sponds to peak hours (morning/late afternoon orearly evening),

• the location corresponds to densely populatedareas,

• people are unused to and unprepared for stormevents in the location forecast,

• a long duration storm event is expected (eg due tosuccessive troughs or frontal waves),

• the gusts are associated with unstable air.

B/ the level is decreased when

• gusts are forecast in the coastal area.

Example 2. RAINFALL WARNINGS

The potential impact matrix is based on precipita-tion amounts forecasts with thresholds for a show-ery regime (high rates but a shorter period given byamount in mm/6h) not associated with thunderyactivity OR for longer lasting episodes with heavyand continuous precipitation given by accumulationin mm/24h:

• Columns in the matrix show precipitation rates,and rows the estimated risk levels. The icons withsymbols and colours represent the potential impactlevel assessed by forecasters.

The risk level is evaluated as a complex function ofseveral factors:



• Meteorological factors as above, but also consi-dering the effects of stability, strong winds or rapidthaw.

• Surface and soil factors such as surface state andsoil moisture along with river flow and height.

• Human and economic factors as above.

DECISION to forecast the IMPACTlevel

The potential impact level depends both on theforecast precipitation amount and rates exceedingthe threshold values AND on the evaluation of therisk level.

The impact level is tuned by forecasters:

A/ the level is increased when

• the timing of the expected precipitation corre-sponds to peak hours (morning/late afternoon orearly evening),

• the location corresponds to more densely-popu-lated areas, in particular in valleys close to therivers or in coastal and estuary areas,

• successive significant rain episodes are forecastin the short and medium range (ie. in the next 5 to7days),

• there are strong river flows and high levels,

• soil is wet or saturated,

• rapid thaw from a dense and deep snow coverover a catchment area is expected.

B/ the level is decreased when

• the soil is dry,

• there are low river flow and level,

• no thaw is expected.

Concluding Remarks

The objective evaluation of risks remains a difficulttask for forecasters and would be very helpful forthe potential impact matrix. The development ofensemble prediction systems for short range fore-casts (like GLAMEPS, or at a larger scale ECMWF-EPS) will be tested in the near future to get bettermeteorological fields and indices giving an indica-tion of the spread of the forecasts and alternativescenarios. We also need the support and the co-operation of authorities and hydrological services toget a more objective assessment of the impact ofweather events on society.

For (very) short range forecasts up to 6 or 12 hoursthe current INCA-BE system seems to be verypromising for severe convective weather events,based on those forecast during the summer of2011. Furthermore a more relevant use of remotesensing data like radar and satellite images (includ-ing SAF nowcasting products) is also required byforecasters for their analysis and their nowcastingand should be helpful to issue warnings.


Communications between the NMHS and Mediaand Civil Protection in CroatiaZoran Vakula, National Meteorological and Hydrological Service, Croatia

Introduction

For many years, cooperation of the NationalMeteorological and Hydrological Service (NMHS)with the media and the National Protection andRescue Directorate (NPRD) - which the Civil Defenceis one part - was at a very high level. The situationpartially changed with the appearance of privateweather companies, along with the strengtheningrole of the internet and the increasing number ofweather forecasts on it. The NMHS maintains asignificant role, although it is no longer the onlysource of meteorological information, especially notof the weather forecast itself, which is particularlyevident when following the press, radio and televi-sion. However, for the NPRD, NMHS is still a "singlevoice" for forecasts and warnings, especially fordangerous weather phenomena. Also, televisionweather forecasts – especially those prepared andpresented by NMHS meteorologists – are amongstthe most popular shows on television.

The Structure of the NationalMeteorological and HydrologicalService

The NMHS of Croatia is the fundamental institutionfor Meteorology and Hydrology in the country. It wasfounded by decree of the Government of thePeople's Republic of Croatia (PRC) on 27th ofAugust 1947. Part of its staff (fifty employees) andequipment was taken from the GeophysicalInstitute and Ministry of Construction of PRC, whichconducted some services in the fields of meteorolo-gy and hydrology (including observations andweather forecasts). Until Croatia became an inde-pendent state (in 1991), the NMHS operated asrepublic institution, and after that as state organiza-tion. Since 1992, Croatia has been a member of theWorld Meteorological Organisation (WMO), and theNMHS peforms international cooperation on behalfof the state. Over the years, as required by thegrowing range and amount of work, the number of

employees has increased to reach a staff of morethan 440. The Institute's internal organization andits scope have been constantly changing and adapt-ing to the increasingly demanding set of require-ments for hydrological and meteorological servicesto meet the demands of the economy, traffic andthe environment.

For many people, the most recognisable part of theNMHS is the Weather Forecast & Analysis Division(WFAD). This consists of the Department of MaritimeMeteorological Centre in Split, a branch office inRijeka which Specializes in Maritime Forecasts, andthe Department of Weather Analysis andForecasting in Zagreb.

WFAD issues a range of products in its area ofexpertise – nowcasting, short-range weather fore-casts and medium range forecasts. There are alsomonthly forecasts (issued twice per month, onTuesdays) and seasonal forecasts (once per month,issued around the 15th). For the general public thereare forecasts for radio, TV, newspapers, web (includ-ing Facebook), SMS, MMS, and a telephone service.There are also special customers like civil services,fire departments, agriculture, roads, engineering,mine-disposal, water and power management andtourism.

Communication to the Media

Forecasts from WFAD appear daily in various media,from traditional media, such as newspapers, radioand television to newer media such as SMS andMMS messaging, web sites and Facebook. In 2010the media accounted for about 52% of directrevenue. The largest part of this was from TV (60%),and the rest in equal proportion came from radio,newspapers and the web.

The NMHS on the Web

NMHS has had a presence on the web almost fromits very beginnings in Croatia, in the first half of



the 1990s, and today has many domains:www.meteo.hr, www.klima.hr, www.prognoza.hr,www.vrijeme.hr, as well as forms without the wwwprefix. Numerous daily meteorological data – bothobserved and forecast, appear on the site. Forexample, there are subjective forecasts from fore-casters in both textual and graphic form for thecurrent and the next day for Croatia - and especiallythe capital Zagreb – as well as text forecasts formariners. Also, there is a 3 day outlook in textualform, written by a forecaster. In addition to thesetypes of subjective forecast, there are also objectiveforecasts - direct model output of models in variousforms – for many locations in Croatia and for thecountry as a whole. Particular attention is paid tothe warnings in the classic, text format in theCroatian language and textual-pictorial form in theMeteoalarm system. All warnings are subjective,written by a forecaster.

Newspapers

NMHS has several decades of experience in fore-casting in the newspapers, since the days whenthere were only a few, to the newspaper boom a fewyears ago, until today, when sales are declining.NMHS does not prepare final products for newspa-pers; all that is required are some elements ofweather reports, which are sent to the newsroomready to be made suitable for printing.

According to market research between October andDecember 2010, which was conducted andpublished by De facto production company, thereare 13 best-selling newspapers in Croatia over thelast 12 years. Of these, 3 do not have forecasts. Ofthe remaining 10, 5 co-operate with HMHS, whilethe other 5 take forecasts from private meteorologi-cal companies. Counting the average number ofdaily newspapers readers in Croatia, the daily fore-cast reaches about 1/3 of readers. It is important tonote that the most widely read and cheapest news-papers account for 44% of respondents to thesurvey, and these newspapers have forecasts froma private company. It is interesting to mention that51% of readers older than 12 years at least read anewspaper, and that 22% of respondents do notread a daily newspaper at all.

Radio

Co-operation between the NMHS and radio also hasa long life. Forecasts have been read by radio news-readers since 1953, and expanded upon since1966, when discussions about weather and fore-

casting started. Today many forms of co-operationand transfer of meteorological information exist –newsreaders reading forecasts, interviews with fore-casters, or forecasters reading forecast. In themeantime, numerous radio stations have appeared,at the national and local levels, especially since1995. According to research results and the radiomarket (also by De facto production, conductedbetween July and September 2011), 12 radiostations in Croatia have the most listeners. Of these12, NMHS works with 6, which cover about 45% ofrespondents, of which the two most listened coverabout 27% of the market.

The forecaster now has 8-10 live radio broadcasts,depending on the day of the week, and 7 pre-recorded forecasts are broadcast on differentstations throughout the day. Most interviews aredone between 6:20 pm to 9:30 am. During theseinterviews – usually around 2 minutes but some-times 10 minutes in duration – in addition to theforecast for the following days, forecasters oftengive a climatological summary of recent weeks, aswell as forecasts for several weeks or even monthsin advance. If, during the daytime, an extraordinarymeteorological event occurs, forecasters mayappear on many different radio stations at theirrequest, regardless of whether the stations havecontracts with NMHS or not.

Television

The has been co-operation between the NMHS andCroatian Television since 1956, when meteorolo-gists began periodic appearances. Regular forecastsin the evening began in 1968. From 1996 forecast-ers have also prepared reports for other televisionprogrammes at the national and local level, some-times also appearing on them. Since 2005 the fore-casters have started to work in shifts on television.Forecasters in the NMHS prepare forecasts for manymedia - newspapers and radio, and text and imagereports for commercial television such as NOVA TV,currently the most popular station in Croatia. Theforecasters at HTV prepare reports only for HTV, from6h to 18h, every day, or in two shifts, with the firstshift preparing programmes from 6h to 14h, and asecond shift in the afternoon and evening. Themeteorologists have a direct connection to theIntranet in the NMHS. They prepare all images andanimations in the TriVis visualisation system, enterthe text forecast in the iNews system and recordthemselves and run the visualization of images andanimations.


There are many TV shows with weather informationand forecasts. The total amount of weather informa-tion, mainly forecasts, during week days is about 16minutes per day, and during the weekend is about10 minutes.

The main weather TV show, just after 20h and themain news, is one of the top 5 most watched showson Croatian television almost every day. The showsare visible on TV and on www.hrt.hr, where there arearchives of the last 20 evening shows, so the fore-casts can be verified.

Communication to Civil Protection

Civil protection is a part of the National Protectionand Rescue Directorate, with which NMHS has along term co-operation. There is a strict standardoperating procedure, which includes provision ofgeneral and specific forecasts and warnings ofhazardous weather events, all in text form, alongwith forecasts for the MeteoAlarm system. TheOperative Department of the National Protectionand Rescue Directorate (Centre 112) has recognizedMeteoAlarm as being beneficial for information ontheir activities and as valuable supplement to ourregular cooperation.

Warnings in Croatia have long history, for sea trafficwarnings have been issued by the MaritimeForecast Department since 1947 via radio (WMOstandards) and also via the internet today in severallanguages (Croatian, English, Italian and German).

Warnings and Forecasts are also issued to theNational Headquarters for Search and Rescue atSea. Warnings of hazardous weather events are anintegral part of many forecasts on radio and televi-sion, and sometimes also in newspapers.

Croatia officially joined the MeteoAlarm system(www.meteoalarm.eu) from 17th July 2009. Everyday forecasters write warnings for 8 climatologicalareas and send them to www.metealarm.eu. The listof parameters that are warned about in Croatia arewind, rain, snow / ice, thunderstorms, extreme lowtemperatures, extreme high temperatures and fog.Integration of hydrological warnings in Croatia stillnot operational but is expected in the future.

Conclusions

The Meteorological and Hydrological Service ofCroatia has a very strong relationship with NationalProtection and Rescue Directorate, including CivilProtection, and with majority of the main media inCroatia.

Cooperation between NMHS and Media and CivilProtection is very important for better understand-ing of weather information and warning people forhazardous weather events.

� Figure 1

LTV shows withweather information,mainly forecasts, onCroatian televisioneach day, withcontents, durationsand style ofbroadcasting of everyshow



IntroductionChallenge or Necessity?

Life brings numerous new challenges, and experi-ence shows that each of us goes before an audienceand becomes a teacher – even if only briefly - onincreasingly frequent occasions. I first heard ofspecial training for meteorology trainers a few yearsago, but this tended to focus on what should betaught, rather than how. In 2014, the WMO plans toorganize training for trainers in the Europeanregion. In this article it would be useful to introducethe training objectives and the benefits for itsparticipants. My own case has shown that meteoro-logical professionals are really not aware of thenature of such training. We would like to invite othercolleagues to dare to participate in them, because,despite the high stress and strain of training, thebenefits are undeniable and obvious.

The WMO Education and TrainingOffice – the Initiator of thisExpansion in Learning

In 2010, I was lucky to participate in the WMOEducation and Training Programme (ETRP) seminar'Train the Trainers'. During two weeks in May, 20meteorologists from meteorological services in 14European countries and several universities gath-ered in the beautiful medieval town of Sibiu inRomania. This workshop was aimed at meteorolo-gists of the WMO Region VI who were one way oranother related to the training of professionals andstudents (Fig. 1).

My colleague Metaxa Konstantara from the HellenicNational Meteorological Service and I then present-ed the results of this workshop to the participantsof the EUMETCAL VI Workshop held in Geneva,recommending that similar courses should beconducted in the EUMETCAL environment. We were

glad that the trainer‘s competence matters were themain focus at the EUMETCAL VII Workshop held inthe beginning of March 2012 at the Met Officeheadquarters in Exeter, United Kingdom. Specialistknowledge was imparted by experts in their fieldfrom WMO-ETRP, EUMETSAT, EUMETCAL and COMET,namely: Luciana Veeck, Jeffrey Wilson, MarkHiggins, Tero Siili, Ian Bell, Patrick Parrish, RogerDeslandes and others. The experts lectured,conducted workshops and talked with more than 50participants from 23 countries, mostly from Europe.

I made a statement about the undeniable benefitsof the WMO-ETRP workshops at the CALMET IXconference in Pretoria in 2011, in view of theupcoming training event for the African region. Fromfeedback I can say that the conference helpeddoubting attendees decide to participate.

What is a Good Lecture?

Each meteorological service attaches increasingimportance to the raising of its specialists‘ qualifi-cations, because meteorologists face new chal-lenges due to rapid technological development andthe need to implement improved methods andapplications. Unfortunately, financial resources arelimited, and increasing use is made of distance orblended learning methods. They are cheaper,because they eliminate travel and subsistenceexpenses of the trainees. Participants only need toregister, and to have a computer with specificallyinstalled programs, headphones, speakers andmicrophone. A course presenter prepares a presen-tation - usually in ppt format - and posts it into avirtual environment. They then connect online to thesession and the lecture is ready to begin. But is thisreally enough? Is it possible to apply the samemethods in a remote lecture as to those delivered inthe classroom and 'live'? How can we make trainingnot only useful but also interesting, and how shouldwe overcome communication barriers and engage

What Forecasters Need to Knowabout Training for TrainersIzolda Marcinonien�, Chief Specialist Meteorological Forecasting DivisionLithuanian Hydrometeorological Service under the Ministry of Environment


the audience? How should we choose assessmentcriteria for the courses and present them in a formof questionnaire, and what is required of a slidepresentation? In short, how does one become acompetent trainer?

Experience shows that many lecturers are self-taught and that firstly they are experts in theirrespective fields, usually in meteorology.Unfortunately, to professionally conduct a lecture,in-depth knowledge of the subject and rich experi-ence are not enough. It also requires pedagogicaland psychological knowledge. You need to haveknowledge of the audience - professionals orstudents, ethnic composition and cultural back-ground; it also is important that the lecture isconducted in their native language.

Lecturers are often faced with a problem of howto focus students’ attention during the lecture(Fig. 2), by the use of elements of performance,presentation design, and so on. Visual andkinaesthetic communication (otherwise called'body language') is very important during a face-to-face type of lecture. During this type of lecture

it is easier to interest students and engage anaudience, to achieve a deeper understanding ofthe mater ial and make it memorable. 'Noemotions, no learning!' might be the credo ofsuch courses. Meanwhile, preparation for lectur-ing at a distance requires much more effort fromthe teacher – it is not always possible to show avideo or animation online, and one must carefullyselect the layout and background of the slidesand any presentation effects. Even with a cameraconnected to PC, it is hard to employ 'bodylanguage’; so one should focus more on the voiceincluding the intonation and pauses. However,this teaching method has its advantages - alecture often is recorded, so both students andteachers have an opportunity to review the mater-ial carefully, objectively evaluate it, and avoid thediscovered mistakes in the future.

Quite often the success of a lecture is evidenced byabundance of students' questions and their quality.A professionally delivered lecture satisfies not onlythe students' curiosity, but also encourages them tofurther deepen their knowledge, and gives a lectur-er personal satisfaction with the work and guide-

� Figure 1



lines on how to improve the work and update thedata (Fig. 3). It means that each presentation is alesson for the trainer as well.

Conclusion – Never Stop Learning!

Clearly knowing how to create the best content ofmeteorological training is very important, but this ismore related to raising of a specialist (usually aforecaster) qualifications rather than the instruc-tor’s teaching kills. This article has another aim - to

show the sources of, and techniques for improving,knowledge transfer techniques to achieve theobjective of delivering a 'good' lecture. Relationsbetween meteorologists in different countries arebecoming closer and more varied, and the numberof training courses and events is rapidly growing.So a logical question arises as to which methods tochoose. My advice would be as follows: if you haveto impart professional knowledge to colleagues,customers or students, but feel you lack knowledgeof how to conduct your lectures, then 'Train theTrainer' workshops are meant specifically for you.

� Figure 2

The problem in the middleand the end of lecture

� Figure 3

The difference betweentrainees with differentmotivation


Introduction

Northern Atlantic atmospheric low frequency varia-bility can be described as mainly evolving through afixed number of recurrent states. (Vautard, 1990)identified four weather regimes which have a meanpersistency about 10 days. Further studies haveshown that these can aid comprehension of largescale dynamics, especially as when also conside-ring interactions with the Atlantic jet (Riviere andOrlanski, 2007). In particular, periods of likelyregime transitions may have high impact on thesequence and density of mid-latitudes cyclones(Michel and Riviere, 2011).

More recently, as concern on climate changeimpacts increases, weather regimes are central tonew developments in statistical downscaling tech-niques. Weather regimes may be characterised byapplying automatic classification tools over longperiods reanalyses. It is then possible to identifylocal climate properties that occur specifically inassociation with particular regimes. Global climatemodels often suffer from a lack of resolution suffi-cient to describe regional to local atmosphericproperties, but they may better reproduce weatherregime distributions. An example of such approachis described in Nuissier et al. (2011) over SouthernFrance, which shows that the conditionnal occur-rence probability of high precipitation events can bededuced from given regime membership.

In the context of weather forecasting, weatherregimes analysis in model forecasts could informforecasters about the next days’ large scale evolu-tion and likely elements of local weatherpredictability in a more synthetic way. The assump-tion is that when the regime prediction skill of amodel is poor, the predictability of local weather isweak.

We present experimental results concerning theforecasting of the 4 Atlantic weather regimes withthe operationnal ensemble forecasting system atMeteo France, PEARP. We also show the results ofoccurrence probability calculations of HeavyPrecipitating Events (HPEs) over SoutheasternFrance in one intense case of 2008 autumn season.

The PEARP Global EnsembleForecasting System atMeteo France

The PEARP ensemble system runs twice a day at06UTC (72h forecast range) and at 18UTC (108hforecast range). The horizontal resolution isT538c2.4, giving an approximate mesh size of15 km over France and 90 km on the opposite sideof the globe. In the vertical, the model has 65levels. The ensemble is composed of 35 members;one unperturbed and 34 having perturbations inthe initial state and in the physics of the model. The

Experimental Daily Forecasts of Northern AtlanticWeather Regimes and Heavy Precipitating Events(HPEs) over Southern France with theMeteo France Global Ensemble System PEARPB. Joly1, N. Girardot2, B. Roulet2, C. Labadie1

1 National Center for Meteorological Research, Toulouse, France,2 National Center of Forecasting, Toulouse, France

June 1, 2012

Abstract



initial perturbations combine Ensemble DataAssimilation (6 members and the assimilationensemble mean) and singular vectors (TL95c1, 65levels) targeted on different areas and using differ-ent metrics and optimization time windows accord-ing to the region (tropical or extra-tropical). Theperturbation of the model is performed, for eachindividual member, by randomly choosing the phys-ical parameterization scheme from a set of 10different ones, with the constraint that each schemeis not used more than 4 times.

A range of products are provided to the forecasters.Some of them are very “classical“ products: probabili-ty maps, quantile maps, postage-stamp maps andspaghetti plots. Fig. 1 shows an example of a 4-quan-tile presentation of one precipitation forecast for thevery intense case on 1st Nov 2011. Others productsare more specific, for example trajectories of mid-lati-tudes cyclones or tropical cyclones. Finally, somerecently developed products are the result of collabo-ration with the research department and are dedicat-ed to the detection of particular severe events.

The Atlantic Weather Regimes

During, the 1990s many studies were devoted toweather regimes characterisation. Over the Atlanticdomain, major studies (Mo and Ghil, 1988; Molteniet al., 1990; Vautard et al., 1988; Vautard, 1990)

identified quasi-stationnaryweather patterns based onclustering techniques. ThenMichelangeli et al. (1995)reviewed the number ofclasses and proposed a clas-sification scheme for theNorth Atlantic consisting of a4 weather regimes. Withinour process of regime calcu-lation, a Principal Compo-nent Analysis is carried outover the whole reanalysis(ERA40) and the classifica-

tion is then performed over the first several compo-nents retained. A distance criterion is then definedin the EOFs subspace. After some iterations thatminimize the variance intra-cluster (k-meansmethod), each day is allocated to a cluster. Thecentre of a cluster is defined as the variable averageover all its members. Fig. 2 shows an illustration ofthe centres of the weather regimes we computed.Even though the Northern Atlantic Oscillation isconstructed simplier like the projection onto thefirst EOF only, weather regimes patterns have simi-larities with the NAO ones. Two broadly equivalentclassification scheme coexist then, the originalVautard names are pointed out in the figure, as wellas the NAO one.

It may be possible to subjectively identify whichregime a current meteorological situation resembles,but we propose here a more objective procedure. Inorder to estimate to which regime a model forecastshould be allocated, we project the forecasts onto theERA40 first representative EOFs (twenty EOFs corre-spond here to 97% of the total variance). Then wecompute the correlation distance in the EOFssubspace between the forecast and the regimecentres for each member at each forecast lead time.In. fig. 3, we can see the results we obtained whenapplying this technique to one PEARP forecast for 19Nov 2011. A membership criterion has been validatedon the reanalysis, showing that under a correlationthreshold the regime membership is predicted with a


� Figure 1

Ensemble quantiles of a 78HPEARP precipitation forecast,from 30th Oct 2011 18H UTC.The quantile are 25% (upperleft), 50% (upper right), 75%(lower left) and Max (lower right).


� Figure 2

Northern Atlantic 4 weatherregimes, composites ofgeopotential height at 500 hPa

� Figure 3

Projection onto the 4 Atlanticweather regimes of the 18 nov2011 18h UTC PEARP forecast.The dashed line is thesignificance threshold forregime membership.

� Figure 4

Evolution of the geopotentialheight at 500 hPa, ARPEGEanalysis from 19 nov 201100h UTC to 23 nov. 201100h UTC


85% score (shown by the dashed blue line). Thepanels in Fig. 3 show the projection onto each of the 4regimes versus the lead time of the forecast. The 34members of the ensemble are drawn in grey lines,whereas the control is the black line, and the mean ofthe ensemble is the red line. The lower the projectionis, the closer the member is to the regime center. Atthe beginning of the forecast, the Blocking regime isobserved. The Atlantic Ridge and NAO- are not likelyto occur all along the forecast. We chose this situationbecause it is a good illustration of a transitionbetween two periods of different regimes, herebetween Blocking and NAO+, which occurs betweenJ+1 and J+2. Fig. 4 shows the corresponding evolutionof the analysed geopotential height at 500 hPa, theblocking is clearly present at the beginning of theperiod over western Europe, but a strong zonal gradi-ent develops towards Iceland by day 3.

On another hand, one can note that the spread ofthe ensemble is naturally increasing with time, andso does the uncertainty of the regime prediction. Bythe way, the increase in spread is not the same forall panels. The anticorrelation of the forecast withNAO- is very strong, with weak spread, wheareas forthe other regimes the spread is much larger fromJ+1. The spread then is not sensitive to the intensityof the correlation.

Forecasting the OccurrenceProbability of HeavyPrecipitating Events

The CYPRIM French research programme aimed toassess the predictability of HPEs that often hit theFrench southeastern region during the autumn

� Figure 5

The two patterns favouring HPE occurence. Composites of geopotential height at 500 hPa (red lines), moisture flux anomaliesat 925 hPa (shaded colours), and winds (over 5m.s−1) at 925 hPa (blue arrows)

season. One section of this programme was aimedat defining the large scale circulations (LSC) thatfavour the occurrence of HPEs. For this purpose, astatistical downscaling technique has beendesigned, a further description of which can befound in Nuissier et al. (2011). In this study, a clas-sification is performed for the autumn days withsignificant rainfall and leads to four typical LSCclasses. Two of them discriminates the HPEs prettywell : the probability of HPE occurrence increases asthe EOFs distance to the center to the LSC is gettinglow. The composites of these two classes are shownin Fig. 5. Two configurations for humidity anomaliesand strong low-level jet hitting the area of interestare underlined. Then, the ability of such atmospher-ic features to represent sufficient conditions for HPEto occur is assessed. A multivariate linear regres-sion based on LSC distance and low level parame-ters shows that more than 25% of the HPEs (aroundtwo hundred betwen 1960 and 2001) coincide withsuch conditions with a skill score of 69%. In otherwords, when these conditions are observed inERA40, a probability of 69% for an HPE to occur isfound. We investigate here the application of thisrelation in one PEARP3 forecast. The regression isapplied on each member of three successive fore-casts (Fig. 6) to examine the persistence of HPEdetection throughout the forecasts. The example onFig. 6 is for the 1-2 November 2008 situation whichis one of the most severe events in recent years. Theleft panel shows the ensemble member forecastterms which validate the criterion in red and theothers in green. It shows that the event was detect-ed since the 29 October 2008 18 UTC forecast andmaintained in the two successive forecasts. In thelast forecast, at the lead time corresponding to 2nov 2008 12-18 UTC, all the members match the


� Figure 7

Precipitation daily amount between 31 oct 18 UTC. and 1 Nov 18 UTC

� Figure 6

Left panel : HPE index on each term of each member of the PEARP3 on 3 successive runs: red box, the criteria are true, green box thecriteria are false.Right panel : the overall HPE occurrence probability taking into account the HPEs index climatological probability and the ensembleprobability


References:

1. Michel C, Riviere G. 2011. The link between rossby wave breakings and weather regime transitions. J. Atmos. Sci. 68:1730–1748.

2. Michelangeli PA, Vautard R, Legras B. 1995. Weather regimes: recurrence and quasi stationarity. J. Atmos. Sci. 52: 1237–1256.

3. Mo K, Ghil M. 1988. Cluster analysis of multiple planetary flow regimes 93(D9): 10 927–10 952.

4. Molteni F, Tibaldi S, Palmer TN. 1990. Regimes in the wintertime circulation over northern extratropics. i: Observational eviden-ce. Quart. J. Roy. Meteor. Soc. 116: 31–67.

5. Nuissier O, Joly B, Joly A, Ducrocq V, Arbogast P. 2011. A numerical study of three catastrophic precipitating events over southernFrance. Part I: Numerical framework and synoptic ingredients. Quart. J. Roy. Meteor. Soc. 137: 1812–1827.

6. Rivière G, Orlanski I. 2007. Characteristics of the atlantic storm-track eddy activity and its relation with the north atlantic oscilla-tion. J. Atmos. Sci. 64: 241–266.

7. Vautard R. 1990.Multiple weather regimes over the North-Atlantic: analysis of precursors and successors. Mon. Wea. Rev. 118:2056–2081.

8. Vautard R, Legras B, Déqué M. 1988. On the source of midlatitude low-frequency variability. Part I: a statistical approach to per-sistence. J. Atmos. Sci. 45(20): 2811–2843.

HPE criterion. As previously stated, the highestprobability that can be reached with the criterion isabout 70%. Then, assuming that the model reliablyreproduces the downscaling function, it is possibleto compute the forecast probability of HPE occur-rence that compounds with the ensemble signal. Itcan be expressed as:

Clim Ens Clim EnsPHPE = PHPE|Red .PRed + PHPE|Green .PGreen

where is the climatological probabilityfor a day to have an HPE when the criterion is red(i.e about 70%), is the probability for a termmember for being red over the three forecasts,

the probability for a day to be an HPE whilethe criterion is green, and the probability fora member term for being green over the three fore-casts. The right panel of Fig. 6 shows the probabilityof HPE occurrence computed in this way. The theo-retical maximum of this value is 69%, and in thiscase it is nearly achieved over the three forecasts.Of course, for other events the probability of HPEcan be very weak, this “index“ cannot be consid-ered as an all-in-one predicting tool. However,considering that the technique is designed to detectquasi sufficient conditions it can be considered as areliable pre-warning index for high impact HPEs.

These results have been derived for one season atthis time, and analysed only roughly. However, theprocedure has been extended experimentally to therest of the year, and it might be interesting toextend the validation period to the whole year.

Conclusion

The projection of an ensemble forecast on theAtlantic weather regimes shows the possibility ofquickly interpreting the main features of the largescale circulation and its associated predictability. Itcan help identify likely transition periods inadvance, of which some may be associated withhigh impact weather. The synthetic view providedby this pro-duct could also make the comparison ofseveral models easier, which consitutes a majorpart of the forecasters role.

The statistical discriminant LSC is explored in theforecast to show the possibility of computing theprobability of occurrence of high impact weather.This computation shows a fair ability to anticipatethe event detection. By means of an appropriatevalidation work by comparing with the observa-tions, this HPE index could represent efficient pre-warning information for this kind of event in themedium range.

C limP HPE|Red

EnsPRed

C limP HPE|GreenEnsPGreen

mise en page 1 › newsletter17 › full.pdfminutesofthe2011annualmeeting,bergen, norway,7th october...

Documents