The European Flood Awareness System (EFAS)
This is a modal window.
Das Video konnte nicht geladen werden, da entweder ein Server- oder Netzwerkfehler auftrat oder das Format nicht unterstützt wird.
Formale Metadaten
| Titel |
| |
| Serientitel | ||
| Anzahl der Teile | 295 | |
| Autor | ||
| Mitwirkende | ||
| Lizenz | CC-Namensnennung 3.0 Deutschland: Sie dürfen das Werk bzw. den Inhalt zu jedem legalen Zweck nutzen, verändern und in unveränderter oder veränderter Form vervielfältigen, verbreiten und öffentlich zugänglich machen, sofern Sie den Namen des Autors/Rechteinhabers in der von ihm festgelegten Weise nennen. | |
| Identifikatoren | 10.5446/43382 (DOI) | |
| Herausgeber | ||
| Erscheinungsjahr | ||
| Sprache |
Inhaltliche Metadaten
| Fachgebiet | ||
| Genre | ||
| Abstract |
| |
| Schlagwörter |
FOSS4G Bucharest 2019201 / 295
15
20
28
32
37
38
39
40
41
42
43
44
46
48
52
54
57
69
72
75
83
85
87
88
101
103
105
106
108
111
114
119
122
123
126
129
130
131
132
137
139
140
141
142
143
144
147
148
149
155
157
159
163
166
170
171
179
189
191
192
193
194
195
196
197
202
207
212
213
214
215
216
231
235
251
252
263
287
00:00
Fahne <Mathematik>Güte der AnpassungSelbst organisierendes SystemInformationKontextbezogenes SystemPhysikalisches SystemAggregatzustandVorlesung/Konferenz
00:31
Selbst organisierendes SystemAggregatzustandSpannweite <Stochastik>Physikalisches SystemEreignishorizontEntscheidungstheorieGlobale OptimierungGruppenoperationKette <Mathematik>InformationWeb ServicesExogene VariableRückkopplungSystemplattformProgrammierumgebungAutomatische HandlungsplanungComputerphysikWasserdampftafelArchitektur <Informatik>Elektronischer FingerabdruckFlash-SpeicherKonditionszahlMereologieNichtlinearer OperatorComputerarchitekturWeb ServicesDifferenteRückkopplungEntscheidungstheorieNachbarschaft <Mathematik>EreignishorizontLuenberger-BeobachterIntegriertes InformationssystemAggregatzustandProgrammierungEchtzeitsystemInverser LimesStatistikDienst <Informatik>InformationPhysikalisches SystemBetriebssystemSoftwareMAPTypentheorieAuthentifikationBenutzerbeteiligungKollaboration <Informatik>Front-End <Software>Servert-TestSchnittmengeSoftwareentwicklerDatenverwaltungExogene VariableGrenzschichtablösungWasserdampftafelInformationsspeicherungWeg <Topologie>WinkelGenerator <Informatik>TropfenPi <Zahl>StandardabweichungMaschinenschreibenTransponderMultiplikationsoperatorPlotterArithmetisches MittelDistributionenraumKlasse <Mathematik>RandwertComputeranimation
06:30
Spannweite <Stochastik>RechenschieberSichtenkonzeptKonditionszahlJust-in-Time-CompilerInternetworkingComputeranimation
06:49
Automatische HandlungsplanungArithmetisches MittelDemoszene <Programmierung>Vorlesung/Konferenz
07:26
EreignishorizontViewerComputeranimationVorlesung/Konferenz
07:49
BrowserClientSummengleichungVolumenMaximum-Entropie-MethodeSchwellwertverfahrenPlastikkarteChatten <Kommunikation>Gerichteter GraphPunktMetropolitan area networkVerkehrsinformationPolygonnetzSampler <Musikinstrument>SinusfunktionNormierter RaumAdressierungFehlermeldungMini-DiscSichtenkonzeptAnfangswertproblemPunktInformationFrequenzBenutzeroberflächeVerkehrsinformationBeobachtungsstudieDifferenteMultiplikationsoperatorDatenfeldMultigraphBenutzerbeteiligungCoxeter-GruppeFlash-SpeicherResultanteTabelleComputeranimation
09:47
VerkehrsinformationNormierter RaumPrädiktor-Korrektor-VerfahrenLokales MinimumPunktFahne <Mathematik>SinusfunktionBildschirmmaskePowerPointRechenschieberSichtenkonzeptFlash-SpeicherDualitätstheoriePunktVerkehrsinformationSoftwareComputeranimationVorlesung/Konferenz
10:12
FrequenzVerkehrsinformationPunktProdukt <Mathematik>p-V-DiagrammNumerisches VerfahrenPrognoseverfahrenVorgehensmodellKette <Mathematik>Lokales MinimumRechenwerkMenütechnikFlash-SpeicherFormation <Mathematik>AuswahlverfahrenInformationsmanagementAlgorithmusFlächeninhaltEreignishorizontFormale GrammatikSchlussregelExogene VariableIndexberechnungGleichheitszeichenGrößenordnungFlächentheorieInformationDienst <Informatik>Inklusion <Mathematik>RückkopplungExplosion <Stochastik>Web ServicesDatentypPhysikalisches SystemInformation RetrievalTypentheorieSoftwareentwicklerRückkopplungProxy ServerFormale GrammatikBildschirmmaskeSpezifisches VolumenEndliche ModelltheoriePunktHyperbelverfahrenParametersystemInformationKategorie <Mathematik>AnalysisSweep-AlgorithmusInhalt <Mathematik>TLSLuenberger-BeobachterEchtzeitsystemDatenverarbeitungssystemDivisionGanze FunktionMultiplikationsoperatorCASE <Informatik>ProgrammierungPackprogrammPhysikalisches SystemInformationsüberlastungInformationsspeicherungProzess <Informatik>PasswortBenutzerbeteiligungGrenzschichtablösungExogene VariableE-MailEnergiedichteMomentenproblemResultanteMeterDruckverlaufGruppenoperationDienst <Informatik>Web ServicesFlash-SpeicherKoordinatenVerkehrsinformationMagnetbandlaufwerkDateiformatSchwellwertverfahrenRohdatenBitBitrateReverse EngineeringWeb SiteViewerComputeranimation
19:38
Keller <Informatik>SoftwareentwicklerProgrammierumgebungCoxeter-GruppeRechenschieberBootenArithmetisches MittelAutomatische HandlungsplanungInformationComputeranimation
20:18
Endliche ModelltheorieRückkopplungProzess <Informatik>MultiplikationsoperatorZweiPhysikalisches SystemExogene VariableWellenpaketVorlesung/Konferenz
21:20
PunktLineares zeitinvariantes SystemAdressierungFehlermeldungExplosion <Stochastik>PunktZweiZeitrichtungFehlermeldungFrequenzBildschirmmaskeDatenbankVorlesung/KonferenzComputeranimation
22:02
InformationPunktDreizehnArithmetisches MittelCoxeter-GruppeEndliche ModelltheorieOrdnung <Mathematik>EchtzeitsystemFehlermeldungComputeranimation
22:49
Coxeter-GruppeOrdnung <Mathematik>Vorlesung/KonferenzComputeranimationBesprechung/Interview
Transkript: Englisch(automatisch erzeugt)
00:09
Okay, good morning. My name is Maurizio Latini and I'm working for the ECMWF and I'm going to talk about the European Flood Awareness System. First of all, some minor information
00:23
about ECMWF. We are an independent intergovernmental organization established in 1975 with 22 member states and 12 cooperating states. The ECMWF has turned 40 in 2015, has more than
00:42
288 employees and collaboration with a lot of research institutes and it provides a lot of more than 40 million observations analyzed every day. Within the ECMWF we are hosting the Copernicus Emergency Management Services. Within the Copernicus Emergency Management
01:06
Services we are running the IFAS, which is a part of the Copernicus program for the flood monitoring all over Europe. Why IFAS? Because within Europe the most severe flooding are
01:27
trans-boundaries, so sometimes these flooding events are leading to incoherent decisions between neighborhood states. So at the beginning the European Commission decided to have
01:45
this trans-boundary operational system to help the member states to have better information about ongoing flooding. This was the idea behind IFAS, which is the first European
02:02
operational monitoring system and can send out early warning notification up to ten days in advance. Plus it sends out to the partners and the European Emergency Response
02:20
Coordination Centre a daily overview about the flooding situation all over Europe. So as I said, it's a probabilistic system, it's operational, so it means that it runs 24-7, 365 days a year. It produces twice a day forecasts, one at 0-0 and the other
02:43
one at 12, and has more than 60 registered partners. And these partners receive observation, sometimes flood notifications and can give back to us feedback on the performances of the system. The IFAS itself is a consortium, so it is composed from different research
03:07
and operational centres. We are ECMWF, we are part of the computational centre, so it means that we are executing the forecast and we are running the information system.
03:21
There is the dissemination centre, which is formed by the SMHI, the Swedish Meteorological Institute, the Rijkswater Stat in the Netherlands and the Slovak Hydrological Institute. And this dissemination centre is analysing the information that IFAS is providing and
03:46
then decides if to send out or not a flood notification. Then we have a consortium for the hydrological collection centre, which is in Spain, and they are collecting historical
04:01
and real-time river discharge. And we have a meteorological data collection centre, which is based in Germany, which collects real-time meteorological data across Europe. In IFAS we have three types of users. We have the anonymous users that can access
04:25
no real-time data, it means forecasts that are older than 30 days, and they have access to a limited amount of information. Then we have the registered users, which are not forecasted, that can access all the real-time data and information, but they cannot receive
04:46
or issue notification. Then we have the registered users and the forecasted users, they have access to all the IFAS infrastructure and they can send out notifications.
05:03
So this is more or less the IFAS architecture for the web part, in which we receive the forecast, we store the forecast in GIS databases, and then we have two types, actually we
05:21
have three types of databases, we have a GIS database, we have an SOS database, we have an oxidation layer, which is mainly Django and Python, we produce WMS services that we deliver to the end-users through OGC standards, we have an authentication
05:47
service based on O2, so the software that we are using for the back-end part is a PostgreSQL and PostGIS, Map Server for the WMS generation, Django and Drupal.
06:10
In the front-end we are using Angular, Leaflet, again Drupal, everything is dockerized and orchestrated with Kubernetes. So as I was telling you, it would be nice if I
06:25
can log, I will try, I don't want to mess up anything, but I want to try to have a live demo, okay, because it's better to show if the internet is working or otherwise
06:49
I have a better plan.
07:27
So we have, we are producing more or less 30 layers a day, which we are delivering
07:56
through the web interface, and we are organizing layers in different categories, as you can
08:03
see, flood summary layers, hydrological layers, initial conditions, flash flood, and a bunch of other layers, plus the static ones. Every layer, for example, the, okay,
08:22
we don't have any, let me go back in time where, every layer is produced twice a day, and we host historical layers, so we can go back in time up to ten years ago. So if I'm going to a situation where we have maybe more floods, okay. So this layer
08:48
here is the reporting points, two years and five years period, which is the one used by the forecaster to issue notification. So how does it work? Every, at zero zero
09:02
and at 12 every day, this layer is produced, is put into the interface, and then the forecaster is analyzing the layer. So through just clicking on a point, you can get all the hydrological information for the point, so we have the discharge hydrographs,
09:26
the period hydrographs, the upstream rainfall, and then the persistence table. So based on this information, the forecasters are analyzing the result of the model, and
09:41
they can decide if they can issue or not a notification. Let me go back to the presentation.
10:10
So as I was saying, we have the reporting points, which is points in the network that are, which the threshold are two, one greater than two years return period, the other one
10:32
required, so these are already shown. Then we have flash floods reporting points, which
10:41
where we, every day we are issuing a flash flood notification because this, the flash flood is much more difficult to be, I'm not an hydrologist so I don't know how to explain it a little bit better, but this kind of notification, these kind of reporting points
11:02
are really difficult to catch, so we have a different model to generate this, and they are produced twice a day like the other ones, and the notification are issued again using this layer. This is the flash flood symbology, and this is what you get if you
11:23
query the layer. We have another type of layer, this is based on observation that comes from radar, and it's generated every 15 minutes, and it shows real time, almost
11:41
real time precipitation-based analysis, and this is how it looks like. This we use a different approach to generate this, we have an in-house WMS service generator which is,
12:01
which reading the netCDF and it's generating through the DCM WF EC chart suite, this layer here, which is an animation by the way, so it can be animated and the user can see how the precipitation are evolving. As I said, we are issuing notification. Notifications
12:28
are divided into three categories, formal, informal, and flash flood notifications. So these are the parameters that the forecasters are using to issue formal or informal notification.
12:40
A notification itself is only an email that is sent to the partner, so the partner can check the forecast, can check his forecast, and then they can decide if they have to take action. So if a notification is just, let's say, it's a helping tool, it's not
13:09
like the official one, so it's up to the partner to decide if they want to take an action based on our notifications. If the criteria for sending notification are not met for the formal
13:24
notification, the forecaster can send an informal notification, which means that okay, something is going to happen, but it's not going to be maybe flooding, but take care and monitor this particular basin. And then we have the flash flood notifications,
13:44
which are the most, let's say, issued ones, because we have a lot of, especially during wintertime, we have a lot of flash flood notifications, and the criteria are here, and these are issued based on administrative reasons, not for basins, plus every day at
14:06
8 a.m. we are sending the daily, the flooding daily report to the emergency and response coordination centre. So it's basically a summary of what happened until the previous
14:20
forecast. So the user that receives notification, what does it see on the web? You have the active notifications that are listed, and for each notification you can get the notification that has been sent to the partner, so you have all these
14:44
hydrological parameters, and the name of the forecaster who issued a notification, so if the partner needs further more information, plus we can query on the notification itself. The forecaster has this, let's say, dashboard to check which are the
15:09
notifications that are actually at the moment active, and they can decide if they can deactivate the notification. Once the notification is deactivated, it goes into
15:21
this other panel, which is a summary of what happened in the last day, and as I was saying, the partner that is receiving the formal notification has the ability and the possibility
15:40
to give us back a feedback. So the feedback is then returned to the system, and we are calculating how well the system has performed, so how precise was the notification issued. So as you can see, we have a colouring system, so it means red, it's really bad,
16:03
so we issued a notification to a partner, but the flooding wasn't happening. And then we have a kind of a rating system to see how well we performed it.
16:21
The daily review, I don't think you can read it. Anyway, it's the situation that was happening before, so in this case, no formal notification was issued, and only a few flash flood notifications were sent to the partners. So this is what
16:43
the European Emergency Response Centre is receiving every day. Users, registered users, can have their own dashboard, so they can create their own viewers to check out, to monitor the situation.
17:06
And if I have time, I can show you how it works on the website. Sharing data. We are currently using the OGC web services to share data for partners and non-registered users.
17:25
So we have two OGC services. We have WMS services for real-time and non-real-time data, and sensor observation services for post-processed reverse discharge.
17:41
SOS services is only accessible by the partners, so they have to be registered, then they will get a partner and they will get a password. WMS services is also accessible from non-registered users. In this case, they will only have access to no real-time forecast, so only forecasts that are older than 30 days. But it's constantly updated, so if you connect
18:03
today, you will get the forecast up to the 30th of July. We offer also a service for downloading the data. So we have two types of data downloads. One is through the Copernicus climate data store, which is the easiest way to get data.
18:26
So simply you go to the CDS, you register, and then you can download the IFAAS data. In this case, the data is available in GRIP2 format or NetCDF. The other way is through the
18:41
ECMWF. ECMWF has this huge archive of meteorological information, which is called the Meteorological Archival and Retrieval System, or MARS, which is a tape system. In this case, you need to be registered into the ECMWF within ECMWF to get the password for getting the MARS.
19:07
And you need to have a little bit of programming skill to download the data. So these are the two ways of downloading our data. Of course, it's raw data,
19:25
but slowly we are now discussing with the partners if we can also provide the models and the data that we elaborated. So last but not least, we are currently looking for developers.
19:46
And this slide is not really correct because the deadline is the 16th of September. So the data has been extended yesterday. So if you are interested, you can
20:00
visit us at our booth in the main theater room and you can get more information. I think I've finished. Thank you for the presentation. Questions?
20:28
Your thoughts using the feedback loop to train your model, like you show that you have a feedback loop based on the notifications? Yes or no. We are using the feedback to train the modeler, not the model.
20:45
So we are getting the responses, we are analyzing that. Well, the scientists are analyzing that and then they are training the model. But it's not an automatic process.
21:00
Also because sometimes the feedback has to be an open questioner, so it has to be analyzed to understand what was the... Plus we have a JIRA system, which is every time you...
21:20
I haven't showed if I have just a second. When there is a point here, the forecaster or whoever is looking at the point can report an error for the point, which means that sometimes the data is not correct.
21:45
So this is what we use to correct the issue. Sometimes the returning periods are wrong or something like that. So we have this form here, and you can say which one was the main problem,
22:15
or you can enter just what you want. And whenever someone is entering this form, we receive a ticket, so we analyze it.
22:25
So if it's a critical error, we tend to fix it. So we rerun. If it's, for example, the model is outputting the wrong data, we are rerunning the forecast and then we are publishing the new one.
22:40
So it's almost in real-time correction, if it's a critical error. Okay, thank you. I think we have to break here. We still have four minutes in order to change rooms. And we'll start with the second presentation. At 9.30.