Marine Data enhanced Virtual Laboratory DEVL #1 - June 18
This is a modal window.
The media could not be loaded, either because the server or network failed or because the format is not supported.
Formal Metadata
| Title |
| |
| Subtitle |
| |
| Alternative Title |
| |
| Title of Series | ||
| Number of Parts | 19 | |
| Author | ||
| Contributors | ||
| License | CC Attribution 3.0 Unported: You are free to use, adapt and copy, distribute and transmit the work or content in adapted or unchanged form for any legal purpose as long as the work is attributed to the author in the manner specified by the author or licensor. | |
| Identifiers | 10.5446/42930 (DOI) | |
| Publisher | ||
| Release Date | ||
| Language |
Content Metadata
| Subject Area | ||
| Genre | ||
| Abstract |
|
Tech Talk9 / 19
00:00
InformationPhysical systemVirtual realityWage labourDisintegrationPresentation of a groupRight angleGroup actionEndliche ModelltheorieCombinational logicProjective planeInformationBitCASE <Informatik>Computer animation
00:58
Product (business)Term (mathematics)Data modelSimulationNumbering schemeBuildingService (economics)Physical systemScientific modellingScale (map)Mathematical analysisMusical ensembleSequenceContinuous functionDigital filterContent (media)InformationCalculationRoute of administrationMathematical optimizationInterpolationSurfaceBoundary value problemCompilation albumVariable (mathematics)HexagonCellular automatonAssembly languageSatelliteParameter (computer programming)File formatFrequencyComputer fileProcess (computing)Inheritance (object-oriented programming)Interface (computing)Codierung <Programmierung>SoftwareUtility softwarePersonal digital assistantMultiplication signDifferent (Kate Ryan album)Web 2.0Pairwise comparisonProcess (computing)Moment (mathematics)Service (economics)Set (mathematics)State observerValidity (statistics)InformationContent (media)Point cloudSoftware testingEndliche ModelltheorieProjective planeNeuroinformatikAreaMathematical analysisSimulationTerm (mathematics)QuicksortProduct (business)InterpolationRight angleMenu (computing)MereologyConfiguration spacePhysical systemOpen sourceMusical ensembleMathematical optimizationKey (cryptography)Numbering schemeSpacetimeBuildingTrailCombinational logicBoundary value problemParameter (computer programming)Profil (magazine)SurfaceFrequencyGastropod shellDialectFunctional (mathematics)Modal logicMetreImage resolutionObservational studyThomas BayesOperating systemInitial value problemOcean currentSoftware developerMathematicsGreatest elementResultantUtility softwareNumberCodeData managementCycle (graph theory)State of matterElectronic mailing listLimit (category theory)Assembly languageComputer programmingSatelliteOperator (mathematics)Function (mathematics)outputBitSelectivity (electronic)Kalman-FilterContinuous functionLevel (video gaming)User interfaceComputer networkSequenceMobile WebGraphical user interfaceDigital video recorderRevision controlCASE <Informatik>Inheritance (object-oriented programming)Dimensional analysisComputer simulationIntegrated development environmentComputer animation
Transcript: English(auto-generated)
00:00
Right, okay, so our virtual laboratory, this data enhanced virtual laboratory, is about building information infrastructure in support of the National Ocean Modelling System. I'm giving the presentation, but two of the guys who are listed below, Angus on the
00:20
right there and Hugo on the left, and other people involved in my group here in the room, are Guillaume Galibert and Sebastian Mancini, and the other members of the project are folk in the Bureau of Meteorology, CSIRO, NCI, Claire's involved a bit,
00:43
South Australia Research Development Institute, Tasmanian Partnership for Advanced Computing, University of New South Wales, University of Western Australia, and in this case, Metocean, New Zealand. So what are we actually trying to do here? So our project aim here is to further develop the marine virtual laboratory to support
01:05
the long-term goal of the ocean research community, namely developing the Australian National Shell Seas Re-analysis, or ANSA, as we call it. So what is a re-analysis? This is a technical term for a product which combines the optimum
01:22
combination of model simulation and observations, maximizing the value of the observational set through data assimilation. Now this project that we're doing isn't going to do the re-analysis. Here the emphasis is on building on our capabilities to assemble the necessary observational data and to
01:41
construct services to enable model ingestion of these data into the data assimilation schemes to be used in the analysis. The actual project to do the re-analysis is probably about a three to four year project to actually complete it. So given that we've only got money till December, that's not
02:01
the realistic goal we even think about doing the re-analysis at this stage. So why do we want to do this? So the growth of the blue economy is predicted to be large by 2025 and it's predicted to grow three times faster than the Australian GDP. So there is a burgeoning interest in the marine
02:24
environment and arguments for this national modeling service system have been around for quite some time and in fact it's a recommendation of the National Marine Science Plan 2015 to 2025. Earlier this year the Bureau of
02:42
Sciences began the operational coastal modeling service and to enable that to really get going then certain groundwork has to be done to provide for such an operational system and this is really what this project is about. It's doing the groundwork to provide the observations and the service
03:01
to enable re-analysis program to be done. So re-analysis is quite simple in concept. It combines combining observations and a model state through an assimilation process to produce an updated model state which essentially
03:21
will bring the model generally speaking closer to the observations and thereby providing a far more realistic simulation than sometimes do come out of models simulations. So there are various ways you can do data assimilation and
03:40
there are ensemble methods such as optimal interpolation and using a Kalman filter. There are variational methods either in three dimensions or in four dimensions, I think three dimensions in time, but each method actually requires a way of reducing the observations that are available in
04:02
the community down to a number that are manageable and this manageable number of observations are called super-obs and you want to do this in such a way that you don't lose the information content of the larger data collection. And then when you actually want to do the analysis there are various ways in which you can do it and on the right here you can see
04:23
four different approaches to doing assimilation, combining the observations into an analysis, running a model and you can see you get step changes to the top version, you can do it sequentially with continuous assimilation and again you get a continuous solution that there are step changes in this
04:43
and then there is non-sequential intermittent assimilation when you put the information in when you actually have it available. Or the bottom one which is what we really would like to aim for which is the non-sequential continuous assimilation to end up with a smooth solution from the results of the assimilation process.
05:03
So in setting up coastal models and this is what the Marine Virtual Laboratory actually does. The Marine Virtual Laboratory as it stands at the moment is a tool for accelerating the development of coastal modeling studies. So in the Marine Virtual Laboratory you can go through a step through a menu which enables you to
05:24
select different community open source models. Choose from a menu of those. I've got a bit of a cold still. And you can choose that. You can define your grid and region of interest from a map. You can then select and you can
05:44
play around with that to get resolution right and things like this that you can then select from another menu to choose from data sets for atmospheric forcing from the initial conditions for ocean boundary conditions and you can tune some of the parameters and at the end of the process you can either submit your configuration
06:04
into the cloud to do a simulation with a particular model or you can actually bundle the data sets that you've collected constructed from the particular model of interest and take them away and use them on a computer somewhere else. But in all those processes there are certain things that you need to do. You need information about
06:25
the geography, about the coastlines, you need information about the sea depths, the bathymetry, you need an initial condition, you need boundary forcing for the ocean and you need surface forcing for meteorology to drive the ocean circulation in the process. And you also need some other information particularly in coastal
06:45
areas. In Marvel at the moment once you've chosen your region of interest in your time period then a web processing service delivers observations that you can use for validation of the simulation. And what we're moving on to now is thinking about how you use the observations available in data assimilation
07:05
processes. Sorry? Okay, plenty of time. So what you actually want in a coastal ocean reanalysis is a variable grid. And this example here which I've chosen as the CSIRO compass model for Storm Bay which is just down here
07:25
in Hobart, shows the kind of resolution that realistically you want to do if you're going to do a sensible reanalysis of coastal processes. So here you've got grid sizes that range between five kilometres down to a hundred metres.
07:41
So our project here intends to acquire and assemble the necessary data to do this reanalysis between 1992 and 2016. We're compiling atmospheric forcing, ocean forcing. Part of the project is providing a coastal discharge set which doesn't exist for Australia at the moment to enable us to do this because fresh water inflow into coastal
08:04
regions is an extremely important forcing function. And we need to assemble the necessary bathymetry and information about tides because tides in coastal areas have to be taken account of when you're doing data assimilation.
08:20
So the second part of this is to develop the services to prepare these model ready observations, the super-obs for use in the intended schemes and the schemes that we're going to use, assemble stuff for ensemble Kalman filter, ensemble optimal interpolation and 4D bar. And there's remote sensing data, satellite information
08:40
along track data. There's in situ data, combination of profile data, mooring data and glider data. And the intention is to assemble all of this information on NCI and to run a service to process those observations to
09:00
produce the super-obs outputs for the necessary inputs to assimilation guns. So we're in the process of looking at designing a GUI, a user interface, which will take a limited number of parameters. It will take the grid that you want your observations gridded to, take the time period of interest and the data
09:20
cycle time that you want to adopt. You'll then choose a particular assimilation scheme. You'll choose observation parameters from a list that we have and then the observations will then be processed. And that's our job. The Australian Ocean Data Network, using either web feature services
09:40
or web processing services. We're still in the process of deciding how the best way approach to go. Then all this then runs on NCI using code that's been developed by Australian data assimilation scientists. One code for ensembles, one code for 4D bar.
10:00
And then your output from the end is these super-obs formatted for the chosen scheme that you will then use to run your particular model. We're aiming for this service to be open. And this is something we're negotiating with NCI at the moment as to how this can actually work, because a lot of these data sets are under different projects in NCI. We want to bring them together and make them available to the community. Some of the key
10:29
involvement of New Zealand in this space that we're contemplating, we have to extend this region of interest over to New Zealand. We still, as I said, mentioned with looking at the best way of delivering the observations in terms of which sort of service would be appropriate.
10:44
And in general, looking at how efficient can we make this production service for the users. And then the latter parts of the project say two things, really. One is about demonstrating the utility of these services through test cases, which our modeling teams are in the process of defining at the moment.
11:04
So it may be they will run different simulation schemes to the one they currently use in their work to look at the comparison between the results, but also look at how efficient it is to construct these super-obs compared to the way they do it manually at the moment. And then when we've piled all those data sets for the
11:23
25-year simulation, then we'll look at how we add these data sets into Marvel as it stands at the moment, which will make the utility of this service much more attractive to our users. And I think that's it. Thank you, Roger.