Upgrade to Pro — share decks privately, control downloads, hide ads and more …

iRIC MI

nkmr
July 03, 2023

iRIC MI

nkmr

July 03, 2023
Tweet

More Decks by nkmr

Other Decks in Science

Transcript

  1. iRIC V4 Release
    June 28, 2023
    iRIC: International River
    Interface Cooperative
    International River Interface Cooperative, since 2010
    “New Developments for iRIC-MI Model
    Coupling”
    Jonathan Nelson
    US Geological Survey- Ret’d
    Now at:
    River Mechanics, Inc
    Arvada, Colorado, USA
    Keisuke Inoue
    Mizuho Instrumentation
    and Research
    Tokyo, Japan
    Kazutake Asahi
    RiverLink Corporation
    Tokyo, Japan

    View Slide

  2. Developments
    in the iRIC
    System
    • Outline of Presentation
    • Coupling iRIC solvers within
    the iRIC-MI system:
    Motivation and structure
    • Some simple examples for
    real world problems
    • Next major iRIC release, but
    alpha versions available now
    for expert users

    View Slide

  3. iRIC-MI = iRIC Model Interface
    • Platform to run solvers connected together
    • Similar to CommonMP and OpenMI but with several
    advantages:
    • Easy to use: iRIC GUI can be used to handle data
    • Development of model is easy
    • Runs on many platforms

    View Slide

  4. Run solvers connected together
    We can connect “Output” socket of a model to “Input” socket of
    another model, and run models exchanging data at every time step
    Ex. Nays2dFlood
    Input socket = Rainfall, discharge from upstream
    Output socket = discharge downstream

    View Slide

  5. Run solvers connected together
    • Input / output data is stored in CGNS files
    • MPI is used for exchanging data between solvers
    • iRIC-MI server controls synchronization and data exchange
    iRIClib iRIClib iRIClib
    Solver1 Solver2 Solver3
    CGNS
    file
    CGNS
    file
    CGNS
    file
    With iRIC:
    MPI = Message Passing Interface
    de-facto standard technology for parallel computing, fast and simple

    View Slide

  6. Run solvers connected together
    • Input / output data is stored in CGNS files
    • MPI is used for exchanging data between solvers
    • iRIC-MI server controls synchronization and data exchange
    iRIC-MI library iRIC-MI library iRIC-MI library
    Solver1 Solver2 Solver3
    iRIC-MI Server
    CGNS
    file
    CGNS
    file
    CGNS
    file
    With iRIC-MI:
    MPI = Message Passing Interface
    de-facto standard technology for parallel computing, fast and simple

    View Slide

  7. Easy to use: iRIC GUI can be used to handle data
    We can run models in the following steps:
    1. Prepare input data for each solvers
    iRIC GUI can be used for preparing calculation condition and grid
    2. Run each solvers separately
    3. Make connections between solvers
    4. Run solvers connected
    5. Visualize result
    iRIC GUI can be used for 2D, 3D visualization and drawing charts

    View Slide

  8. Development of models is easy
    • Language: FORTRAN, C/C++, Python
    • Easy to define “input” and “output” sockets, using definition.xml
    • Use iRIC-MI library for implementing I/O
    No wrapper needed for object-oriented programming languages
    Easy to migrate solvers from iRIC to iRIC-MI

    View Slide

  9. Runs on many platforms
    • PCs with Windows / Mac OS X / Linux
    • Virtual machines in Cloud of Amazon, Google, or MS.
    • Super computers
     Can be used to solve real problems

    View Slide

  10. The components of iRIC-MI:
    • iRIC-MI library = Extended version of iRIClib with:
    • Function for “Input sockets”
    • Function for “Output sockets”
    • Function to synchronize mode
    • iRIC GUI
    • iRIC-MI server:
    • The program that works as “conductor”.
    • Watches the time of each model, and make a model wait until others catch up
    • Emit commands to models in the platform to exchange data
    • iRIC-MI GUI: Model connecting interface
    • Connect “Input sockets” and “Output sockets” of components
    • Saves information about connection to “iricmi-project.xml”

    View Slide

  11. Short demonstration of using iRIC-MI

    View Slide

  12. iRIC-MI Characteristics
    • End user ready (no programming required, this is not
    just another package of callable routines for
    programmers)
    • Interface support for setting up all coupled models
    (each model has its own interface, but they all look
    consistent- learn one, not 20)
    • No limit on type or number of coupled models running
    concurrently

    View Slide

  13. iRIC-MI Characteristics (cont)
    • No limit on multiple-core usage by individual models or
    overall
    • XML for defining model inputs/outputs and coupling
    scheme (invisible to user, but makes new model inclusion
    straightforward)
    • Any coordinate system: structured (including non-
    orthogonal, multigrid, etc) and unstructured (holes, local
    resolution, any element shape)
    • Coupling for global values, boundary conditions, and/or
    individual grid nodes, faces, or cells

    View Slide

  14. Developments
    in the iRIC
    System
    • Outline of Presentation
    • Coupling iRIC solvers within
    the iRIC-MI system:
    Motivation and structure
    • Some simple examples for
    real world problems
    • Next major iRIC release, but
    alpha versions available now
    for expert users

    View Slide

  15. Example 1: Simple 1-d to 2-d coupling for channel model: BCs for short-reach 2-d model come from long-reach 1-D model
    With this coupling, we can run 1-d models
    Over very long reaches and paste in 2-d
    models in regions of particular interest, or
    to be run only when certain conditions are
    met. In this case, the discharge is from
    data, but could be from discharge forecasts
    or from a precip runoff model driven by
    precip forecasts- see below.
    2d model
    location

    2d model
    1d model

    View Slide

  16. Example 2: Rainfall-runoff model coupled to 2-d quasi-steady model: Discharge for the 2d model comes from the RRI simulation (at just 1 point)
    With this coupling, we can drive flow
    forecasts with spatial detail using only
    rainfall forecasts or measurements.
    Because the model is quasisteady, it can
    run at huge time steps (10 minutes in this
    case) the approach is computationally very
    fast. I assume a baseflow of 200m3/s, but
    this will soon be supplied from a coupled
    groundwater model with time variation.

    View Slide

  17. Example 3: Rainfall-runoff model coupled to 2-d unsteady model: Discharge for the 2d model comes from the RRI simulation (at just 1 point)
    With this coupling, we can drive flow
    forecasts with spatial detail using only
    rainfall forecasts or measurements.
    Because the model is unsteady, it runs at
    small time steps (1s in this case) the
    approach is computationally slow, but
    includes the effect of the discharge wave
    and predicts local unsteady effects like
    vortex shedding at separation points.

    View Slide

  18. Example 3: Rainfall-runoff model coupled to 2-d unsteady model: Discharge for the 2d model comes from the RRI simulation (at 10 points)
    Here’s a more realistic case where we couple the rainfall runoff model to a 2-d fully unsteady urban inundation
    model at 10 upstream points. In the coupling system, this is still only boundary condition coupling, although the
    inundation model also handles in-domain rainfall. In this case, our colleague Kazutake Asahi digitized the locus of
    maximum inundation for this three-day flood (blue dashes) which can be compared to the prediction. Not bad,
    considering no calibration.

    View Slide

  19. Example 4 (slide 2): Rainfall-runoff model coupled to 2-d unsteady model: Discharge for the 2d model comes from the RRI simulation (at 10 points)
    A closer view of the inundation area digitized from aerial photos and predicted by the coupled
    modeling approach based on rainfall alone, a precip-runoff model, and a 2-d inundation model.

    View Slide

  20. Future Visions
    • Making solvers more reusable and customizable
    • New applications:
    • Environmental evaluation
    • Coupled models of rain-fall, river flows, ground water
    • Geochemistry
    • Machine learning
    • Easier for new developers to join the community

    View Slide

  21. Developments
    in the iRIC
    System
    • Outline of Presentation
    • Coupling iRIC solvers within
    the iRIC-MI system:
    Motivation and structure
    • Some simple examples for
    real world problems
    • Next major iRIC release, but
    alpha versions available now
    for expert users

    View Slide

  22. 22
    Questions?
    www.i-ric.org

    View Slide