1 / 96

Jean-Luc de Kok Flemish Institute for Technological Research (VITO)

TOWARDS A GENERIC MODEL LIBRARY. Jean-Luc de Kok Flemish Institute for Technological Research (VITO). Outline Introduction: state of affairs Why a generic model library ? What is a generic model library ? Requirements of generic MBBs SSA Extend models: what do have so far?

Télécharger la présentation

Jean-Luc de Kok Flemish Institute for Technological Research (VITO)

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. TOWARDS A GENERIC MODEL LIBRARY Jean-Luc de Kok Flemish Institute for Technological Research (VITO)

  2. Outline Introduction: state of affairs Why a generic model library ? What is a generic model library ? Requirements of generic MBBs SSA Extend models: what do have so far? Designing the library: steps to take The Online Data Portal: ready for testing Data management Extend-PCraster interface: a quick outlook Appraisal Step Discussion

  3. Introduction

  4. Situation Feb 2009: based on SR and review of Extend models : 14 working Extend models covering range of topics - eutrophication / waterquality (9) - fisheries / shell species (4) - hydrology (4) - tourism (3) - economics / social dimension (4) 4 sites with custom libraries, 6 use H blocks, 6 do not , 2 use a mix

  5. Note 1: Science can (at least for some time) do without a model library, but a good model library cannot do without science That’s why the design of the library is a joint task for the SSAs and supporting WPs

  6. Note 2: • Model review is technically oriented, and not • meant to critisize the scientific quality

  7. Why a generic model library ?

  8. Several advantages: easier to build & expand your models easier to maintain your models easier to exchange (parts of) models with colleagues creating next generation models (post Spicosa) knowledge management in a practical format iterative model improvement rather than “reinventing the wheel over and over again ...”

  9. What is a generic model library ?

  10. Collection of reusable model components:

  11. more unique/ less generic systems SPICOSA toolbox Library SYSTEM SUB SYSTEM DISCIPLINARY MODEL STATE VARIABLE PROCESS OPERATORS, MATHEMATICAL FUNCTIONS, PRIMITIVES system model output & indicators environmental model economic model RK4 integrate Area plotter SPICOSA model blocks Library hydrologic model ecologic model 1D advection fish growth 1D diffusion state variable Nitrogen state variable Plankton return on investment PCRASTER interface process denitrification process mineralisation run off module ExtendSim value Library

  12. What should be in the model and • what not to make it generic ?

  13. Example: ballistic bullet trajectory ? position at time t (x,y)t initial velocity Vy at t = 0 gravitational accelation g Vx bullet mass M initial position at t = 0 (x,y)0

  14. Model block BulletTrajectoryOnEarth : • Outside model: mass M • initial position • initial velocity • Inside model: gravitational acceleration • (universal, model-specific constant)

  15. Model block BulletTrajectoryInUniverse : • Outside model: mass M • initial position • initial velocity • gravitational acceleration

  16. Conclusions: • The more specific a model is the more • is inside it • 2. More generic models receive more data from outside

  17. Rules of thumb: • time-dependent input variable: you must use • an in-connector • changeable constants: you can use an in-connector or DBread block • model-specific information: place inside model but make sure you document it

  18. Requirements for reusable • model components or MBBs

  19. reusable in other simulation studies and Extend models • complementary like jigsaw puzzle pieces • self-explaining and well documented (not only in report) • encapsulated, i.e. implementation is hidden • easy to adapt, expand and maintain • manageable in size • constructed following guidelines of Model Delivery System • robust and reliable (i.e. tested for e.g. fool proofness) • scientific maturity

  20. Typical examples of reusable MBB: standard Extend blocks clear function – limited nr of connectors - reusable Plotter block Data fitter block

  21. Swaney, D.P., Scavia, D., Howarth, R.W., Marino, R.M. (2008) Estuarine classification and response to nitrogen loading: Insights from simple ecological models. Estuarine Coastal and Shelf Science 77, 253 - 263.

  22. + : clear function, limited nr of connectors • : some changeable parameters in holding tanks remark: model could allow for variable nr of age classes note: im- and exporting data slows down your simulation !

  23. changeable parameters better read from database

  24. typical example of reuse of a block in a model

  25. +: limited number of connectors, clear function

  26. here documentation could be added

  27. -: DB table names not generic

  28. SSA Extend models • What material do we have • already to design the MBBs? • - • Some examples

  29. Catch-effort relationship:

  30. Two-layer circulation

  31. Fisheries income:

  32. Box model for solving 1D advection-diffusion equation:

  33. Beach attractiveness:

  34. Eutrophication:

  35. Estuarine dispersal of N:

  36. For a tentative list of candidate MBBs: see D8.8 – Appendix A • Please check/complete yourself

  37. How to do it ? • Deriving the generic MBBs • from your Extend models

  38. STEP1 Identification of model building blocks Is the information reusable? Use your model to identify model blocks Use duplicated model patterns to identify model blocks Define functions of the blocks based on state variables no yes STEP2 Remove model specific parameters and data Is the block manageable in size? Put all parameters in the dialog Use connectors for input and output data Separate data and model Limit the number of connectors and user defined parameters no yes STEP3 Construct the block Is the block self explaining? Encode in ModL or group generic extend blocks in a H-block Pay special attention to full documentation of the block Use the guidelines in deliverable D8.4 & D8.8 no yes STEP4 Implement the block Is the block working at all times? Testing, testing, testing Organize sessions with “foolish” test users no

  39. Creating a new library : manual p. 494 • Note: do not use the name of an existing Extend library

  40. Suggestion to work with five sublibraries: • economic processes (income, profit etc) • ecological processes (bacterial growth, fish pop dyn) • social & policy processes (attractiveness, policy acceptance etc) • physical processes (diffusion, circulation etc) • non-domain specific tools (integration, regression etc)

  41. General design recommendations: • start with the function of the block: what should it do? • ask yourself: can someone else redesign my model? • limit number of connectors • remove changeable parameters from model • document the block: use the help functionality • separate data and model as much as possible; • use connectors and Extend Database functionality • for details see D8.8

  42. Extend gives you several ways • to design your model:

  43. 1. Hierarchical or H blocks: can be placed in a library: • See Extend manual, page 540 ff.

  44. Creating a new H block : manual p. 543

  45. ... or selecting part of your model and making it a H block :

  46. Opening the structure window (double click + Alt key) • you can further design the H block (see page 540 ff): here more info could be added

  47. After making changes and closing the structure window • Extend will ask how you want to save the H block, for example • in a library or only in your model:

More Related