A QA/QC protocol for modelling applications · Outcome: A draft QA/QC protocol to be prepared. It...

19
A QA/QC protocol for modelling applications P. Thunis & L. Tarrasón Berlin, February 2020

Transcript of A QA/QC protocol for modelling applications · Outcome: A draft QA/QC protocol to be prepared. It...

  • A QA/QC protocol for modelling applications

    P. Thunis & L. Tarrasón

    Berlin, February 2020

  • Current situation

    No prerequisites on the actual model to be used as long as the model complies with the

    Modelling Quality Objectives (MQO)

    FAIRMODE &

    CEN TC264 / WG43

    Is this sufficient to ensure quality of modelling applications?

  • 1. Do you have a documented protocol (or set of consistency checks) that you currently apply to

    your modelling chain (emissions, meteorology, air quality modelling...) to ensure that the final

    assessments are of sufficient quality? If yes, what type of protocol (or actions) do you

    apply?

    2. According to you, which are the key elements that are currently missing (or elements where

    harmonization is highly needed) that should be considered in priority in this protocol?

    3. To support the elaboration of a common QA/QC protocol, we plan to use the composite

    mapping platform as a main tool to identify the main issues, identify the inconsistencies and

    assess if the proposed solutions are efficient. In this context, do you have proposals on the

    suggested usage of this platform?

    Questionnaire sent to NCP (summer 2019)

  • 1. What type of protocol (or actions) do you apply?

    • Some protocol exist for very specific situations (e.g. industrial plant) but are not general enough

    • Lots of statistical analysis and plots but no real documented protocol

    • Model inter-comparison whenever possible

    • QA/QC procedures (e.g. checks on emissions) are existing but not a real protocol

    • Input data are fixed by law

    2. Which are the key elements that are currently missing?

    • Summary performance plots and statistics

    • Checks on emissions

    • Checks on meteorology

    • Structuring and automation of the process

    3. Use of the composite mapping platform – looking for suggestions…

    Responses to the questionnaire

  • 1. What content for the protocol? (Documentation and/or assessment oriented?)

    2. Which process for elaborating the protocol? (collect best practices?)

    Outcome: A draft QA/QC protocol to be prepared.

    It should outline best practices for QA/QC of modelling applications with an initial

    scope on assessment purposes (forecast and planning will be considered at a

    later stage)

    Madrid discussions and outcome (autumn 2019)

  • Draft document

  • • For which pollutants? NO2, PM and O3

    • For which scales? All relevant to the AAQD

    • Models? All types (Gaussian, Eulerian, Lagrangian,

    CDF…) relevant to the AAQD

    Scope (assessment purposes)

  • Documentation (overall)

    • Model details

    • Input data

    • Intended applications

    • Evaluation databases

    • Code verification

    • …

    Documentation (purpose)

    • Model resolution

    • Model calibration

    • Data requirement

    • …

    Evaluation

    (QA/QC tests)

    • Performance evaluation

    • Sensitivity analysis

    • Input data analysis

    • Model inter-comparison

    • Benchmark tests

    • …

    Scope: ensuring quality through documentation and evaluation tests

    Overall aspects Specific aspects of implementations

  • Modelling results originate from fit-for-purpose models

    We will follow the recommendations from CEN TC264/43 (reflected in

    the FAIRMODE guidance document on “MQO and benchmarking”) on

    specific aspects of the evaluation process. E.g. which measurements to

    use? how frequent should the assessment be?, what to do when

    sufficient measurements are not available?...

    We will apply the station classification and station types currently

    proposed by the IPR for evaluation (QA/QC tests)

    Requisites and assumptions

  • Three main sources:

    1. Documentation required in the context of the “Implementing Provisions

    for reporting” mechanism

    2. Documentation filed within the former “Model Documentation System”

    3. National / regional / local protocols

    a. French audit for the AASQAs (presented by F. Tognet in Madrid)

    b. Any other available source of information?

    Documentation

    Composite mapping platform

  • MDS descriptive topics Sub-topics

    Basic information

    Model name

    Model version and status

    Latest date of revision

    Contact information

    Level of knowledge needed to operate model

    Intended field of application

    Model type and dimension

    Model description summary

    Model limitations/approximations

    Resolution

    Temporal resolution

    Horizontal resolution

    Vertical resolution

    Schemes Advection & Convection

    Turbulence

    Deposition

    Chemistry

    Solution technique

    Input Availability and Validation of Input data

    Emissions

    Meteorology

    Topography

    Initial condition

    Boundary conditions

    Data assimilation options

    Output quantities

    User interface availability

    User community

    Previous applications

    Documentation status Ranking levels 1 – 5

    Validation and evaluationRanking levels 1 – 5

    Model inter-comparison

    Frequently asked questions

    Extract from Model Documentation SystemExtract from French AASQA audit

    Documentation (examples)

  • • All sources of documentation to be merged

    according to most relevant and link to

    composite mapping

    Documentation examples IPR and composite mapping

  • Main sources:

    1. Outcome of the Madrid’s discussions

    2. QA/QC protocol available from modelling groups (currently only from the

    CHIMERE group!)

    QA/QC proposed evaluation process

  • Main QA/QC protocol steps:

    STEP 1: Passing the Modelling Quality objective (MQO)

    STEP 2: How well is my model capturing the spatial variability?

    - Fulfilling CEN spatial Model Performance Criteria (MPC)

    - Incremental assessment (rural vs. urban vs. street)

    STEP 3: How well is my model capturing the temporal variability?

    For each station type, assess the seasonal, day/night and week/week-

    end behavior when appropriate

    STEP 4: Input consistency (meteorology, emissions…)

    STEP 5: Ex-post assessment (repeat step 1 on multiple cases + delta)

    STEP 6: Multi-pollutants

    QA/QC proposed evaluation (main steps)

  • MQI(i) =Oi −Mi

    )β𝑈(Oi

    FAIRMODE QA/QC evaluation (indicators)

    STEP 2: Spatial indicators STEP 3: Temporal indicatorsSTEP 1: MQO

  • FAIRMODE QA/QC evaluation (indicators)

    STEP 5: Ex-post assessment STEP 6: Multi-pollutants aspectsSTEP 4: Input consistency

    Emissions

    Meteorology

    Bound. Cond.

    Learn from the

    compared behavior of

    different pollutants

  • • Is the proposed approach to QA/QC modelling protocol appropriate? Is

    there something missing?

    • Documentation - What actions should be put in place for identifying the

    most appropriate documentation requests? Should FAIRMODE use the

    composite mapping metadata for modelling as a test? How do you

    recommend to link to the IPR?

    • Evaluation tests – Are the proposed tests adequate? Is there something

    missing? Should FAIRMODE initiate a series of feasibility exercises for

    the evaluation tests to be included in the protocol?

    Discussion in plenary

  • • Outcome from today’s plenary discussion to be incorporated in the draft

    protocol

    • First draft document of the QA/QC protocol to be circulated by March

    2020 -

    • Review from FAIRMODE participants suggestion on documentation,

    QA/QC tests… June 2020

    • Selection of documentation requirements Next technical

    • Selection of evaluation tests based on information received Next

    technical

    Timing and follow-up

  • Thank you

    © European Union 2020

    Unless otherwise noted the reuse of this presentation is authorised under the CC BY 4.0 license. For any use or reproduction of elements that are not owned by the

    EU, permission may need to be sought directly from the respective right holders.

    Slide xx: element concerned, source: e.g. Fotolia.com; Slide xx: element concerned, source: e.g. iStock.com

    https://creativecommons.org/licenses/by/4.0/