Titre de la présentation

Titre de la présentation

Comparison of ensembles of atmospheric dispersion simulations Lessons learnt from the CONFIDENCE project about uncertainty quantification 5th NERIS workshop 5 April 2019 Session: Challenges in estimating the source term and operational radiological picture (on-site versus off- IRSN/FRM-414 ind 3 site) MEMBER OF Spyros Andronopoulos, Poul Astrup, Peter Bedwell, I. KORSAKISSOK, WP1 Karine Chevalier-Jabet, Hans De Vries, Gertie Geertsema, Florian Gering, Thomas Hamburger, Heiko Klein, Susan Leadbetter, Anne members Mathieu, Tamas Pazmandi, Raphal Prillat, Csilla Rudas, Andrey Sogachev, Peter Szanto, Jasper Tomas, Chris Context Input uncertainties Uncertainty propagation Perspectives Context In case of an accidental release of radionuclides Atmospheric dispersion models are used to forecast the health and

environmental impact A tool for decision making: countermeasures (evacuation, sheltering, stable iodine intake) A tool to reconstruct the contamination events combining simulation and measurements Results are subject to many uncertainties It can include stochastic uncertainties (i.e physical randomness), epistemological uncertainties (lack of scientific knowledge), ambiguities (ill-defined meaning), value uncertainties (when the required endpoint is ill-defined), judgemental uncertainties (e.g. setting of parameter values in codes), computational uncertainties (i.e. inaccurate calculations), modelling errors (i.e. however good the model is, it will not fit the real world perfectly) We should also address social and ethical uncertainties, in the analysis of risk and in decision making I. KORSAKISSOK Comparison of ensembles of atmospheric dispersion simulations: lessons learnt from the CONFIDENCE project about uncertainty quantification 5 th NERIS workshop 5 April 2019 IRSN MEMBER OF 2 Context Input uncertainties Uncertainty propagation Perspectives The CONFIDENCE project COping with uNcertainties For Improved modelling and DEcision making in Nuclear emergenCiEs The CONFIDENCE Project will perform research focussed on uncertainties in the area of emergency management and long-term rehabilitation. It concentrates on the early and transition phases of an emergency, but considers also longer-term decisions made during these phases.

Duration 3 years: 1.1.2017 31.12.2019 31 partners from 17 European countries Budget: 6.201.026 , request to EC: 3.252.487 Part of CONCERT 7 work packages (WPs) WP1: uncertainties in the pre and early release phase (atmospheric dispersion simulations) WP2, WP3: data assimilation, measurements, radioecological models I. KORSAKISSOK Comparison of ensembles of atmospheric dispersion simulations: lessons learnt from the CONFIDENCE project about uncertainty quantification 5 th NERIS workshop 5 April 2019 IRSN WP4, WP5: stakeholders, transition phase to long-term recovery WP6: visualization and decision-making WP7: education and training MEMBER OF 3 Context Input

uncertainties Uncertainty propagation Perspectives WP1: uncertainties in atmospheric dispersion simulations 1.1 Analyzing and ranking sources of uncertainties (Lead: IRSN) 1. 2. 3. 4. Using ensemble data for meteorological uncertainties (Lead: UK MetOffice) Using meteorological measurements to reduce uncertainties (Lead: EEAE) Uncertainties related to source term (Lead: IRSN) Uncertainties related to models (Lead: PHE) 1.2 Uncertainty propagation and analysis (Lead: IRSN) 5. 6. Simulation and comparisons to observations for the Fukushima case Simulation for the synthetic European case studies 1.3 Emergency response and dose assessment 7. 8. Food chain uncertainty propagation (Lead: BfS) Recommendations and operational methodology in an emergency context (Lead: PHE) DONE 2017 DONE 2018 / IN PROGRESS I. KORSAKISSOK Comparison of ensembles of atmospheric dispersion simulations: lessons learnt from the CONFIDENCE project about uncertainty quantification 5 th NERIS workshop 5 April 2019 IRSN

IN PROGRESS 2019 MEMBER OF 4 Context Input uncertainties Uncertainty propagation Perspectives How to quantify the uncertainty of data? Model Model parameters parameters Experts judgment, literature review pdf I. KORSAKISSOK Comparison of ensembles of atmospheric dispersion simulations: lessons learnt from the CONFIDENCE project about uncertainty quantification 5 th NERIS workshop 5 April 2019 IRSN MEMBER OF 5 Context Input uncertainties Uncertainty propagation Perspectives How to quantify the uncertainty of data? Model Model parameters parameters

Using meteorological forecast ensembles pdf Input Input :: meteo meteo I. KORSAKISSOK Comparison of ensembles of atmospheric dispersion simulations: lessons learnt from the CONFIDENCE project about uncertainty quantification 5 th NERIS workshop 5 April 2019 IRSN MEMBER OF 6 Context Input uncertainties Uncertainty propagation Perspectives How to quantify the uncertainty of data? Model Model parameters parameters Past-accident analysis (Fukushima ) literature review pdf Input Input :: meteo meteo Emergency : May rely on experts judgment / ensemble of ST Input:

Input: source source term term European projects I 131 I. KORSAKISSOK Comparison of ensembles of atmospheric dispersion simulations: lessons learnt from the CONFIDENCE project about uncertainty quantification 5 th NERIS workshop 5 April 2019 IRSN MEMBER OF 7 Context Input uncertainties Uncertainty propagation Perspectives Further on input uncertainties http://www.concert-h2020.eu/en/Publications I. KORSAKISSOK Comparison of ensembles of atmospheric dispersion simulations: lessons learnt from the CONFIDENCE project about uncertainty quantification 5 th NERIS workshop 5 April 2019 IRSN MEMBER OF 8 Context Input uncertainties Uncertainty propagation

WP1 Case study: Borssele Meteorological Perspectives Indicative plume trajectories based on analysed weather between 10m and 500m, for a release at 12 UTC 11/01/17, and associated rain (cumulated on one hour). scenario Ensemble (KNMI), 10 members, 2,5 km resolution 72-hours forecast, 1-hour time step 11-13 January 2017: Easy case (established wind direction), rain Long release scenario: ensemble (FASTNET) Duration 72 hours Extracted from a database built with ASTEC severe accident code Release time 11 January at 06 UTC without uncertainties (a) 137Cs Second major release = opening of the venting containment system Aerosols are filtered for the second release Representative of model uncertainties (release phase) Short release scenario Duration 4 hours - Release time 11 January at 12 UTC 8 radionuclides, no kinetics (b) 131I Representative of uncertainties in the pre-release phase I. KORSAKISSOK Comparison of ensembles of atmospheric dispersion simulations: lessons learnt from the CONFIDENCE project about uncertainty quantification 5 th NERIS workshop 5 April 2019 IRSN

MEMBER OF 9 Context Input uncertainties Uncertainty propagation Perspectives Short release scenario Release time 11 January at 12 UTC +/- 6 hours Release height 50m +/- 50m Released quantity X [1/3, 3] Radionuclide Xe-133 I-131 Activity(Bq) 3.51E18 2.25E16 2.84E16 1.37E16 2.69E15 6.37E14 2.06E15 Participant IRSN BfS MetOffice/ PHE Number of simulations 100 (Monte Carlo) 150 I-132

Te-132 Cs-134 Cs-136 Cs-137 Ba-137m 2.78E14 Source perturbations Release height Release time Released quantity [0, 100m] uniform [-6h, 6h] uniform [1/3, 3] uniform [0m, 50m, 100m] T0 + [-6h, -3h, 0h, +3h, +6h] [50m] T0 + [-6h, 0h, +6h] 90 [x1/3, x1, x3] EEAE 50 [50m] T0 + [-6h, -3h, 0h, +3h, +6h]

MTA EK 150 [0m, 50m, 100m] T0 + [-6h, -3h, 0h, +3h, +6h] 650 RIVM DTU 10 [0m, 25m, 50m, 75m, 100m] [-6h, +6h] with a time step of 1 hour (13 steps) - I. KORSAKISSOK Comparison of ensembles of atmospheric dispersion simulations: lessons learnt from the CONFIDENCE project about uncertainty quantification 5 th NERIS workshop 5 April 2019 IRSN - MEMBER OF 10 Context Input uncertainties Uncertainty propagation Perspectives Short release scenario Endpoints: consequences computed at T0+24h Ground deposition of 137Cs and 131I Post-Chernobyl reference level: 37 kBq/m2 for 137Cs

Other levels: 10 kBq/m2 for 137Cs, 131I Effective dose and inhalation thyroid dose for 1-year old child 10, 50, 100 mSv How to use ensemble results? Deterministic: one simulation 137 Cs deposition, threshold 37 kBq/m2 Probabilistic: 137Cs ground deposition for N simulations N maps of deposition: postage stamp Median (or 25th, 75th percentile) of the N deposition maps For a given threshold t Maximum distance D above threshold N maximum distances Di above t Map of probability of exceeding t I. KORSAKISSOK Comparison of ensembles of atmospheric dispersion simulations: lessons learnt from the CONFIDENCE project about uncertainty quantification 5 th NERIS workshop 5 April 2019 IRSN Mean 361 km Range 16586 km MEMBER OF 11 Context Input uncertainties Uncertainty propagation

Perspectives Cs deposition (kBq/m2) at T0+24h - UK MetOffice 137 Short release: postage stamp T0 = 06:00 UTC T0 = 18:00 UTC Different meteorological fields T0 = 12:00 UTC I. KORSAKISSOK Comparison of ensembles of atmospheric dispersion simulations: lessons learnt from the CONFIDENCE project about uncertainty quantification 5 th NERIS workshop 5 April 2019 IRSN MEMBER OF 12 Context Input uncertainties Uncertainty propagation Perspectives Short release: probability maps Maps of probability of threshold exceedance For a threshold of 37 kBq/m for the 137Cs deposition Example of UK MetOffice (NAME model) Mean 557 km Range 517585 km 10 simulations (meteorology only) Mean 361 km

Range 16586 km 90 simulations (meteorology + source term) With source perturbations Maximum distance of threshold exceedance is lower Surface covered by low probabilities is larger I. KORSAKISSOK Comparison of ensembles of atmospheric dispersion simulations: lessons learnt from the CONFIDENCE project about uncertainty quantification 5 th NERIS workshop 5 April 2019 IRSN MEMBER OF 13 Context Input uncertainties Uncertainty propagation Short release: median of (kBq/m2) RIMPUFF Gaussian puff BfS SINAC Gaussian puff Cs deposition 137 RIMPUFF Gaussian puff DTU NAME Lagrangian MetOffice Perspectives

DIPCOT Gaussian puff EEAE NPK-puff Gaussian puff LdX Eulerian IRSN RIVM 10 simulations (meteorology only) Differences may come from: Type of models (Gaussian, Eulerian, Lagrangian) Different wet deposition schemes Diffusion coefficients MTA EK Modelling domain, interpolation I. KORSAKISSOK Comparison of ensembles of atmospheric dispersion simulations: lessons learnt from the CONFIDENCE project about uncertainty quantification 5 th NERIS workshop 5 April 2019 IRSN MEMBER OF 14 Context Input uncertainties Uncertainty propagation Perspectives Short release: box plots Maximum distance from the source (km) For a threshold of 37 kBq/m for the 137Cs deposition 10 simulations (meteorology only)

10-650 simulations (meteorology + source term) Larger variability (boxes size) with ST perturbations Inter-model variability not totally encompassed by the range of variation I. KORSAKISSOK Comparison of ensembles of atmospheric dispersion simulations: lessons learnt from the CONFIDENCE project about uncertainty quantification 5 th NERIS workshop 5 April 2019 IRSN MEMBER OF 15 Context Input uncertainties Uncertainty propagation Perspectives Long release: box plots Maximum distance from the source (km) and surface (km 2) For a threshold of 10 kBq/m for the 131I deposition Maximum distances Surfaces Inter-model variability mostly encompassed within the range of each ensemble Surfaces are less dependent on outliers and may be more reliable I. KORSAKISSOK Comparison of ensembles of atmospheric dispersion simulations: lessons learnt from the CONFIDENCE project about uncertainty quantification 5 th NERIS workshop 5 April 2019 IRSN MEMBER OF 16 Context Input uncertainties Uncertainty propagation

Perspectives Conclusions Influence of source perturbations Importance of taking into account source perturbations Larger ensembles spread More perturbations induce lower distance above a given threshold Inter-model variability Less important when overall uncertainties are larger Some models or configurations may be more appropriate to the case Part of this variability may be taken into account An uncertainty assessment with only one model will always be partial Uncertainty assessment Lower threshold induces higher distances / probability Surface above threshold (instead of distance) limits the effect of outliers Importance of choosing correctly the threshold and percentile I. KORSAKISSOK Comparison of ensembles of atmospheric dispersion simulations: lessons learnt from the CONFIDENCE project about uncertainty quantification 5 th NERIS workshop 5 April 2019 IRSN MEMBER OF 17 Context Input uncertainties Uncertainty propagation Perspectives

Uncertainties in an emergency context Our knowledge of uncertainties will always be partial Deep uncertainties, lack of information Have to tackle the main sources of uncertainties! Avoid false confidence in probabilistic results Computational time: how many members are needed to correctly represent uncertainties? How to reduce computational time? Reducing the number of members: clustering techniques, adaptive sampling Model reduction: emulators, model assumptions Adaptation to the endpoint: domain size and resolution How to include uncertainties in output products for decision makers? I. KORSAKISSOK Comparison of ensembles of atmospheric dispersion simulations: lessons learnt from the CONFIDENCE project about uncertainty quantification 5 th NERIS workshop 5 April 2019 IRSN MEMBER OF 18 Context Input uncertainties Uncertainty propagation Perspectives Next steps Case studies in progress Ensemble runs for Borssele more challenging case front passage ongoing Fukushima and Western Norway ongoing (not synchronised between participants)

Other tasks in the remaining year Food chain uncertainty propagation Recommendations and operational methodology in an emergency context Dissemination Training course (May 2019, Slovak Republic) Special CONFIDENCE session at HARMO 19th (June 2019, Belgium) Dissemination workshop (December 2019, Slovak Republic) https://www.eu-n eris.net/ Thank you for your attention! I. KORSAKISSOK Comparison of ensembles of atmospheric dispersion simulations: lessons learnt from the CONFIDENCE project about uncertainty quantification 5 th NERIS workshop 5 April 2019 IRSN MEMBER OF 19

Recently Viewed Presentations

  • Using Excel: Descriptive Statistics

    Using Excel: Descriptive Statistics

    Scatter Diagrams (Cont.) Then you will get the "Simple Scatterplot" menu. Select one variable from the list on the left for each axis. For each variable, click the purple arrow button next to the appropriate X or Y Axis for...
  • This file has the right logos!!

    This file has the right logos!!

    Authors found the structure difficult. Drummond was right about everything but . . . I moved. John Williams moved. ... "Randomization was alternated every 10 patients, such that the first 10 patients were assigned to early atropine and the next...
  • Planning your project EPF1 Aims and objectives 4.

    Planning your project EPF1 Aims and objectives 4.

    A Gantt chart is a popular project planning tool. Gantt charts illustrate the start and finish dates of the tasks that make up the project. The advantage over a basic timeline is that the chart allows tasks to be overlapped...
  • Mobile Health Provider Program - VA Mobile | VA Mobile

    Mobile Health Provider Program - VA Mobile | VA Mobile

    The VA's Mobile Health Provider Program is designed to equip care team members with mobile technology to enhance care delivery for Veterans. Launched in 2014, this program aims to improve the health of Veterans by leveraging the power of mobile...
  • The Colonies Come of Age SECTION 1 England

    The Colonies Come of Age SECTION 1 England

    NEXT 4 SECTION The Colonies and Britain Grow Apart Problems Resulting from the War • Colonists feel threatened by British troops stationed in colonies • Prime Minister George Grenville sets policies to pay war debt • Parliament passes Sugar Act...
  • ERA 6: The Development of Industrial U.S. 1870-1900

    ERA 6: The Development of Industrial U.S. 1870-1900

    Theodore Dreiser Sister Carrie (1900) factory girl's experience 18 yr. old country girl who moves to Chicago…is used by men…eventually uses men to become a Broadway actress." Progressive Journalists
  • English 299C: Film as Narrative Art - nwsccmoodledemo.com

    English 299C: Film as Narrative Art - nwsccmoodledemo.com

    English 299B: Film as Narrative Art Mr. Kelley The Seventh Seal (Ingmar Bergman, 1956) In my film, the crusader returns from the Crusades as the soldier returns from war today.
  • Mine Health and Safety Act

    Mine Health and Safety Act

    Federal Regulation Becomes a Reality 1940 to 1977 ©June 15, 2007 Dr. Bradley C Paul Myths and Realities of Regulatory Structures Myth - Federal Regulation is Better Than State Bureau of Mines Brought 30 years of knowledge and 40s brought...