taking the right decisions with uncertain models

Report
Nuclear emergency management:
taking the right decisions with
uncertain models
Catrinel Turcanu, Johan Camps & Benny Carlé
[email protected] / [email protected]
Society and Policy Support
Institute Environment, Health and Safety
Belgian Nuclear Research Centre
1
Workshop “All models are wrong…”, Groningen, 14-16/03/2011
Outline
 Context
 Nuclear emergency
model uncertainties
 Conclusions
Model for evaluation of nuclear emergencies
for the Doel NPP site
2
Nuclear emergencies & models
Use of
models
for the
protection
of people
in
emergency
situations
&
preparedness
phase:
 Atmospheric transport and dispersion models
(concentrations, deposition)
 Dose models (dose adults, children, thyroid, .)
 Food models (concentrations, dose)
…
3
The modelling problem:
Inversion layer
Height-dependent
wind velocity
Atmospheric
turbulence
rain
Dry
deposition
Wash-out
Irradiation
Inhalation
Ingestion
4
Irradiation
shielding
How are decisions taken?
Legislation
 Reference band of dose values – calculated based
on model predictions (or measurements or both)
 Action levels on specific actions (Belgian levels)
14/03/2011
Action
General sheltering
(up to 24h)
Indicative ranges for intervention
(dose – mSv)
5-15 (effective dose)
10-50 (thyroid equivalent dose)
Stable iodine prophylaxis
General evacuation
50-150 (effective dose integrated 1
week)
 New recommendations: 20-100 mSv/y, all pathways
 Range ≠ uncertainty !
5
Uncertainties
 Modelling assumptions
 Simplifications of reality
 Parameter uncertainty
 Calibration of model parameters
 Input data
 Meteorology
 Source term
6
Uncertainties from modelling
assumptions
Average plume
simple
Meandering plume
Conservative
calculation
Best
estimate
Fluctuating plume
complex
Source models:
Torben Mikkelsen, Risø
7
Model intercomparison
Standard conditions
Experimentally validated: factor 2-3 within experiments
8
Model intercomparison
Very specific conditions
Scenario 4
9
10
8
-3
TICair (Bq s m )
10
7
10
6
10
SCK•CEN_A
SCK•CEN_B
HOTSPOT
RODOS (Atstep)
ARGOS (Rimpuff)
Rimpuff (Ismode=1)
NPKPuff4
5
10
4
10
0
5
10
15
Distance (km)
9
20
25
Model intercomparison
realistic scenario
Noodplan Doel
JRODOS
Rimpuff (ARGOS)
TIC [Bq s/m3]
same color scale
10
Potential problems related to the
resolution of the calculation grid
For the same scenario:
Inner grid cell: 1 km
 no sheltering
Inner grid cell: 100m
 sheltering
>1km
11
11
Parameter uncertainty:
Cs-137 in milk
1000
Messungen
Bq/kg
100
Modell-Vorhersage
10
1
0,1
0
days after
200
400deposition
600
800
1000
Zeit in Tagen nach Deposition (April 30, 1986)
12
1200
Source: Florian Gering
Model uncertainties in radiological
assessments
Malcolm Crick, IAEA
13
Source: Malcolm Crick
Uncertainties in input data (1):
meteorology
Source: Marc de Cort
14
Uncertainties in the input data (2)
with on-site single rain
gauge data
with multiple rain
gauge data
With rain-radar data
Example: Input data precipitation
15
Effect of conservative approach for
treatment of rain
10
Scenario 2
10
9
-2
Ground deposition (Bq m )
10
8
10
7
10
SCK•CEN_A
SCK•CEN_B (depletion off)
SCK•CEN_B (depletion on)
HOTSPOT
RODOS (Atstep)
RODOS (Rimpuff)
ARGOS (Rimpuff)
NPKPuff4
6
10
5
10
0
10
20
16
30
Distance (km)
40
50
Example: conservative approach
Tihange, core melt,
rupture primary circuit
Standard weather conditions
17
Example: conservative approach
Tihange, partial core melt,
rupture primary circuit
Standard weather conditions
18
Example: conservative approach
Tihange, core melt,
rupture primary circuit
unstable weather conditions
19
Conclusions
 Complex problem
 Often models extended beyond validated
range
 Difficult to obtain realistic uncertainties on
calculations
 Even more difficult to communicate these
uncertainties to decision-makers
 Best estimate often replaced by conservative
approach
 But … conservative estimates may lead to
unfeasible countermeasures
20

similar documents