Modelling the environmental dispersion of radionuclides Jordi Vives i Batlle Centre for Ecology and Hydrology, Lancaster, October 2011

Modelling the environmental dispersion of radionuclides

Jordi Vives i Batlle

Centre for Ecology and Hydrology, Lancaster, October 2011

www.ceh.ac.uk/PROTECT

Dispersion models available in the ERICA Tool Other types of dispersion models that are

available

And, along the way…

Key parameters that drive dispersion models for radioactivity in the environment

Applicability to different scenarios/circumstances

Lecture plan


Often the receptor is not at a point of emission but is linked via an environmental pathway (dilution)

Need to predict media concentrations when (adequate) data are not available

Dispersion models are the tools required to make this connection

What reasons to use models?


Part I - Dispersion modelling in ERICA


Designed to minimise under-prediction (conservative generic assessment)

A default discharge period of 30 y is assumed (estimates doses for the 30th year of discharge)

Currently being upgraded

IAEA SRS Publication 19


Gaussian plume model version depending on the relationship between building height, HB & cross-sectional area of the building influencing flow, AB

Assumes a predominant wind direction and neutral stability class

Key inputs: discharge rate Q & location of source / receptor points (H, HB, AB and x)

Atmospheric dispersion


A Gaussian plume model for an elevated release is as follows:

C x y z

Q

u

y z H

z y y

S

z

, , exp

2 2 210

2

2

2

2

where C = the air concentration (Bq/m3) or its time integral Bq.s/m3

Q = release rate (Bq/s) or total amount released (Bq) u10 = wind speed at 10 m above the ground (m/s) z = standard deviation of the vertical Gaussian distribution (m) y = standard deviation of the horizontal Gaussian distribution (m) HS = effective release height (m) x, y, z = rectilinear coordinates of the receptors

Importance of Release Height

Basic dispersion equation

Effective stack height


a) H > 2.5HB (no building effects)

b) H 2.5HB & x > 2.5AB½(airflow in the wake zone)

c) H 2.5HB & x 2.5AB½(airflow in the cavity

zone). Two cases: source / receptor at same building surface not at same surface

(a) (b)

(c)

Not generally applicable at > 20 km from stack

Conditions for the plume


Wind speed and direction 10 minute average from 10 m wind vane & anemometer

Release height Precipitation

10 minute total rainfall (mm) Stability or degree of turbulence (horizontal and vertical

diffusion) Manual estimate from nomogram using time of day, amount

of cloud cover and global radiation level

Atmospheric boundary layer (time-dependent) Convective and or mechanical turbulence Limits the vertical transport of pollutants

Key parameters


Based on the recommendations of the Working

Group on Atmospheric Dispersion (NRPB-R91, -

R122, -R123, -R124)

Gaussian plume model

Meteorological conditions specified by: Wind speed Wind direction Pasquill-Gifford stability classification

Implemented in PC CREAM and CROM

R91 aerial dispersion model


Model assumes constant meteorological and topographical conditions along plume trajectory

Prediction accuracy < 100 m and > 20 km limited Source depletion unrealistic (deposition modelling &

transfer factors are uncertain) Developed for neutral conditions Does not include

Buildings Complex terrain e.g. hills and valleys Coastal effects

R91 - model limitations


Freshwater Small lake (<

400 km2) Large lake

(≥400 km2) Estuarine River

Marine Coastal Estuarine

No model for open ocean waters

Surface water dispersion


Based on analytic solution of the advection diffusion equation describing transport in surface water for uniform flow conditions at steady state

Processes included: Flow downstream as transport (advection) Mixing processes (turbulent dispersion) Concentration in sediment / suspended particles estimated from

ERICA Kd at receptor (equilibrium) Transportation in the direction of flow No loss to sediment between source and receptor

In all cases water dispersion assumes critical flow conditions, by taking the lowest in 30 years, instead of the rate of current flow

The only difference between RNs in predicted water concentrations as material disperses is decay by their different radiological half-lives.

Processes and assumptions


The river model assumes that both river discharge of radionuclides such as water harvesting is done in some of the banks, not in the midstream

The estuary model is considered an average speed of the current representative of the behaviour of the tides.

Condition for mixing is x > 7D and (y-y0)<< 3.7xconcentration in sediment is assumed to be concentration in water x Kd

Lz = distance to achieve full vertical mixing

• Kd = Activity concentration on sediment (Bq kg-1) xxxxxActivity concentration in seawater (Bq L-1)

Rivers and coastal waters


Assumes a homogeneous concentration throughout the water body Expected life time of facility is required as input

Small lakes and reservoirs


Simple environmental and dosimetric models as well as sets of necessary default data: Simplest, linear compartment models Simple screening approach (robust but conservative) Short source-receptor distances Equilibrium between liquid and solid phases - Kd

More complex / higher tier assessments: Aerial model includes only one wind direction Coastal dispersion model not intended for open waters e.g.

oil/gas marine platform discharges Surface water models assume geometry (e.g. river cross-

section) & flow characteristics (e.g. velocity, water depth) which do not change significantly with distance / time

End of pipe mixing zones require hydrodynamic models

Limitations of IAEA SRS 19


Part II: PC CREAM as a practical alternative for dispersion modelling


Consequences of Releases to the Environment Assessment Methodology

A suite of models and data for performing radiological impact assessments of routine and continuous discharges

Marine: Compartmental model for European waters (DORIS)

Seafood concentrations => Individual doses => Collective doses.

Aerial: Radial grid R-91 atmospheric dispersion model with (PLUME) with biokinetic transfer models (FARMLAND)

Ext. & internal irradiation => foodchain transfer (animal on pasture e.g. cow & plant uptake models) => dose

Collective dose model PC CREAM


Compartmental - marine model

(continuous discharge)

Radial grid - atmospheric model

Sellafield

Dundalk

Dublin

Irish SeaNorth West

Irish SeaNorth

Irish SeaNorth East

Irish SeaSouth East

Liverpool And Morecambe Bays

CumbrianWaters

Irish SeaWest

Localcompart.

Marine and aerial dispersion


Marine model (DORIS) => improvement Has long-range geographical resolution Incorporates dynamic representation of water /

sediment interaction Aerial model (PLUME) => no improvement

Still a gaussian dispersion model unsuitable for long distances > 20 km

Also assumes constant meteorological conditions Does not correct for plume filling the boundary layer

Degree of improvement of the models


Part III: Other alternative dispersion models


Uncertainty associated with the application of aquatic SRS models: Models generally conservative. From factor of 2 to 10 difference with respect to a dynamic

model. Uncertainty associated with the application of a

Gaussian plume model for continuous releases: About a factor of 4 or 10 for a flat and complex terrain

respectively. At distances < 2.5 times the square root of the frontal area

of the building, the model provides conservative results. For distances of about 2.5 the above, the model tends to

underpredict for wind speeds above 5-m s-1.

Effects of using different models


For aerial, PC-Cream is no improvement to SRS 19 For marine, PC cream has a dynamic compartment

model Effect of using such a fully dynamic model:

In periods where concentrations in compartments increase, dynamic model estimates of transfer will be lower than for equilibrium model (‘build-up effect’)

In period where environmental concentrations decrease, dynamic model estimates higher than equilibrium model (‘memory effect’)

Diffcult to generalise, but differences could be up to a factor of 10.

Effects of using different models (2)


Aerial modelling


Include deviations from idealised Gaussian plume model

Include turbulence data rather than simplified stability categories to define boundary layer

Include particulate vs gases and chemical interactions Model includes the effects on dispersion from:

Complex buildings Complex terrain & coastal regions

Advanced models: ADMS, AERMOD Gaussian in stable and neutral conditions Non-Gaussian (skewed) in unstable conditions

New-generation plume models


Modified Gaussian plume model Gaussian in stable and neutral conditions Skewed non-Gaussian in unstable conditions

Boundary layer based on turbulence parameters Model includes:

Meteorological preprocessor, buildings, complex terrain Wet deposition, gravitational settling and dry deposition Short term fluctuations in concentration Chemical reactions Radioactive decay and gamma-dose Condensed plume visibility & plume rise vs. distance Jets and directional releases Short to annual timescales

UK ADMS


Stack with building

0 100 200 300 400 500 600 700 800 900 1000

M etres

NOx Concentration (ug/m 3)

-400

-300

-200

-100

0

100

200

300

400

Met

res

23456789101112131415161718192021222324

Concentration plot

Types of output


Marine modelling


Allow for nonequilibrium situations e.g. acute release into protected site

Advantages: Resolves into a large geographical range Results more accurate (if properly calibrated)

Disadvantages: Data and CPU-hungry (small time step and grid sizes

demand more computer resources) Run time dependent on grid size & time step Requires specialist users Post-processing required for dose calculation (use as input

to ERICA)

Geographically-resolving marine models


Input requirements: Bathymetry, wind fields, tidal velocities, sediment distributions, source term

Type of output: a grid map / table of activity concentration (resolution dependent on grid size)

All use same advection/dispersion equations, differences are in grid size and time step

Types of model: Compartmental: Give average solutions in compartments

connected by fluxes. Good for long-range dispersion in regional seas.

Finite differences: Equations discretised and solved over a rectangular mesh grid. Good for short-range dispersion in coastal areas

Estuaries a special case: Deal with tides (rather than waves), density gradients, turbidity, etc.

Model characteristics


Finite differences Compartmental



Long-range marine models (regional seas): POSEIDON - N. Europe (similar to PC-CREAM model but

redefines source term and some compartments - same sediment model based on MARINA)

MEAD (in-house model available at WSC) Short-range marine models (coastal areas):

MIKE21 - Short time scales (DHI) - also for estuaries Delft 3D model, developed by DELFT TELEMAC (LNH, France) - finite element model COASTOX (RODOS PV6 package)

Estuarine models DIVAST ( Dr Roger Proctor) ECoS (PML, UK) - includes bio-uptake

Some commonly available models


Two-dimensional depth averaged model for coastal waters

Location defined on a grid - creates solution from previous time step

Hydrodynamics solved using full time-dependent non-linear equations (continuity & conservation of momentum)

Large, slow and complex when applied to an extensive region

Suitable for short term (sub annual) assessments

A post processor is required to determine biota concentrations and dose calculations

DHI MIKE 21 model


2 km grid

Applies advection - dispersion equations over an area and time

Generates activity concentration predictions in water and sediment

Has been combined with the ERICA methodology to make realistic assessments of impact on biota

Marine Environmental Advection Dispersion (MEAD)


Distribution of fine grained bed sediment

Distribution of suspended particles (modelled)

MEAD input data - sediment


020406080

100120140160180200

1950 1960 1970 1980 1990 2000 2010

Model predictions

Measured activity in crabs

0

5

10

15

20

25

30

35

40

1950 1960 1970 1980 1990 2000 2010

Model predictionsMeasured activity in winkles

0

100

200

300

400

500

600

700

800

900

1950 1960 1970 1980 1990 2000 2010

Model predictions

Measured activity in mussels

0

200

400

600

800

1000

1200

1400

1950 1960 1970 1980 1990 2000 2010

Model predictionsMeasured activity in codMeasured activity in plaice

60Co in winkles 137Cs in cod / plaice

99Tc in crab 241Am in mussels

Could be used to derive CFs for use in ERICA

MEAD output - Cumbrian coast


-6.0 -5.5 -5.0 -4.5 -4.0 -3.5 -3.053.0

53.5

54.0

54.5

55.0

0.00

0.01

0.10

1.00

5.00

10.00

50.00

100.00

1000.00

-6.0 -5.5 -5.0 -4.5 -4.0 -3.5 -3.053.0

53.5

54.0

54.5

55.0

0.00

0.01

0.10

1.00

5.00

10.00

100.00

1000.00

5000.00

Cumbria fishingarea

-6.0 -5.5 -5.0 -4.5 -4.0 -3.5 -3.053.0

53.5

54.0

54.5

55.0

0.00

0.01

0.10

1.00

5.00

10.00

50.00

100.00

1000.00

-6.0 -5.5 -5.0 -4.5 -4.0 -3.5 -3.053.0

53.5

54.0

54.5

55.0

0.00

0.01

0.10

1.00

5.00

10.00

100.00

1000.00

5000.00

Cumbria fishingarea

Predicted distribution of 137Cs in seawater in 2000

Predicted distribution of 137Cs in bed sediments in 2000

MEAD - Long-range results


Extra modules for extra processes More complex water quality issues e.g.

eutrophication Wave interactions Coastal morphology Particle and slick tracking analysis Sediment dynamicsD isso lv e d

Su sp e n d e dSe d im e n t

A d so rp tio n

D e so rp tio nR e su sp e n sio n

D e p o s itio n

A d v e c tio n

D if fu s io n

D if fu s io n

D e so rp tio n

A d so rp tio n

D e p o site dSe d im e n t

A d v e c tio n

Discharge

ModelMaker biokinetic models

Dynamic interactions with sediment

SpeciationDynamic uptake in

biota

Seawater_Pu_III_IVSeawater_Pu_V_VI

Suspended_load

Reduction

Oxidation

Desorption_sed Adsorption_sed

Adsorption_coll

Adsorption_susp

DepositionRemobilisation Coagulation

Desorption_susp

SedimentationBioturbation

Flushing_oxidisedFlushing_reduced

Flushing_colloidal

Flushing_susp

Burial

Discharge

Influx_Pu_III_IV

Influx_Pu_V_VI

Influx_Pu_particulate

Surface_sediment

Bottom_sedimentMiddle_sediment

Far_field

Colloidal

Pelagic_fish

Crustaceans

Molluscs

Benthic_fish

More complex process models


River and estuary modelling


Advantages: Large geographical range Consider multiple dimensions of the problem (1 - 3D) Considers interconnected river networks Results more accurate (if properly calibrated)

Disadvantages - same as marine models: Data hungry Run time dependent on grid size & time step Requires a more specialised type of user CPU-hungry (as time step and grid size decreases it demands

more computer resources) Post-processing required for dose calculation (use as input to

ERICA)

River and estuary models


Input requirements: Bathymetry, rainfall and catchment data, sediment properties, network mapping, source term

Type of output: activity concentration in water and sediment, hydrodynamic data for river

All use same advection/dispersion equations as marine but differences in boundary conditions

Generally models solve equations to: Give water depth and velocity over the model domain. Calculate dilution of a tracer (activity concentration)



Can be 1D, 2D or 3D models 1D river models: River represented by a line in

downstream direction - widely used 2D models have some use where extra detail is required 3D models are rarely used unless very detailed process

representation is needed Off-the-shelf models:

MIKE11 model developed by the DHI, Water and Environment (1D model)

VERSE (developed by WSC) MOIRA (Delft Hydraulics)

Research models: PRAIRIE (AEA Technology) RIVTOX & LAKECO (RODOS PV6 package)

Common models


MIKE11 - Industry standard code for river flow simulation

River represented by a line in downstream direction

River velocity is averaged over the area of flow

Cross sections are used to give water depth predictions

Can be steady flow (constant flow rate) or unsteady flow

Use of cross sections can give an estimate of inundation extent but not flood plain velocity

Example - MIKE 11


Convert rainfall over the catchment to river flow out the catchment

Represent the processes illustrated, however in two possible ways: Simple “black box” type model such as empirical

relationship from rainfall to runoff (cannot be used to simulate changing conditions)

Complex physically based models where all processes are explicitly represented

Example: DHI MIKE-SHE

Catchment modelling


ERICA uses the IAEA SRS 19 dispersion models to work out a simple, conservative source - receptor interaction

SRS 19 has some shortcomings PC-CREAM can be used as an alternative to the SRS-

19 marine model There are further off-the-shelf models performing

radiological impact assessments of routine and continuous discharges ranging from simple to complex

Key criteria of simplicity of use and number of parameters need to be considered – must match complexity to need

Conclusions


Summary of key pointsSRS19 model PC Cream Orther modelsMarine DORIS Marine + point in coast + Large compartment box model + compartmental models for large areas

+ requires very few parameters + Dynamic transfer to water and sediments

+ Grid models for fine resolutions (small areas)

- no offshore dispersion - requires more parameters + Dynamic / time-variable discharges - very simple equilibrium model (Kd based)

- Does not work well at fine resolution

- parameter hungry (bathimetry, gridding, etc)

River, lake, reservoir N/A River, lake, reservoir + very simple 1D model + 2D - 3D models - only models riverbanks + Full representation of hydrodynamics

- Simple average flow conditions + Can deal with tides, concentration gradients

- very simple equilibrium model (Kd based) + Dynamic / time-variable discharges - Simple linear river + Complex river networks

- parameter hungry (bathimetry, gridding, etc)

Aerial PLUME AERMOD, ADMS, etc. + limited range 100 m to 20 km - Same as SRS19 + Non Gaussian for unstable conditions + constant meteorology + Buildings and terrain + Gaussian plume, still conditions + Solute modelling + Complex meteorology


Model Organisation LinkADMS 4 CERC http://www.cerc.co.uk/environmental-software/ADMS-model.htmlAERMOD EPA http://www.epa.gov/scram001/dispersion_prefrec.htm#aermod

(Freeware)DELFT 3D DELFT

Hydraulicshttp://delftsoftware.wldelft.nl/index.php?option=com_content&task=blogcategory&id=13&Itemid=34

DIVAST CardiffUniversity

http://hrc.engineering.cf.ac.uk/

EcoS 3 PML http://www.pml.ac.uk/

HEC-RAS HEC(USACE)

http://www.hec.usace.army.mil/software/hec-ras/hecras-download.html (Freeware)

IAEA SRS 19 IAEA www-pub.iaea.org/MTCD/publications/PDF/Pub1103_scr.pdfISIS Halcrow http://www.halcrow.com/isis/isisfree.asp (Freeware)

http://www.halcrow.com/isis/default.asp (Professional edition)MEAD WSC http://www.westlakes.org (in-house model)MIKE11 DHI http://www.dhigroup.com/Software/WaterResources/MIKE11.aspxMIKE21 DHI http://www.dhigroup.com/Software/Marine/MIKE21.aspxMIKE3 DHI http://www.dhigroup.com/Software/Marine/MIKE3.aspxMIKE-SHE DHI http://www.dhigroup.com/Software/WaterResources/MIKESHE.aspxMOIRA-PLUS EU MOIRA

programmehttp://user.tninet.se/~fde729o/MOIRA/Software.htm (Freeware)

PC CREAM 08 HPA http://www.hpa.org.uk/web/HPAweb&HPAwebStandard/HPAweb_C/1195733792183

POSEIDON CEPN http://www.cepn.asso.fr/en1/logiciels.htmlPRAIRIE AEA

Technologyhttp://www.aeat.co.uk/

R91 NRPB http://www.admlc.org.uk/NRPB-R91.htmRODOS PV6(COASTOX,RIVTOX &LAKECO)

EU RODOSprogramme

http://www.rodos.fzk.de/rodos.html (Freeware, password protected)

TELEMAC 2 &3D

SOGREAH http://www.telemacsystem.com/index.php?option=com_jdownloads&Itemid=31&task=viewcategory&catid=3&lang=en (Freeware)

VERSE WSC http://www.westlakes.org (in-house model)

Links to alternative models

Documents

Modelling the environmental dispersion of radionuclides Jordi Vives i Batlle Centre for Ecology and Hydrology, Lancaster, October 2011