Istituto di Scienze dell’Atmosfera e del Clima C.N.R. - Torino

D. Anfossi , S. Trini Castelli, G. Belfiore

Modelli LagrangianiModelli Lagrangiani

1. Studio e applicazione delle tecnologie consolidate a scenari reali, corrispondenti a periodi recenti e confronto dei dati modellistici con le misure effettuate dall'Agenzia Regionali per l'Ambiente. (ALPNAP)

2. Studio di nuovi approcci modellistici relativi alla chiusura delle equazioni fondamentali: momenti di ordine elevato nei modelli Euleriano e sperimentazione di nuovi approcci di tipo Lagrangiano (MSS).

3. Analisi dell'influenza dell'input meteorologico dei modelli di dispersione tramite l'interfacciamento di codici meteorologici prognostici (HARMO12).

Model system MSS

MicroSprayMicroSwift

prognostic (mass consistent) wind interpolator over complex terrain accounting for complex terrain

and buildings

LPD model derived from SPRAY; it accounts for the presence of buildings, other obstacles, complex terrain, and

possible occurrence of low wind speed

any kind of source configuration, with emission in any direction and any initial velocity

M S Sallows taking into

account:

negatively, positively or neutral emissions in presence of obstacles

dispersion of dense and/or light gas, accidental releases and possible terrorist attacks in

urban areas.

from: Venetsanos et al. (2003), Journal of Hazardous Materials

Accidental release of hydrogen in Stockholm

500 meters

plume phase

normal dispersionas passive scalar

gravity spreading

gravity spreading

Equations

energy conservation

mass conservationss

a

p uEbudt

d

2

222 bwuNBbudt

dps

a

ps

vertical momentum conservationsps

a

p ubBbwudt

d 22

Y horizontal momentum conservationaspsa

p vuEvbudt

d

2

aspsa

p uuEubudt

d

2

X horizontal momenta conservation

five unknowns p, up, vp, wp, b where:

eubE 2 entrainmentz

gN a

a

2

a

aegB

Uauu se 21 entrainment velocity

222

ppps wvuu

222

aaaa wvuU where

and 1.0 6.0

Plume spread at ground

When a dense plume reaches the ground an horizontal momentum is generated by the weight of the plume

itself that tends to spread the plume

This heavy gas induced outflow velocity depends on the bulk properties of the dense plume, i.e. how

density varies in 3 D and requires all the particles position to be accounted for

Thus the movement of each particle depends on the characteristics of the ‘ensemble’

A hybrid algorithm used

To each particle is assigned an horizontal speed

bulkbulkg HgU 2

with:a

abulkbulk

bulk= ‘bulk’ density of the plume above the particle

H = ’bulk’ height of the plume above the particle

direction of the spread

cosggs UU

singgs UV where is randomly picked from a uniform [0°-360°] distribution

is chosen at emission time and kept by the particle

How to compute Hbulk and bulk ?

1) when a particles ‘reaches’ the ground 2) all the particles pi belonging to a dx dy column are accounted for

np

ipbulk i

znp

H1

1

np

ipbulk inp 1

1

Trial 8 Instantaneous Release – Downrange

Courtesy of Dr. Jim McQuaid

Trial 8 Instantaneous Release – Overhead

Courtesy of Dr. Jim McQuaid

SPRAY

HARMO12-2008

Chlorine accident - Macdona, TX, USA

June 28, 2004 two trains collision

Picture from Railroad Accident Report NTSB/RAR-06/03

Left graph refers to the Cl2 concentration versus distance; right graph plots Cl2 cloud width and height, both to the model-simulated concentration of 20 ppm versus

distance

MSS results compared to the Macdona accident simulations from six widely-used models (Hanna, 2007). Continuous lines indicate present results, vertical bars show

the variability (max, min) of the six models, circles locate their median

50 m

23.3 m26.2 m

47 m

MERCURE - MSS

Building

•Xo = 50 m from release

•H = 47 m, Lx =23.3 m, Ly = 26.2 minitial momentum directed vertically, w = 1.14 m/s

emission height = 10 m

initial density ratio (plume/air) = 2.0

initial emission diameter = 2.17 m

gas emission rate = 10 kgs-1

neutral stratification, logarithmic wind profile

low wind at 10 m = 1.5 m/s

higher wind at 10 m = 5 m/s 2 flow regimes

MSS

MERCURE Wind(z=10m) = 1.5 m/s

Iso-surface 0.01 kg/kg

MERCURE

MSS

Wind(z=10m) = 5.0 m/s

Iso-surface 0.01 kg/kg

MERCURE

MSS

Wind(z=10m) = 5.0 m/s

Vertical cut

1.E+02

1.E+03

1.E+04

1.E+05

1.E+06

1.E+02 1.E+03 1.E+04 1.E+05 1.E+06

Observed Concentration (ppm)

Pre

dic

ted

Co

nc

en

tra

tio

n (

pp

m)

ti6ti7ti8ti9ti12ti13ti18ti19tc45tc47bu2abu2ibu3abu3ibu4abu4ibu5abu5ibu6abu6ibu7abu7ibu8abu8ibu9abu9ico3aco3ico5aco5ico6aco6i

MicroSpray dense-gas model evaluation

(TIsland-Ist + TIsland-Cont + Burro+Coyote)

SCIPUFF dense-gas model evaluation

Paris 16th October 2007

HARMO12-2008

Plot plan of the Kit Fox site

consists of 52 trials

with CO2 gas releases, puffs & continuous plumes 

samplers (one reading per second) were installed at four arcs:

25, 50, 100, and 225 m

8 masts with wind speed measurements

HARMO12-2008

Validation of MSS against Kit Fox field data

Statistical evaluation of comparisons between observations and predicted data includes: geometric mean bias (MG),

geometric variance (VG)

factor of 2 (FA2)

Kit Fox experiment

Overall URA Continuous

URAPuff

ERP Continuous

ERPPuff

52 experiments

12 experiments

21 experiments

6 experiments

13 experiments

MG 1.04 1.42 0.95 1.19 0.87

VG 1.20 1.20 1.15 1.25 1.29

FA2 88 % 92 % 99 % 83% 83 %

CONCLUSIONS

Preliminary results suggest that MicroSpray

performs correct and fast simulations of dense gas dispersion in real field situations.

Computation timeType of model

Some hoursCFD model

Few minutesMicroSpray

RAMS-MIRS configuration

Example of a typical configuration for a simulation of the meteo fields using the prognostic code RAMS up to 1 km resolution, 4 nested domains

grid 1: 64 km horizontal resolutiongrid 2: 16 km horizontal resolutiongrid 3: 4 km horizontal resolution grid 4: 1 km horizontal resolution

Vertical grid: vertical stretched layers, 0 –15/20000 m, first layer 50 m depth (first level at ~25 m)

RAMS is initialised with the ECMWF (0.5o lat/lon) analysis fields.

Nudging at the lateral boundaries of the outer grid every 6 hours.

Mesoscale

Regional to local scale

Downscaling from RMS to MINERVE

mass consistent model

Simulation of the meteo fields using the diagnostic code MINERVE up to ~ 100 m resolution, in subdomainstypically 10-20 km x 10-20 km sizeMINERVE gets as input the hourly RAMS 3D gridded dynamical and thermal fields and…

- interpolates the mean input fields on its 3D computational domain-performs and objective analysis: application of mass conservation in every domain cell

Regional scale

Local scale

Advantages of RAMSMINERVE downscaling:

-possibility of including local measurements-possibility of including more detailed topograhy data

An example of how RAMS_MIRS + MINERVE

works for wind field in complex terrain

from ALPNAP Alpine Space Project

RAMS MINERVE

For its nature, MINERVE is not designed to account for the prognostic turbulence fields, and the Lagrangian turbulent variables are thus calculated in SPRAY from parameterisations defined for flat terrain (ex. Hanna, 1982).

In this work we investigate whether a proper interpolation from the coarser-resolution prognostic 3D-gridded turbulence fields, like diffusion coefficients, turbulent kinetic energy and its dissipation, might be used in complex and inhomogeneous terrain.

In this way, the shortcoming of using parameterised turbulent fields might be overcome by coupling MINERVE with a module, which calculates the turbulence fields on the high-resolution diagnostic grid by interpolating from the coarser prognostic grid.

What is this work about

What we compare here

RAMS is run with four nested grids, where the third (G3) and the fourth (G4) grids have respectively 1 km and 250 m resolution.

RAMS fields on G4 at 250 m are considered the ‘truth’ versus which to test other two combinations.

The G3 turbulence fields from the 1-km grid are bilinearly interpolated on the 250-m mesh points, originating the turbulence dataset G3_INTP to be checked as an alternative to flat-terrain parameterisations.

A downscaling of the mean flow to 250 m with MINERVE, using in input the 1-km resolution grid RAMS G3 fields, is done. MINERVE wind fields at 250 m are then used to calculate the surface layer and boundary layer parameters entering the turbulence calculation in the standard configuration, that is applying the Hanna (1982) parameterisation

We consider three different turbulence closure schemes in RAMS……

The MY 2.5 scheme (as in RAMS)

Vertical diffusion coefficients from the TKE equation in boundary layer approximation:

with

502

20

.xhminhorizm S)xC(,KmaxK

Horizontal diffusion coefficients from the deformation scheme as in El-anis…

340750 /Ahmin xK.K with

εPz

EK

zdt

dEE

21

EE E)(lSK 2

21

mm E)(lSK 2 1

2

32E

l

kz+

kzl

1

dzE

dzEzal

are functions depending on the set of empirical constants = (0.92, 16.6, 0.74,

10.1, 0.08) and on the shear and buoyancy terms (ref. to Mellor and Yamada (1974,1982)).

Em S,S C,B,A,B,A 2211

Closure length scales: lB,A,B,A,l,,l 22112211

The turbulence closures used in RAMS_MIRS

The EL_(iso)anis scheme

Vertical diffusion coefficients from the 3D TKE (E) equation:

εPjx

EK

jxdt

dEE

lEcK /m

21

d

/

l

Ec 23

c c E empirical coefficients

mEE KK with

502

20

.xhminhorizm S)xC(,KmaxK

Horizontal diffusion coefficients from a deformation scheme

340750 /Ahmin xK.K with

ρ0 air density, Cx dimensionless coefficient, Δx grid spacing

S2 horizontal strain rate, KA user-specified coefficient of order 1.

lld

l

kz+

kz

1

dzE

dzEzal

The turbulence closures used in RAMS_MIRS

The case considered

Susa

Point at 970 m

Grid 4

Grid 3

Torino

North-West Italian Alpine region around Torino

Altitudes……

G4 970 m

G3, 4 points:

NW 772 mNE 598 m SE 780 mSW 939 m

Dashed blue: values interpolated from Grid 3Solid orange: values calculated on Grid 4

Distributions of TKE for G3_INTP and G4 values (h < 1450 m)

MY2.5 EL_iso EL_anis

Red: RAMS G3_intpBlue: RAMS G4Green: (RAMS G3 mean flow ) MINERVE+Hanna

A critical case in complex terrain, TKE

MY 2.5 EL-anis15 GMT

Conclusions on the turbulence closure analisys

Interpolated values of TKE from 1 km resolution grid (G3_INTP) result to be overall representative of the TKE

values simulated on a 250 m grid (G4).

The spread between the two sets of TKE values, G3_INTP and G4 are probably mainly due to the fact

that the G3 points, on which the interpolation procedure is applied, may be characterized by even significantly

different altitudes

The methodology seems to be feasible, also in complex terrain and in critical locations and asks for further

investigation.

Grazie per la Grazie per la vostra vostra

atattenzionetenzione