EURAD-2 Deli e able 16.1 – S a e-o - he-a on high ideli y nume ical simula ions o s ongly coupled
p ocesses o eposi o y sys ems (Ini ial epo )
Dissemina ion le el: Public
Da e o issue o his epo : 29/09/2025 1
Deli e able 16.1: S a e-o - he-a on high ideli y
nume ical simula ions o s ongly coupled
p ocesses o eposi o y sys ems
(Ini ial epo )
Wo k Package 16
Co- unded by he Eu opean Union unde G an Ag eemen n°101166718
EURAD-2 Deli e able 16.1 – S a e-o - he-a on high ideli y nume ical simula ions o s ongly coupled
p ocesses o eposi o y sys ems (Ini ial epo )
Dissemina ion le el: Public
Da e o issue o his epo : 29/09/2025 2
Documen in o ma ion
P ojec Ac onym
EURAD-2
P ojec Ti le
Eu opean Pa ne ship on Radioac i e Was e Managemen -2
EC g an ag eemen No.
101166718
Wo k Package Ti le
High ideli y nume ical siMula ions o s ongly coupled p ocesses
o Eposi o y syS ems (HERMES)
Deli e able No.
16.1
Deli e able Ti le
S a e-o - he-a on high ideli y nume ical simula ions o s ongly
coupled p ocesses o eposi o y sys ems (Ini ial epo )
Lead Bene icia y
PSI / SCK-CEN
Con ac ual Deli e y Da e
Sep embe 2025
Ac ual Deli e y Da e
Sep embe 2025
Dissemina ion le el
Public
Au ho s
E ic Laloy (SCK-CEN), Se gey Chu ako (PSI), A ila Baksay (TS
Ene con), Jan B ezina (TUL), Anne-Ca he ine Dieudonné (TUDel ),
Ola Koldi z (UFZ & DUT), Nikolaos P asianakis (PSI), Ja ie
Sampe (UDC), Luis Mon eneg o (UDC), Alba Mon (UDC), Ryan
San oso (FZJ), Yuankai Yang (FZJ), Jenna Poonoosamy (FZJ),
Gesa Zie le (BGR), F ancesco F eddi (UNIPR), Lo enzo Mingazzi
(UNIPR), Xa i Pin ado (Mi a Enginee ing Oy), Lau a Asensio
(UCLM), Vicen e Na a o (UCLM), Romin Dagnelie (CEA Pa is-
Saclay), Clémen Lynde (CEA Pa is-Saclay), Anne-Julie Tine (UL),
Fab ice Gol ie F (UL), Ma co De Lucia (GFZ), Guangjing Chen (SC
CEN), E san Demi e (Amphos 21), Jan S ebel (TUL), Michal Bé eš
(IGN), Zol án Bő hi (WSP Hunga y Consul ing Z ), Tamás Olasz
(WSP Hunga y Consul ing Z ), Jian u Shao (LaMcube), Zhan Yu
(LaMcube)
To be ci ed as:
Laloy E., Chu ako S. e al. (2025): S a e-o - he-a on high ideli y nume ical simula ions o s ongly
coupled p ocesses o eposi o y sys ems (Ini ial epo ). Final e sion as o 29/09/2025 o deli e able
D16.1 o he Eu opean Pa ne ship EURAD-2. EC G an ag eemen n°:101166718.
Disclaime
All in o ma ion in his documen is p o ided "as is" and no gua an ee o wa an y is gi en ha he
in o ma ion is i o any pa icula pu pose. The use , he e o e, uses he in o ma ion a i s sole isk
and liabili y. Views and opinions exp essed a e howe e hose o he au ho (s) only and do no
necessa ily e lec hose o he Eu opean Union o Eu opean A omic Ene gy Communi y. Nei he he
Eu opean Union no he g an ing au ho i y o he indi idual Colleges o EURAD-2 can be held
esponsible o hem.
Acknowledgemen
This documen is a deli e able o he Eu opean Pa ne ship on Radioac i e Was e Managemen 2
(EURAD-2). EURAD-2 is co- unded by he Eu opean Union unde G an Ag eemen N° 101166718.
EURAD-2 Deli e able 16.1 – S a e-o - he-a on high ideli y nume ical simula ions o s ongly coupled
p ocesses o eposi o y sys ems (Ini ial epo )
Dissemina ion le el: Public
Da e o issue o his epo : 29/09/2025 3
S a us o deli e able
By
Da e
Deli e ed (Lead Bene icia y)
SCK-CEN
16/07/2025
Ve i ied (WP Leade )
S. Chu ako
18/07/2025
Re iewed (Re iewe s)
S. Chu ako , O. Koldi z,
N. P asianakis, A. Baksay
30/06/2025
Re iew (KM WPL)
A. Ta omi
25/09/2025
App o ed (PMO)
V. Ma uzas
26/09/2025
Submi ed o EC (Coo dina o )
And a
01/10/2025
EURAD-2 Deli e able 16.1 – S a e-o - he-a on high ideli y nume ical simula ions o s ongly coupled
p ocesses o eposi o y sys ems (Ini ial epo )
Dissemina ion le el: Public
Da e o issue o his epo : 29/09/2025 4
Execu i e Summa y
Despi e he con inuous inc ease o compu a ional powe , nume ical models o was e
eposi o ies emain limi ed in dimensions, esolu ion, and p ocess coupling. In e se
modelling, such as da a in e p e a ion, design op imisa ion, and unce ain y analysis
equi e compu a ionally in ensi e i e a i e o wa d modelling o de i e op imal
pa ame e s. Bo h o wa d and in e se modelling can be g ea ly accele a ed by using
high- ideli y su oga e models. Machine lea ning-based su oga es a e especially
p omising o The mal-Hyd o-Mechanical-Chemical (T-H-M-C) model componen s,
mul iscale da a exchange, big da a educ ion, and o de i e cons i u i e ela ions om
la ge da ase s.
WP HERMES aims a he de elopmen o high- ideli y nume ical models o
simula ions o s ongly coupled T-H-M-C p ocesses in eposi o y nea ield, eposi o y
design op imisa ion, and in e p e a ion o mock-up expe imen s using a combina ion o
physics-based models and accele a ed compu ing assis ed wi h machine lea ning and
a i icial in elligence.
This ini ial S a e-o - he-A epo desc ibes he majo coupled T-H-M-C p ocesses in
geological eposi o y sys ems and he on ie o ela ed model de elopmen . Pa icula
ocus is placed on he analysis o exis ing app oaches and open esea ch ques ions
wi h espec o u he de elopmen s o coupled codes and models o ealis ic mul i-
scale simula ion o eposi o y sys ems. These include he use o machine lea ning and
a i icial in elligence o he accele a ion o compu e codes; sensi i i y analysis,
in e se modelling and op imisa ion; so wa e enginee ing and collabo a i e pla o ms
o model de elopmen . By he end o he WP-HERMES, his epo will be
complemen ed wi h he knowledge acqui ed du ing he cou se o he WP HERMES
RD&D p og am.
Keywo ds
Coupled p ocesses; nume ical simula ion; eposi o y; adioac i e was e; T-H-M-C;
Machine lea ning; A i icial in elligence.
EURAD-2 Deli e able 16.1 – S a e-o - he-a on high ideli y nume ical simula ions o s ongly coupled
p ocesses o eposi o y sys ems (Ini ial epo )
Dissemina ion le el: Public
Da e o issue o his epo : 29/09/2025 5
Con en s
Execu i e Summa y ................................................................................................................................. 4
Keywo ds ................................................................................................................................................. 4
1. Lis o igu es ................................................................................................................................... 7
2. In oduc ion o he scope o HERMES WP .......................................................................................... 8
2.1. The ole o modelling in he design and pe o mance assessmen o nuclea was e eposi o y
sys ems ................................................................................................................................................ 8
2.1.1 The wha ’s and why’s o eposi o y sys ems and coupled p ocesses .................................... 8
2.1. 2 Why nume ical models ma e ............................................................................................... 9
2.1.3 Needs o u u e de elopmen o high- ideli y nume ical modelling ools o coupled p ocesses
....................................................................................................................................................... 10
2.2. Desc ip ion o eposi o y ele an Fea u e-E en s-P ocesses (FEPs) in a gillaceous hos ocks
........................................................................................................................................................... 12
2.3. An o e iew o he s a e-o - he-a nume ical model .................................................................. 15
2.3.1 Physics and chemis y-based desc ip ion o coupled phenomena and scales .................... 15
2.3.2 Applica ion o machine lea ning ........................................................................................... 17
2.3.3 Model alida ion and benchma king ..................................................................................... 19
2.4. F on ie s o ealis ic eposi o y modelling, design and op imisa ion ........................................... 20
2.4.1 Op imisa ion, unce ain y analysis, and in e se modelling ................................................... 20
2.4.2 Abs ac ion and simpli ica ion me hods ................................................................................ 21
2.4.3 Global sensi i i y analyses ................................................................................................... 21
2.4.4 So wa e de elopmen and HPC in as uc u e .................................................................... 22
2.4.5 Fu u e o ien ed collabo a i e pla o ms ................................................................................ 23
3. La es de elopmen s and open gaps in nume ical simula ion o coupled p ocesses ....................... 25
3.1. Imp o ing models’ ideli y and compu a ional e iciency o s ongly coupled T-H-M-C p oblems
and subp oblems ............................................................................................................................... 25
3.1.1 Coupled p ocesses ............................................................................................................... 25
3.1.2 Modelling scale ..................................................................................................................... 26
3.1.3 Simula ion me hods .............................................................................................................. 28
3.1.4 Su oga e modelling ............................................................................................................. 30
3.1.5 Benchma k sui es ................................................................................................................. 39
3.2. Me hods and ools o unce ain y analysis o THCM models .................................................... 40
3.2.1 Upscaling .............................................................................................................................. 40
3.2.2 Sensi i i y analysis ............................................................................................................... 41
3.2.3 In e se modelling ................................................................................................................. 43
3.3. In eg a ion be ween da a and compu ing ools .......................................................................... 44
3.3.1 Collabo a i e pla o ms / model and da a hubs .................................................................... 44
EURAD-2 Deli e able 16.1 – S a e-o - he-a on high ideli y nume ical simula ions o s ongly coupled
p ocesses o eposi o y sys ems (Ini ial epo )
Dissemina ion le el: Public
Da e o issue o his epo : 29/09/2025 6
3.3.2 Da a and model in eg a ion .................................................................................................. 47
3.4. Applica ion o p ocess-based models o ield expe imen s and eposi o y sys ems ................. 48
3.4.1 Real ime simula ion o expe imen s .................................................................................... 48
3.4.2 Was e package in eg i y and nea - ield ................................................................................ 50
3.4.3 Reposi o y design and op imisa ion and digi al wins’ de elopmen .................................... 54
4. Summa y ........................................................................................................................................... 59
5. Re e ences ........................................................................................................................................ 60
EURAD-2 Deli e able 16.1 – S a e-o - he-a on high ideli y nume ical simula ions o s ongly coupled
p ocesses o eposi o y sys ems (Ini ial epo )
Dissemina ion le el: Public
Da e o issue o his epo : 29/09/2025 7
1. Lis o igu es
Figu e 1 – Summa y o coupled p ocesses and hei app oxima e empo al ex en in a eposi o y nea -
ield loca ed in a low pe meabili y a gillaceous hos ock (middle). Semi quali a i e conside a ion o main
physical-chemical p ocesses and in si u condi ions based on a Swiss eposi o y concep o SF/HLW
(le ) and cemen i ious eposi o y o L/ILW ( igh ) (Leupin e al., 2016a, 2016b). EGTS s ands o
Enginee ed Gas T anspo Sys em. Whe eas exac du a ion o he ac i e phase o peak ield alues o
speci ic p ocesses (e.g. peak empe a u e o du a ion o e-sa u a ion phase, e c. indica ed by he colou
in ensi y) depend on de ails o eposi o y design and he p ope ies o he was e, he de ailed sequence
o he physical p ocess and hei in e dependence is cha ac e is ic o a wide ange o cu en eposi o y
concep s. ............................................................................................................................................... 12
Figu e 2 – Compu a ional app oach o ex ac ing po e-scale dynamics in o ma ion. This wo k low is
summa ised om Mah ous e al. (2022) and Maes & Menke (2021). ................................................... 32
Figu e 3 – Ske ch o he p oposed EURAD-2 model-hub. .................................................................... 45
Figu e 4 – Example o he Digi al Twin o he URL Mon Te i displaying he expe imen loca ions in he
labo a o y, embedding simula ion esul s and a se ious game o educa ional and knowledge ans e
pu poses (G aebling e al., 2023). ......................................................................................................... 47
Figu e 5 – Rep esen a ion o he laye -by-laye app oach (a e Lemmens e al. (2024)) in which only
he ou e (supe icial) laye o s eel is a po ous medium and can co ode. .......................................... 51
EURAD-2 Deli e able 16.1 – S a e-o - he-a on high ideli y nume ical simula ions o s ongly coupled
p ocesses o eposi o y sys ems (Ini ial epo )
Dissemina ion le el: Public
Da e o issue o his epo : 29/09/2025 8
2. In oduc ion o he scope o HERMES WP
1
The EURAD SRA de ines join esea ch needs whe e Eu opean collabo a ion o e s
added alue beyond wha indi idual na ional p og ams migh do (EURAD Bu eau.
(2023)). Accu a e p ocess based nume ical simula ions o coupled p ocesses has
been iden i ied as key ool o he eposi o y sa e y demons a ion and op imisa ion o
eposi o y design. In alignmen wi h he SRA’s goals o building a collabo a i e and
sus ained knowledge base, he WP HERMES is se up o suppo s op imiza ion o
eposi o y designs (e.g. layou , ma e ials, ba ie sys ems) wi h p edic i e ools.
Facili a es upscaling om lab/ ield expe imen s o eposi o y scale. P omo es
knowledge sha ing ia open modelling/da a in as uc u e, enhancing anspa ency
and ep oducibili y. Builds expe ise in ad anced modelling echniques (including ML‐
assis ed / su oga e modelling) ac oss Eu ope.
2.1. The ole o modelling in he design and pe o mance assessmen o
nuclea was e eposi o y sys ems
Sa e geological disposal o adioac i e was e elies on a combina ion o enginee ed
and na u al ba ie s ep esen ing a so-called mul iple ba ie app oach (Ap ed & Ahn,
2017). Na u al ba ie s, e.g., he hos ocks, a e chosen o p o ide s able hyd o-
chemical-geo- ec onic condi ions and o slowdown a po en ial mig a ion o
adionuclides in o he biosphe e. Depending on he he mo-hyd o-chemo-mechanical
condi ions p o ided by he hos ocks, he sys em o enginee ed ba ie s con aining
he was e is op imised o ensu e he mechanical in eg i y o was e packages, and o
delay possible elease o soluble adionuclides in o he hos ocks and biosphe e
(IAEA, 2020).
2.1.1 The wha ’s and why’s o eposi o y sys ems and coupled p ocesses
Acco ding o (IAEA, 2009), adioac i e was e is classi ied in o six ca ego ies: Exemp
was e (EW), Ve y low le el was e (VLLW), Low le el was e (LLW), In e media e le el
was e (ILW), High le el was e (HLW). Because o hei physical-chemical p ope ies
and adio oxici y le el di e en concep s a e used in design o was e eposi o ies o
Spen (nuclea ) Fuel/High Le el Was e (SF/HLW), Low Le el Was e and In e media e
Le el Was e (L/ILW) (IAEA, 2009). The ollowing discussion is mainly ocused on he
enginee ing solu ions o SF/HLW and L/ILW.
The SF/HLW is also e e ed o as hea emi ing was e. Depending on he in en o y o
spalla ion p oduc s and he p edisposal his o y o he SF, i s he mal ou pu can ha e
signi ican implica ions o he ime e olu ion o he he mo-hyd o-chemo-mechanical
1
Based on Chu ako , S.V., Cla e , F., Idia , A. e al. Posi ion pape on high ideli y simula ions o coupled p ocesses, mul i-
physics and chemis y in geological disposal o nuclea was e. En i on Ea h Sci 83, 521 (2024).
h ps://doi.o g/10.1007/s12665-024-11832-7
EURAD-2 Deli e able 16.1 – S a e-o - he-a on high ideli y nume ical simula ions o s ongly coupled
p ocesses o eposi o y sys ems (Ini ial epo )
Dissemina ion le el: Public
Da e o issue o his epo : 29/09/2025 9
condi ions in he eposi o y. The combina ion o he mal pulse, mechanical s ain, po e
p essu e build-up, solu es and mois u e anspo leads o complex ansien condi ions
(Seyedi e al., 2017). These p ocesses a e coupled o chemical g adien s and
he e ogeneous eac ion on s e ol ing in he enginee ed ba ie sys em (EBS) and
e en in he adjacen hos ocks (Bilds ein O, 2019; Leupin e al., 2016a).
L/ILW is ypically immobilised using a cemen -based ma ix (Ojo an e al., 2019). O he
binde s such as bi umen and geopolyme s ha e also been used o a e being
conside ed. Cemen i ious ma e ials a e known o hei du abili y, shielding capabili y,
mechanical pe o mance and he e o e an impo an ma e ial in many eposi o y
designs. These ma e ials ha e inhe en ly high pH, which would be in chemical
disequilib ium wi h common hos ock ypes conside ed o he disposal o adioac i e
was e (Gauche & Blanc, 2006; Wilson e al., 2021). On a imescale o se e al
hund eds o housands o yea s, a cemen i ious ba ie is expec ed o unde go complex
chemical in e ac ion wi h he su ounding hos ocks (Blanc e al., 2024 ). In addi ion,
he deg ada ion o ce ain was e ypes and componen s o he EBS may lead o gas
elease which a ec s he sa u a ion s a e o he eposi o y and he po e luid p essu e
(Le asseu e al., 2024; Mo eno e al., 2001; Polle e al., 2016; Wendling e al., 2019).
Gas elease can he e o e ha e a signi ican impac on he long- e m hyd o-chemo-
mechanical e olu ion o he ba ie sys em (Leupin e al., 2016b).
Conside ing he sho - and long- e m dynamics o he in-si u condi ions, he desc ip ion
o eposi o y sys em is di ided in o nea - and a - ield domains. Nea - ield comp ises
he was e, he enginee ed ba ie sys em and he adjacen hos ocks a ec ed by he
eposi o y. Recen ly, he e m disposal cell has been used o desc ibe he was e
package - ba ie s - adjacen hos ock; whe eas nea - ield may e e also o he
combina ion o he enginee ed eposi o y sys em and adjacen hos ock (Jacques e
al., 2024). The a - ield comp ises he dis an pa o hos ocks and he biosphe e,
which, in essence, could be conside ed o be una ec ed by he T-H-M-C phenomena
in he eposi o y e en in he long e m.
2.1. 2 Why nume ical models ma e
The pe o mance and sa e y assessmen s o he eposi o ies ely, among a ious o he
aspec s, on model-based desc ip ions and simula ions o possible eposi o y e olu ion
scena ios. Due o he complexi y o he eposi o y sys ems and he long- ime scales
in ol ed, modelling is he only way o e alua e he long- e m e olu ion o he eposi o y
in si u condi ions. Fo he same easons eliable model-based p edic ions o eposi o y
e olu ion a e challenging. The oo s o he challenges a e mul i- old and ela ed o:
EURAD-2 Deli e able 16.1 – S a e-o - he-a on high ideli y nume ical simula ions o s ongly coupled
p ocesses o eposi o y sys ems (Ini ial epo )
Dissemina ion le el: Public
Da e o issue o his epo : 29/09/2025 16
In he po e-scale app oxima ion, spa ial dis ibu ion o indi idual phases is aken in o
accoun explici ly. The phase is a solid o luid wi h dis inc chemical and physical
p ope ies. Depending on he model, hese phases could be po e space illed wi h
mul i-phase mul i-componen luids and solids. The in e ac ion be ween phases is
di ec ly de ined a phase bounda ies (in e aces) and may lead o he al e a ion o he
po e space subjec o dissolu ion/p ecipi a ion eac ions. Conside ing he ac ha any
phase may ha e in insic he e ogenei ies, he po e scale app oxima ion emba ks on
he “con inuum scale” app oach when desc ibing p ope ies o he indi idual phases in
he sys em.
In he molecula scale models, he in e ac ion be ween ions and molecules in luid and
solid phases is aken in o accoun explici ly. Once again, depending on he le el o
abs ac ion and de ail, hese models ange om he explici conside a ion o elec onic
s uc u e based quan um mechanical app oach o coa se g ained ones in which la ge
molecula segmen s a e desc ibed as e ec i e in e ac ion si es (pa icles). Fu he
simpli ica ions a e possible employing e ec i e media app oach such as he dielec ic
con inuum app oach o sol en .
The choice o he op imal model and scale o he sys em desc ip ion should be d i en
by he scien i ic ques ion. Con inuum scale modelling neglec s spa ial dis ibu ion o
indi idual phases o he ma e ial, which can be c i ical i phase changes ha e s ong
e ec on ma e ial p ope ies. Po e scale modelling, on he o he hand, is capable o
acking he in e aces be ween indi idual phases and p o ide he e olu ion o sys em
pa ame e s a REV scale. Molecula scale models can deli e he p ope ies o
indi idual phases and phase in e aces. Thus, u he de elopmen o he
compu a ional ools and he models should be aimed a mul iscale model coupling in
which la ge scale models de ine bounda y condi ions whe eas he small-scale models
p o ide ma e ial p ope ies necessa y o he accu a e sys em desc ip ion (Molins &
Knabne , 2019). Majo limi a ions o mul iscale modelling amewo ks a e ela ed o
compu a ional e o and he c oss-scale model coupling, which would no be sol ed
by inc ease o compu a ional esou ces alone. New e icien algo i hms and he scale
coupling-decoupling schemes should be in he i s p io i ies o u u e de elopmen s.
Coupled eac i e anspo models o eposi o y sys ems
The so-called eac i e anspo models include he coupling be ween luid low, mass
anspo and geochemis y. Such modes a e well es ablished (S ee el, 2019).
Nume ous codes exis a he Da cy o con inuum scale (S ee el e al., 2015). Se e al
compu a ional benchma ks ele an o deep eposi o ies ha e been published (Aguila
e al., 2021; Idia , 2019; Ma y e al., 2015; Poonoosamy e al., 2021). The
co esponding models ha e been applied o simula e he e olu ion o ba ie s in deep
geological eposi o ies du ing he ecen EU p ojec s (e.g. EU CEBAMA (Du o e al.,
2020) and wo k packages ACED (Jacques e al., 2024) and DONUT in EURAD (Cla e
e al., 2022). In EURAD WP-ACED, he componen s o eposi o y sys ems a e
EURAD-2 Deli e able 16.1 – S a e-o - he-a on high ideli y nume ical simula ions o s ongly coupled
p ocesses o eposi o y sys ems (Ini ial epo )
Dissemina ion le el: Public
Da e o issue o his epo : 29/09/2025 17
modelled a he scales o in e aces, was e packages and disposal cell, e.g. (Blanc e
al., 2024 ; De Wind e al., 2024; Lemmens e al., 2023; Mon e al., 2023; Mon eneg o
e al., 2023; Wi eb ood e al., 2024a; Wi eb ood e al., 2024b ).
Mo e ecen ly, coupled eac i e anspo models ha e been implemen ed a he po e-
scale as well. Such modes a e able o ela e po e-scale mic os uc u al changes due
o geochemical eac ions o e ec i e pa ame e s ha can be applied o he con inuum
scale. Examples o such app oach, among many o he s, a e gi en in (Molins e al.,
2017; Pa el e al., 2018; P asianakis e al., 2017; P asianakis e al., 2018; Seigneu e
al., 2017; Va zina e al., 2020).
In si u condi ions and ma e ial p ope ies o he EBS sys em e ol e due o he he mal,
hyd aulic, mechanical, biochemical and chemical g adien s ha exis wi hin and
be ween he di e en eposi o y componen s. Wi hin he EURAD p ojec , wo wo k
packages ha e ocussed on hese changes, i.e. geochemis y induced changes
(ACED WP) and coupled chemo-mechanical p ocesses (MAGIC) (Cla e e al., 2022).
Fo he geochemical e ec s on he EBS p ope ies, (Nee , 2022) and (Deissmann e
al., 2021) ga e an o e iew o p ocesses and a ailable models a he in e ace scale.
Se e al coupled eac i e anspo models ha e been used o e alua e he e olu ion
a he in e ace o wo ma e ials (e.g. (Idia , La iña, Kosakowski, e al., 2020; Ma y
e al., 2015; Sa age e al., 2010) see also (Bilds ein e al., 2019) o a ecen o e iew).
Recen coupled eac i e anspo models ha e been success ul in coupling all ele an
ma e ials in a disposal cell con aining high le el i i ied was e: nuclea glass-s eel-
cemen /ben oni e-hos ock (g ani e o clay) and epo ed in (De Wind e al., 2024;
Mon eneg o e al., 2023). Also ecen ly, hyd o-chemo-mechanical modelling has been
used o s udy he e olu ion o he Cigéo eposi o y closu e sys ems based on
ben oni e-based sealing componen s su ounded by cemen i ious ma e ials (Idia ,
La iña, Cochepin, e al., 2020; La iña e al., 2023).
Chemo-mechanical couplings in massi e conc e e in as uc u e ha e been add essed
in WP MAGIC using mul iscale simula ions anging om nano o cell scale models
(Dauzè es e al., 2022). Reac i e anspo codes and mechanical codes can be
sequen ially coupled o model cemen i ious ma e ial damage due o ca bona ion (Socié
e al., 2023). The hyd o-chemo-mechanical a ia ional phase- ield ac u e app oach,
o example, is capable o handling chemical eac ions, as well as he esul ing ma e ial
dissolu ion and/o p ecipi a ion caused by hyd a ion o deg ada ion (such as
ca bona ion) o ac u ed cemen i ious ma e ials (Zhang e al., 2018). Po e scale
simula ions a e used o in es iga e mic os uc u e e olu ion and o es ima e he
e ec i e mechanical pa ame e s o he media (Shen e al., 2020).
2.3.2 Applica ion o machine lea ning
The compu a ional ime in a mul i-physics modelling amewo k is o en domina ed by
only a ew o e en jus a single p ocess. Fo example, i is well known ha mos o he
compu a ion ime in coupled T-H-C eac i e anspo models is aken by he solu ion
EURAD-2 Deli e able 16.1 – S a e-o - he-a on high ideli y nume ical simula ions o s ongly coupled
p ocesses o eposi o y sys ems (Ini ial epo )
Dissemina ion le el: Public
Da e o issue o his epo : 29/09/2025 18
o he geochemical equa ions. Recen ly, se e al e o s ha e been made o eplace
hese compu a ionally expensi e ou ines by cheap su oga e models o en based on
machine lea ning echniques o look-up ables. Huang e al. (2018) desc ibed he
complex geochemis y o ageing cemen i ious was e esul ing om ca bona ion and
alkali-silica eac ions ia a look-up able app oach o a mul i-phase anspo model o
a conc e e s uc u e. Fu he mo e, emula ing he geochemical models using machine
lea ning echniques and by inco po a ing hem in o eac i e anspo models ha e
been es ed and implemen ed in some ecen s udies (De Lucia e al., 2017; Demi e
e al., 2023; Huang e al., 2018; Ja nieks e al., 2016; Laloy & Jacques, 2022; Leal e
al., 2020; P asianakis e al., 2020). P asianakis e al. (2020) ha e shown se e al
examples o success ul use o machine lea ning o upscaling po e-scale models
(la ice Bol zmann model) o a con inuum (Da cy)-scale eac i e anspo model.
Applica ions a he Da cy-scale show a gain in compu a ional ime o abou an o de o
magni ude, while geochemical calcula ions can be accele a ed be ween one and ou
o de s o magni ude. Al e na i ely, a su oga e can be made o eplace he comple e
model including all coupled physical p ocesses. Recen applica ions in he ield o
coupled eac i e anspo a e o u anium anspo (Laloy & Jacques, 2019) and
elec o-kine ic bio emedia ion (Sp oca i & Rolle, 2021). Machine lea ning echniques
a e also becoming popula o app oxima ing he s ess-s ain cons i u i e beha iou
o geoma e ials, including a i icial neu al ne wo ks and gene ic p og amming (Gao,
2018). (G a e al., 2010) ained ecu en neu al ne wo ks (RNN) wi h ime-dependen
da a o assess he long- e m beha iou o a ein o ced conc e e s uc u e subjec o
mechanical and en i onmen al loads. (Capuano & Rimoli, 2019) and (Loga zo e al.,
2021) also used RNN wi h ime-his o y p edic ion capabili y o eplace inelas ic
homogenisa ion (i.e. mul iscale) beha iou o ma e ials consis ing o a so
elas oplas ic ma ix wi h s i elas ic inclusions (e.g. conc e e). (Con i e al., 2018)
applied concep s om da a science o ma e ials science wi h he so-called da a-d i en
pa adigm, consis ing o e o mula ing he classical ini ial-bounda y- alue p oblems
di ec ly om ma e ial da a.
Many nume ical algo i hms a e ma u e enough o p o ide coupled desc ip ion o
mul iphysics p ocess, bu a e ypically limi ed a a speci ic leng h- and ime scale. The
T-H-M-C p ocesses ha go e n he eposi o y e olu ion a e in insically mul iscale,
comp ising phenomena om he a omis ic le el (e.g. chemical so p ion and ions
mobili y), o he mac oscopic le el. Fu he mo e, small-scale p ocesses can ha e a
s ong e ec on he sys em e olu ion a eposi o y scale. In addi ion, he nume ical
algo i hms use di e en p og amming languages (e.g. Fo an, C/C++, CUDA, Py hon,
o name a ew). Me ging he di e en codes is a e y challenging ask. Machine
lea ning echniques can be used o suppo he communica ion ac oss he algo i hms
since he mul i-dimensional complex ou pu o he models may be ep esen ed by
eg essing a simple ma hema ical objec , e.g. neu al ne wo ks. A he same ime, and
o speci ic applica ions, i seems easonable o c ea e su oga e models, ained on
he ull physical algo i hms, each a he espec i e leng h scale and o subsequen ly
in eg a e hem in o de o accele a e he o e all mul iscale calcula ions.
EURAD-2 Deli e able 16.1 – S a e-o - he-a on high ideli y nume ical simula ions o s ongly coupled
p ocesses o eposi o y sys ems (Ini ial epo )
Dissemina ion le el: Public
Da e o issue o his epo : 29/09/2025 19
The implemen a ion o accele a ed nume ical me hods and so wa e enginee ing
should go hand in hand and uned o he u u e o ien ed de elopmen o High
Pe o mance Compu ing (HPC) in as uc u e. This is no always he case o exis ing
scien i ic so wa e (G annan e al., 2020). Fo example, G aphical P ocesso Uni s
(GPU) based HPC sys ems ha e domina ed he HPC landscape since nea ly 10 yea s
(www. op500.o g). In con as , a he ew mul i-physics codes a e able o ake ull
ad an age o such a chi ec u e. Ou assessmen sugges s ha cu en ly he
de elopmen o coupled codes s ays behind he p og ess in ha dwa e de elopmen .
Full exploi a ion o HPC in as uc u e would enable as e and mo e complex
calcula ions han cu en ly easible. This would p o ide a basis o de elopmen o
digi al wins o T-H-M-C coupled p ocesses, and o in eg a ion o da a collec ed p io
and du ing he design, ope a ional and/o he pos closu e phases o eposi o y
sys ems.
2.3.3 Model alida ion and benchma king
Benchma king plays an impo an ole in he concep ual and nume ical de elopmen
o coupled T-H-M-C codes, whe e conse a ion laws (con inuum mechanics),
he modynamics (e.g., equa ions o s a e), ma e ial beha iou based on highly non-
linea cons i u i e laws, and chemis y (law o mass ac ion, Gibbs ene gy) need o be
conside ed simul aneously. Al hough he benchma king o coupled p ocess models
has made g ea p og ess in ecen decades, pa icula ly in he analysis o labo a o y
expe imen s, a high deg ee o unce ain y emains om s uc u al o p ac ical
applica ions (modelling o eal eposi o ies). The benchma king o coupled p ocesses
s a ed in he 1980s wi h he INTRACOIN, HYDROCOIN and INTRAVAL ini ia i es
(He be e al., 1988; Konikow e al., 1997; La sson, 1992). Unlike p e ious
benchma king ini ia i es, DECOVALEX is an ongoing and g owing p ojec ha has
uni ed a la ge communi y o modelle s o mo e han 25 yea s o he de elopmen o
coupled models and hei alida ion agains expe imen s (Bi kholze e al., 2018;
Bi kholze e al., 2019; Chen e al., 2009; Jing e al., 1995). DECOVALEX ocuses
mainly on nea ield he mo-hyd o-mechanical p ocesses. Recen ly a pe o mance
assessmen ask has been in eg a ed in o i s po olio. In his espec , he
benchma king ini ia i es MoMaS and SeS Bench o eac i e anspo p ocesses ha e
been launched in he pas (Aguila e al., 2021; Bilds ein e al., 2021; Ca ay ou e al.,
2010; Poonoosamy e al., 2021; S ee el e al., 2015). The combina ion o T-H-M and
eac i e anspo ini ia i es is s ill no a ailable, which also shows a lack o a uly
mul idisciplina y benchma king ini ia i e uni ing geomechanics and geochemis y.
Cu en ly, his limi a ion s ill hinde s he ull po en ial o T-H-M-C code de elopmen
(Koldi z e al., 2018). The e is a la ge body o li e a u e desc ibing indi idual
benchma king e o s. A mo e sys ema ic way o o ganising T-H-M/C has been
de eloped in a book se ies linked o da a eposi o ies whe e inpu iles and ela ed
code e sions a e s o ed (Koldi z e al., 2012). Mo e ecen ly, online e sions o
benchma k collec ions ha e become a ailable, e.g. ia he benchma king galle ies
(h ps://www.opengeosys.o g/docs/benchma ks/) and GeoML (h ps://geoml.eu/),
EURAD-2 Deli e able 16.1 – S a e-o - he-a on high ideli y nume ical simula ions o s ongly coupled
p ocesses o eposi o y sys ems (Ini ial epo )
Dissemina ion le el: Public
Da e o issue o his epo : 29/09/2025 20
allowing easie ep oduc ion o benchma ks e en o e ing collabo a i e in e ac i e
wo k en i onmen ia Jupy e Labs (see also Fu u e o ien ed collabo a i e pla o ms
below). In au ho s’ opinion, suppo and de elopmen o such ini ia i es a Eu opean
le el (i.e. wi hin EURAD esea ch p og am) and i s le e age by associa ed
in e na ional esea ch g oups ha e high added alue o he p og ess o scien i ic
collabo a ion owa ds FAIR (Findable, Accessible, In e ope able, Reusable) esea ch
da a p inciple (Wilkinson e al., 2016) and quali y assu ed code de elopmen .
Besides benchma king o nume ical T-H-M-C coupled codes, equally impo an is he
alida ion s udies o he implemen ed models wi h a ailable expe imen al da a
(Addassi e al., 2022; Peleg í e al., 2023). Fo ben oni e, which is an essen ial EBS
componen , ecen ly compiled expe imen al da abases a e a ailable in he li e a u e
and can be used as a means o collabo a i e T-H-M-C model alida ion (Cab e a e
al., 2023; Tha che e al., 2017).
2.4. F on ie s o ealis ic eposi o y modelling, design and op imisa ion
In his sec ion, we explo e he p esen challenges o coupled p ocess simula ions,
mul i-physics, and chemis y ele an o he geological disposal o nuclea was e.
Speci ically, we discuss he de elopmen and applica ion o ad anced T-H-M-C
models o in e se modelling, eposi o y design op imisa ion and sensi i i y analysis o
model pa ame e s. These a e, in ou opinion, he key a eas o u he de elopmen ,
d i en by he needs o eposi o y implemen a ion. Aiming o exploi he ull po en ial o
he mos powe ul compu a ional esou ces, we add ess he necessi y o he
cohe ence be ween he design o so wa e packages and he HPC a chi ec u e.
Finally, we elabo a e on he ad an ages in he use o collabo a i e pla o ms o code
de elopmen .
2.4.1 Op imisa ion, unce ain y analysis, and in e se modelling
P ocess-based nume ical simula ions a e he basis o in-dep h sys em unde s anding,
analysis o expe imen al obse a ions and hei upscaling. Despi e he con inuous
g ow h o he compu a ional esou ces, he ealism o he models applied in he
simula ions o eposi o y sys ems emains limi ed in e ms o dimensions, ime-space
esolu ion and p ocess couplings. In e p e a ion o expe imen al da a, sa e y and cos -
d i en design op imisa ion, and model unce ain y analysis belong o he class o
in e se p oblems. Nume ical solu ion o in e se p oblems implies i e a i e o wa d
modelling un il he solu ion con e ges o he op imal pa ame e se (e.g. sa is ac o y
desc ip ion o expe imen al da a, cos -sa e y op imisa ion o eposi o y design, o
unce ain y analysis).
Fo bo h o wa d and in e se p oblems, o de s o magni ude imp o emen in he
compu a ional e iciency can be ob ained by eplacing he physical based sol e s o i s
componen s wi h high ideli y su oga e models (sec ion 3.2). Pa icula ly p omising a e
EURAD-2 Deli e able 16.1 – S a e-o - he-a on high ideli y nume ical simula ions o s ongly coupled
p ocesses o eposi o y sys ems (Ini ial epo )
Dissemina ion le el: Public
Da e o issue o his epo : 29/09/2025 21
he su oga e models based on machine lea ning o speci ic aspec s o T-H-M-C
coupled models, da a exchange be ween models a di e en scales, educ ion o big
da a and ex ac ion o cons i u i e ela ions om la ge nume ical, expe imen al, and
moni o ing da ase s (Elodie e al., 2020; Hu e al., 2024; Hu e al., 2023; Ringel e al.,
2024).
2.4.2 Abs ac ion and simpli ica ion me hods
Fo a gi en nume ical simula ion p oblem wi h mul iple p ocesses, some p ocesses o
elemen s om he concep ual model migh ha e only a limi ed e ec on a model ou pu
(F an z, 1995). Model abs ac ion e e s o a sys ema ic me hod o educe he
complexi y o he compu a ional bu den o he model while main aining he alidi y o
he simula ion esul s wi h espec o he ques ion ha he simula ion is being used o
add ess. Model abs ac ion educes he simula ed sys em o i s essen ial componen s
and p ocesses h ough a simpli ica ion o concep ual (sub-)models, selec ion o
signi ican p ocesses and app op ia e ime and spa ial scales o mo e compu a ionally
e icien implemen a ions (o speci ic model componen s and p ocesses). In i s mos
ex eme o m, he model is s ipped down o a single componen which jus ep oduces
he desi ed ou pu om he inpu in a compu a ionally e icien way (Go ae s e al.,
2022).
The classi ica ion o echniques desc ibed in (Go ae s e al., 2022) a e based on
p e ious wo k by (Pachepsky, 2006) and (Raza i e al., 2012). A g oup o simpli ica ion
me hods is labelled lowe ideli y nume ical models and uses s a egies as a p e-
de ined hie a chy o models, delimi ing he inpu domain, he scale change and
educing nume ical accu acy. A second g oup is based on s a is ically de i ed
su oga e models which we e discussed abo e. Hie a chy o models a e de eloped o
low and anspo in ac u ed po ous media (Be e e al., 2019), T-H-M p ocesses
conce ning mechanical ba ie in eg i y (Pi z e al., 2023), di usion in cha ged po ous
media (Hennig & Kühn, 2021), o su ace complexa ion modelling (A o a e al., 2018).
Reducing he compu a ional cos s can be done by lowe ing he dimensionali y o he
p oblem (e.g. (Idia , 2019)) o sub-g id scale e inemen (e.g. (Fins e le e al., 2020;
Ma ine , 2020)), bu educ ion o chemical sys em o coupled p ocesses (e.g., he
ele ance o po osi y eedback in (Aguila e al., 2020) a e o he op ions as well. A
ecen example o upscaling me hods is o ep esen ac u ed nuclea glass as an
e ec i e medium (Repina e al., 2020).
2.4.3 Global sensi i i y analyses
T ea men o unce ain ies in he pe o mance assessmen o deep geological disposal
has been ecognized as an impo an opic o mo e han h ee decades (E. J. Bonano
& R. M. C anwell, 1988). In o de o quan i y he e ec o pa ame e a ia ion on
p edic ed sys em pe o mance, he local sensi i i y analysis (LSA) and global
EURAD-2 Deli e able 16.1 – S a e-o - he-a on high ideli y nume ical simula ions o s ongly coupled
p ocesses o eposi o y sys ems (Ini ial epo )
Dissemina ion le el: Public
Da e o issue o his epo : 29/09/2025 22
sensi i i y analysis (GSA) me hods a e ele an . While LSA de e mine he impac o
small inpu pe u ba ions a ound nominal alues on he model ou pu , GSA conside s
simul aneously he whole combina ional a ia ion ange o he inpu s. While bo h
me hods could p o ide ele an in o ma ion o T-H-M-C coupling, GSA accoun o
non-linea i y and in e ac ions among pa ame e s in sys em esponses in a mo e obus
manne (Chaudh y e al., 2021; Delchini e al., 2021; Nguyen e al., 2009; Wainw igh
e al., 2013). Recen ly, GSA ha e also been used o ackle eac i e anspo p oblems
and adionuclide mig a ion (Ayoub e al., 2020). Su oga e models can also be used o
deciphe unce ain y p opaga ion (Sochala e al., 2022) and o sensi i i y analyses.
The use o su oga e models can ci cum en an issue ela ed o GSA. Indeed, GSA
equi es many model e alua ions o achie e sa is ac o y accu acy, which will lead o a
g ea challenge in compu a ional e o s o la ge models, which is o pa icula
ele ance o complex coupled T-H-M-C p ocesses in eposi o ies. E en unning
su oga e models can be ema kably challenging and e en compu a ionally p ohibi i e
in he case o in ensi e simula ions and la ge-dimensional sys ems and necessi a e
he use o educed space su oga es (Voh a e al., 2019).
2.4.4 So wa e de elopmen and HPC in as uc u e
So wa e enginee ing plays an inc easingly impo an ole in scien i ic p ojec s. The
main d i e s a e he complexi y o he asks o be ackled, e.g. mul i-physics, mul i-
scale app oaches o coupled p ocesses, eal-wo ld applica ion wi h complex
geome ies and he associa ed compu a ional e o equi ing he applica ion o high-
pe o mance compu ing echnologies (Pa k e al., 2021; T inche o e al., 2017). We
belie e ha , mee ing hese challenges, la ge in e na ional mul idisciplina y
de elopmen eams a e needed o dis ibu ed de elopmen , so open-sou ce p ojec s
based on e sion con ol, con inuous in eg a ion and code e iew ha e become a
ounda ion o ecen esea ch so wa e (Bilke e al., 2019; Bjo ge e al., 2022; Fio ina
e al., 2022). Quali y-assu ance o he so wa e o sa e y assessmen applica ions in
nuclea was e managemen is pa icula ly impo an , and equi e anspa ency,
aceabili y and ep oducibili y o esul s. New so wa e echnologies a e cu en ly
making hei way in o geoscience applica ions, such as con aine echnologies o
po abili y o complex so wa e p ojec s, au oma ed wo k lows o sol ing complex
asks in ol ing la ge applica ion da a, o au oma ed benchma king wo k lows
(Lehmann e al., 2023). Mode n so wa e p ojec s also make in ensi e use o
ecosys ems such as Py hon o Julia and in eg a e hem in o he en i e wo k low, i.e.
p e- and pos -p ocessing as well as simula ion ke nels. This equi es app op ia e
applica ion in e aces (APIs) such as ogs ools (e.g. (Buchwald e al., 2021)).
P o essional so wa e de elopmen and deploying echnologies such as Vi ual Reali y
a e key o he success ul implemen a ion o digi al win concep s o nuclea was e
managemen in he u u e (Koldi z e al., 2023).
EURAD-2 Deli e able 16.1 – S a e-o - he-a on high ideli y nume ical simula ions o s ongly coupled
p ocesses o eposi o y sys ems (Ini ial epo )
Dissemina ion le el: Public
Da e o issue o his epo : 29/09/2025 23
The design and de elopmen o ha dwa e o cu en and u u e HPC sys ems a e
shaped by se e al physical and economical challenges (G annan e al., 2020). Mode n
HPC sys ems need signi ican amoun o ene gy supply o ope a ion and cooling,
meaning ha maximizing ene gy e iciency in hei ope a ion and usage is mo e
impo an han e e . The minia u ised design o he ha dwa e is eaching he physical
limi s o pe o mance ha can be achie ed on each indi idual compu e chip. The new
gene a ion o HPC (www. op500.o g) inc easingly combines Cen al P ocessing Uni
(CPUs) wi h dedica ed accele a o s based on GPUs. Such hyb id CPU/GPU sys ems
a e ene gy e icien and p o ide, in heo y, many mo e lops pe wa o consumed
ene gy (Ash a e al., 2018). New p og amming models and so wa e op imisa ion a e
indispensable o be able o exploi he heo e ical pe o mance o such sys ems,
conside ing he di e en ha dwa e a chi ec u e be ween CPUs and GPUs; GPUs a e
composed o hund eds o housands o co es ocusing on pa allel p ocessing and high
h oughpu a lowe clock speed(Fio e e al., 2018). As a ma e o ac , many widely
used so wa e packages and scien i ic applica ions de eloped o CPU-only sys ems
a e ha dly able o bene i om he ha dwa e po en ial o e ed by hyb id CPU/GPU
sys ems, since dedica ed GPU compa ible sou ce code has o be gene a ed and
compiled.
The use o a i icial in elligence (AI), dynamic da a p ocessing and in eg a ion o da a
in o he models a e o he changing aspec s o scien i ic compu ing. T adi ional physics
d i en modelling accep ela i ely small inpu da ase s desc ibing ini ial and bounda y
condi ion. Con a y, he AI and dynamic da a in eg a ion s a egies o en equi e he
managemen o massi e da a s eams composed o millions o he e ogeneous
da ase s, meaning ha inpu and ou pu algo i hm o dis ibu ed ile sys ems mus be
op imised o da a p ocessing. Whe eas he high p ocessing speed o da a on CPU o
GPU pe mi s mo e complex ensemble simula ions o mul i-physics models, i also
means gene a ing e e -la ge da ase s o pos p ocessing and analysis.
2.4.5 Fu u e o ien ed collabo a i e pla o ms
The landscape o compu a ional algo i hms ele an o machine lea ning is e y b oad
and he espec i e so wa e is upda ed a a e y as pace. This makes he consis en
ins alla ion and use o hese ools o be a e y challenging ask o scien is s wi hou
scien i ic compu ing expe ise, who howe e a e in e es ed o use and explo e he
machine lea ning po en ial. Mo eo e , in collabo a i e p ojec s wi h many esea ch
pa ne s he e is no common wo king space which could allow o p og am and es
algo i hms in a collabo a i e way. Cen alizing he e o s in p o iding an open access
web-se e based collabo a i e pla o m educes duplica ion o wo k and ein en ion
o he wheel. As a esponse o hese challenges, he www.geoml.eu open pla o m has
been launched ecen ly om he EURAD-DONUT wo kpackage and PREDIS p ojec
as a ehicle o enhance collabo a ion, educa ion, join code de elopmen and
demons a ion o esul s. A jupy e lab se e , ha ing p e-ins alled all necessa y
EURAD-2 Deli e able 16.1 – S a e-o - he-a on high ideli y nume ical simula ions o s ongly coupled
p ocesses o eposi o y sys ems (Ini ial epo )
Dissemina ion le el: Public
Da e o issue o his epo : 29/09/2025 24
compu a ional packages, allows o p og am, sha e and es nume ical codes o ypical
classes o p oblems wi hou he need o local compu a ional esou ces, o o he
ins alla ion o compu a ional en i onmen s. Such pla o ms p o ide means o
communica e esul s o he scien i ic communi y and he public in o m o online
in e ac i e demons a o s signi ican ly enhancing he ou each o scien i ic esul s.
Simila pla o m has been ecen ly launched o coupled p ocess modelling as well
(e.g. h ps://www.opengeosys.o g/docs/benchma ks/).
EURAD-2 Deli e able 16.1 – S a e-o - he-a on high ideli y nume ical simula ions o s ongly coupled
p ocesses o eposi o y sys ems (Ini ial epo )
Dissemina ion le el: Public
Da e o issue o his epo : 29/09/2025 25
3. La es de elopmen s and open gaps in nume ical simula ion o
coupled p ocesses
3.1. Imp o ing models’ ideli y and compu a ional e iciency o s ongly
coupled T-H-M-C p oblems and subp oblems
3.1.1 Coupled p ocesses
This sec ion includes an o e iew o he THM p ocesses and modelling app oaches
associa ed wi h gas anspo in di e en eposi o y concep s h oughou he eposi o y
li e ime; T-H-M-C o canis e /bu e in e ac ions; THC o eac i e anspo associa ed
wi h canis e and i on co osion and glass dissolu ion; he modynamic consis ency o
coupled models.
The modelling o eposi o y sys ems, in pa icula so-called combined eposi o ies,
whe e bo h in e media e and low-le el was e as well as high-le el was e a e o be
s o ed, o en equi es he possibili y o ully coupled THM models. As he ma e o
p ac ical implemen a ion THM and RTP codes (S ee el e al., 2015), a e o en
de eloped by di e en communi ies. The cu en ly conside ed combined eposi o y
concep s o F ance (ILW and HLW in CIGEO) and Swi ze land (I/LLW and SF/HLW in
Nö dlich Läge n) all equi e a ope a ional coupling o hese models (Chu ako , 2024).
Speci ic coupling be ween chemical pe u ba ions and mig a ion
Bo h adionuclide and co-con aminan mig a ion a e conside ed in he sa e y
assessmen o was e disposal. Al hough he o ganic and saline co-con aminan s
eleased by was e packages do no con ibu e di ec ly o he adiological dose, hei
e ec on he mig a ion o adionuclides mus be well documen ed (Descos es e al.,
2017; F alo a e al., 2021). In addi ion o he mechanical and chemical pe u ba ions
induced by he co-con aminan s (pH, Eh, I.S., dissolu ion, gas gene a ion, e c.), he e
is a s ong coupling be ween chemical pe u ba ions and anspo p ope ies
(Poonoosamy e al., 2021). Fo example, he di usion d i en by he g adien o species
ac i i ies is modi ied along he mig a ion pa h o co-con aminan s (Dagnelie e al.,
2015; Hennig & Kühn, 2021). The desc ip ion o he di usion o co-con aminan s in
clay ocks appea s o be su icien ly documen ed o accoun o he pe u ba ions in
mul i-componen o mul i-species models discussed below (Guo e al., 2020).
Expe imen al esul s om in-si u expe imen s conduc ed in unde g ound esea ch
labo a o ies can also be used as inpu da a o benchma k and sensi i i y analyses
ega ding his coupling (Dagnelie e al., 2023).
Assessing he pe o mance o he enginee ed ba ie sys em (EBS) equi es he use
o high- ideli y eac i e anspo models o simula e he eac ions occu ing a he
EURAD-2 Deli e able 16.1 – S a e-o - he-a on high ideli y nume ical simula ions o s ongly coupled
p ocesses o eposi o y sys ems (Ini ial epo )
Dissemina ion le el: Public
Da e o issue o his epo : 29/09/2025 32
3.1.4.3 Su oga es o mechanics and hyd aulics ope a ing di ec ly om images o
ma e ials
Inco po a ing po e-scale dynamics in o eac i e anspo simula ions is c ucial o
enhancing hei accu acy and p edic i e capabili ies (P asianakis e al., 2020; Um e
al., 2018). This in eg a ion is ypically achie ed by upscaling po e-scale simula ion
esul s o de i e e ec i e p ope ies (pe meabili y, e ec i e di usi i y, e ec i e
eac ion a es, e c.), which a e hen embedded in o he eac i e anspo sol e
(Löna z e al., 2023; Poonoosamy e al., 2022; P asianakis e al., 2020). Howe e ,
conduc ing po e-scale simula ions on eal po e ne wo ks de i ed om compu ed
omog aphy (CT) scans o co es o d ill cu ings is compu a ionally expensi e. This
challenge a ises om he ine disc e isa ion equi ed o cap u e he complex
geome ies o hese po e ne wo ks (Maes & Menke, 2021; Mah ous e al., 2022;
Mehmani & Tchelepi, 2017). To educe hese compu a ional cos s, simula ions a e
o en pe o med on ep esen a i e subdomains ha cap u e essen ial ea u es o he
en i e po e ne wo k, see Figu e 2 (Mah ous e al., 2022; Menke e al., 2021). While his
app oach educes compu a ional demands, conduc ing simula ions on mul iple
ep esen a i e subdomains is s ill necessa y o ga he su icien s a is ics o accu a ely
es ima ing e ec i e p ope ies (Mah ous e al., 2022). Consequen ly, he o e all
p ocess emains compu a ionally expensi e. A s a egy is o use su oga e models ha
map po e ne wo k images (co esponding o he size o he ep esen a i e subdomain)
o e ec i e p ope ies, he eby enabling ex ensi e po e-scale simula ions ac oss he
en i e domain.
-
Figu e 2 – Compu a ional app oach o ex ac ing po e-scale dynamics in o ma ion. This
wo k low is summa ised om Mah ous e al. (2022) and Maes & Menke (2021).
Machine lea ning has ecen ly eme ged as a p omising su oga e modelling app oach
o his ask (Alqah ani e al., 2021; Elmo sy e al., 2023; Elmo sy & Zhao, 2022; Liu e
al., 2022; Menke e al., 2021; Niu e al., 2020; San os e al., 2021; Sia ashi e al., 2022;
Wang e al., 2022; Wu e al., 2019; Yong e al., 2024). These models a e ained in a
da a-d i en manne by lea ning he inpu -ou pu ela ionship be ween po e ne wo k
images and e ec i e p ope ies. Gi en ha he inpu s a e image-based, many s udies,
EURAD-2 Deli e able 16.1 – S a e-o - he-a on high ideli y nume ical simula ions o s ongly coupled
p ocesses o eposi o y sys ems (Ini ial epo )
Dissemina ion le el: Public
Da e o issue o his epo : 29/09/2025 33
such as hose by Jiang e al. (2023), Elmo sy and Zhao (2022), Liu e al. (2022),
Sia ashi e al. (2022), Wang e al. (2022), Alqah ani e al. (2021), Menke e al. (2021),
San os e al. (2021), San os e al. (2020), Kam a a e al. (2020), Niu e al. (2020) and
Wu e al. (2019) u ilize con olu ional neu al ne wo k (CNN) a chi ec u es o e ec i ely
map he po e ne wo k images o he e ec i e p ope ies. Howe e , CNN models o en
equi e a la ge numbe o aining samples o ensu e physically consis en p edic ions
(Elmo sy e al., 2023; Elmo sy & Zhao, 2022; Kam a a e al., 2020). Fo example,
Elmo sy and Zhao (2022) used 57,500 samples (p io o da a augmen a ion) o ain a
CNN model o pe meabili y p edic ion.
San os e al. (2020) employ a CNN model, ained in a da a-d i en manne , o map
po e ne wo k images o physical s a es, speci ically eloci y ields, a he han di ec ly
mapping hem o e ec i e p ope ies such as pe meabili y. This app oach ensu es ha
physics is accu a ely cap u ed by he CNN model be o e pe o ming any upscaling o
de i e pe meabili y. Despi e he e o o inco po a e physics in o he upscaling p ocess
using machine lea ning, his me hod s ill demands subs an ial compu a ional
esou ces. In he epo ed s udy aining he model ook as much as 12 hou s.
To imp o e he compu a ional e iciency, Liu e al. (2022) p oposed o educe he
numbe o aining samples by combining a Random Fo es (RF) wi h a Neu al Ne wo k
(NN). Thei app oach eplaces he con olu ional a chi ec u e wi h an RF me hod, which
is used o classi y REV samples in o se e al g oups based on key po e s uc u al
pa ame e s. These s uc u al ea u es a e hen inpu in o a NN model o p edic
e ec i e eac ion a es. By keeping he NN a chi ec u e simple, his app oach educes
he equi ed numbe o aining samples. Howe e , his app oach in oduces
complexi y in de ining he po e s uc u al ea u es o each g oup. When add essing
complex po e-scale phenomena, he numbe o equi ed ea u es migh inc ease
subs an ially, demanding addi ional compu a ional esou ces o obus ly de e mine
classi ica ion ea u es.
Elmo sy e al. (2023) conside ed ano he app oach by employing Physics-In o med
Neu al Ne wo ks (PINN) me hod, whe e physics a e embedded in o he loss unc ion
du ing he aining o he NN model (Raissi e al., 2019). This app oach begins by
pe o ming po e-scale simula ions on small domains o ob ain local pe meabili y
alues, which se e as a educed ep esen a ion o he po e ne wo k image, simila o
he ole o he RF algo i hm in Liu e al. (2022). These local pe meabili y alues a e
hen used as inpu s o he NN model o p edic he pe meabili y o la ge domains. By
inco po a ing physics in he o m o analy ical cons ain s imposed o he p edic ed
pe meabili ies, his me hod signi ican ly educes he numbe o aining samples
equi ed o model aining, om 57,500 samples (as used in Elmo sy and Zhao (2022))
o 15,336 samples o a simila da ase .
EURAD-2 Deli e able 16.1 – S a e-o - he-a on high ideli y nume ical simula ions o s ongly coupled
p ocesses o eposi o y sys ems (Ini ial epo )
Dissemina ion le el: Public
Da e o issue o his epo : 29/09/2025 34
Howe e , he compu a ional e o s o conduc ing small-scale po e-scale simula ions
emains signi ican . Fu he mo e, he PINN su e s om nume ical ins abili y du ing
aining ( anishing o exploding g adien s), which complica es he aining p ocess and
makes i challenging o assign op imum weigh s o he e ms in he loss unc ion
(Chuang & Ba ba, 2022; San oso & Wellmann, 2022). Consequen ly, his can esul in
physically inconsis en p edic ions (Chuang & Ba ba, 2022; San oso & Wellmann,
2022). F om hese h ee s udies, wo key s a egies eme ge o e icien ly cons uc ing
su oga e models o accu a ely cap u e po e-scale dynamics: (1) inco po a ing physics
in o su oga e model de elopmen and (2) educing he dimensionali y o po e ne wo k
images be o e eeding hem in o he su oga e models. Fo he i s s a egy, in addi ion
o using he PINN, o he me hods include:
• he non-in usi e educed basis (NI-RB) me hod, which ep esen s a s a e as a
linea combina ion o basis unc ions and coe icien s, whe e he basis unc ions
main ain he s uc u e o he p oblem, and he coe icien s a e p edic ed using a
NN model (Hes ha en & Ubbiali, 2018),
• en o cing physics in o he s uc u e o machine lea ning models, such as Fou ie
Neu al Ope a o (FNO) (Li e al., 2020), Kolmogo o -A nold Ne wo ks (KAN)
(Liu e al., 2024), and Physics encoded Neu al Ne wo k (PeNN) (Meinde s e
al., 2024), and
• blending nume ical simula ions wi h a U-Ne neu al ope a o , whe e nume ical
simula ions a e conduc ed o only a ew ini ial ime s eps, and he U-Ne neu al
ope a o ex apola es esul s o u u e ime s eps based on he simula ion
ou pu (Oommen e al., 2024).
Fo he second s a egy, besides using RF algo i hm, o he me hods include:
• he Ka hunen-Loè e me hod, which pe o ms dimensionali y educ ion based
on he a iance o each pixel in an image (O’Lea y-Rosebe y e al., 2022),
• he Ac i e Subspaces me hod, which educes dimensionali y by iden i ying he
impo ance o each pixel ela i e o an objec i e unc ion (O’Lea y-Rosebe y e
al., 2022), and
• he Au oencode me hod, which uses an encoding-decoding app oach o
dimensionali y educ ion (Shams e al., 2020; Tschannen e al., 2018).
In eg a ing hese wo key s a egies in o a common wo k low should enable he e icien
use o machine lea ning me hods o p edic ing e ec i e p ope ies om po e ne wo k
images.
Du ing he explo a ion phase o analyzing possible loca ions o he deep geological
eposi o ies, i is common o ex ac and analyze d ill co es which p o ide signi ican
in o ma ion abou he composi ion and hyd aulic p ope ies o he hos ock. A
p omising app oach o accele a ing he modelling wo k lows is he possibili y o ex ac
EURAD-2 Deli e able 16.1 – S a e-o - he-a on high ideli y nume ical simula ions o s ongly coupled
p ocesses o eposi o y sys ems (Ini ial epo )
Dissemina ion le el: Public
Da e o issue o his epo : 29/09/2025 35
he mine alogical con en o d ill co es om hei espec i e high esolu ion images. As
demons a ed in (Boige e al., 2024), ele an o he Swiss deep d illing explo a ion
p og am, i is possible o co ela e he pe ophysical p ope ies o ocks using machine
lea ning (models build on con olu ional neu al ne wo ks) and d ill co e images as inpu .
The achie ed accu acy o selec ed samples was equi alen o labo a o y XRD
measu emen s on samples om he co es. Such me hodology can gene a e almos
ins an ly e y high- esolu ion bo e-hole geological models, o eac i e anspo
simula ions, upon a ailabili y o d ill co e images educing he ime and esou ces
needed o labo a o y analysis.
3.1.4.4 Su oga e geochemical models o eplace he geochemical sol e in eac i e
anspo models in o de o speedup compu a ions
Recen ly, he accu acy o a ious machine lea ning (ML) me hods was e alua ed in
he con ex o ba ch chemical eac ions by (P asianakis e al., 2025). A na u al ollow-
up ques ion is compu a ional pe o mance and cumula i e accu acy o ML emula ion
o geochemical modelling in eac i e anspo simula ions, and mo e speci ically wi h
an in e ace (e.g. clay / cemen ). The complexi y will be in eg a ed sequen ially, s a ing
wi h a simple eac i e anspo sys em wi hou an in e ace. I is planned o e alua e
he bes possible pe o mance o accele a ed eac i e anspo on his simple sys em.
Wi h he democ a isa ion o a i icial in elligence, so did he p oli e a ion o new
a chi ec u es and echniques o dealing wi h eg ession asks. A ew s udies ha e
compa ed a ious ype o machine lea ning me hods in he con ex o eac i e
anspo , anging om mo e adi ional MLPs (mul ilaye pe cep ons) (Demi e e al.,
2023; Ja nieks e al., 2016; Laloy & Jacques, 2022; P asianakis e al., 2025) o mo e
ecen a chi ec u es such as ans o me s (Sil a e al., 2024). O e all, i seems ha a
pu ely da a-d i en app oach lacks he p ecision and s abili y equi ed du ing eac i e
anspo . This is pa icula ly due o he au o- eg essi e na u e o he p oblem and
su oga e models. An al e na i e is o ely on adi ional geochemical sol e s bu
educe hei cos by building a da abase o chemical con igu a ions which can hen be
used o pe o m an in e pola ion as s udied by Leal e al. (2020).
On hese bases, a hyb id app oach, combining bo h su oga e models (such as neu al
ne s) and adi ional geochemical sol e simula ions appea s o be a p omising way
o wa d. This would equi e o de elop me ics o es ima es du ing a eac i e anspo
simula ion o a model’s d i , which when iden i ied would be eplaced by a adi ional
geochemical sol e e alua ion. Se e al me ics can be es ed, o ins ance localised o
global mass o cha ge balance; ins an aneous e o be ween su oga e and
geochemical sol e a disc e e poin s in space and ime, e c.
EURAD-2 Deli e able 16.1 – S a e-o - he-a on high ideli y nume ical simula ions o s ongly coupled
p ocesses o eposi o y sys ems (Ini ial epo )
Dissemina ion le el: Public
Da e o issue o his epo : 29/09/2025 36
Go ae s (2022), wi hin EURAD WP-ACED, p o ided a comp ehensi e compila ion o
model abs ac ion echniques, which includes a g oup o simpli ica ion me hods
ca ego ised as lowe - ideli y nume ical models. These me hods employ s a egies such
as p ede ined model hie a chies, delimi ing inpu domains, scaling changes, and
educing nume ical accu acy. A second g oup o me hods is based on s a is ically
de i ed su oga e models (o me amodels), which aim o simula e he inpu -ou pu
ela ionships o complex models h ough s a is ical ela ionships, wi hou conside ing
he unde lying physics. Su oga e models can be used o emula e ull eac i e anspo
models, eplace he geochemical sol e in eac i e anspo codes, and subs i u e
compu a ionally expensi e p ocess models. Go ae s (2022) demons a ed se e al
cases in which he geochemical model was eplaced by a da a-d i en su oga e model.
Sampe (2024), also wi hin EURAD WP-ACED, p esen ed he applica ion o model
abs ac ion echniques o assessing chemical e olu ion o he HLW and ILW a
disposal cell scales in g ani ic and clay ocks. They in oduced a me amodel o a
simple geochemical sys em ep esen ing s eel/ben oni e in e ac ions and he
p ecipi a ion o co osion p oduc s in an HLW disposal cell in g ani e, based on
Gaussian P ocesses. Sampe (2024) also p oposed a su oga e model o an ILW
disposal cell in g ani e o eplace he geochemical sol e in a eac i e anspo code,
u ilizing ei he a ained deep neu al ne wo k (DNN) o a k-nea es neighbo (kNN)
eg esso .
The Random Fo es Gaussian P ocesses (RF-GP) echnique combines a RF model o
disc imina e be ween a ew “cha ac e is ic geochemical egions” wi h he
co esponding GPs o make a p edic ion. The echnique equi es a low numbe o
pa ame e s, signi ican ly educing he likelihood o model o e i ing and p o iding
es ima es o he accu acy o su oga e p edic ions. The main limi a ions o RF-GP a e
he la ge compu a ional esou ces and memo y equi emen s o aining models wi h
la ge da ase s, pa icula ly o N > 10,000 poin s. To make he me hod applicable o
la ge sys em he e is a need o: 1) Imp o e he pe o mance o RF-GP in ela ion o
he me ics o he cemen case; 2) Enhance he es ima ion o e y low aqueous and
so bed u anium concen a ions in he u anium benchma k case; and 3) Le e age he
con idence in e als o ou pu es ima es p o ided by RF-GP. The bene i s o GPs
should also be explo ed in ela ion o he p ese a ion o mass and cha ge balance.
Se e al u u e ML-based eac i e anspo benchma king cases could be conside ed,
in addi ion o hose al eady ini ia ed in EURAD WP-DONUT. These would in ol e
geochemical sys ems o inc easing complexi y and size, as well as sys ems a eac i e
in e aces, such as glass/co osion p oduc s, co osion p oduc s/cemen , co osion
p oduc s/clay, clay/cemen , and conc e e/g ani e.
EURAD-2 Deli e able 16.1 – S a e-o - he-a on high ideli y nume ical simula ions o s ongly coupled
p ocesses o eposi o y sys ems (Ini ial epo )
Dissemina ion le el: Public
Da e o issue o his epo : 29/09/2025 37
Ja nieks e al. (2016) we e p obably he i s o poin ou ha eplacing he solu ion o
geochemical p ocesses in coupled eac i e anspo models wi h da a-d i en
su oga es ep esen ed a p omising way o educe he compu a ional bu den
associa ed wi h his class o simula ions. In pa icula , he imp o emen s o heo e ical
unde s anding and echnical implemen a ions o many mul i a ia e gene al
app oxima o s by he e ile ields o machine lea ning and a i icial in elligence, such
as Mul i-Laye Pe cep ons (MLP), we e es ima ed o each h oughpu s o many
o de s o magni ude highe han a geochemical sol e , a p ice o mode a e accu acy
loss, also p o i ing om ha dwa e accele a o s o e en inc ease hei compu a ional
e iciency bo h in he aining and in he in e ence phase. Howe e , he numbe o
simula ions needed o aining su oga es and he aining p ocess i sel could p o e
cumbe some and leng hy in a pu ely da a-d i en app oach; mo eo e , a su oga e
such as ANN (a i icial neu al ne wo k) is a all e ec s a "black box” and canno be
in e p e ed by domain expe s. Many a emp s o a ain explainable AI and o hyb idize
a i icial in elligence’s me hods wi h physical o domain knowledge ensued in he las
yea s, such as PINN (Physics In o med NN), bu wi h no applica ion o geochemis y
ye . De Lucia and Kühn (2021) showed ha e en o qui e simple geochemical
sys ems pu e da a-d i en app oach ailed o cap u e many non-linea i ies, such as a
mine al anishing o s a ing o p ecipi a e, and possess limi ed ex apola ion capaci y
ou side he aining pa ame e space. Le e aging knowledge abou he ela ionships
be ween a iables (s oichiome y o eac ions, mass ac ion laws, mass and cha ge
balance) hey cons uc ed a hyb id su oga e which iden i ied egions in he pa ame e
space wi h he same numbe o deg ees o eedoms, and p oceeded o u ilize simple
mono a ia e o low-dimensional eg ession echniques as p edic o s. The same
p inciples we e employed by De Lucia (2024) in applica ion o u anium su ace
complexa ion on clay. Besides he su oga e modelling wi h AI/ML me hods, o he
algo i hmical imp o emen s we e in es iga ed. Among hose he caching in hash
ables and euse o p e ious geochemical simula ions we e demons a ed o p o ide
signi ican speedup in ini ially homogeneous, ad ec ion-domina ed eac i e anspo
simula ions (De Lucia e al., 2021), in pa icula , he massi ely pa allel RTM POET
implemen ed caching in e icien Dis ibu ed Hash Tables in he con ex o HPC.
Replacing he geochemical componen o a eac i e anspo code by a ained ML
algo i hm is a p omising way o accele a e eac i e anspo simula ions. This has
al eady been demons a ed by Laloy and Jacques (2022) o simple o mode a ely
complica ed cemen i ious sys ems. Howe e , Laloy and cowo ke s also shown ha o
cemen sys ems o mode a e complexi y al eady, issues appea wi h (1) he
ep esen a i eness o he aining se used o lea n he ML algo i hm pa ame e s and
(2) he accumula ion o ML-based p edic ion e o s du ing he eac i e anspo
simula ion. Indeed, while he geochemical pa ame e space in which he eac i e
anspo is occu ing is ypically a small subspace o he pa ame e space used o
gene a e he aining se , he su oga e geochemical needs o be highly accu a e o e
ha subspace. Fu he mo e, seemingly small indi idual su oga e p edic ions o
EURAD-2 Deli e able 16.1 – S a e-o - he-a on high ideli y nume ical simula ions o s ongly coupled
p ocesses o eposi o y sys ems (Ini ial epo )
Dissemina ion le el: Public
Da e o issue o his epo : 29/09/2025 38
ce ain poin s in space and ime can accumula e o e he cou se o he eac i e
anspo simula ion, he eby leading o la ge biases in he ML-accele a ed eac i e
anspo esul s.
These p oblems can appea a wo spa ial scales o in e es : (1) he con inuum scale,
whe e he Richa ds equa ion applies o a iably-sa u a ed low modelling, and (2) he
po e scale whe e he la ice-Bol zmann me hod (LBM) is classically employed o
simula e anspo . In he la e case, he YANTRA LBM sol e by KIT IMB/MPA can
be used. A ew 1D and 2D cemen i ious sys ems o inc easing geochemical complexi y
should be de ised o in es iga e hese p oblems and hei po en ial solu ions.
Rega ding he geochemical su oga e model echnique, deep neu al ne wo ks (NNs)
ha e been used a lo , mainly due o hei abili y o handle mul i-ou pu eg ession
p oblems and hei la ge compu a ion speed using mode n lib a ies and GPU ha dwa e
(Demi e e al., 2023; Laloy & Jacques, 2022; P asianakis e al., 2020). Ye i appea s
ha pu ely black-box NNs alone may no be handle he high nonlinea i y o he
geochemical p ocesses happening du ing eac i e anspo in cemen sys ems wi h a
su icien ly high accu acy (Laloy & Jacques, 2022). As w i en a ew imes al eady in
his epo , d awing inspi a ion om he physics-in o med neu al ne wo k (PINN)
amewo k, whe e some physics cons ain s a e added o he NN aining loss unc ion,
migh be a p omising way o wa d. In addi ion, blending ML p edic ions wi h ull
geochemical calcula ions o some g id nodes a some speci ic imes migh p o e
use ul o co ec he eac i e anspo ajec o y while s ill e aining an a ac i e speed
gain compa ed o a ully physics-based simula ion. This could be done andomly (Sil a
e al., 2024) hough ideally, a c i e ion could be de ised o decide when and whe e
igge ing a ull geochemical calcula ion.
3.1.4.5 Su oga e models o model in-si u expe imen s
The physics-in o med neu on ne wo ks (PINNs) a e p omising ools o sol ing
e icien ly coupled he mo-hyd omechanical p oblems. Some speci ic PINNs ha e
been de eloped o sol ing non-iso he mal mul iphase po omechanics he mo-
hyd omechanical p oblems (Amini e al., 2023). Howe e , mos p e ious PINNs a e
de o ed o elas ic po ous media. In he con ex o geological disposal in clayey
o ma ions, he hos ocks exhibi plas ic de o ma ion and induced damage and
c acking. Va ious cons i u i e models ha e been de eloped o such ma e ials (Chen
e al., 2024; Zhao e al., 2022). Fo modelling induced damage and c acking p ocess,
di e en ypes o nume ical me hods a e a ailable. In pa icula , he so-called phase-
ield me hod has been applied o modelling in-si u he mo-hyd omechanical
expe imen s in he unde g ound esea ch labo a o y o ANDRA (Shao e al., 2025; Yu,
Shao, Du eau, e al., 2024; Yu, Shao, & Vu, 2024). New PINNs model o mula ions
could help o accele a e induced damage and c acking simula ion.
EURAD-2 Deli e able 16.1 – S a e-o - he-a on high ideli y nume ical simula ions o s ongly coupled
p ocesses o eposi o y sys ems (Ini ial epo )
Dissemina ion le el: Public
Da e o issue o his epo : 29/09/2025 39
Du ing he EURAD “Moni o ing Equipmen and Da a T ea men o Sa e Reposi o y
Ope a ion and S aged Closu e” (EURAD-MODATS) p ojec , digi al wins we e
de eloped o simula e empe a u e e olu ion in he Full-Scale Emplacemen (FE)
expe imen a Mon Te i (Hu & P ings en, 2023; Hu e al., 2024; Hu e al., 2023). A
e y high esolu ion geome ical model was de eloped o inco po a e all
he e ogenei ies o he sys em and o in eg a e he senso signals. A physical model o
he empe a u e ield was implemen ed while he humidi y e olu ion and sa u a ion was
in eg a ed in he p ojec by using he eal- ime senso da a and machine lea ning. By
combining hese echniques i was possible o inc ease signi ican ly he accu acy o
he simula ions and o p o ide in o ma ion abou possible aul y senso da a.
Accele a ion o calcula ions and a sensi i i y analysis was also conduc ed using
su oga e models. While hese models can se e as an accompanying eal- ime
assis an s hey lack p edic i e capabili y mainly due o he lack o a hyd ological
physical model, which is he na u al o eseen model ex ension.
3.1.5 Benchma k sui es
Benchma king is ge ing mo e and mo e impo an as he complexi y o nume ical
codes inc eases. As numbe o coupled p ocesses inc eases, he benchma king
p ocedu e becomes mo e expensi e. Fu he mo e, o es he di e en p ocess
coupling, a hie a chical app oach is equi ed, i.e. s a ing wi h indi idual p ocesses and
inc easing complexi y in an sys ema ic way (Lehmann e al., 2023; Mollaali e al., 2023;
Pi z e al., 2023). Ano he challenge is he au oma ion o benchma king p ocedu es,
i.e., au oma ed es ing wo k lows in Gi eposi o ies. Pos -p ocessing should also be
au oma ed and ailed benchma ks should be epo ed o he co esponding au ho wi h
code e ision eques . Jupy e no ebooks p o ide a powe ul en i onmen o
au oma ed benchma king p ocedu es and can e en be un on web pla o ms (e.g.
Binde ) o in e ac i e p ocessing. La ge benchma king p ojec s such as DECOVALEX
o wi hin EURAD should ake ad an age o hese new echnologies o au oma ed
benchma king, which will equi e addi ional and conce ed e o s. As examples,
OpenGeoSys (h ps://www.opengeosys.o g/docs/benchma ks/) and FEniCS
(h ps:// enicsp ojec .o g/) al eady o e comp ehensi e benchma k collec ions as
Jupy e no ebook ia web pla o ms.
Sampe (2024) p esen ed a benchma k s udy on mul iphase low and eac i e
anspo modelling in he con ex o adioac i e was e disposal. The s udy ocused on
a 1D column o unsa u a ed ben oni e h ough which wa e , d y ai and CO₂(g) low.
The model included se e al key p ocesses, including aqueous complexa ion,
dissolu ion/p ecipi a ion o calci e and gypsum, ca ion exchange and gas dissolu ion.
The benchma k in ol ed ou modelling codes: INVERSE-FADES-CORE V2, DuMuX,
TOUGHREACT and iCP, and was es ed o e six inc easingly complex scena ios.
These anged om conse a i e ace anspo unde a iable unsa u a ed condi ions
o mo e complica ed simula ions in ol ing wa e low, gas di usion, mine al eac ions
and ca ion exchange. The esul s showed ha all codes p oduced simila o e all
ends. Howe e , small di e ences we e obse ed in conse a i e ace anspo ,
EURAD-2 Deli e able 16.1 – S a e-o - he-a on high ideli y nume ical simula ions o s ongly coupled
p ocesses o eposi o y sys ems (Ini ial epo )
Dissemina ion le el: Public
Da e o issue o his epo : 29/09/2025 40
CO₂(g) concen a ions and pH alues nea he no- low bounda y. Mos o hese
di e ences diminished wi h ime and we e la gely esol ed when all codes used he
same Debye-Hückel app oach o aqueous specia ion.
Recen ly, P asianakis e al. (2025) p esen ed a benchma king exe cise ha aims a
p o iding a se o e e ence da a and models o de eloping and applying ML
echniques o su oga e geochemical calcula ions. Se e al well-known geochemical
specia ion codes we e used o gene a e chemical equilib ium da a o wo main
benchma ks: (i) cemen chemis y wi h inc easing le el o complexi y and (ii) u anium
so p ion on a clay mine al. The pe o mance o di e en ML su oga e echniques was
hen e alua ed in e ms o hei nume ical e iciency and accu acy. The esul s
highligh ed ha depending o he es case and used ML echnique, he e iciency gain
p o ided by ML models can be o wo o se e al o de s o magni ude, compa ed o he
s anda d use o he geochemical sol e s. A op ion o “wa m es a ” o ini ial guess
p o ide u he e iciency gains o he ORCHESTRA geochemical sol e (Meeussen,
2003), wi hou howe e su passing he e iciency o he op-e icien su oga e models.
The exac pe o mance gains du ing ealis ic eac i e anspo simula ions emain o
be in es iga ed. In p ac ical compu a ions ac ual compu a ional speedup s ongly
depends on he coupling and he e iciency o he low sol e and he sys em unde
conside a ion.
3.2. Me hods and ools o unce ain y analysis o THCM models
3.2.1 Upscaling
One o he majo di icul ies in modelling he beha iou o a eposi o y and assessing
i s easibili y is ela ed o he exis ence o se e al leng h and ime scales ha a e
cha ac e is ic o he di e en p ocesses in ol ed. These scales ange om nanosized
po osi y p esen in clay laye s o he la ge-scale he e ogenei ies o he eposi o y and
he a ield. Simila ly, in c ys alline ocks, he size o indi idual ac u es ollows a
powe -law dis ibu ion (Bonne e al., 2001). These he e ogenei ies a e ex emely
di icul o model because o he wide ange o scales in ol ed, and di e en nume ical
s a egies ha e eme ged o model he mac oscopic beha iou o clay o ma ions. A
i s solu ion o his scale dispa i y p oblem is o apply upscaling echniques (e.g.
homogenisa ion, olume a e aging, momen ma ching echniques) ha ela e he
mic oscopic pa ame e s o he obse able mac oscopic ones. This is achie ed by
de eloping mac oscopic anspo equa ions exp essed in e ms o he spa ial (o
ensemble) a e ages o he p ocesses hemsel es. The coun e pa is he di icul y o
dealing wi h complex physical couplings, including non-equilib ium si ua ions, non-
linea i y o in e acial dynamics. Al e na i ely, nume ical upscaling based on di ec
nume ical simula ions (DNS) can also be used (Pazdniakou e al., 2018; Tine e al.,
2020). This implies ha he balance equa ions a e assumed o be known a he scale
o in e es . The e ec i e coe icien s (e.g. ela i e pe meabili y, damage a iable,
he mal conduc i i y, dispe si i y) and he s a e laws desc ibing hei e olu ion, which
EURAD-2 Deli e able 16.1 – S a e-o - he-a on high ideli y nume ical simula ions o s ongly coupled
p ocesses o eposi o y sys ems (Ini ial epo )
Dissemina ion le el: Public
Da e o issue o his epo : 29/09/2025 41
appea in he mac oscopic sys em o equa ions, a e hen de i ed di ec ly by in eg a ion
o he DNS explici ly desc ibing he couplings. The ad an age is o bene i om a ine
desc ip ion a lowe scales o he coupled p ocesses, bu also o he a chi ec u e o he
he e ogenei ies and hei connec i i y.
To desc ibe he s ochas ic he e ogenei y in ac u ed ocks, se e al app oaches a e
commonly used:
1. Con inuum Fields (CF), whe e ock p ope ies a e modelled as scala o
enso - alued andom co ela ed ields.
2. Disc e e F ac u e Ne wo ks (DFN), whe e a sys em o disc e e ac u es is
sampled om a s ochas ic model o indi idual ac u e p ope ies.
3. Disc e e F ac u e-Ma ix (DFM), which couples he con inuum and disc e e
ac u e ep esen a ions, p o iding a lexible in e pola ion be ween he wo.
Since he ac u e sizes in he DFN desc ip ion o en ollow a powe -law dis ibu ion,
he co ela ed ields in he CF desc ip ion a e o en assumed o ha e a co esponding
powe -law co ela ion (Neuman, 2008). In o de o use DFM o cons uc a lexible
mul i- ideli y model, explici upscaling om he DFN o he CF desc ip ion is equi ed.
To achie e his, con olu ional neu al ne wo ks ained on DNS samples ha e been
used (Špe lík e al., 2024).
3.2.2 Sensi i i y analysis
Reac i e anspo simula ions a e ypically compu a ionally in ensi e ope a ions and
sensi i i y analyses (e.g. Sobol analysis) ypically equi e se e al housands o such
simula ions. In Laloy and Jacques (2019) i was shown ha i is possible o c ea e
su oga e models o he ull sol e s (emula o s) and acco dingly pe o m he sensi i i y
analysis based on hese models. Se e al me hods o c ea e he su oga e models we e
es ed, wi h Deep Neu al Ne wo ks (DNN) and Gaussian P ocesses (GP) being he
mos p omising, comple ing he sensi i i y analysis signi ican ly as e wi h easonable
accu acy.
Chu ako e al. (2024) highligh ed he impo ance o unce ain y ea men in he
eac i e anspo models used o pe o mance assessmen in deep geological
disposal sys ems, employing local sensi i i y analysis (LSA) and global sensi i i y
analysis (GSA) me hods. Sampe J (2025) emphasised ha he ou pu s o eac i e
anspo models a e ela ed o inpu pa ame e s h ough complex nonlinea
ela ionships. The impac o pa ame e unce ain y on model ou pu s is add essed
h ough sensi i i y analysis. Sophis ica ed me hods o sensi i i y analysis ha e been
de eloped o nume ical models based on he analysis o model esul s. LSA me hods
quan i y sensi i i y in he icini y o speci ic pa ame e se s, while GSA me hods
e alua e model sensi i i y ac oss selec ed pa ame e in e als. GSA me hods include
Mo is’s "elemen a y e ec s" and i s a ian s (Campolongo, 2007; Mo is, 1991), as
well as Sobol indices (Homma, 1996; Rabi z, 1999; Sobol, 1990; Sobol, 1993; Sobol,
EURAD-2 Deli e able 16.1 – S a e-o - he-a on high ideli y nume ical simula ions o s ongly coupled
p ocesses o eposi o y sys ems (Ini ial epo )
Dissemina ion le el: Public
Da e o issue o his epo : 29/09/2025 48
compu a ional needs e ol ed, newe pla o ms began o inco po a e mo e lexible
in eg a ion s a egies. MOSSCO and Hyd oCouple, o example, used s anda dised
da a o ma s such as Ne CDF and XML, allowing models o in e ace wi h a wide ange
o ex e nal da ase s. Hyd oCouple, in pa icula , was designed o in eg a e eal- ime
hyd ological da a om senso ne wo ks, making i highly sui able o ope a ional wa e
managemen applica ions. Wi h he ise o cloud-based compu ing, pla o ms such as
C3F and Salome (Asch 2016) ha e in oduced API-d i en da a in eg a ion me hods.
By using REST ul APIs, hese pla o ms allow models o dynamically in e ac wi h
emo e da a sou ces, elimina ing he need o p e-con igu ed da a pipelines. This
app oach signi ican ly imp o es he scalabili y and adap abili y o coupled simula ions,
allowing esea che s o inco po a e eal- ime da a s eams om di e se sou ces.
Recen ad ances include machine lea ning and deep lea ning echniques o imp o e
da a assimila ion. Two app oaches a e o pa icula in e es : da a usion (AI echniques
o me ge he e ogeneous da ase s (e.g. sa elli e image y, senso measu emen s and
clima e models) o imp o e simula ion accu acy) and au oma ed ea u e ex ac ion
(deep lea ning me hods o ex ac ele an pa e ns om la ge da ase s o suppo eal-
ime model adjus men s). Se e al open sou ce lib a ies p o ide eusable ools o
in eg a ing AI-d i en da a handling in o simula ion amewo ks. Finally, an example
om a comple ely di e en ield is a collabo a i e and e y ad anced p ojec called
IMAS (ITER In eg a ed Modelling and Analysis Sui e), a so wa e in as uc u e ha
p o ides a s anda d amewo k o da a exchange and code in e acing o he scien i ic
exploi a ion o ITER and o he okamaks (Imbeaux e al., 2015).
3.4. Applica ion o p ocess-based models o ield expe imen s and eposi o y
sys ems
3.4.1 Real ime simula ion o expe imen s
Gas anspo can be simula ed wi h CODE_BRIGHT (COuple De o ma ion BRIne,
Gas and Hea T anspo ) assuming single ac u es embedded in con inuous ini e
elemen s (A nedo e al., 2013; A nedo e al., 2008; Oli ella, 2024). These ac u es
migh open and close as unc ion o he s ains, which depend on he s ess s a e.
These ac u es can ep esen discon inui ies ela ed o bedding planes induced by
ock sedimen a ion o bu e ma e ial compac ion, con ac be ween di e en ma e ials,
c acks induced by he mal desicca ion o discon inui ies de e mined by an app op ia e
mechanical cons i u i e model ha de e mines hem om s ess and s ain o ien a ion.
The e can also be p e e en ial gas low pa hs in ma e ials ha do no ha e ac u es
embedded, like he bu e . In his case, he low pa hs migh appea due o he
he e ogenei y o he ma e ial (Rod iguez-Dono e al., 2023). Bo h mechanisms o gas
low migh be combined (Nogh e ab e al., 2024; Rod iguez-Dono e al., 2024).
EURAD-2 Deli e able 16.1 – S a e-o - he-a on high ideli y nume ical simula ions o s ongly coupled
p ocesses o eposi o y sys ems (Ini ial epo )
Dissemina ion le el: Public
Da e o issue o his epo : 29/09/2025 49
T-H-M-C modelling in CODE_BRIGHT was ini ially enabled by he implemen a ion o
a chemical sub ou ine (Gens e al., 2010; Guima aes e al., 2009). The calcula ion
p ocess was sol ing he THM balance equa ions (ene gy, solids, ai , wa e and
momen um) and he chemical equa ions (balance equa ions ela ed o kine ics, and
equilib ium) sepa a ely. The HMC equa ions we e sol ed oge he in a second e sion
o he chemical implemen a ion in CODE_BRIGHT. The code was applied o s udy he
swelling due o he dissolu ion o anhyd i e and he p ecipi a ion o gypsum (Ramon e
al., 2017). Cu en ly, he chemical equa ions a e implemen ed in CODE_BRIGHT and
sol ed oge he wi h he THM equa ions (Ramon, 2024).
Full-Scale In-Si u Sys em Tes (FISST) a e cu en ly ongoing in he ONKALO®
demons a ion a ea, pa o he Finnish spen nuclea uel disposal acili y unde
commissioning phase (Posi a). Di e en es s we e ca ied ou o he cha ac e isa ion
o he clay ma e ials, Wyoming ben oni e in bu e (Top ak e al., 2024) and I alian and
Bulga ian ben oni es in back ill. THM modelling has al eady been pe o med
(Tsi sopoulos e al., 2023). T-H-M-C modelling can be ca ied ou using a THC
compu e code combined wi h a mechanical (M) compu e code. Ru q is (2014),
Zheng e al. (2015) and (Zheng & Fe nández, 2023) and ha e pe o med T-H-M-C
simula ions combining he THC compu e code TOUGHREACT and he M compu e
code FLAC. The mechanical coupling can be limi ed o ce ain bounda y condi ions
p oblems. Tes in isocho ic condi ions can be simula ed wi h CRUNCHFLOW (Yus es
e al., 2017) o ORCHESTRA (Jenni e al., 2021), whe e double po osi y models
capable o simula e complex THC p ocesses a e combined wi h simple mechanical
and empi ical laws ha p edic chemically induced swelling o in e laye s and calcula e
he swelling p essu e.
When alida ing nume ical models by compa ing hei esul s wi h measu ed da a, he
in luence o geological and geo echnical bounda y condi ions like inhomogenei ies,
ac u es, localised ea u es becomes e iden a all scales (Le asseu e al., 2024).
Wi h la ge scales, hese in luencing ac o s become e en mo e appa en due o he
inc easing complexi y o he sys em on he one hand, bu also due o he limi ed da a
a ailable. Howe e , compa ing measu emen s and modelling esul s on he ield scale
o e s he oppo uni y o achie e an enhanced p ocess and sys em unde s anding by
iden i ying essen ial e ec s and bounda y condi ions like di e en suppo sys ems o
en ila ion and imp o ing model app oaches and se -up acco dingly, as done e.g. in
he DECOVALEX p ojec (Bi kholze e al., 2019).
In-si u expe imen s, such as hose ca ied ou in unde g ound labo a o ies like he Mon
Te i ock labo a o y, enable a comp ehensi e da a basis o his kind o in es iga ions.
Some examples can be ound in Lee e al. (2021) and G aupne e al. (2025). P ocess
a iables such as de o ma ion, wa e con en , po e p essu e and o he s a e moni o ed
o e he long e m and a comp ehensi e cha ac e isa ion o he hos ock p o ides a
EURAD-2 Deli e able 16.1 – S a e-o - he-a on high ideli y nume ical simula ions o s ongly coupled
p ocesses o eposi o y sys ems (Ini ial epo )
Dissemina ion le el: Public
Da e o issue o his epo : 29/09/2025 50
aluable basis o a sound sys em unde s anding (Bossa e al., 2017). Imp o ed
in es iga ions o coupled hyd aulic-mechanical e ec s a e planned in he con ex o
HERMES. In addi ion o he u he de elopmen o classical physical modelling
me hods, machine lea ning (ML) app oaches o e a aluable oppo uni y o use he
collec ed long- e m in-si u da a as an ex ension o classical geophysical app oaches
and hus make mo e ealis ic s a emen s o ele an a eas, e.g. in he nea ield o
exca a ions (Hu e al., 2023; Zie le e al., 2024).
Unde he coo dina ion o ANDRA, a numbe o in-si u expe imen s ha e been
conduc ed in he unde g ound esea ch labo a o y a Bu e (A mand e al., 2017;
Bumbiele , 2020; Conil e al., 2020; Plua, 2023; Tou chi, 2021). These expe imen s
p o ide aluable expe imen al da a on he e olu ions o po e liquid p essu e, gas
p essu e, empe a u e and de o ma ion as well as he dis ibu ion o damaged zones.
Some pa s o hose esul s ha e been used in he p e ious phases o Deco alex
p ojec s (Yu, Shao, Du eau, e al., 2024; Yu, Shao, & Vu, 2024; Yu e al., 2021). In he
cu en phase o Deco alex, some new in-si u es s will be conduc ed, mainly de o ed
o gas injec ion. I is expec ed o de elop and imp o e nume ical models o modelling
hyd omechanical coupling p ocesses in pa ially sa u a ed po ous media wi h mul i-
phase luid low. A se ies o new eal ime h ee-dimensional simula ions o in-si u
expe imen s will hen be ealised, wi h o he pa icipan s o he Deco alex p ojec .
3.4.2 Was e package in eg i y and nea - ield
To assess he adioac i e was e disposal sys ems, he modelling o he chemical
e olu ion o a disposal cell is an impo an poin . The modelling o he inne mos pa s
o he disposal sys em gi es insigh in he du abili y o he di e en nea - ield
componen s bu as well in he chemical backg ound o modelling he specia ion and
mig a ion o adionuclides. These models can se e as aluable inpu s o e alua ing
pe o mance (du abili y) and sa e y (chemical backg ound). Reac i e anspo models
(RTM) play a c ucial ole in comp ehending and assessing he in e play o he mal,
hyd ological, and geochemical p ocesses wi hin hese con ainmen ba ie s. These
ba ie s a e an icipa ed o endu e a ious empe a u e anges and geochemical
condi ions while needing o uphold hei in eg i y o o e 100,000 yea s.
A comp ehensi e o e iew on he modelling o he was e package deg ada ion is
p esen ed in Jacques e al. (2024). This epo con ains he compila ion o ecen
modelling ac i i ies ca ied ou in he amewo k o EURAD. The documen p esen s
concep ual and ma hema ical models ha illus a e he geochemical e olu ion a
in e aces in ol ing s eel. I ou lines a p ac ical app oach o concep ualizing s eel
co osion wi hin coupled eac i e anspo models. Addi ionally, i co e s model
concep s, implemen a ions, and esul s o in e aces wi h ei he clay o cemen , bo h
EURAD-2 Deli e able 16.1 – S a e-o - he-a on high ideli y nume ical simula ions o s ongly coupled
p ocesses o eposi o y sys ems (Ini ial epo )
Dissemina ion le el: Public
Da e o issue o his epo : 29/09/2025 51
in he p esence and absence o glass. The models de ail a ious expe imen al se ups
and condi ions, wi h da a spanning om a ew mon hs o se e al yea s.
Du ing he co osion o s eel in he eposi o y en i onmen , an ini ial ae obic co osion
phase is ollowed by an anae obic co osion phase. The a e o he co osion du ing
he ae obic phase can be desc ibed wi h a cons an a e cons an and a i s -o de
kine ic a e wi h espec o oxygen.
When used in eac i e anspo models, di e en app oaches a e possible:
• Flux bounda y model – he s eel is no explici ly p esen in he anspo domain,
bu implici ly a a bounda y wi h a kine ic Fe- lux owa ds he domain. Based on
he solu ion chemis y o he con ac ing node (o he back ill ma e ial in con ac
wi h he s eel), he s eel co osion a e is calcula ed and he calcula ed amoun
o Fe is added o he solu ion (and, i ele an , eac an s a e emo ed). The
o ma ion o co osion p oduc s can hus only occu in he back ill ma e ial,
di usion is no possible om back ill in o he s eel as in he o he wo app oach
in which (pa o ) he s eel is ep esen ed as a po ous medium;
• Po ous medium ep esen a ion – he s eel is ep esen ed as a po ous medium
wi h a (low ini ial) po osi y. Co osion occu s h oughou he comple e s eel
c oss-sec ion eleasing Fe in he aqueous phase in he s eel and subsequen
di usion owa ds he su ounding ma e ials.
• Laye -by-laye co osion (Figu e 5) – he c oss-sec ion o he s eel is di ided
in o a numbe o small laye s. Co osion only occu s in he ou e mos laye ( he
supe icial laye ); co osion in deepe laye s canno s a i i on emains in he
supe icial laye . Thus, he s eel is ep esen ed as a po ous medium only in he
supe icial laye ; deepe laye s a e no po ous and a e impe meable.
Figu e 5 – Rep esen a ion o he laye -by-laye app oach (a e Lemmens e al. (2024)) in
which only he ou e (supe icial) laye o s eel is a po ous medium and can co ode.
In case o modelling S eel-Clay in e aces he o e all goal is o imp o e and e alua e
he modelling app oaches o complex coupled Fe co osion and Fe-clay in e ac ion
p ocesses wi h coupled eac i e anspo codes (Wi eb ood e al., 2024b).
Ad ancemen s ha e been made wi h espec o:
• Ini ial unsa u a ed condi ions in he ben oni e wi h esa u a ion (unde a
empe a u e g adien ) and swelling aken in o accoun
• Ini ial ae obic condi ions in ben oni e wi h subsequen ansi ion o anae obic
condi ions
EURAD-2 Deli e able 16.1 – S a e-o - he-a on high ideli y nume ical simula ions o s ongly coupled
p ocesses o eposi o y sys ems (Ini ial epo )
Dissemina ion le el: Public
Da e o issue o his epo : 29/09/2025 52
• Modelling he Fe a e in he s eel-ben oni e sys em including:
• P ecipi a ion in he s eel co osion zone
• So p ion by ca ion exchange and su ace complexa ion
• Elec on ans e om Fe(II) o s uc u al Fe(III) in ben oni e
In case o modelling was e packages h ee app oaches we e ou lined:
1. Mixing ank app oach – This app oach accoun s o he geochemical p ocesses in he
was e d ums o package bu neglec s any spa ial aspec o anspo p ocess
(Kosakowski, 2020; Wieland, 2020, 2018). I is a as and simple me hod o assess
mass balances and gi es some indica ions on impac o ma e ial. The app oach
includes bo h he modynamic equilib ium eac ions (cemen ca bona ion p ocesses)
and ime-dependen deg ada ion p ocesses (co osion, o ganic ma e ial deg ada ion,
agg ega e eac ions).
2. Models ocused on wa e and gas low – The close in e ac ion be ween wa e
consump ion, p oduc ion and ing ess, and gas p oduc ion, p essu e and low a he
was e package scale (mul iple was e d ums in a con aine ) is simula ed wi h a
mul iphase eac i e anspo model wi h look-up ables o geochemical p ocesses
( ollowing he app oach desc ibed in Huang e al. (2018, 2021). The model accoun s
o he p ecise geome y and ma e ial dis ibu ion o each was e d um and o he
comple e was e package. An exac geome ical ep esen a ion o (bulk) s eel was e
and in ill mo a is no possible, he inne pa o he was e package is abs ac ed by a
consump ion a e o wa e induced by s eel co osion
3. Coupled eac i e anspo models – This ype o model links de ailed geochemical
calcula ions pe o med wi h a geochemical sol e wi h anspo calcula ions wi h a
anspo sol e . The codes used in his case do no allow mul iphase calcula ions,
he e o e simpli ying assump ion o ully sa u a ed condi ions needed o be aken.
One o he longes - unning p ojec s aiming o es ablish be e unde s anding and
modelling o T-H-M-C (The mo-Hyd o-Mechanical-Chemical) p ocesses is he
DECOVALEX (DE elopmen o COupled THM models and hei VALida ion agains
EXpe imen s) in e na ional esea ch p ojec which is cu en ly in i s nine h phase
(DECOVALEX-2027). The la es comple ed phase publishing inal epo s is
DECOVALEX-2023.
Task C o DECOVALEX aimed a building 3D nume ical models ocusing on he hea
induced po e p essu e changes o he FE (Full-scale Emplacemen ) expe imen
ca ied ou in he Mon Te i Unde g ound Rock Labo a o y (URL) in Swi ze land
(G aupne , 2024). The models by se e al in e na ional eams we e de eloped in a
s epwise ashion ge ing mo e complex in each subsequen s ep. E en in s ep 0 in
which 2D models we e de eloped, some di e ences be ween esul s we e obse ed
esul ing om model sizes and disc e isa ion, and model concep ualisa ion and
assump ions. I demons a ed ha compa ison wi h expe imen s and wi h o he models
p o ides indica ion on unce ain y in model p edic ions. In s ep 1 3D models we e
de eloped and a e an ini ial s age expe imen da a we e p o ided o he modelling
eams o model calib a ion by solely changing ma e ial p ope y pa ame e s. In s ep 2
EURAD-2 Deli e able 16.1 – S a e-o - he-a on high ideli y nume ical simula ions o s ongly coupled
p ocesses o eposi o y sys ems (Ini ial epo )
Dissemina ion le el: Public
Da e o issue o his epo : 29/09/2025 53
u he ea u es (like EDZ and sho c e e) and p ocesses we e added o he models,
and wi h hese a be e ag eemen wi h expe imen da a was achie ed. S ep 3
ep esen s he p edic i e modelling phase, in which models de eloped in s eps 2 and
3 we e used o model he u u e e olu ion o he sys em, and hei ag eemen will be
compa ed wi h da a obse ed in-si u la e on.
Task D o DECOVALEX ollowed a simila app oach, bu he models we e applied o
he Ho onobe ull-scale EBS (Enginee ed Ba ie Sys em) expe imen a he Ho onobe
URL in Japan (Sugi a, 2024). The main di e ence was he inclusion o chemical
p ocesses o model ben oni e-base bu e and back ill ma e ial beha iou . S ep 1 deal
wi h labo a o y in es iga ions, while s ep 2 ex ended i o he URL condi ions. Al hough
his ask achie ed o alida e se e al app oaches applied in he models, and
empe a u e dis ibu ion in he bu e was simula ed well by all eams, wa e con en
showed good ag eemen only on he ou e side o he bu e , esul s o he inne (close
o he hea -emi ing was e package) and middle pa s we e o . Rega ding mechanical
s esses and de o ma ions in he bu e , e en he in-si u measu ed da a by he senso s
could be p oblema ic, and simula ion esul s show di e ences based on whe he
elas ic o elas oplas ic phenomena we e conside ed.
DECOVALEX-2023 was he i s phase in which pos -closu e pe o mance assessmen
(PA) models and me hods we e compa ed (which is con inued in he cu en phase).
Task F did no include in-si u measu emen s, and was spli in o F1, conside ing
c ys alline (Ma ine , 2024), and F2 conside ing sal hos ocks (LaFo ce, 2024). Fo
F1, eams de ined he FEPS (Fea u es, E en s and P ocesses) o be conside ed and
he se o pe o mance measu es agains which modelling esul s had o be compa ed
o. Benchma k p oblems we e designed o compa e he capabili ies o me hods and
models applied o ac u e low and o anspo a di e en scales. The eams we e
able o ap ly model low and anspo in ac u ed media using ei he DFN (Disc e e
F ac u e Ne wo k) and ECPM (Equi alen Con inuous Po ous Medium) modelling
app oaches, o by combining he wo. The di e ences obse ed be ween modelling
esul s a e mainly due o how p ocesses a e conside ed wi hin he eposi o y.
Ano he in e na ional p ojec ocusing on in es iga ing he beha iou o EBS a ull-
scale is he FEBEX (Full-scale Enginee ed Ba ie EXpe imen ) p ojec ca ied ou a
he G imsel URL in Swi ze land. I has also been unning in se e al phases, he cu en
ones, FEBEXe and FEBEX-DP, deal wi h ea ly- ime coupled p ocesses and he
disman ling o he ins alled hea e sys em, espec i ely. Al hough hey ocus on he in-
si u measu emen s, bu also p o ide he oppo uni y o de elop nume ical modelling
ools o T-H-M-C coupled modelling. Sanchez (2023) in es iga ed he e olu ion o he
sys em by using a THM coupled model’s p edic ions compa ed o measu ed
expe imen al da a. They ound ha hei model was able o cap u e he ansien
p ocess well. Kiczka (2024) in es iga ed he eac i e anspo phenomena occu ing
du ing he ea ly-s age o eposi o y closu e in a s eel / ben oni e sys em. They
de eloped a THC eac i e anspo model o desc ibe i on co osion unde bo h
ae obic and anae obic condi ions, he anspo o O2 in gas and liquid phases and he
EURAD-2 Deli e able 16.1 – S a e-o - he-a on high ideli y nume ical simula ions o s ongly coupled
p ocesses o eposi o y sys ems (Ini ial epo )
Dissemina ion le el: Public
Da e o issue o his epo : 29/09/2025 54
chemical e olu ion o he s eel/ben oni e in e ace. They succeeded in modelling he
accumula ion o goe hi e unde ae obic condi ions, and hen he o ma ion o Fe(II) and
Fe(II)/Fe(III) co osion p oduc s unde anae obic condi ions. The model was able o
quali a i ely ep oduce he conca e shape o he Fe accumula ion on in he ben oni e
as obse ed in he expe imen .
The sensi i i y analysis s udy by Swile e al. (2020) in es iga es he anspo o I-129
using an upscaled con inuum model ha inco po a es was e packages, he exca a ion
damaged zone (EDZ), back ill, and a disc e e ac u e ne wo k (DFN). A nes ed
sampling echnique is employed, u ilizing 25 DFN ealisa ions and 40 pa ame e
samples pe DFN, esul ing in a o al o 1,000 simula ions. The DFN is cha ac e ised
using ou g aph me ics o assess ac u e ne wo k connec i i y, while eigh addi ional
model pa ame e s in luence adionuclide anspo . Fi e key quan i ies o in e es
(QoIs) a e analysed, including peak I-129 concen a ion in he aqui e and anspo
luxes. New unce ain y quan i ica ion me hods ha e been applied o he same case
wi hin he la es JOSA g oup epo (Swile e al., 2025).
Reac i e anspo models ocused on he in eg i y o was e packages o HLW o he
chemical e olu ion o he nuclea glass/s eel in e ace a he scale and imescales o a
HLW disposal cell a e limi ed (Bilds ein e al., 2019; Bilds ein e al., 2012; Bilds ein e
al., 2007; De Wind e al., 2006). Modelling he dissolu ion o i i ied was e is a c i ical
aspec o p edic ing he chemical e olu ion in HLW disposal cells in g ani e o clay.
Mon eneg o L (2023), De Wind e al. (2024) and Sampe e al. (2024) p esen ed non-
iso he mal mul icomponen eac i e anspo models o he long- e m geochemical
e olu ion in HLW disposal cells in g ani e and clay. These models ad ance p e ious
app oaches by conside ing he in eg i y o he was e package in e ms o glass
dissolu ion, as well as he in e ac ions be ween glass, s eel, and co osion p oduc s.
3.4.3 Reposi o y design and op imisa ion and digi al wins’ de elopmen
The sa e y case o he pos -ope a ional phase is go e ned by unde lying p inciples as
hey a e o example laid down in he cu en ICRP ecommenda ions (Valen in, 2007).
Any exposu e o wo ke s o he gene al public has o be jus i ied and op imisa ion o
he ecei ed dose is equi ed o ensu e ha as ew pe sons a e exposed wi h as low a
dose as easonably achie able (ALARA p inciple, see o example Oudiz (1986)). By
employing high- ideli y modelling o coupled p ocesses and in a u he s ep by se ing
up digi al wins o a eposi o y sys em his demand o op imisa ion can be ul illed
(Be nie e al., 2017). Se e al s udies al eady demons a e he po en ial o
op imisa ion o he eposi o y design o example conce ning he handling o decay
hea . Kim e al. show in hei s udies dealing wi h he op imised a angemen o spen
uel elemen s and he he mal limi o he bu e ha he e iciency o a disposal sys em
can be inc eased up o a ac o o 2.5, while in o he s udies he possibili y o damage
EURAD-2 Deli e able 16.1 – S a e-o - he-a on high ideli y nume ical simula ions o s ongly coupled
p ocesses o eposi o y sys ems (Ini ial epo )
Dissemina ion le el: Public
Da e o issue o his epo : 29/09/2025 55
due o inc eased he mo-mechanical s ess is modelled (Kim, 2024; Lee e al., 2021;
Ru q is & Tsang, 2024).
Equally impo an p inciples o he sa e y assessmen o he pos -closu e phase a e
he demons abili y and he managemen o unce ain ies. In bo h aspec s digi al wins
can be o immense alue, as hei applica ion will help o demons a e o he egula o
and he gene al public he sa e y and sa is ac o y pe o mance o he disposal sys em.
(Vi ando e al., 2024). Fo he managemen o unce ain ies many ac o s will need o
be conside ed, bu one ele an aspec is he e olu ion o he clima e. As a change in
he en i onmen can igge a change in he edox and he e o e e en ion condi ions
in soil and sedimen , clima e change can ha e a subs an ial impac on he unce ain ies
o anspo o edox-sensi i e adionuclides as Se-79. To manage and minimize he
unce ain y emana ing om hese changes i is essen ial o b eak down global clima e
models o be applicable on egional o si e scale as has al eady been shown in he
ecen IAEA MODARIA p ojec (Lindbo g & Ikonen, 2020).
The p ecise es ima ion o po en ial adionuclide eleases om he nea - ield o a
adioac i e was e eposi o y is an impo an sou ce o unce ain ies in he pos -closu e
sa e y assessmen . Thei managemen equi es he de elopmen o models ha can
desc ibe he elease and anspo o adionuclides om was e packages h ough
enginee ed ba ie s. A amewo k o building hese models is he compa men
app oach, which uses a ne wo k o esis ances and capaci ies o simula e anspo
(C ossland e al., 2005). The s a e o each compa men is in luenced by local
p ocesses such as decay o adionuclides o p ogeny wi h po en ially di e en chemical
p ope ies and exchange p ocesses like ad ec i e and di usi e anspo . The
eposi o y sys em is di ided in o se e al sys em componen s, each cha ac e ised by
ime-dependen physical a iables ha can change o e ime. Examples o hese
p ocesses include hea anspo , wa e up ake, and chemical decomposi ion, which
can occu in he bu e zone. To c ea e a concep ual model o hese coupled
p ocesses, g aphical ep esen a ions such as In e ac ion Ma ices can be used, and
so wa e ools can help in c ea ing compa men models and conduc ing simula ions.
SKB used he so wa e ool Ecolego (simila o No malysa) o c ea ing compa men
models o hei sa e y case. Such concep ual app oaches and compa men al models
can o m he basis o u he high- ideli y models and digi al wins (Ås and, 2022).
Recen coupled eac i e anspo models ha e success ully in eg a ed all ele an
ma e ial and nea - ield p ocesses in HLW disposal cells con aining nuclea glass, s eel,
cemen /ben oni e, and hos ock (g ani e o clay), as epo ed in De Wind e al. (2024)
and (Mon eneg o e al., 2023). Addi ionally, hyd o-chemo-mechanical modelling has
been employed o s udy he e olu ion o he Cigéo eposi o y closu e sys ems, which
a e based on ben oni e-based sealing componen s su ounded by cemen i ious
EURAD-2 Deli e able 16.1 – S a e-o - he-a on high ideli y nume ical simula ions o s ongly coupled
p ocesses o eposi o y sys ems (Ini ial epo )
Dissemina ion le el: Public
Da e o issue o his epo : 29/09/2025 56
ma e ials (Idia , La iña, Cochepin, e al., 2020; La iña, 2023). Chemo-mechanical
couplings in massi e conc e e in as uc u e ha e been add essed in WP MAGIC
h ough mul iscale simula ions anging om nano o cell scale models (Dauzè es e
al., 2022). Sequen ial coupling o eac i e anspo and mechanical codes has been
used o model cemen i ious ma e ial damage due o ca bona ion (Socié e al., 2023).
Fo example, he hyd o-chemo-mechanical a ia ional phase ield ac u e app oach
can handle chemical eac ions, along wi h he esul ing ma e ial dissolu ion and/o
p ecipi a ion caused by hyd a ion o deg ada ion (such as ca bona ion) o ac u ed
cemen i ious ma e ials (Zhang, 2018). Po e-scale simula ions a e also employed o
in es iga e mic os uc u e e olu ion and es ima e he e ec i e mechanical pa ame e s
o he media (Shen, 2020).
IAEA (2020) s a es ha he design p ocess will con inue e en du ing ope a ion as
op imisa ion will be equi ed du ing his long pe iod, due o echnical ad ances, new
in o ma ion and consequen changes. An unambiguous design basis is he e o e
desi able ha can be used o b idge he gap be ween di e en disciplines and
s akeholde s. This design basis should be con inuously upda ed and ex ended, as
mo e and mo e in o ma ion becomes a ailable, and would se e as he base o he
op imisa ion. Sa e y assessmen s play an impo an ole in he design op imisa ion,
which commonly use complex modelling sys ems.
I is desi able o use he la es echnologies o in eg a e he a ious mo ing pa s o he
design and sa e y assessmen s in o one digi al sys em. Digi al wins could p o ide a
solu ion o his p oblem. Digi al win is a “digi al ep esen a ion o an ac i e unique
p oduc , ha comp ises i s selec ed cha ac e is ics, p ope ies, condi ions, and
beha io s by means o models, in o ma ion, and da a wi hin a single o e en ac oss
mul iple li e cycle phases” (Cha i, 2019). Li e a u e o digi al wins in adioac i e was e
disposal is limi ed compa ed o decommissioning ac i i ies (Koldi z e al., 2023). The
PLEIADES H2020 p ojec 3 (Szöke I., 2021) ocuses on digi alisa ion o
decommissioning ac i i ies, including some aspec s o on-si e was e managemen . An
impo an pa o hese digi al wins is he in eg a ion o BIM (Building In o ma ion
Modelling) wi h GIS. BIM-like sys ems in digi al wins could also be used o he
cons uc u e, ope a ion and closu e o geological eposi o ies. One o he majo
challenges iden i ied he e is he in eg a ion o BIM wi h geological en i onmen s
(Jacques, 2023).
The PREDIS p ojec (PRe-DISposal managemen o adioac i e was e) was ano he
o e on o digi al solu ions and da a science in adioac i e managemen . As a way o
op imisa ion o disposal, a digi al pla o m was de eloped o in o ming decisions
ela ed o was e p e-disposal phases u ilizing da a gene a ed by digi al wins o was e
packages including hei his o y o s a e and p edic ed u u e s a es (Jacques, 2023).
EURAD-2 Deli e able 16.1 – S a e-o - he-a on high ideli y nume ical simula ions o s ongly coupled
p ocesses o eposi o y sys ems (Ini ial epo )
Dissemina ion le el: Public
Da e o issue o his epo : 29/09/2025 57
Wi hin he PREDIS p ojec a was e package digi al win was de eloped, ele an o he
p edisposal managemen o adioac i e was e packages. The mixing ank concep was
used o he modelling while he geochemical eac ions we e modelled using GEMS.
Machine lea ning algo i hms we e ained on he physico-chemical model and we e
used o p edic he geochemical and mechanical e olu ion o he was e packages (Hu
& Dähn, 2024). The as calcula ion o he e olu ion o was e packages can allow o
pe o m op imisa ion asks bo h du ing p edisposal as well as du ing he inal disposal
o he was e a he deep geological eposi o y.
Digi al wins o he long- e m disposal a e cu en ly much less ad anced. Commonly,
p ocess-based models a e used o his pu pose. An in eg a ed amewo k has been
de eloped by Nakabayashi and Sugiyama (2018) using a p obabilis ic app oach ha
can assess long- e m sa e y based on dose cons ain s and di ec ly ansla e i in o he
op imisa ion o design. Hyb id physics- and AI-based (da a- ained) models a e also
epo edly gaining ac ion o he ansi ional phase be ween ope a ion and long- e m
disposal (Koldi z e al., 2023). Digi al wins o moni o ing sys ems a e being
de eloped, which inco po a e nuclea was e modules (Di Gio anni e al., 2023).
In eg a ion o moni o ing da a in o pos -closu e digi al wins is essen ial, an example
being he concen a ions o oxygen which due o he co osion o he canis e di ec ly
in luence he elease o adionuclides (Koldi z e al., 2023). Long- e m da a and
knowledge managemen he e o e mus be in he cen e o a en ion o digi al wins
o disposal.
E o s ha e been made o inco po a e machine lea ning o de elop su oga e models
and emula o s o he mo e e icien calcula ions o a ious physical p ocesses.
Jacques (2023) epo 1-4 o de s o magni ude speedup o use cases such as cemen
deg ada ion and u anium adso p ion on clay. The main challenge highligh ed he e is
ha he da a ha is used o he aining o he machine lea ning algo i hm may no be
consis en wi h he use ’s speci ic case.
Machine lea ning solu ions o op imisa ion and model-based design we e explo ed in
he EURAD p ojec ’s DONUT (De elopmen and imp o emen O NUme ical me hods
and Tools o modelling coupled p ocesses) wo k package as well (Cla e e al., 2024).
New nume ical me hods we e de eloped o he op imisa ion o p ocess-based
modelling, mainly o he chemical e olu ion o he sys em and adionuclide anspo
ela ed p ocesses as eac i e anspo modelling commonly poses he highes
compu a ional bu den due o i s complexi y.
The main challenges iden i ied o digi al wins and op imisa ion o he disposal a e
ela ed o da a unce ain y and eliabili y. As he ime and spa ial scales o adioac i e
was e disposal a e la ge and a e conside ed in a ying de ail, meaning ul in eg a ion
EURAD-2 Deli e able 16.1 – S a e-o - he-a on high ideli y nume ical simula ions o s ongly coupled
p ocesses o eposi o y sys ems (Ini ial epo )
Dissemina ion le el: Public
Da e o issue o his epo : 29/09/2025 64
COMSOL. (2024). COMSOL 6.3 - COMSOL Mul iphysics Re e ence Manual.
h ps://doc.comsol.com/6.3
Conil, N., Vi el, M., Plua, C., Vu, M. N., Seyedi, D., & A mand, G. (2020). In si u in es iga ion
o he THM beha io o he Callo o-Ox o dian clays one. Rock Mechanics and Rock
Enginee ing, 53, 2747–2769.
Con i, S., Mülle , S., & O iz, M. (2018). Da a-D i en P oblems in Elas ici y. A chi e o Ra ional
Mechanics and Analysis, 229(1), 79–123. h ps://doi.o g/10.1007/s00205-017-1214-0
C ossland, I., Pinedo, P., Kessle , J., To es-Vidal, C., & Wal e s, B. (2005). “Re e ence
Biosphe es” o solid adioac i e was e disposal: he BIOMASS Me hodology. Jou nal
o en i onmen al adioac i i y, 84(2), 135–149.
Dagnelie, R. V., Wechne , S., Laba , M., Daumas, S., Le Milbeau, C., Le y, Y., & Lundy, M.
(2023). In si u di usion o o ganic compounds in And a's unde g ound labo a o y: a 4-
yea insigh om he'DRO'expe imen .
Dagnelie, R. V. H., A noux, P., Enaux, J., Radwan, J., Ne ie, P., Page, J., Coelho, D., &
Robine , J. C. (2017). Pe u ba ion induced by a ni a e plume on di usion o solu es in
a la ge-scale clay ock sample. Applied Clay Science, 141, 219–226.
h ps://doi.o g/10.1016/j.clay.2017.02.025
Dagnelie, R. V. H., A noux, P., Radwan, J., Lebeau, D., Ne ie, P., & Beaucai e, C. (2015).
Pe u ba ion induced by EDTA on HDO, B - and Eu di usion in a la ge-scale clay ock
sample. Applied Clay Science, 105, 142–149.
h ps://doi.o g/10.1016/j.clay.2014.12.004
Damiani, L. H., Kosakowski, G., Glaus, M. A., & Chu ako , S. V. (2020). A amewo k o
eac i e anspo modeling using FEniCS-Reak o o: go e ning equa ions and
benchma king esul s. Compu a ional Geosciences, 24(3), 1071–1085.
h ps://doi.o g/10.1007/s10596-019-09919-3
Dauzè es, A., Helson, O., Chu ako , S. V., Mon oya, V., Zghondi, J., Nee , E., Shao, J.,
Che kouk, A., Mijnendonckx, K., Sellie , A., Deissmann, G., A nold, T., Laca iè e, L.,
G i a, M., Vidal, T., Neji, M., Bou bon, X., Ib ahim, L., Seigneu , N.,…Vila asa, V.
(2022). Ini ial S a e o he A on he chemo-mechanical e olu ion o cemen i ious
ma e ials in disposal condi ions. . Final e sion as o 09/11/2022 o deli e able D16.1
o he HORIZON 2020 p ojec EURAD. EC G an ag eemen no: 847593.
De C aen, M., Van Gee , M., Hon y, M., Wee jens, E., & Sillen, X. (2008). Ex en o oxida ion
in Boom Clay as a esul o exca a ion and en ila ion o he HADES URF:
Expe imen al and modelling assessmen s. Physics and Chemis y o he Ea h, 33,
S350–S362. h ps://doi.o g/10.1016/j.pce.2008.10.032
De Lucia, M. (2024). DecT ee: a physics-based geochemical su oga e o su ace
complexa ion o u anium on clay. Ad ances in Geosciences, 65, 45–54.
De Lucia, M., Kempka, T., Ja nieks, J., & Kühn, M. (2017, Ap 23–28). In eg a ing su oga e
models in o subsu ace simula ion amewo k allows compu a ion o complex eac i e
anspo scena ios.Ene gy P ocedia [Eu opean geosciences union gene al assembly
2017, egu di ision ene gy, esou ces & en i onmen (e e)]. Gene al Assembly o
Eu opean-Geosciences-Union-Ene gy-Resou ces-and-En i onmen -Di ision, Vienna,
AUSTRIA.
De Lucia, M., & Kühn, M. (2021). DecT ee 1. 0–chemis y speedup in eac i e anspo
simula ions: pu ely da a-d i en and physics-based su oga es. Geoscien i ic Model
De elopmen , 14(7), 4713–4730.
De Lucia, M., Kühn, M., Lindemann, A., Lübke, M., & Schno , B. (2021). POET ( 0. 1): speedup
o many-co e pa allel eac i e anspo simula ions wi h as DHT lookups.
Geoscien i ic Model De elopmen , 14(12), 7391–7409.
De Wind , L., Lecle cq, S., & an de Lee, J. (2006). Assessing he du abili y o nuclea glass
wi h espec o silica con olling p ocesses in a clayey unde g ound disposal. MRS
Online P oceedings Lib a y (OPL), 932, 93.91.
De Wind , L., Sampe , J., Cochepin, B., Mon oya, V., Ga cia, E., Ga ibay-Rod igues, J., Mon,
A., Mon eneg o, L., Sampe , A., Raimbaul , L., & E., V. (2024). In eg a ed eac i e
EURAD-2 Deli e able 16.1 – S a e-o - he-a on high ideli y nume ical simula ions o s ongly coupled
p ocesses o eposi o y sys ems (Ini ial epo )
Dissemina ion le el: Public
Da e o issue o his epo : 29/09/2025 65
anspo models o assessing he chemical e olu ion a he disposal cell scale. Final
e sion as o 21-05-2024 o deli e able D2.17 o he HORIZON 2020 p ojec EURAD.
EC G an ag eemen no: 847593.
De Wind , L., Spyche , N. (2019). Reac i e anspo modeling: a key pe o mance assessmen
ool o he geologic disposal o nuclea was e, elemen s: In . Mag, Mine al. Geochem.
Pe ol. 15, 99.
Deissmann, G., Ai Mouheb, N., Ma in, C., Tu e o, M. J., To es, E., Ku s en, B., Wee jens,
E., N., A. M., C., M., Tu e o, M. J., To es, E., Ku s en, B., Wee jens, E., Jacques, D.,
Cue as, J., Sampe , J., Mon eneg o, L., Lei o, M., Some uo i, M., & Ca pen, L.
(2021). Expe imen s and nume ical model s udies on in e aces. Final e sion as o
12.05.2021 o deli e able D2.5 o he HORIZON 2020 p ojec EURAD, EC G an
ag eemen no: 847593.
Delchini, M. O. G., Swile , L. P., & Le eb e, R. A. (2021). Ex ension o he NEAMS wo kbench
o pa allel sensi i i y and unce ain y analysis o he mal hyd aulic pa ame e s using
Dako a and Nek5000. Nuclea Enginee ing and Technology, 53(10), 3449–3459.
h ps://doi.o g/10.1016/j.ne .2021.04.005
Demi e , E., Coene, E., I aola, A., Na di, A., Aba ca, E., Idia , A., de Paola, G., & Rod íguez-
Mo illas, N. (2023). Imp o ing he Pe o mance o Reac i e T anspo Simula ions
Using A i icial Neu al Ne wo ks. T anspo in Po ous Media, 149(1), 271–297.
h ps://doi.o g/10.1007/s11242-022-01856-7
Dep ulski, R. C., Dymi owska, M., & Kondo, D. (2022). Modelling non-local elas ici y in 1D
ib a ing ods using Co ec ed Smoo hed Pa icle Hyd odynamics me hod. Eu opean
Jou nal o Mechanics-A/Solids, 91, 104403.
Descos es, M., Poin eau, I., Radwan, J., Poonoosamy, J., Lacou , J.-L., Menu , D., Ve cou e ,
T., & Dagnelie, R. V. (2017). Adso p ion and e a ded di usion o EuIII-EDTA− h ough
ha d clay ock. Jou nal o Hyd ology, 544, 125–132.
De ille, E., Mon a nal, P., Lo h, L., & Cha an , C. (2009). Alliances: simula ion pla o m o
adioac i e was e disposal.
Di Gio anni, M., Campanile, L., D'Ono io, A., Ma one, S., Ma ulli, F., Romoli, M., Sabba ese,
C., & Ve de, L. (2023). Suppo ing he de elopmen o digi al wins in nuclea was e
moni o ing sys ems. P ocedia Compu e Science, 225, 3133–3142.
Dinç Gögüs, Ö., & Schol ès, L. (2018). Disc e e Analysis o Damage and Shea Banding in
A gillaceous Rocks. Rock Mechanics and Rock Enginee ing, 51. 10.1007/s00603-017-
1397-6.
Duchaine, F., Jau é, S., Poi ou, D., Quéme ais, E., S a elbach, G., Mo el, T., & Gicquel, L.
(2015). Analysis o high pe o mance conjuga e hea ans e wi h he OpenPALM
couple . Compu a ional Science & Disco e y, 8(1), 015003.
Du o, L., Al maie , M., Hol , E., Mäde , U., Cla e , F., G ambow, B., Idia , A., Valls, A., &
Mon oya, V. (2020). Con ibu ion o he esul s o he CEBAMA p ojec o dec ease
unce ain ies in he Sa e y Case and Pe o mance Assessmen o adioac i e was e
eposi o ies. Applied Geochemis y, 112, A icle 104479.
h ps://doi.o g/10.1016/j.apgeochem.2019.104479
Elmo sy, M., El‐Dakhakhni, W., & Zhao, B. (2023). Rapid pe meabili y upscaling o digi al
po ous media ia physics‐in o med neu al ne wo ks. Wa e Resou ces Resea ch,
59(12), e2023WR035064.
Elmo sy, M., El‐Dakhakhni, W., & Zhao, B. (2022). Gene alizable pe meabili y p edic ion o
digi al po ous media ia a no el mul i-scale 3D con olu ional neu al ne wo k. Wa e
Resou ces Resea ch, 58(3), e2021WR031454.Elodie C, Au élie T, Alaa C, Pie e B,
Guillaume H, Ma c L (2020) Senso s posi ion op imiza ion o moni o ing he
con e gence o adioac i e was e s o age unnel, Nuclea Enginee ing and Design 367.
h ps://doi.o g/doi:10.1016/j.nucengdes.2020.110778
Elodie, C., Au élie, T., Alaa, C., Pie e, B., Guillaume, H., & Ma c, L. (2020). Senso s posi ion
op imiza ion o moni o ing he con e gence o adioac i e was e s o age unnel.
Nuclea Enginee ing and Design, 367, A icle 110778.
h ps://doi.o g/10.1016/j.nucengdes.2020.110778
EURAD-2 Deli e able 16.1 – S a e-o - he-a on high ideli y nume ical simula ions o s ongly coupled
p ocesses o eposi o y sys ems (Ini ial epo )
Dissemina ion le el: Public
Da e o issue o his epo : 29/09/2025 66
E n, A., S ephansen, A. F., & Zunino, P. (2009). A discon inuous Gale kin me hod wi h
weigh ed a e ages o ad ec ion–di usion equa ions wi h locally small and aniso opic
di usi i y. IMA Jou nal o Nume ical Analysis, 29(2), 235–256.
EURAD Bu eau. (2023): Upda e o he EURAD S a egic Resea ch and Knowledge
Managemen Agenda (SRA), o deli e able D1.9 o he HORIZON 2020 p ojec
EURAD. EC G an ag eemen no: 847593.
Fins e le, S., Mulle , R. A., G imsich, J., Apps, J., & Bal ze , R. (2020). Pos -Closu e Sa e y
Calcula ions o he Disposal o Spen Nuclea Fuel in a Gene ic Ho izon al D illhole
Reposi o y. Ene gies, 13(10), A icle 2599. h ps://doi.o g/10.3390/en13102599
Fio e, S., Bakhouya, M., & Sma i, W. W. (2018). On he oad o exascale: Ad ances in high
pe o mance compu ing and simula ions - An o e iew and edi o ial. Fu u e Gene a ion
Compu e Sys ems- he In e na ional Jou nal o Escience, 82, 450–458.
h ps://doi.o g/10.1016/j. u u e.2018.01.034
Fio ina, C., Cli o d, I., Kelm, S., & Lo enzi, S. (2022). On he de elopmen o mul i-physics
ools o nuclea eac o analysis based on OpenFOAM®: s a e o he a , lessons
lea ned and pe spec i es. Nuclea Enginee ing and Design, 387, A icle 111604.
h ps://doi.o g/10.1016/j.nucengdes.2021.111604
F alo a, L., Le è e, G., Madé, B., Ma sac, R., Tho y, E., & Dagnelie, R. V. (2021). E ec o
o ganic compounds on he e en ion o adionuclides in clay ocks: Mechanisms and
speci ici ies o Eu (III), Th (IV), and U (VI). Applied Geochemis y, 127, 104859.
F an z, F. K. (1995, Dec 03–06). A axonomy o model abs ac ion echniques. [1995 win e
simula ion con e ence p oceedings]. 1995 Win e Simula ion Con e ence, A ling on,
Va.
F ugie , P., Mine , Y., Rajmohan, N., Godon, N., Gin, S. (2018). Modeling glass co osion wi h
GRAAL. Ma e , Deg ad. 2, 35.
Gao, W. (2018). A comp ehensi e e iew on iden i ica ion o he geoma e ial cons i u i e
model using he compu a ional in elligence me hod. Ad anced Enginee ing
In o ma ics, 38, 420–440. h ps://doi.o g/10.1016/j.aei.2018.08.021
Gauche , E. C., & Blanc, P. (2006). Cemen /clay in e ac ions - A e iew: Expe imen s, na u al
analogues, and modeling. Was e Managemen , 26(7), 776–788.
h ps://doi.o g/10.1016/j.wasman.2006.01.027
Gens, A., Guima äes, L. d. N., Oli ella, S., & Sánchez, M. (2010). Modelling he mo-hyd o-
mechano-chemical in e ac ions o nuclea was e disposal. Jou nal o Rock Mechanics
and Geo echnical Enginee ing, 2(2), 97–102.
Gobbe , M. K., Chu chill, A., Wang, G., & Seidman, T. I. (2009). COMSOL Mul iphysics o
e icien solu ion o a ansien eac ion-di usion sys em wi h as eac ion. P oceedings
o he COMSOL Con e ence,
Go ae s, J., Jacques, D., Sampe , J., Nee , E., Mon oya, V. (2022). Model abs ac ion
me hods o upscaling and in eg a ion o p ocess knowledge in eac i e anspo
models o geological disposal o adioac i e was e. Final e sion o 10/01/2022 o
deli e able D2.18 o he HORIZON 2020 p ojec EURAD, EC G an ag eemen 847593.
G aebling, N., Al haus, M., Şen, Ö. O., Reimann, T., Cajuhi, T., Scheue mann, G., Koldi z, O.,
& Rink, K. (2023). “Feels Like an Indie Game”—E alua ion o a Vi ual Field T ip
P o o ype on Radioac i e Was e Managemen Resea ch o Uni e si y Educa ion.
IEEE Compu e G aphics and Applica ions, 44(1), 13–24.
G a , W., F ei ag, S., Kaliske, M., & Sicke , J. U. (2010). Recu en Neu al Ne wo ks o
Unce ain Time-Dependen S uc u al Beha io . Compu e -Aided Ci il and
In as uc u e Enginee ing, 25(5), 322–333. h ps://doi.o g/10.1111/j.1467-
8667.2009.00645.x
G annan, A., Sood, K., No is, B., & Dubey, A. (2020). Unde s anding he landscape o
scien i ic so wa e used on high-pe o mance compu ing pla o ms. In e na ional
Jou nal o High Pe o mance Compu ing Applica ions, 34(4), 465–477, A icle
1094342019899451. h ps://doi.o g/10.1177/1094342019899451
EURAD-2 Deli e able 16.1 – S a e-o - he-a on high ideli y nume ical simula ions o s ongly coupled
p ocesses o eposi o y sys ems (Ini ial epo )
Dissemina ion le el: Public
Da e o issue o his epo : 29/09/2025 67
G aupne , B., Tha che , K., Newson, R. (2024). DECOVALEX-2023 Tack C Final Repo .
G aupne , B. J., Tha che , K., F iedenbe g, L., Guo, R., Hadgu, T., Hou, W., Kaise , S., Kim,
T., Pa k, C.-H., & Lee, C. (2025). An in e na ional s udy on THM modelling o he ull-
scale hea e expe imen a Mon Te i labo a o y. Geomechanics o Ene gy and he
En i onmen , 41, 100631.
Guima aes, L. D., Gens, A., Sanchez, M., & Oli ella, S. (2009, No 23–25). Coupled THMC
modeling o unsa u a ed swelling clays: Cons i u i e o mula ion and bounda y alue
p oblems. [Unsa u a ed soils: Theo e ical and nume ical ad ances in unsa u a ed soil
mechanics]. 4 h Asia Paci ic Con e ence on Unsa u a ed Soils, Newcas le,
AUSTRALIA.
Guo, J., Chen, C. P., Liu, Z., & Yang, X. (2024). Dynamic neu al ne wo k s uc u e: A e iew
o i s heo ies and applica ions. IEEE T ansac ions on Neu al Ne wo ks and Lea ning
Sys ems.
Guo, R. P., Tha che , K. E., Seyedi, D. M., & Plua, C. (2020). Calib a ion o he he mo-hyd o-
mechanical pa ame e s o he Callo o-Ox o dian clays one and he modelling o he
ALC expe imen . In e na ional Jou nal o Rock Mechanics and Mining Sciences, 132,
A icle 104351. h ps://doi.o g/10.1016/j.ij mms.2020.104351
Han, Y., Huang, G., Song, S., Yang, L., Wang, H., Wang Y. (2022) Dynamic Neu al Ne wo ks:
A Su ey. IEEE T ansac ions on Pa e n Analysis & Machine In elligence, ol. 44, no,
11, pp. 7436-7456, No . 2022. h ps://doi.o g/doi:10.1109/TPAMI.2021.3117837.
Hansen, N. K. S. (2004). E alua ing he CMA E olu ion S a egy on Mul imodal Tes Func ions,
in: Yao, X., Bu ke, E.K., Lozano, J.A., Smi h, J., Me elo-Gue ós, J.J., Bullina ia, J.A.,
Rowe, J.E., Tino, P., Kabán, A., Schwe el, H.-P. (Eds.), Pa allel P oblem Sol ing om
Na u e - PPSN VIII, Sp inge , Be lin, Heidelbe g, pp. 282-291.
h ps://doi.o g/10.1007/978-3-540-30217-9_29.
Hapla, V., Ho ak, D., Pospisil, L., Ce mak, M., Vasa o a, A., & Sojka, R. (2015). Sol ing
con ac mechanics p oblems wi h PERMON. In e na ional Con e ence on High
Pe o mance Compu ing in Science and Enginee ing,
Heinze, J., Meisel, T., Zill, F., Buchwald, J., Jäschke, M., Kiszku no, F.K., Kiszku no, F., Bilke,
L. . (2025). OGSTools. In h ps://zenodo.o g/ eco ds/14731492
Hennig, T., & Kühn, M. (2021). Su oga e Model o Mul i-Componen Di usion o U anium
h ough Opalinus Clay on he Hos Rock Scale. Applied Sciences-Basel, 11(2), A icle
786. h ps://doi.o g/10.3390/app11020786
He be , A. W., Jackson, C. P., & Le e , D. A. (1988). Coupled g oundwa e - low and solu e
anspo wi h luid densi y s ongly dependen upon concen a ion. Wa e Resou ces
Resea ch, 24(10), 1781–1795. h ps://doi.o g/10.1029/WR024i010p01781
Hes ha en, J. S., & Ubbiali, S. (2018). Non-in usi e educed o de modeling o nonlinea
p oblems using neu al ne wo ks. Jou nal o Compu a ional Physics, 363, 55–78.
Hiems a, T., & Van Riemsdijk, W. H. (1996). A su ace s uc u al app oach o ion adso p ion:
he cha ge dis ibu ion (CD) model. Jou nal o colloid and in e ace science, 179(2),
488–508.
Homma, T. S. A. (1996). Impo ance measu es in global sensi i i y analysis o nonlinea
models. Reliab, Eng. Sys . Sa . 52 (1), 1-17. h ps://doi.o g/10.1016/ 0951-
8320(96)00002-6.
Hu, G., & P ings en, W. (2023). Da a-d i en machine lea ning o disposal o high-le el nuclea
was e: A e iew. Annals o Nuclea Ene gy, 180, 109452,
h ps://doi.o g/10.1016/j.anucene.2022.109452.
Hu, G., P asianakis, N., Chu ako , S., & P ings en, W. (2024). Pe o mance analysis o da a-
d i en and physics-in o med machine lea ning me hods o he mal-hyd aulic
p ocesses in Full-scale Emplacemen expe imen . Applied The mal Enginee ing, 245,
A icle 122836. h ps://doi.o g/10.1016/j.appl he maleng.2024.122836
Hu, G., Schoenball, M., & P ings en, W. (2023). Machine lea ning-assis ed hea anspo
modelling o ull-scale emplacemen expe imen a Mon Te i unde g ound labo a o y.
In e na ional Jou nal o Hea and Mass T ans e , 213, A icle 124290.
h ps://doi.o g/10.1016/j.ijhea mass ans e .2023.124290
EURAD-2 Deli e able 16.1 – S a e-o - he-a on high ideli y nume ical simula ions o s ongly coupled
p ocesses o eposi o y sys ems (Ini ial epo )
Dissemina ion le el: Public
Da e o issue o his epo : 29/09/2025 68
Hu, G., Mi on, G. D., P ings en W., & Dähn, R. (2024). Digi al win and su oga e model o
long- e m geochemical p ocesses in nuclea was e managemen . In In e na ional
Con e ence on Nuclea Enginee ing (Vol, 88285, p. V008T09A018). Ame ican Socie y
o Mechanical Enginee s. h ps://doi.o g/10.1115/ICONE31-135796.
Huang, Y., & Shao, J.-F. (2013). A mic omechanical analysis o ime-dependen beha io
based on subc i ical damage in clays ones. In e na ional Jou nal o Damage
Mechanics, 22(6), 773–790.
Huang, Y. H., Shao, H. B., Wieland, E., Koldi z, O., & Kosakowski, G. (2018). A new app oach
o coupled wo-phase eac i e anspo simula ion o long- e m deg ada ion o
conc e e. Cons uc ion and Building Ma e ials, 190, 805–829.
h ps://doi.o g/10.1016/j.conbuildma .2018.09.114
Huang, Y. H., Shao, H. B., Wieland, E., Koldi z, O., & Kosakowski, G. (2021). Two-phase
anspo in a cemen ed was e package conside ing spa io- empo al e olu ion o
chemical condi ions. npj Ma e ials Deg ada ion, 5(1), A icle 4.
h ps://doi.o g/10.1038/s41529-021-00150-z
IAEA. (2009). Classi ica ion o Radioac i e Was e. IAEA Sa e y S anda ds Se ies No. GSG
‐
1,
IAEA, Vienna
IAEA. (2020). Design P inciples and App oaches o Radioac i e Was e Reposi o ies. IAEA
Nuclea Ene gy Se ies No. NW-T-1.27, 75 pp.
Idia , A., Bisschop, J., Caballe o, A., & Lu a, P. (2012). A nume ical and expe imen al s udy
o agg ega e-induced sh inkage c acking in cemen i ious composi es. Cemen and
Conc e e Resea ch, 42(2), 272–281. h ps://doi.o g/10.1016/j.cemcon es.2011.09.013
Idia , A., La iña, M., Cochepin, B., & Pas eau, A. (2020). Hyd o-chemo-mechanical modelling
o long- e m e olu ion o ben oni e swelling. Applied Clay Science, 195, A icle 105717.
h ps://doi.o g/10.1016/j.clay.2020.105717
Idia , A., La iña, M., Kosakowski, G., Cochepin, B., Meeussen, J. C. L., Sampe , J., Mon, A.,
Mon oya, V., Munie , I., Poonoosamy, J., Mon eneg o, L., Deissmann, G., Rohmen, S.,
Damiani, L. H., Coene, E., & Na es, A. (2020). Reac i e anspo modelling o a low-
pH conc e e / clay in e ace. Applied Geochemis y, 115, A icle 104562.
h ps://doi.o g/10.1016/j.apgeochem.2020.104562
Idia , A. S., B. (2019). Modelling o conc e e deg ada ion - Hyd ochemical p ocesses. Repo
o he sa e y e alua ion SE-SFL. SKB, Repo R-19-11.
Ikonen, K. (2009). The mal Dimensioning o Spen Fuel Reposi o y. . POSIVA Wo king Repo
2009-69, Finland.
Imbeaux, F., Pinches, S., Lis e , J., Bu a and, Y., Caspe , T., Du al, B., Guille mine , B.,
Hosokawa, M., Houlbe g, W., & Huynh, P. (2015). Design and i s applica ions o he
ITER in eg a ed modelling & analysis sui e. Nuclea Fusion, 55(12), 123006.
Jacques, D., Koldi z O., Szőke I., Chu ako S., Ga cía D. , Laloy E-, Mon oya V., P asianakis
N., Sampe J. (2023). Digi aliza ion and Digi al Twins in Long Te m Managemen o
Radioac i e Was e. . ASME 2023 In e na ional Con e ence on En i onmen al
Remedia ion and Radioac i e Was e Managemen ., S u ga , Ge many.
Jacques, D., Nee , E., & Deissmann, G. (2024). Upda ed S a e o he A on he assessmen
o he chemical e olu ion o ILW and HLW disposal cells. Final e sion as o 15.05.2024
o deli e able D2.2 o he HORIZON 2020 p ojec EURAD. EC G an ag eemen no:
847593.
Ja nieks, J., De Lucia, M., D ansch, D., & Sips, M. (2016, Ap 17–22). Da a-d i en su oga e
model app oach o imp o ing he pe o mance o eac i e anspo
simula ions.Ene gy P ocedia [Eu opean geosciences union gene al assembly 2016].
Gene al Assembly o he Eu opean-Geosciences-Union, Vienna, AUSTRIA.
Jenni, A., Meeussen, J. C., Pakkanen, T. A., Hi i, J. T., Akinwunmi, B., Yus es, Á., Na a o,
V., López-Vizcaíno, R., Muu i, E., & Niskanen, M. (2021). Coupling o chemical and
hyd omechanical p ope ies in ben oni e: a new eac i e anspo model. Applied Clay
Science, 214, 106274.
EURAD-2 Deli e able 16.1 – S a e-o - he-a on high ideli y nume ical simula ions o s ongly coupled
p ocesses o eposi o y sys ems (Ini ial epo )
Dissemina ion le el: Public
Da e o issue o his epo : 29/09/2025 69
Jenni, A., We sin, P., Thoenen, T., Baeyens, B., Fe a i, A., Gimmi, T., Mäde , U., Ma shall,
P., Hummel, W., & Leupin, O. (2019). Ben oni e back ill pe o mance in a high-le el
was e eposi o y: a geochemical pe spec i e.Technical Repo 19-03, NAGRA.
Jia, Y., Bian, H., Su, K., Kondo, D., & Shao, J.-F. (2010). Elas oplas ic damage modeling o
desa u a ion and esa u a ion in a gilli es. In e na ional Jou nal o Nume ical and
Analy ical Me hods in Geomechanics, 34(2), 187–220.
Jiang, F., Guo, Y., Tsuji, T., Ka o, Y., Shimokawa a, M., Es eban, L., Seyyedi, M., Pe ukhina,
M., Lebede , M., & Ki amu a, R. (2023). Upscaling pe meabili y using mul iscale X‐ ay‐
CT images wi h digi al ock modeling and deep lea ning echniques. Wa e Resou ces
Resea ch, 59(3), e2022WR033267.
Jiang, J., & Younis, R. M. (2016). Hyb id coupled disc e e- ac u e/ma ix and mul icon inuum
models o uncon en ional- ese oi simula ion. SPE Jou nal, 21(03), 1009–1027.
Jiang, P., Zhou, Q., & Shao, X. (2020). Su oga e Model-Based Enginee ing Design and
Op imiza ion. Sp inge Singapo e. h ps://doi.o g/10.1007/978-981-15-0731-1
Jing, L., Tsang, C. F., & S ephansson, O. (1995). DECOVALEX-An in e na ional co-ope a i e
esea ch p ojec on ma hema ical models o coupled THM p ocesses o sa e y
analysis o adioac i e was e eposi o ies [A icle]. In e na ional Jou nal o Rock
Mechanics and Mining Sciences, 32(5), 389–398.
h ps://doi.o g/h ps://doi.o g/10.1016/0148-9062(95)00031-B
Kam a a, S., Tahmasebi, P., & Sahimi, M. (2020). Linking mo phology o po ous media o hei
mac oscopic pe meabili y by deep lea ning. T anspo in Po ous Media, 131(2), 427–
448.
Kiczka, M., Al -Epping, P., Hadi, J., We sin, P., Leupin, O.X., Diamidis, N. (2024). Reac i e
anspo modelling o i on ben oni e in e ac ion in he FEBEX in si u expe imen .
Applied Geochemis y. Applied Geochemis y 170.
h ps://doi.o g/10.1016/j.apgeochem.2024.106057
Kiczka, M., Pekala, M., Maanoja, S., Muu i, E., & We sin, P. (2021). Modelling o solu e
anspo and mic obial ac i i y in di usion cells simula ing a ben oni e ba ie o a spen
nuclea uel eposi o y. Applied Clay Science, 211, A icle 106193.
h ps://doi.o g/10.1016/j.clay.2021.106193
Kim, K.-I., Lee, C., Cho, D., Ru q is , J. (2024). Enhancemen o disposal e iciency o deep
geological eposi o ies based on h ee design ac o s − Decay hea op imiza ion,
inc eased he mal limi o he bu e and double-laye concep . Tunnelling and
Unde g ound Space Technology, 153(106017).
h ps://doi.o g/h ps://doi.o g/10.1016/j. us .2024.106017
King, F., & Kolá , M. (2019). Li e ime P edic ions o Nuclea Was e Disposal Con aine s.
Co osion, 75(3), 309–323. h ps://doi.o g/10.5006/2994
Koldi z, O., Gö ke, U. J., Shao, H., & Wang, W. (2012). The mo-hyd o-mechanical-chemical
p ocesses in po ous media: Benchma ks and examples. Lec u e No es in
Compu a ional Science and Enginee ing 86, Sp inge , Be lin, Heidelbe g, 86, 391 pp.
Koldi z, O., Jacques, D., Cla e , F., Be and, J., Chu ako , S. V., Debayle, C., Diaconu, D.,
Fuzik, K., Ga cia, D., G aebling, N., G ambow, B., Hol , E., Idia , A., Lei a, P., Mon oya,
V., Niede lei hinge , E., Olin, M., P ings en, W., P asianakis, N. I.,…Wendling, J.
(2023). Digi alisa ion o nuclea was e managemen : p edisposal and disposal.
En i onmen al Ea h Sciences, 82(1), 42. h ps://doi.o g/10.1007/s12665-022-10675-4
Koldi z, O., Nagel, T., Shao, H., Wang, W., & Baue , S. (2018). The mo-hyd o-mechanical-
chemical p ocesses in ac u ed po ous media: Modelling and benchma king. F om
benchma king o u o ing. . Te es ial En i onmen al Sciences, Sp inge Na u e,
Cham, XXIII, 301 pp.
Konikow, L. F., San o d, W. E., & Campbell, P. J. (1997). Cons an -concen a ion bounda y
condi ion: Lessons om he HYDROCOIN a iable-densi y g oundwa e benchma k
p oblem. Wa e Resou ces Resea ch, 33(10), 2253–2261.
h ps://doi.o g/10.1029/97w 01926
EURAD-2 Deli e able 16.1 – S a e-o - he-a on high ideli y nume ical simula ions o s ongly coupled
p ocesses o eposi o y sys ems (Ini ial epo )
Dissemina ion le el: Public
Da e o issue o his epo : 29/09/2025 70
Kosakowski, G., Huang, Y. and Wieland, E. (2020). In luence o ma e ial he e ogenei ies,
p ocess couplings and agg ega e eac i i y on he geochemical e olu ion o he L/ILW
eposi o y. (Nag a A bei sbe ich e Issue.
K ause, M. J., Kumme lände , A., A is, S. J., Kusumaa maja, H., Dapelo, D., Klemens, F.,
Gaed ke, M., Ha en, N., Mink, A., & T unk, R. (2021). OpenLB—Open sou ce la ice
Bol zmann code. Compu e s & Ma hema ics wi h Applica ions, 81, 258–288.
LaFo ce, T. (2024). DECOVALEX-2023 Task F2 Final Repo .
Laloy, E., & Jacques, D. (2019). Emula ion o CPU-demanding eac i e anspo models: a
compa ison o Gaussian p ocesses, polynomial chaos expansion, and deep neu al
ne wo ks. Compu a ional Geosciences, 23(5), 1193–1215.
h ps://doi.o g/10.1007/s10596-019-09875-y
Laloy, E., & Jacques, D. (2022). Speeding Up Reac i e T anspo Simula ions in Cemen
Sys ems by Su oga e Geochemical Modeling: Deep Neu al Ne wo ks and <i>k</i>-
Nea es Neighbo s. T anspo in Po ous Media, 143(2), 433–462.
h ps://doi.o g/10.1007/s11242-022-01779-3
Laloy, E., & V ug , J. A. (2012). High‐dimensional pos e io explo a ion o hyd ologic models
using mul iple‐ y DREAM (ZS) and high‐pe o mance compu ing. Wa e Resou ces
Resea ch, 48(1).
La ion, Y., Chen, G., Zlo nik, S., Díez, P., See ha am, S., & Massa , T. J. (2022). Using
educed basis app oxima ion o e icien su oga e-based in e se iden i ica ion o
he mo-hyd o-mechanical pa ame e s om an in si u hea ing es . Rock Mechanics and
Rock Enginee ing, 55(9), 5739–5757.
La sson, A. (1992). The in e na ional p ojec s INTRACOIN, HYDROCOIN and INTRAVAL.
Ad ances in wa e esou ces, 15(1), 85–87. h ps://doi.o g/10.1016/0309-
1708(92)90034-y
La , J., Malaspinas, O., Kon axakis D., Pa migiani A., Lag a a D., B ogi F., & Chopa d, B.
(2021). Palabos: pa allel la ice Bol zmann sol e . Compu e s & Ma hema ics wi h
Applica ions, 81, 334-350, h ps://doi.o g/10.1016/j.camwa.2020.03.022.
La iña, M., Peleg í, J., Idia , A., Pas eau, A., Michau, N., Talandie , J., & Cochepin, B. (2023).
Long- e m E olu ion o Ben oni e-Based Seals o Closu e o a Radioac i e Was e
Reposi o y in Clays one: A Hyd o-Chemo-Mechanical Modelling Assessmen .
T anspo in Po ous Media. h ps://doi.o g/10.1007/s11242-023-01989-3
Leal, A. M. M., Kyas, S., Kulik, D. A., & Saa , M. O. (2020). Accele a ing Reac i e T anspo
Modeling: On-Demand Machine Lea ning Algo i hm o Chemical Equilib ium
Calcula ions. T anspo in Po ous Media, 133(2), 161–204.
h ps://doi.o g/10.1007/s11242-020-01412-1
Lee, C., Lee, J., & Kim, G. Y. (2021). Nume ical analysis o coupled hyd o-mechanical and
he mo-hyd o-mechanical beha iou in bu e ma e ials a a geological eposi o y o
nuclea was e: Simula ion o EB expe imen a Mon Te i URL and FEBEX a G imsel
es si e using Ba celona basic model. In e na ional Jou nal o Rock Mechanics and
Mining Sciences, 139, A icle 104663. h ps://doi.o g/10.1016/j.ij mms.2021.104663
Lehmann, C., Bilke, L., Buchwald, J., G aebling, N., G unwald, N., Heinze, J., Meisel, T., Lu,
R., Naumo , D., Rink, K., Sen, O. Ö., Selze , P., Shao, H., Wang, W., Zill, F., Nagel,
T., & Koldi z, O. (2023). OpenWo kFlow - De elopmen o an open-sou ce syn hesis-
pla o m o sa e y in es iga ions in he si e selec ion p ocess. G undwasse , submi ed.
Lemmen, C., Ho meis e , R., Klingbeil, K., Nase moaddeli, M. H., Ke imoglu, O., Bu cha d, H.,
Kös e s, F., & Wi z, K. W. (2018). Modula Sys em o Shel es and Coas s (MOSSCO
1. 0)–a lexible and mul i-componen amewo k o coupled coas al ocean ecosys em
modelling. Geoscien i ic Model De elopmen , 11(3), 915–935.
Lemmens, K., Debu e, M., Fe and, K., Goe hals, J., Liu, S., & Ma y, N. (2023). Resul s o
modelling and ecommenda ions o u u e expe imen al and nume ical wo k (HLW
disposal sys em). Final e sion as o 18.03.2024 o deli e able D2.14 o he HORIZON
2020 p ojec EURAD. EC G an ag eemen no: 847593.
Leupin, O. X., Smi h, P., Ma schall, P., Johnson, L., Sa age, D., Cloe , V., Schneide , J., &
Senge , R. (2016a). High-le el was e eposi o y-induced e ec s. NTB-14-13, 137pp.
EURAD-2 Deli e able 16.1 – S a e-o - he-a on high ideli y nume ical simula ions o s ongly coupled
p ocesses o eposi o y sys ems (Ini ial epo )
Dissemina ion le el: Public
Da e o issue o his epo : 29/09/2025 71
Leupin, O. X., Smi h, P., Ma schall, P., Johnson, L., Sa age, D., Cloe , V., Schneide , J., &
Senge , R. (2016b). Low- and in e media e-le el was e eposi o y-induced e ec s.
NTB-14-14, 109pp.
Le asseu , S., Collin, F., Dymi owska, M., Ha ing on, J., Jacops, E., Koldi z, O., Ma schall,
P., No is, S., Sillen, X., Talandie , J., T uche, L., & Wendling, J. (2024). S a e o he
A on Gas T anspo in Clayey Ma e ials – Upda e 2023. Deli e able D6.2 o he
HORIZON 2020 p ojec EURAD, Wo k Package Gas. EC G an ag eemen no: 847593.
Li, Z., Ko achki, N., Azizzadenesheli, K., Liu, B., Bha acha ya, K., S ua , A., & Anandkuma ,
A. (2020). Fou ie neu al ope a o o pa ame ic pa ial di e en ial equa ions. a Xi
p ep in a Xi :2010.08895.
Lindbo g, T., & Ikonen, A. T. (2020). De elopmen o a common amewo k o add essing
clima e and en i onmen al change in pos -closu e adiological assessmen s o
adioac i e was e–a new epo om IAEA (TecDoc 1904). Jou nal o Radiological
P o ec ion, 40(4), E31.
Liu, M., Kwon, B., & Kang, P. K. (2022). Machine lea ning o p edic e ec i e eac ion a es in
3D po ous media om po e s uc u al ea u es. Scien i ic Repo s, 12(1), 5486.
Liu, Z., Wang, Y., Vaidya, S., Ruehle, F., Hal e son, J., Soljačić, M., Hou, T. Y., & Tegma k,
M. (2024). Kan: Kolmogo o -a nold ne wo ks. a Xi p ep in a Xi :2404.19756.
Loga zo, H. J., Capuano, G., & Rimoli, J. J. (2021). Sma cons i u i e laws: Inelas ic
homogeniza ion h ough machine lea ning. Compu e Me hods in Applied Mechanics
and Enginee ing, 373, A icle 113482. h ps://doi.o g/10.1016/j.cma.2020.113482
Löna z, M. I., Yang, Y., Deissmann, G., Bosbach, D., & Poonoosamy, J. (2023). Cap u ing he
dynamic p ocesses o po osi y clogging. Wa e Resou ces Resea ch, 59(11),
e2023WR034722.
Lykkegaa d, M., Dodwell, T.J., Fox. C., Mingas G., Scheichl, R. (2023). Mul ile el Delayed
Accep ance MCMC. SIAM/ASA Jou nal on Unce ain y Quan i ica ion, 11(1), 1-30,
h ps://doi.o g/10.1137/22M1476770.
Maalal, O., P a , M., & Lasseux, D. (2021). Po e ne wo k model o d ying wi h Kel in e ec .
Physics o Fluids, 33(2).
Maes, J., & Menke, H. P. (2021). GeoChemFoam: Di ec modelling o mul iphase eac i e
anspo in eal po e geome ies wi h equilib ium eac ions. T anspo in Po ous Media,
139(2), 271–299.
Mah ous, M., Cu i, E., Chu ako , S. V., & P asianakis, N. I. (2022). Pe ophysical ini ializa ion
o co e-scale eac i e anspo simula ions on Indiana limes ones: Po e-scale
cha ac e iza ion, spa ial au oco ela ions, and ep esen a i e elemen a y olume
analysis. Jou nal o Pe oleum Science and Enginee ing, 213, 110389.
Mánica, M. A., Gens, A., Vauna , J., A mand, G., & Vu, M.-n. (2022). Nume ical simula ion o
unde g ound exca a ions in an indu a ed clay using non-local egula isa ion. Pa 2:
sensi i i y analysis. Géo echnique, 72(12), 1113–1128.
Ma ine , P., Leone, R., S ein, E. . (2024). DECOVALEX-2023 Task F1 Final Repo .
Ma ine , P.E., Be g, T. M.; Chang, K. W.; Debussche e, B. J.; Leone, R. C.; Seidl D. T. (2020).
Su oga e Model De elopmen o Spen Fuel Deg ada ion o Reposi o y Pe o mance
Assessmen . Spen Fuel and Was e Disposi ion, SAND2020-10797 R.
Ma y, N. C. M., Bilds ein, O., Blanc, P., Cla e , F., Cochepin, B., Gauche , E. C., Jacques, D.,
La igue, J. E., Liu, S. H., Maye , K. U., Meeussen, J. C. L., Munie , I., Poin eau, I., Su,
D. Y., & S ee el, C. I. (2015). Benchma ks o mul icomponen eac i e anspo ac oss
a cemen /clay in e ace. Compu a ional Geosciences, 19(3), 635–653.
h ps://doi.o g/10.1007/s10596-014-9463-6
Meeussen, J. C. (2003). ORCHESTRA: An objec -o ien ed amewo k o implemen ing
chemical equilib ium models. En i onmen al science & echnology, 37(6), 1175–1182.
Mehmani, Y., & Tchelepi, H. A. (2017). Minimum equi emen s o p edic i e po e-ne wo k
modeling o solu e anspo in mic omodels. Ad ances in wa e esou ces, 108, 83–
98.
EURAD-2 Deli e able 16.1 – S a e-o - he-a on high ideli y nume ical simula ions o s ongly coupled
p ocesses o eposi o y sys ems (Ini ial epo )
Dissemina ion le el: Public
Da e o issue o his epo : 29/09/2025 72
Meinde s, M. B., Yang, J., & Linden, E. . d. (2024). Applica ion o physics encoded neu al
ne wo ks o imp o e p edic abili y o p ope ies o complex mul i-scale sys ems.
Scien i ic Repo s, 14(1), 15015.
Menke, H. P., Maes, J., & Geige , S. (2021). Upscaling he po osi y–pe meabili y ela ionship
o a mic opo ous ca bona e o Da cy-scale low wi h machine lea ning. Scien i ic
Repo s, 11(1), 2625.
Mi chell, R., Menzenbach, R., Ke s ing, K., & Mund , M. (2023). Sel -expanding neu al
ne wo ks. a Xi p ep in a Xi :2307.04526.
Mokos, A., Sa o, Y., Niceno, B., Chu ako , S.V., P asianakis, N.I. (2024). Nuclea e boiling
wi hin a uel assembly a ec ed by CRUD. EPJ Web o Con e ences, 302, 03003,
h ps://doi.o g/10.1051/epjcon /202430203003.
Molins, S., & Knabne , P. (2019). Mul iscale App oaches in Reac i e T anspo Modeling. In J.
L. D uhan & C. Tou nassa (Eds.), Reac i e T anspo in Na u al and Enginee ed
Sys ems (Vol. 85, pp. 27–48). h ps://doi.o g/10.2138/ mg.2019.85.2
Molins, S., T ebo ich, D., Mille , G. H., & S ee el, C. I. (2017). Mine alogical and anspo
con ols on he e olu ion o po ous media ex u e using di ec nume ical simula ion.
Wa e Resou ces Resea ch, 53(5), 3645–3661. h ps://doi.o g/10.1002/2016w 020323
Mollaali, M., Koldi z, O., Hu, M., Pa k, C. H., Pa k, J. W., McDe mo , C. I., Chi enden, N.,
Bond, A., Yoon, J. S., & Zhou, J. (2023). Compa a i e e i ica ion o hyd o-mechanical
ac u e beha io : Task G o in e na ional esea ch p ojec DECOVALEX–2023.
In e na ional Jou nal o Rock Mechanics and Mining Sciences, 170, 105530.
Mon, A., Sampe , J., Mon eneg o, L., Tu e o, M. J., To es, E., Cue as, J., Fe nández, R., &
De Wind , L. (2023). Reac i e anspo models o he geochemical in e ac ions a he
i on/ ben oni e in e ace in labo a o y co osion es s. Applied Clay Science, 240, A icle
106981. h ps://doi.o g/10.1016/j.clay.2023.106981
Mon eneg o, L., Sampe , J., Mon, A., De Wind , L., Sampe , A. C., & Ga cía, E. (2023). A non-
iso he mal eac i e anspo model o he long- e m geochemical e olu ion a he
disposal cell scale in a HLW eposi o y in g ani e. Applied Clay Science, 242, A icle
107018. h ps://doi.o g/10.1016/j.clay.2023.107018
Mo eno, L., Skagius, K., Soede g en, S., & Wibo gh, M. (2001). P ojec Sa e. Gas ela ed
p ocesses in SFR. Sweden. S ensk Kä nb änslehan e ing AB.
Mo is, M. D. (1991). Fac o ial sampling plans o p elimina y compu a ional expe imen s.
Technome ics 33161-33174, h ps://doi.o g/10.1080/00401706.1991.10484804.
Mos a a, A., Schol ès, L., & Gol ie , F. (2023). Po e-scale hyd o-mechanical modeling o gas
anspo in coal ma ix. Fuel, 345, 128165.
Nakabayashi, R., & Sugiyama, D. (2018). Me hodology o op imize adia ion p o ec ion in
adioac i e was e disposal a e closu e o a disposal acili y based on p obabilis ic
app oach. Jou nal o Nuclea Science and Technology, 55(3), 335–347.
Nasika, C., Díez, P., Ge a d, P., Massa , T. J., & Zlo nik, S. (2023). Disc e e empi ical
in e pola ion o hype ‐ educ ion o hyd o‐mechanical p oblems in g oundwa e low
h ough soil. In e na ional Jou nal o Nume ical and Analy ical Me hods in
Geomechanics, 47(5), 667–693.
Na a o Gami , V., & Asensio Sánchez, L. (2024). A mixed ini e elemen / ini e olume
o mula ion o g anula ben oni e mix u es.
Na a o Gami , V., Asensio Sánchez, L., Yus es Real, Á., Mo ena Bo ja, G. d. l., & Pin ado,
X. (2016). Swelling and mechanical e osion o MX-80 ben oni e.
Na a o, V., Asensio, L., Alonso, J., Yus es, A., & Pin ado, X. (2014). Mul iphysics
implemen a ion o ad anced soil mechanics models. Compu e s and Geo echnics, 60,
20–28.
Na a o, V., Asensio, L., Gha bieh, H., De la Mo ena, G., & Pulkkanen, V.-M. (2019).
De elopmen o a mul iphysics nume ical sol e o modeling he beha io o clay-
based enginee ed ba ie s. Nuclea Enginee ing and Technology, 51(4), 1047–1059.
Na a o, V., Pucci, A., Tengblad, E., Casini, F., & Asensio, L. (2023). Symbolic algeb a
in eg a ion o soil elas oplas ic models. Compu e s and Geo echnics, 164, 105834.
EURAD-2 Deli e able 16.1 – S a e-o - he-a on high ideli y nume ical simula ions o s ongly coupled
p ocesses o eposi o y sys ems (Ini ial epo )
Dissemina ion le el: Public
Da e o issue o his epo : 29/09/2025 73
Nee , E. J., D.; Deissmann, G. (2022). Ini ial S a e o he A on he assessmen o he chemical
e olu ion o ILW and HLW disposal cells. D 2.1 o he HORIZON 2020 p ojec EURAD,
EC G an ag eemen no: 847593.
Neuman, S. P. (2008). Mul iscale ela ionships be ween ac u e leng h, ape u e, densi y and
pe meabili y. Geophysical Resea ch Le e s, 35(22).
Nguyen, T. S., Bo gesson, L., Chijima su, M., He nelind, J., Jing, L. R., Kobayashi, A., &
Ru q is , J. (2009). A case s udy on he in luence o THM coupling on he nea ield
sa e y o a spen uel eposi o y in spa sely ac u ed g ani e. En i onmen al Geology,
57(6), 1239–1254. h ps://doi.o g/10.1007/s00254-008-1565-9
Niu, Y., Mos aghimi, P., Shabaninejad, M., Swie ojanski, P., & A ms ong, R. T. (2020). Digi al
ock segmen a ion o pe ophysical analysis wi h educed use bias using
con olu ional neu al ne wo ks. Wa e Resou ces Resea ch, 56(2), e2019WR026597.
Nogh e ab, B., Damians, I., Oli ella, S., & Gens, A. (2024). Coupled hyd o-gas-mechanical 3D
modeling o LASGIT expe imen . Geomechanics o Ene gy and he En i onmen , 40,
100623.
Noi iel, C., & Soulaine, C. (2021). Po e-Scale Imaging and Modelling o Reac i e Flow in
E ol ing Po ous Media: T acking he Dynamics o he Fluid-Rock In e ace. T anspo
in Po ous Media, 140(1), 181–213. h ps://doi.o g/10.1007/s11242-021-01613-2
O’Lea y-Rosebe y, T., Villa, U., Chen, P., & Gha as, O. (2022). De i a i e-in o med p ojec ed
neu al ne wo ks o high-dimensional pa ame ic maps go e ned by PDEs. Compu e
Me hods in Applied Mechanics and Enginee ing, 388, 114199.
Ojo an, M. I., Lee, W. E., & Kalmyko , S. N. (2019). An In oduc ion o Nuclea Was e
Immobilisa ion. (3 d ed. ed.). Else ie ; Ams e dam, The Ne he lands.
Oli ella, S., Vauna J., Rod iguez-Dono A. (2024). CODE_BRIGHT 2024 use 's guide.
Depa men o Ci el and En i onmen al Enginee ing, Technical Uni e si y o Ca alonia.
Ba celona, Spain.
Oommen, V., Shukla, K., Desai, S., Ding e ille, R., & Ka niadakis, G. E. (2024). Re hinking
ma e ials simula ions: Blending di ec nume ical simula ions wi h neu al ope a o s. npj
Compu a ional Ma e ials, 10(1), 145.
Oudiz, A., Locha d, J., Lomba d, J., C o , J., Fleishman, A., Webb, G. (1986). Wha is Ala a.
CEPN Cen e d'é ude su l'E alua ion de la P o ec ion dans le domaine Nucléai e.
Pachepsky, Y. Gube , A., an Genuch en, M.T., Nicholson, T.J., Cady, R.E., Simunek, J.,
Schaap, M.G. (2006). Model Abs ac ion Techniques o Soil-Wa e Flow and
T anspo . NUREG/CR-6884.
Pa doen, B., & Collin, F. (2017). Modelling he in luence o s ain localisa ion and iscosi y on
he beha iou o unde g ound d i s d illed in clays one. Compu e s and Geo echnics,
85, 351–367.
Pa k, H. D., Hammond, G. E., Valocchi, A. J., & LaFo ce, T. (2021). Linea and nonlinea
sol e s o simula ing mul iphase low wi hin la ge-scale enginee ed subsu ace
sys ems. Ad ances in wa e esou ces, 156, A icle 104029.
h ps://doi.o g/10.1016/j.ad wa es.2021.104029
Pa el, R. A., Pe ko, J., Jacques, D., De Schu e , G., Ye, G., & Van B eugel, K. (2018). A h ee-
dimensional la ice Bol zmann me hod based eac i e anspo model o simula e
changes in cemen pas e mic os uc u e due o calcium leaching. Cons uc ion and
Building Ma e ials, 166, 158–170. h ps://doi.o g/10.1016/j.conbuildma .2018.01.114
Pa el, R.A., Chu ako , S.V., P asianakis, N.I. (2021). A mul i-le el po e scale eac i e anspo
model o he in es iga ion o combined leaching and ca bona ion o cemen pas e.
Cemen and Conc e e Composi es, 115, 103831,
h ps://doi.o g/10.1016/j.cemconcomp.2020.103831.
Pazdniakou, A., & Dymi owska, M. (2018). Mig a ion o Gas in Wa e Sa u a ed Clays by
Coupled Hyd aulic‐Mechanical Model. Geo luids, 2018(1), 6873298.
Pazdniakou, A., Tine , A.-J., Gol ie , F., Kalo, K., Gabo eau, S., & Gai e, P. (2018). Nume ical
e iciency assessmen o he la ice Bol zmann model o digi al nano-po ous ock
applica ions. Ad ances in wa e esou ces, 121, 44–56.
EURAD-2 Deli e able 16.1 – S a e-o - he-a on high ideli y nume ical simula ions o s ongly coupled
p ocesses o eposi o y sys ems (Ini ial epo )
Dissemina ion le el: Public
Da e o issue o his epo : 29/09/2025 80
Xu, H., Zheng, L. G., Ru q is , J., & Bi kholze , J. (2021). Nume ical s udy o he chemo-
mechanical beha io o FEBEX ben oni e in nuclea was e disposal based on he
Ba celona expansi e model. Compu e s and Geo echnics, 132, A icle 103968.
h ps://doi.o g/10.1016/j.compgeo.2020.103968
Xu, T., Spyche , N., Sonnen hal, E., Zheng, L., & P uess, K. (2012). TOUGHREACT use ’s
guide: A simula ion p og am o non-iso he mal mul iphase eac i e anspo in
a iably sa u a ed geologic media, e sion 2.0. Ea h Sciences Di ision, Law ence
Be keley Na ional Labo a o y, Be keley, USA.
Yong, W. P., Menke, H., Maes, J., Geige , S., Baka , Z. A. A., Lewis, H., Buckman, J., Bonnin,
A., & Singh, K. (2024). Mul iscale Upscaling S udy o CO2 S o age in Ca bona e Rocks
Using Machine Lea ning and Mul iscale Imaging. O sho e Technology Con e ence
Asia,
Yu, Z., Shao, J.-F., Du eau, G., Wang, M., Vu, M.-n., & Plua, C. (2024). Nume ical modeling
o gas injec ion induced c acking wi h unsa u a ed hyd omechanical p ocess in he
con ex o adioac i e was e disposal. Tunnelling and Unde g ound Space Technology,
146, 105609.
Yu, Z., Shao, J.-F., & Vu, M.-N. (2024). A he mo-hyd omechanical damage model and i s
applica ion o a deep geological adioac i e eposi o y. Compu e s and Geo echnics,
170, 106306.
Yu, Z., Shao, J., Du eau, G., Vu, M.-N., & A mand, G. (2021). Nume ical modeling o
de o ma ion and damage a ound unde g ound exca a ion by phase- ield me hod wi h
hyd omechanical coupling. Compu e s and Geo echnics, 138, 104369.
Yus es, A., Jenni, A., Asensio, L., Pin ado, X., Koskinen, K., Na a o, V., & We sin, P. (2017).
Compa ison o he hyd ogeochemical and mechanical beha iou s o compac ed
ben oni e using di e en concep ual app oaches. Applied Clay Science, 141, 280–291.
Zhang, H. Z., Sa ija, B., Schlangen, E. (2018). Towa ds unde s anding s ochas ic ac u e
pe o mance o cemen pas e a mic o leng h scale based on nume ical simula ion.
Cons Build Ma e 183:189-201, h ps://doi.o g/10.1016/j.conbuildma .2018.06.167.
Zhang, N. L., Luo, Z. F., Chen, Z. X., Liu, F. S., Liu, P. L., Chen, W. Y., Wu, L., & Zhao, L. Q.
(2023). The mal-hyd aulic-mechanical-chemical coupled p ocesses and hei
nume ical simula ion: a comp ehensi e e iew. Ac a Geo echnica, 18(12), 6253–6274.
h ps://doi.o g/10.1007/s11440-023-01976-4
Zhao, J., Shen, W., Shao, J., Liu, Z., & Vu, M. (2022). A cons i u i e model o aniso opic clay-
ich ocks conside ing mic o-s uc u al composi ion. In e na ional Jou nal o Rock
Mechanics and Mining Sciences, 151, 105029.
Zheng, L., & Fe nández, A. M. (2023). P edic ion o long- e m geochemical change in ben oni e
based on he in e p e a i e THMC model o he FEBEX in si u es . Mine als, 13(12),
1522.
Zheng, L. E., Ru q is , J., Bi kholze , J. T., & Liu, H. H. (2015). On he impac o empe a u es
up o 200 deg ees C in clay eposi o ies wi h ben oni e enginee ba ie sys ems: A
s udy wi h coupled he mal, hyd ological, chemical, and mechanical modeling.
Enginee ing Geology, 197, 278–295. h ps://doi.o g/10.1016/j.enggeo.2015.08.026
Zie le, G., Cajuhi, T., Cos abel, S., Fu che, M., & Maßmann, J. (2024). Wa e con en e olu ion
in he EDZ o Opalinus Clay: A me hodic app oach o a compa a i e in e p e a ion o
measu emen s and modelling. Rock Mechanics and Rock Enginee ing, 57(6), 4443–
4472.
