Cbcilt: Sçäk Fs

aefJt^pv=pçÑíï~êÉ
cbcilt o
cáåáíÉ=bäÉãÉåí=pìÄëìêÑ~ÅÉ=cäçï
C=qê~åëéçêí=páãìä~íáçå=póëíÉã
tÜáíÉ=m~éÉêë
sçäK=fs
aefJt^pv=dãÄe
`çéóêáÖÜí=åçíáÅÉW
kç=é~êí=çÑ=íÜáë=ã~åì~ä=ã~ó=ÄÉ=éÜçíçÅçéáÉÇI=êÉéêçÇìÅÉÇI=çê=íê~åëä~íÉÇ=ïáíÜçìí=ïêáííÉå=éÉêãáëëáçå=çÑ=íÜÉ=
ÇÉîÉäçéÉê=~åÇ=ÇáëíêáÄìíçê=aefJt^pv=dãÄeK
`çéóêáÖÜí=EÅF=OMMRI=OMMSI=OMMTI=OMMV=aefJt^pv=dãÄe=_Éêäáå=J=~ää=êáÖÜíë=êÉëÉêîÉÇK=
aefJt^pv=~åÇ=cbcilt=~êÉ=êÉÖáëíÉêÉÇ=íê~ÇÉã~êâë=çÑ=aefJt^pv=dãÄeK
aefJt^pv=dãÄeI===
t~äíÉêëÇçêÑÉê=píê~≈É=NMRI==aJNOROS=_ÉêäáåI=dÉêã~åó
mÜçåÉW=HQVJPMJST==VV=VUJMI=c~ñW=HQVJPMJST=VV=VUJVV
bJj~áäW=ã~áä]ÇÜáJï~ëóKÇÉ
áá=ö=tÜáíÉ=m~éÉêë=J=sçäK=fs
`çåíÉåíë
`çåíÉåíë
NK=oÉ~ÅíáîÉ=ãìäíáJëéÉÅáÉë=íê~åëéçêíK=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K R
NKN= fåíêçÇìÅíáçå=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=KR äÉãë =K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K= PQ

NKO= qÜÉçêó =K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=KR NKU= bñ~ãéäÉë K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K= PR
NKOKN= oÉ~Åíáçå=âáåÉíáÅë K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=KR NKUKN= pÉêá~äJé~ê~ääÉä=êÉ~Åíáçåë=~åÇ=Åçãé~êáëçå=íç=íÜÉ=~å~äóíJ
NKOKNKN= póãÄçäáÅ=êÉ~Åíáçå=ëíçáÅÜáçãÉíêó K=K=K=K=K=K=K=K=K=K=K=K=K=K=KR áÅ~ä=ëçäìíáçå=Äó=pìå=Éí=~äK K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K= PR
NKOKNKO= bñ~ãéäÉë =K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=KS NKUKNKN= pí~íÉãÉåí=çÑ=íÜÉ=Na=éêçÄäÉã =K=K=K=K=K=K=K=K=K=K=K=K=K=K=K= PR
NKOKO= dçîÉêåáåÖ=íê~åëéçêí=Éèì~íáçåë=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=KT NKUKNKO= ^å~äóíáÅ~ä=ëçäìíáçå=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K= PS
NKOKOKN= _~ëáÅ=Ä~ä~åÅÉ=ëí~íÉãÉåíëK=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=KT NKUKNKP= kìãÉêáÅ~ä=~å~äóëáë K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K= PT
NKOKOKO= dÉåÉê~äáòÉÇ=Ä~ä~åÅÉ=Éèì~íáçåë=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=KU NKUKNKQ= oÉëìäíë K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K= PT
NKOKOKP= cáå~ä=Éèì~íáçåë=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=KV NKUKNKR= Pa=ãìäíáJëéÉÅáÉë=íê~åëéçêí=ïáíÜ=ÑáêëíJçêÇÉê=ÇÉÖê~Ç~J
NKOKP= oÉ~Åíáçå=ê~íÉë=oâK=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=KNN íáçå=êÉ~Åíáçå=âáåÉíáÅë=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K= PV
NKOKPKN= aÉÖê~Ç~íáçå=íóéÉ=âáåÉíáÅë K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=KNN NKUKO= o~íÉJäáãáíÉÇ=ÇÉëçêéíáçå=~åÇ=ÇÉÅ~óW=`çãé~êáëçå=íç=
NKOKPKO= ^êêÜÉåáìë=íóéÉ=âáåÉíáÅë =K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=KNO cêó=Éí=~äKÛë=~å~äóíáÅ~ä=ëçäìíáçå K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K= QO
NKOKPKP= jçåçÇ=íóéÉ=âáåÉíáÅë K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=KNO NKUKOKN= pí~íÉãÉåí=çÑ=íÜÉ=Na=éêçÄäÉã =K=K=K=K=K=K=K=K=K=K=K=K=K=K=K= QO
NKOKPKQ= m~ê~ääÉä=áêêÉîÉêëáÄäÉ=~åÇ=êÉîÉêëáÄäÉ=êÉ~Åíáçåë K=K=K=K=K=KNQ NKUKOKO= kìãÉêáÅ~ä=~å~äóëáë K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K= QP
NKP= kìãÉêáÅ~ä=pçäìíáçå K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=KNQ NKUKOKP= oÉëìäíë K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K= QQ
NKPKN= cáåáíÉJÉäÉãÉåí=Éèì~íáçå=ëóëíÉãK=K=K=K=K=K=K=K=K=K=K=K=K=K=KNQ NKUKP= qïçJëáíÉ=ÉèìáäáÄêáìãLâáåÉíáÅ=ëçêéíáçå=ïáíÜ=ÇÉÖê~Ç~J
NKPKO= ^Ç~éíáîÉ=éêÉÇáÅíçêJÅçêêÉÅíçê=íáãÉJëíÉééáåÖ=ëçäìíáçå= íáçåW=`çãé~êáëçå=íç=pq^kjla=~å~äóíáÅ~ä=ëçäìíáçåë =
ëíê~íÉÖóK=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=KNS =K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K= QQ
NKQ= oÉä~íÉÇ=cbcilt=aá~äçÖë K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=KOM NKUKPKN= mêçÄäÉã=Ñçêãìä~íáçå=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K= QQ
NKQKN= jìäíáJëéÉÅáÉë=éêçÄäÉã=Åä~ëë=ëéÉÅáÑáÅ~íáçåK=K=K=K=K=K=K=KOM NKUKPKO= pí~íÉãÉåí=çÑ=~=Na=íÉëí=Å~ëÉ K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K= QS
NKQKO= péÉÅáÉë=ëÉäÉÅíçêK=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=KOM NKUKPKP= kìãÉêáÅ~ä=~å~äóëáë K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K= QT
NKQKP= jìäíáJëéÉÅáÉë=ã~íÉêá~ä=Ç~í~=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=KON NKUKPKQ= oÉëìäíë K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K= QU
NKR= oÉ~Åíáçå=háåÉíáÅë=bÇáíçê K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=KON NKV= pÉèìÉåíá~ä=~åÇ=kçåJëÉèìÉåíá~ä=`Üäçêáå~íÉÇ=pçäîÉåíë=
NKRKN= mêÉÅçãéáäÉÇ=ê~íÉ=ÉñéêÉëëáçåë =K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=KOO aÉÖê~Ç~íáçå=ìåÇÉê=s~êá~ÄäÉ=^ÉêçÄáÅJ^å~ÉêçÄáÅ=`çåJ
NKRKNKN= aÉÖê~Ç~íáçå=íóéÉ=âáåÉíáÅë K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=KOO Çáíáçåë K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K= QV
NKRKNKO= ^êêÜÉåáìë=íóéÉ=âáåÉíáÅë =K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=KOP NKVKN= fåíêçÇìÅíáçå =K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K= QV
NKRKNKP= jçåçÇ=íóéÉ=âáåÉíáÅë K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=KOQ NKVKO= `çåÅÉéíì~ä=éêçÄäÉã=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K= RM
NKRKO= rëÉêJÇÉÑáåÉÇ=ê~íÉ=ÉñéêÉëëáçåëW=cbj^qeba=ÉÇáíçê =KOQ NKVKP= bèì~íáçåë=çÑ=ã~ëë=íê~åëéçêí=ïáíÜ=ëÉèìÉåíá~ä=~åÇ=åçåJ
NKS= j~ëë=j~íÉêá~ä=a~í~=çÑ=jìäíáJpéÉÅáÉë=qê~åëéçêí =K=K=KOT ëÉèìÉåíá~ä=êÉ~Åíáçå =K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K= RM
NKT= kçíÉ=íç=s~êá~ÄäÉJaÉåëáíó=cäçï=~åÇ=jìäíáJpéÉÅáÉë= NKVKQ= oÉ~Åíáçå=ê~íÉë =K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K= RN
qê~åëéçêí =K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=KPN NKVKR= kìãÉêáÅ~ä=ãçÇÉä K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K= RO
NKTKN= jìäíáJëéÉÅáÉë=ÇÉåëáíó=êÉä~íáçå =K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=KPN NKVKS= páãìä~íáçå=êÉëìäíë =K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K= RQ
NKTKO= léíáçå=ëÉííáåÖë K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=KPO NKVKT= ^ÅâåçïäÉÇÖÉãÉåíë K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K= RU
NKTKP= jìäíáéäÉ=ëéÉÅáÉëJÇÉéÉåÇÉåí=ÇÉåëáíó=ê~íáç =K=K=K=K=K=K=KPO NKNM= pìãã~êó=~åÇ=`çåÅäìëáçåëK=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K= RU
NKTKQ= `çåëí~åí=ëçäìí~ä=ÇÉåëáíó=ê~íáç =K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=KPP kçí~íáçå=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K= RV
NKTKR= sáëÅçëáíó=êÉä~íáçåë=Ñçê=ãìäíáJëéÉÅáÉë=íê~åëéçêí=éêçÄJ oÉÑÉêÉåÅÉë K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K= SN
cbcilt=ö=ááá
`çåíÉåíë
OK=m~ê~ãÉíÉê=Éëíáã~íáçå=çÑ=íê~åëáÉåí=Ñäçï=éêçÄäÉãë=ïáíÜ=mbpq=áå=cbcilt=K=K=K= SP
OKN= fåíêçÇìÅíáçå=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=KSP OKOKP= táÅÜáí~=éìãé=íÉëí K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K= TQ

OKO= qÉëí=bñ~ãéäÉë K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=KSR OKOKQ= cäççÇï~îÉ=éêçÄäÉã K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K= TU
OKOKN= qÜÉáëD=ïÉää=éêçÄäÉã K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=KSR OKP= `çåÅäìÇáåÖ=oÉã~êâë =K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K= UM
OKOKO= _êÉóÉää=éìãé=íÉëí =K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=KST oÉÑÉêÉåÅÉë K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K= UN
PK=_ÉåÅÜã~êâáåÖ=î~êá~ÄäÉJÇÉåëáíó=Ñäçï=~åÇ=íê~åëéçêí=áå=éçêçìë=ãÉÇá~=Åçåí~áåáåÖ=
~å=áåÅäáåÉÇ=Ñê~ÅíìêÉ=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K= UP
PKN= fåíêçÇìÅíáçå=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=KUP PKOKP= páãìä~íáçå=êÉëìäíëK=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K= US

PKO= s~êá~ÄäÉJÇÉåëáíó=cäçï=áå=~=mçêçìë=j~íêáñ=ïáíÜ=~=QRçJ PKP= pìãã~êó=~åÇ=`çåÅäìëáçåëK=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K= UV
fåÅäáåÉÇ=cê~ÅíìêÉ K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=KUP PKQ= kçí~íáçå =K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K= UV
PKOKN= pí~íÉãÉåí=çÑ=íÜÉ=Oa=éêçÄäÉã K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=KUP oÉÑÉêÉåÅÉë K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K= VM
PKOKO= dçîÉêåáåÖ=Éèì~íáçåëK=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=KUR
QK=kìãÉêáÅ~ä=ëáãìä~íáçå=çÑ=î~êá~ÄäÉJÇÉåëáíó=ãìäíáÇáÑÑìëáîÉ=ÑáåÖÉêáåÖ=ÅçåîÉÅíáçå=
éÜÉåçãÉå~=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K= VP
QKN= fåíêçÇìÅíáçå=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=KVP QKQKO= _çìåÇ~êó=~åÇ=áåáíá~ä=ÅçåÇáíáçåëI=éÉêíìêÄ~íáçå=çÑ=áåáíá~ä=

QKO= _~ëáÅ=bèì~íáçåë K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=KVQ ëçäìíÉ=áåíÉêÑ~ÅÉ =K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K= NMO
QKP= cáåÖÉê=jçÇÉä=mêçÄäÉãK=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=KVR QKR= oÉëìäíë K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K= NMP
QKQ= kìãÉêáÅ~ä=jçÇÉäáåÖ K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=KNMM QKS= `çåÅäìëáçåë K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K= NNQ
QKQKN= pé~íá~ä=~åÇ=íÉãéçê~ä=ÇáëÅêÉíáò~íáçåI=áíÉê~íáçå=ëíê~íÉÖó= kçí~íáçå =K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K= NNQ
~åÇ=ìëÉÇ=ëçäîÉêë=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=KNMM oÉÑÉêÉåÅÉë K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K= NNR
pìÄàÉÅí=fåÇÉñ K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K= NNT
^ìíÜçê=fåÇÉñ=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K= NON
áî=ö=tÜáíÉ=m~éÉêë=J=sçäK=fs
oÉ~ÅíáîÉ=ãìäíáJëéÉÅáÉë=íê~åëéçêí
H.-J. G. Dierscha, A. Pönitza, D. Etcheverryb & Y. Rossierb

N
a
WASY Institute for Water Resources Planning and Systems Research, Berlin, Germany
b
Sita Remediation SAS, Meyzieu, France
bined with conditional (if-else) expressions allowing a
N oÉ~ÅíáîÉ=ãìäíáJëéÉÅáÉë=íê~åëéçêí
NKN fåíêçÇìÅíáçå high degree of freedom in formulating the chemical

reaction kinetics. FEFLOW incorporates a fast code
Starting with release 5.2, FEFLOW5 solves reactive interpreter for the user-defined reactions. The computa-
multi-species transport processes in multi-dimensional tional performance is comparable to precompiled code.
(2D, 3D) porous media under variably saturated, vari- Precompiled code versions of reaction kinetics are
able-density and nonisothermal conditions involving available for reactions of the degradation, Arrhenius
an arbitrary number of species. A given species can be and Monod type.
either mobile, i.e., associated with a fluid (aqueous)
phase, or immobile, i.e., associated with a solid phase.
Chemicals in a fluid phase are subject to advection and NKO qÜÉçêó
dispersion, while in a solid phase there is no advection
and dispersion. NKOKN oÉ~Åíáçå=âáåÉíáÅë
NKOKNKN póãÄçäáÅ=êÉ~Åíáçå=ëíçáÅÜáçãÉíêó
FEFLOW provides enhanced tools in introducing
and editing reactive multi-species transport problems. The basis of the chemical modeling represents the
A powerful new reaction kinetics editor, FEMATHED, equations of reactions r which can be written in their
allows the user to freely define chemical or biochemi- general stoichiometric symbolic form4 (symbols are
cal reactions in a fully graphics-based and interactive summarized in the Appendix Notation):
manner without requiring any programming or code
compiling. Furthermore, the rate equations can be com-
ν 1r A 1 + ν 2r A 2 + … + ν N°r A N° → ν ( N° + 1 )r B ( N° + 1 ) + ν ( N° + 2 )r B ( N° + 2 ) + … + ν Nr B N (1-1)
cbcilt=ö=R
NK=oÉ~ÅíáîÉ=ãìäíáJëéÉÅáÉë=íê~åëéçêí
for N° < N ( N° = number of reactants) C 6 H 6 + 7.5O 2 → 6CO 2 + 3H 2 O (1-3)

and r = 1, …, Nr
(iii) Degradation of BTEX using multiple electron
which is related and quantified by the stoichiometric acceptors
coefficients ν kr . The algebraic stoichiometric num- The biodegradation of BTEX can occur via five differ-
bers ν kr are subject to: ent degradation pathways1,17,23: aerobic respiration,
denitrification, iron reduction, sulfate reduction and
ν kr < 0 for 1 ≤ k ≤ N° (reactants) methanogenesis. Accordingly, the following instanta-
ν kr > 0 for N° < k ≤ N (products). neous five reactions are given ( N r = 5 ):
In (1-1) A and B represent chemical species (reactants C 6 H 6 + 7.5O 2 → 6CO 2 + 3H 2 O ⎫

and products, respectively) and the symbol → iden- ⎪
-
6NO 3
+
+ 6H + C 6 H 6 → 6CO 2 + 6H 2 O + 3N 2 ⎪
tifies the direction of the irreversible (nonequilibrium) ⎪
2+ ⎪
reactions.
+
30Fe(OH) 3 + 60H + C 6 H 6 → 6CO 2 + 78H 2 O + 30Fe ⎬ (1-4)
⎪
3.75 SO 4 + 7.5H + C 6 H 6 → 6CO 2 + 3H 2 O + 3.75 H 2 S ⎪
2- +
⎪
C 6 H 6 + 4.5 H 2 O → 2.25 CO 2 + 3.75 CH 4 ⎪
⎭
NKOKNKO bñ~ãéäÉë
To illustrate the symbolic reaction stoichiometry (iv) Leaching of low-grade uranium ores
(1-1) we exemplify the following reactions occurring Two principal types of low-grade uranium ores are ura-
in different applications of subsurface mass transport ninite (UO2) and pitchblende (U3O8). Typical reaction
simulations. equations may be written as16
(i) Pyrite oxidation 1 2- - 4-

UO 2 + --- O 2 + CO 3 + 2HCO 3 → UO 2 (CO 3 ) 3 + H 2 O ⎫
2 ⎪
One mechanism involves oxidation of pyrite by O2. ⎬ (1-5)
1
U 3 O 8 + --- O 2 + 3CO 3 + 6HCO 3 → 3UO 2 (CO 3 ) 3 + 3H 2 O ⎪
2- - 4-
Another possible mechanism for the oxidation of pyrite 2 ⎭
is the reaction with Fe(III) as the oxidant. These reac-
tions for the pyrite oxidation read19: consisting of four reactants ( N° = 4 ) for each reaction.
7 +2 2- ⎫ +
FeS 2 + --- O 2 + H 2 O → Fe + 2SO 4 + 2H
2 ⎪ (v) Radionuclide decay chain of uranium
⎬ (1-2)
3+ 2+ 2- +⎪
FeS 2 + 14Fe + 8H 2 O → 15Fe + 2SO 4 + 16H ⎭
The radionuclide decay of 238U occurs in the following
decay series of serial and parallel reactions (note that U
- uranium, Th - thorium, Pa - protactinium, Ra -
(ii) Aerobic biodegradation of BTEX radium, Rn - radon, Po - polonium, Pb - lead, Bi - bis-
The overall aerobic reaction stoichiometry for a fuel muth, At - astatine, Tl - thallium)12:
hydrocarbon (e.g., benzene) can be written as1,3
S=ö=tÜáíÉ=m~éÉêë=J=sçäK=fs
NKO=qÜÉçêó
⎫
⎪
⎪
⎪
⎬ (1-6)
⎪
⎪
⎪
⎭
NKOKO dçîÉêåáåÖ=íê~åëéçêí=Éèì~íáçåë fluid f and solid s phases of a porous medium can be

concisely written in the following form4,5,8,14, assuming
NKOKOKN _~ëáÅ=Ä~ä~åÅÉ=ëí~íÉãÉåíë that each species k exists (per definition) only in one
The mass conservation of chemical species in the phase ( f or s ):
⎧ α = f, s (fluid, solid)
∂ α α α α α α ⎪ α α
( ε C ) – ∇ ⋅ ( D k ⋅ ∇C k ) + ∇ ⋅ ( q C k ) – ε α Q k = R k ⎨ q = D dk ≡ 0 for α = s (solid) (1-7)
∂t α k ⎪
⎩ k = 1, …, N
α ⎫
R k = ε α R hom + R het
k k ⎪
⎪
N ⎪
∑ Rk = 0 ⎪
⎬ (1-8)
k=1 ⎪
α α⎪
α α α q ⊗q ⎪
D k = ( ε α D d + β T q )I + ( β L – β T ) -------------------
α ⎪
k
q ⎭
α
where each species, labeled by the subscript k , is asso- Dk = tensor of hydrodynamic dispersion of
ciated with a particular phase α ∈ ( f, s ) , where f and s species k of α -phase;
α
indicate the fluid and the solid phase, respectively. In Dd
k
= coefficient of molecular diffusion of
(1-7) we have species k of α -phase;
I = unit tensor;
α
Ck = concentration of species k of α -phase; k = species indicator, k = 1, …, N ;
cbcilt=ö=T
N = total number of chemical species; express the volume fraction ε α for the fluid phase
α
Qk = zero-order nonreactive production term of α = f and for one solid phase α = s in a porous
α -phase; medium as
α α
q = ε α v , Darcy flux of α -phase;
α εf = sf ε εs = 1 – ε (1-9)
v = pore velocity of α -phase;
Rk = bulk rate of chemical reaction of species k ; with
α
R hom = homogeneous rate of reaction of species k
k
within the α -phase; 0 < sf ≤ 1 (1-10)
R het = heterogeneous rate reaction of species k
k
associated with different phases;
where ε is the porosity (void space) of the porous
α = phase indicator, α = f, s, … ;
medium, and s f is the saturation referring to the fluid
β L, β T = longitudinal and transverse dispersivity of
(liquid) f-phase. For saturated media we have s f ≡ 1
porous medium, respectively.
and ε f = ε . Using (1-9) the balance equations (1-7) can
εα = volume fraction of α -phase;
be written for the f- and s-phases in their specific forms
according to
Considering the fluid phase as a wetting liquid in an
unsaturated porous medium, it is often convenient to
∂ f f f f f f
( s εC ) – ∇ ⋅ ( D k ⋅ ∇C k ) + ∇ ⋅ ( q C k ) – s f Q k = R k fluid phase f and species k
∂t f k
(1-11)
∂ s s
( ε C ) – Qk = Rk solid phase s and species k
∂t s k
where (1-11) can also be subjected to a reversible equilibrium

reaction with a species, say m, which is associated with
f f
Qk = εQ k , bulk zero-order nonreactive the solid phase s, so that we find the adsorption
production term of fluid phase and fluid isotherms8
species k;
s s
Qk = ε s Q k , bulk zero-order nonreactive
production term of solid phase and solid f s →
νk Ak + νm Am ←
s f
νk Ak + νm Am (1-12)
species k;
f s
NKOKOKO dÉåÉê~äáòÉÇ=Ä~ä~åÅÉ=Éèì~íáçåë where A k is the fluid species, A m is the sorbed (solid)
species, and the symbol → ← identifies reversible reac-
Let us assume that a species k of a fluid phase f for tions at a chemical equilibrium. The law of mass action
U=ö=tÜáíÉ=m~éÉêë=J=sçäK=fs
NKO=qÜÉçêó
(LMA)8 can be used to explicitly express the (sorbed) ing to8

s
solid species C m as a function of the (dissolved) fluid
f 1–ε
species C k , viz.8, ℜ k = 1 + ----------- χ k (1-16)
ε
s f
Cm = χk ⋅ Ck (1-13)
with the adsorption function χ k
f
where χ k = χk ( Ck )
is the adsorption function, which
f ⎧κ Henry
can be dependent on Ck . ⎪
⎪ k1
⎪ --------------------- Langmuir
As stated in8 for equilibrium reactions (adsorptions) χk = ⎨ 1 + k Cf (1-17)
⎪ 2 k
under unsaturated conditions the solid volume fraction
⎪ ( b2 – 1 )
ε s should be subdivided into chemically active and ⎪ b 1 C fk Freundlich
inactive parts ε s = ε s + ε s . It is shown8 that
active inactive
⎩
εs = f ( ε f )ε s ≈ s f ε s . Then, the balance equations for
active
species k and m can be written as where κ, k 1, k 2, b 1, b 2 correspond to sorption coeffi-

cients.
∂ f f f f f f
( s εC ) – ∇ ⋅ ( D k ⋅ ∇C k ) + ∇ ⋅ ( q C k ) – s f Q k = R k∗
∂t f k
(1-14)
∂ s s
NKOKOKP cáå~ä=Éèì~íáçåë
( ε s active C m ) – ε sactive Q = R m∗
∂t m
In summation we use the following generalized
(Note that we identify the chemical rates by * for spe- mass balance equations for chemically reacting species
cies having additionally equilibrium reactions.) Using k in the fluid phase f and solid phase s (species
k = 1, …, N are subjected to nonequilibrium reac-
the adsorption isotherms (1-13) the following balance
equation results after summation of both equations tions):
(1-14)
∂ f f f f f
( s εℜ C ) – ∇ ⋅ ( D k ⋅ ∇C k ) + ∇ ⋅ ( q C k ) (1-15)
∂t f k k
f s f
– s f ( Q k + Q m ) = R k ∗ + R m∗ ( C k )
⎧
⎪
⎨
⎪
⎩
⎧
⎪
⎪
⎨
⎪
⎪
⎩
Qk Rk
where the retardation factor ℜ k is introduced accord-
cbcilt=ö=V
∂ f f f f f
( s εℜ C ) – ∇ ⋅ ( D k ⋅ ∇C k ) + ∇ ⋅ ( q C k ) – s f Q k = R k fluid phase f and species k ⎫
∂t f k k ⎪
⎬ (1-18)
∂ s
solid phase s and species k ⎪
s
( ε C ) – Qk = Rk
∂t s k ⎭
1–ε ⎫
ℜ k = 1 + ----------- χ k ⎪
ε ⎪
f f⎬ (1-19)
f f f q ⊗ q- ⎪
D k = ( s f εD d + β T q )I + ( β L – β T ) ---------------
f ⎪
k
q ⎭
For depth-integrated 2D horizontal formulations of and unconfined aquifers result:

the basic equations (1-18) and (1-19) the following
specific multi-species transport equations in confined
∂ f f f f f
fluid phase f and species k ⎫
)
( εℜ k C k ) – ∇ ⋅ ( D k ⋅ ∇C k ) + ∇ ⋅ ( q C k ) – Q k = BR k
∂t ⎪
⎬ (1-20)
∂ s
solid phase s and species k ⎪⎭
s
)
( ε BC ) – Q k = BR k
∂t s k
1–ε ⎫
ℜ k = B ⎛ 1 + ----------- χ k⎞ ⎪
⎝ ε ⎠
⎪
f f⎬
(1-21)
f f f q ⊗q ⎪
D k = ( εBD d + β T q )I + ( β L – β T ) ---------------
-
f ⎪
k
q ⎭
and
q = Bq ⎫⎪
f f
⎪
)
Q k = BQ k ⎬ (1-22)
s
⎪
Q k = BQ k ⎪⎭
s
)
where B denotes the aquifer thickness (cf. FEFLOW’s Reference Manual5).
NM=ö=tÜáíÉ=m~éÉêë=J=sçäK=fs
NKO=qÜÉçêó
NKOKP oÉ~Åíáçå=ê~íÉë=oâ NKOKPKN aÉÖê~Ç~íáçå=íóéÉ=âáåÉíáÅë

For a given number of irreversible reactions For degradation type kinetics the rate r r can be
r = 1, …, N r the reaction rate R k for a species k can developed in a polynomial representation of low order,
be expressed in the following general form4,8,14 viz.,
Nr N nm
α
Rk = ∑ νkr rr ( k = 1, …, N ) (1-23)
rr = ∑ B rm ( C m ) α = f or s (1-25)
m=1
r=1
where
where B rm is a rate constant and n m ≥ 0 corresponds to
Nr = number of reactions; an exponent of species m. For such a type the reaction
rr = rate of reaction associated with the type of rate R k (1-23) can be written in the general form
reaction r;
N
ν kr = stoichiometric number of species k and α nm
reaction r.
Rk = ∑ κm ( Cm ) α = f or s
m=1 (1-26)
Reactive systems can be broadly classified into simple α n1 α n2 α nN

= κ 1 ( C 1 ) + κ 2 ( C 2 ) + …+ κ N ( C N )
and complex kinetic systems. The former consists of
unimolecular and bimolecular reactions while the latter
encompasses opposing, concurrent, and consecutive where κ m ( m = 1, …, N ) are bulk rate constants that
reactions. According to the mechanism of a given reac- can be dependent on ε and s f .
tion, the functional form of r r can be very complicated.
A typical constitutive representation of r r has a

functional
reaction rate
α α Monod type
r r = r r ( C 1 , …, C N, s f, ε ) α = f or s (1-24)
Arrhenius type
which can be developed by a polynomial expression of

α 1st order degradation type
low order in terms of concentrations C k for simple
kinetic systems or more complex rate expressions of
concentration(s)
higher order, cf. Fig. 1.1. We can distinguish the fol-
lowing classes of rate expressions. Figure 1.1 Variation of reaction rate with concentration.
cbcilt=ö=NN
NKOKPKO ^êêÜÉåáìë=íóéÉ=âáåÉíáÅë enzyme kinetics equation8. Its mathematical represen-

tation can be written as
The Arrhenius type kinetics are expressed by a
polynomial representation of higher order for the rate
α nn
rr , N n N° ⎛ B rn ( C n ) ⎞
α r ⎜ -⎟
rr = ∑ A rm ( C r ) ∏
---------------------------------------
⎜ nn α n n⎟
(1-29)
m=1 n = 1 ⎝ D rn + E rn ( C n ) ⎠
N nm N nm
1 α 2 α β nn
rr = ∑ B rm ( C m ) + ∑ Brm ( Cm ) ( Cn ) + … (1-27)
m=1 m, n where A rm, B rn, D rn, E rn are rate constants. In (1-29)
α
N
N α nm
C n represents the concentration of the growth-limiting
+ ∏ B rm ( C m ) substrate. The term B rn is the maximum specific
m=1 growth rate and the term D rn denotes the half-satura-
α
tion constant and is defined as the concentration C n at
1 N
introducing a number of rate constants B rm, …, B rm . which the microorganisms grow at half the maximum
Accordingly, we find for the reaction rate R k (1-23) the specific growth rate B rn . The half-saturation constant
following expression may be viewed as a measure of the affinity the microor-
ganisms have for the growth-limiting substrate: (1) the
lower the value of Drn , the greater the capacity to grow
α n1 α n1 α n2
Rk = κ1 ( C1 ) + κ2 ( C1 ) ( C2 ) + … (1-28) rapidly in an environment with low concentrations of
growth-limiting substrate, and (2) the lower the value
N
α nm
of D rn , the lower the growth-limiting substrate concen-
+ κN ∏ ( Cm ) α = f or s tration at which the maximum specific growth rate B rn
m=1 is attained.
where κ m ( m = 1, …, N ) are bulk rate constants that Using (1-29) with (1-23) the following expression
can be dependent on ε and s f . For the species expo- for the reaction rate R k is obtained:
nents n m ≥ 0 is assumed.
NKOKPKP jçåçÇ=íóéÉ=âáåÉíáÅë
Monod type kinetics describe more complex bio-
chemical reaction systems. Monod was the first to rec-
ognize that the growth rate of a microbial population is
restricted by the concentration of the growth-limiting
substrate. Monod established that the form of the rela-
tionship was analogous to the Michaelis-Menten
NO=ö=tÜáíÉ=m~éÉêë=J=sçäK=fs
NKO=qÜÉçêó
n n n
⎛ b1 ( Cα ) 2 ⎞ ⎛ b1 ( Cα ) 2 ⎞ ⎛ b1 ( Cα ) 3 ⎞
α n1 ⎜ α n1 ⎜
1 11 2
R k = κ 1 ( C 1 ) ⎜ ------------------------------------- - + κ 2 ( C 1 ) -------------------------------------- --------------------------------------⎟
⎟ 1 21 2 ⎟ ⎜ 22 3
(1-30)
1 1 α n 2⎟ ⎜ 1 1 α n 2⎟ ⎜ 1 1 α n 3⎟
⎝ d 11 + e 11 ( C 2 ) ⎠ ⎝ d 21 + e 21 ( C 2 ) ⎠ ⎝ d 22 + e 22 ( C 3 ) ⎠
n n n α nn
⎛ b1 ( Cα ) 2 ⎞ ⎛ b1 ( Cα ) 3 ⎞ ⎛ b1 ( Cα ) 4 ⎞ N° ⎛ 1
b Nn ( C n ) ⎞
α n1 ⎜ ⎟ ⎜ ⎟ ⎜ ⎟ α n1 ⎜
1 31 2 32 3 33 4
- -------------------------------------- -------------------------------------- + … + κ N ( C 1 )
+ κ 3 ( C 1 ) ⎜ -------------------------------------
1
----------------------------------------⎟
∏
1 1 α n 2⎟ ⎜ 1 1 α n 3⎟ ⎜ 1 1 α n 4⎟ ⎜ 1 1 α n n⎟
⎝ d 31 + e 31 ( C 2 ) ⎠ ⎝ d 32 + e 32 ( C 3 ) ⎠ ⎝ d 33 + e 33 ( C 4 ) ⎠ n = 2 ⎝ d Nn + e Nn ( C n ) ⎠
n α n1 ⎞ ⎛ α n3 ⎞
⎛ b2 ( Cα ) 1 ⎞ n2 ⎛
2
b 21 ( C 1 )
2
b 22 ( C 3 )
α n2
+ κ 1 ( C 2 ) ⎜ -------------------------------------
-⎟ + κ 22 ( C α2 ) ⎜ ------------------------------------- -⎟ ⎜ -------------------------------------
-⎟
2 11 1
⎜ 2 2 α n 1⎟ ⎜ 2 2 α n 1⎟ ⎜ 2 2 α n 3⎟
⎝ d 11 + e 11 ( C 1 ) ⎠ ⎝ d 21 + e 21 ( C 1 ) ⎠ ⎝ d 22 + e 22 ( C 3 ) ⎠
n n n n
⎛ b2 ( Cα ) 1 ⎞ ⎛ b2 ( Cα ) 3 ⎞ ⎛ b2 ( Cα ) 4 ⎞ N° ⎛ b2 ( Cα ) n ⎞
α n2 n2
+ κ 3 ( C 2 ) ⎜ -------------------------------------
-⎟ ⎜ -------------------------------------
-⎟ ⎜ -------------------------------------
-⎟ + … + κ 2N ( C α2 ) ∏ ⎜⎜ ----------------------------------------⎟
2 31 1 32 3 33 4 Nn n
⎜ 2 2 α n 1⎟ ⎜ 2 2 α n 3⎟ ⎜ 2 2 α n 4⎟ 2 2 α n n⎟
⎝ d 31 + e 31 ( C 1 ) ⎠ ⎝ d 32 + e 32 ( C 3 ) ⎠ ⎝ d 33 + e 33 ( C 4 ) ⎠ n = 1, n ≠ 2 ⎝ d Nn + e Nn ( C n ) ⎠
+…
n n n
Nr
nN ⎛ bNr ( Cα ) 1 ⎞ nN ⎛ Nr
b 21 ( C 1 )
α 1 ⎞⎛ Nr
b 22 ( C 2 )
α 2 ⎞
+
α
κ1 ( CN )
r
⎜ -------------------------------------
11 1
-⎟ + κ N ( C
α
)
r
⎜ -------------------------------------
- ⎟ ⎜ -------------------------------------
-⎟
r ⎜ Nr N α n 1⎟
2 Nr ⎜ Nr N α n 1⎟ ⎜ N N α n 2⎟
⎝ d 11 + e 11r ( C 1 ) ⎠ ⎝ d 21 + e 21r ( C 1 ) ⎠ ⎝ d 22r + e 22r ( C 2 ) ⎠
n n n n
Nr
nN ⎛ bNr ( Cα ) 1 ⎞ ⎛ bNr ( Cα ) 2 ⎞ ⎛ bNr ( Cα ) 3 ⎞ N
nN N° ⎛ bNr ( Cα ) n ⎞
⎜ -------------------------------------
-⎟ ⎜ -------------------------------------
-⎟ ⎜ -------------------------------------
-⎟ + … + κ N r ( C N ) ∏ ⎜⎜ ----------------------------------------⎟
α r 31 1 32 2 33 3 α r Nn n
+ κ3 ( CN )
r ⎜ Nr Nr α n 1⎟ ⎜ N r Nr α n 2⎟ ⎜ N r Nr α n 3⎟ r Nr Nr α n n⎟
⎝ d 31 + e 31 ( C 1 ) ⎠ ⎝ d 32 + e 32 ( C 2 ) ⎠ ⎝ d 33 + e 33 ( C 3 ) ⎠ n = 1, n ≠ N r ⎝ d Nn + e Nn ( C n ) ⎠
or in general terms
n
Nr N n N° ⎛ bm ( Cα ) i ⎞
∏ ⎜⎜ -----------------------------------
-⎟
m α m ni i
Rk = ∑ ∑ κn ( Cm )
m m α n i⎟
α = f or s (1-31)
m = 1n = 1 i = 1, i ≠ m ⎝ d ni + e ni ( C i ) ⎠
m m m m
with bulk rate constants κ n , b ni, d ni, e ni ( m = 1, …, N r )( n = 1, …, N )( i = 1, …, N° ), which can be dependent on
m m
ε and s f . For the species exponents n n ≥ 0 is assumed. In (1-31) b ni is a maximum specific growth rate and d ni
is a half-saturation constant.
cbcilt=ö=NP
NKOKPKQ m~ê~ääÉä=áêêÉîÉêëáÄäÉ=~åÇ=êÉîÉêëJ and assuming that the decay rate constants ϑ k are iden-
áÄäÉ=êÉ~Åíáçåë tical for the fluid and solid phases, the balance equa-
tions (1-33a) and (1-33b) can be simply added. This
Let us consider the case when irreversible and yields the following overall species mass balance equa-
reversible reactions occur together. For instance, spe- tion
cies are dissolved in the fluid phase f and undergo a
decay reaction. At the same time, these species may be ∂ f f f f f
adsorbed at the solid phase s in equilibrium reactions. ( s εℜ C ) – ∇ ⋅ ( D k ⋅ ∇C k ) + ∇ ⋅ ( q C k ) – s f Q k (1-35a)
∂t f k k
As exemplified for a reactive eight-species system with = Rk ( k = A, B, C, D )
a degradation-type kinetics we can write
Af Bf Cf Df with the reaction rates of a degradation type
f f
Rk = sf ε ( ϑk – 1 ℜk – 1 Ck – 1 – ϑk ℜk Ck ) (1-35b)
Es Fs Gs Hs (1-32)
( k = A, B, C, D )
Accordingly, the following balance equations result for

få= íÜÉ=Å~ëÉ= çÑ= ëáåJ
the mobile species k of the fluid phase f where ℜ k denotes the retardation factor defined by (1-
ÖäÉJëéÉÅáÉë= íê~åëJ
16).
éçêí= EN = 1F= íÜÉ
∂ f f f f f f
êÉ~ÅíáåÖ= ëéÉÅáÉë= áë= ~äï~óë ( s εC ) – ∇ ⋅ ( D k ⋅ ∇C k ) + ∇ ⋅ ( q C k ) – s f Q k (1-33a)
ÅçåëáÇÉêÉÇ= ~ë= ãçÄáäÉK= qÜÉ ∂t f k Similar to (1-35b), rate expressions can be derived
f f
êÉ~Åíáçå=ê~íÉ=R1=áë=íÜÉå = sf ε ( ϑk – 1 Ck – 1 – ϑk Ck ) ( k = A, B, C, D ) for more complex parallel reaction kinetics when all or
only part of the species undergo a reversible equilib-
f
====== R 1 = – s f εϑ 1 ℜ 1 C 1 rium reaction (adsorption isotherms). Here, the corre-
and for the immobile species m associated with the sponding species of the solid phase are substituted via
solid phase s the adsorption isotherms (1-13) and the balance equa-
kçíáÅÉI=íÜÉ=ÇÉÅ~ó tions for the fluid and solid species are added, resulting
∂ s s
ê~íÉ= ϑ k = Å~å= ÄÉ ( s ε C ) – sf εs Qm (1-33b) in modified rate expressions R k that incorporate the
∂t f s m
ÉñéêÉëëÉÇ= Äó= íÜÉ s s retardation factor ℜ k .
êÉ~Åíáçå= Ü~äÑJäáÑÉ= t k, 1 ⁄ 2 = çÑ = sf εs ( ϑm – 1 Cm – 1 – ϑm Cm ) ( m = E, F, G, H )
ëéÉÅáÉë=â=áå=íÜÉ=ÑçêãW
ln2
========== ϑ k = --------------
where ϑ k are decay rate constants. Using the adsorp- NKP kìãÉêáÅ~ä=pçäìíáçå
t k, 1 ⁄ 2 tion isotherms (1-13) with
NKPKN cáåáíÉJÉäÉãÉåí=Éèì~íáçå=ëóëíÉã
s f
ïÜÉêÉ= ln2 = 0.693 K Cm = χk ⋅ Ck ( k = A, B, C, D ) ( m = E, F, G, H ) (1-34) FEFLOW uses the finite-element method to solve
the final mass balance equations (1-18) of species k in
NQ=ö=tÜáíÉ=m~éÉêë=J=sçäK=fs
NKP=kìãÉêáÅ~ä=pçäìíáçå
the fluid f or solid phase s. It ends up with the following C so that an efficient numerical solution strategy is
matrix system: required here. One possibility would be the solution of
the coupled matrix system (1-36a) in a direct and
dC simultaneous manner. Although mathematically rigor-
S ⋅ ------- + K ⋅ C – R = 0 (1-36a)
dt ous, practical implementation of that approach has
or been limited and is not generally applicable to large,
geometrically complex and multidimensional problems
⎧ dC 1 ⎫ because of the significant memory and/or computa-
⎪ --------- ⎪
⎪ dt ⎪ tional burden. The size of the coefficient matrices S
S1 0 0 … 0 ⎪ dC ⎪ K1 0 0 … 0 ⎧ C1 ⎫
⎪ ---------2 ⎪ ⎪ ⎪ and K in the discretized system (1-36a) grows as a
0 S2 0 … 0 ⎪ dt ⎪ 0 K2 0 … 0 ⎪ C2 ⎪
⎪ ⎪ ⎪ ⎪
(1-36b)
product of the number of nodes NP and the number of
0 0 S3 … 0 ⎨ dC 3 ⎬ + 0 0 K3 … 0 ⎨ C3 ⎬
⎪ --------- ⎪ ⎪ ⎪ species N. In general, the direct approach involves
… ⎪ dt ⎪ … ⎪…⎪
0 0 0 … SN
⎪ … ⎪
0 0 0
⎪ ⎪
… KN ⎩ CN ⎭
solving a NP × N system of nonlinear equations at each
⎪ ⎪
⎪ dC N ⎪ time step. Furthermore, the system for a simultaneous
⎪ ---------- ⎪
⎩ dt ⎭ solution can be ill-conditioned due to the significantly
different scales of the processes involved. Alterna-
tively, in order to reduce the computational require-
⎧ R1 ⎫ 0
⎪ ⎪ ⎧ ⎫ ments, a decoupled (or split-operator) solution strategy
⎪ R2 ⎪ ⎪0⎪
⎪ ⎪ ⎪ ⎪
– ⎨ R3 ⎬ = ⎨ 0 ⎬
is preferred, in which the species equations are solved
⎪ ⎪ ⎪ ⎪ sequentially by using efficient iteration techniques.
⎪…⎪ ⎪… ⎪
⎪ ⎪
R
⎩0⎭ Kanney et al.13 discuss different strategies of such split-
⎩ N⎭
operator approaches. Among a variety of split-operator
techniques the sequential iterative approach (SIA) have
where S k = S k ( C ) ( k = 1, …, N ) are the nonlinear proven superior and powerful. In FEFLOW, we prefer
storage matrices including retardation effects, K k an adaptive error-controlled SIA strategy which is
( k = 1, …, N ) are the ’conductance’ matrices encom- based on an efficient predictor-corrector time-stepping
passing advection and dispersion effects, and technique. In contrast to a common SIA technique the
R k = R k ( C ) ( k = 1, …, N ) are the chemical rate vec- transport equations with the reaction terms are solved
tors, which represent nonlinear dependencies on the in an adaptive full time interval using predictor solu-
total concentration vector C = { C1, C 2, C 3, …, CN } tions to linearize the nonlinear reaction terms. The
according to the considered reaction kinetics. Further- overall iteration control is fully embedded in a time-
more, we note that there is no conductance K k ≡ 0 for a marching strategy via a sophisticated error-based time-
species k belonging to a solid phase s. step adaptation.
The matrix system (1-36a) can be highly nonlinear

mainly due to the dependence of the reaction rate R on
cbcilt=ö=NR
NKPKO ^Ç~éíáîÉ= éêÉÇáÅíçêJÅçêêÉÅíçê n

⎛S k(C) Ck 1 · n
+ K k⎞ ⋅ C k = S k ( C ) ⋅ ----------- + ⎛ --- – 1⎞ C k
n+1
íáãÉJëíÉééáåÖ=ëçäìíáçå=ëíê~íÉÖó --------------
⎝ θΔt n ⎠ θΔt n ⎝ θ ⎠
(1-39)
The predictor-corrector solution strategy with auto-
n+1
matically adapted time stepping has been shown very + Rk (C) ( k = 1, …, N )
cost-efficient and robust for classes of nonlinear sys-
tems such as variable density9 and/or variably
saturated6 problems. We also apply this technique for where n+1 identifies the new time solution and
solving the present chemically reactive systems. θ ∈ ( 1--2-, 1 ) is a weighting factor used for the trapezoid
rule (TR) of second-order accuracy and the backward
We rewrite the composed matrix system in separate Euler (BE) schemes of first-order accuracy in time t.
matrix systems for each species k as follows
The system is nonlinear due to the dependence on
· C = { C 1, C 2, C 3, …, C N } and can be appropriately lin-
Sk ( C ) ⋅ Ck + Kk ⋅ Ck – Rk ( C ) = 0 ( k = 1, …, N ) (1-37)
earized by using the following adaptive predictor-cor-
rector time-stepping strategy:
or
(Step 0 - Initialization) Computation of the initial
S 1 ( C 1, …, C N ) ⋅ C 1 + K 1 ⋅ C 1 – R 1 ( C 1, …, C N ) = 0 ⎫⎪
· · 0
acceleration vectors C k for n = 0 (once per k-species
· ⎪
S 2 ( C 1, …, C N ) ⋅ C 2 + K 2 ⋅ C 2 – R 2 ( C 1, …, C N ) = 0 ⎪ equation)
⎪
·
S 3 ( C 1, …, C N ) ⋅ C 3 + K 3 ⋅ C 2 – R 3 ( C 1, …, C N ) = 0 ⎬ (1-38)
⎪ 0 · 0 0 0
… ⎪
⎪
Sk ( C ) ⋅ Ck = –Kk ⋅ Ck + Rk ( C ) ( k = 1, …, N ) (1-40)
·
S N ( C 1 , … , C N ) ⋅ C N + K N ⋅ C N – R N ( C 1 , …, C N ) = 0 ⎪
⎭
and guessing of an initial time step Δt 0 . The initial sys-
tems (1-40) are solved with the initial concentration
· dC 0 0 0 0 0
with Ck = --------k- ( k = 1, …, N ) . vector C = { C1, C 2, C 3, …, C N } .
dt
The nonlinear matrix systems (1-37) are discretized in (Step 1 - Predictor solutions) Perform explicit predic-
time t by using implicit time-stepping techniques. Con- tor solutions for all species k by using the Adams-
sidering Ck within the time interval ( t n, t n + Δt n ) , Bashforth (AB) or the forward Euler (FE) algorithm,
where the subscript n denotes the time plane and Δt n is respectively:
a variable time step length, the implicit approximation
of (1-37) yields ⎧ C n + Δt Δt n
-------n- 2 + --------------
n Δt n · n – 1
- C· – --------------
-C
n + 1, p ⎪ k 2 Δt n – 1 k Δt n – 1 k
Ck = ⎨ (1-41)
⎪ n · n
⎩ C k + Δt n C k ( k = 1, …, N )
NS=ö=tÜáíÉ=m~éÉêë=J=sçäK=fs
(Step 2 - Corrector solutions) Do corrector solution (Step 5 - Prediction of new time step) Predict the
via the TR or BE schemes (1-39) by applying the pre- potential new k-specific time-step lengths from the
dictor solution error estimates (1-44) for each species k:
n + 1, p n + 1, p n + 1, p n + 1, p n + 1, p
C = { C1 , C2 , C3 , …, C N } from
(1-41) to linearize the species equations as ⎛ δ ⎞1 ⁄ λ
Δt n + 1, k = Δt n ⎜ ----------------
-⎟ ( k = 1, …, N ) (1-45)
⎝ d nk + 1 ⎠
n + 1, p
⎛ Sk ( C ) ⎞ n+1
⎜ ----------------------------- + K k⎟ ⋅ C k (1-42)
⎝ θΔt n ⎠ where λ is 3 for the AB/TR and 2 for the FE/BE
n
Ck 1 · n scheme, δ is a user-specified error tolerance
) ⋅ ----------- + ⎛⎝ --- – 1⎞⎠ C k
n + 1, p
= Sk ( C –4 –3
θΔt n θ ( δ = 10 – 10 is typical), and . is a norm to be
chosen as the weighted RMS norm or, alternatively, the
n+1 n + 1, p
+ Rk (C ) ( k = 1, …, N ) maximum norm, respectively:
⎧ 1⁄2
⎛ NP ⎞
(Step 3 - Updated time derivatives) Update the new ⎪ 1 1 ⎜ n + 1 2⎟
max NP ⎜ ∑ ki
⎪ -----------
- ------- d
⎟
time-derivative vectors by inverting the TR and BE n+1 ⎪ Ck ⎝ i=1 ⎠
dk = ⎨ (1-46)
schemes, respectively: ⎪
1 -⎛
⎪ ----------- n+1⎞
max d ( k = 1, …, N )
⎪ C max ⎝ i ki ⎠
⎧ 2 n+1 n · n ⎩ k
⎪ -------- ( C k – Ck ) – Ck
· n+1 Δt
⎪ n
Ck = ⎨ (1-43) max
1
⎪ ------- n+1 n where Ck is the maximum concentration of species k
-(C – C k ) ( k = 1, …, N )
⎪ Δt n k at the time plane n+1, NP is the total number of spa-
⎩ n+1
tially discretized points (nodes), and d ki denotes the
error of species equations k at the node i
(Step 4 - Error estimation) Compute the local trunca- ( i = 1, …, NP ) .
tion error for the AB/TR or FE/BE scheme, respec-
tively: (Step 6 - Tactic of adaptive time stepping and error
control) The following criteria are used to monitor the
⎧ n+1 n + 1, p progress of the nonlinear solution:
⎪C k – Ck
-------------------------------------
-
⎪ Δt n – 1⎞
n+1 ⎪ 3 ⎛ 1 + -------------- - 1. If
dk = ⎨ ⎝ Δt n ⎠ (1-44)
⎪
⎪1 n + 1 n + 1, p Δt n + 1, k ≥ Δt n (1-47a)
⎪ --2- ( C k – Ck ) ( k = 1, …, N )
⎩
cbcilt=ö=NT
n+1
the solution C k for the species equation k is accurate
Δt n + 1 = min ( Δt n + 1, k ) (1-47e)
within the error bound defined by δ and the increase of
k
the time step is always accepted.
2. If It is important to note that the error tolerance δ is

the only user-specified parameter to control the entire
ζΔt n ≤ Δt n + 1, k ≤ Δt n (1-47b) nonlinear and transient solution process. The starting-
up phase is still influenced by the initial time step Δt 0
where ζ is typically 0.85, the kth solution Ck is
n+1 which should be kept small. In practice, two further
accepted but the time step is not changed, i.e., constraints for the time-step size have shown to be use-
Δt n + 1, k = Δt n . ful. Firstly, the time step should not exceed a maximum
max
measure, i.e., Δt n + 1 ≤ Δt . Secondly, the rate for
3. If changing the time-step size Ξ = Δt n + 1 ⁄ Δt n can also
max
be limited, e.g., Ξ ≤ Ξ (say, 2 or 3).
Δt n + 1, k < ζΔt n (1-47c)
The predictor-corrector strategy fully monitors the
n+1 nonlinear and transient solution process via the time
the solution C k cannot be accepted within the truncation error estimation in which the size of the time
required error tolerance δ and has to be rejected. The step is cheaply and automatically varied in accordance
proposed new time step size (1-45) has to be reduced with the overall accuracy requirements. The time step
according to is increased whenever possible and decreased only
2
when necessary. It is evident to note that by monitoring
* Δt n ⎛ δ ⎞ the temporal accuracy requirements, at the same time
Δt n + 1, k = ------------------- ⎜ ----------------
-⎟ (1-47d)
Δt n + 1, k ⎝ d n + 1 ⎠ the solution strategy provides an efficient control of the
k
nonlinearities of the species transport equation system
via the predictor solutions. Due to the power of the pre-
and the solution is repeated for the time plane n+1 with dictor-corrector strategy any additional iterative feed-
*
Δt n = Δt n + 1, k . back within the adapted time steps can be avoided.
4. If the criteria (1-47a) and (1-47b) are satisfied by all

n+1
species equations and the solutions C k can be
accepted for all species k within the required error tol-
erance δ , the new time step is determined from the
minimum of the k-specific time step lengths, viz.,
NU=ö=tÜáíÉ=m~éÉêë=J=sçäK=fs
Set MASS (or THERMOHALINE for non-isothermal) transport
Set multi-species option
Input number of species
Open species list editor for phase definition and naming
Species IDs automatically countered
Phase definition: fluid or solid
Naming of species
Figure 1.2 FEFLOW’s Problem Classifier dialog with selected multi-species mass transport, and Chemical
Species List Editor shown for 7 species.
cbcilt=ö=NV
NKQ oÉä~íÉÇ=cbcilt=aá~äçÖë NKQKO péÉÅáÉë=ëÉäÉÅíçê

If multi-species transport is defined a Species Selec-
NKQKN jìäíáJëéÉÅáÉë= éêçÄäÉã= Åä~ëë tor (Fig. 1.3) appears in proper contexts of data input,
ëéÉÅáÑáÅ~íáçå evaluation and drawing of data and results. It displays
In FEFLOW’s Problem Editor the Problem Classi- the selected species ID (with the phase association and
fier dialog allows the introduction of multi-species the name) which applies to all input, evaluation, graph-
mass transport as shown in Fig. 1.2. This is possible for ical output and export. The selector is used to switch
the problem classes MASS or THERMOHALINE between the different species IDs.
TRANSPORT, both under saturated or variably satu-
rated conditions. The problem class THERMOHA-
LINE can be used to define a nonisothermal (thermally
coupled) multi-species transport problem.
In the dialog the number of species N must be set. Figure 1.3 FEFLOW’s species selector is used to browse
This is an arbitrary positive integer (practically, up to a through all defined species.
maximum of 10,000 species is possible). The default
option (N = 1) is single-species transport. Multi-species The species selector automatically appears for ini-
transport is considered for a number N greater than tial data, boundary data and material data of multi-spe-
unity. In this case the Species List Editor (see Fig. 1.2) cies mass transport. All input applies to the species ID
can be opened to name the species and to define the currently shown by the selector. In the problem editor
phase association. At the same time the species list edi- there are also additional options to copy the problem
tor automatically introduces the so-called species IDs data for different species IDs (Fig. 1.4).
( ID = 1, …, N ) , which are used to identify the selected
species. (Note that the species ID can be considered as
the species indicator k = 1, …, N used in the above
theoretical developments). Furthermore, a species con-
α
centration C k = ID can be associated either with a fluid
phase ( α = f ) or with the solid phase ( α = s ) , which is
related to the governing phase-specific mass transport
equations (1-18). In the mobile fluid phase f the species
are dissolved and subject to dispersion and advection,
while species in the immobile solid phase s have no dis-
persion and advection.
Figure 1.4 Copy of input data for different species.
OM=ö=tÜáíÉ=m~éÉêë=J=sçäK=fs
NKR=oÉ~Åíáçå=háåÉíáÅë=bÇáíçê
irrelevant and input is disabled for those parameters.

If the selector is applied a redraw of the underlying Moreover, for the solid phase the solid volume fraction
data (e.g., boundary conditions, material data, plotted ε s = 1 – ε is to be entered instead of porosity ε , which
results, diagram curves) is immediately performed. must be entered for species associated with the fluid
phase f, cf. Fig. 1.5.
NKQKP jìäíáJëéÉÅáÉë=ã~íÉêá~ä=Ç~í~ A new button identified by the Σ symbol launches

the Reaction Kinetics Editor, where the reaction rates
Entering the mass-material menu in FEFLOW’s
R k are edited (as thoroughly described in Chapter 1.5).
problem editor the material data are related to the A Rate parameter (see Fig. 1.5) appearing in the mate-
selected species ID. If the selected species is associated rial menu can be used in the Reaction Kinetics Editor
with a solid phase s sorption parameter, molecular dif- to refer to distributed (or constant) rate coefficients.
fusion, longitudinal and transverse dispersivities are
porosity ε
appearing as additional
green variable Porosityk solid volume fraction εs
green variable SolidFrack
sorption coefficient κ (or k1, b1)
green variable Sorptionk irrelevant for solid phase
reaction kinetics editor reaction kinetics editor
rate parameter to be used

in the reaction kinetics editor rate parameter to be used
as green variables Ratek in the reaction kinetics editor
as green variables Ratek
Figure 1.5 Mass-material menus for species of a fluid phase f (left) and a solid phase s (right).
NKR oÉ~Åíáçå=háåÉíáÅë=bÇáíçê two principal classes of reaction expressions that can

be applied. First, the kinetics editor includes precom-
FEFLOW provides a powerful graphical tool for piled expressions for which no additional code inter-
editing the rate expression R k (1-23). Here, there are pretation is needed. The provided formulae refer to
cbcilt=ö=ON
simplified expressions of the degradation (1-26), NKRKN mêÉÅçãéáäÉÇ=ê~íÉ=ÉñéêÉëëáçåë

Arrhenius (1-28) and Monod (1-31) reaction types for a
limited number of reaction terms. NKRKNKN aÉÖê~Ç~íáçå=íóéÉ=âáåÉíáÅë
Pushing the Σ button in the material editor (Fig.
Second (at present available on Windows PC and 1.5) opens the Reaction Kinetics Editor with the default
LINUX platforms), FEFLOW’s FEMATHED kinetics option of a degradation-type kinetics as illustrated in
editor enables the input of user-defined reaction-rate Fig. 1.6.
expressions which the user can freely edit in an interac-
tive, graphics-based and flexible manner without any
limitations on the algebraic structure of the rate formu-
lae. The user-defined kinetics provides important addi-
tional features not available in the precompiled
formulae, such as
• freely editable algebraic formulae of arbitrary

structure and complexity;
• combination with conditional if-else statements;
• use of porosity (or solid-volume fraction) parame-
ters in the rate expressions;
• use of the saturation variable for unsaturated prob-
lems in the rate expressions;
• use of the temperature variable for nonisothermal
multi-species transport in the rate expressions.
Rate expressions stemming from user-defined kinetics

requires a code interpreter at run time. FEFLOW pro-
vides such a interpreter with a high computational
speed. Practically, when comparing precompiled and
user-defined rate expressions for a common number of Figure 1.6 FEFLOW’s Reaction Kinetics Editor for pre-
reaction terms of the formulae, comparable simulation compiled rate expression of degradation-type kinetics.
times have been found. Consequently, the user-defined
kinetics should be the preferred option providing a In mathematical terms it represents a reduced variant of
maximum of flexibility at a high computational speed. the general rate expression (1-26) in the following
form:
OO=ö=tÜáíÉ=m~éÉêë=J=sçäK=fs
8 α ∈ ( f, s ) more than eight terms are needed the user-defined reac- báíÜÉê= áå= íÜÉ= ê~íÉ
α ni Åçåëí~åíë= ki= çê= áå
tion kinetics (see Section 1.5.2 below) must be used.
Rk = ∑ k i Rate m ( C m ) k = 1, …, N
m ∈ ( 1, …, N )
(1-48)
íÜÉ=ê~íÉ=ÅçÉÑÑáÅáÉåíë
i=1
Ratem= ~= Ñ~Åíçê= Ñçê= éçêçëáíó
NKRKNKO ^êêÜÉåáìë=íóéÉ=âáåÉíáÅë Eçê= êÉí~êÇ~íáçåF= ãìëí= ÄÉ
where k i are rate constants, Rate m are rate coefficients, áåÅäìÇÉÇI=ÉKÖK
α
C m are the concentrations of species m associated with In contrast to degradation-type kinetics we classify
phase α ∈ ( f, s ) , and n i are exponents of the concentra- a reaction as an Arrhenius type if products of different =========== k i = εℜ i k i
tions, where n i ≥ 0 is commonly assumed. The rate species concentration are to be considered. The pre-
expression is related to the selected species ID = k (in compiled rate expression provides for a maximum of ïÜÉêÉ= k i = áë= íÜÉ= áåíêáåëáÅ
the example of Fig. 1.6 the ID is 2 and the rate expres- four reaction terms as shown in Fig. 1.7. The meaning ê~íÉ=Åçåëí~åíK=pìÅÜ=íóéÉ=çÑ
sion is formulated for R 2 ). The indicator k is the ID set ãçÇáÑáÅ~íáçå= áë= åçí= åÉÅÉëJ
of the colored variables and coefficients is identical to
ë~êó= áÑ= éêÉÑÉêêáåÖ= íÜÉ
by the species selector (cf. Fig. 1.3). In the symbolism that described in Section 1.5.1.1.
cbj^qeba= âáåÉíáÅë= ÉÇáíçê
of the formula five different types appear: ~ë= ÇáëÅìëëÉÇ= áå= pÉÅíáçå
NKRKO= ïÜáÅÜ= éêçîáÇÉë= ~ÇÇáJ
• green variables (Rate1, Rate2, .., RateN): These íáçå~ä= î~êá~ÄäÉë= ëìÅÜ= ~ë
rate coefficients are related to the mass material éçêçëáíóI= ëçêéíáçå= ~åÇ= çíÜJ
distributions prescribed in the material editor (see Éêë= íç= Ñçêãìä~íÉ= íÜÉ= êÉ~ÅJ
Fig. 1.5) for each species ID. They can be treated íáçå= ê~íÉ= R k = áå= ~= ÖÉåÉê~ä
also as spatially variable parameters. ï~óK
• blue variables (C1, C2, ..., CN): These are the
internal system variables of concentrations of the kçíÉ= ÑìêíÜÉê= íÜ~í= éêÉÅçãJ
related species. éáäÉÇ= ê~íÉ= ÉñéêÉëëáçåë= ~êÉ
Åçããçåäó= áå~ééäáÅ~ÄäÉ= Ñçê
• black coefficients and exponents (k1, k2, ..., kN;
ìåë~íìê~íÉÇ= ÅçåÇáíáçåë
n1, n2, ..., nN): These are editable variables to ÄÉÅ~ìëÉ= íÜÉ= ÇÉéÉåÇÉåÅó= çÑ
assign rate factor and rate signs as well as concen- R k = çå= íÜÉ= ë~íìê~íáçå= s f = áë
tration exponents, respectively. They are input in åçí=ÉãÄçÇáÉÇK=cçê=ìåë~íìêJ
the dialog. ~íÉÇ= éêçÄäÉãë= íÜÉ
• red indices m ∈ ( 1, …, N ) : They are selectable in cbj^qeba=ÉÇáíçê=ëÜçìäÇ=ÄÉ
the dialog and relate the concentration and accom- ìëÉÇ=ïÜÉêÉ= R k =Å~å=~äëç=ÄÉ
panying variables to a species ID. ÉñéêÉëëÉÇ= Äó= íÜÉ= ë~íìê~J
íáçå=î~êá~ÄäÉK
The precompiled degradation formula is prepared
for eight terms. Commonly, this is sufficient for a wide qÜÉ=ë~ãÉ=áë=íêìÉ=Ñçê=åçåáëçJ
Figure 1.7 FEFLOW’s Reaction Kinetics Editor for pre- íÜÉêã~ä= éêçÄäÉãëK= rëáåÖ
class of problems (e.g. typical in sequential decay compiled rate expression of Arrhenius-type kinetics. cbj^qeba=íÜÉ=íÉãéÉê~íìêÉ
chain reactions). Reaction terms can be suppressed by Å~å=ÄÉ=áåÅçêéçê~íÉÇ=áå= R k K
specifying zero rate factors in the dialog ( k i = 0 ). If
cbcilt=ö=OP
The used Arrhenius reaction type represents a reduced If the number of terms is not sufficient or the structure
variant of the general form of (1-28) according to: of the provided algebraic formula is unsuitable the user
should turn to user-defined reaction kinetics (see Sec-
4 i α ∈ ( f, s ) tion 1.5.2 below), where more flexibility in the formu-
α n1 α nj
Rk = ∑ ki Ratem ( Cm ) ∏ ( Cm ) k = 1, …, N
m ∈ ( 1, …, N )
(1-49) lation of the Monod type expression is available.
i=1 j=2
where k i are rate constants, Rate m are rate coefficients,

α
C m are the concentrations of species m associated with
phase α ∈ ( f, s ) , and n j are exponents of the concentra-
tions, where n j ≥ 0 is commonly assumed.
NKRKNKP jçåçÇ=íóéÉ=âáåÉíáÅë
For the precompiled Monod type kinetics a reduced
variant of the general rate law (1-31) has been imple-
mented based on the following expression
n
3 i ⎛ b ( Cα ) j + 1 ⎞
α n1
∏ ⎜⎜ ---------------------------------------
-⎟
ij m
Rk = ∑ k i Rate m ( C m )
α n j + 1⎟
(1-50)
i=1 j = 1 ⎝ d ij + e ij ( C m ) ⎠
α ∈ ( f, s )
k = 1, …, N
m ∈ ( 1, …, N )
Figure 1.8 FEFLOW’s Reaction Kinetics Editor for pre-

where k i, b ij, d ij, e ij are rate constants, Rate m are rate compiled rate expression of Monod-type kinetics.
α
coefficients, Cm are the concentrations of species m
associated with phase α ∈ ( f, s ) , and n j are exponents
of the concentrations, where n j ≥ 0 is commonly
assumed. NKRKO rëÉêJÇÉÑáåÉÇ= ê~íÉ= ÉñéêÉëëáçåëW
cbj^qeba=ÉÇáíçê
This specific Monod reaction formula (1-50) can be
User-defined type kinetics are accessed via the
edited by the Reaction Kinetics Editor as shown in Fig.
powerful FEMATHED graphical formula editor that is
1.8. It consists of three terms for which the colored
used to freely design rate expressions and algebraic
variables, coefficients and species indicators can be set.
OQ=ö=tÜáíÉ=m~éÉêë=J=sçäK=fs
relationships for reactions. It can fully replace the pre- variables, coefficients or constants appear in red if
compiled formulae of Section 1.5.1. Moreover, reac- syntactically erroneous or incompletely assigned.
tion expressions different from the precompiled types
and more complex formulae become easy to formulate The major elements of FEMATHED are illustrated
and apply. in Fig. 1.9. The editor supports standard word-process-
ing shortcuts and key accelerations such as Control-C
The FEMATHED user-defined kinetics editor for copy, Control-X for cut or Control-V for paste of
appears as the dialog shown in Fig. 1.9. Variables and expression parts or characters. All input may be entered
coefficients are identified by different colors: by keyboard. For instance, the concentration variable
C 3 for species 3 can be selected from the list of blue
• green variables (Rate1, Rate2, .., RateN; Porosity1, variables available on the toolbar. Alternatively, enter-
Porosity2, ..., PorosityN or SolidFrac1, SolidFrac2, ing the key sequence C Down 3 automatically retrieves
..., SolidFracN; and Sorption1, Sorption2, ..., Sorp- and sets the variable. Any number of auxiliary vari-
tionN): They represent rate coefficients, values of ables and parentheses can be used to modify the for-
porosity (or solid fraction for solid phase species), mula. Moreover, if-else statements can be introduced in
and sorption coefficients ( κ for Henry, k 1 for Fre- the formula (cf. Fig. 1.9). Arithmetic and logical opera-
undlich, and b 1 for Langmuir adsorption iso- tions can be used and combined in a practically arbi-
therms) that are related to the mass material data trary manner. The edited rate formulae for the species
input in the material editor (see Fig. 1.5) for each are automatically stored in the fem-dataset of the prob-
species ID (= k). The green variables can be lem. A separate file-access interface exists directly in
selected from a list in the editor or by a direct key the FEMATHED editor to save and load the formula
input. They can be spatially variable. expression in XML format. This can be useful to
• blue variables (C1, C2, ..., CN and additionally exchange formulae for different species equations or to
Saturation for unsaturated problems as well as save variants of rate expressions in a separate database.
Temperature for thermohaline transport): These
are the internal system variables of concentrations An additional feature of FEMATHED is the testing
of the related species, the saturation and tempera- of formulae by activating the built-in code interpreter.
ture, which can be freely used in the formula Pressing the right mouse button invokes a context
design of the rate expression. The blue variables menu that includes the option Tools for experienced
can be selected from a list in the toolbar of the edi- users. Using this option the edited formula can be
tor (Fig. 1.9). tested by entering numerical values for the blue and
• Other coefficients, constants and exponents can be green variables and activating the interpreter (see Fig.
arbitrarily introduced in the formulae by using the 1.10). Note that this option is intended for testing the
editor. Once completely defined they appear in built-in interpreter outside a simulation run. During the
black color in the formula expression, otherwise problem simulation the blue and green variables are
internally generated when solving the species mass
cbcilt=ö=OR
transport equations and remain hidden from the user. symbolical correctness and completeness of the edited
rate formulae. If any check fails the simulation is
During the start-up phase of a simulation with user- rejected, the user is alerted and FEMATHED must be
defined reaction expressions the built-in code inter- used to remove the error or conflict in the formulae.
preter performs a number of checks with regard to the
List of blue variables

(species concentrations,
saturation, temperature)
selectable
List of green variables

(rate coefficients, porosity)
Expression input/export
selectable
as XML string
Arithmetic operations
selectable Logical operations
selectable
Conditional operations
(if-else statements)
selectable
Example of a rate
using a conditional Monod
type relationship defined Editor window scrollable
by variable r and a stoichio-
metric yield coefficient Y
Figure 1.9 FEFLOW’s Reaction Kinetics Editor FEMATHED with user-defined kinetics.
OS=ö=tÜáíÉ=m~éÉêë=J=sçäK=fs
NKS=j~ëë=j~íÉêá~ä=a~í~=çÑ=jìäíáJpéÉÅáÉë=qê~åëéçêí
NKS j~ëë= j~íÉêá~ä= a~í~= çÑ

jìäíáJpéÉÅáÉë=qê~åëéçêí
The following tables relate the mass material data
available in FEFLOW in the context of multi-species
transport conditions to the theoretical framework
developed above. The species k corresponds to the ID
shown by the species selector. The material data differ
depending on the phase association of the selected spe-
cies k.
select specific
option
activate code
interpreter
Figure 1.10 FEMATHED editor with testing the built-in

code interpreter.
Table 1.1 Material conditions for 3D, vertical or axisymmetric mass transport of
species k at fluid phase f
Item Symbol Unit Default Equations
Porosity ε 1 0.3 1-9, 1-18
Henry sorption:
Sorption coefficient κ 1 0 1-17
Freundlich sorption:
1 – b2
Sorption coefficient b1 ( mg ⁄ l ) 0 1-17
Sorption exponent b2 1 0 1-17
Langmuir sorption:
cbcilt=ö=OT
species k at fluid phase f (continued)
Sorption coefficient (numerator) k1 1 0 1-17
Sorption coefficient (denominator) k2 l ⁄ mg 0 1-17

–9 2 –1
Molecular diffusion Dd 10 m s 1 1-19
k
Longitudinal dispersivity βL m 5 1-19
Transverse dispersivity βT m 0.5 1-19

–4 –1
Rate Rate k 10 s 0 1-48, 1-49, 1-50
Σ Rk –3 –1
gm d 0 1-23, 1-48, 1-49,
(Reaction kinetics editor) 1-50 and others
–3 –1
Source(+)/sink(-) Qk gm d 0 1-15, 1-18
In-transfer rate in –1 0 see FEFLOW Refer-
ΦC md
k ence Manual5
Out-transfer rate out –1 0 see FEFLOW Refer-

ΦC md
k ence Manual5
Table 1.2 Material conditions for 2D horizontal mass transport of species k at

fluid phase f
Aquifer thickness B m 1.0 1-22

(only relevant confined conditions)
Porosity ε 1 0.3 1-9, 1-18
Henry sorption:
Sorption coefficient κ 1 0 1-17
Freundlich sorption:
OU=ö=tÜáíÉ=m~éÉêë=J=sçäK=fs
NKS=j~ëë=j~íÉêá~ä=a~í~=çÑ=jìäíáJpéÉÅáÉë=qê~åëéçêí

fluid phase f (continued)
Sorption coefficient b1 1 – b2 0 1-17

( mg ⁄ l )
Sorption exponent b2 1 0 1-17
Langmuir sorption:
Sorption coefficient (numerator) k1 1 0 1-17
Sorption coefficient (denominator) k2 l ⁄ mg 0 1-17
Molecular diffusion Dd –9 2 –1 1 1-19

k 10 m s
Longitudinal dispersivity βL m 5 1-19
Transverse dispersivity βT m 0.5 1-19

–4 –1
Rate Rate k 10 s 0 1-48, 1-49, 1-50
Σ Rk gm d
–3 –1 0 1-23, 1-48, 1-49,
–2 –1
)
In-transfer rate in – 1 a) 0 see FEFLOW Refer-

ΦC md
ence Manual5
Out-transfer rate out – 1 b) 0 see FEFLOW Refer-

ΦC md
ence Manual5
a) For confined conditions and for unconfined conditions at simultaneous application of integral boundary conditions of 3rd
2 –1
kind the unit for the ’In-transfer rate’ is: m d
b) For confined conditions and for unconfined conditions at simultaneous application of integral boundary conditions of 3rd
2 –1
kind the unit for the ’Out-transfer rate’ is: m d
cbcilt=ö=OV
species k at solid phase s
Solid volume fraction εs 1 0.3 1-9, 1-18

–4 –1
Rate Rate k 10 s 0 1-48, 1-49, 1-50
Σ Rk –3 –1
gm d 0 1-23, 1-48, 1-49,
s –3 –1
In-transfer rate in –1 0 see FEFLOW Refer-

ΦC md
k ence Manual5
out –1
Out-transfer rate ΦC md 0 see FEFLOW Refer-
k ence Manual5

solid phase s
Aquifer thickness B m 1.0 1-22

(only relevant to confined conditions)
Solid volume fraction εs 1 0.3 1-9, 1-18

–4 –1
Rate Rate k 10 s 0 1-48, 1-49, 1-50
Σ Rk –3 –1
gm d 0 1-23, 1-48, 1-49,
s –2 –1
)

in – 1 a)
In-transfer rate ΦC md 0 see FEFLOW Refer-
ence Manual5
Out-transfer rate out – 1 b) 0 see FEFLOW Refer-

ΦC md ence Manual5
PM=ö=tÜáíÉ=m~éÉêë=J=sçäK=fs
NKT=kçíÉ=íç=s~êá~ÄäÉJaÉåëáíó=cäçï=~åÇ=jìäíáJpéÉÅáÉë=qê~åëéçêí
a) For confined conditions and for unconfined conditions at simultaneous application of integral boundary conditions of 3rd
2 –1
kind the unit for the ’In-transfer rate’ is: m d
b) For confined conditions and for unconfined conditions at simultaneous application of integral boundary conditions of 3rd
2 –1
kind the unit for the ’Out-transfer rate’ is: m d
NKT kçíÉ= íç= s~êá~ÄäÉJaÉåëáíó The EOS (1-52) is developed as follows (cf.1,5,10)
cäçï= ~åÇ= jìäíáJpéÉÅáÉë
qê~åëéçêí ∂ρ
f
∂ρ
f
∂ρ
f
∑ --------
f f f
dρ = --------f dp + - dC k + -------- dT (1-54)
∂p ∂C
f ∂T f f
f
C k, T k k p , Ck
NKTKN jìäíáJëéÉÅáÉë=ÇÉåëáíó=êÉä~íáçå
f
p,T
f ⎛ 1 ∂ρ f⎞ ⎛ 1 ∂ρ f ⎞ ⎛ 1 ∂ρ f⎞
The fluid density ρ is composed of a number of
∑
f f f f f
= ⎜ ----f --------f⎟ ρ dp + ⎜ ----f ---------f ⎟ ρ dC k + ⎜ ----f --------⎟ ρ dT
miscible chemical species k with a partial fluid density ⎝ ρ ∂p ⎠ k
⎝ ρ ∂C k⎠ ⎝ ρ ∂T ⎠
f
⎧
⎪
⎨
⎪
⎩
⎧
⎪
⎨
⎪
⎩
⎧
⎪
⎨
⎪
⎩
ρ k (mass of species k per unit volume of fluid), so that γ αk –β
f f where γ is the fluid compressibility, and α k and β are

ρ = ∑ ρk (1-51)
the volumetric species and thermal expansion coeffi-
k
cient, respectively. If we assume that γ , α k and β are
f constant, the integration of (1-54) immediately leads to
holds for the mixture10, where ρ k stands for all species the EOS for the density in the common form1
f
present in the fluid. Moreover, the fluid density ρ can
be influenced by the temperature T (e.g., density
∑ α (C
f f f f
γ ( p – p0 ) + – C k0 ) – β ( T – T 0 )
decreases when temperature increases) and by the fluid k k
f f f k
pressure p (density increases when pressure increases ρ = ρ0 e (1-55)
f
due to compressibility). In a formal manner, ρ is to be
regarded as a dependent thermodynamic variable for where suitable reference values appear for the density
which a constitutive relationship, or equation of state f f f
(EOS), holds ρ0 = ρ f f at reference pressure p 0 , reference con-
p 0, C k0, T 0
f
f f f
ρ = ρ ( p , ρ k, T )
f
(1-52)
centration C k0 and reference temperature T 0 .
f The EOS for the fluid density (1-55) is often lin-

The partial fluid density ρ k appears here as species-
early approximated by
related density, which is nothing more than the mass
f
concentration Ck
f f
Ck ≡ ρk (1-53)
cbcilt=ö=PN
f f f f f f NKTKP jìäíáéäÉ= ëéÉÅáÉëJÇÉéÉåÇÉåí

ρ = ρ 0 1 + γ ( p – p 0 ) + ∑ α k ( C k – C k0 ) (1-56)
ÇÉåëáíó=ê~íáç
k
In FEFLOW, the dimensionless fluid density differ-
– β ( T – T0 )
ence ratio5 α is used, which is defined for a multi-spe-
cies problem as
where γ , α k and β are considered constant1,10. f f

α k = α k ( C ks – C k0 ) (1-57)
f f
NKTKO léíáçå=ëÉííáåÖë where Cks is the maximum concentration and Ck0 is
the reference concentration of species k. The species-
To handle variable-density problems for multi-spe- related maximum and reference concentrations are to
cies transport FEFLOW differs two principal options: be input in the specific reference dialog as seen in Fig.
1.12. This has to be done for each species k.
• Multiple species-dependent density ratio
• Constant solutal density ratio (default)
which will be explained further below. These options

can be set in the Specific option settings dialog as
shown in Fig. 1.11.
(a)
Figure 1.12 Input of maximum and reference concentra-

tion for selected species k.
(b)
With the multiple species-dependent density ratio
α k the following relationship for the fluid density
expansion in multi-species applications results and is
implemented in FEFLOW, viz.,
Figure 1.11 Setting of (a) multiple and (b) con-

stant density ratio.
PO=ö=tÜáíÉ=m~éÉêë=J=sçäK=fs
NKT=kçíÉ=íç=s~êá~ÄäÉJaÉåëáíó=cäçï=~åÇ=jìäíáJpéÉÅáÉë=qê~åëéçêí
αk the fluid density with respect to a concentration

f f f f f f
ρ = ρ 0 1 + γ ( h – h 0 ) + ∑ ---------------------------
f f
- ( C k – C k0 ) (1-58) increase is the same for all species k, so that
k ( C ks – C k0 )
α ≈ αk = α1 = α2 = … = αN ( k = 1, …, N ) (1-59)
– β ( T – T0 )
where α represents an overall constant solutal density

Note that the hydraulic head h is used in (1-58) for ratio. In FEFLOW, the constant α is defined as5
f
which a specific fluid compressibility γ = γρ 0 g is
f f
given (g = gravitational acceleration). α = α ( Cs – Co )
(1-60)
f f f f
α = [ ρ ( Cs ) – ρ0 ] ⁄ ρ0
If the multiple density ratio option is set the
required multiple density ratio α k has to be input for with
each species k in the flow material menu. The species
f f f f
selector appears and the multiple density ratio α k is ρ ( C s ) ≈ ρ 0 + ωC s (1-61)
related to the selected species as shown in Fig. 1.13.
where ω is a constant density factor, often approxi-
f
mated by a value of 0.7 for water, and C s is the overall
maximum concentration given by
f max
Cs = ∑ Ck (1-62)
k
max
where C k is the maximum concentration of species
f
k. In (1-60), Co represents an overall reference concen-
f
tration that corresponds to the reference density ρ 0 . It
may be determined by
f f
C o = min( C k ) (1-63)
Figure 1.13 Input of multiple density ratio αk k
for selected species k.
Using (1-60) and (1-59) in (1-58), the following
modified relationship for the fluid density expansion
NKTKQ `çåëí~åí=ëçäìí~ä=ÇÉåëáíó=ê~íáç results and is applied in FEFLOW for the constant den-
Frequently, it can be assumed that the expansion of sity ratio option
cbcilt=ö=PP
f f f f α - f f
f ∑
ρ = ρ 0 1 + γ ( h – h 0 ) + ----------------------
f
( Ck – C0 ) (1-64) f
where the overall fluid concentration C is defined as
( Cs – Co ) k
f f
– β ( T – T0 ) C = ∑ Ck (1-67)
k
f f with summation over all dissolved species k in the fluid

where α, C s, C 0, γ, β, T 0 represent input parameters to
be set in the material data menu and the reference data phase. The empirical relationship (1-66) is given in
dialog. detail in FEFLOW’s Reference manual5 (cf. Fig. 1.14a)
(a)
NKTKR sáëÅçëáíó= êÉä~íáçåë= Ñçê= ãìäíáJ
ëéÉÅáÉë=íê~åëéçêí=éêçÄäÉãë
Generally, in multi-species transport problems the
constitutive relationship for the dynamic viscosity of
f
fluid μ is developed as a function of the species con-
f
centration Ck and the temperature T , viz.,
f f f (b)
μ = μ ( C k, T ) (1-65)
FEFLOW provides in the Specific option settings dia-

log two options to specify the empirical relationship for
f f
μ in dependence on the variables C k and T :
• Standard
• User-defined
Figure 1.14 Setting of (a) standard and (b) user-defined
viscosity relationship.
as exhibited in Fig. 1.14.
f
In the standard case (Fig. 1.14a) μ is developed as Alternatively, in the case of a user-defined fluid vis-
f
a function of the overall fluid concentration C and the cosity the FEMATHED viscosity editor is available
f
temperature T according to (see Fig. 1.14b) to specify arbitrary formulae for μ of
(1-65) in dependence on each species concentrations
f
f f f C k and temperature T without restrictions. For exam-
μ = μ (C , T) (1-66)
PQ=ö=tÜáíÉ=m~éÉêë=J=sçäK=fs
NKU=bñ~ãéäÉë
ple, the fluid viscosity should be developed for two dimensional (1D) multi-species transport problems
species according to with serial and parallel reaction kinetics. As an exam-
ple, the following reaction network is considered (Fig.
f 1.16) consisting of five species (N = 5). The species B
f f ∂μ f f
μ = μ 0 + ∑ --------f- ( C k – C 0 ) (1-68)
has three daughter species C1, C2, C3.
k ∂C k
f f
where ∂μ ⁄ ∂Ck are constant coefficients of viscosity C1
f
variability for each species, and μ 0 is the reference
fluid viscosity at the reference species concentration C2
f A B
C 0 . With the FEMATHED viscosity editor such a user-
defined relation can simply input as displayed in Fig.
1.15. C3
Figure 1.16 Serial-parallel reaction network by Sun et al.2
The reaction network of Fig. 1.16 can be decom-

posed into three serial reaction chains: A → B → C1 ,
A → B → C 2 and A → B → C 3 . Accordingly, the fol-
lowing system of transport equations is considered:
2
∂C A ∂ C ∂C A ⎫
---------- – D ------------A- + v ---------- = –kA CA ⎪
∂t ∂x
2 ∂x ⎪
⎪
Figure 1.15 Example of a two-species viscosity relation- ∂C B ∂ C
2
∂C ⎪
ship by using the FEMATHED viscosity editor. ---------- – D ------------B- + v ---------B- = yB kA CA – kB CB ⎪
∂t ∂x
2 ∂x ⎪
⎪
2 ⎪
∂C C ∂ CC ∂C C ⎪
NKU bñ~ãéäÉë ------------1 – D ---------------1 + v ------------1 =
∂t ∂x
2 ∂x
yC kB CB – kC CC ⎬
1 1 1
⎪
(1-69)
⎪
NKUKN pÉêá~äJé~ê~ääÉä= êÉ~Åíáçåë= ~åÇ ∂C C
2
∂ CC ∂C C ⎪
------------2 – D ---------------2 + v ------------2 = y C k B C B – k C C C ⎪⎪
Åçãé~êáëçå= íç= íÜÉ= ~å~äóíáÅ~ä ∂t ∂x
2 ∂x 2 2 2
⎪
ëçäìíáçå=Äó=pìå=Éí=~äK 2 ⎪
∂C C ∂ CC ∂C C ⎪
NKUKNKN pí~íÉãÉåí=çÑ=íÜÉ=Na=éêçÄäÉã ------------3 – D ---------------3 + v ------------3 = yC kB CB – kC CC ⎪
∂t ∂x
2 ∂x 3 3 3⎭
Sun et al.20 present analytical solutions for one-
cbcilt=ö=PR
The right-hand sides of (1-69) represent the reaction C A ( 0, t ) = 1.0 t>0

rates R A, R B, R C1, R C2, R C3 . In (1-69) D = αv is a con-
C k ( 0, t ) = 0 ( k = B, C 1, C 2, C 3 ) t > 0 (1-70b)
stant dispersion coefficient ( α represents dispersivity),
x is the one-dimensional coordinate, v is a constant C k ( ∞, t ) = 0 ( k = A, B, C 1, C 2, C 3 ) t>0
pore velocity, k k ( k = A, B, C1, C2, C3 ) are first-order
rate constants, and y k are corresponding stoichiometric NKUKNKO ^å~äóíáÅ~ä=ëçäìíáçå
yields. All species are diluted chemicals in a mobile
fluid phase. The transport parameters assumed for this The equation system (1-69) can be written for spe-
problem are shown in Tab. 1.5. There is no need to cies k
specify the porosity ε .
L ( Ck ) = yk kk – 1 Ck – 1 – kk Ck (1-71)
Table 1.5 Problem parameters used
where L ( . ) represents the differential operator. Sun et
Parameter Symbol Value
al.20 introduce the auxiliary variable a k as
Length of column, m L 40.0
Dispersivity coefficient, m α 10.0 k–1 k–1 yi + 1 ki
Pore velocity, md-1 v 0.4
ak = Ck + ∑ ∏ ---------------- C j
ki – kk
∀k = 2, 3, …, N (1-72)
j=1 i=j
Rate constant of A, d-1 kA 0.2
Rate constant of B, d-1 kB 0.1 Substituting (1-72) in (1-71) yields the reactive trans-
port equations in terms of a k , viz.,
Rate constant of C1, d-1 kC 0.02
1
Rate constant of C2, d-1 kC 0.02 L ( ak ) = –kk ak ∀k = 2, 3, …, N (1-73)

2
Rate constant of C3, d-1 kC 0.02

3
Stoichiometric yield of A → B yB 0.5 Note that for k = 1 the transport equation (1-73) in
terms of the first auxiliary variable is identical to the
Stoichiometric yield of B → C1 yC 0.3
1 original equation (1-71) since a k = Ck . The substituted
Stoichiometric yield of B → C2 yC 0.2 equations (1-73) can be easily solved by the basic ana-
2
Stoichiometric yield of B → C3 yC 0.1 lytical formula

3
The following initial and boundary conditions are a ko x – t v + 4k k D

2
vx
used: a k ( x, t ) = ------- exp ⎛⎝ -------⎞⎠ exp ( – ψ k x )erfc ---------------------------------------
2 2D 2 Dt
2
(1-74)
C k ( x, 0 ) = 0 ( k = A, B , C 1 , C 2 , C 3 ) x ≥ 0 (1-70a) x + t v + 4k k D
+ exp ( ψ k x )erfc ---------------------------------------
2 Dt
PS=ö=tÜáíÉ=m~éÉêë=J=sçäK=fs
NKU=bñ~ãéäÉë
where quadrilateral bi-linear finite elements. The AB/TR pre-

dictor-corrector adaptive time-marching scheme is
2
v - kk applied, where the RMS error tolerance (1-45) corre-
ψk = --------- + ----
2 D
–4
sponds to δ = 10 . The initial time step is
4D –3
∞ (1-75) Δt o = 10 d.
2 2
erfc ( ξ ) = 1 – erf ( ξ ) = ------- ∫ exp ( – τ ) dτ
π The reaction kinetics for the present problem is of a
ξ
degradation type. We employ both the precompiled rate
expressions and the user-defined rate formulae by
and a ko is the initial condition in terms of the auxiliary using FEMATHED, where the rates R k are specified
variable and ξ is a dummy variable. simply as follows:
The solutions of all concentrations Ck in the real R 1 = – Rate 1 ⋅ C 1 ⎫

untransformed domain can be determined by a succes- ⎪
R 2 = 0.5 ⋅ Rate 1 ⋅ C 1 – Rate 2 ⋅ C 2 ⎪
sive substitution process using (1-72) in a reverse way ⎪
R 3 = 0.3 ⋅ Rate 2 ⋅ C 2 – Rate 3 ⋅ C 3 ⎬ (1-78)
⎪
k–1 k–1 R 4 = 0.2 ⋅ Rate 2 ⋅ C 2 – Rate 4 ⋅ C 4 ⎪
yi + 1 ki
⎪
Ck = ak – ∑ ∏ ---------------- C j
ki – kk
∀k = 2, 3, …, N (1-76) R 5 = 0.1 ⋅ Rate 2 ⋅ C 2 – Rate 5 ⋅ C 5 ⎭
j=1 i=j
In (1-78) the parameters Rate k represent the reaction

where a k is the solution from (1-74).
constants k k of Tab. 1.5. The used species IDs are
linked to the species names and phases as summarized
NKUKNKP kìãÉêáÅ~ä=~å~äóëáë in Tab. 1.6.
Using FEFLOW for the above 5-species transport
problem the column is discretized with a spatial incre- NKUKNKQ oÉëìäíë
ment of Δx = 0.133333 m. Because the infinite bound-
A comparison of the FEFLOW results with the ana-
ary conditions in (1-70b) cannot be applied in a
lytical solutions gives perfect agreements as exhibited
numerical context we extend the column up to 80 m
in Fig. 1.17. The FEFLOW simulation takes 72 time
(instead of L = 40 m) and use a Neumann-type bound-
steps of variable length. Comparing simulation perfor-
ary condition at the outlet boundary section as
mance between the FEMATHED-interpreted and the
∂C k ( 2L, t )
precompiled formula options for this problem we
– D -------------------------- = 0 ( k = A, B, C 1, C 2, C 3 ) t > 0 (1-77) found practically the same CPU times. Note that the
∂x
precompiled degradation formula always executes
eight terms (cf. Fig. 1.6) while the FEMATHED code
Accordingly, the double domain is discretized by 600 interpreter does not need to consider any extra terms.
cbcilt=ö=PT
1.0 0.20
0.8
0.15
exact exact
numerical numerical
Concentration A
Concentration B
0.6
0.10
0.4
0.05
0.2
0.0 0.00
0 10 20 30 40 0 5 10 15 20 25 30 35 40
Distance [m] Distance [m]
0.06 0.04
0.03
0.04
Concentration C1
Concentration C2
0.02
0.02 exact
exact numerical
0.01
numerical
0.00 0.00
0 5 10 15 20 25 30 35 40 0 5 10 15 20 25 30 35 40
Distance [m] Distance [m]
0.020
0.015 Figure 1.17 Concentration profiles for the five

species across the column after t = 40 days of
Concentration C3
reactive transport: Comparison of the exact (ana-

0.010
lytical) solutions with FEFLOW results.
exact
numerical
0.005
0.000
0 5 10 15 20 25 30 35 40
Distance [m]
PU=ö=tÜáíÉ=m~éÉêë=J=sçäK=fs
NKU=bñ~ãéäÉë
Table 1.6 Species IDs for Sun et al.’s problem First-order reaction rates for the sequential reaction
kinetics C 1 → C2 → C 3 → C 4 are given as follows
ID ( = k ) Phase Name
R1 = –k1 C1 ⎫
1 fluid A ⎪
R2 = k1 C1 – k2 C2 ⎪
2 fluid B ⎬ (1-79)
R3 = k2 C2 – k3 C3 ⎪
3 fluid C1 ⎪
R4 = k3 C3 – k4 C4 ⎭
4 fluid C2
5 fluid C3 where k k ( k = 1, 2, 3, 4 ) are rate constants, which are

listed together with the remaining parameters in Tab.
1.7. All species are considered mobile in the fluid
NKUKNKR Pa=ãìäíáJëéÉÅáÉë=íê~åëéçêí=ïáíÜ= phase; porosity ε does not play a role.
ÑáêëíJçêÇÉê=ÇÉÖê~Ç~íáçå=êÉ~Åíáçå=
âáåÉíáÅë Table 1.7 Problem parameters for the 3D problem
Sun et al.21 have extended their analytical approach
to 3D problems for homogeneous parameters and Parameter Symbol Value
steady-state flow regimes. The solutions are demon-
Length, m L 100.0
strated for a four-species transport in a 3D aquifer of
100 m × 41 m × 25 m . For the finite-element analysis Width, m 41.0
the domain is discretized by 440,000 hexahedral tri-lin-
Thickness, m B 25.0
ear elements consisting of 462,825 nodes as shown in
Fig. 1.18. Longitudinal dispersivity, m βL 1.5
Transverse dispersivity, m βT 0.3
Pore velocity, md-1 v 0.2
Rate constant of C1, d-1 k1 0.05
Figure 1.18 Discretized 3D aquifer used in FEFLOW. Boundary and initial conditions are the same as those
cbcilt=ö=PV
used in the 1D example (Section 1.8.1.1). The obtained domain is considered semi-infinite. Furthermore, Sun
results are shown in Figs. 1.19 and 1.20. The concen- et al.21 used an alternative dispersion model, where dif-
tration contours reveal differences between the analyti- ferent transverse dispersivities in the horizontal and
cal and numerical solutions. In the finite-element vertical directions are applied. In the FEFLOW simula-
analysis the aquifer is finite and natural Neumann-type tions the isotropic Bear-Scheidegger dispersion model
boundary conditions (zero concentration gradients) are (1-19) is preferred with only one transverse dispersion
applied at the outer border faces of the discretized parameter (Tab. 1.7).
domain. Unlike, in the analytical solution the aquifer
C1 C2
C3 C4
Figure 1.19 FEFLOW results of the three-dimensional 0.01 isosurface concentration for the four species C1,
C2, C3 and C4 after t = 400 days.
QM=ö=tÜáíÉ=m~éÉêë=J=sçäK=fs
NKU=bñ~ãéäÉë
analytical solution FEFLOW
C1
0.2
0.8
0.4
0.1 0.01
C2
0.4 0.2 0.1 0.05 0.01
C3
0.25 0.2 0.1 0.05 0.01
C4
0.01
0.15 0.1 0.05 0.02
Figure 1.20 Comparison of Sun et al.21’s analytical solution (left) with FEFLOW results (right): Concentra-
tion contours of the four species C1, C2, C3 and C4 in the xy-plane at z = 13 m and t = 400 days.
cbcilt=ö=QN
NKUKO o~íÉJäáãáíÉÇ= ÇÉëçêéíáçå= ~åÇ described using first-order kinetics, where the rate of
ÇÉÅ~óW= `çãé~êáëçå= íç= cêó= Éí mass transfer of contaminant from the solid phase to
the aqueous phase depends on the concentration gradi-
~äKÛë=~å~äóíáÅ~ä=ëçäìíáçå
ent between the two phases and a single rate coeffi-
NKUKOKN pí~íÉãÉåí=çÑ=íÜÉ=Na=éêçÄäÉã cient. The following 1D two-species transport
equations are considered (written in the present nota-
Fry et al.11 studied rate-limited desorption and first-
tion), which is a one-site kinetic model15,25 with linear
order decay on the feasibility of in situ bioremediation
kinetic sorption and decay in the aqueous phase:
of contaminated groundwater by using analytical solu-
tions. The conceptual model is shown in Fig. 1.21.
∂C ∂S ∂C
2
∂ C ⎫
ε ------- + ε s ------ + q ------- – D ---------2 = – εϑC ⎪
∂t ∂t ∂x ∂x ⎪
⎬ (1-80)
∂S s ⎪
ε s ------ = ε s α ( ρ K d C – S ) ⎪
∂t ⎭
or
∂C ∂C ∂ C
2 ⎫
ε ------- + q ------- – D ---------2 = R c ⎪
∂t ∂x ∂x ⎪
⎬ (1-81a)
Figure 1.21 Conceptual model of reacting contaminant ⎪
∂S
transport in groundwater by Fry et al.11. ε s ------ = R s ⎪
∂t ⎭
A remedial pump-and-treat scheme is considered with

assuming conditions of one-dimensional, steady-state
groundwater flow through a homogeneous and isotro- s ⎫
R c = – ( εϑ + ε s ρ αK d )C + ε s αS ⎪
pic aquifer. The modeled portion of the aquifer is ⎬ (1-81b)
s
bounded by injection and extraction wells (see control Rs = εs α ( ρ Kd C – S ) ⎪
volume drawn in Fig. 1.21). ⎭
The study concerns a method of restoration of a where

contaminated aquifer domain, where organic com-
pounds are degraded by indigenous or introduced C = aqueous concentration (at fluid phase);
microorganisms. Degradation of the contaminant is S = sorbed concentration (at solid phase);
represented by a first-order decay, where the rate of D = β L q , hydrodynamic dispersion coefficient;
degradation is a function of the contaminant concentra- q = εv , Darcy velocity;
tion in the aqueous (fluid) phase. Desorption is v = pore velocity;
QO=ö=tÜáíÉ=m~éÉêë=J=sçäK=fs
NKU=bñ~ãéäÉë
t = time;
x = distance; A test case is considered for which the used parame-
ε = porosity; ters are listed in Tab. 1.8.
εs = ( 1 – ε ) , solid volume fraction;
s
ρ = solid density;
= equilibrium distribution coefficient; Table 1.8 Problem parameters used
Kd
α = first-order desorption rate constant;
Parameter Symbol Value
βL = longitudinal dispersivity;
ϑ = first-order decay rate constant; Length of domain, m L 10.0
The aquifer is initially contaminated and concentra- Longitudinal dispersivity, m βL 1.0

tions C and S are uniform throughout the control vol- Darcy velocity, md -1
q 0.04
ume. Furthermore, the sorbed and aqueous phases are
initially in linear equilibrium as described with the dis- Equilibrium distribution coeffi- Kd 0.68
cient, cm3g-1
tribution coefficient Kd . These initial conditions are
s
stated as Solid density, g cm-3 ρ 2.67
C ( x, 0 ) = C o 0≤x≤L Porosity ε 0.4

⎫
S ( x, 0 ) = S o 0 ≤ x ≤ L ⎪⎬ (1-82) Solid volume fraction εs 0.6
⎪
s
So = ρ Kd Co ⎭ Desorption rate constant, d-1 α 0.01
Decay rate constant, d-1 ϑ 0.1

where C o is the aqueous concentration at t = 0 , S o is
the sorbed concentration at t = 0 , and L is the length
of the control volume (Fig. 1.21). NKUKOKO kìãÉêáÅ~ä=~å~äóëáë
Fry et al.’s two-species bioremediation problem as
At the control-volume inlet ( x = 0 ) the contaminant
stated above is simulated by FEFLOW. The 1D domain
flux due to advection and dispersion is zero for all
of length L = 10 m is discretized by 300 quadrilateral
times. At the control-volume exit ( x = L ) the concen-
bi-linear finite elements, i.e., the spatial increment is
trations are uniform with distance. Accordingly, the
Δx = 0.03333 m. The AB/TR predictor-corrector adap-
following boundary conditions hold:
tive time marching scheme is applied, where the maxi-
–4
mum error tolerance (1-46) corresponds to δ = 10 .
∂C
– D ------- ( 0, t ) + qC ( 0, t ) = 0 t > 0⎫ –5
The initial time step is Δt o = 10 d. Additionally, a
∂x ⎪
⎬ (1-83) maximum time step length of 0.5 d and a rate for
∂C
------- ( L, t ) = 0 t > 0⎪ changing the time-step size Ξ of 2 have proven useful
∂x ⎭
cbcilt=ö=QP
for the present problem simulation. Due to the bound- NKUKOKP oÉëìäíë
ary conditions (1-83) the divergence form7 of the gov-
The FEFLOW results for the problem are compared
erning transport equations has to be used, which allows
with the analytical solutions which are presented by
the input of the total (advective plus dispersive) mass
Fry et al.11. As shown in Fig. 1.22 very good agreement
flux at a boundary.
with the analytical results is obtained. The FEFLOW
simulation takes 2039 time steps. With respect to the
The reaction kinetics for the present problem is of a
computational performances we also tested the pre-
degradation type. We prefer the FEMATHED input for
compiled formulae for the degradation-type rate
the rates (1-81b), which are specified as follows (note
against the preferred FEMATHED formulation. Again
that species ID 1 represents the aqueous species with
we found the same CPU times in both simulations,
concentration C ≡ C 1 and species ID 2 represents the
illustrating the power of the FEMATHED code inter-
sorbed species with concentration S ≡ C 2 ):
preter.
R 1 = – ( Porosity 1 ⋅ Rate 1 + Rate 2 ⋅ Rb ⋅ Kd ) ⋅ C 1 (1-84a)
+ Rate 2 ⋅ SolidFrac 2 ⋅ C 2
Rb = SolidFrac2 ⋅ 2.67 Kd = 0.68 NKUKP qïçJëáíÉ= ÉèìáäáÄêáìãLâáåÉíáÅ
ëçêéíáçå= ïáíÜ= ÇÉÖê~Ç~íáçåW
R 2 = SolidFrac2 ⋅ Rate 2 ⋅ ( Rs ⋅ Kd ⋅ C 1 – C 2 ) (1-84b) `çãé~êáëçå=íç=pq^kjla=~å~J
Rs = 2.67 Kd = 0.68 äóíáÅ~ä=ëçäìíáçåë
NKUKPKN mêçÄäÉã=Ñçêãìä~íáçå
The parameters in (1-84a) and (1-84b) are related to
the notation used in (1-81b) as follows: R 1 ≡ R c , The two-site sorption concept presumes that sorp-
s
R 2 ≡ R s , Porosity 1 ≡ ε , Rate 1 ≡ ϑ , Rate 2 ≡ α , Rs ≡ ρ , tion or exchange sites in soils can be classified into two
s
Rb ≡ ε s ρ , Kd ≡ K d , and SolidFrac 2 ≡ ε s . The species fractions: one fraction (Type-1) on which sorption is
IDs are linked to the species names and phases as sum- assumed to be instantaneous, and another fraction
marized in Tab. 1.9. (Type-2) on which sorption is considered to be time-
dependent. The resulting two-site kinetic model inter-
acts with a solid phase composed of such different con-
Table 1.9 Species IDs for Fry et al.’s problem stituents as soil minerals, organic matter and various
oxides. Studies in transport of pesticides indicate that
ID ( = k ) Phase Name the two-site kinetic model may well be suitable25.
1 fluid C
2 solid S
QQ=ö=tÜáíÉ=m~éÉêë=J=sçäK=fs
NKU=bñ~ãéäÉë
a) b)
100
10-1 t = 200 days
10-2
C/Co, S/So
10-3
t = 1000 days
10-4
10-5
0 0.2 0.4 0.6 0.8 1.0
DISTANCE (x/L)
Figure 1.22 Aqueous ( C ⁄ C o , solid lines) and sorbed ( S ⁄ S o , dashed lines) concentrations versus distance x at times
t = 200 days and t = 1000 days: a) Fry et al.’s11 analytical solution (pore volumes = tv ⁄ L ), b) FEFLOW results.
The derivation proceeds in the same fashion as for By using the equilibrium sorption (1-85) the Type-1
the one-site sorption model. We introduce two different concentration S 1 can be eliminated (expressed by C )
sorbed concentrations S 1 and S 2 , where the first is for from the 3-species basic equations and only 2 species
Type-1 at equilibrium sites and the second is for Type-2 (namely C and S 2 ) have to be solved. Assuming a lin-
at kinetic sites. Because Type-1 sites are always at ear degradation for all species C , S 1 and S 2 , as well as
equilibrium, sorption onto these sites is given by an a Henry-type sorption for S 1 , we found the following
adsorption function similar to Eq. (1-13), viz., 2-species model equations for a two-site kinetic
sorption22,25 with degradation written in the present
S 1 = fχC (1-85) notation:
∂C
where C is the aqueous concentration at fluid phase, f ε f ℜ m ------- + q ⋅ ∇C – ∇ ⋅ ( D ⋅ ∇C ) = R c ⎫
∂t ⎪
is the fraction of exchange sites assumed to be at equi- ⎬ (1-87a)
∂S 2
librium and χ is a sorption function. The kinetic part ε s -------- = R s ⎪⎭
S 2 is subjected to a kinetic relationship in a form ∂t
S 2 → ( 1 – f )χC (1-86) with
cbcilt=ö=QR
μf = decay coefficient of diluted species C ;

(1 – f) ⎫
R c = – αε s --------------- κ + ε s κμ s + ε f μ f C + αε s S 2 ⎪
f 1 ⎪ Note that the two-site adsorption model (1-87b)
⎬ (1-87b)
(--------------
1 – f )- ⎪ reduces to the one-site fully kinetic adsorption model
R s = αε s κC – ε s ( α + μ s )S 2 ⎪
f 2
comparable to (1-81b) if f → 0 , where the ( 1 – f )κ ⁄ f
⎭ s
terms in (1-87b) have to be replaced by ( 1 – f )ρ Kd .
and8
NKUKPKO pí~íÉãÉåí=çÑ=~=Na=íÉëí=Å~ëÉ
1–ε
ℜ m = 1 + ----------- κ ⎫⎪ We solve the above two-site kinetic sorption equa-
ε (1-87c)
⎬ tions for a 1D domain (column) of length L , for which
s
κ = f ρ Kd ⎪
⎭ analytical solutions are available18,22,25. To compare to
where analytical solutions the following dimensionless
parameters are to be defined:
C = aqueous concentration (at fluid phase f);
ℜm 1–ε κ
S1 = Type-1 sorbed concentration (at solid β = -------- ℜ = 1 + ⎛ -----------⎞ ---
ℜ ⎝ ε ⎠f
phase s); (1-88)
S2 = Type-2 sorbed concentration (at solid L qL
ω = α ( 1 – β )ℜ --- P = ---------
phase s); v D
D = hydrodynamic dispersion;
q = ε f v , Darcy velocity; where v = v is the constant 1D pore velocity. With
v = pore velocity; given parameters ℜ , β and ω the model parameters
t = time; κ , f and α can be specified as
εf = s f ε , fluid volume fraction;
εs = ( 1 – ε ) , solid volume fraction; εβ ( ℜ – 1 ) – ( 1 – β )ε ⎫
ε = porosity; κ = --------------------------------------------------- ⎪
1–ε ⎪
sf = saturation; ⎪
κ ( 1 – ε )
ρ
s
= solid density; f = --------------------
- ⎪
( ℜ – 1 )ε ⎬ (1-89)
Kd = equilibrium distribution coefficient; ⎪
ω ⎪
f = fraction of exchange sites; α = --------------------------
v ⎪
ℜm = retardation factor; ( 1 – β )ℜ --- ⎪
L ⎭
α = first-order kinetic rate coefficient;
s
κ = f ρ K d , Henry adsorption coefficient;
μs = decay coefficient of sorbed species S 1 ; Note that α is only defined if β < 1 . Equations (1-87a)
1
μs = decay coefficient of sorbed species S 2 ; with (1-87b) are solved for an initially solute-free col-
2
QS=ö=tÜáíÉ=m~éÉêë=J=sçäK=fs
NKU=bñ~ãéäÉë
umn subject to a pulse-type input boundary condition. In accordance with (1-93) and in using (1-89) the com-
The initial and boundary conditions are stated as plete dataset is listed in Tab. 1.10.
C ( x, 0 ) = 0 0≤x≤L⎫
(1-90) Table 1.10 Problem parameters used
S 2 ( x, 0 ) = 0 0 ≤ x ≤ L ⎬⎭
and Parameter Symbol Value
Length of domain, m L 10.0

∂C ⎧ qC o 0 < t ≤ to
– D ------- ( 0, t ) + qC ( 0, t ) = ⎨ Pulse duration, d to 300
∂x ⎩0 t ≥ to
(1-91)
Input concentration, mg l-1 Co 1.0
∂C-
------ ( L, t ) = 0 t>0
∂x Longitudinal dispersivity, m βL 2.128
-1
Pore velocity, md v 0.1
where C o is the input concentration, D = D = β L q
is the dispersion coefficient, q = ε f v = ε f v is the Darcy velocity, md-1 q 0.04
Darcy flux and t o is the time duration of the applied
Porosity ε 0.4
solute pulse.
Solid volume fraction εs 0.6
We consider the 1D column for steady-state flow
Henry coefficient κ 0.16667
( q = const ) and saturated conditions ( ε f = ε ). Further-
more, we assume that all decay coefficients are the Fraction of exchange site f 0.16667
same, i.e., μ f = μ s1 = μ s2 = μ . Accordingly, a dimen-
Kinetic rate coefficient, d-1 α 0.004
sionless decay parameter ψ is defined as
Decay rate coefficients, d-1 μ 0; 0.001; 0.003;
0.006; 0.01
ψ = μL ⁄ v (1-92)
The test case is considered for the dimensionless

NKUKPKP kìãÉêáÅ~ä=~å~äóëáë
parameters as
The two-site equilibrium/kinetic problem as stated
β = 0.5 above is simulated by FEFLOW. The 1D domain of
⎫
ℜ = 2.5 ⎪ length L = 10 m is discretized by 300 quadrilateral bi-
⎪ linear finite elements, i.e., the spatial increment is
ω = 0.5 ⎬ (1-93)
⎪ Δx = 0.03333 m. The AB/TR predictor-corrector adap-
P = 4.7 ⎪
tive time marching scheme is applied, where the RMS
ψ = 0; 0.1; 0.3; 0.6; 1.0 ⎭ –4
error tolerance (1-46) corresponds to δ = 10 . The
cbcilt=ö=QT
–5
initial time step is Δt o = 10 d. The final simulation NKUKPKQ oÉëìäíë
time is 800d. Due to the boundary conditions (1-91) the
The FEFLOW results for the problem are compared
divergence form7 of the governing transport equations
with the analytical solutions which are evaluated by
has to be used, which allows the input of the total
using the STANMOD package18. We simulate the
(advective plus dispersive) mass flux at a boundary.
breakthrough characteristics for C and S 2 measured at
the effluent boundary at x = L for different decay
To model the reaction kinetics we prefer the
parameters ψ . The plots are related to dimensionless
FEMATHED input for the rates (1-87b), which are
aqueous and sorbed concentrations C∗ and S 2∗ ,
specified as follows (note that species ID 1 represents
respectively, defined as
the aqueous species with concentration C ≡ C 1 and spe-
cies ID 2 represents the sorbed species with concentra- C
tion S 2 ≡ C2 ): C∗ = ------ ⎫
Co ⎪
⎪
R 1 = – ( Rate 2 ⋅ SolidFrac 2 ⋅ g ⋅ K (1-94a) S2 ⎬ (1-95)
S 2∗ = --------------------- ⎪
+ SolidFrac 2 ⋅ K ⋅ Rate 1 + Porosity 1 ⋅ Rate1 ) ⋅ C 1 1 – -f
--------- κC o ⎪
+ Rate 2 ⋅ SolidFrac 2 ⋅ C 2 f ⎭
f = 0.16667 g = ( 1 – f ) ⁄ f K = Sorption 1
Figure 1.23 reveals a good agreement with the analyti-
R 2 = Rate 2 ⋅ SolidFrac 2 ⋅ g ⋅ K ⋅ C 1 (1-94b)
cal solutions.
– SolidFrac 2 ⋅ ( Rate 2 + Rate1 ) ⋅ C 2
f = 0.16667 g = ( 1 – f ) ⁄ f K = Sorption 1
The parameters in (1-94a) and (1-94b) are related to

the notation used in (1-87b) as follows: R 1 ≡ R c ,
R2 ≡ Rs , Porosity 1 ≡ ε , Rate 1 ≡ μ , Rate 2 ≡ α ,
Sorption1 ≡ κ , and SolidFrac 2 ≡ ε s . The species IDs are
linked to the species names and phases as summarized
in Tab. 1.11.
Table 1.11 Species IDs for two-site sorption problem
ID ( = k ) Phase Name
1 fluid C
2 solid S2
QU=ö=tÜáíÉ=m~éÉêë=J=sçäK=fs
NKV=pÉèìÉåíá~ä=~åÇ=kçåJëÉèìÉåíá~ä=`Üäçêáå~íÉÇ=pçäîÉåíë=aÉÖê~Ç~íáçå=ìåÇÉê=s~êá~ÄäÉ
^ÉêçÄáÅJ^å~ÉêçÄáÅ=`çåÇáíáçåë
a) b)
0.8 0.5
ψ =0 ψ =0
STANMOD
0.1 FEFLOW
0.4
0.6 0.1
0.3 STANMOD
FEFLOW
Concentration C*
Concentration S*2
0.3
0.6 0.3
0.4
0.2
1.0
0.6
0.2
0.1
1.0
0.0 0.0
0 200 400 600 800 0 200 400 600 800
Time [days] Time [days]
Figure 1.23 FEFLOW results versus STANMOD solutions18 for effluent breakthrough history of a) aqueous C∗ and b) sorbed S 2∗ concentra-
tions for different decay parameters ψ at P = 4.7 , β = 0.5 , ω = 0.5 , ℜ = 2.5 and t o = 300 d.
NKV pÉèìÉåíá~ä= ~åÇ= kçåJ released during the dehalogenation can be useful to
ëÉèìÉåíá~ä= `Üäçêáå~íÉÇ locate the areas where dehalogenation occurs and to
estimate degradation rates. This example simulation
pçäîÉåíë= aÉÖê~Ç~íáçå issues from a benchmark within the MACAOH
ìåÇÉê= s~êá~ÄäÉ= ^ÉêçÄáÅJ (Modélisation de l'atténuation des composés organo-
^å~ÉêçÄáÅ=`çåÇáíáçåë chlorés dans les aquifères) project of the French Envi-
ronment and Energy Management Agency (ADEME)
NKVKN fåíêçÇìÅíáçå with various university and private partners, namely,
BURGEAP, IFARE, IFP, IMFS, and IMFT. The project
In contrast to nonsequential (aerobic) degradation
focuses on chlorinated solvents, specifically PCE (per-
of chlorinated solvents, sequential dehalogenation is
chloroethylene), TCE (trichloroethylene), DCE (cis-
performed by anaerobic bacteria that cannot work
and trans-1,2-dichloroethylene) and VC (vinyl chlo-
under aerobic conditions. Both mechanisms can occur
ride). The aim of the benchmark was to evaluate the
in the same contaminant plume depending on oxygen
state of the art in the numerical simulation of the natu-
and nitrate concentrations. Monitoring the chloride
cbcilt=ö=QV
ral degradation of chlorinated solvents in aquifers in +

H , 2e
-
-
+
H , 2e
-
-
+
H , 2e
-
-
France. The results of this project will be published in PCE → TCE + Cl → DCE + Cl → VC + Cl
2006. (1-96)
The concentration of nitrate is supposed to decrease

NKVKO `çåÅÉéíì~ä=éêçÄäÉã independently of other species, as soon as the oxygen
A mixture of PCE and TCE is injected continuously concentration reaches zero, following a simple first-
in an initially uncontaminated 1D domain containing order degradation law.
dissolved oxygen, nitrate, and chloride. The initial aer-
obic conditions do not allow the degradation of PCE by
anaerobic bacteria, but a slow complete mineralization NKVKP bèì~íáçåë= çÑ= ã~ëë= íê~åëéçêí
of TCE is considered. Daughter products of TCE dur- ïáíÜ= ëÉèìÉåíá~ä= ~åÇ= åçåJ
ing the mineralization (H2O and CO2) are not simu- ëÉèìÉåíá~ä=êÉ~Åíáçå
lated, except chloride ions. The oxygen concentration
The one-dimensional equation of transport for the
decreases as a consequence of the aerobic bacteria res-
homogeneous reaction of nonretarded parent and
piration. This behavior continues as long as the concen-
daughter species can be written as
tration of oxygen remains above a critical level. The
reactions take place only in the fluid phase. Reaction 2
∂C k ∂ Ck ∂C k anae ae
and sorption with the solid phase are neglected. ε --------- – D -----------
∂t 2
+ q --------- = ε ( δ k k k C k – ( 1 – δ k )k k C k )
∂x
∂x
2
(1-97)
∂C j ∂ Cj ∂C
Wiedemeier et al.24 explain that anaerobic bacteria ∂t 2
- + q --------j = ε ( δ j k anae
ε -------- – D ----------
∂x j
anae ae
C j – δ i a k, j k k C k – ( 1 – δ j )k j C j )
∂x
cannot work at oxygen concentrations greater than 0.5
mg l-1. When no more oxygen remains in water, aero-
bic bacteria use nitrate. After Wiedemeier et al.24, where ε denotes porosity, k the parent species (PCE to
anaerobic bacteria can start the sequential reductive DCE), j the daughter product (TCE to VC), C k and C j
dechlorination of chlorinated solvents under nitrate the concentration of species k and j , t the time, q the
concentrations smaller than 1 mg l-1. Darcy velocity, D the mechanical dispersion
( D = β L q with β L as the longitudinal dispersivity),
ae ae anae anae
In the MACAOH benchmark, it was assumed that kk , kj , kk and k j the first order decay constants of
the sequential degradation starts in the presence of species k and j under aerobic and anaerobic condi-
nitrate as soon as the oxygen concentration reaches tions, respectively, a k, j the stoichiometric yield for the
zero. Thus, once oxygen reaches sufficiently low con- degradation of species k to produce species j , δ k and
centration anywhere in the domain, the following δ j functions equal to 0 for degradation in aerobic con-
reductive sequential degradation of chlorinated sol- ditions and equal to 1 for degradation in anaerobic con-
vents starts ditions.
RM=ö=tÜáíÉ=m~éÉêë=J=sçäK=fs
NKVKQ oÉ~Åíáçå=ê~íÉë Oxygen
Aerobic bacteria do not use oxygen in definite propor-
Chlorinated solvents tions during their respiration. The oxygen consumption
The reaction rates appear on the right-hand sides of (1- was arbitrarily defined in the benchmark as follows
97). Under aerobic conditions, there is no sequential
ae
degradation and the reaction rates R PCE,TCE,DCE,VC of chlo- ∂C O2 ∂C TCE ∂C DCE ∂C VC
rinated solvents in aerobic conditions simplify to ----------- = 4.5 ------------- + 4 -------------
- + 3.5 -----------
- (1-100)
∂t ∂t ∂t ∂t
ae ⎫
R PCE = 0 ⎪ From the conceptual model, the terms in DCE and VC
ae ae ⎪ are superfluous because they are not present in the sys-
R TCE = – ε ( k TCE C TCE ) ⎪
⎬ (1-98) tem at the initial state and because TCE does not
ae
R DCE = 0 ⎪ degrade into those compounds in aerobic conditions.
⎪
ae
R VC = 0 ⎪ Thus in this example simulation there cannot be DCE
⎭ or VC under aerobic conditions and it follows from
(1-100) that
Under anaerobic conditions, the sequential degrada-
anae ae ae ae ae
tion from PCE to VC lead to reaction rates R PCE,TCE,DCE,VC R O2 = 4.5R TCE + 4R DCE + 3.5R VC (1-101)
defined as follows
By definition there is no oxygen in anaerobic condi-
anae anae ⎫ tions, so
R PCE = – ε ( k PCE C PCE ) ⎪
anae anae anae ⎪
R TCE = – ε ( k TCE C TCE – a PCE,TCE k PCE C PCE ) ⎪ anae
⎬ (1-99) R O2 = 0 (1-102)
anae anae anae
R DCE = – ε ( k DCE C DCE – a TCE,DCE k TCE C TCE ) ⎪
⎪
= – ε ( k VC C VC – a DCE,VC k DCE C DCE ) ⎪⎭
anae anae anae
R VC Chloride
Chloride is released into the groundwater during the
dehalogenation of chlorinated solvents. The relation
Except for PCE that has no parent species, all reaction
rates are made of an independent degradation term and ∂C Cl- ∂C TCE ∂C DCE ∂C VC
of a production term dependent on the degradation of ----------- = -1.068 ------------- – 0.712 -------------
- – 0.552 -----------
- (1-103)
∂t ∂t ∂t ∂t
the parent species. Note that because FEFLOW can
handle simultaneously homogeneous and heteroge-
is assumed under aerobic conditions, i.e., in terms of
neous reactions for the same species, users have to
reaction rates
include the porosity ’manually’ in the definition of the
reaction rates for a homogeneous reaction. ae ae ae ae
R Cl- = -1.068R TCE – 0.712R DCE – 0.552R VC (1-104)
cbcilt=ö=RN
uniform, are equal to the boundary condition for oxy-

For reductive dechlorination, different dechlorination gen, nitrate and chloride, and to zero for all chlorinated
kinetics is assumed, viz., solvents. The first order decay rates and stoichiometric
coefficients are given in Tab. 1.12.
∂C Cl- ∂C PCE ∂C TCE ∂C DCE
----------- = -0.208 ------------
- – 0.262 ------------- – 0.356 -------------
- (1-105)
∂t ∂t ∂t ∂t
∂C VC Table 1.12 First order decay rates and
– 0.552 -----------
- stoichiometric coefficients
∂t
and anae ae
Species k kk kk Unit a k, j Unit
anae anae anae anae
R Cl- = -0.208R PCE – 0.262R TCE – 0.356R DCE (1-106) PCE 0.03 0 d-1 aimless
anae
– 0.552R VC TCE 0.09 0.009 d-1 a PCE,TCE = 0.792 -
-1
DCE 0.009 0.15 d a TCE,DCE = 0.738 -
Nitrate
Nitrate is supposed to degrade at a given independent VC 0 0.24 d-1 a DCE,VC = 0.644 -
rate if the oxygen concentration is zero, which leads to oxygen 0 0 d-1 undefined
the following reaction rates
nitrate 0.1 0 d-1 undefined
ae
R NO3- = 0 chloride 0 0 d-1 undefined
(1-107)
anae anae
R NO3- = – ε ( k NO3- C NO3- )
FEFLOW’s FEMATHED Reaction Kinetics Editor

allows to easily define complex reaction rates. Particu-
NKVKR kìãÉêáÅ~ä=ãçÇÉä larly useful here is the ’if otherwise’ construct to
One-dimensional steady flow and transient trans- switch between aerobic and anaerobic behavior. It
port are supposed. The species simulated are PCE, allows the combination of aerobic and anaerobic reac-
TCE, DCE, VC, oxygen, nitrate, and chloride. The tion rates in one composite reaction rate R k for each
aquifer length is 250 m, Darcy velocity of water species k. The reaction rates are entered in FEFLOW as
0.4 m d-1. For all species, porosity ε is 0.4, longitudi- listed in Tab. 1.13.
nal dispersivity β L is 1 m, and retardation factor ℜk is
1. Steady Dirichlet transport boundaries of 3 mg l-1 for Another ’if otherwise’ construct is used to stop the
PCE, 5 mg l-1 for TCE, 0 mg l-1 for DCE and VC, 10 consumption of oxygen at an arbitrary residual concen-
mg l-1 for oxygen, 20 mg l-1 for nitrate, and 15 mg l-1 tration of 0.05 mg l-1. If the oxygen consumption was
for chloride, are applied at x = 0 . Initial conditions are not bounded by the user, the oxygen concentration
would continue to decrease and would lead to negative
RO=ö=tÜáíÉ=m~éÉêë=J=sçäK=fs
concentrations. In contrast, the degradation of nitrate order reaction that can only lead to positive concentra-
does not need to be bounded because it is a simple first- tions.
Table 1.13 Reaction rates as defined in FEFLOW
Reaction rate R k Comment
anae
⎧ – ε ( k PCE C PCE ) if C O2 < 0.1 ae
R PCE = ⎨ k PCE = 0
⎩0 otherwise
anae anae
⎧ – ε ( k TCE C TCE – a PCE,TCE k PCE C PCE ) if C O2 < 0.1
⎪
R TCE = ⎨ anae
k TCE
⎪ – ε ⎛ ---------
-C ⎞ otherwise
⎩ ⎝ 10 TCE⎠
⎧ –ε ( k C – a anae anae degradation is supposed

⎪ DCE DCE TCE,DCE k TCE C TCE ) if C O2 < 0.1
R DCE = ⎨ anae 0.150
under aerobic conditions
⎪ – ε ( k DCE ------------- ) otherwise
⎩ 0.009
anae no VC decay in anaerobic

⎧ – ε ( 0 – a DCE,VC k DCE C DCE ) if C O2 < 0.1
R VC = ⎨ conditions ( kanae
VC = 0 ) but
⎩ – ε ( k ae
VC C VC ) otherwise decay supposed under aero-
bic conditions
⎧ anae
k TCE
⎪ ⎛ 0.15 anae ae ⎞
R O2 = ⎨ ε ⎝ – 4.5 10 C TCE – 4.0 0.009 k DCE C DCE – 3.5 k VC C VC⎠
---------- ------------- if C O2 > 0.05
⎪
⎩0 otherwise
anae decay of nitrate starts under

⎧ – ε ( k NO3 - C NO3 - ) if C O 2 < 0.1
R
NO3
- = ⎨ anaerobic conditions
⎩0 otherwise
anae anae anae neither VC nor DCE under

⎧ ε ( 0.208 k PCE C PCE + 0.262 k TCE C TCE + 0.356 k DCE C DCE ) if C O 2 < 0.1
R
Cl
- = ⎨ aerobic conditions
⎩ ε ( 1.068 k ae
TCE C TCE ) otherwise
FEFLOW accepts only one value of first-order decay son why it is sometimes necessary to correct the value
rate for one species, but in this example, the decay rates of the rate in the Reaction Kinetics Editor to match the
differ in aerobic and anaerobic conditions. It is the rea- decay rate for the given conditions. The reference
cbcilt=ö=RP
decay rates are here the rates in anaerobic conditions. It The computation time is a function of the number of
means for example for TCE that the decay rate has to species simulated, because the transport problem has to
be divided by 10 in the kinetics editor for aerobic con- be solved for each species (here N = 7 species) at each
ditions. VC is the exception. It is necessary to enter the time step. Figure 1.25 shows the evolution of the time
ae anae
value k VC ≠ 0 instead of k VC = 0 in the first order steps during the computation. While the solution pro-
decay rate, to make the decay rate of VC available to ceeds relatively fast at the beginning, the speed of com-
the definition of the oxygen consumption. putation decreases drastically once the sequential
degradation starts. The sudden decrease in time step
length at about 233 days corresponds exactly to the
NKVKS páãìä~íáçå=êÉëìäíë time at which VC reaches the outlet leading to a strong
concentration gradient at the outlet for this species.
The problem was solved by FEFLOW for steady
flow and transient transport for 365 days in a two-
dimensional (pseudo one-dimensional) model as
shown in Fig. 1.24, with an automatic forward Euler/
backward Euler time-integration scheme without
upwind. The maximal time-step increase Ξ is 1.1 and
max
the maximal time step Δt is limited at 0.5 days to be
sure that the Courant number does not exceed 0.5. In
order to reproduce as closely as possible the simulation
conditions of other numerical codes (finite differ-
ences), the mesh is built on the basis of linear quadrilat-
eral elements. Models with 250, 500 and 1000
elements were solved.
Figure 1.25 Time steps Δt n as a function of computation

Ci(x=0,t)
time for 250 elements.
250 m
Figure 1.24 Pseudo one-dimensional mesh with 250 lin-

The simulation results of the 250-elements model are
ear quadrilateral elements. Vertical exageration 500:1. shown in Figs. 1.26, 1.27, and 1.28.
Identifier 1 and 2 represent observation points.
The computing time for the 250-elements model on

a 1.5 GHz Intel Pentium4 processor with 1.25 GByte
RAM is about 13 min.
RQ=ö=tÜáíÉ=m~éÉêë=J=sçäK=fs
PCE
TCE
DCE
VC
oxygen
nitrate
chloride
Figure 1.26 Concentration fringes for all species k at 150 days. Darcy velocity = 0.4 m d-1, adsorption = 0.
cbcilt=ö=RR
(a) O2 NO3- Cl-
25
20
15
ci [mg l ]
-1
10
0
0 50 100 150 200 250
x [m]
(b) 6
4
ci [mg l ]
-1
0
0 50 100 150 200 250
x [m]
PCE TCE DCE VC
Figure 1.27 Concentration profiles along the x-axis at 150 days, for (a) oxygen, nitrate, chloride, and (b) chlori-
nated solvents. Darcy velocity = 0.4 m d-1, adsorption = 0.
RS=ö=tÜáíÉ=m~éÉêë=J=sçäK=fs
O2 NO3- Cl- Cl_ae_analytic

(a)
O2_analytic NO3_analytic Cl_anae_analytic
25
20
15
ci [mg l ]
-1
10
0
0 50 100 150 200 250
-5
x [m]
6
(b)
5
4
ci [mg l ]
-1
0
0 50 100 150 200 250
x [m]
PCE TCE DCE VC

PCE_analytic TCE_analytic DCE_analytic VC_analytic
Figure 1.28 Concentration profiles along the x-axis at steady state (365 days), for (a) oxygen, nitrate, chloride, and
(b) PCE, TCE, DCE, and VC. Darcy velocity = 0.4 m d-1, adsorption = 0. Dashed lines represent analytical solutions.
cbcilt=ö=RT
Simulation results (a) show the extent of the aerobic on a self-made code named VisualRFlow. A participant
zone versus the anaerobic zone. At 150 days, steady using RT3D had serious dispersion problems. Another
state is reached from x = 0 to x = 65 m for all species. participant developed his own code and got a much
Two separate aerobic zones appear from x = 0 to 65 m narrower anaerobic zone. All these codes use finite dif-
and from x = 165 m to the outlet. Anaerobic conditions ferences. FEFLOW as the only finite-element code
are found between these two zones, allowing the degra- used in this benchmark has proven a versatile and user-
dation of nitrate and the sequential degradation of PCE friendly simulation system that allows the definition of
into TCE, DCE and VC. Under aerobic conditions, the multi-species and multi-reaction transport problems of
fast increase in chloride is a result of the complete min- arbitrary complexity.
eralization kinetics of TCE. After 365 days, the anaero-
bic zone extends from x = 65 m to the outlet. The
sequential degradation of chlorinated solvents leads to NKVKT ^ÅâåçïäÉÇÖÉãÉåíë
an accumulation of VC. The analytical solutions are
obtained for each species, for the aerobic domain first, D. Etcheverry and Y. Rossier, the authors of the
then for the anaerobic domain, by applying the decou- Chapter 1.9, wish to thank the French Environment and
pling solution of Sun et al.21 to Bear's one-dimensional Energy Management Agency (ADEME) for financial
support (Décision d'attribution d'une aide financière
transport and reaction analytical solution. No method
no. 0472C0004). The ADEME and its partners for the
was found to solve the problem analytically in transient
MACAOH project (BURGEAP, IFARE, IFP, IMFS,
state, because the solution must include a switching
and IMFT) are thanked for allowing publication of
term between aerobic and anaerobic rates of reactions
these numerical results.
as a function of the oxygen concentrations, which are
varying in space and in time. On the other hand, steady
state offers by definition a stable space and time limit
between aerobic and anaerobic conditions. Thus, it is NKNM pìãã~êó=~åÇ=`çåÅäìëáçåë
possible to solve first the aerobic domain, and to take
the concentrations calculated at the end of the domain In this paper, the details for the multi-species trans-
as boundary conditions for the calculations in the port in FEFLOW are reported. FEFLOW can be used
anaerobic domain. To avoid the development of a solu- to simulate any type of reaction kinetics with an arbi-
tion for parallel reactions, the contributions of each trary number of mobile or immobile species in 2D and
chlorinated solvent to the production of chloride under 3D applications. Multi-species transport is also avail-
anaerobic conditions are solved separately and added able for unsaturated problems, variable-density and
to the chloride concentration obtained at steady state at nonisothermal problems.
the end of the aerobic zone. The results given by
FEFLOW for all species are quasi identical to the ana- The paper describes the basic theory of the multi-
lytical solutions. A participant of the benchmark species transport. Reaction models are formulated for
obtained quasi identical results on PHAST (USGS) and degradation, Arrhenius and Monod kinetics. The
RU=ö=tÜáíÉ=m~éÉêë=J=sçäK=fs
NKNM=pìãã~êó=~åÇ=`çåÅäìëáçåë
numerical approach for solving the resulting nonlinear benchmark within the MACAOH project of the French
equation systems in the present finite-element context Environment and Energy Management Agency with
is thoroughly described. FEFLOW provides powerful some university and private partners.
adaptive predictor-corrector schemes with fully auto-
matic time marching. FEFLOW’s multi-species transport option is useful
for analyzing different and complex chemical reaction
The reaction-kinetics editor FEMATHED repre- problems in the subsurface. Radionuclide transport and
sents a new and powerful formula editor for an interac- biochemical processes in natural attenuation of con-
tive and graphical input of the required rate expressions taminants are typical applications. Predictions from the
at any level of complexity. The formulae can be com- simulations can be used for screening remediation
bined with if-else statements. There is no need for an alternatives and environmental risks.
additional compiling and linking of the kinetics expres-
sions. A sophisticated built-in code interpreter is avail-
able in FEFLOW which has been shown to be as fast as kçí~íáçå
precompiled formulae. FEFLOW provides precom-
piled rate expressions for degradation, Arrhenius and Roman letters
Monod kinetics.
A variable rate constant;
To illustrate and benchmark FEFLOW’s multi-spe- a ML
–3
transformed concentration variable;
cies transport functionality, a number of reactive trans-
a k, j 1 stoichiometric yield;
port example problems are posed and solved. The first
B L aquifer thickness;
example presents reactive chain applications in 1D and
B variable rate constant;
3D, typical for radionuclide decay. The second exam-
ple presents a comprehensive kinetic model for analyz- b variable rate constant;
– 3 1 – b2
ing rate-limited contaminant one-site kinetic sorption b1 ( ML ) Freundlich sorption coefficient;
reactions, typical for in-situ bioremediation problems. b2 1 Freundlich sorption exponent;
–3
Comparisons to analytical solutions reveal very good C ML concentration;
–3
agreements with the numerical FEFLOW results. The Cs ML overall maximum concentration;
–3
third example is focussed on a two-site equilibrium/ C ML nodal concentration vector;
kinetic sorption process combined with linear degrada- 2 –1
D L T tensor of hydrodynamic dispersion;
tion suitable for pesticide transport. For this type of D
3 –1
L T depth-integrated dispersion tensor;
problems good agreements with the STANMOD ana- Dd
2 –1
L T coefficient of molecular diffusion;
lytical solutions are shown. The fourth example con-
D variable rate constant;
cerns more complex sequential and non-sequential
d 1 error vector;
chlorinated solvents degradation under variable aero-
d variable rate constant;
bic-anaerobic conditions. This simulation issues from a
cbcilt=ö=RV
E variable rate constant; y 1 stoichiometric yield;

e variable rate constant;
–2
g LT gravitational acceleration; Greek letters
h L hydraulic head;
3 –1
I 1 unit tensor; α L M solutal expansion coefficient;
–1 –1
K T nodal conductance matrix; α T desorption rate constant;
3 –1
Kd L M equilibrium distribution coefficient; α 1 fluid density difference ratio;
–1 –1
k T rate constant; β Θ thermal expansion constant;
k1 1 Langmuir 1st sorption coefficient; β L, β T L longitudinal and transverse disper-
3 –1 sivity, respectively;
k2 L M Langmuir 2nd sorption coefficient;
–1 2
L L length; γ M LT fluid compressibility;
–1
N total number of chemical species; γ L specific fluid compressibility;
N° number of reactants; Δt T temporal increment;
Nr number of reactions; Δx L spatial increment;
n 1 concentration exponent; δ 1 error tolerance;
–1 –2 ε 1 porosity, void space;
p ML T pressure;
–3 –1 εα 1 volume fraction of α -phase;
Q ML T nonreactive production term;
–3 –1 εs 1 solid volume fraction;
Q ML T bulk nonreactive production term;
–2 –1 –1
ϑ T decay rate constant;
)
Q ML T depth-integrated bulk nonreactive

production term; κ 1 Henry sorption coefficient;
–1
q LT Darcy velocity vector; κk variable bulk rate constant of species k;
2 –1 –1 –1
q L T depth-integrated Darcy flux; μ ML T dynamic viscosity;
–3 –1 –3
R ML T bulk rate of chemical reaction; ρ ML density;
–3 –1
R ML T nodal bulk rate vector; Ξ 1 rate for changing time step size;
ℜ 1 retardation factor; ν kr 1 stoichiometric number;
ℜ L depth-integrated retardation factor; χ 1 adsorption function;
–3 –1
r ML T rate of reaction; ω 1 density factor;
–3 –1
S ML sorbed concentration; ∇ L Nabla (vector) operator;
S 1 nodal storage matrix;
s 1 saturation; Subscripts
T Θ temperature;
t T time; α phase indicator;
–1
v LT pore velocity vector; b bulk;
x L spatial coordinate; f fluid phase;
SM=ö=tÜáíÉ=m~éÉêë=J=sçäK=fs
NKNM=pìãã~êó=~åÇ=`çåÅäìëáçåë
het heterogeneous; ID identifier of species;

hom homogeneous; LMA law of mass action;
i species index; MACAOH Modélisation de l'atténuation des
j species index; composés organo-chlorés dans les
k species indicator; aquifères;
m species index; PCE perchloroethylene;
n time plane or species index; RMS root-mean square;
o initial value; SIA sequential interactive approach;
r reaction; TCE trichloroethylene;
s solid phase; TR trapezoid rule;
VC vinyl chloride;
Superscripts
oÉÑÉêÉåÅÉë
ae aerobic;
1. Bear, J. and Bachmat, Y. Introduction to modeling of transport
anae anaerobic; phenomena in porous media. Kluwer Acad. Press, Dordrecht,
α phase indicator; 1991.
f fluid phase; 2. Clement, T.P. RT3D - A modular computer code for simulating
reactive multi-species transport in 3-dimensional groundwater
n time plane; systems. Documentation. Version 1.0, Pacific Northwest National
max maximum value; Laboratory, Richland, Washington 99352, 1997.
p predictor value; 3. Clement, T.P., Sun, Y., Hooker, B.S. and Petersen, J.N. Modeling
s solid phase; multi-species reactive transport in groundwater aquifers. Ground-
water Monitoring & Remediation Journal 18 (1998)2, 79-92.
4. Diersch, H.-J.G., Modeling and numerical simulation of geohy-
Abbreviations drodynamic transport processes (in German). Reprint, WASY
Ltd. Berlin, 1991.
AB Adams-Bashforth; 5. Diersch, H.-J.G. Interactive, graphics-based finite-element simu-
lation system FEFLOW for modeling groundwater flow, contam-
AB/TR Adams-Bashforth/trapezoid rule inant mass and heat transport processes. Release 5.2, Reference
scheme; Manual, WASY Ltd., Berlin, 2005.
BE backward Euler; 6. Diersch, H.-J.G. On the primary variable switching technique for
BTEX bezene-toluene-xylene mixture; simulating unsaturated-saturated flows. FEFLOW White Papers
Vol. I, Chapter 1, WASY Ltd. 2002, Berlin, pp. 9-66.
DCE cis- and trans-1,2-dichloroethylene;
7. Diersch, H.-J.G. About the difference between the convective
EOS equation of state: form and the divergence form of the transport equation. FEFLOW
FE forward Euler; White Papers Vol. I, Chapter 6, WASY Ltd. 2002, Berlin, pp. 119-
FE/BE forward Euler/backward Euler 129.
scheme;
cbcilt=ö=SN
8. Diersch, H.-J.G. Chemical reactions. FEFLOW White Papers Vol. 21. Sun, Y., Petersen, J.N. and Clement, T.P. Analytical solutions for
I, Chapter 10, WASY Ltd. 2002, Berlin, pp. 191-206. multiple species reactive transport in multiple dimensions. Jour-
9. Diersch, H.-J.G. Coupled groundwater flow and transport: Ther- nal of Contaminant Hydrology 35 (1999), 429-440.
mohaline and 3D convection systems. FEFLOW White Papers 22. Toride, N., Leij, F.J. and van Genuchten, M.Th. The CXTFIT
Vol. I, Chapter 17, WASY Ltd. 2002, Berlin, pp. 307-349. code for estimating transport parameters from laboratory or field
10. Diersch, H.-J.G. and Kolditz, O. Variable-density flow and trans- tracer experiments. Vers. 2.1, US. Salinity Lab. Riverside, Cal.,
port in porous media: approaches and challenges. Adv. Water Research Report No. 137, 1999.
Resour. 25 (2002), 899-944. 23. Wiedemeier, T.H., Swanson, M.A., Wilson, J.T., Kampbell, D.H.,
11. Fry, V.A., Istok, J.D., and Guenther, R.B. An analytical solution Miller, R.N. and Hansen, J. Approximation of biodegradation rate
to the solute transport equation with rate-limited desorption and constants for monoaromatic hydrocarbons (BTEX) in groundwa-
decay. Water Res. Resour. 29 (1993)9, 3201-3208. ter. Groundwater Monitoring & Remediation Journal 16 (1996)3,
12. Garcia-Talavera, M. Laedermann, J.P., Decombaz, M., Daza, M.J. 186-194.
and Quintana, B. Coincidence summing corrections for the natu- 24. Wiedemeier, T.H., Swanson, M.A., Moutoux, D.E., Kinzie Gor-
ral decay series in γ-ray spectrometry. J. Radiation and Isotopes don E., Wilson, J.T., Wilson B.H., Kampbell, D.H., Haas, P.E.,
54 (2001), 769-776. Miller, R.N., Hansen, J.E. and Chapelle, F.H. Technical protocol
13. Kanney, J.F., Miller, C.T. and Kelley, C.T. Convergence of itera- for evaluating natural attenuation of chlorinated solvents in
tive split-operator approaches for approximating nonlinear reac- ground water. EPA/600/R-98/128. U.S. Environmental protection
tive transport problems. Adv. Water Resour. 26 (2003)3, 247-261. agency, 1998.
14. Lichtner, P.C., Continuum formulation of multicomponent-mul- 25. van Genuchten, M.Th. and Wagenet, R.J. Two-site/two-region
tiphase reactive transport. In: Reactive transport in porous media, models for pesticide transport and degradation: Theoretical
ed. by Lichtner et al., Reviews in Mineralogy, Vol. 34, Mineral- development and analytical solutions. Soil Sci. Soc. Am. J. 53
ogical Soc. Am., Washington DC, 1996, pp. 1-81. (1989)5, 1303-1310.
15. Lindstrom, F.T., Pulsed dispersion of trace chemical concentra-
tions in a saturated sorbing porous medium. Water Res. Resour.
12 (1976), 229-238.
16. Nguyen, V.V., Gray, W.G., Pinder, G.F., Botha, J.F. and Crerar,
D.A. A theoretical investigation on the transport of chemicals in
reactive porous media. Water Res. Resour. 18 (1982)4, 1149-
1156.
17. Rifai, S.H., Newell, C.J. Miller, R., Taffinder, S. and Rounsaville,
M. Simulation of natural attenuation with multiple electron
acceptors. Bioremedation 3 (1995)1, 53-58.
18. STANMOD (STudio of ANalytical MODels): Computer software
for evaluating solute transport in porous media using analytical
solutions of convection-dispersion equation. http://
www.ars.usda.gov/services/software/software.htm.
19. Stumm, W. and Morgan, J.J. Aquatic chemistry. Wiley-Inter-
science, New York, 1981.
20. Sun, Y., Petersen, J.N., Clement, T.P. and Skeen, R.S., Develop-
ment of analytical solutions for multispecies transport with serial
and parallel reactions. Water Res. Resour. 35 (1999)1, 185-190.
SO=ö=tÜáíÉ=m~éÉêë=J=sçäK=fs
m~ê~ãÉíÉê=Éëíáã~íáçå=çÑ=íê~åëáÉåí=Ñäçï=éêçÄäÉãë
ïáíÜ=mbpq=áå=cbcilt
Jan Christian Kaiser

O
ITUPH Institut für Theoretische Umweltphysik, Munich, Germany
for observable physical quantities, will profit from the

O m~ê~ãÉíÉê=Éëíáã~íáçå=çÑ=íê~åëáÉåí=Ñäçï=éêçÄäÉãë=ïáíÜ=mbpq=áå=cbcilt
OKN fåíêçÇìÅíáçå application of concepts and methods of data assimila-

tion. In this analysis technique observed information
The integration of the PEST package for parameter (i.e. measured gauge curves of hydraulic heads) is used
estimation (Doherty et al.6) into the FEFLOW to improve the model state (i.e. FEFLOW's material
simulator4 was accomplished in FEFLOW's version 4.8 parameters) and to enhance the predictive power of the
for stationary flow processes. The coupling has been model. Two basic approaches to data assimilation are
realized via FEFLOW's interface manager (see known (Bouttier and Courtier1):
Gründler8) using the method of direct implementation.
The PEST module has access to FEFLOW's address • Sequential assimilation, which considers only
space and is called within its graphical user interface. observations from the past until the time of analy-
Data exchange is facilitated during run time with call- sis. This is the case for models of meteorological
back functions. A description of the PEST-FEFLOW weather forecast or flood forecast models in real-
coupling is given in Kaiser11. time applications.
• Retrospective assimilation, where observations
In FEFLOW's version 5.2 the range of applicability from any conceivable time range can be used for
of parameter estimation with PEST has been extended model calibration and validation. Groundwater
to transient flow problems. The extension required only models are typically treated with this approach.
minor changes in the PEST graphical user menu for
observations, since the time-dependent observation Both approaches can use the observed information
curves are treated as power functions, for which either intermittently or continuously in time. In the
FEFLOW provides a comprehensive data management. intermittent mode, observations are grouped in time
intervals, which are much smaller than the total simula-
From a theoretical point of view the analysis of tion time but much larger than the time scales of the
transient simulation models, which produce predictions underlying physical processes. Usually, this mode is
cbcilt=ö=SP
OK=m~ê~ãÉíÉê=Éëíáã~íáçå=çÑ=íê~åëáÉåí=Ñäçï=éêçÄäÉãë=ïáíÜ=mbpq=áå=cbcilt
technically convenient. In the continuous mode obser- n

T –1
vations over the whole simulation time of the model
are considered to improve the model state, which is
J(p) = ∑ ( o i ( t ) – M i ( t, p ) ) Ri ( oi ( t ) – M i ( t, p ) ) (2-1)
i=1
physically more realistic.
by summing over all n observation points. The vector
Most subsurface flow models are conveniently
treated with retrospective data assimilation in the con-
⎛ o (t ) ⎞
tinuous mode. To make use of the PEST algorithm to ⎜ i 1 ⎟
minimize the objective function, the model parameters oi ( t ) = ⎜ … ⎟ (2-2)
⎜ ⎟
must be constant in time. The general data assimilation ⎝ oi ( tm ) ⎠
concept includes time dependent model parameters as
well. However, in this case the Kalman filter technique
should be used for the parameter update. This tech- contains the recorded measurements of the hydraulic
nique is mathematically equivalent to the minimization head for m time steps of the gauge curve pertaining to
of the objective function†) (Bouttier and Courtier1). observation point i . For the vector M i ( t, p ) an analo-
gous expression is valid. Measurement errors are mod-
In the present implementation complete use of the eled here in a simplified way. They are treated as
observed information is made by considering measure- uncorrelated and identical for a given observation
ments at each observation point i of the model and at point. After normalization they are represented as
each time step t j . Measured hydraulic heads o i ( t j ) can weights of the gauge curves. Note, that for each obser-
be compared with model predictions M i ( t j, p ) by using vation point an individual sequence of time steps
the callback function OnTimeStepConstraint() exists, which may or may not differ from the sequences
of the interface manager. This functions ensures, that of other observation points. The set of FEFLOW mate-
the hydraulic head is calculated for the whole finite ele- rial parameters p min minimizes the objective function
ment grid at a prescribed time step irrespectively of the J ( p ) and can be considered as the optimal parameter
selected algorithm for time stepping. The vector p con- set of the parameter estimation problem. A full descrip-
tains the time-constant material parameters of tion of the mathematical framework of PEST and of the
FEFLOW, which are updated in the estimation process. algorithms to minimize the objective function is given
in Doherty et al.6.
Now the objective function J ( p ) can be constructed
with the residual vectors o i ( t ) – M i ( t, p ) and the covari- In the following section four simple test examples
–1
ance matrix R i of the measurement errors are presented to demonstrate the reliability of the
implementation of PEST in FEFLOW for transient
flow problems. The examples cover both confined and
unconfined aquifers with measured and synthetically
†) called 4D variational assimilation (4D VAR) in Bouttier generated observation data. In Theis' well problem for
and Courtier1.
SQ=ö=tÜáíÉ=m~éÉêë=J=sçäK=fs
OKO=qÉëí=bñ~ãéäÉë
an unconfined aquifer synthetically generated gauge explained in Tab. 2.1.

curves are used to retrieve the material parameters of a
FEFLOW benchmark problem. The pump tests of The time stepping algorithms of the FEFLOW
Breyell and Wichita have been conducted in a confined groundwater simulator have been successfully tested
and an unconfined aquifer. They were analyzed already against the Theis benchmark (Diersch5) on the finite
by Langguth and Voigt12 with graphical evaluation element grid of Fig. 2.1. Here the same grid with the
methods. Here they have been reanalyzed with PEST in same values for the simulation parameters (Tab. 2.1) is
FEFLOW to compare the results of both approaches. used to test the implementation of the PEST algorithm
Finally, a generic floodwave problem has been set up to for the estimation of parameters in transient flow prob-
assess the results of a parameter estimation, which is lems. The simulation was done with automatically con-
influenced by the consequences of non-uniqueness. trolled time steps based on the first order (FE/BE)
predictor-corrector techniques.
OKO qÉëí=bñ~ãéäÉë
OKOKN qÜÉáëD=ïÉää=éêçÄäÉã
Theis' problem of the lowering of a water table by a
pumping well is a famous benchmark problem for
numerical algorithms since it possesses an analytical
solution couched in the well-known well function
(Theis14). The complete solution of the governing par-
tial differential equation of second order
2
∂h- ∂--------h 1--- -----
∂h-
--S- ----- – – = 0 (2-3) Figure 2.1 FEFLOW grid with PEST observation points
T ∂t ∂r 2 r ∂r for Theis' well problem.
for a confined aquifer in a radial symmetric geometry The coordinates of the observation points on the
with the boundary conditions of a constant flux q into FEFLOW grid are given in Tab. 2.2. At these points the
a pumping well of infinitely small radius and a constant drawdown of the Theis problem has been recorded. For
head h o at the outer fringe, which is very far away PEST 15 time steps at each observation point have
from the well, has been derived by Theis by exploiting been selected with an approximately constant distance
the analogy to a heat conduction problem (Carslaw and on a logarithmic scale (Fig. 2.3). The PEST estimation
Jaeger2). The initial condition h o ( 0, r ) is a constant problem has been set up in one zone which comprised
head of zero at t = 0 in the whole plane, except at the all nodes of the FEFLOW grid. The transmissivity T
wellbore where h ( 0, r b ) = -0.01 m . All symbols are and the storage coefficient S have been estimated
cbcilt=ö=SR
simultaneously using the initial values of 10-4 m2s-1 The course of the PEST objective function during
and 10-3, respectively. The optimization control param- the optimization is shown in Fig. 2.2. With the starting
eters of PEST have been left at their default values. values for T and S the objective function reaches some
104 for in total 60 time steps from 4 observation points.
–6
Table 2.1 Simulation parameters for Theis' well After 26 model runs it has been minimized to 2.5 ⋅ 10
problem by passing through 10 orders of magnitude. The model
runs no. 27-39 are used by PEST to verify that a mini-
Name Symbol Unit Value mum of the objective function has been actually
reached.
wellbore radius rb m 0.3048
flow initial h ( 0, r ) m 0
flow initial at wellbore h ( 0, r b ) m -0.01

Q
well pumping rate q = -----------
2πr b
m2d-1 638.75425
fixed head at outer h ( t, 304.8 m ) m 0

boundary
–4
transmissivity T m2s-1 9.2903 ⋅ 10
storage coefficient S - 0.001
storage compressibility So - 0
Table 2.2 Coordinates and distance from the

pumping well of the observation points for Theis'
Figure 2.2 Minimization of the PEST objective func-
well problem tion for Theis' well problem.
Obser- Coordinates Distance

vation –4
point no. x[m] y[m] r[m] The estimated values from PEST are 9.2902 ⋅ 10
2 -1 –4
m s and 1.0002 ⋅ 10 for T and S , respectively. For
1 9.64 0 9.64 both parameters PEST produced highly accurate point
estimates which deviate only in the fifth digit from the
2 76.98 12.16 77.93
true values of Tab. 2.1. The correlation coefficient for
3 150.28 23.75 152.15 T and S is -0.57 and indicates a relatively weak corre-
lation favouring an effective parameter estimation. Fig-
4 223.57 35.33 226.34
ure 2.3 exhibits a perfect agreement of 'measured' and
SS=ö=tÜáíÉ=m~éÉêë=J=sçäK=fs
predicted hydraulic heads at the observation points OKOKO _êÉóÉää=éìãé=íÉëí

which should have been expected with the highly pre-
cise point estimates from PEST. This figure also dem- The pump test of the Breyell waterworks (Langguth
onstrates that after some 0.1 d the hydraulic head and Voigt12) has been evaluated on the basis of the ana-
becomes constant at all observation points owing to a lytical results of Theis14. The test has been conducted
simulation system of finite extent. On the other hand, in a confined aquifer in the Lower Rhine region near
with growing distance from the pumping well, the the city of Krefeld. The sequence of geological layers
begin of the drawdown is delayed. Therefore, the infor- has been subdivided in an upper unconfined aquifer
mation from observation point 4 alone is not sufficient which is separated by a layer of very impermeable clay
to estimate the material parameters, because of the from the lower unconfined aquifer. During the test the
–2
short drawdown phase. However, with the drawdown pumping rate was held constant at Q = 2.67 ⋅ 10
3 -1
curve of observation point 1 alone a very precise m s . The drawdown has been observed at three
parameter estimation is always possible. gauges: gauge 11b at a distance of 7.4 m, gauge 3b at a
distance of 23 m and gauge 6b at a distance of 139.6 m
from the well. The recorded drawdown curves, copied
from Langguth and Voigt12, are given in the Tables 2.3,
2.4 and 2.5. For gauge 11b also the rerise has been
recorded after the pump was stopped at t = 0.1690 d.
The measured values are shown in Tab. 2.6. All mea-
sured curves for the Breyell pump test are depicted in
Fig. 2.5.
Table 2.3 Measured drawdown at gauge
11b with r = 7.40 m for the Breyell
pump test
Time Drawdown
[10-4 d] [m]
1.968 0.30
Figure 2.3 Comparison of measured (symbols) and pre- 2.431 0.40
dicted (full lines) hydraulic heads for the four observa-
tion points (OP) of Theis' well problem. 3.125 0.50
4.051 0.70
6.134 0.93
cbcilt=ö=ST
Table 2.3 Measured drawdown at gauge Table 2.4 Measured drawdown at gauge
11b with r = 7.40 m for the Breyell 3b with r = 23 m for the Breyell pump
pump test (continued) test
Time Drawdown Time Drawdown

[10-4 d] [m] [10-4 d] [m]
9.838 1.03 1.389 0.05
19.44 1.20 2.083 0.10
25.46 1.28 2.546 0.15
32.52 1.35 4.167 0.25
38.19 1.40 6.250 0.35
52.08 1.47 9.375 0.45
62.50 1.50 14.12 0.55
83.33 1.56 26.16 0.70
125.0 1.64 40.04 0.80
166.7 1.71 50.23 0.85
243.1 1.79 63.54 0.90
347.2 1.86 79.75 0.95
451.4 1.90 99.65 1.00
625.0 1.99 125.0 1.05
960.6 2.07 164.4 1.11
1690 2.18 208.3 1.15
250.0 1.20
319.4 1.25
402.8 1.30
506.9 1.35
SU=ö=tÜáíÉ=m~éÉêë=J=sçäK=fs
Table 2.4 Measured drawdown at gauge Table 2.5 Measured drawdown at gauge
3b with r = 23 m for the Breyell pump 6b with r = 139.60 m for the Breyell
test (continued) pump test (continued)
Time Drawdown Time Drawdown

[10-4 d] [m] [10-4 d] [m]
611.1 1.39 555.6 0.59
937.5 1.48 746.5 0.65
1342 1.55 1001 0.71
1638 1.58 1325 0.76
1609 0.80
Table 2.5 Measured drawdown at gauge

Table 2.6 Measured rerise at gauge
6b with r = 139.60 m for the Breyell
11b after the pump stop at
pump test
t = 0.169 d
Time Drawdown
Time Drawdown
[10-4 d] [m]
[10-4 d] [m]
27.78 0.14
1690 2.18
41.67 0.20
1709 1.03
69.44 0.25
1716 0.95
90.28 0.29
1721 0.90
131.9 0.35
1727 0.85
173.6 0.40
1763 0.69
236.1 0.45
1775 0.65
319.4 0.50
1806 0.59
451.4 0.55
1830 0.55
cbcilt=ö=SV
Table 2.6 Measured rerise at gauge Langguth and Voigt12 have given estimates for the
11b after the pump stop at transmissivity T and the storage compressibility S
t = 0.169 d (continued) using graphical evaluation methods of Theis14, and
Cooper and Jacob3, respectively. Theis' method is
Time Drawdown based on the graphical matching of a modified well
[10-4 d] [m] function with scaled curves of the recorded drawdown.
Cooper and Jacob3 use the exponential behavior of the
1873 0.47
drawdown which dominates the process asymptotically
1898 0.44 for sufficiently large times. A detailed description of
both methods is not the subject of this paper, it is given
2002 0.35
in Langguth and Voigt12. All estimates for T and S are
2072 0.30 summarized in Tab. 2.7. The averages over all esti-
–4 –4
mates are 97 ⋅ 10 m2s-1 for T and 3.7 ⋅ 10 for S .
2148 0.26
The corresponding maximal relative deviations are
2225 0.23 24 % for T and -43 % for S . Hence, the storage com-
pressibility S shows a larger variability and can only
2396 0.17
be estimated with less precision.
2697 0.09
Table 2.7 Summary of estimates for the transmissivity T and the storage compressibility S of the
Breyell pump test from Langguth and Voigt12
Gauges used for Transmissivity T Storage compressibility S

estimation [10-4 m2s-1] [10-4]
Theis’ superposition method
11b & 3b & 6b together 110 2.4
Time-drawdown-method (Cooper and Jacob3)
6b alone 120 2.1
3b alone 94 4.6
11b alone 89 3.7
Distance-drawdown-method (Cooper and Jacob3)

11b & 3b & 6b together 90 4.8
TM=ö=tÜáíÉ=m~éÉêë=J=sçäK=fs
Table 2.7 Summary of estimates for the transmissivity T and the storage compressibility S of the
Breyell pump test from Langguth and Voigt12 (continued)
Gauges used for Transmissivity T Storage compressibility S

estimation [10-4 m2s-1] [10-4]
Distance-time-drawdown-method (Cooper and Jacob3)
11b & 3b & 6b together 89 4.6
Rerise method (Jacob9)
11b alone 84 n.a.
In FEFLOW the pump test is modeled in a 2D prob-

lem for confined aquifers on a mesh with 2138 3-noded
triangles. The outline of the mesh consists of a circle
segment with an opening angle of 30° and a radius of
1500 m. The mesh is based on two superelements. The
first superelement covers the zone near the well up to
the radius of 150 m with a higher spatial resolution of
the finite elements. This zone is shown in Fig. 2.4
together with the location of the pumping well and the
three gauges 6b, 3b, and 11b, respectively. The pump-
ing well is modeled with a boundary condition of the
4th kind at the tip of the circle segment. On the outer
boundary a constant head of 0 m is prescribed with a
boundary condition of the 1st kind. The initial head on
all nodes was set to zero.
cbcilt=ö=TN
Figure 2.4 Finite element grid near the pumping well at the tip with gauges 11b at r = 7.40 m, 3b at r = 23.00 m and
6b at r = 139.60 m.
With PEST the transmissivity T and the storage of estimation runs has been conducted. Firstly, the
compressibility S have been estimated simultaneously drawdown curves were used separately to produce the
in one zone which covered the whole mesh. The initial parameter estimates. Then the drawdown curves of the
–4 –4
values were chosen to be 10 m2s-1 and 4 ⋅ 10 , Tables 2.3 to 2.5 have been used together in one esti-
respectively. The control parameters for the optimiza- mation run. Also the rerise curve of Tab. 2.6 was used
tion process have been left at their default values. Also in a separate run. Finally, a PEST problem has been set
the correlation structure of the two estimated parame- up using all recorded 75 observation points from the
ters has been calculated. To allow for a direct compari- Tables 2.3 to 2.6.
son with the results from Langguth and Voigt12 a series
TO=ö=tÜáíÉ=m~éÉêë=J=sçäK=fs
Table 2.8 contains the point estimates of the six the pumping well. Langguth and Voigt12 observe the
PEST estimation runs for the transmissivity T and the same trend. The PEST estimates for S differ by some
storage compressibility S . Owing to the narrow width factor of four whereas in Langguth and Voigt12 the
of the objective function near the minimum, the 95 % spread is markedly lower. For both T and S the aver-
confidence intervals (CI) are mostly around one digit. ages over all estimates agree well with PEST point esti-
Only for the PEST run using gauge 6b alone, which has mates of the last line of Tab. 2.8 which were obtained
the largest distance from the pumping well, the CI were by using all information of the 75 measured time steps
notably larger. In this run, which needed 73 model calls in one estimation run.
to converge, the correlation coefficient between T and In Fig. 2.5 the curves of the measured and predicted
S is almost one. In all other PEST runs the correlation drawdown are compared. To calculate the predicted
coefficient is negative and around 30 model calls were drawdown the point estimates of the last line of Tab.
necessary. The absolute value is minimal if the time 2.8 were used. The agreement is very good except for
steps for the rerise in gauge 11b is included in the esti- the late phase of the rerise in gauge 11b. During the
mation. Small correlation coefficients favour the stabil- pump test the pore structure of the aquifer may change
ity of the optimization process. so that drawdown and rerise cannot be described with
one pair of constant parameter values for T and S .
The point estimates for T exhibit a similar variabil- However, in general, with PEST in FEFLOW the
ity as those from Langguth and Voigt12 in Tab. 2.7. The results for the Breyell pump test of Langguth and
estimates of T decrease with decreasing distance from Voigt12 have been fully confirmed.
Table 2.8 Parameter estimates and statistical information of the PEST estimation runs for the
Breyell pump test
Storage Number of
Gauges used for Transmissivity T Objective Correlation
compressibility S observation J ⁄ N obs
estimation [10-4 m2s-1] function J coefficient
[10-4] points Nobs
–3 –3
6b alone 117 2.35 1.7 ⋅ 10 14 3.0 ⋅ 10 0.99
–2 –3
3b alone 90.7 5.12 1.9 ⋅ 10 24 5.7 ⋅ 10 -0.78
–1 –2
11b alone 79.8 7.91 2.0 ⋅ 10 21 2.1 ⋅ 10 -0.82
–1 –2
6b & 3b & 11b 88.2 5.03 3.7 ⋅ 10 59 1.0 ⋅ 10 -0.77
–2 –3
11b rerise only 96.8 1.81 2.5 ⋅ 10 17 9.3 ⋅ 10 -0.65
–1 –3
all observation 92.0 4.31 4.7 ⋅ 10 75 9.2 ⋅ 10 -0.55
points
cbcilt=ö=TP
Figure 2.5 Measured (symbols) and predicted (lines) drawdown for the gauges 6b
(diamonds, dotted line), 3b (squares, broken line) and 11b (circles, full line) of the
Breyell pump test.
OKOKP táÅÜáí~=éìãé=íÉëí with a water-filled height of 8.20 m and a maximal

drawdown of 1.80 m at a gauge with a distance of
The Wichita pump test (Langguth and Voigt12) has 15.00 m from the pumping well. This value was
been conducted in an unconfined aquifer. If the recorded after 18 d when the well was operated at an
assumption of Dupuit (1863) is valid, the well formulas –2
average pumping rate of Q = 6.31 ⋅ 10 m3s-1. Values
of Theis14 can also be applied to aquifers with free
for in total six gauges, which have been arranged in
water tables with sufficient accuracy. In practical appli- two rows in the north and the south of the pumping
cations the maximal drawdown must be small (i.e. < 15 well, are given in Tab. 2.9. Langguth and Voigt12 have
% of the original water-filled height of the aquifer).
evaluated the pump test with a graphical method of
However, this is not the case for the Wichita pump test
TQ=ö=tÜáíÉ=m~éÉêë=J=sçäK=fs
Cooper and Jacob3. To account for the deviation from instead, where B denotes the original water-filled
Dupuit's assumption they corrected the recorded draw- height of the aquifer (Jacob10).
2
down s and used a lower drawdown s' = s – s ⁄ 2B
Figure 2.6 Finite element grid for the Wichita pump test with the location of the northern gauges 1-3 and the southern
gauges 4-6.
For the simulation of the pump test with PEST in pumping well at the tip. The pumping well was mod-
FEFLOW the same circle segment as for the Breyell eled with a well boundary condition of the 4th kind. At
pump test was used with a radius of 1500 m and an the outer radius a constant head boundary condition of
opening angle of 30°. A cut of the grid near the tip with 0 m has been prescribed. The problem has been set up
the location of the six gauges of Tab. 2.9 is shown in in 2d with an unconfined aquifer of -8.20 m bottom ele-
Fig. 2.6. Owing to the radial symmetry the results vation. The initial head at all nodes has been set to
depend only on the distance between the gauge and the zero. The final simulation time was 20 d.
cbcilt=ö=TR
mated simultaneously with the start values of 10-2 ms-1

and 0.2, respectively.
Table 2.9 Distance to pumping well and
observed drawdown at t = 18 d for the 6
The minimum of the objective function at 0.016
gauges of the Wichita pump test
was reached in 7 optimization iterations after 35 model
runs. In Tab. 2.10 the PEST point estimates are com-
Distance to
Drawdown pared with the results of the graphical evaluation
Gauge no. pumping well
[m] method of Langguth and Voigt12. For the hydraulic con-
[m]
ductivity the point estimate of PEST lies close to the
northern row estimate of the graphical evaluation method using the
corrected drawdown. The PEST estimate of the storage
1 15.00 1.80 compressibility overestimates value from the method
2 30.70 1.40 with the corrected drawdown by some ten percent. The
agreement of the estimates for both material parame-
3 57.70 1.04 ters from PEST is better with the estimates from the
southern row method using the corrected drawdown. This agreement
justifies the application of this method from hindsight.
4 14.95 1.67
5 30.60 1.31 The very large 95 % CI of the PEST point estimates

for the storage compressibility indicate a broad mini-
6 57.90 0.97 mum of the objective function. The estimated material
parameters show a large negative correlation of almost
-99 %. Both observations indicate that the Wichita
For PEST the six gauges of Tab. 2.9 were used as pump test poses a more difficult estimation problem
observation points. The definition of the time-depen- than the Breyell pump test.
dent power functions in FEFLOW requires at least two
time steps, but at the gauges measurements had been In Fig. 2.7 the predicted drawdown curves for the
taken only at one time step of 18 d. Therefore, to obtain three northern gauges are shown. Those for the south-
six valid power functions to each time step the initial ern gauges have been omitted because they are almost
hydraulic head 0 m at t = 0 was added. Note, that with identical within the drawing accuracy. The predicted
this measure no additional information has been intro- curves pass between measured points of the northern
duced into the estimation problem, because the initial and southern gauges which is an indication of the plau-
condition is already known without a need to run the sibility of the estimation result.
model. It does not depend on changes of the model
parameter. The two material parameters hydraulic con-
ductivity and storage compressibility have been esti-
TS=ö=tÜáíÉ=m~éÉêë=J=sçäK=fs
Table 2.10 Estimates for the flow parameters of the Wichita pump test
Graphical evaluation (Langguth and

Parameter PEST in FEFLOW
Voigt12)
(95 % confidence
intervals in
with measured with corrected
Name Unit brackets)
drawdown drawdown
conductivity [10-4 ms-1] 23 28 27 (15;47)
storage compressibility [-] 0.44 0.30 0.34 (0.05;24)
Figure 2.7 Measured observation points at t = 18 d for both the northern row (open sym-
bols) and the southern row (full symbols) of gauges, predicted gauge curves only for the
northern row.
cbcilt=ö=TT
OKOKQ cäççÇï~îÉ=éêçÄäÉã 2.8. At the right edge a time-dependent boundary con-

dition of the 3rd kind has been applied. The floodwave
In this example a passing floodwave in a river is is modeled by a linear rise of the water table in the river
used to estimate the hydraulic properties of the river of 3 m during one day, followed by a subsequent drop
bed and the connected aquifer from the response of the to the initial value on the next day. It is shown in Fig.
hydraulic head. In FEFLOW a (quasi-) 1D problem has 2.9. At the left edge a no-flow boundary condition is
been set up for an unconfined aquifer of 10 m thickness applied.
on a finite element grid of 100 m length, using 100
4-noded quadrilateral mesh elements as shown in Fig.
Figure 2.8 Finite element grid for the floodwave problem with the location of the observation points 1 (at 1 m), 2 (at 5
m) and 3 (at 10 m); both shape and position of the time-dependent floodwave boundary condition of the 3rd kind are
indicated.
The observed gauge curves at the observation points in various combinations as indicated in Tab. 2.11. Note,
1 at 1 m, 2 at 5 m and 3 at 10 m distance from the right that the stationary estimation problem (for a confined
edge were calculated with the values of 10-5 ms-1, 5 d-1 aquifer) is ill-posed since only a head boundary is pre-
and 0.2 for the three material parameters conductivity, scribed. Now the conductivity is only defined up to an
transfer coefficient and storativity. They are shown in integration constant and cannot be estimated (Sun13,
Fig. 2.9 for the first ten days of the simulation. The Kaiser11). However, in a transient problem the conduc-
actual simulation time was 50 d where for each obser- tivity alone can be estimated with a relative error of
vation point 59 time steps have been recorded. Here to one percent (Tab. 2.11). In the estimation run for the
the transfer coefficients for inflow and outflow identi- transfer coefficient alone the point estimate exceeds the
cal values were always assigned. The initial head was true value by more than ten percent. Moreover, the true
set to zero at all nodes. With growing distance from the value is not included in the 95 % CI of the point esti-
river boundary, the maximum of the hydraulic head mate. If the conductivity and the transfer coefficient are
appears with a certain delay and the response to the estimated together the accuracy of the point estimate
floodwave becomes broader and weaker. for the conductivity remains unchanged. But for the
transfer coefficient the relative error is reduced to three
With PEST the conductivity, the transfer coefficient percent. When the conductivity and the storativity are
and the storativity have been estimated separately and estimated together, both point estimates fall signifi-
TU=ö=tÜáíÉ=m~éÉêë=J=sçäK=fs
cantly below the true values. When all three material

parameters are estimated the results become even
worse. Obviously, for this combination of parameters
there exists an issue with uniqueness.
Figure 2.10 Objective functions for the PEST optimization

runs to estimate the transfer coefficient alone (crosses), the
conductivity alone (triangles), the transfer coefficient and
conductivity together (diamonds), the conductivity and stor-
ativity together (squares), and all three parameters conduc-
tivity, storativity and transfer coefficient together (circles).
Figure 2.9 Hydraulic head curves at the right floodwave
boundary and at the observation points 1, 2 and 3 with dis-
tances of 1 m, 5 m and 10 m.
cbcilt=ö=TV
Table 2.11 Input parameter and parameter estimates of PEST in FEFLOW for the floodwave
problem, 95 % CI in brackets
Conductivity Transfer coefficient Storativity Objective function

[10-5 m s-1] [d-1] [-] [10-3 m2]
input values 1.0 5.0 0.2 n.a.
parameter estimates
conductivity 1.01 (1.00;1.02) 5.0 fixed 0.2 fixed 3.37
transfer coefficient 1.0 fixed 5.61 (5.55;5.66) 0.2 fixed 3.49
conductivity & transfer 1.01 (1.00;1.02) 5.16 (4.89;5.44) 0.2 fixed 3.48
coefficient
conductivity & storativity 0.754 (0.737;0.771) 5.0 fixed 0.149 (0.147;0.151) 3.51
all 3 parameters 0.481 (0.467;0.496) 2.49 (2.25;2.75) 0.095 (0.092;0.098) 3.63
OKP `çåÅäìÇáåÖ=oÉã~êâë various graphical evaluation methods based on Theis'

analytical solution (Langguth and Voigt12). Owing to
The ability of the PEST software package to esti- the simplification of the real hydrogeological setting,
mate material parameters of FEFLOW flow problems which produces unknown systematic errors, the pre-
has been tested with four simple examples. dicted parameter values show a higher variability. This
variability is reflected in the point estimates from both
For Theis' analytical problem the transmissivity and the PEST optimization process and the graphical evalu-
the storage coefficient have been estimated with excel- ation methods in a similar way. The agreement of the
lent precision. estimated parameter values is better for the transmis-
sivity than for the storage compressibility. Both estima-
In the Breyell pump test for a confined aquifer tion methods produced parameter estimates with
recorded drawdown data has been used to determine sufficient accuracy.
the transmissivity and the storage compressibility with
UM=ö=tÜáíÉ=m~éÉêë=J=sçäK=fs
OKP=`çåÅäìÇáåÖ=oÉã~êâë
The Wichita pump test has been conducted in an inant mass and heat transport processes. Release 5.2, Reference
unconfined aquifer and has been evaluated by Lang- Manual, WASY Ltd., Berlin, 2005.
6. Doherty, J., Brebber, L. and Whyte, P. PEST - model independent
guth and Voigt12 using the same approach as for an con-
parameter estimation, User's Manual, Watermark Computing,
fined aquifer with and without corrected drawdown. Corinda, Australia, 1994.
With PEST it could be shown that the evaluation with 7. Dupuit J. Etudes thèoriques et pratiques sur le mouvement des
the corrected drawdown produced a more accurate eaux dans les canaux dècouvertes et à travers les terrains per-
result. In this example the use of the graphical method mèables. 2nd edition., Dunod, Paris, 1863.
has been pushed to the limit of applicability, whereas 8. Gründler R. Interface manager - extensions and programming
interface for FEFLOW, WASY Ltd, Berlin, 2005.
with PEST the Wichita pump test has been evaluated
9. Jacob, C.E. The recovery method for determining the coefficient
straight forwardly. of transmissibility. In: Bentall, R., Methods of determining per-
meability, transmissibility and drawdown, Geol. Survey Water-
The numerical floodwave problem for the estima- Supply Paper 1536-I, 283-292, 1963.
tion of three material parameters showed that there was 10. Jacob, C.E. Determining the permeability of water-table aquifers.
no unique set of parameters to reproduce the measured In: Bentall, R., Methods of determining permeability, transmissi-
bility and drawdown, Geol. Survey Water-Supply Paper 1536-I,
curves of the hydraulic head. Although the deviations
245-271, 1963.
from the true parameters are not excessive this example 11. Kaiser, J.C. Kopplung von FEFLOW mit dem Programm PEST
demonstrated that the estimation of material parame- zur Parameterkalibrierung (Coupling FEFLOW to the program
ters with PEST does not automatically lead to reliable PEST for parameter calibration). In: 3. Fachtagung ’Grafik-
results. gestützte Grundwassermodellierung’, Conference proceedings,
WASY Ltd, Berlin, Germany, 51-65, 1998.
12. Langguth, H.-R. and Voigt, R. Hydrogeologische Methoden. in
German, Springer, Berlin, 1980.
oÉÑÉêÉåÅÉë 13. Sun, N.-Z. Inverse problems in groundwater modeling. Kluwer
Academics Publishers, Dordrecht, 1994.
1. Bouttier, F. and Courtier P. Data assimilation concepts and meth- 14. Theis C.V. The relation between lowering of the piezometric sur-
ods, Meteorological Training Course Lecture Series, Lecture face and the rate and duration of discharge of a well using
Notes of the European Centre for Meteorological Weather Fore- groundwater storage. Trans. Amer. Geophys. Union 16 (1935),
cast (ECMWF), 2002. 519-524.
2. Carslaw, H.S. and Jaeger, J.C. Conduction of heat in solids. 2nd
ed. Oxford: University Press, 1959.
3. Cooper, H.H. and Jacob, C.E. A generalized graphical method for
evaluating formation constants and summarizing well-field his-
tory. Trans. Am. Geoph. Union 27 (1946), 526-534.
4. Diersch, H.-J.G. Interactive, graphics-based finite-element simu-
inant mass and heat transport processes. Release 5.2, User’s
Manual, WASY Ltd., Berlin, 2005.
5. Diersch, H.-J.G. Interactive, graphics-based finite-element simu-
cbcilt=ö=UN
UO=ö=tÜáíÉ=m~éÉêë=J=sçäK=fs
_ÉåÅÜã~êâáåÖ= î~êá~ÄäÉJÇÉåëáíó= Ñäçï= ~åÇ= íê~åëJ
éçêí=áå=éçêçìë=ãÉÇá~=Åçåí~áåáåÖ=~å=áåÅäáåÉÇ=Ñê~ÅJ
íìêÉ
H.-J. G. Dierscha & F. Cornatonb
P
a
b
Centre d’Hydrogéologie, Université de Neuchâtel, Switzerland
single fractures of arbitrary inclination embedded in a
P _ÉåÅÜã~êâáåÖ=î~êá~ÄäÉJÇÉåëáíó=Ñäçï=~åÇ=íê~åëéçêí=áå=éçêçìë=ãÉÇá~=Åçåí~áåáåÖ=~å=áåÅäáåÉÇ=Ñê~ÅíìêÉ
PKN fåíêçÇìÅíáçå low-permeable porous matrix. They simulated the

problems by using the control volume finite element
The physical basis, modeling strategies and numeri- code FRAC3DVS9 for 2D schematizations. Their
cal solutions of variable-density flow and transport results are well documented and appropriate for model
processes in porous media are thoroughly described testing. We shall benchmark the 45o-inclined fracture
and discussed in the recent review papers by Diersch problem against FEFLOW and the research code
and Kolditz2 as well as by Simmons et al.8. These Ground Water (GW) developed by F. Cornaton1.
papers are mainly focused on porous media, however,
fractured media are not of specific concern. Indeed,
there is a lack in studying density effects in fractured PKO s~êá~ÄäÉJÇÉåëáíó=cäçï=áå=~
media, particularly in single fractures. Laboratory mçêçìë=j~íêáñ=ïáíÜ=~=QRçJ
experiments in fractured media have not been carried fåÅäáåÉÇ=cê~ÅíìêÉ
out. Previous numerical investigations7 were limited to
a regular fracture network consisting of only vertical PKOKN pí~íÉãÉåí=çÑ=íÜÉ=Oa=éêçÄäÉã
and horizontal fractures, embedded in a porous matrix.
Shikaze et al.7 have shown that density plumes in a The single fracture problem is shown in Fig. 3.1.
fracture network may develop in a highly irregular The inclined fracture is discretized by 1D inclined pipe
fashion and are extremely difficult to predict. But, the fracture elements. The left and right boundaries are
growth of instabilities in single arbitrarily inclined assumed to be impermeable. The top and bottom
fractures, located in a low-permeability porous matrix, boundaries are modeled as open boundaries with a con-
remained open to be solved. stant hydraulic head h (set to zero). A contaminant
source of constant concentration C = Cs overlies
Recently, Graf and Therrien5 have presented a groundwater of initial concentration C = Co , where
numerical study on density-driven solute transport in C o = 0.0 < C s = 1.0 .
cbcilt=ö=UP
PK=_ÉåÅÜã~êâáåÖ=î~êá~ÄäÉJÇÉåëáíó=Ñäçï=~åÇ=íê~åëéçêí=áå=éçêçìë=ãÉÇá~=Åçåí~áåáåÖ=~å
áåÅäáåÉÇ=Ñê~ÅíìêÉ
dictor-corrector Adams-Bashforth/trapezoid rule (AB/
–4
contaminant
source
C = Cs h = 0 TR) time stepping2 with a RMS tolerance error of 10 .
No upwinding is preferred in all simulations.
gravity, g
Table 3.1 Model parameters
no flow no flow
observation Quantity5 Value
point discretized
1D fracture
------ = 0
10 m
------ = 0
elements Width 12 m
∂h
∂n
∂h
∂n
Initials:
Height 10 m
6m
------- = 0
fracture
------- = 0
φ = 45ο h ( x, t = 0 ) = 0
Simulation time 20 yr
∂C
∂n
∂C
C ( x, t = 0 ) = C o
∂n
Freshwater density, ρ o 1000 kg m-3
12 m Maximum fluid density, 1200 kg m-3

∂C ρ ( Cs )
------- = 0 h = 0
∂n 4
Fluid dynamic viscosity, μ o 3.545 ⋅ 10 kg m-1 yr-1
o
Figure 3.1 Single 45 -inclined fracture in a porous – 25
matrix; 2D geometry, boundary and initial conditions. Fluid compressibility, α fl 4.42 ⋅ 10 kg-1 m yr2
– 24
Matrix compressibility, α m 2.51 ⋅ 10 kg-1 m yr2
The simulations cover a time of 20 years. The 15
model parameters are summarized in Table 3.1, where Gravitational acceleration, g 9.75 ⋅ 10 m yr-2
FEFLOW-specific symbolic3,4 is used. It is assumed Tortuosity, τ 0.1
that the porous matrix is isotropic and homogenous and – 15
that the entire aquifer is completely saturated. The Matrix permeability, k 10 m2
parameters in FEFLOW units can be found in Table Matrix porosity, ε 0.35
3.2.
Matrix longitudinal dispersiv- 0.1 m
ity, β L
Graf and tested different fracture slopes φ
Therrien5
and mesh refinement levels ( l = 1, …, 5 ) . The present Matrix transverse dispersiv- 0.005 m
study focusses on the 45o-inclined fracture problem at ity, β T
the highest grid level l = 5 , consisting of 12,221 nodes
Fracture dispersivity, β L 0.1 m
and 24,000 triangles as shown in Fig. 3.2. We use two
time stepping strategies: (1) in agreement to Graf and Fracture aperture, b 50 μm
Therrien5 a fully implicit time step marching scheme
Free-solution molecular diffu- 0.15768 m2 yr-1
(combined with a Picard iteration) with a constant time
sion coefficient, D
step length of 0.2 yr, (2) alternatively, an adaptive pre-
UQ=ö=tÜáíÉ=m~éÉêë=J=sçäK=fs
PKO=s~êá~ÄäÉJÇÉåëáíó=cäçï=áå=~=mçêçìë=j~íêáñ=ïáíÜ=~=QRçJfåÅäáåÉÇ=cê~ÅíìêÉ
Table 3.2 FEFLOW parameters
Quantity Value
Density ratio, α 0.2

–5
Bulk compressibility, S o 1.743 ⋅ 10 m-1
–9
Matrix conductivity, 8.7216 ⋅ 10 m s-1
K = ( kρ o g ) ⁄ μ o
Matrix porosity, ε 0.35

– 10
Matrix diffusion coefficient, 5 ⋅ 10 m2 s-1
Dd
Matrix longitudinal dispersiv- 0.1 m
ity, β L
Matrix transverse dispersiv- 0.005 m Figure 3.2 2D triangular finite element mesh with 1D pipe
ity, β T fracture elements used for FEFLOW and GW1 simulations.
Fracture dispersivity, β L 0.1 m
–5
Fracture area, bB 5 ⋅ 10 m2
PKOKO dçîÉêåáåÖ=Éèì~íáçåë
–6
Fracture compressibility, S 4.4 ⋅ 10 m-1
The porous medium and fracture flow are transient.
–5
Hydraulic aperture, b corr a) 5.374 ⋅ 10 m Accordingly, the conservation of mass and momentum
–9 can be described by the following equations (symbols
Fracture diffusion coefficient, 5 ⋅ 10 m2 s-1
are summarized below under Notation):
D
a) Aperture b has to be corrected by the factor f ⁄ f o = 1.0747952 due ∂h
S ------ + ∇ ⋅ q = Q EOB (3-1)
to a different viscosity magnitude, where f = ρg ⁄ μ and ∂t
f o = ρ o g ⁄ μ o = 7.55 ⋅ 10 m s , see Appendix D in4.
6 –1 –1
q = – K ( ∇h + αC∗ e ) (3-2)
The inclined fracture is modeled by 100 1D pipe with
fracture elements fitted to the edges of the correspond-
ing triangular elements (Fig. 3.3). For the flow in the ⎧ So bulk compressibility in the porous matrix
fracture the Hagen-Poiseuille law is applied. Fluid vis- S = ⎨ (3-3)
⎩S compressibility in the fracture
cosity is considered independent of the concentration
μ = μ o = const. and
cbcilt=ö=UR
kρ o g⎞ where
⎧⎛
⎪ ⎝ K = -----------μo ⎠
- I Darcy law in the porous matrix
⎪ q = εv (3-9)
K = ⎨ 2 (3-4)
⎪ b corr ρ o g and
⎪ -------------------- I Hagen-Poiseuille law in the fracture
⎩ 12μ o
⎧ε porous matrix
ε = ⎨ (3-10)
where ⎩1 fracture
C – Co
C∗ = ------------------ (3-5) Note that the term Q EOB C in the solute transport equa-
Cs – Co
tion (3-7) is commonly negligible6 even in the case of
using the EOB approximation QEOB in the flow equa-
corresponds to a normalized solute concentration. Note tion (3-1).
that in (3-1) the term of Extended Oberbeck-Boussinesq
(EOB) approximation Q EOB is included, which is
defined as
PKOKP páãìä~íáçå=êÉëìäíë
∂C∗
Q EOB = – α ⎛ ε ---------- + q ⋅ ∇C∗⎞ (3-6) For the 45o-inclined fracture problem the results
⎝ ∂t ⎠
obtained by Graf and Therrien5 and by FEFLOW in
form of computed concentration distributions as well
If the standard Oberbeck-Boussinesq (OB)
approximation2 is applied, which is in agreement to the as velocity fields and pathline patterns at 2, 4 and 10 yr
assumptions done by Graf and Therrien5 , we have to simulation time are shown in Fig. 3.3. It reveals how
solve with Q EOB ≈ 0 . the solutes migrate from the fracture into the adjoining
porous matrix mainly governed by hydrodynamic dis-
The solute transport of a conservative tracer in the persion and to a small degree by convection.
porous matrix and the fracture is governed by the
convection-dispersion equation (solute mass balance) in
the following form: As a typical feature of the problem two convection
cells form above and below the fracture with increasing
∂C extent in time. Both cells move downward in time.
ε ------- + q ⋅ ∇C – ∇ ⋅ ( D ⋅ ∇C ) + Q EOB C = 0 (3-7)
∂t Note that the cell above the fracture moves faster
downward than the lower cell. Both convection cells
with the tensor of hydrodynamic dispersion remain separated by the high-conductive fracture,
therefore, acts as a barrier to convection.
⊗ q-
⎧ ( εD d + β T q )I + ( β L – β T ) q------------ porous matrix
⎪ q
D = ⎨ (3-8)
⊗ q-
⎪ DI + β q------------ fracture
⎩ L q
US=ö=tÜáíÉ=m~éÉêë=J=sçäK=fs
PKO=s~êá~ÄäÉJÇÉåëáíó=cäçï=áå=~=mçêçìë=j~íêáñ=ïáíÜ=~=QRçJfåÅäáåÉÇ=cê~ÅíìêÉ
Graf & Therrien5 FEFLOW
2 years
4 years
10 years
Figure 3.3 Computed concentration distributions and velocity/pathline field after 2, 4 and 10 yr simulation. Comparison of FEFLOW results obtained
by an AB/TR time stepping (right) to findings by Graf and Therrien5 modeled by a fully implicit constant time stepping (left). OB approximation is used.
cbcilt=ö=UT
At a first glance, FEFLOW and FRAC3DVS agree
very well. However, as already seen in Fig. 3.3 the 1.0
advance of solute transport in the fracture seems

slightly faster at early times in the FRAC3DVS predic- 0.8
tions compared to the FEFLOW results. Indeed, this

can be confirmed if evaluating precisely the break-
Concentration [-]
0.6
through curves of solute at the observation point as
shown in Fig. 3.4. While the FEFLOW curves for
Graf & Therrien
adaptive time stepping (taking 236 steps) and for con- 0.4
GW adapt
stant time steps (100 implicit steps with each of 0.2 yr FEFLOW const
FEFLOW adapt
length) provide reasonably close solutions, Graf and 0.2
Therrien’s breakthrough curve is apparently advanced
at early times. Due to the high velocity contrasts
between matrix and fracture, the influence of early 0.0
0 5 10 15 20
times on the spreading of solute in the depth is crucial Time [yr]
and requires further model comparisons. Figure 3.4 Breakthrough curves at the observation point.
Comparison of Graf & Therrien’s results5 to GW1 (with
adaptive time stepping) and FEFLOW (with constant and
The problem was also simulated by using the GW adaptive time stepping) in using OB approximation.
finite-element simulator1. The GW results provide a
nearly perfect agreement with the FEFLOW predic- 0.20
tions (conf. Figs. 3.4, 3.5 and 3.6). As evidenced in Fig.
3.4 FEFLOW’s and GW’s breakthrough curves are
very close. This could be confirmed by using both 0.15
adaptive and constant time stepping strategies. Note
Hydraulic head [m]

further that the type of solving the resulting sparse
equation systems did not influence the outcome. Direct 0.10
and iterative equation solvers were tested in FEFLOW.
GW
Additionally, the Extended Boussinesq approximation FEFLOW
(EOB)2 is also performed. As indicated in Fig. 3.7 the 0.05
breakthrough curve for the EOB is slightly shifted in

advance compared to FEFLOW’s OB solution, how-
ever, remains further behind Graf and Therrien’s OB 0.00
0 5 10 15 20
solution. Time [yr]
Figure 3.5 History of hydraulic head h at the observation

point. FEFLOW versus GW1 results (adaptive time step-
ping) in using OB approximation.
UU=ö=tÜáíÉ=m~éÉêë=J=sçäK=fs
PKP=pìãã~êó=~åÇ=`çåÅäìëáçåë
PKP pìãã~êó=~åÇ=`çåÅäìëáçåë
The 45o-inclined fracture benchmark problem
firstly proposed and simulated in Graf and Therrien5 by
using FRAC3DVS9 has been recomputed by using
FEFLOW and, additionally, by the research code GW1.
The major outcome of the comparison analysis is the
following:
• In principle, Graf and Therrien’s results5 could be

confirmed.
• There is a nearly perfect agreement between
FEFLOW and GW solutions.
• FRAC3DVS produces a clearly advanced solute
breakthrough in the fractures at early times com-
Figure 3.6 Computed solute concentration contours at 15 pared to the FEFLOW and GW simulations.
yr: FEFLOW versus GW1 results in using OB approxima-
tion.
It has been found that the discrepancies between
1.0
FRAC3DVS and FEFLOW or GW are not attributed to
different time stepping strategies, Boussinesq approxi-
mations and different sparse matrix solvers. Further-
0.8
more, more spatially refined meshes did not change
notably anymore the solutions because the mesh con-
vergence is practically achieved at the analyzed mesh
Concentration [-]
0.6
refinement level.
0.4 Graf & Therrien, OB
FEFLOW OB
FEFLOW EOB
PKQ kçí~íáçå
0.2
Roman letters
0.0
0 5 10 15 20
Time [yr] B L thickness;
Figure 3.7 Breakthrough curves at the observation point. b L aperture of fracture;
Comparison between OB approximation and EOB approxi- b corr L = b f ⁄ f o , corrected aperture of
mation. Adaptive time stepping is used for FEFLOW’s fracture;
solutions.
cbcilt=ö=UV
–3
C ML solute concentration; Greek letters
C∗ 1 normalized solute concentration;
–3
Co ML initial or reference solute α 1 = ( ρ ( C s ) – ρ o ) ⁄ ρ o , density ratio;
concentration; –1 2
α fl M LT fluid compressibility coefficient;
–3 –1 2
Cs ML maximum or boundary solute con- αm M LT matrix compressibility coefficient;
centration;
2 –1 β L, β T L longitudinal and transverse disper-
D L T tensor of hydrodynamic dispersion; sivity, respectively;
2 –1
D L T free-solution molecular diffusion ε 1 porosity of porous matrix;
coefficient;
2 –1 ε 1 generalized void space;
Dd L T = τD , medium molecular diffu- –3
ρ ML fluid density;
sion; –3
ρo ML = ρ ( C o ) , reference fluid density;
e 1 = –g ⁄ g , gravitational unit –1 –1
vector; μ ML T dynamic viscosity of fluid;
–1 –1
f L T
–1 –1
= ρg ⁄ μ , relation factor; μo ML T = μ ( Co ) , reference dynamic
–1 –1 viscosity of fluid;
fo L T = ρ o g ⁄ μ o , reference relation
factor; τ 1 tortuosity of porous medium;
g LT
–2
gravity vector; φ ° fracture slope angle;
–2
g LT = g , gravitational acceleration;
h L hydraulic head; Abbreviations
I 1 unit (identity) tensor;
K LT
–1
generalized hydraulic conductivity 1D one-dimensional;
tensor; 2D two-dimensional;
–1 EOB extended Oberbeck-Boussinesq
K LT hydraulic conductivity;
k L
2
permeability of porous medium; approximation;
l mesh refinement level, l = 1, … ; OB Oberbeck-Boussinesq approxima-
–1 tion;
Q EOB T extended Oberbeck-Boussinesq
approximation term; RMS root-mean square;
–1
q LT specific flux vector;
–1
S L = α fl ρ o g , fluid compressibility;
So L
–1
= [ εαfl + ( 1 – ε )α m ]ρ o g , bulk oÉÑÉêÉåÅÉë
compressibility;
–1 1. Cornaton, F. Ground Water - a 3-D ground water flow and trans-
S L generalized compressibility: port finite element simulator. Centre for Hydrogeology, Univer-
–1
v LT fluid velocity vector; sity of Neuchâtel, Neuchâtel, Switzerland, 2006.
2. Diersch, H.-J.G. and Kolditz, O. Variable-density flow and trans-
port in porous media: approaches and challenges. Adv. Water
VM=ö=tÜáíÉ=m~éÉêë=J=sçäK=fs
PKQ=kçí~íáçå
Resour. 25 (2002), 899-944. (also published in FEFLOW White

Papers Vol. II, Chapter 1, pp. 5-101).
3. Diersch, H.-J.G. FEFLOW Reference manual, WASY Ltd. 2002,
Berlin.
4. Diersch, H.-J.G. Discrete feature modeling of flow, mass and heat
transport processes by using FEFLOW. FEFLOW White Papers
Vol. I, Chapter 9, WASY Ltd. 2005, Berlin, pp. 149-195.
5. Graf, T. and Therrien, R. Variable-density groundwater flow and
solute transport in porous media containing nonuniform discrete
fractures. Adv. Water Resour. 28 (2005), 1351-1367.
6. Kolditz, O., Ratke, R., Diersch, H.-J.G. and Zielke, W. Coupled
groundwater flow and transport: 1. Verification of variable-den-
sity flow and transport models. Adv. Water Resour. 21 (1998), 27-
46.
7. Shikaze, S.G., Sudicky, E.A. and Schwartz, F.W. Density-depen-
dent solute transport in discretely-fractured geologic media: is
prediction possible? J. Contaminant Hydrol. 34 (1998)10, 273-
291.
8. Simmons, C.T, Fenstemaker, T.R. and Sharp Jr, J.M. Variable-
density groundwater flow and solute transport in heterogeneous
porous media: approaches, resolutions and future challenges. J.
Contaminant Hydrol. 52 (2001)11, 245-275.
9. Therrien, R. and Sudicky, E.A. Three-dimensional analysis of
variably saturated flow and solute transport in discretely-frac-
tured porous media. J. Contaminant Hydrol. 23 (1996), 1-44.
cbcilt=ö=VN
VO=ö=tÜáíÉ=m~éÉêë=J=sçäK=fs
kìãÉêáÅ~ä= ëáãìä~íáçå= çÑ= î~êá~ÄäÉJÇÉåëáíó= ãìäíáJ
ÇáÑÑìëáîÉ=ÑáåÖÉêáåÖ=ÅçåîÉÅíáçå=éÜÉåçãÉå~
H.-J. G. Diersch & V. Clausnitzer

Q
cases: (1) If the slower diffusing species is on top, a

Q kìãÉêáÅ~ä=ëáãìä~íáçå=çÑ=î~êá~ÄäÉJÇÉåëáíó=ãìäíáÇáÑÑìëáîÉ=ÑáåÖÉêáåÖ=ÅçåîÉÅíáçå=éÜÉåçãÉå~
QKN fåíêçÇìÅíáçå parcel of fluid perturbed downward across the interface

of the two species takes on mass from the surrounding
Fingering instabilities can play an important role in faster diffusing species so that the parcel continues to
the spreading of heavy leachates with or without tem- fall. Likewise, parcels of fluid perturbed upward con-
perature influences. Density-driven convection pro- tinue to rise. This mode of convection is termed dou-
cesses in the subsurface are of growing concern in ble-diffusive finger convection (DDFC) or multispecies
relation to contamination and geothermal problems. finger convection and is characterized by patterns
There is both a theoretical and a practical need to driven by long, narrow columns (fingers) of rising and
model variable-density flow, multispecies mass and falling fluid. (2) Alternatively, if the species with the
heat transport phenomena with multiple buoyancy and higher diffusivity is on top, a parcel of fluid perturbed
multidiffusive effects (see Diersch and Kolditz5). downward across the interface diffuses mass outward
to the surrounding fluid more rapidly than it gains mass
A striking feature of such processes is that convec- from the lower-diffusivity species. The parcel of fluid
tion can occur in initially density-stable configurations then becomes less dense than the surrounding fluid,
containing two or more components (solutal species, moves upward, and overshoots its original position
heat) with diffusivities that make opposing contribu- before repeating the motion. This mode of convection
tions to vertical density gradients. A basic requirement is termed oscillatory double-diffusive convection and
is in diffusivities differences of the components which can lead to well-mixed convecting layers separated by
can result in buoyant instabilities capable of initiating sharp contrasts in fluid density (e.g., forming staircase-
convection. The initial distribution and concentration type patterns5).
of each species (or temperature of heat) strongly influ-
ence such type of convective pattern. Consider the situ- In reality, convective fingers can also be created as a
ation where two solutions are layered with the denser result of a difference in diffusivities between heat and
solution on bottom, we can differ in two principal solute, which as been firstly detected in oceans. In this
cbcilt=ö=VP
QK=kìãÉêáÅ~ä=ëáãìä~íáçå=çÑ=î~êá~ÄäÉJÇÉåëáíó=ãìäíáJÇáÑÑìëáîÉ=ÑáåÖÉêáåÖ=ÅçåîÉÅíáçå=éÜÉJ
åçãÉå~
case double-diffusive finger convection is termed ther-
mohaline convection. While thermohaline convection Recently, the Hele-Shaw experiments collected by
is responsible on large-scale circulation in oceans, it Pringle et al.9 were successfully simulated by Hughes
has also been recognized that subsurface environments et al.7 using a modified version of the SUTRA code
(porous media and fractures) are favorable to DDFC. It (Voss and Provost11, Hughes and Sanford6) that com-
can be important at deep circulation in marine and ter- bines Galerkin finite-element and integrated finite-dif-
restrial alluvial basins, for interaction of groundwater ference techniques. Their modeling results have
and surface water, and in transport of dissolved solutes prompted our own numerical experiments based on the
from solid waste landfills. commercial finite-element simulator FEFLOW.
DDFC processes were studied by Cooper et al.2 and The data set of Pringle et al.9 is well-suited for code
Pringle et al.9 via Hele-Shaw experiments using a light verification of DDFC numerical models because,
transmission technique that provides high-resolution unlike most previous experimental Hele-Shaw data
concentration fields. From a near perturbation-free ini- sets, it is of sufficient spatial and temporal resolution to
tial layering of a lighter sucrose solution over a dense allow accurate comparisons of simulated and measured
salt solution, upward and downward moving fingers convective fingering. In addition, computational high-
quickly form at the interface between the two solutions. resolution results obtained by different numerical
Particularly, the recent experimental data obtained by approaches and full-field images from the experimental
Pringle et al.9 provide a suited baseline for use in the data set allow qualitative comparison of the evolving
development and evaluation of numerical models. flow field and quantitative comparison of mass transfer
rates.
Numerical models must play an increasing role in a
better understanding of DDFC phenomena in porous
media. As already argued by Cooper et al.2 a limitation QKO _~ëáÅ=bèì~íáçåë
in finger growth due to large-scale circulation con-
trolled by inertial forces as observed in ordinary fluids Following the experimental and numerical studies
(nonporous media) does not seem to exist in porous by Pringle et al.9 and Hughes et al.7, respectively, the
systems characterized by low Reynolds numbers. An focus is on DDFC phenomena with solutes. Accord-
intriguing possibility is that the merging and subse- ingly, variable-density multispecies convection pro-
quent formation of conduits along which fingers travel cesses involving multiple buoyancy and multidiffusive
could be repeated at larger and larger scales. Cooper et effects require the solution of the following coupled
al.2 concluded that larger and greater-spaced conduits equation system for fluid mass conservation, Darcy
for mass transport may naturally evolve in porous law, and mass balance assuming isothermal, nonreac-
media, leading to growth bounded on a much larger tive, and fully saturated conditions:
scale than has been observed in any laboratory experi-
ments.
VQ=ö=tÜáíÉ=m~éÉêë=J=sçäK=fs
QKP=cáåÖÉê=jçÇÉä=mêçÄäÉã
∂h ⎫
S ------ + ∇ ⋅ q = Q + Q EOB
∂t ⎪
⎪
ρ – ρo ⎪
q = – Kf μ ⎛ ∇h – --------------- e⎞ ⎬ (4-1)
⎝ ρo ⎠ ⎪
⎪
∂ ( εC ) + ∇ ⋅ ( qC ) – ∇ ⋅ ( D ⋅ ∇C ) = R ( k = 1, ..., N ) ⎪
k k k k k ⎭
∂t
with the constitutive relationships
αk μo μo ⎫
ρ = ρ 0 1 + ∑ ---------------------------
- ( C – C k0 ) f μ = -------------- ≈ ----------------------------------------------------
⎪
( C ks – C k0 ) k μ ( Ck )
k μ o + ∑ ϑ k ( C k – C k0 ) ⎪
⎪
k ⎬ (4-2)
d q⊗q ⎛ ⎞ ⎛ ∂C k⎞ ⎪⎪
Dk = ( εD k + β T q )I + ( β L – β T ) ------------- Q EOB = – q ⋅ ⎜ ∑ α k ∇C k⎟ – ε ⎜ ∑ α k ---------⎟ ⎪
q ⎝ k ⎠ ⎝ k ∂t ⎠
⎭
where h denotes hydraulic head, q the Darcy flux vec- respectively, t the time, and I the unit tensor. Sub-
tor, C k solute concentration of species k (with N as script 0 identifies reference values.
the maximum number of species), C ks is the maximum
concentration and Ck0 is the reference concentration of
species k, respectively, S the storage coefficient due to QKP cáåÖÉê=jçÇÉä=mêçÄäÉã
fluid and skeleton compressibility, Q the sink/source
term, Q EOB the Extended Oberbeck-Boussinesq term, Pringle et al.9 used a Hele-Shaw cell to study the
ρ the fluid density, K the tensor of hydraulic conduc- temporal and spatial distribution of DDFC phenomena
tivity, μ the dynamic fluid viscosity, f μ the viscosity of two solutes initially in a density-stable configuration
relation function, e the gravitational unit vector, ε the –3
with a mean interface thickness of about 1 ⋅ 10 m.
porosity, D k the tensor of hydrodynamic dispersion of The Hele-Shaw cell was filled with a sucrose solution
species k (assuming Fickian-based dispersive mass over a denser sodium chloride solution (NaCl). The 2D
flux), R k the reaction term, α k the density difference domain measures 0.2541 m (cell length L) by 0.1625 m
ratios (as dimensionless concentration expansion coef- (cell height H), see Fig. 4.1. The Hele-Shaw cell is
ficients), ϑ k the coefficients of fluid viscosity change, inclined at an angle of 25° relative to horizontal. To
d
D k the coefficients of molecular diffusivity of species visualize sodium chloride concentrations and quantify
k , β L, β T the longitudinal and transverse dispersivities, convective motion, a dye tracer with a low concentra-
cbcilt=ö=VR
åçãÉå~
tion was mixed with the sodium chloride. The dye had defined with the smaller diffusivity in the numerator
a negligible effect on fluid density. Accordingly, three such that 0 < Le < 1 .
species ( N = 3 ) have to be considered: sucrose
( k = s ), sodium chloride ( k = c ), and dye ( k = dye ). To maintain the full physical equivalence to the
The used parameters are summarized in Table 4.1. Note experimental and numerical studies done by Pringle et
that NaCl concentrations are not mapped perfectly by al.9 and Hughes et al.7 the most important physical
the dye tracer because the diffusivity of sodium chlo- quantities characterizing the DDC fingering problem
ride is approximately 2.5 times greater than the diffu- are the Turner number R ρ given by 1.22, the Lewis
sivity of the dye (Table 4.1). Because the motion is number Le according to 0.3303 and one Rayleigh
convective through most of the experiment, Pringle et number given for sodium chloride as Rac = 26,460 .
al.9 suggested the diffusivity differences had little The remaining quantities can be directly derived from
impact on the mapping of sodium chloride concentra- these characteristic numbers. Note that by using the
tions over the length of time of the experiment. dimensionless density difference ratios α k in the fluid
density ρ of (4-2), the density expansion becomes
The parameters in Table 4.1 are related to the independent of the real values of species concentra-
dimensionless Rayleigh numbers Rak (k = s, c), Lewis tions and the maximum concentrations C ks can be arbi-
number Le and Turner number R ρ as given in Table trarily chosen. In agreement to the physical experiment
4.2, where Rac = Ras Le R ρ . The Le number is C ks and C k0 are chosen as listed in Table 4.3
L = 0.2541 m
Sucrose solution (0.05234 kg/kg)

Dsucrose = 4.878 . 10-10 m2/s
x2
H = 0.1625 m
x1
Chloride solution (0.03463 kg/kg), dyed

DNaCl = 1.477 . 10-9 m2/s
to
ti
Aperture 1.77.10-4 m, inclination 25o Concentration of Concentration of Density ρ

sucrose chloride
Figure 4.1 Hele-Shaw study experiment by Pringle et al.9.
VS=ö=tÜáíÉ=m~éÉêë=J=sçäK=fs
Table 4.1 Parameters of the three-species finger Table 4.2 Rayleigh, Lewis and Turner numbers
model problem
Number Value
Parameter Value Unit
hydraulic conductivity, –2 –1 α s KH
1.07838 ⋅ 10 ms sucrose Rayleigh number, Ra s = --------------
d
65,664
K = kρ o ( g sin 25° ) ⁄ μ o εD s
–9 2 α c KH
intrinsic permeability, k 2.61 ⋅ 10 m chloride Rayleigh number, Rac = --------------
- 26,460
d
εD c
–3
reference density, ρ o 998.0 kg m
d d
–3 –1 –1 Lewis number, Le = D s ⁄ D c 0.3303
reference viscosity, μ o 10 kg m s
–1
storativity, S 0.0 m α
Turner number, R ρ = -----c 1.22
αs
–1
sucrose density expansion, α s 0.0182787 kg kg
–1
chloride density expansion, α c 0.022302 kg kg
–1
dye density expansion, α dye 0.0 kg kg Table 4.3 Maximum and reference
d – 10 2 –1 concentrations
sucrose diffusivity, D s 4.878 ⋅ 10 m s
d –9 2 –1
chloride diffusivity, Dc 1.477 ⋅ 10 m s Number Value Unit
d – 10 2 –1
dye diffusivity, D dye 5.670 ⋅ 10 m s
3 –3 –3
porosity, ε 1.0 m m sucrose concentration, C ss 52.34 kg m
dispersivity, β L, β T 0.0 m –3
chloride concentration, C cs 34.63 kg m
–3 2 –1
viscosity change to sucrose, ϑ s 2.75 ⋅ 10 m s
–3 2 –1 –3
viscosity change to chloride, ϑ c 1.59 ⋅ 10 m s dye concentration, C dyes 0.25 kg m
2 –1
viscosity change to dye, ϑ dye 0.0 m s
–3
reference concentrations, 0.0 kg m
C k0 ( k = s, c, dye )
cbcilt=ö=VT
åçãÉå~
It has been shown in stability analysis8 and Hele- Pringle et al.9 a total of 300 images of the evolving
Shaw experiments1 as the Turner number R ρ decreases concentration field were collected. A sequence of dye
–1
from the stability boundary at Le , the system transi- concentrations from the experiment is shown in Fig.
d 2
tions from being diffusion-dominated to convection- 4.3. Time is presented as dimensionless ( t∗ = tD c ⁄ H ) .
dominated. The corresponding stability and instability The measured time stages ( t and t∗ ) are listed in Table
domains in the Rayleigh parameter space are shown in 4.4.
Fig. 4.2. The current situation with a Turner number of
1.22 is clearly located in the DDFC domain with Table 4.4 Measured time stages
increasing mass fluxes and finger velocities. DDFC
–1
exists in the range 1 < R ρ < Le . For R ρ < 1 there is no t∗ t t
more an initially density-stable stratification of the sol- Stages
[-] [sec] [hr]
utes and the system becomes gravitationally instable.
–5
a Early stage 4.03 ⋅ 10 720.5 0.20
30000 –4
b 1.31 ⋅ 10 2342.1 0.65
UNSTABLE
Mature stage:
Gravitational instability –4
c vertical growth 2.21 ⋅ 10 3951.1 1.10
Pringle et al. (2002)
of fingers –4
d 3.22 ⋅ 10 5756.8 1.60
Line of neutral –4
20000
gravitational stability Rρ = 1
e Fingers reach 4.23 ⋅ 10 7562.5 2.10
top and bottom
–4
Ras Le
f boundaries 5.24 ⋅ 10 9368.2 2.60

Rρ = 1.22 –4
DOUBLE-DIFFUSIVE g 6.04 ⋅ 10 10798.5 3.00
FINGER CONVECTION –4
10000 h 7.25 ⋅ 10 12961.8 3.60
–4
i Roundown 7.85 ⋅ 10 14034.5 3.90
stage –3
STABLE j 1.03 ⋅ 10 18414.7 5.12
-1
Stability boundary Le –3
0 k 1.77 ⋅ 10 31644.6 8.79
30000 20000 10000 0
–3
Rac l 3.17 ⋅ 10 56674.2 15.7
Figure 4.2 Stability and instability domains in the Ray-

leigh parameter space with the location of the current dou-
ble-diffusive finger problem at Rρ = 1.22.
In the 16-hour Hele-Shaw experiments done by
VU=ö=tÜáíÉ=m~éÉêë=J=sçäK=fs
a b c
d e f
g h i
j k l
-5
Figure 4.3 Hele-Shaw observation results from Pringle et al.9 for the dye component at (a) t∗ = 4.03 ⋅ 10 , (b)
-4 -4 -4 -4 -4
t∗ = 1.31 ⋅ 10 , (c) t∗ = 2.21 ⋅ 10 , (d) t∗ = 3.22 ⋅ 10 , (e) t∗ = 4.23 ⋅ 10 , (f) t∗ = 5.24 ⋅ 10 , (g)
-4 -4 -4 -3 -3
t∗ = 6.04 ⋅ 10 , (h) t∗ = 7.25 ⋅ 10 , (i) t∗ = 7.85 ⋅ 10 , (j) t∗ = 1.03 ⋅ 10 , (k) t∗ = 1.77 ⋅ 10 , and (l)
-3 d 2
t∗ = 3.17 ⋅ 10 . ( t∗ = tD c ⁄ H dimensionless time). Color sequence black-blue-green-yellow-orange-red
depicts normalized dye concentration from 0 to 1.
cbcilt=ö=VV
åçãÉå~
As seen in a sequence of concentration fields in Fig. QKQ kìãÉêáÅ~ä=jçÇÉäáåÖ
4.3 there are interesting features in the behavior of the
DDC system. Due to the initially perturbed solution QKQKN pé~íá~ä=~åÇ=íÉãéçê~ä=ÇáëÅêÉíáò~J
interface an array of distinct fingers rapidly grow in
íáçåI= áíÉê~íáçå= ëíê~íÉÖó= ~åÇ
unison at the early time stage (Fig. 4.3a,b). These fin-
gers begin to interact with one another causing a re- ìëÉÇ=ëçäîÉêë
organization of the initial uniform finger structure (Fig. To solve successfully the DDFC problem a suffi-
4.3c,d). A typical feature at this stage is a large number ciently fine spatial discretization is fundamental.
of very small fingers with a wide variation in vertical Because most transfer in a DDFC system is a result of
extent. As convection proceeds, small-scale fingers convection, small finger dimensions may evolve.
continuously emerge from the region of the initial solu- Damping effects by artificial numerical dispersion
tion interface referred to as the finger generation zone should be hold down on a lowest level to resolve accu-
by Cooper et al.2. These newly generated fingers add to rately the minimum finger dimension occurring in a
the structural intricacy of the field by growing, and in DDFC simulation. A further important point in DDFC
many cases, merging with, and convecting up through computations refers to arising numerical perturbations
the stems of early formed neighbors. The generation of which can affect the evolution of DDFC5. It is to be
new finger pairs as the tips of some upward and down- expected that uniform and aligned structured meshes
ward growing fingers can also be observed (Fig. 4.3c- with square elements can minimize uncontrollable
–4
f). At t∗ = 4.23 ⋅ 10 (Fig. 4.3e), the fastest growing numerical perturbations during the simulation.
fingers reach the top and bottom boundaries of the cell
and begin to spread laterally forming more dense (at Quadrilateral meshes with different resolution were
the bottom) and less dense (at the top) ‘clouds’ of fluid studied by Hughes et al.7. It can be recognized as a
(Fig. 4.3f-h). Within the finger generation zone, far stepwise global refinement ϒ l of meshing according to
from the boundaries, new fingers continue to form
from isolated pockets of nearly pristine solution ϒl l = 0, 1, 2, … (4-3)
located about the initial solution interface (Fig. 4.3g-j).
Finally, at late time, the finger structure becomes ‘tree- where l is the refinement level. In the global refine-
like’ with a branching pattern that has greater lateral ment ϒ l of the mesh starting with the 2D discretization
travel than a early time. This final convective structure ϒ 0 each quadrilateral is subdivided into four equally
remains long after motion has stopped, diffusion now sized quadrilaterals. The number of quadrilaterals NE
acting to slowly uniformize the field (Fig. 4.3l). and number of nodes NP then increase according to the
refinement level l = 0, 1, … :
( 10 + 2l )
NE = 41 ⋅ 2
(4-4)
(2 + l)
NP = NE + 105 ⋅ 2 +1
NMM=ö=tÜáíÉ=m~éÉêë=J=sçäK=fs
QKQ=kìãÉêáÅ~ä=jçÇÉäáåÖ
be simulated by Hughes et al.7 due to their computa-

Table 4.5 summarizes the mesh properties up to level 4. tional limitations.
Table 4.5 Meshes according to refinement levels l In the present FEFLOW simulations we recompute
the DDFC problem in using Hughes et al.’s finest
Used Number of Number of Spatial 671,744-element mesh at l = 2 . Additionally,
Level
FEFLOW elements nodes increment FEFLOW simulations are performed on a further
l
mesh (NE) (NP) [mm] refined mesh having the refinement level l = 3 . In the
following FEFLOW simulations we denote these
meshes as mesh A consisting of 1024 . 656 quadrilat-
0 - 41,984 42,405 0.992
(256 . 164) (257 . 165) eral elements (673,424 nodes) and mesh B consisting
of 2048 . 1312 quadrilateral elements (2,690,337
nodes), see Table 4.5. Mesh A is comparable to the fin-
1 - 167,936 168,777 0.496
(513 . 329)
est spatial discretization used by Hughes et al.7. Note
(512 . 328)
that the high-resolution mesh B is more refined than
the length scales of in the Hele-Shaw experiment. The
2 mesh A 671,744 673,425 0.248 spatial increment in mesh B with 0.124 mm is smaller
(1024 . 656) (1025 . 657)
than the pixel size of the Hele-Shaw experiment with
0.154 mm.
3 mesh B 2,686,976 2,690,337 0.124
(2048 . 1312) (2049 . 1313)
It is important to note that mesh B requires 64-bit
execution because the needed RAM is on the order of 4
4 - 10,747,904 10,754,625 0.062 GB. In the present study we prefer a fully implicit for-
(4096 . 2624) (4097 . 2625) ward Euler/backward Euler (FE/BE) adaptive error-
controlled predictor-corrector time stepping strategy5
and parallel computations. A sequential iterative
Picard-type strategy for the flow and transport equa-
Hughes et al.7 simulated meshes at levels l of 0, 1
tions (4-1) is performed as thoroughly described in
and 2. Their computations with the finest mesh at
Diersch4. The adaptive time stepping and iterative pro-
l = 2 agreed rather well with Pringle et al.’s Hele-
cedure is controlled via a RMS error criterion using
Shaw experiments. They found that the coarser dis- –4
10 as dimensionless tolerance length. While the flow
cretizations with l = 0 and l = 1 are inappropriate to
equations are solved by the algebraic multigrid solver
model the finger development with a reasonable accu-
SAMG (Stüben10, Diersch3), the transport equations are
racy. However, even their finest discretization at l = 2
solved by an iterative BiCGSTAB equation solver with
with a spatial increment of 0.248 mm is still larger than
an incomplete LU preconditioning. Both solvers are
the pixel size with 0.154 mm of the Hele-Shaw experi- – 12
applied with a reduced stop criteria of 10 to termi-
ment by a factor of 1.6. More refined meshes could not
cbcilt=ö=NMN
åçãÉå~
nate iterations in solving the sparse finite-element QKQKO _çìåÇ~êó= ~åÇ= áåáíá~ä= ÅçåÇáJ
matrix equation systems. Table 4.6 summarizes the íáçåëI=éÉêíìêÄ~íáçå=çÑ=áåáíá~ä=ëçäJ
numerical features used in the FEFLOW simulations.
ìíÉ=áåíÉêÑ~ÅÉ
All external boundary faces represent no-flux con-
Table 4.6 Numerical features used in the ditions both for fluid flow and for species mass trans-
FEFLOW simulations port. This is automatically satisfied by natural (zero-
value) Neumann-type boundary conditions and no
Feature Description specifications are required. But, there is one exception.
Because the storativity S is zero in the flow equation
Spatial discretization Structured meshes A and B of quad- (4-1) there is no more a regular time-derivative term
rilateral bi-linear finite elements and the flow equations should be linked with a
Dirichlet-type boundary condition to stabilize the
Temporal discretiza- Implicit predictor-corrector FE/BE numerical solution. It is sufficient to specify at least
tion time stepping method with RMS one node with an arbitrary head value h. While Hughes
error criterion of 10–4 et al.7 specified both the upper left and upper right cor-
ner nodes with values for pressure and species concen-
Iteration method Sequential iterative Picard-type ter- trations, in our simulations only the node at the centre
minated with RMS error criterion of of the mesh is specified with a hydraulic head h of 0.0,
–4
10
no extra boundary conditions are introduced for the
species concentrations.
Iterative sparse Flow equations (symmetric):
matrix solvers SAMG with stop criteria 10–12
At initial time t o the three species are distributed as
Transport equations (unsymmetric):
BiCGSTAB with stop criteria 10–12 follows within the 2D domain ( 0 ≤ x 1 ≤ L, 0 ≤ x 2 ≤ H ) in
using incomplete LU precondition- a layered configuration, where the solute interface is
ing located at x 2 = 0 (see Fig. 4.1):
Parallelization Assembly of finite element matrices ⎧ H

OpenMP for SAMG equation solver ⎪ C ss ⎛ 0 ≤ x 1 ≤ L, 0 ≤ x 2 ≤ ----⎞
⎪ ⎝ 2⎠
Auto-parallelizing compiler options C s ( x, t o ) = ⎨ (4-5a)
⎪ ⎛ H-
used for BiCGSTAB equation
⎪ C s0 ⎝ 0 ≤ x 1 ≤ L, – --- ≤ x 2 < 0⎞⎠
solver 2
⎩
Address space 64-bit
NMO=ö=tÜáíÉ=m~éÉêë=J=sçäK=fs
QKR=oÉëìäíë
DO FOR ALL INTERFACE NODES I {

⎧ H
⎪ C c0 ⎛ 0 ≤ x 1 ≤ L, 0 < x 2 ≤ ----⎞ RN1 = RANDOM NUMBER BETWEEN 0 AND 1
⎪ ⎝ 2⎠
C c ( x, t o ) = ⎨ (4-5b) IF (RN1 < 0.5) {
I I
⎪ C ⎛ 0 ≤ x ≤ L, – H ---- ≤ x 2 ≤ 0⎞ C s ( x 1, x 2 = 0, t o ) = 0.01 ⋅ RN1 ⋅ C ss
⎪ cs ⎝ 1 2 ⎠ }
⎩
ELSE {
I I
C s ( x 1, x 2 = 0, t o ) = C ss
⎧ H
⎪ C dye0 ⎛ 0 ≤ x 1 ≤ L, 0 < x 2 ≤ ----⎞ }
⎪ ⎝ 2⎠
C dye ( x, t o ) = ⎨ (4-5c)
⎪C ⎛ H- RN2 = RANDOM NUMBER BETWEEN 0 AND 1
⎪ dyes ⎝ 0 ≤ x 1 ≤ L, – --- ≤ x 2 ≤ 0⎞
2 ⎠ IF (RN2 < 0.5) {
⎩ I I
C c ( x 1, x 2 = 0, t o ) = 0.01 ⋅ RN2 ⋅ C cs
}
The present finger convection problem is very sen- ELSE {
sitive with respect to perturbations. Pringle et al.9 I I
C c ( x 1, x 2 = 0, t o ) = C cs
expended significant effort in minimizing initial pertur- }
bations for the Hele-Shaw experiment. Although the }
thickness of the solute interface was small (about 1
mm), perturbations at the start of the experiment could I I
where x 1 and x 2 correspond to the x 1 - and x 2 -coordi-
not be avoided. They were seeds for initial finger nates of node I.
developments. For the numerical simulation a control
of such type of initial seeds for finger developments is To have comparable initial perturbations for both
needed. This should be mimicked by the following ran- mesh A and B the perturbation procedure at the inter-
dom procedure as proposed by Hughes et al.7. face nodes is firstly applied to mesh A. This perturba-
tion pattern is then inherited to mesh B when mesh A is
Random noise with a mean of zero and maximum globally refined from level l = 2 to level l = 3 by
amplitude of 0.5% of maximum initial concentrations subdivision (each quadrilateral element is subdivided
C ks is applied for both sucrose and sodium chloride at into four quadrilateral elements). However, in a last
the initial solution interface. Dye concentrations at the case we will deviate from this inherited perturbation
interface are not perturbed. To develop initial perturba- and present simulations with initial perturbations
tions for sucrose and sodium chloride their nodal con- directly generated on the refined mesh B.
centrations at nodes sharing the solute interface at
x 2 = 0 are modified as follows:
QKR oÉëìäíë
Table 4.7 summarizes the performance results
cbcilt=ö=NMP
åçãÉå~
attained in the FEFLOW simulations for meshes A and already indicated by Hughes et al.7 the experimental
B. While the run time of mesh A required about 1 day vertical finger evolution appears to be slightly ahead of
on a double processor machine (6 GB RAM, 3 GHz the simulated fingers.
clock speed, MS XP Enterprise x64 Edition), mesh B
took a run time of about one week. It is interesting to
note that the parallel-version of the SAMG solver have Table 4.7 FEFLOW performance results
accelerated the simulation time by a factor of about two
as indicated in Table 4.7. The simulation of mesh A and Number of
B required 3205 and 3626 adaptive time steps, respec- Mesh adaptive CPU timea)
tively. The time step history for the mesh B simulation time steps
is plotted in Fig. 4.4.
A 3205 ~1 day
10-4
(~2.5 days)b)
10-5
10-6 B 3626 ~7.5 days

(~16 days)b)
10-7
Δt*
10-8 a) Approximate total simulate time on a double

processor server with 3 GHz clock speed, 6 GB
10-9 RAM, MS Windows XP Enterprise x64 Edition,
two single-core Intel Xeon CPUs. Executable
10-10 and libraries are compiler-optimized using
options of auto-parallelization.
10-11 b) SAMG solver without OpenMP parallelization.
10-12 -11
10 10-10 10-9 10-8 10-7 10-6 10-5 10-4 10-3 10-2 A more quantitative comparison can be done by
t* using the vertical mass flux exemplified for the dye
Figure 4.4 Time stepping history for mesh B. concentrations. A normalized mass transfer of dye
upward across the centerline of the cell can be defined
according to
Figure 4.6 shows the FEFLOW-simulated dye con-
M∗ = M ⁄ M o (4-6)
centrations for mesh A at the same dimensionless times
of Pringle et al.9 (cf. Table 4.4). The results agree rather
well with the computations obtained by Hughes et al.7 where M is the dye mass above the centerline of the
as displayed in Fig. 4.7. Qualitatively, the numerical Hele-Shaw cell at time t∗ and M o is the total dye mass
results are similar to the experimental results as seen in in the cell. Numerical results compare reasonably well
Fig. 4.8 in comparison to the mesh B results. As to observed values of M∗ as depicted in Fig. 4.5. As
NMQ=ö=tÜáíÉ=m~éÉêë=J=sçäK=fs
QKR=oÉëìäíë
also seen there FEFLOW’s and Hughes et al.7’s results stant viscosity solution produces a slightly faster finger
agree very well. Their agreement with the Hele-Shaw development as for the case with a variable (concentra-
–3
experiment is acceptable until t∗ = 1 ⋅ 10 . After tion-dependent) viscosity.
–3
t∗ = 1 ⋅ 10 , the simulated mass transfer M∗ is less
than observed mass transfer. Larger percent errors at Table 4.8 Measured vs. simulated M*
early times are an artifact of small M∗ values and rep-
resent small absolute differences in mass transfer (e.g., M*
0.011 observed and 0.017 simulated) influenced by the
initial perturbation at the interface nodes for the given t* Hele-Shaw
spatial discretization. Note further that the simulated FEFLOW
experiment
mass transfer M∗ did not changed anymore if using a mesh B
Pringle et al.9
more refined mesh (cf. mesh A and mesh B results in
–5
Fig. 4.5). It indicates that the numerical accuracy with a 4.03 ⋅ 10 0.01 0.02
respect to the mass transfer is sufficiently achieved at a –4
b 1.31 ⋅ 10 0.05 0.05
lower refinement level as given for mesh A. Table 4.8
–4
compares the measured mass transfer M∗ against the c 2.21 ⋅ 10 0.10 0.08
FEFLOW results obtained for mesh B. –4
d 3.22 ⋅ 10 0.15 0.12
A comparison of the finger evolution for the two –4
e 4.23 ⋅ 10 0.20 0.16
meshes A and B are exhibited in Fig. 4.9. It reveals a –4
slightly faster finger development for the more refined f 5.24 ⋅ 10 0.25 0.20
mesh B compared to the coarser mesh A. While for g 6.04 ⋅ 10
–4
0.30 0.23
mesh A at the front of the fingers small wiggles in the
–4
numerical solution could be observed at early times h 7.25 ⋅ 10 0.35 0.28
(indicated by white color spots in the fringed distribu- –4
i 7.85 ⋅ 10 0.40 0.30
tions of Fig. 4.9 left), the solution for mesh B is fully
–3
wiggle-free. j 1.03 ⋅ 10 0.45 0.36
–3
We also studied the influence of the Oberbeck- k 1.77 ⋅ 10 0.50 0.41
Boussinesq approximation and the fluid viscosity on l 3.17 ⋅ 10
–3
0.51 0.42
the simulation results. Noticeable but not significant
differences exist in the simulated finger patterns when
comparing the solutions with and without the Ober- The influence of the initial perturbation on the fin-
beck-Boussinesq approximation as seen in Fig. 4.10. ger development can be seen in Fig. 4.12 resulting for a
More influence on the finger pattern results from the mesh B simulation. On the one hand, it shows the fin-
fluid viscosity effect. As evidenced in Fig. 4.11 a con- ger pattern evolving from a perturbation which has
cbcilt=ö=NMR
åçãÉå~
been inherited from the coarser mesh A. On the other remain comparable. In this case the growth of fingers
hand, it also exhibits the case of initial perturbations in the vertical extent is slightly slower because the per-
which has been directly generated on the refined mesh turbation of the interface nodes on the refined mesh has
B. As revealed the perturbation significantly affects the a smaller scale (nodal spacing is smaller) compared to
local finger structure of the convection system while the perturbation inherited from the coarser mesh (spac-
the average quantities, such as the mass transfer, ing of the perturbed nodes is larger).
Pringle et al.’s experimental and Hughes et al.’s numerical data FEFLOW results
100
M*
10-1
Mesh A
Mesh B
10-2
10-6 10-5 10-4 10-3 10-2
t*
Figure 4.5 Normalized mass transfer across the center line M*. Comparison of observed data taken by Pringle et al.9
and numerical results by Hughes et al.7 (left) with FEFLOW results computed for meshes A and B (right).
NMS=ö=tÜáíÉ=m~éÉêë=J=sçäK=fs
QKR=oÉëìäíë
a b c
d e f
g h i
j k l
-5
Figure 4.6 FEFLOW results simulated with mesh A for the dye component at (a) t∗ = 4.03 ⋅ 10 , (b)
-4 -4 -4 -4 -4
∗ ∗ ∗ ∗ ∗
t = 1.31 ⋅ 10 , (c) t = 2.21 ⋅ 10 , (d) t = 3.22 ⋅ 10 , (e) t = 4.23 ⋅ 10 , (f) t = 5.24 ⋅ 10 , (g)
-4 -4 -4 -3 -3
t∗ = 6.04 ⋅ 10 , (h) t∗ = 7.25 ⋅ 10 , (i) t∗ = 7.85 ⋅ 10 , (j) t∗ = 1.03 ⋅ 10 , (k) t∗ = 1.77 ⋅ 10 , and (l)
-3 d 2
t∗ = 3.17 ⋅ 10 . ( t∗ = tD c ⁄ H dimensionless time). Color sequence blue-green-yellow-orange-red depicts
normalized dye concentration from 0 to 1.
cbcilt=ö=NMT
åçãÉå~
-5
Figure 4.7 Simulated results by Hughes et al.7 for the dye component at (a) t∗ = 4.03 ⋅ 10 , (b)
-4 -4 -4 -4 -4
t∗ = 1.31 ⋅ 10 , (c) t∗ = 2.21 ⋅ 10 , (d) t∗ = 3.22 ⋅ 10 , (e) t∗ = 4.23 ⋅ 10 , (f) t∗ = 5.24 ⋅ 10 , (g)
-4 -4 -4 -3 -3
t∗ = 6.04 ⋅ 10 , (h) t∗ = 7.25 ⋅ 10 , (i) t∗ = 7.85 ⋅ 10 , (j) t∗ = 1.03 ⋅ 10 , (k) t∗ = 1.77 ⋅ 10 , and (l)
-3 d 2
t∗ = 3.17 ⋅ 10 . ( t∗ = tD c ⁄ H dimensionless time). Color sequence black-blue-green-yellow-orange-red
depicts normalized dye concentration from 0 to 1. Figure taken from Hughes et al.7.
NMU=ö=tÜáíÉ=m~éÉêë=J=sçäK=fs
QKR=oÉëìäíë
Hele-Shaw experiments (Pringle et al., 2002) FEFLOW with mesh B (NE = 2,686,976)
-4
t∗ = 1.31 ⋅ 10
-4
t∗ = 3.22 ⋅ 10
-4
t∗ = 7.25 ⋅ 10
0 1
Figure 4.8 Comparison of Hele-Shaw experiments from Pringle et al.9 to FEFLOW results simulated with mesh
d 2
B for the dye component at different times t∗ . ( t∗ = tD c ⁄ H dimensionless time). Case of variable fluid viscos-
ity f μ = μ o ⁄ μ ( C k ) . Color sequence black-blue-green-yellow-orange-red depicts normalized dye concentration
from 0 to 1.
cbcilt=ö=NMV
åçãÉå~
Mesh A (NE = 671,774) Mesh B (NE = 2,686,976)
-4
t∗ = 3.22 ⋅ 10
-4
t∗ = 7.25 ⋅ 10
-3
t∗ = 3.17 ⋅ 10
0 1
d 2
Figure 4.9 FEFLOW results simulated for the dye component at different times t∗ . ( t∗ = tD c ⁄ H dimension-
less time). Comparison between mesh A (left) and mesh B (right) for the case of variable fluid viscosity
f μ = μ o ⁄ μ ( C k ) . Color sequence black-blue-green-yellow-orange-red depicts normalized dye concentration
from 0 to 1.
NNM=ö=tÜáíÉ=m~éÉêë=J=sçäK=fs
QKR=oÉëìäíë
Oberbeck-Boussinesq approximation Extended Oberbeck-Boussinesq approximation
-4
t∗ = 3.22 ⋅ 10
-4
t∗ = 7.25 ⋅ 10
-3
t∗ = 3.17 ⋅ 10
0 1
Figure 4.10 FEFLOW results simulated with mesh A for the dye component at different times t∗ .
d 2
( t∗ = tD c ⁄ H dimensionless time). Comparison of Oberbeck-Boussinesq approximation Q EOB ≡ 0 (left) to the
Extended Oberbeck-Boussinesq approximation Q EOB ≠ 0 (right) for the case of constant fluid viscosity f μ ≡ 1 .
Color sequence black-blue-green-yellow-orange-red depicts normalized dye concentration from 0 to 1.
cbcilt=ö=NNN
åçãÉå~
Constant fluid viscosity Variable fluid viscosity
-4
t∗ = 3.22 ⋅ 10
-4
t∗ = 7.25 ⋅ 10
-3
t∗ = 3.17 ⋅ 10
0 1
Figure 4.11 FEFLOW results simulated with mesh A for the dye component at different times t∗ .
d 2
( t∗ = tD c ⁄ H dimensionless time). Comparison of constant fluid viscosity f μ ≡ 1 (left) to the variable fluid vis-
cosity case f μ = μ o ⁄ μ ( C k ) (right). Color sequence black-blue-green-yellow-orange-red depicts normalized dye
concentration from 0 to 1.
NNO=ö=tÜáíÉ=m~éÉêë=J=sçäK=fs
QKR=oÉëìäíë
Initial perturbation inherited from mesh A Initial perturbation generated on mesh B
-4
t∗ = 1.31 ⋅ 10
-4
t∗ = 3.22 ⋅ 10
-4
t∗ = 7.25 ⋅ 10
Figure 4.12 FEFLOW results simulated with mesh B for the dye component at different times t∗ .
d 2
( t∗ = tD c ⁄ H dimensionless time). Influence of perturbation applied to the initial solution interface. Case of
variable fluid viscosity f μ = μ o ⁄ μ ( C k ) . Color sequence blue-green-yellow-orange-red depicts normalized dye
concentration from 0 to 1.
cbcilt=ö=NNP
åçãÉå~
QKS `çåÅäìëáçåë cient simulator in 2D applications. Even 3D DDFC
applications seem to be possible in the future.
The FEFLOW simulator provides the capability to
simulate complex DDFC phenomena for fine mesh res-
olutions. For each species (and in addition temperature kçí~íáçå
in case of need) multiple density expansions and vis-
cosity change parameters can be specified. The viscos- Roman letters
ity relationship can be freely defined for species (and
–3
temperature) via a flexible formula editor. FEFLOW is C ML concentration;
parallelized and uses an efficient algebraic multigrid 2 –1
D L T tensor of hydrodynamic dispersion;
solver to speed-up the computations. D
d 2 –1
L T coefficient of molecular diffusion;
e 1 gravitational unit vector;
The FEFLOW simulations have been tested for a
fμ 1 fluid viscosity relation function;
2D DDFC problem against the Hele-Shaw experiment –2
g LT gravitational acceleration;
done by Pringle et al.9 and the numerical study per-
H L cell height;
formed by Hughes et al.7, where good agreements
could be found. Structured meshes (A and B) of higher h L hydraulic (piezometric) head;
resolution at a refinement level of 2 and 3 consisting of I 1 unit vector;
–1
673,425 and 2,690,337 nodes were applied. The finest K LT tensor of hydraulic conductivity;
–1
discretization B with 0.124 mm is smaller than the K LT constant hydraulic conductivity;
2
pixel size of the Hele-Shaw experiment with 0.154 k L intrinsic permeability;
mm. L L cell length;
Le Lewis number;
Within the extensive numerical investigations we l mesh level;
studied the influence of following features: M M mass;
M∗ 1 dimensionless mass;
• mesh effects,
N maximum number of species;
• computational acceleration by parallelization,
NE number of elements;
• viscosity effects,
• mass transfer across the centerline, NP number of nodes (points);
–1
• extensions to the Oberbeck-Boussinesq approxi- Q T sink/source term;
–1
mation, Q EOB T extended Oberbeck-Boussinesq
• perturbation of the initial solution interface. approximation term;
–1
q LT Darcy velocity;
–3 –1
For this complex and difficult problem class of convec- R ML T reaction term;
tion phenomena FEFLOW has shown a robust and effi- Ra Rayleigh number;
NNQ=ö=tÜáíÉ=m~éÉêë=J=sçäK=fs
QKS=`çåÅäìëáçåë
Rρ Turner number; Abbreviations

–1
S L storage coefficient / storativity;
t T time; 2D two-dimensional;
t∗ 1 dimensionless time; 3D three-dimensional;
x L global Cartesian coordinate vector; CPU central processing unit
x 1, x 2 L 2D Cartesian coordinates; DDC double-diffusive convection;
DDFC double-diffusive finger convection;
Greek letters FE/BE forward Euler/backward Euler
scheme;
α 1 density difference ratio; GB gigabyte;
β L, β T L longitudinal and transverse disper- EOB extended Oberbeck-Boussinesq
sivity, respectively; approximation;
ε 1 porosity, void space; RAM random access memory;
2 –1 RMS root-mean square;
ϑ L T coefficient of fluid viscosity change;
–1 –1
μ ML T dynamic viscosity of fluid;
–3
ρ ML fluid density;
ϒ global refinement; oÉÑÉêÉåÅÉë
–1
∇ L Nabla (vector) operator;
1. Cooper, C.A., Glass, R.J. and Tyler, S.W. Experimental investiga-
tion of stability boundary for double-diffusive finger convection
Subscripts in a Hele-Shaw cell. Water Resour. Res. 33 (1997), 517-526.
2. Cooper, C.A., Glass, R.J. and Tyler, S.W. Effect of buoyancy ratio
on the development of double-diffusive finger convection in a
c chloride;
Hele-Shaw cell. Water Resour. Res. 37 (2001)9, 2323-2332.
dye dye; 3. Diersch, H.-J.G. Using and testing the algebraic multigrid equa-
k species indicator; tion solver SAMG in FEFLOW. White Papers Vol. III, pp. 25-37,
0 reference / initial; 2004, WASY Berlin, Germany.
4. Diersch, H.-J.G. Reactive multi-species transport. White Papers
s sucrose;
Vol. IV, pp. 5-56, 2005, WASY Berlin, Germany.
s maximum; 5. Diersch, H.-J.G. and Kolditz, O. Variable-density flow and trans-
port in porous media: approaches and challenges. Adv. Water
Superscripts Resour. 25 (2002), 899-944. (also published in FEFLOW White
Papers Vol. II, Chapter 1, 2004, pp. 5-101, WASY Berlin, Ger-
many).
d diffusive; 6. Hughes, J.D. and Sanford, W.E. SUTRA-MS, A version of
I nodal index; SUTRA modified to simulate heat and multiple-solute transport.
USGS, Open-File Report 2004-1207, Reston, Virginia, 2004,
141pp.
cbcilt=ö=NNR
åçãÉå~
7. Hughes, J.D., Sanford, W.E. and Vacher, H.L. Numerical simula-
tion of double-diffusive finger convection. Water Resour. Res. 41
(2005), W01019, doi:10.1029/2003WR002777.
8. Nield, D.A. and Bejan, A. Convection in porous media. Springer,
New York, 1992.
9. Pringle, S.E., Glass, R.J. and Cooper, C.A. Double-diffusive fin-
ger convection in a Hele-Shaw cell: An experiment exploring the
evolution of concentration fields, length scales and mass transfer.
Transp. Porous Media 47 (2002)2, 195-214.
10. Stüben, K. User’s manual SAMG, Release 2.1, Fraunhofer Insti-
tute SCAI, 2002, St. Augustin, Germany.
11. Voss, C.I. and Provos, A.M. SUTRA - A model for saturated-
unsaturated, variable-density groundwater flow with solute or
energy transport, U.S. Geol. Surv. Water Resour. Invest. Rep., 02-
4231, 2002, 260 pp.
NNS=ö=tÜáíÉ=m~éÉêë=J=sçäK=fs
pìÄàÉÅí=fåÇÉñ
pìÄàÉÅí=fåÇÉñ ê~íÉ=QP
ÇÉÅÜäçêáå~íáçå=RM
^ ÇÉÖê~Ç~íáçå=NNI=OOI=QOI=RM
^Ç~ãëJ_~ëÜÑçêíÜ=NSI=UQ ÇÉåáíêáÑáÅ~íáçå=S
~Çëçêéíáçå=VI=NQ ÇÉåëáíó=PN
~ÉêçÄáÅ=RM ÇáÑÑÉêÉåÅÉ=ê~íáç=PO
~å~ÉêçÄáÅ=RM î~êá~ÄäÉ=UP
~å~äóíáÅ~ä=ëçäìíáçå=PS ÇÉåëáíó=ê~íáç
~èìáÑÉê Åçåëí~åí=ëçäìí~ä=PO
Pa=PV ãìäíáéäÉ=ëéÉÅáÉëJÇÉéÉåÇÉåí=PO
íÜáÅâåÉëë=NM ÇÉëçêéíáçå
ìåÅçåÑáåÉÇ=NM ê~íÉJäáãáíÉÇ=QO
^êêÜÉåáìë=NOI=OOI=OP ÇáëéÉêëáçå
~ëëáãáä~íáçå ÜóÇêçÇóå~ãáÅ=TI=US
êÉíêçëéÉÅíáîÉ=SP ÇáëíêáÄìíáçå=ÅçÉÑÑáÅáÉåí=QPI=QS
ëÉèìÉåíá~ä=SP ÇáîÉêÖÉåÅÉ=Ñçêã=QQI=QU
~ñáëóããÉíêáÅ=OT ÇçìÄäÉJÇáÑÑìëáîÉ=ÑáåÖÉê=ÅçåîÉÅíáçå=VP
aìéìáí=~ëëìãéíáçå=TQ
_
ÇóÉ=VS
Ä~ä~åÅÉ=Éèì~íáçåë=U
ÄÉåÅÜã~êâ=QV b
Ñê~ÅíìêÉ=éêçÄäÉã=UP Éèì~íáçå=çÑ=ëí~íÉ=PN
j^`^le=RM ÉèìáäáÄêáìã=UI=NQ
ÄáçÇÉÖê~Ç~íáçå=S Éêêçê=NT
ÄáçêÉãÉÇá~íáçå=QO Åçåíêçä=NT
_qbu=S bìäÉê
Ä~Åâï~êÇ=NS
`
Ñçêï~êÇ=NS
ÅÜäçêáÇÉ=RNI=VS
Éñé~åëáçå
ÅçãéêÉëëáÄáäáíó=PN
ëçäìí~ä=PP
ÅçåÇìÅí~åÅÉ=NR
íÜÉêã~ä=PN
ÅçåîÉÅíáçå
bñíÉåÇÉÇ=lÄÉêÄÉÅâJ_çìëëáåÉëè=~ééêçñáã~íáçå=US
ÇÉåëáíóJÇêáîÉå=VP
Éñíê~Åíáçå=QO
íÜÉêãçÜ~äáåÉ=VQ
î~êá~ÄäÉJÇÉåëáíó=ãìäíáëéÉÅáÉë=VQ c
ÅçåîÉÅíáçå=ÅÉääë=US cbj^qeba=ÉÇáíçê=OQ
ÑáåÖÉê=VR
a
ÖÉåÉê~íáçå=òçåÉ=NMM
a`b=RM
é~ííÉêå=NMR
ÇÉÅ~ó
ÑáåáíÉJÉäÉãÉåí=ãÉíÜçÇ=NQ
ÅçÉÑÑáÅáÉåí=QS
ÑäççÇï~îÉ=éêçÄäÉã=TU
cbcilt=ö=NNT
co^`Pasp=UP ãáåáãáò~íáçå=SQ
Ñê~ÅíìêÉ=UP ãçäÉÅìä~ê=ÇáÑÑìëáçå=T
áåÅäáåÉÇ=UP jçåçÇ=NOI=OOI=OQ
cêÉìåÇäáÅÜ k
~Çëçêéíáçå=ÑìåÅíáçå=V åáíê~íÉ=RO
d åçåÉèìáäáÄêáìã=V
dt=UP åçêã
e ã~ñáãìã=NT
e~ÖÉåJmçáëÉìáääÉ=UR ojp=NT
Ü~äÑJäáÑÉ=NQ l
Ü~äÑJë~íìê~íáçå=NO lÄÉêÄÉÅâJ_çìëëáåÉëè=~ééêçñáã~íáçå=USI=NMR
eÉäÉJpÜ~ï=ÉñéÉêáãÉåí=VQ ÉñíÉåÇÉÇ=VR
eÉåêó çÄàÉÅíáîÉ=ÑìåÅíáçå=SQ
~Çëçêéíáçå=ÅçÉÑÑáÅáÉåí=VI=QS çÄëÉêî~íáçå=éçáåíë=SQ
Üçêáòçåí~ä=Ñçêãìä~íáçåë=NM léÉåjm=NMO
f çñóÖÉå=RMI=RN
áåíÉêÑ~ÅÉ=ã~å~ÖÉê=SQ m
áåíÉêéêÉíÉê=OOI=OR é~ê~ãÉíÉê=Éëíáã~íáçå=SP
h m`b=RM
h~äã~å=ÑáäíÉê=SQ éÉêÑçêã~åÅÉ=NMP
âáåÉíáÅ=ãçÇÉä éÉêíìêÄ~íáçå=NMM
çåÉJëáíÉ=QO áåáíá~ä=NMP
íïçJëáíÉ=QQ mbpq=SP
âáåÉíáÅë=RI=NN éÉëíáÅáÇÉë=QQ
éÜ~ëÉ
i
ÑäìáÇ=T
i~åÖãìáê
áããçÄáäÉ=OM
~Çëçêéíáçå=ÑìåÅíáçå=V
áåÇáÅ~íçê=U
ä~ï=çÑ=ã~ëë=~Åíáçå=U
ãçÄáäÉ=OM
äÉ~ÅÜáåÖ=S
ëçäáÇ=T
iÉïáë=åìãÄÉê=VT
éêÉÇáÅíçêJÅçêêÉÅíçê=NS
j éêçÇìÅíë=S
ã~ëë éìãé=íÉëí=SR
ÅçåëÉêî~íáçå=T _êÉóÉää=ST
ã~ëë=íê~åëÑÉê=NMQ táÅÜáí~=TQ
ã~íêáñ=ëóëíÉã=NR o
ãÉíÜ~åçÖÉåÉëáë=S
ê~ÇáçåìÅäáÇÉ=ÇÉÅ~ó=S
jáÅÜ~ÉäáëJjÉåíÉå=NO
ê~åÇçã=åçáëÉ=NMP
ãáÅêççêÖ~åáëãë=NO
NNU=ö=tÜáíÉ=m~éÉêë=J=sçäK=fs
o~óäÉáÖÜ=åìãÄÉê=VT q
êÉ~Åí~åíë=S q`b=RM
êÉ~Åíáçå íÉãéÉê~íìêÉ=OPI=OR
^êêÜÉåáìë=OP qÜÉáëD=éêçÄäÉã=SR
ÅÜ~áåë=PR íÜÉêãçÜ~äáåÉ=ÅçåîÉÅíáçå=VQ
ÇÉÖê~Ç~íáçå=OO íÜÉêãçÜ~äáåÉ=íê~åëéçêí=OMI=OR
Ü~äÑJäáÑÉ=NQ qìêåÉê=åìãÄÉê=VT
áêêÉîÉêëáÄäÉ=NN
s
âáåÉíáÅë=R
î~êá~ÄäÉë
âáåÉíáÅë=ÉÇáíçê=ONI=OUI=OVI=PM
Ää~Åâ=OPI=OR
jçåçÇ=OQ
ÄäìÉ=OPI=OR
é~ê~ääÉä=PR
ÖêÉÉå=OPI=OR
ê~íÉë=NN
êÉÇ=OPI=OR
êÉîÉêëáÄäÉ=U
s`=RM
ëÉêá~ä=PR
îáëÅçëáíó=PQI=UR
ëíçáÅÜáçãÉíêó=S
ëí~åÇ~êÇ=PQ
ëóãÄçäáÅ=R
ìëÉêJÇÉÑáåÉÇ=PQ
ìëÉê=ÇÉÑáåÉÇ=OOI=OS
îáëÅçëáíó=ÉÇáíçê=PQ
êÉ~Åíáçå=âáåÉíáÅë=ÉÇáíçê=ON
îçáÇ=ëé~ÅÉ=U
êÉ~Åíáçå=ê~íÉ=NN
îçäìãÉ=Ñê~Åíáçå=U
éêÉÅçãéáäÉÇ=OO
ìëÉêJÇÉÑáåÉÇ=OQ
êÉÑáåÉãÉåí=äÉîÉä=NMM
êÉí~êÇ~íáçå=VI=NQI=NR
Ñ~Åíçê=QS
oÉóåçäÇë=åìãÄÉê=VQ
oqPa=RU
p
p^jd=NMNI=NMQ
ë~íìê~íáçå=U
ëéÉÅáÉë
fa=OM
äáëí=ÉÇáíçê=OM
ëÉäÉÅíçê=OM
ëéäáíJçéÉê~íçê=NR
pq^kjla=QQ
ëíçáÅÜáçãÉíêó=S
ëìÅêçëÉ=VS
prqo^=VQ
cbcilt=ö=NNV
NOM=ö=tÜáíÉ=m~éÉêë=J=sçäK=fs
^ìíÜçê=fåÇÉñ
^ìíÜçê=fåÇÉñ pìå=PRI=PSI=PVI=RUI=TU
q
_ qÜÉáë=SRI=ST
_çìííáÉê=SP qÜÉêêáÉå=UP
` s
`~êëä~ï=SR sçáÖí=SRI=TM
`ççéÉê=TMI=VQ sçëë=VQ
`çêå~íçå=UP t
`çìêíáÉê=SP táÉÇÉãÉáÉê=RM
a
aáÉêëÅÜ=SRI=UPI=VP
açÜÉêíó=SPI=SQ
c
cêó=QOI=QQI=QRI=QS
d
dê~Ñ=UP
dêΩåÇäÉê=SP
e
eìÖÜÉë=VQ
g
g~ÅçÄ=TMI=TR
g~ÉÖÉê=SR
h
h~áëÉê=SPI=TU
h~ååÉó=NR
hçäÇáíò=UPI=VP
i
i~åÖÖìíÜ=SRI=TM
m
mêáåÖäÉ=VQ
mêçîçëí=VQ
p
p~åÑçêÇ=VQ
pÜáâ~òÉ=UP
páããçåë=UP
píΩÄÉå=NMN
cbcilt=ö=NON
^ìíÜçê=fåÇÉñ
NOO=ö=tÜáíÉ=m~éÉêë=J=sçäK=fs

Cbcilt: Sçäk Fs

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Cbcilt: Sçäk Fs

Uploaded by

Copyright:

Available Formats

aefJt^pv=pçÑíï~êÉ

NKN= fåíêçÇìÅíáçå=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=KR äÉãë =K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K= PQ

OKN= fåíêçÇìÅíáçå=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=KSP OKOKP= táÅÜáí~=éìãé=íÉëí K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K= TQ

PKN= fåíêçÇìÅíáçå=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=KUP PKOKP= páãìä~íáçå=êÉëìäíëK=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K= US

QKN= fåíêçÇìÅíáçå=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=KVP QKQKO= _çìåÇ~êó=~åÇ=áåáíá~ä=ÅçåÇáíáçåëI=éÉêíìêÄ~íáçå=çÑ=áåáíá~ä=

pìÄàÉÅí=fåÇÉñ K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K=K= NNT

H.-J. G. Dierscha, A. Pönitza, D. Etcheverryb & Y. Rossierb

NKN fåíêçÇìÅíáçå high degree of freedom in formulating the chemical

for N° < N ( N° = number of reactants) C 6 H 6 + 7.5O 2 → 6CO 2 + 3H 2 O (1-3)

In (1-1) A and B represent chemical species (reactants C 6 H 6 + 7.5O 2 → 6CO 2 + 3H 2 O ⎫

(i) Pyrite oxidation 1 2- - 4-

NKOKO dçîÉêåáåÖ=íê~åëéçêí=Éèì~íáçåë fluid f and solid s phases of a porous medium can be

where (1-11) can also be subjected to a reversible equilibrium

(LMA)8 can be used to explicitly express the (sorbed) ing to8

species k and m can be written as where κ, k 1, k 2, b 1, b 2 correspond to sorption coeffi-

where the retardation factor ℜ k is introduced accord-

For depth-integrated 2D horizontal formulations of and unconfined aquifers result:

where B denotes the aquifer thickness (cf. FEFLOW’s Reference Manual5).

NKOKP oÉ~Åíáçå=ê~íÉë=oâ NKOKPKN aÉÖê~Ç~íáçå=íóéÉ=âáåÉíáÅë

Reactive systems can be broadly classified into simple α n1 α n2 α nN

A typical constitutive representation of r r has a

which can be developed by a polynomial expression of

NKOKPKO ^êêÜÉåáìë=íóéÉ=âáåÉíáÅë enzyme kinetics equation8. Its mathematical represen-

Af Bf Cf Df with the reaction rates of a degradation type

Accordingly, the following balance equations result for

The matrix system (1-36a) can be highly nonlinear

NKPKO ^Ç~éíáîÉ= éêÉÇáÅíçêJÅçêêÉÅíçê n

2. If It is important to note that the error tolerance δ is

4. If the criteria (1-47a) and (1-47b) are satisfied by all

Set MASS (or THERMOHALINE for non-isothermal) transport

Set multi-species option

Input number of species

Open species list editor for phase definition and naming

Species IDs automatically countered

Phase definition: fluid or solid

NKQ oÉä~íÉÇ=cbcilt=aá~äçÖë NKQKO péÉÅáÉë=ëÉäÉÅíçê

Figure 1.4 Copy of input data for different species.

irrelevant and input is disabled for those parameters.

NKQKP jìäíáJëéÉÅáÉë=ã~íÉêá~ä=Ç~í~ A new button identified by the Σ symbol launches

reaction kinetics editor reaction kinetics editor

rate parameter to be used

NKR oÉ~Åíáçå=háåÉíáÅë=bÇáíçê two principal classes of reaction expressions that can

simplified expressions of the degradation (1-26), NKRKN mêÉÅçãéáäÉÇ=ê~íÉ=ÉñéêÉëëáçåë

• freely editable algebraic formulae of arbitrary

Rate expressions stemming from user-defined kinetics

where k i are rate constants, Rate m are rate coefficients,

Figure 1.8 FEFLOW’s Reaction Kinetics Editor for pre-

List of blue variables

List of green variables

NKS j~ëë= j~íÉêá~ä= a~í~= çÑ

Figure 1.10 FEMATHED editor with testing the built-in

Item Symbol Unit Default Equations

Porosity ε 1 0.3 1-9, 1-18

Sorption coefficient κ 1 0 1-17

Sorption exponent b2 1 0 1-17

Item Symbol Unit Default Equations

Sorption coefficient (numerator) k1 1 0 1-17

Sorption coefficient (denominator) k2 l ⁄ mg 0 1-17

Longitudinal dispersivity βL m 5 1-19

Transverse dispersivity βT m 0.5 1-19

Out-transfer rate out –1 0 see FEFLOW Refer-

Table 1.2 Material conditions for 2D horizontal mass transport of species k at