Unde s anding Unce ain y in Ma ke -Media ed Responses o US
Oilseed Biodiesel Demand: Sensi i i y o ILUC Emission Es ima es o
GLOBIOM Pa ame ic Unce ain y
Neus Escoba ,*Hugo Valin, S e an F ank,*Diana Galpe in, Ch is ophe M. Wade, Leopold Ringwald,
Daniel Tanne , Niklas Hinkel, Pe Ha lík, Jus in S. Bake , Sha yn Lie, and Ch is ophe Ramig
Ci e This: h ps://doi.o g/10.1021/acs.es .3c09944
Read Online
ACCESS Me ics & Mo e A icle Recommenda ions *
sı Suppo ing In o ma ion
ABSTRACT: The li e cycle g eenhouse gas (GHG) emissions o
bio uels depend on unce ain es ima es o induced land use change
(ILUC) and subsequen emissions om ca bon s ock changes.
Demand o oilseed-based bio uels is associa ed wi h pa icula ly
complex ma ke and supply chain dynamics, which mus be
conside ed. Using he global pa ial equilib ium model GLOBIOM,
his s udy explo es he unce ain y in ma ke -media ed impac s and
ILUC- ela ed emissions om inc easing demand o soybean
biodiesel in he Uni ed S a es in he pe iod 2020−2050. A one-a -
a- ime (OAT) analysis and a Mon e Ca lo (MC) analysis a e
pe o med o assess he sensi i i y o modeled ILUC-GHG emissions in ensi ies (gCO2e/MJ) o a ying key economic and
biophysical model pa ame e s. Addi ionally, he in luence o he app oach on he simula ion o u u e ILUC e ec s is explo ed using
wo al e na i e ILUC-GHG me ics: a compa a i e-s a ic app oach o 2030 and a ecu si e-dynamic app oach using model ou pu s
h ough 2050. We ind ha p ojec ed ILUC-GHG alues la gely a y based on which ege able oils eplace di e ed soybean oil,
ma ke esponses o cop oduc s, and he ca bon con en o land con e ed o ag icul u al use. These a e all, in u n, subjec o
decision unce ain y h ough he choice o he modeling app oach and he ime ho izon conside ed o each ILUC-GHG me ic.
Gi en he longe simula ion pe iod, ILUC-GHG emission unce ain y anges inc ease unde he ecu si e-dynamic app oach (42.4
±25.9 gCO2e/MJ) compa ed o he compa a i e-s a ic app oach (40.8 ±20.5 gCO2e/MJ). The combina ion o MC analysis wi h
o he echniques such as Bayesian Addi i e Reg ession T ees (BART) is powe ul o unde s anding model beha io and cla i ying
he sensi i i y o ma ke esponses, ILUC, and associa ed GHG emissions o speci ic model pa ame e s when simula ed wi h global
economic models. The BART e eals ha biophysical pa ame e s gene a e mo e linea ILUC-GHG esponses o changes in assumed
pa ame e alues while changes in economic pa ame e s lead o mo e nonlinea ILUC-GHG esul s as mul iple e ec s a he
in e play o ood, eed, and uel uses o e lap. The choice o he ecu si e-dynamic me ic allows cap u ing he longe - e m e olu ion
o ILUC while gene a ing addi ional unce ain ies de i ed om he baseline de ini ion.
KEYWORDS: bio uels, clima e change, economic model, g eenhouse gas, in e na ional ade, li e cycle assessmen , spillo e , anspo
1. INTRODUCTION
The po en ial con ibu ion o bio uels o clima e change
mi iga ion has ecei ed signi ican a en ion in he li e a-
u e.
1−3
The es ima ion o land use change (LUC) emissions
associa ed wi h bio uels has been deba ed in ensi ely,
mo i a ed pa ially by he ac ha po en ial u u e LUC
impac s canno be measu ed bu only modeled.
4−8
LUC can
e e o ei he land con e sion o g ow bio uel eeds ocks
(o en e e ed o as di ec LUC) o he subsequen land
ans o ma ion o suppo o he uses globally, such as ood and
eed ( e e ed o as indi ec LUC). Induced land use change
(ILUC) he eina e designa es he combina ion o he wo
componen s, cap u ing he o e all ma ke -media ed displace-
men o land uses in esponse o an inc eased demand o
biomass o uel.
9,10
ILUC esul s in ne g eenhouse gas
(GHG) emissions when global ne land ca bon s ocks a e los
h ough successi e land con e sions.
11,12
The wo me hods commonly used o quan i y he GHG
emissions in ensi y o bio uels a e a ibu ional li e cycle
assessmen (ALCA) and consequen ial LCA (CLCA). ALCA
conside s impac s om eeds ock p oduc ion up o combus-
ion, o en including di ec LUC.
13−16
Quan i ying indi ec
LUC emissions equi es economic easoning unde con-
Recei ed: May 29, 2024
Re ised: No embe 27, 2024
Accep ed: Decembe 2, 2024
Published: Decembe 18, 2024
A iclepubs.acs.o g/es
© 2024 The Au ho s. Published by
Ame ican Chemical Socie y A
h ps://doi.o g/10.1021/acs.es .3c09944
En i on. Sci. Technol. XXXX, XXX, XXX−XXX
This a icle is licensed unde CC-BY-NC-ND 4.0
Downloaded ia 62.42.100.239 on Decembe 20, 2024 a 20:00:16 (UTC).
See h ps://pubs.acs.o g/sha ingguidelines o op ions on how o legi ima ely sha e published a icles.
sequen ial app oaches.
17,18
Fo bio uels, CLCA aims o
ep esen how inc eased demand o c ops causes ag icul u al
land expansion, c op displacemen , in ensi ica ion, and changes
o c op and li es ock p oduc ion emissions, depending on
ela i e p oduc i i y and p ice adjus men s ac oss sec o s and
egions.
19,20
CLCA o en applies economic modeling o
de e mine he ex en and loca ion o land con e sion and he
ecosys ems and land uses a ec ed.
21−23
In a e iew o he
science o bio uel LCA, he US Na ional Academies o Science,
Enginee ing, and Medicine ound ha CLCA is p e e able
when seeking o unde s and he consequences o decisions o
ac ions ha al e o e all quan i ies o bio uel consumed.
24
Recen ALCA s udies show ha ege a ion loss emains a la ge
con ibu o o he GHG emission in ensi y o bio uel
eeds ocks, showing ha di ec LUC emissions can be highe
han ILUC emission alues o bio uels p oduced in speci ic
loca ions when using spa ially explici ca bon s ock and land
use da a.
25−27
Global economic models p o ide consis en amewo ks o
assess ILUC impac s o bio uels unde CLCA app oaches.
28−30
These models cap u e key economic mechanisms ha shape
he global dis ibu ion o land uses, such as yield and c opland
a ea esponses o changes in land a ailabili y and ade.
31
Fo
ins ance, he Compu able Gene al Equilib ium (CGE) model
GTAP-BIO
32,33
includes se e al ood and non ood bio uel
sec o s −and hei cop oduc s− o simula e economy-wide
e ec s and land co e change due o expanded bio uel
p oduc ion. GLOBIOM
34,35
is a Pa ial Equilib ium (PE)
model ep esen ing ag icul u e and o es y sec o s, wi h a
spa ially explici land use ep esen a ion. Bo h models ha e
been widely used o he analysis o bio uel policies.
32,36−40
ILUC impac s o bio uels ha e been analyzed ex ensi ely,
ini ially o he so-called i s -gene a ion bio uels
41−43
and
mo e ecen ly o non ood eeds ocks.
44−47
Pa icula ly, he
ILUC emission es ima ion o eeds ocks o a ia ion bio uel
p oduc ion as pa o he Ca bon O se ing and Reduc ion
Scheme o In e na ional A ia ion (CORSIA) uses GTAP-BIO
and GLOBIOM o p opose a ha monized ILUC emission
in ensi y alue pe eeds ock, including non ood c ops and
esidues.
48
In he con ex o he Uni ed S a es (US), s udies
ha e mainly ocused on co n e hanol
42,43,49
and soybean
biodiesel,
5,50,51
wi h he impac s o oilseed p oduc ion and
ade ecen ly gaining mo e a en ion.
52−55
As in e na ional
demand o oilseeds and ege able oil o a ious uses
con inues o inc ease wo ldwide, he e a e conce ns ha
hese inc easing demands impac o es and na u al ege a ion
loss, especially in he opics.
56−59
Global combined biodiesel
and enewable diesel p oduc ion expanded om a ound 20.2
billion li e s in 2010 o a ound 52.7 billion li e s in 2021.
60
Following a 5- old inc ease in US biodiesel supply be ween
2010 o 2021, he US is now he second la ges biodiesel
p oduce behind he EU, accoun ing o 20% o global
p oduc ion.
60−63
Mos US biodis illa e uel p oduc ion o e
his pe iod has been sou ced om ege able oils, wi h soybean
oil ep esen ing he plu ali y o biodis illa e eeds ock.
Es ima ing ILUC emission in ensi ies o ege able oil-based
uels has di e se sou ces o unce ain y including he choice o
he modeling amewo k and accompanying assump ions o
quan i y cu en and u u e e ec s, e.g., baseline yea , analy ic
ho izon (decision unce ain y) and he lack o ull unde -
s anding o he unde lying complex land use dynamics
(epis emic unce ain y).
42
The la e is ela ed o he inpu
pa ame e s and speci ic alues needed o simula e he
p ocesses being modeled and is also e e ed as pa ame ic
unce ain y.
64,65
CLCA modeling equi es nume ous da a and
inpu pa ame e s, which ep esen and in luence bo h
economic and biophysical dynamics, o cap u e esponses
ac oss in e linked ood, eed, and uels ma ke s. Key dynamics
o e alua ing ILUC e ec s include c opland in ensi ica ion/
ex ensi ica ion as well as impac s on land use and managemen
and li es ock p oduc ion.
66
P e ious wo k ound ha , among
he pa ame e s de e mining long- e m ILUC e ec s, he mos
decisi e ela e o cos s o land ans o ma ion, endogenous
p oduc i i y esponses, and c opland a ailabili y.
66
Pas unce ain y analyses o bio uels ocused on he e ec s o
al e na i e yield elas ici ies o c op p ices, demand elas ici ies,
ade elas ici ies, and land ans o ma ion (expansion)
elas ici ies o land en s.
37,40,67
S udies pe o ming sensi i i y
analyses wi h he CGE model GTAP-BIO
32
highligh he ole
o he elas ici y go e ning land con e sion be ween c opland,
pas u eland, and managed o es land.
37,68
When assessing
pa ame ic unce ain y in GTAP-BIO wi h MC analysis,
64
he yield elas ici y o p ice, he A ming on ade elas ici ies, he
GHG emissions ac o (EF) o c opland- o-pas u e con-
e sion, and he p oduc i i y o newly con e ed c opland we e
ound o con ibu e mos o he a iance in ILUC emission
alues o US soybean biodiesel.
The size and complexi y o global economic models pose
challenges o designing, implemen ing, unning, and in e p e -
ing la ge-scale sensi i i y analyses in ended o p opaga e
unce ain y om pa ame e s o model ou comes. As such,
only a ew s udies ha e de o ed e o s o execu e hese
me hods.
37,44,49,64,65
Using he GLOBIOM global economic
modeling amewo k,
69
his s udy aims o assess he sensi i i y
o wo di e en measu es o ILUC-GHG emissions in ensi ies
(he eina e e e ed o as “ILUC-GHG alues”) o US
soybean biodiesel o he unce ain y in key model pa ame e s
in luencing ILUC esponses. This is done by combining a one-
a -a- ime (OAT) analysis, whe e each o he selec ed
pa ame e s is a ied indi idually o e a p ede e mined ange;
and a Mon e Ca lo (MC) analysis, whe e all pa ame e s a e
gi en speci ic p obabili y dis ibu ions and a ied andomly
and simul aneously. Addi ionally, we assess he in luence o he
choice o he ime ho izon and ILUC emission accoun ing
p ocedu e by using wo di e en app oaches o de i e u u e
ILUC-GHG alues: a compa a i e-s a ic app oach, in which
modeled changes in 2030 a e amo ized o e 25 yea s o
biodiesel p oduc ion, and a ecu si e-dynamic app oach using
model ou pu s h ough 2050. Thus, he goal o he s udy is o
be e unde s and GLOBIOM model beha io and he ole o
ce ain pa ame e s and assump ions when es ima ing ILUC-
GHG alues. O e all, his s udy con ibu es o he unde -
s anding o he ac o s d i ing ma ke -media ed esponses o
inc eased demand o biobased p oduc s, while iden i ying
a eas o u u e esea ch o a be e ep esen a ion o oilseed
bio uels in global economic models.
2. METHODS
2.1. Modeling F amewo k. The global ecu si e-dynamic
PE model GLOBIOM
40,69
is applied o assess he e ec s o an
inc eased demand o soybean biodiesel in he US in he
pe iod 2020−2050, he eina e called shock. GLOBIOM
compu es a global equilib ium in ag icul u al and o es
p oduc ma ke s in 10-yea ime s eps h ough he pe iod
2000−2050 by choosing land use and p ocessing ac i i ies ha
maximize wel a e subjec o esou ce, echnological, demand,
En i onmen al Science & Technology pubs.acs.o g/es A icle
h ps://doi.o g/10.1021/acs.es .3c09944
En i on. Sci. Technol. XXXX, XXX, XXX−XXX
B
and policy cons ain s. The model does no ep esen ma ke -
media ed e ec s o bio uel policies on sec o s o he economy
o he han land use ( o es y and ag icul u e, including c op
and li es ock). Fo his analysis, he model is agg ega ed in o
57 wo ld egions.
34,36
The baseline ep esen s economic
de elopmen s acco ding o he Sha ed Socioeconomic Pa h-
way 2 scena io,
70
while his o ical yields we e ecalib a ed so
ha hey include 2017−2019 da a.
71
Vege able oil p ices we e
also ecalib a ed o align wi h ela i e p ice p ojec ions bo h
o he US e sus he es o he wo ld (ROW), espec i ely
based on da a om he Uni ed S a es Depa men o
Ag icul u e (USDA) and he Food and Ag icul u e O gan-
iza ion (FAO) Fu he de ails a e included in Sec ion S1.1 o
he elec onic Suppo ing In o ma ion (ESM).
2.2. Scena ios and ILUC Me ics. This s udy models a
soybean biodiesel consump ion shock in he US in oduced in
2021 o p og essi ely each a o al addi ional demand o 126.9
PJ/yea in 2030, which co esponds o 1 billion gallons o
gasoline equi alen , o oughly a doubling o cu en US
consump ion. This is compa ed o a baseline ha keeps global
bio uel olumes cons an h ough 2050 a 2020 le els.
To unde s and he unce ain y b ough abou by he ILUC
emission accoun ing p ocedu e, wo al e na i e me hods a e
applied o calcula e ILUC-GHG alues, he eina e e e ed o
as compa a i e-s a ic and ecu si e-dynamic (see equa ions in he
Appendix). The o me p ojec s he ILUC-GHG alues in
2030 − he yea in which he ull bio uel manda e is eached −
o e a 10-yea pe iod. This app oach ies o isola e he e ec s
o he bio uel shock in he sho e m, assuming he impac s
will be equally dis ibu ed h oughou he amo iza ion pe iod
o 25 yea s. Al e na i ely, he ecu si e-dynamic app oach
conside s a cons an demand o 126.9 PJ/yea abo e baseline
le els un il 2050, he eby cap u ing ma ke and land use
de elopmen s in he longe e m. This co esponds o bio uel
olumes equi alen o 25 yea s o o al manda ed biodiesel
consump ion in 2020−2050 (Figu e S1 in ESM), in line wi h
he 25-yea amo iza ion pe iod unde he compa a i e-s a ic
app oach.
Unde bo h app oaches, ILUC-GHG alues a e calcula ed
conside ing CO2e emissions/ emo als om changes ac oss
ca bon pools acco ding o IPCC guidelines (IPCC 2006) (eq
1). This e e s o ne changes in abo e- and below-g ound
biomass, dead wood, li e , and ha es ed wood p oduc s, while
addi ional emissions a ise om pea land oxida ion (eq 2) and
changes in soil o ganic ca bon (SOC) ela i e o e e ence
soils. In GLOBIOM, ne changes in ca bon s ocks a e he
esul o he land ansi ions simula ed, ul ima ely leading o
na u al ege a ion loss, na u al ege a ion e e sion, and
c opland expansion. The compa a i e-s a ic ILUC-GHG
alue conside s changes in land use- ela ed emissions o
2030 when he bio uel consump ion manda e is eached (eq
3), whe eas he ecu si e-dynamic ILUC-GHG alue conside s
cumula i e emissions up o 2050 (eq 4). The compa a i e-
s a ic ILUC-GHG alue is o en used in scien i ic li e a u e
looking a sho - e m esponses o a bio uel shock
32,39,72
and
hence acili a es compa ison o ou esul s, while he ecu si e-
dynamic ILUC-GHG alue gi es addi ional insigh s on how
he bio uel emissions impac s e ol e o e ime.
49,73
2.3. Sensi i i y Analysis. Sensi i i y analysis is applied o
iden i y how a ia ion in inpu pa ame e alues causes
a ia ion in ou pu a iables. In his case, sensi i i y in
ILUC-GHG alues is assessed by changing inpu pa ame e s
inc emen ally and indi idually, as well as s ochas ically and
simul aneously, in OAT and MC analyses, espec i ely. The
wo analyses a e combined o unde s and he model’s
pe o mance o ILUC-GHG alue es ima ion, shedding ligh
on he ole o he a ied pa ame e s in de e mining
unce ain y. Ele en model pa ame e s and hei associa ed
p obabili y dis ibu ions we e chosen based on expe knowl-
edge and expe ience om p e ious GLOBIOM s udies.
40,74,75
They include se en economic and ou biophysical pa ame e s
(Table S1 in ESM).
To ini ially unde s and he in luence o hese pa ame e s on
es ima ed ILUC-GHG alues, he OAT analysis a ies each
pa ame e alone while keeping he es a he cen al alue, i.e.,
a he model de aul . Fo each key pa ame e , eigh al e na i e
alues we e conside ed, ou below and ou abo e he cen al
alue, co e ing he pa ame e anges de ailed in ESM Table S1.
Fo each o he 99 di e en pa ame e combina ions (eigh
al e na i es o all 11 pa ame e s, and a cen al case wi h all
de aul alues), baseline and shock scena ios we e un o
calcula e bo h ILUC-GHG me ics conside ed.
MC simula ion is a s ochas ic echnique ha p oduces a
ange o esul s om which a p obabili y dis ibu ion o
modeled esul s can be in e ed. MC analysis in ol es sol ing
he model ac oss a numbe o uns ha is su icien ly high
ela i e o he sample size used o inpu pa ame e s.
76
In each
un, alues o speci ic model inpu pa ame e s a e andomly
selec ed om de ined p obabili y dis ibu ions o each
s ochas ic inpu pa ame e . In his s udy, 1,000 uns a e
pe o med o bo h he shock and baseline scena ios o
es ima e he di e ence in model ou comes o each scena io
pai wi h he same combina ion o inpu pa ame e s. In
con as wi h he OAT analysis, se e al pa ame e alues a e
d awn independen ly ac oss egions (subs i u ion elas ici y−
ege able oils), commodi ies ( ade elas ici y− ege able oils), o
land ypes (land expansion in o na u al ege a ion) o each MC
un. This yields 72 indi idual p obabili y dis ibu ions in o al,
conside ing he 11 pa ame e s a ied and he speci ic egions
and p oduc s o which hey apply.
Las ly, o iden i y he ole o each pa ame e in MC analysis,
Bayesian Addi i e Reg ession T ees (BART) analysis
77
was
applied ex-pos o app oxima e he unc ional o m o
GLOBIOM ou comes wi h espec o he pa ame e s unde
in es iga ion. Mo e speci ically, he BART model is ained on
he sampled economic and biophysical shock alues as
co a ia es, in o de o explain he a ia ion in ILUC-GHG
alues ob ained ac oss all MC uns. Mo e de ails on he
app oach a e de ailed in he ESM (Sec ion S1.6).
3. RESULTS
3.1. Cen al Case. This sec ion p esen s he cen al case
(wi h de aul model pa ame e alues) o con ex ualize he
subsequen sensi i i y analysis and summa ize impo an
model dynamics in esponse o he shock. GLOBIOM
es ima es ha he US biodiesel consump ion shock leads o
h ee key ma ke -media ed e ec s ha de e mine he ILUC
e ec s and ela ed GHG in ensi y alues in he cen al case: 1)
global ege able oil demand o non uel uses dec eases, 2)
o he ege able oils (mainly palm oil) subs i u e o soybean oil
o non uel uses globally, and 3) he egional dis ibu ion o
soybean p oduc ion shi s a he ma gin. The i s and second
e ec s a e in eg al o unde s anding he ILUC implica ions o a
US soybean biodiesel shock in he cen al case.
The US sou ces mos o he addi ional soybean oil needed
o supply he bio uel shock (a o al o 3.5 M ) in 2030 h ough
En i onmen al Science & Technology pubs.acs.o g/es A icle
h ps://doi.o g/10.1021/acs.es .3c09944
En i on. Sci. Technol. XXXX, XXX, XXX−XXX
C
inc eased domes ic p oduc ion (1.9 M ), educ ion o ne ade
(−1.2 M ), and domes ic educ ion o non uel uses o soybean
oil (−0.3 M ) (Figu e 1). This signi ican ly a ec s global
ege able oil ma ke s and land use dynamics in non-US egions
(Figu e S2).
The inc eased US soybean oil demand and he dec eased
expo s aise global p ices o soybean oil and, subsequen ly,
o he ege able oils (by 6% and 3% in 2030, espec i ely).
While global supply o ege able oils inc eases in 2030 (2.1
M ), demand o ood and o he non uel applica ions dec eases
(−1.3 M ); bo h because ege able oil p ices ise, especially o
ood uses, and because ege able oils a e no pe ec
subs i u es. Palm and apeseed oils pa ly compensa e o
dec eases in soybean oil consump ion (−3.1 M ) o non uel
applica ions globally, wi h addi ional 1.8 M consumed in
2030. This s ongly a ec s GHG emissions om ILUC in
GLOBIOM, as addi ional palm oil o igina es om Sou heas
Asia (SEA), whe e palm expansion in he pas en ailed
ain o es con e sion and pea land oxida ion.
As he US inc eases domes ic soybean p oduc ion in
esponse o he shock (+5.7 M , + 1.4 Mha ha es ed a ea
in 2030), domes ic c ushing inc eases a ailabili y o soybean
meal (+7.7 M ), demanded o animal eeding in in e na ional
ma ke s. While p ices o soybean oil inc ease 27.1% in he US
and 5.5% globally in 2030, p ices o soybean meal dec ease
compa ed o he baseline by 7.8% domes ically and 3.3%
globally. The inc eased a ailabili y o ela i ely cheap US
soybean meal o li es ock eeding pu s p essu e on less
p oduc i e p oduce s in Asia bu also educes demand o
soybeans and soybean meal om o he leading expo e s,
mainly A gen ina and B azil, who, oge he , accoun o a ound
60% o he global ma ke and expo s o soybeans. As wo ld
a e age soybean p ices dec ease, soybean a ea expansion in
Sou h Ame ica (SAM) in 2020−2030 is mo e limi ed and
declines by 1.4 pe cen age poin s, o 33.8%, compa ed o
35.2% expansion in he baseline. This ansla es in o lowe
c opland expansion in SAM (+0.3 Mha, + 0.28% in 2030) wi h
he shock compa ed o he baseline.
In 2050, c opland expands in SAM by 0.1 Mha (+30.8% in
2020−2050 s +30.7% in he baseline) due o inc eased
demand o o he ag i- ood c ops globally, mainly h ough he
li es ock ebound e ec , which he ein desc ibes he inc eased
p o i abili y o li es ock sec o s globally gi en he inc eased
a ailabili y o ela i ely cheap meals in he ma ke , and he
associa ed inc ease in animal p oduc s consump ion. As a
esul , animal p oduc ion sys ems in ensi y (especially pig,
poul y, and dai y), demanding mo e eed concen a es and
o he g ains o complemen eed a ions. Global ce eal demand
o animal eed inc eases by 1.3 M as well as ela ed ha es ed
a eas (+0.2 Mha in 2030) (Figu e 1). G assland a eas also
expand in egions wi h ela i ely ex ensi e li es ock p oduc ion
sys ems, e.g., US, SAM, and Sou he n Asia (SAS).
ILUC e ec s inc ease global GHG emissions ela i e o he
baseline. The compa a i e-s a ic ILUC-GHG alue is es i-
ma ed a 29.4 gCO2e/MJ (Figu e 1) wi h he mos impo an
emissions sou ce being pea land oxida ion (36.0 gCO2e/MJ),
ollowed by na u al land con e sion (16.3 gCO2e/MJ), and
o gone na u al land e e sion (2.4 gCO2e/MJ). These
emissions a e pa ly compensa ed by enhanced ca bon
seques a ion om ag icul u al biomass (−22.6 gCO2e/MJ)
and SOC (−2.7 gCO2e/MJ). Palm expansion in SEA
con ibu es signi ican ly o ILUC- ela ed emissions, especially
h ough na u al land con e sion and pea land oxida ion.
Simul aneously, palm plan a ions seques e mo e ca bon pe
hec a e han annual c ops he eby gene a ing some ca bon sink
in ag icul u al biomass. The size o he o al biomass e ec
om palm depends on he ype o land ha palm plan a ions
a e expanding on o. I palm plan a ions eplace ain o es s,
he e is ne loss in o al biomass ca bon. Since he model
es ima es g ea e expansion on o p e iously undis u bed
pea lands, pea land emissions associa ed wi h palm expansion
ou weigh he biomass seques a ion e ec . These esul s in he
cen al case scena io a e highe han he GTAP-BIO alue o
a ound 20 gCO2e/MJ o US soybean biodiesel.
32,55,78
No able
di e ences in assump ions be ween hese s udies include
bio uel olumes, shock yea , and bio uel p ocessing e iciencies.
Howe e , as discussed u he below, his 20 gCO2e/MJ esul
alls wi hin he ange o GLOBIOM esul s desc ibed in ou
sensi i i y analysis.
Figu e 1. Absolu e changes in global ha es ed a ea (le ) and global land co e (cen e ) when compa ing he shock scena io o he baseline
scena io in 2030 and 2050; and co esponding ILUC-GHG alues unde compa a i e-s a ic and ecu si e-dynamic app oaches ( igh ), espec i ely,
b oken down by GHG sou ce. Each “x” ep esen s he ne alue ac oss ca ego ies depic ed. Di e en mul ic opping in ensi ies ac oss c ops and
wo ld egions explain he di e ence in o al ha es ed a ea s o al c opland changes.
En i onmen al Science & Technology pubs.acs.o g/es A icle
h ps://doi.o g/10.1021/acs.es .3c09944
En i on. Sci. Technol. XXXX, XXX, XXX−XXX
D
Global GHG emissions inc ease o e ime, p ima ily due o
he inc easing p oduc ion and use o palm oil o subs i u e o
dec eased US soybean oil a ailabili y o ood and o he
non uel uses (Figu e S2 in ESM). The ecu si e-dynamic
ILUC-GHG alue, which, in con as o he compa a i e-s a ic
alue, accoun s o hese lagged e ec s o he shock as he
model ecu si ely sol es imes eps h ough 2050, is 33.6
gCO2e/MJ (Figu e 1). Mos emissions come om pea land
oxida ion (38.8 gCO2e/MJ) while na u al land con e sion and
associa ed biomass losses con ibu e wi h 9.6 gCO2e/MJ.
Ag icul u al biomass g ow h o se s emissions by 25.4 gCO2e/
MJ, mos ly h ough inc eases in oil palm a eas ollowed by
co n and o he ce eals o he li es ock ebound e ec (Figu e
1) (see eq 1 in Appendix).
3.2. Sensi i i y Analysis. 3.2.1. One-a -a-Time Sensi i -
i y Analysis. In he OAT analysis, some, bu no all, o he
pa ame e al e a ions p oduce conside able a ia ion in he
ILUC-GHG alues ela i e o he cen al case (Figu e 2). Fo
he compa a i e-s a ic ILUC-GHG alue (mean ±s anda d
de ia ion, gCO2e/MJ), he la ges a ia ions a e associa ed
wi h he ade elas ici y o ege able oils (35.8 ±13.6), he
expansion esponse o palm in o pea land (29.4 ±12.3), and he
EF o ca bon seques a ion in biomass in palm plan a ions (29.9
±12.3). These esul s unde pin he indings o he cen al case
and highligh he impo ance o land use spillo e s in he SEA
and SAM egions as majo con ibu o s o ILUC-GHG alues
and hei unce ain y. The demand elas ici y o ege able oils
(38.3 ±10.7) and he subs i u ion elas ici y among ege able oils
(33.0 ±10.7), a e ound o impac ILUC-GHG alues as hese
de e mine how much palm oil is subs i u ed in he in e na-
ional ma ke in esponse o changes in p ices, he eby al e ing
whe e oilseed p oduc ion expands a he ma gin. Land
expansion in o na u al ege a ion (35.3 ±10.4) was ound o
impac ILUC-GHG alue anges by a ec ing land use
expansion dynamics mainly in he US and SEA. The es o
pa ame e s assessed ha e mino e ec s in he compa a i e-
s a ic ILUC-GHG alue, mainly de e mining c opland a ea
equi emen s o mee oilseed demand o biodiesel uses and
u he c op demand o animal eed (yield elas ici y,demand
elas ici y o animal p oduc s, and exogenous yield p ojec ion); as
well as he GHG emissions in ensi y o o es land co e loss
and pea land oxida ion.
The anges o ecu si e-dynamic ILUC-GHG alues widen
o mos pa ame e s, pa icula ly, he economic pa ame e s
which go e n adjus men s in in e na ional ade and oilseed
ma ke s. The ade elas ici y o ege able oils (44.2 ±18.8
gCO2e/MJ) shows he la ges a ia ions. The ole o
pa ame e s de e mining he magni ude o he li es ock
ebound e ec in ensi ies wi h he ecu si e-dynamic ILUC-
GHG alue. Fo ins ance, he yield elas ici y (44.7 ±16.1), and
he demand elas ici y o animal p oduc s (44.0 ±14.1) ha e
sizable impac s on he ILUC-GHG alue anges. The demand
elas ici y o ege able oils (34.2 ±14.1) and he land expansion
in o na u al ege a ion (53.5 ±18.0) emain e y in luen ial,
while he subs i u ion elas ici y among ege able oils (32.1 ±3.5)
ma e s less o ILUC-GHG alue a iabili y ela i e o he
compa a i e-s a ic ILUC-GHG alue.
3.2.2. Mon e Ca lo Analysis. To es sensi i i y o he
GLOBIOM esul s o di e en assump ions, he OAT
es ima es p esen ed abo e we e complemen ed wi h a ull
MC analysis a ge ing he same 11 pa ame e s in he model.
Unde a MC analysis, each pa ame e is assigned an assumed
Figu e 2. Top and bo om panels show he dis ibu ion o ILUC-GHG alues (gCO2e/MJ) o each pa ame e in he OAT analysis o he
compa a i e-s a ic and ecu si e-dynamic ILUC-GHG alues. Ba s ep esen he 10 h and 90 h pe cen ile. Whiske s ep esen he minimum and
maximum alues. Solid e ical lines ep esen he ILUC-GHG alue in he cen al scena io.
En i onmen al Science & Technology pubs.acs.o g/es A icle
h ps://doi.o g/10.1021/acs.es .3c09944
En i on. Sci. Technol. XXXX, XXX, XXX−XXX
E
p obabili y dis ibu ion and a la ge numbe o simula ions is
pe o med, based on a numbe o simul aneous andomized
d aws om each dis ibu ion o pa ame e s. This me hod
allows es ing o he o e all sensi i i y o he model esul s o
pa ame ic unce ain y. Howe e , his app oach does no ouch
upon he decision unce ain y ela ed o model design.
The e o e, hese esul s do no ep esen a comp ehensi e
p obabilis ic es ima e o ILUC-GHG alues o his modeling
amewo k, bu a he a p obabilis ic es ima e speci ic o he
cu en model s uc u e. While he OAT iden i ies he e ec o
a ying indi idual pa ame e s on ILUC-GHG alues, he MC
analysis allows de i a ion o unce ain y anges when all es ed
Figu e 3. Dis ibu ion o compa a i e-s a ic and ecu si e-dynamic ILUC-GHG alues (gCO2e/MJ) and unde lying emissions sou ces ob ained
om he MC analysis. Boxes show alues be ween he 10 h and 90 h pe cen iles. The uppe whiske is he maximum and he lowe whiske is he
minimum alue when excluding ou lie s acco ding o he “1.5 ule;” an es ima e is conside ed an ou lie i i is < Q1−1.5 x IQR o > Q3 + 1.5 x
IQR, whe e IQR is he in e qua ile ange. This is mean o exclude esul s ob ained om he combina ion o he mos ex eme alues o he
pa ame e s, gi en he linea p og amming o mula ion o he GLOBIOM model.
Figu e 4. Dis ibu ion o egional compa a i e-s a ic (uppe panels) and ecu si e-dynamic (lowe panels) ILUC-GHG alues (gCO2e/MJ)
ob ained om he MC analysis (le ) and decomposi ion o he co esponding s anda d de ia ion by emissions sou ce ( igh ). The boxes in he box
and whiske s plo show alues be ween he 10 h and 90 h pe cen iles; he uppe whiske is he maximum and he lowe whiske is he minimum
alue when excluding ou lie s acco ding o he “1.5 ule.” An es ima e is conside ed an ou lie i i is < Q1−1.5 x IQR o > Q3 + 1.5 x IQR, whe e
IQR is he in e qua ile ange. This is mean o exclude esul s ob ained om he combina ion o he mos ex eme alues o he pa ame e s, gi en
he linea p og amming o mula ion o he GLOBIOM model. Sou h Ame ica includes B azil, A gen ina, and es o Sou h Ame ica.
En i onmen al Science & Technology pubs.acs.o g/es A icle
h ps://doi.o g/10.1021/acs.es .3c09944
En i on. Sci. Technol. XXXX, XXX, XXX−XXX
F
pa ame e s a e andomly and simul aneously a ied. The MC
analysis inds ILUC-GHG alues o US soybean biodiesel
be ween he 10 h and 90 h pe cen iles anging om 15.1 o
67.7 gCO2e/MJ in compa a i e-s a ic and om 8.4 o 77.4
gCO2e/MJ in ecu si e-dynamic app oaches (Figu e 3). The
ull compa a i e-s a ic ( ecu si e-dynamic) ILUC-GHG alue
ange spans om a minimum alue o −17.0 (−23.7) gCO2e/
MJ o a maximum alue o 98.7 (112.8) gCO2e/MJ when
excluding ou lie s, wi h bo h dis ibu ions gene ally symme ic.
The mean alues and sp ead o alues sligh ly inc ease in he
ecu si e-dynamic ILUC-GHG alue (42.4 ±25.9 gCO2e/MJ)
ela i e o he compa a i e-s a ic one (40.8 ±20.5 gCO2e/
MJ), mainly h ough inc eased emissions (and associa ed
unce ain y) om pea land oxida ion (Figu e 3). Mos ILUC-
GHG alues all on he posi i e side o he dis ibu ion, wi h
98.8% and 94.7% o uns being abo e ze o in he compa a i e-
s a ic and ecu si e-dynamic ILUC-GHG alues, espec i ely.
Emissions om na u al land con e sion and pea land
oxida ion a e he mos in luen ial emissions sou ces, showing
a wide dis ibu ion han all o he s (Figu e 3), wi h a s anda d
de ia ion o 13.5 and 11.4 gCO2e/MJ, espec i ely, in he
compa a i e-s a ic app oach; 18.2 and 12.3 gCO2e/MJ in he
ecu si e-dynamic one. Na u al land con e sion con ibu es
47% and 58%; and pea land emissions con ibu e 34% and
26% o he o al a iance o he compa a i e-s a ic and
ecu si e-dynamic ILUC-GHG alues, espec i ely. Despi e
he la ge a ia ion, he dis ibu ions o hese wo sou ces o
emissions emain posi i e wi hin he 10 h and 90 h pe cen iles,
which is consis en wi h indings om he cen al case and
OAT analyses (see Sec ions 3.1 and 3.2.1). Con e sely, he
dis ibu ion o ag icul u al biomass emissions alls en i ely o
he le o he y-axis in Figu e 3, indica ing ne global ca bon
seques a ion in ag icul u al biomass. This s ems om a
signi ican po ion o he added ag icul u al biomass coming
om palm plan a ions. Ag icul u al biomass is a he na owly
dis ibu ed a ound he mean alue wi h a con ibu ion o o al
a iance o 7% (6%) in he compa a i e-s a ic ( ecu si e-
dynamic) se up. The emaining emissions om SOC and
na u al land e e sion span om nega i e o posi i e and a e
cen e ed a ound ze o. SOC con ibu es 12% (10%) o he o al
a iance in he compa a i e-s a ic ( ecu si e-dynamic) ILUC-
GHG alues, while na u al land e e sion plays a ma ginal ole
(<1% o he a iance). As o o al ILUC-GHG alues, all
emissions sou ces’ dis ibu ions a e mo e sp ead in he
ecu si e-dynamic se up han in he compa a i e-s a ic one,
including SOC (0.3 ±7.6 s 2.7 ±6.7 gCO2e/MJ), na u al
land con e sion (26.2 ±18.2 s 27.9 ±13.5 gCO2e/MJ),
pea land oxida ion (37.4 ±12.3 s 34.3 ±11.4 gCO2e/MJ),
and ag icul u al biomass (−22.8 ±5.7 s −21.9 ±5.1 gCO2e/
MJ). This esul is due o he inc eased unce ain y in he wo
main mechanisms d i ing ILUC-GHG alues, namely he
subs i u ion in ege able oil and eed eeds ock ma ke s, as well
as in he subsequen li es ock ebound e ec .
The wides unce ain y anges in bo h compa a i e-s a ic
and ecu si e-dynamic ILUC-GHG alues, espec i ely, a e
ound o SEA (38.0 ±15.3; 41.8 ±18.1 gCO2e/MJ) and
SAM (−7.2 ±7.4; −5.3 ±14.1 gCO2e/MJ), while mean
ILUC-GHG alues o he US emain a he low and na ow
(4.7 ±3.7; 6.4 ±4.4 gCO2e/MJ) (Figu e 4). SEA espec i ely
con ibu es 73.5% and 54.8% o he compa a i e-s a ic and
ecu si e-dynamic ILUC-GHG alue a iance. Emissions om
pea land oxida ion a e again a key d i e o he a ia ion,
con ibu ing 54% (48%) o he a iance o SEA’s compa a i e-
s a ic ( ecu si e-dynamic) ILUC-GHG alue, ollowed by
na u al land con e sion wi h 25% (34%). Fo SAM, egional
ILUC-GHG alue a ia ion is less p onounced han o SEA
(Figu e 4). ILUC-GHG alues o SAM a e nega i e in 71%
(60%) o he MC simula ion uns in he compa a i e-s a ic
( ecu si e-dynamic) se up. These nega i e ILUC-GHG alues
co espond o hose MC simula ions wi h lowe de o es a ion
a es in B azil and A gen ina, coupled wi h highe assumed EF
om o es biomass loss alues, which ansla es in o ne ca bon
gains compa ed o he baseline. MC uns ha esul in
c opland expansions in SAM a e la gely showing highe
demand elas ici ies o ege able oils, highe yield p ojec ion
esponses, lowe ege able oil subs i u ion elas ici ies, and
lowe yield elas ici ies (Sec ion S2.4). This leads o a
ma ginally g ea e demand o soybean, co n, and soybean
Figu e 5. Ma ginal e ec s on he ILUC-GHG alue (gCO2e/MJ) om he a ia ion o key model pa ame e s in he MC analysis. Only he mos
in luen ial pa ame e s a e shown. The bands a ound he lines indica e he 95% con idence in e al o he ma ginal e ec . The alues in he x-axis
indica e he pe cen ile wi hin he de ined dis ibu ions o shi e alues applied o he pa ame e s as desc ibed in Table S1. Ag icul u al biomass
emissions ac o : EF o ag icul u al biomass in palm plan a ions; o es emissions ac o : EF om o es biomass loss; palm o pea expansion esponse:
expansion esponse o palm in o pea land.
En i onmen al Science & Technology pubs.acs.o g/es A icle
h ps://doi.o g/10.1021/acs.es .3c09944
En i on. Sci. Technol. XXXX, XXX, XXX−XXX
G
oil om SAM used o mee demand o ood, eed, and o he
uses in egions ou side o he US, while palm oil con inues o
ha e a limi ed ma ke pene a ion in he global mix.
The BART analysis shows ha he EF om o es biomass
loss causes he wides ma ginal a ia ion o ILUC-GHG alues
unde he compa a i e-s a ic app oach wi h a 95% con idence
in e al; while ha e y pa ame e and especially he expansion
esponse o palm in o pea land cause he sha pes a ia ion in
ecu si e-dynamic ILUC-GHG alues, as he la e di ec ly
in luences unce ain y in pea land emissions. The ma ginal
e ec s o he EF o ag icul u al biomass in palm plan a ions and
he pea land EF also become ma ked unde he wo app oaches
(Figu e 5). The yield p ojec ion esponse has a mo e i egula
e ec , especially unde he compa a i e-s a ic app oach, gi en
he ole ha his pa ame e plays in de e mining c opland a ea
dynamics in US and SAM. On he one hand, he highes shi e
alues esul in ne c opland inc eases in SAM; on he o he
hand, he lowes alues imply la ge a eas con e ed in o soy in
US. Unde he ecu si e-dynamic app oach, yields end o
inc ease o all egions and c ops o e he pe iod conside ed,
con ibu ing o mi iga e global c opland a ea expansion and
hus ILUC as he shi e inc eases. The ma ginal e ec o he
economic pa ame e s inc eases unde he ecu si e-dynamic
app oach, hough emains mo e mode a e han ha o he
biophysical ones. The ma ginal e ec o he ege able oil
subs i u ion elas ici y becomes especially unce ain wi h he
highes alues. Unlike he OAT analysis, BART unde lines he
mo e p ominen ole o biophysical pa ame e s in d i ing
ILUC esul s unce ain y when p opaga ed wi h MC
simula ions.
4. DISCUSSION
4.1. Majo Ma ke -Media ed Responses. Inc easing
demand o bio uels p oduced om oilseeds gene a es di e se
ma ke -media ed e ec s and c oss-sec o al in e ac ions, which
can ul ima ely ansla e in o in ensi ied compe i ion o land
among all uses and inc eased land p ices.
72,79−81
These
cha ac e is ics make he po en ial spillo e e ec s o ege able
oil demand shocks complex, es ablishing he need o
sys ema ic unce ain y analysis when es ima ing ILUC-GHG
alues wi h global economic models. Ou s udy imp o es he
unde s anding o he esponses ac oss ege able oil and oilseed
ma ke s and p o ides addi ional e idence ha modeled
soybean biodiesel ILUC-GHG alues a e highly sensi i e o
bo h economic and biophysical pa ame e s. OAT analysis and
he combina ion o BART and MC analyses help iden i y
model pa ame e s ela ed o ege able oils ma ke s, and in
pa icula GHG emissions om oil palm expansion, as key
de e minan s o ILUC-GHG alues and anges. Resul s
highligh he ole o pa icula egions, SEA and SAM, whe e
land use spillo e s om he US soybean oil biodiesel
consump ion shock a e signi ican . The le el o es ima ed
c opland expansion and associa ed p essu es on na u al
ecosys ems in SEA and SAM in modeled esul s depends
signi ican ly on he assumed ege able oil subs i u ion elas ici y −
he ease wi h which ma ke s may shi consump ion om one
ege able oil o ano he in esponse o demand shocks−and
he ege able oil demand elas ici y − he ease wi h which
consump ion o ege able oils o e all may dec ease when
p ices ise. These pa ame e s al e how much addi ional
soybean and palm oil is p oduced in SAM and SEA
espec i ely, he eby con ibu ing o he in e play be ween
ege able oil demand in he US and enhanced p oduc ion
ab oad.
Ac oss he MC uns, GLOBIOM consis en ly es ima es
inc easing palm oil use as he p ima y subs i u e o di e ed
soybean oil om ood and o he uses in in e na ional ma ke s,
causing na u al land con e sion and pea land emissions in
SEA. This is consis en wi h se e al o he CLCA s udies ha
iden i y palm oil om SEA as he mos cos -compe i i e bu
emission in ensi e
82
ma ginal sou ce o ege able oil eeds ock
in he ma ke .
19,45,83,84
This ma ke -media ed e ec is also
iden i ied in GTAP-BIO s udies,
55
especially in scena ios wi h
ela i ely highe subs i u ion elas ici ies ha only conside soy
oil-palm oil subs i u ion. The sha e o Malaysia and Indonesia
in he es ima ed ILUC-GHG emissions alue o soybean
biodiesel (17.5 g CO2e/MJ, + 2 billion gallons) is 78%; in ou
case i is <50% (see Figu e 1). I o he ege able oils and
animal a s a e aken in o accoun , mos o he addi ional
demand o soybean oil is di e ed o hose p oduced in he
US o in egions o he han he SEA.
55
This highligh s he ole
o he elas ici y o subs i u ion among he a ious ypes o
biodiesel eeds ocks, especially when o he kinds o ege able
oils and animal a s a e conside ed. In his way, we could
expec lowe ILUC-GHG alues when including mo e
eeds ocks a ailable in he US, such as allow o used cooking
oil, as combined wi h g ea e assumed elas ici ies. This would
allow o nonpalm ege able oils and animal a s o subs i u e
o soybean oil mo e han in ou esul s, hence dec easing
demand o palm oil and land con e sion in SEA.
This di e ence among indings highligh s he decision
unce ain y exis ing among models, which may be pa ially
ela ed o he elas ici y o subs i u ion among he a ious ypes
o ege able oils and animal a s in he US; g ea e assumed
elas ici y may allow o nonpalm ege able oils and animal a s
o subs i u e o soybean oil mo e so han we ha e ound in
hese esul s. Con e sely, GLOBIOM gene ally es ima es a
slowing a e o na u al land con e sion in SAM. B azil loses
ma ke sha e in global soybean ma ke s h ough he educed
demand o B azilian soybean meal o eed applica ions due o
he inc eased a ailabili y o ela i ely cheap US soybean meal.
The MC analysis s ill inds a mino i y o uns (6.3% in 2030,
and 45.9% in 2050), in which c opland expands in SAM,
inc easing ILUC-GHG alues. This e ec eme ged ac oss
di e se a ays o inpu pa ame e alues and could no be
explained by any one speci ic pa ame e . Since bo h SAM and
SEA include ca bon- ich o es s and na u al lands, any
c opland expansion in hese egions will impac he ILUC-
GHG alue o US soybean biodiesel (Figu e S7) and
associa ed unce ain y anges (Figu e 4). This highligh s he
impo ance o he ou biophysical pa ame e s o he MC
analysis, as hese de e mine he ne GHG emissions pe
hec a e when o es is con e ed in o c opland and he o al
pea land emissions om palm expansion. GLOBIOM con-
sis en ly es ima es addi ional c opland expansion in majo
g ain-p oducing egions ou side he US o mee inc eased
demand o animal p oduc s when soybean meal a ailabili y
inc eases. This li es ock ebound e ec
78
is also subjec o
unce ain y in he pa ame e s conside ed and con ibu es
addi ional GHG emissions in egions such as SAM, SAS, and
he US, especially unde he ecu si e-dynamic app oach in
which he impac s o unce ain y in hese pa ame e s a e
accoun ed o o e mul iple decades.
Me hodological Conside a ions. The ma ke -media ed
esponses summa ized abo e highligh he in eg a ion o
En i onmen al Science & Technology pubs.acs.o g/es A icle
h ps://doi.o g/10.1021/acs.es .3c09944
En i on. Sci. Technol. XXXX, XXX, XXX−XXX
H
globalized ood, eed, and uel ma ke s. Simila e ec s could be
expec ed when simula ing inc eased demand o o he oilseeds,
especially hose wi h meal cop oduc ion, such as apeseed.
40
Unce ain ies pe sis in es ima es o p ojec ed u u e impac s
om ege able oil-based uels. Ou esul s highligh he need
o be e empi ical es ima es o c i ical economic pa ame e s
such as elas ici ies o ege able oil subs i u ion and consume
demand elas ici ies o ege able oil p ices. Fu he e o s a e
needed o e ine es ima es on biophysical pa ame e s, such as
EFs om land con e sion bene i ing om imp o ing spa ially
explici da a on yields, land a eas, and associa ed ca bon
s ocks.
31
Inc eased a ailabili y o emo e sensing da a and
associa ed p ocessing capaci ies can help o be e cha ac e ize
he inhe en unce ain y ela ed o he a iabili y o he
modeled sys ems, e.g., spa ial a iabili y in ca bon s ocks.
25
,
42
Howe e , gi en he size and ma hema ical complexi y o global
economic models, epis emic unce ain y will always emain a
challenge when es ima ing and in e p e ing ILUC-GHG
alues.
MC analysis is o en used o assess sensi i i y and
unce ain y in esul s om global economic models. Howe e ,
he numbe o a ied model pa ame e s ends o be limi ed.
This s udy uses MC analysis o explo e he ILUC-GHG alue
sensi i i y o pa ame ic unce ain y, whe e bo h he numbe o
pa ame e s and hei indi idual alue anges and dis ibu ions
ha e been chosen in ad ance, essen ially de ining he assumed
likelihood ha each pa ame e akes a alue wi hin a ce ain
ange. Al hough his is a common p ac ice o unce ain y
analysis,
14,49,85
o he impo an pa ame e s may be o e looked
by his app oach. I should also be no ed ha o many
pa ame e s, he p obabili y dis ibu ion is no e y well-known,
such ha epis emic unce ain y may no be ully ep esen ed.
The decision o which pa ame e s o include and how o shape
hei s ochas ic dis ibu ion o alues is a c i ical analy ical
choice by he modele . This highligh s he impo ance o
anspa ency ega ding he choice o pa ame e s, model
s uc u e and unde lying assump ions o CLCA, while
assessing ade-o s on di e en empo al and spa ial
scales.
37,42,43,86
The MC analysis p esen ed helps unde s and
model beha io when speci ic model inpu s a e modi ied
a bi a ily, as a mean o p opaga e unce ain y in ILUC-GHG
alues. Howe e , a mo e comp ehensi e unce ain y analysis
would equi e a p e ious cha ac e iza ion o an expec ed ange
o possible beha io s o da a and would need o ca y ha
cha ac e iza ion h ough o a ange o implied possible
ou comes.
64,87
Despi e he conside a ions abo e, his s udy co e s a
comp ehensi e se o pa ame e s and dis ibu ions in he
MC simula ion based on ecen li e a u e and expe judgmen
(Table S1 in ESM). Se en economic pa ame e s and ou
biophysical pa ame e s a e a ied, expanding he numbe o
pa ame e s a ied ela i e o p e ious GLOBIOM unce ain y
s udies.
40
The assumed dis ibu ions could be e ined u he ,
o ins ance, by de ining he elas ici ies based on econome ic
me hods and empi ic app oaches, o by using imp o ed, ine -
scale, and mo e up- o-da e emo e sensing da a o de e mine
he s ochas ic dis ibu ion o biophysical pa ame e s ac oss he
globe. This is ou o he scope o he p esen s udy bu should
be add essed in u u e wo k. Gi en he nonlinea na u e o
ILUC-GHG alues, ou MC analysis shows ha including
mo e pa ame e s does no necessa ily yield wide ILUC-GHG
alue anges, as many pa ame e s in e ac , especially he
economic ones ha a ec se e al egions and commodi ies
simul aneously. This s udy applies BART as a ool o be e
unde s and he ole o each pa ame e in he a ia ion o
ILUC-GHG alues, aiding he in e p e a ion o MC analysis
esul s om global economic models. S ill, MC analysis p o es
use ul o p opaga e unce ain y in inpu pa ame e s and ob ain
p obabili y dis ibu ions o ILUC and o he model-based
ou comes.
86,88
O he c ucial modeling choices include he magni ude o
eeds ock consump ion simula ed exogenously, he bio uel
shock size, and he baseline used as coun e ac ual.
18,40
Fu he mo e, he ime ame o e which a bio uel shock is
e alua ed can ha e a signi ican impac on esul ing ILUC-
GHG alue.
86
These choices explain di e ences in he mean
ILUC-GHG alue es ima ed in he MC analysis in his s udy
(be ween 40.8 and 42.7 gCO2e/MJ, depending on he
app oach) as compa ed o p e ious assessmen s such as P ussi
e al.
48
The compa a i e-s a ic app oach may be mo e
app op ia e when one seeks o unde s and nea - e m ma ke
dynamics, as i elies on only assump ions ele an o he nex
immedia e u u e ime s ep and p o ides be e compu a ional
e iciency. The ecu si e-dynamic app oach may be mo e
app op ia e when seeking o unde s and longe - e m bioeco-
nomic dynamics, such as impac s on ILUC-GHG emissions, as
i be e cap u es he inhe en empo al a ia ion and
unce ain y in such es ima es. Mos ecen li e a u e has
conside ed only one o hese app oaches, wi h he p edom-
inan ocus on he compa a i e-s a ic app oach. The wo
me ics apply di e en empo al scopes, hence p o iding
di e en insigh s since esul s a y depending on he baseline
de elopmen s and unde lying assump ions. Fo example, he
ecu si e-dynamic ILUC-GHG alue (up o 2050) also
conside s o he impo an d i e s o e ime, e.g., he ac
ha yields a e la ge in 2050, which educes he impac o he
bio uel shock. This s udy’s explici compa ison o compa a i e-
s a ic and ecu si e-dynamic app oaches in an o he wise
consis en modeling amewo k demons a es he me hodo-
logical choices modele s should conside in he ace o model
unce ain y, pa icula ly ela ed o he ea men o ime. When
elying on bio uel CLCA modeling o in o m decision-making,
modele s should acknowledge which unce ain ies a e and a e
no ep esen ed in hei suppo ing analyses, including ela ed
o assump ions implici o u ilized me hods o quan i ying
ILUC-GHG impac s.
■APPENDIX
CONV: na u al land co e con e sion including abo e- and
below-g ound biomass, dead wood, li e , and ha es ed wood
p oduc s
REV: na u al land co e e e sion o o es eg ow h
BIOM: ca bon seques a ion in ag icul u al biomass o
bioene gy c ops
SOC: soil o ganic ca bon
PEAT: GHG emissions om pea land oxida ion
= [ +
+ ] ×
LUC M CO CONV M C BIOM M C
REV M C
( ) ( ) ( )
( ) 44
12
y y y
y
2
(1)
En i onmen al Science & Technology pubs.acs.o g/es A icle
h ps://doi.o g/10.1021/acs.es .3c09944
En i on. Sci. Technol. XXXX, XXX, XXX−XXX
I
