110530 ecofys gbep iluc report with e4tech

Indirect effects of biofuel production
Overview prepared for GBEP

A SUSTAINAB LE ENE RGY SUPPLY FOR EVERYONE

Indirect effects of biofuel production
Overview prepared for GBEP

By: Bart Dehue, Stijn Cornelissen and Daan Peters
Date: 18 May 2011

Project number: PSUPNL102012

© Ecofys 2011
by order of: Dutch Ministry of Infrastructure and Environment and the
United Kingdom’s Department of Energy and Climate Change

E COFYS N E TH ERLAN DS BV, A PRI VATE LIMITED LIABILITY COM PANY INCORPORATED UNDER TH E LAW S OF T H E N E THERLAN DS H AVING ITS OFFICIAL SE AT AT U TRECH T
AN D REGI STERED WITH THE TRADE REGI STER OF TH E C HAM BE R OF C OMMERCE IN MIDDEN NEDERLAND UNDER FILE NUMBER 30161191


Foreword

This report was prepared to provide input to discussion on the indirect effects of
biofuel production and consumption in the Global Bioenergy Partnership (GBEP).
Financial support was kindly provided by the United Kingdom’s Department of Energy
and Climate Change and the Dutch Ministry of Infrastructure and Environment. This
document does not necessarily reflect the views of GBEP, the Government of the
United Kingdom or the Government of the Netherlands.

19 May 2011 i


Table of contents

1 Introduction to indirect effects of biofuel production ............................... 1

1.1 The importance of bioenergy ................................................................... 1

1.2 Bioenergy and sustainability.................................................................... 1

1.3 About this report ................................................................................... 4

2 Quantitative review of work on pathway-specific indirect effects............. 6

2.1 Key findings .......................................................................................... 6

2.2 Quantification of indirect effects .............................................................. 7

2.3 Key assumptions in quantifying indirect effects.........................................11

2.4 Overview of reviewed quantitative work on indirect effects ........................14

2.5 Review of pathway-specific indirect effects – understanding the differences 16

2.6 Discussion on key assumptions in E4tech-study........................................40

2.7 Conclusion ...........................................................................................50

3 Mitigation of unwanted indirect effects .................................................. 53

3.1 Possible ways to mitigate unwanted indirect effects from bioenergy ............53

3.2 Summary of existing mitigation initiatives................................................54

3.3 Detailed analysis of mitigation initiative: RFS ...........................................56

3.4 Detailed analysis of mitigation initiative: LCFS..........................................56

3.5 Detailed analysis of mitigation initiative: EU Renewable Energy Directive.....57

3.6 Detailed analysis of mitigation initiative: Low Indirect Impacts Biofuels .......58

3.7 Conclusion on current mitigation measures ..............................................60

Reference sources ........................................................................................ 63

Appendix A Emissions from additional fertilizer ...................................... 66

Appendix B The debate on the effectiveness of expanding on ‘idle land’
and its links to the RCA initiative ............................................................ 68

19 May 2011 ii


1 Introduction to indirect effects of biofuel production

Bioenergy plays an important role in decarbonising our economy and stimulating
sustainable development. The production and consumption of bioenergy will have both
direct and indirect effects. This chapter sets out the key indirect effects of bioenergy
production and consumption and the main mechanisms causing these indirect effects.
It thereby lays the foundation for the subsequent detailed analyses on the size of
indirect effects and how these can be mitigated.

1.1 The importance of bioenergy

Bioenergy production has seen a sharp growth in recent years. Key drivers include
reduction of greenhouse gas emission, energy security and rural development. While
large scale bioenergy production has met concerns about sustainability, it is important
to note that bioenergy will play an important role in a decarbonised economy. This is
due to the fact that for several sectors with a large and growing energy demand few
alternatives exist. This includes aviation, shipping, road freight transport and
industries requiring high temperature heating. Furthermore, many developing
countries are currently heavily reliant on traditional use of biomass for heating and
cooking. A transition to more sustainable forms of bioenergy can help achieve
development and climate change mitigation goals. Developing a successful and
sustainable bioenergy sector is therefore of key importance to the decarbonisation of
the economy and to sustainable development around the world.

1.2 Bioenergy and sustainability

While bioenergy is a key form of renewable energy for low carbon sustainable
development, the recent increase in bioenergy production has led to concerns about
the sustainability of such large scale production. Key concerns include deforestation
for energy crops, greenhouse gas (GHG) emissions from land-use change, impacts on
the local environment and competition with food. On the other hand, biofuel
production can contribute to economic development and energy security, e.g. through
job creation, value addition and displacement of fossil fuel imports. The effects of
bioenergy production and consumption can be divided into direct effects and indirect
effects.

1.2.1 Direct effects
The direct effects of bioenergy production and consumption are a direct result of the
activities needed to produce the bioenergy, for example the effects of crop cultivation
on soil, air and water. As long as the location of production and processing is known,
these direct effects can be monitored. GBEP currently develops indicators for a wide
spectrum of such direct effects, including economic, energy security, social and
environmental effects.

19 May 2011 1


One of the main direct effects is direct land-use change (LUC). A direct LUC occurs
when new areas (e.g. forest areas or grasslands, see circles A in Figure 1-1) are taken
into production to produce the additional feedstock demand for bioenergy. This can
have both positive and negative consequences on aspects such as biodiversity, carbon
stocks and livelihoods.

Direct LUC effects and other direct effects of crop production can generally be
measured and attributed to the party that caused them. These properties make direct
LUC relatively easy to control. The development of voluntary certification schemes
such as the Roundtable on Sustainable Palm Oil and the Round Table on Responsible
Soy aim to prevent negative direct effects from crop cultivation. However, as long as
not all worldwide production is controlled by such certification schemes, effective
enforcement of land-use planning, or alternative control mechanisms, such
mechanisms are not able to fully control indirect effects.

1.2.2 Indirect effects
Much of the feedstock use for bioenergy today is sourced from existing plantations,
especially since many of today’s biofuel feedstocks are food and feed crops. In this
case, no direct effects take place during the feedstock cultivation phase, but so-called
indirect effects can take place. The main indirect effects of additional bioenergy
feedstock demand are1:
• Indirect land-use change, explained in more detail below;
• Rise in agricultural commodity prices, with potential consequences for food
security;
• Demand-induced yield increases – where the additional demand for the feedstock
triggers additional yield increases (Ecofys 2009b).

The indirect effect that is currently dominating much of the debate on the
sustainability of biofuels is indirect land-use change (ILUC). ILUC can occur when
existing plantations (see circle B) are used to cover the feedstock demand of
additional biofuel production. This displaces the previous productive function of the
land (e.g. food production). This displacement can cause an expansion of the land use
for biomass production to new areas (e.g. to forest land or to grassland, see circles
B’and B’’) if the previous users of the feedstock (e.g. food markets) do not reduce
their feedstock demand and any demand-induced yield increases are insufficient to
produce the additional demand. Where this indirect LUC will take place is uncertain
and is out of control of the bioenergy sector.

1
For a detailled discussion on these indirect effects and a review of existing modeling work that aims to
quantify the sizes of these idirect effects, see “Summary of approaches to accounting for indirect impacts”
(Ecofys 2009b).

19 May 2011 2


Figure 1 - 1: Illustration of the displacement mechanisms that may cause indirect land-use change.
Adapted from (Dehue 2006)

Direct land-use
change

Forest
A

B’
Existing
?
Plantations Indirect land-use
change
F B ?
B’’

Grassland

1.2.3 Key characteristics of indirect effects
Since land requirements are a key concern for environmental, social and economic
sustainability issues, controlling direct and indirect LUC effects is a major challenge to
ensure a sustainable energy crop production. Several key characteristics of ILUC are
summarised in the Box below. Any mechanism aiming to resolve indirect effects will
need to take these complexities into account.

1.2.4 Note on direct LUC caused by food, feed and fibre sectors
This report is focussed on the indirect effects of bioenergy. It should be noted that
unwanted effects from indirect LUC from bioenergy manifest itself through unwanted
direct LUC for the production of agricultural products for other sectors such as the
food and feed sector. Preventing unwanted direct LUC would thus eliminate unwanted
indirect LUC altogether and is the optimal long term solution for unwanted LUC.
However, because of the international characteristics of ILUC and the competition for
land between different sectors, this mitigation measure requires global implementation
for all land-based sectors to be effective. Until this is achieved, and if biofuels are to
meet their policy goals such GHG savings, intermediate solutions will need to be
implemented that acknowledge the lack of control of sustainability in other biomass
consuming sectors.

19 May 2011 3


Box 1: Key characteristics of indirect land-use change

• Displacement effects act across national border. Commodities such as palm oil, soy oil
and sugarcane are traded on a global scale. Therefore, displacement effects act across
borders. Achieving effective national land-use planning in some producing countries
should therefore not be taken as full protection against indirect effects. If, for example,
Indonesia were to prevent further deforestation through effective land-use planning,
sourcing increasing amounts of palm oil from Indonesia for the energy sector may still
cause indirect land-use change in other producing countries such as Malaysia.

• Displacement effects act across substituting crops. This is caused by the fact that
different crops can substitute each other to some extent. For example, if the EU diverts
more rapeseed oil production from food to feed then it is likely to increase its imports
of vegetable oils. This could be rapeseed oil but could also be a different vegetable oil
as different vegetable oils are to some degree substituting products. Thoenes (2007)
states that “EU palm oil imports have already doubled during the 2000-2006 period,
mostly to substitute for rapeseed oil diverted from food to fuel uses.”
• Competition for land connects also non-substituting crops. Another reason why
displacement effects act across crops is that different (non-substituting) crops can
compete for the same agricultural land. A recent example of this occurred in 2008
when high maize prices led farmers in the US to plant more maize and less soy (USDA
2010), which could trigger soy expansion in other world regions.

1.3 About this report

This report intended to inform the GBEP workstream on indirect effects of bioenergy.
Its aim is to provide an up-to-date insight into the science relating to ILUC from
bioenergy: the mechanisms that cause ILUC; the various approaches to quantifying
ILUC; the extent to which these approaches converge or diverge (and why); and the
various approaches that are being developed to mitigate ILUC. A thorough
understanding of this science will be important for the work of GBEP – namely the
development of voluntary criteria and indicators for the effects of bioenergy
production and consumption, including such criteria and indicators for the indirect
effects of bioenergy production.

In order to provide an up-to-date insight into the science of ILUC, this report is
structured as follows:
• Chapter 1 gives a general introduction into the mechanisms that cause indirect
effects and discusses the key characteristics of such indirect effects. It lays the
foundation for a more detailed analysis of ILUC in subsequent chapters.
• Chapter 2 reviews some of the key efforts to quantify the indirect effects of
bioenergy production. It explains what the main approaches are to quantify
indirect effects and quantitatively reviews a selection of existing ILUC
quantification work. This quantitative review does not merely compare the

19 May 2011 4


differences in outcomes but aims to explain those differences by looking at the
underlying assumptions in more detail. This report thereby does not aim to
identify the correct number for ILUC but aims to generate an understanding of
what causes the studies to find different numbers, and what therefore are the key
parameters that determine the size of ILUC. This again provides important
insights for GBEP indicators for indirect effects.
• Chapter 3 reviews the mitigation options for unwanted indirect effects. It thereby
provides an insight into what measures can be taken to prevent unwanted indirect
effects from bioenergy production, over different time scales, by different actors.

Focus on GHG emissions from ILUC from biofuels
This report discusses the indirect effects of bioenergy production and consumption.
This chapter introduced the three main indirect effects (indirect LUC – with effects on
biodiversity, GHG emissions and other environmental, social and economic factors;
changes in agricultural commodity prices; and demand-induced yield increases). It
should be noted however that most work on indirect effects has focussed on the GHG
emissions from ILUC from liquid biofuel production. Especially Chapter 2 will therefore
contain much more information on the GHG emissions from ILUC than on the other
indirect effects, although section 2.4.1. focuses on biodiversity losses. Also, the focus
of the reviewed literature is on liquid biofuels specifically as opposed to bioenergy in
general.

19 May 2011 5


2 Quantitative review of work on pathway-specific indirect effects

In this section, we give a quantitative review of work on pathway-specific indirect
effects. As this is a complex issue, we first present a summary of our key findings in
Section 2.1. Section 2.2 and 2.3 explain the methodologies used for quantification of
the indirect effects and the key assumptions they rely on. Section 2.4 introduces the
studies we reviewed in this report. Section 2.5 provides the detailed results of the
quantitative review and explains the differences found between studies. Section 2.6
draws a conclusion on the current state of pathway-specific quantification of indirect
effects.

2.1 Key findings

140 (I)LUC found in this review
Typical value (RED)

120

100

80
gCO2eq/MJ

60

40

20

0
Palm oil [I]
Wheat [I]

Rapeseed [I]
IFPRI

IFPRI

IFPRI

Palm oil [E]
Wheat [E]

Rapeseed [E]

Sunflower
Öko/IFEU

Öko/IFEU
EPA

EPA

EPA
ICONE

Fossil fuel
Searchinger

E4tech

E4tech

Sugar beet
CARB

CARB

CARB

-20

-40 Ethanol from Ethanol from Biodiesel from soy Other pathways
corn sugarcane

Figure 2 - 2 Graphical representation of the emissions caused by (I)LUC, direct and indirect land
use change, for different biofuel pathways and different studies. For reference, typical
non-land-use change emissions for the different pathways and a fossil reference from
the EU Renewable Energy Directive (RED) have been added. Pathways labeled [E] are
from the E4tech study, pathways labeled [I] are from the IFPRI study. Note that
(I)LUC emissions found for ethanol from wheat in the E4tech study are negative.

19 May 2011 6


Using Figure 2 - 3, we present our key findings for this quantitative review:
1 Within each pathway, there is no clear consensus on the size of total
emissions from direct and indirect land-use change. A slight trend can be
seen where sugarcane generally has the lowest emissions from land-use change
(4-46 gCO2eq/MJ), followed by corn (27-103 gCO2eq/MJ, with the second highest
value being 54 gCO2eq/MJ), followed by soy (32-75 gCO2eq/MJ). In the other
pathways, a similar trend is visible: ethanol pathways score better than biodiesel
pathways. However, due to the large ranges in the results it would be premature
to draw firm conclusions, based on the studies reviewed in this report, on the
(I)LUC from ethanol versus biodiesel.
2 The differences between specific studies within a certain pathway are
attributable to differences in quantification methodology and key
assumptions on e.g. co-products, agricultural intensification, reduced demand in
other sectors and carbon stocks of converted land, as described in Sections 2.2
and 2.3.
3 These differences can be made reasonably clear by our framework for
quantitative comparison of intermediate results as presented in Section 2.5.
Unfortunately, in many instances the data to derive these intermediate results are
not available and, more importantly, a clear explanation of the causes of the
differences between feedstocks is not given by the authors. An exception to this
rule was the E4tech study, which contained a very transparent and well
documented causal-descriptive approach.
4 In general, most studies find the emissions from (I)LUC for most pathways to
be significant when compared to e.g. a fossil reference value of 80-90
gCO2eq/MJ fuel. An exception to this rule is the ethanol from wheat scenario in
the E4tech study, that finds negative (I)LUC emissions. However, the assumptions
and errors made in that scenario (see Section 2.5.4 for details) are critically
discussed in Section 2.6 of this review.

2.2 Quantification of indirect effects

Different methods are used to quantify the indirect effects of biofuels. Most do follow a
general four-step approach though. This is described in Section 2.2.1. This approach
can be executed using equilibrium modelling or a causal-descriptive approach. These
are described in Section 2.2.2 and 2.2.3.

2.2.1 Stepwise approach for quantification of indirect effects
Indirect effects occur through a series of market mechanisms, as described in Section
1.2. This means that they cannot be quantified by direct monitoring. Therefore,
studies on quantification of indirect effects always use a quantification framework that
estimates future indirect effects, e.g. an equilibrium model or a causal-descriptive
approach.

19 May 2011 7


Each study has its own exact methodology for quantification of indirect effects of
additional biofuel demand. However, a general four-step approach is commonly found
in these methodologies. This approach is visualized in Figure 2 - 3.

Figure 2 - 3 Four-step approach to quantifying the indirect effects of additional biofuel demand. In
the reviewed studies the focus was on the greenhouse gas emissions caused by
indirect land use change.

First a certain additional biofuel demand to be analyzed is chosen. Usually this is
based upon a certain biofuel mandate, but it can also be based on general market
expectations on competition with other fuels. In the studies reviewed in this report
pathway-specific indirect effects were investigated. This means that the additional
biofuel demand is supplied in whole by one biofuel pathway to calculate the indirect
effects of that pathway specifically. A pathway always consists of a certain fuel and
feedstock and sometimes of a specific region, e.g. production of ethanol from
sugarcane in Brazil. Then the market response to this demand is calculated by
modelling expected interdependent changes in commodity prices, crop yields and
cropland areas due to the additional biofuel demand. In other words, the additional
biofuel demand will come from one of three sources: 1) an additional yield increase
that results from the additional biofuel demand, 2) an increase in agricultural land,
and 3) a reduction in consumption in other sectors (step 1 in the above figure).

The area expansion, after correcting for additional yield increases and co-products, is
the component causing a LUC. To further quantify the effects of this LUC, information
is gathered on types of LUC (step 2 in the above figure) and the corresponding
changes in carbon stocks and biodiversity (step 3 in the above figure). Finally, a time
horizon, to which these indirect impacts are allocated and, in certain studies, a
discount rate are chosen in order to enable comparison on, for example, a per year or
per MJ fuel basis (step 4 in the above figure).

The approach visualized in Figure 2 - 3 can be used to quantify different indirect
effects of biofuels. This is further explained in Table 2 - 1. As the focus of the reviewed
studies was on the greenhouse gas emissions caused by ILUC, these studies do not
contain specific analysis of food/feed consumption and biodiversity impacts.

19 May 2011 8


Table 2 - 1 Details of the methodology to quantify the indirect effects of biofuels. As the focus of
the reviewed studies was on the GHG emissions caused by ILUC, these studies do not
contain specific analysis of food/feed consumption and biodiversity impacts.

Step Description Reviewed
impacts
1. Market Global agroeconomic equilibrium models or causal-descriptive Information on
response approaches are used to assess the effect of additional biofuel global
demand, for example by introducing a biofuel mandate, on the food/feed
market. Effects to accommodate the biofuel demand are usually prices and
separated in three categories: food/feed

• Expansion of agricultural land. consumption

• Intensification of agricultural production; e.g. higher
yield per harvest, increased number of harvests per
year

• Higher commodity prices, crowding out consumers of
the same commodity in other markets, leading to
reduced consumption, e.g. for food/feed.

2. LUC From step 1, it is known what amount of expansion of
agricultural land can be expected. Also, the location is usually
available on a country/region level. In this step a prediction is
made on which types of land will be converted to agricultural
land. One method used for this purpose is satellite analysis of
historical LUC trends.
3a. Once the amount and the type of LUC is known from steps 1. Information on
Biodiversity and 2., the biodiversity and carbon stocks impacts can be biodiversity
3b. Carbon assessed, making use of information sources on the carbon loss
stocks stocks and biodiversity values present in the LU-types that are
converted – e.g. IPCC data-sources on carbon stocks. For the
carbon impact, an additional step is needed, see step 4.
4. Time Although the carbon emissions quantified in step 3b largely take Information on
allocation place upon conversion, they are usually allocated to the GHG life cycle GHG
balance of biofuels over time. Different allocating mechanisms balance
have been suggested. Different discount rates have also been
suggested to allow comparison of emissions occurring over
different time periods.

2.2.2 Methodologies using global agro-economic equilibrium models
In many cases where indirect effects of additional biofuel production are assessed,
global agro-economic equilibrium models are used. These models predict the market
response to additional biofuel demand by calculating equilibrium states for the global
market using a complex set of mathematical equations that relate to e.g. international
trade, agricultural economics and policy and energy markets. Underlying these

19 May 2011 9


equations is a vast set of assumptions e.g. on trade elasticities, agricultural yield
developments and demands from other sectors such as the food sector.

By comparing scenarios with and without the additional biofuel demand, they deduce
changes in the system that can be attributed to the additional biofuel demand. An
advantage of this approach using equilibrium models is that it is very comprehensive.
A disadvantage is that the complex nature of the models can make their assumptions
and results rather non-transparent.

It is also important to remark that the land-use changes identified by the models
implicitly contain both direct and indirect land-use changes. These two effects can not
be separated in the models. This is caused by the way the models operate: land-use
changes are measured by comparing the land use in the additional biofuel scenario to
the land use in the reference scenario. The model does not explicitly specify whether
the land that underwent a land-use change is now in use for biofuel feedstock
production or for other purposes such as food production. Therefore, the nature of the
land-use change could be either direct or indirect. However, it is certain that all land-
use changes were directly or indirectly triggered by the additional biofuel demand, as
that is the only difference between scenarios. Following from this observation is that
when ILUC is mentioned in a model study or a study reviewing model studies, actually
a total of direct and indirect land-use change is meant. We will refer to this as (I)LUC.

Studies that use equilibrium models can use either general or partial equilibrium
models. General equilibrium models calculate an equilibrium state for a system
including all (relevant) economic markets. Partial equilibrium models calculate an
equilibrium state for one specific sector, e.g. the energy sector or agricultural sector.
The state of all other sectors is assumed constant. General equilibrium models are
therefore more comprehensive, but can in turn also include more uncertainties in
assumptions.

2.2.3 Methodologies using a causal-descriptive approach
In an effort to reduce the intransparency caused by some agro-economic equilibrium
models, other methodologies use a causal-descriptive approach. This is a bottom-up
approach where a causal chain of events following the additional production of biofuels
is constructed. Assumptions that lead to the steps of this chain of events are backed
by historic data, projections for the future or expert opinions. This makes these
approaches more transparent and more easily discussed. A drawback is that these
approaches require some amount of simplification as they lack the comprehensive
scope and the computational power of the agro-economic equilibrium models.

Causal-descriptive approaches can have a retrospective or a predictive nature. For
example, in the ICONE study introduced in Section 2.4 the authors looked back at a
historical period and tried to assign the observed land-use changes to biofuels and
other drivers of agricultural development. In the E4tech study and in the ILUC Factor

19 May 2011 10


approach by the Öko-Institut and IFEU, both are also introduced in Section 2.4, the
authors try to analyse possible future land-use change scenarios using a causal chain
of events. An advantage of the retrospective approach is that in principle one only
needs historical statistics, without the need for making future predictions. At the same
time, this could also be a weakness, as it makes the implicit assumption that historical
trends reflect future trends, which is not necessarily correct due to changes in e.g.
technology and policy.

2.3 Key assumptions in quantifying indirect effects

From the methodology of quantification presented in Section 2.1, it becomes clear that
each methodology is a complex computational framework with its own regional
divisions and assumptions. These differences in methodology setup and assumptions
lead to differences in outcomes ranging from minor to major. The main assumptions
associated with each of these steps are thus determining factors in the eventual
outcomes. These are discussed here along the lines of the four methodological steps
described in Table 2 - 1:

1 Market response to additional biofuel demand
This step contains the following main assumptions to quantify the resulting cropland
expansion, intensification of production and reduction of demand in other sectors.
• The choice of feedstock for the additional biofuel demand; e.g. including biofuel
pathways with a high biofuel yield per hectare or biofuels from residues and
wastes leads to lower land-use change.
o Choosing biofuel pathways with a high biofuel yield per hectare means that
in principle less area is needed to accommodate the additional biofuel
demand and thus that indirect impacts are lower. As biofuel yields per
hectare from high yielding pathways can be multiple times higher than
those of low yielding pathways this can have a large effect. However, co-
products should be taken into account as well; this is discussed in the next
assumption. This report focuses on pathway-specific indirect effects, so
every pathway’s effect is quantified separately.
o Choosing biofuel pathways from residues and wastes that put no strain on
the economic system, so no indirect impacts occur. This absence of indirect
impacts of residues and wastes is a model assumption. In practice, some
feedstocks classified as residue or waste in the EU Renewable Energy
Directive (RED) can have indirect effects.2 An example would be tallow from
animal rendering, which can currently be used as process fuel by the
renderers as well as for oleo-chemical applications. If that tallow were now
to be used for biofuel production, the renderers would need a replacement
process fuel, for example heavy fuel oil. In this report, no biofuel pathways
from residues and wastes are included.

2
Recently a report was issued by the UK Renewable Fuel Agency on this matter: Ecometrica, Methodology
and Evidence Base on the Indirect Greenhouse Gas Effects of Using Wastes, Residues, and By-products for
Biofuels and Bioenergy, November 2009.

19 May 2011 11


• Treatment of co-products of biofuel production.
o Most biofuel feedstock crops do not only have biofuel as an end-product,
but also produce one or more co-products. For example, corn used for
ethanol production also yields residual dry distiller’s grains and solubles
(DDGS) commonly used as animal feed. In soy biodiesel production the
biofuel co-product, soy meal commonly used as animal feed, is even the
main product in terms of volumes. These co-products can be accounted for
by assuming they displace a certain amount of other commodities on the
markets, usually animal feed. Assumptions on what product is displaced
and where and how this displaced product would have been produced, have
a significant effect on the outcome of the methodology. For example, the
DDGS from corn ethanol production is often assumed to replace one third of
the original corn demand for biofuels.
• Relation between agricultural intensification and commodity prices
and/or demand.
o This relation determines the amount of additional biofuel demand that is
met through additional agricultural intensification. Any amount met this
way does not need to be met by cropland expansion and thus no (I)LUC
occurs for that amount. Note that intensification may also lead to changes
in GHG emissions. The studies reviewed in this report do not take this into
account.
o This relation is very difficult to quantify: historical data on yield and prices
are distorted by many other factors, making it extremely challenging to
identify the actual causal relationship between demand or price increases
on the one hand and yield increases on the other hand. Studies on this
matter are often inconclusive.
• Relation between food/feed demand and commodity prices.
o This relation determines the amount of additional biofuel demand that is
met through reductions in food/feed demand. Any amount met this way
does not need to be met by cropland expansion and thus no (I)LUC occurs
for that amount. In addition, it determines the indirect impacts on food
/feed consumption.
• Relative yield of new land taken into production
o Some studies assume3 that the land best suited for agriculture is already in
production. Any land that is used for cropland expansion therefore has a
lower average yield than current cropland, which increases the amount of
land expansion needed.

3
Exact assumptions on this issue and their effect in the methodology are often unclear.

19 May 2011 12


2 LUC caused by cropland expansion
• Assumptions of types of LUC caused by cropland expansion.
o Since carbon stocks and biodiversity values between different land types,
e.g. forest, grassland, savannah, can differ significantly, the type of land-
use change assumed to occur because of cropland expansion is a key
parameter in quantifying indirect impacts. Conversion of specific high-
carbon subcategories of these land types such as wet- and peatlands are
not included in all studies. This can cause uncertainties in the carbon stock
values found in step 3.
Many studies rely on historical satellite data for their assumptions on the
type of land that is converted by expanding cropland. It should be noted
that the reliability of such satellite data, and especially their suitability to
observe land-use change is questioned by some experts.
o Secondary indirect effects can occur in the cattle sector, which is not
included in all methodologies. For example, when a biofuel is produced from
sugarcane previously used for the food sector, sugarcane might expand
onto pasture used for cattle grazing. This cattle pasture might in turn be
replaced by pasture expansion into a forested area. This secondary effect is
only quantified when the methodology accounts for these changes in the
cattle sector.
3 Current carbon stocks and biodiversity values of land used for cropland expansion
• Assumptions on carbon stocks and biodiversity values of land types affected
by cropland expansion.
o Even when the types of LUC are known from the second step, the carbon
stock and biodiversity value of the land affected by LUC is still an important
assumption, as different values are used within the different model studies.
Some studies also include foregone carbon sequestration of land taken into
crop production that would otherwise likely have increased in carbon stock,
e.g. young forests or fallow cropland.
4 Time allocation of GHG emissions of LUC
• Assumptions on time allocation of GHG emission effects.
o Most GHG emissions of LUC occur soon after the conversion of land takes
place, for example by burning the original vegetation. However, in the final
result of the GHG life cycle analysis of the biofuel, these emissions from
LUC need to be allocated over time to the biofuel produced on the land. The
amount of years chosen for that varies between the studies and often no
clear rationale provided. This amount of years chosen commonly varies
between 20 to 100 years, and has a significant effect on the indirect GHG
impacts of biofuel. In addition, some studies incorporate discussion on
discount rates that can be used to attach gradually less value to GHG
emissions and/or savings as they occur later in time. This is comparable to
the use of discount rates for the value of money in time in economic theory.

19 May 2011 13


Quantification of these assumptions and their effect on model outcomes, where
possible, forms the core of the quantitative comparison made in Section 2.5.

2.4 Overview of reviewed quantitative work on indirect effects

In this report, only studies are included that have performed pathway-specific
quantification of indirect effects. The reviewed studies, their character and the way
their results are reviewed are presented in the list below and Table 2 - 2:
• CARB, California’s Low Carbon Fuel Standard (LCFS) Final Statement of Reasons,
December 2009
• E4tech, A causal descriptive approach to modelling the GHG emissions associated
with the indirect land use impacts of biofuels, Final Report, November 2010.
• EPA, Renewable Fuel Standard Program (RFS2) Final Rule, March 2010
• ICONE, An Allocation Methodology to Assess GHG Emissions Associated with Land
Use Change, Final Report, September 2010
• IFPRI, Global trade and environmental impact study of the EU biofuels mandate,
March 2010
• Öko-Institut and IFEU, the ILUC Factor contained in Sustainable Bioenergy:
Current Status and Outlook, March 2009
• Searchinger et al., Use of U.S. Croplands for Biofuels Increases Greenhouse Gases
Through Emissions from Land Use Change, Sciencexpress, February 2008
Table 2 - 2 Character of the reviewed studies and the way they were used in this review.

Study Character Use in this review
CARB Equilibrium model study with results Quantitative review on corn (Section
for ethanol from sugarcane and corn 2.5.1), sugarcane (2.5.2) and soy
and biodiesel from soy use in the US. (2.5.3).
E4tech Causal-descriptive study with results Quantitative review on sugarcane
for use ethanol and biodiesel in the EU (Section 2.5.2), soy (2.5.3), palm oil,
from a variety of pathways. See Table wheat and rapeseed oil (all 2.5.4).
2 - 3 for the specific scenario per crop
included in this study.
EPA Equilibrium model study with results Quantitative review on corn (Section
for ethanol from sugarcane and corn 2.5.1), sugarcane (2.5.2) and soy
and biodiesel from soy use in the US. (2.5.3).
ICONE Causal-descriptive study with results Quantitative review on sugarcane
from ethanol from sugarcane in Brazil. (2.5.2).
IFPRI Equilibrium model study with results Quantitative review on corn (Section
for use of ethanol and biodiesel in the 2.5.1), sugarcane (2.5.2) and soy
EU from a variety of pathways. (2.5.3) and other pathways (2.5.4).
Öko- Causal-descriptive study with a Quantitative review on sugarcane
Institut/IFEU generalised methodology that can be (2.5.2) and soy (2.5.3).
used for a variety of pathways.
Searchinger Equilibrium model study with results Quantitative review on corn (Section
for ethanol from corn use in the US. 2.5.1).

19 May 2011 14


In the E4tech study different scenarios are presented for each pathway. They can
have significantly different results on GHG emissions from indirect land use change.
These scenarios differ on the assumptions used as input, for example on co-product
effect, land use change types and carbon stocks. In the E4tech study none of these
scenarios is presented as most likely or most accurate. However, we had to choose
one of the scenarios for each pathway to include in this review.

We did this by including the scenario that had the most common denominators in its
assumptions. For example, if there are five scenarios for a certain pathway, and four
of them assume historical deforestation rates to continue, we chose to include one of
those four. By applying this logic to all the different assumptions of the scenarios, we
were able to select a scenario with the most common denominators in its assumptions
for each pathway. These scenarios, including a short description of their most
important assumptions are listed in Table 2 - 3.

Table 2 - 3 Overview of the scenario included in this review per pathway from the E4tech study.
Please refer to the E4tech study for more details.

Pathway Scenario Most important assumptions
Biodiesel from palm oil 4 No demand induced yield increase; Historical
deforestation rates; 33% of expansion on peat
land; single plantation lifetime.
Biodiesel from rapeseed oil 1 High share produced in EU; inclusion of demand
induced yield increase; no displacement of
Ukrainian food rapeseed; historical deforestation
rates; 100% use of rapeseed meal as animal
feed; high substitution values for co-products.
Biodiesel from soybean oil 1 China: 50%/50% substitution of soy oil by palm
oil and rapeseed oil; high emission factor scenario
used for palm oil.
Ethanol from wheat 4 Increased production all in EU; high demand
induced yield increase; historical deforestation
rates; 100% use of DDGS as animal feed.
Ethanol from sugarcane 7 Low demand projections; some production in USA
and Africa; inclusion of demand induced yield
increase; historical pasture displacement in Brazil
attributed to sugarcane expansion; historical
deforestation rates.

2.4.1 Quantifying biodiversity losses

All studies reviewed in this report focus on the GHG emissions from (I)LUC resulting
from an additional demand for biofuels. None of the studies includes a detailed
analysis of the impacts of (I)LUC on biodiversity. However, as explained in Section

19 May 2011 15


2.2.1 indications on the impacts on biodiversity are available from the existing studies.
After all, all of the reviewed studies, except the ILUC-factor from the Öko -Institute,
provide the following information:
The amount of (I)LUC that takes place, usually on a country/region level (step 2 in
the 4-step process described in Section 2.2.1). This information is summarised for
each crop in each study in the tables of Section 2.5 in ha/toe.
The type of land converted as a result of (I)LUC (step 3 in the 4-step process
described in Section 2.2.1. Combined with the information on the country/region
where the (I)LUC takes place, this provides information on the type of biome or
ecosystem that is converted.
Therefore, even though the reviewed studies did not explicitly analyse the impacts on
biodiversity, they contain valuable information on the potential biodiversity impacts of
additional biofuel demand.
Another indication of the impacts on biodiversity of additional biofuel demand is given
by the recent “Rethinking Global Biodiversity Strategies” (PBL 2010). While this study
is not specifically focussed on the impacts of biofuels or bioenergy it does find in its
results that increasing the amount of bioenergy, without additional measures to
minimise impacts on biodiversity, has a negative impact on global biodiversity by 2050
even if it reduces global warming. Interestingly the study also finds that if additional
measures are taken alongside bioenergy, such as additional yield increases, the total
impacts on biodiversity are positive.
To increase the understanding of the impacts of biofuels on biodiversity, it is
recommended that future biofuel/bioenergy (I)LUC studies attempt to quantify these
biodiversity impacts next to the impacts on GHG emissions.

2.5 Review of pathway-specific indirect effects – understanding the
differences

This section provides the quantitative review of indirect effects for different pathways:
ethanol from corn (Section 2.5.1), ethanol from sugarcane (Section 2.5.2), biodiesel
from soy (Section 2.5.3) and biofuels from other pathways (Section 2.5.4).

2.5.1 Ethanol from corn
The review is split in three stages: first, we provide a condensed but comprehensive
overview of quantitative results from all studies in our review in one table. Then we
explain the details of each study. Finally, we explain the differences between studies,
using the details described earlier.

Summary of important results and assumptions of the reviewed studies
The table on the next pages summarizes the key results and assumptions of the
various studies.

19 May 2011 16


Table 2 - 4 Pathway-specific results from different studies on GHG emissions from (I)LUC for ethanol from corn. Please refer to the text for detailed
discussion of the reviewed studies and their differences.

ETHANOL FROM CORN Unit CARB (LCFS) EPA (RFS2) IFPRI Searchinger
Treatment of co-products
Included, but 33% displacement of
25-45% displacement of original
5
- Included in the model. replacement rates original feedstock by co-
feedstock by co-products.4
are unclear. products.
Additional demand from biofuels for energy crops results in
Biofuel-induced agricultural Included for both crops Included, but extent
% Included, with a significant but 0% (assumption)
intensification and pastures.7 unclear.
unquantifiable effect. (Rough
Included. (Estimate for Included, but extent
Reduced demand in other sectors % estimate: ~50%)6 20% (estimate9)
total mandate: ~10%)8 unclear.
Included, but extent
Included, but extent unclear. Included, but extent
Cropland expansion % unclear. (Rough 80%
(Rough estimate: ~50%)6 unclear.
10
estimate: ~90%)
Extent of cropland expansion
Cropland expansion ha/toe 0.15 0.26 Unspecified. 0.38

4
The co-product is assumed to replace corn on a ~1:1 weight basis. As generally 25%-45% of the corn ends up in the co-product, this range is used to estimate the
co-product displacement effect.
5
The co-product is assumed to replace a combination of corn and soy meal on a 1:1 to 1.2:1 weight basis. The effect of this replacement is endogeneously calculated
in the model.
6
From communication between Ecofys and the authors we know this is included. They estimated that these effects would be significant, which is supported by the
sensitivity of the final outcome of the study to changes in assumptions in these areas. Based on data reported in the study (corn yield of 151.3 bushel/acre; ethanol
yield of 2.8 gallon/bushel) and assuming a 1/3 co-product displacement, we calculated a very rough estimate of ~50% for these two effects combined.
7
Both the use of pastures and that of cropland can be intensified in the model in response to demand/price changes. The magnitude of this effect is not clear from the
data, but could very well be significant. As an example: international pasture reduction equals 33% of cropland expansion in the corn scenario. As this effect could also
partially be caused by reduced demand for cattle products, it can not only be attributed to pasture intensification.
8
The study provides data for global reduced food consumption for the total RFS2 mandate. Based on this data we made a rough estimate on the size of this reduction
relative to the additional production of land-using biofuels.
9
This value was an estimate as reported by Searchinger during personal communication with Ecofys in 2009 and is therefore indicative.
10
Based on data reported in the study (corn yield of 183 bushel/acre; ethanol yield of 2.6 gallon/bushel) and assuming a 1/3 co-product displacement, we calculated a
very rough estimate of ~10% for biofuel-induced agricultural intensification and reduced demand in other sectors combined. This means ~90% of cropland expansion.

19 May 2011 17

A SUSTAIN AB LE ENE RGY SUPPLY FOR EVERYONE

ETHANOL FROM CORN Unit CARB (LCFS) EPA (RFS2) IFPRI Searchinger

GHG effect of cropland expansion

Weighted average of emissions from
tCO2eq/ha 244 126 Unspecified. 351
crop expansion
Project horizon for emissions from
Years 30 30 20 30
(I)LUC
gCO2eq/MJ
Emissions from (I)LUC 30 27 54 103
fuel
Emissions from (I)LUC (assuming a
gCO2eq/MJ
standardized project horizon of 20 45 41 54 155
fuel
years; for comparison only)

19 May 2011 18


Detailed discussion of reviewed studies
CARB
The CARB study includes a co-product replacement for corn, where the resulting co-
product is assumed to replace corn used as animal feed.

The modelling done in the study also accounts for the effects of biofuel-induced
agricultural intensification and reduced demand in other sectors. We expect that these
effects are significant from communication with the authors, from their sensitivity
analysis and from the relatively low value of 0.15 ha/toe cropland expansion. It is
unclear how large these effects are. A very rough estimate based on the available data
indicates that they could combine to compensate ~50% of additional biofuel demand.

The model includes a land-use module that automatically identifies the type of land
being converted to cropland as either “forest land” or “pasture land”. For corn it is
found that ~20% of cropland expansion occurs on forest land and ~80% on pasture
land. It is not clear from the data whether this includes any cattle knock-on effect
where displaced pastures are reclaimed in other areas. This leads to a moderate value
for associated emissions of 244 tCO2eq/ha. The emissions are allocated to a 30-year
time period.

EPA
The EPA study includes a co-product replacement for corn, but this is not a pure
replacement of corn itself. Instead, the co-product replaces a mix of corn and soybean
meal. As this happens endogeneously in the model the exact effect cannot be derived
from the available reports.

It is also clear that biofuel-induced agricultural intensification and reduced demand in
other sectors are included in the model. Both cannot be quantified exactly based on
the available reports, but are expected to play a significant role. For example, the
study provides data for global reduced food consumption for the total RFS2 mandate.
Based on this data we made a rough estimate on the size of this reduction relative to
the additional production of land-using biofuels being ~10%. Biofuel-induced
agricultural intensification occurs both on cropland and on pastures in the EPA model.
The magnitude of this effect is not clear from the data, but could very well be
significant. As an example: international pasture reduction equals 33% of cropland
expansion in the corn scenario. As this effect could also partially be caused by reduced
demand for cattle products, it cannot only be attributed to pasture intensification.

A very rough estimate based on the available data indicates that in total biofuel-
induced agricultural intensification and reduced demand in other sectors in the corn
scenario could combine to compensate ~10% of additional biofuel demand.

The resulting cropland expansion value of 0.26 ha/toe is in the middle of the range
found in other studies and consists of separately quantified US (“domestic”) and non-

19 May 2011 19


US (“international”) cropland expansion. This is due to the fact that EPA used a
combination of a model specific for the US and an international model for their
calculations. For corn 42% of the expansion is domestic, 58% is international.

The GHG emission effect of cropland expansion found in the EPA study is very low
compared to the other studies. The US model endogeneously calculates the GHG
emission effect of domestic cropland expansion. For corn, negative emissions of -4
gCO2eq/MJ are found, meaning that the land-use changes lead to carbon
sequestration. The causes for this effect are unclear because they occur within the
model and are not explained in detail by the authors.

The international model cannot endogeneously calculate the GHG emission effect of
international cropland expansion. Therefore, satellite data is used to predict which
land types are converted and Winrock data for carbon stocks of these land types are
used to calculate the associated emissions. For corn, relatively large conversion of
high carbon stock biomes in the Brazilian Amazon are found, leading to significant
emissions.

On average, combining domestic and international land-use changes, a relatively low
value of 126 tCO2eq/ha is found. The time horizon for the emissions is set at 30 years.

IFPRI
No detailed intermediate results are available for the pathway-specific calculations
performed by IFPRI. Therefore, no breakdown or discussion of their 54 gCO2eq/MJ
value can be provided. The exception is the projection on time horizon for the
emissions at 20 years, where the other studies use 30 years. If a 30-year horizon
were to be used, the value would be 36 gCO2eq/MJ.

Communication with the author indicated that IFPRI may perform an updated analysis
on pathway specific ILUC values for EU biofuel demand. We hope this updated version
will provide more insights in the assumptions and intermediate results of the study,
which will allow for a better comparison with other studies.

Searchinger
Searchinger finds a relatively high value of 103 gCO2eq/MJ ethanol fuel. Two
important reasons for that are his assumption that there will be no biofuel-induced
agricultural intensification and the relatively high carbon stocks he assumes for
converted land. In his paper he explains these two assumptions:

1 The study assumes that potential agricultural intensification due to higher demand
and/or price is neutralised by the yield loss associated with the taking into
production of marginal lands that are less suitable for agriculture.
2 Searchinger assumes that all deforestation historically found in a certain area is
caused by cropland expansion, provided that cropland expansion has historically

19 May 2011 20


been larger than deforestation. Through this assumption the study finds large LUC
in forested areas, which are high in carbon compared to other land types.

Matching comparison and discussion of the reviewed studies – understanding
the differences
All studies find significant GHG emissions associated with (I)LUC for ethanol from
corn, with the results spanning 27 – 103 gCO2eq/MJ ethanol fuel. Within this range of
values, Searchinger finds a particularly high value. Two important reasons for that are
the study assumes that any biofuel-induced agricultural intensification is cancelled out
by lower yield on the new land taken into cultivation, and the relatively high carbon
stocks assumed for converted land.

The IFPRI, CARB and EPA values, when compared on a 30 year time horizon, are
relatively similar at 36, 30 and 27 gCO2eq/MJ respectively. No details are available for
the IFPRI analysis, so we cannot comment on the way this value was achieved.

When comparing the details of the CARB and the EPA value it is striking that, although
they reach a similar end value, the intermediate values are quite different. The EPA
cropland expansion is almost twice as high as that of CARB: 0.26 ha/toe instead of
0.15 ha/toe. The reason for this cannot be pinpointed as co-product replacement,
biofuel-induced agricultural intensification and reduced demand in other sectors all
occur endogeneously in the model. Not enough detailed data are available to analyse
their effect in the two studies.

The difference in cropland expansion values is compensated by the EPA value for
emissions per hectare of cropland expansion being about half that of CARB: 126
tCO2eq/ha instead of 244 tCO2eq/ha. The low value for EPA can be explained by the
fact that a significant part of the cropland expansion occurs in the US, which leads to a
negative emission effect, meaning that the land-use changes lead to carbon
sequestration. As this happens endogeneously in the model, the precise reasons
cannot be explained. These negative emissions compensate for the non-US land use
changes, which cause significant emissions in the forest biomes of the Amazon,
leading to the overall low value of 126 tCO2eq/ha.

2.5.2 Ethanol from sugarcane

19 May 2011 21


Table 2 - 5 Pathway-specific results from different studies on GHG emissions from (I)LUC for ethanol from sugarcane. Please refer to the text for detailed

Öko-
ETHANOL FROM
Unit CARB (LCFS) E4tech EPA (RFS2) ICONE IFPRI11 Institut/IFEU
SUGARCANE
(ILUC Factor)
Co-products only
Co-products are all Included, as
Co-product is Co-product is used have economic value,
used for process Not gross yield
- used for energy for energy no agricultural
energy, no agricultural considered. numbers are
generation.12 generation.13 feedstock
feedstock replacement. used.14
replacement.

Biofuel-induced Included for both
agricultural % Included, with an 26% crops and 29% (Total mandate:
intensification unquantifiable effect. pastures.16 33%)
92% 50%17
(Rough estimate: 0% (Excluded) Included.
Reduced demand in other ~0%)15 8% (Total mandate:
% (Estimate for total
sectors 24%)
mandate ~10%)18
Included, but
Included, but extent
extent unclear. 64% (Total mandate:
Cropland expansion % unclear. (Rough 74% 8% 50%
(Rough estimate: 44%)
estimate: ~100%)15
~40%)19

11
No detailed intermediate data was presented for the sugarcane pathway-specific calculations in the IFPRI study. However, as 69% of the increased demand for
biofuels from energy crops in IFPRI’s EU mandate scenario included in the same report was met by sugarcane, some indicative values could be derived from that
scenario. In this column, these indicative sugarcane-specific values are presented where possible. Behind them, between brackets, are the same values but then
pertaining to the total mandate including all biofuel pathways in their EU mandate scenario.
12
The co-product credit therefore materialises by replacement of fossil energy sources, outside the scope of the land-use analysis.
13
The co-product credit therefore materialises by replacement of fossil energy sources, outside the scope of the land-use analysis.
14
The ILUC Factor methodology uses gross cropland yields. This means that the yield of all products from one hectare, including co-products is used for the
calculations. The GHG effects are thus automatically allocated to both products and co-products on an energy content basis.
15
From communication between Ecofys and the authors we know this is included. They estimated that these effects would be significant for the combination of all
pathways. For sugarcane specifically, the contribution might be low: based on data reported in the study (cane yield of 75.13 ton/ha) and an assumed ethanol yield of
0.075 ton ethanol/ton cane and assuming no co-product displacement, we calculated a very rough estimate of ~0% for these two effects combined.
16
data, but could very well be significant. As an example: international pasture reduction equals 38% of cropland expansion in the sugarcane scenario. As this effect
could also partially be caused by reduced demand for cattle products, it can not only be attributed to pasture intensification.
17
This number is derived from the “medium ILUC risk level” used in the ILUC Factor methodology. In addition to biofuel-induced agricultural intensification and
reduced demand in other sectors, it also includes use of “set-aside and abandoned land” which the authors deem to have no associated ILUC.
18
19
Based on data reported in the study (ethanol yield 600 gallon/acre) and assuming no co-product displacement, we calculated a very rough estimate of ~60% for
biofuel-induced agricultural intensification and reduced demand in other sectors combined. This means ~40% of cropland expansion.

19 May 2011 22


Öko-
ETHANOL FROM
Unit CARB (LCFS) E4tech EPA (RFS2) ICONE IFPRI11 Institut/IFEU
SUGARCANE
(ILUC Factor)
Unknown. (Total
Cropland expansion ha/toe 0.29 0.22 0.14 0.03 0.13
mandate: 0.11)


Weighted average of
Unknown. (Total
emissions from crop tCO2eq/ha 202 125 34 380 270
mandate: 133)
expansion
Project horizon for 20 (Total mandate:
Years 30 30 30 30 20
emissions from (I)LUC 20)
gCO2eq/MJ 18 (Total mandate:
Emissions from (I)LUC 46 22 4 8 42
fuel 18)
Emissions from (I)LUC
(assuming a standardized
gCO2eq/MJ
project horizon of 20 69 33 6 12 18 42
fuel
years; for comparison
only)

19 May 2011 23


CARB
The modelling done in the study accounts for the effects of biofuel-induced agricultural
intensification and reduced demand in other sectors. It is unclear from the
documentation how large these effects are. As the cropland expansion found by CARB
is relatively large compared to the other studies, it can be expected that these effects
are smaller than the 37-92% found in the other studies. Indeed, a very rough
estimate based on the available data indicates that in total biofuel-induced agricultural
intensification and reduced demand in other sectors do not have a significant
compensation effect (~0%) in the sugarcane scenario.

being converted to cropland as either “forest land” or “pasture land”. For sugarcane it
is found that ~30% of cropland expansion occurs on forest land and ~70% on pasture
where displaced pastures are reclaimed in other areas. All emissions are allocated to a
30-year time period.

E4Tech
[We reviewed scenario 7 of the sugarcane scenarios of E4Tech, please refer to Section
2.4 for details.]

The E4tech study includes no co-product replacement for sugarcane that affects the
land-use calculations as the co-product (bagasse) is used to generate energy.

Biofuel induced agricultural intensification plays a significant role, resulting in a 26%
reduced need for area expansion. This intensification was based on a combination of
the Lywood (2009a) method for certain regions with an estimate for regions where
this method was deemed unfeasible. The use of the Lywood method can lead to high
values for biofuel induced agricultural intensification, as explained in Section 2.6.1.

Analysis of reduced demand in other sectors was not included in the E4Tech work, so
by default contributes 0%.

The scenario reviewed by us includes a knock-on effect for pasture displaced by
sugarcane in Brazil, leading to an expansion of pasture in other regions of Brazil.

The resulting area expansion is in the middle of the range of other studies at 0.22
ha/toe. The emissions caused by this expansion are on the lower end of the range of
the other studies at 125 tCO2eq/ha. This low number is partly caused by an error in
the carbon stock numbers for land converted to pasture in Brazil (as a result of
pasture displacement by sugarcane). The effect of this error is that the ILUC value of
Brazilian sugarcane ethanol is underestimated by 50%. Note that Brazil supplies only
55% of total additional sugarcane ethanol in the scenario reviewed by us, so the error

19 May 2011 24


in the average emissions per hectare of land converted is less than 50%. As discussed
in more detail in Section 2.6.8, the effect on the average ILUC emissions from
sugarcane ethanol is between 3 and 10 gCO2eq/MJ according to the authors.
Furthermore, the carbon stock numbers are based on Winrock data that found
medium levels of deforestation. Combined with the error mentioned above and the
fact that sugarcane is treated as a perennial crop leading to more carbon
sequestration than annual crops, the overall value for the emissions per hectare is
relatively low.

The reviewed E4Tech-scenario finds a midrange value of 22 gCO2eq/MJ GHG emissions
from (I)LUC associated with sugarcane from ethanol. When the error in the carbon
stock numbers used in Brazil is taken into account, the value is 25-32 gCO2eq/MJ.

EPA
The EPA study includes no co-product replacement for sugarcane that affects the land-
use calculations as the co-product is used to generate energy. The co-product credit is
applied in another step of the total emissions analysis.

Biofuel-induced agricultural intensification and reduced demand in other sectors are
included in the model. Neither can be quantified exactly, but we expect they play a
significant role. For example, the study provides data for global reduced food
consumption for the total RFS2 mandate. Based on this data we made a rough
estimate on the size of this reduction relative to the additional production of land-
using biofuels being ~10%.

Biofuel-induced agricultural intensification occurs both on cropland and on pastures in
the EPA model. The magnitude of this effect is not clear from the data, but seems
significant. As an example: international pasture reduction equals 38% of cropland
expansion in the sugarcane scenario. As this effect could also partially be caused by
reduced demand for cattle products, it cannot only be attributed to pasture
intensification.

A rough estimate based on the available data indicates that in total biofuel-induced
agricultural intensification and reduced demand in other sectors in the sugarcane
scenario could combine to compensate ~60% of additional biofuel demand. The
resulting cropland expansion value of 0.14 ha/toe is in the middle of the range found
in other studies.

The GHG emission effect of cropland expansion of 34 tCO2eq/ha found in the EPA
study is extremely low compared to the other studies. The model cannot
endogeneously calculate the GHG emission effect of international cropland expansion.
Therefore, satellite data is used to predict which land types are converted and
emissions factors calculated by Winrock for carbon stocks of these land types are used
to calculate the associated emissions. For sugarcane, the authors comment that these
low values are “due largely to our projection that sugarcane crops would expand onto

19 May 2011 25


grassland in South and Southeast Brazil, which results in a net sequestration because
sugarcane sequesters more biomass carbon than the grasslands it would replace”. No
significant indirect effects of this expansion of sugarcane onto grasslands, such as the
displacement of cattle which may lead to cattle expansion in the Amazon region, are
observed in the EPA modelling.

As a last variable, the time horizon for the emissions is set at 30 years.

ICONE
The ICONE study is a retrospective causal-descriptive methodology that attributes
historic deforestation of three types of forest biomes, observed in Brazil between 2005
and 2008 through satellite data, to the agricultural sector developments in the same
period. The role of sugarcane is explicitly analyzed such that deforestation emissions
can be attributed to the expansion of sugarcane for ethanol. In contrast to the
agroeconomic models it distinguishes between direct and indirect LUC.

The ICONE study finds a relatively low value of 8 gCO2eq/MJ GHG emissions from
(I)LUC associated with sugarcane from ethanol. The main reason for the low value of
GHG emissions from direct LUC is that sugarcane expansion rarely leads to direct
conversion of natural vegetation. Satellite image data is presented as evidence for
this. For the emissions for indirect LUC the situation is more complicated and is
discussed below.

Important causes of the low value for emissions from indirect LUC have to do with
methodological choices on calculating the indirect LUC emissions that the authors have
made:
1 Crops/cattle displaced by sugarcane are first accommodated within the same
region as where they are displaced by sugarcane by relocating them to land that
has become available through yield increases or area reduction of other crops.
Thereby, the crops/cattle displaced by sugarcane receive priority over crops/cattle
displaced by other activities. The report does not explain this prioritization.
This leads to the situation where indirect LUC caused by sugarcane only takes
place within the region in which the sugarcane expands. These are typically the
regions where little deforestation occurs, while most deforestation occurs in the
Amazone where sugarcane does not expand directly. This results in the finding
that sugarcane causes low emissions from ILUC.
In summary, the situation arises that a) sugarcane predominantly displaces cattle,
b) cattle predominantly expands in the Amazon region, but c) none of the
expansion of cattle in the Amazon region is linked to the displacement of cattle by
sugarcane.
2 Knock-on effects of crops/cattle displaced by sugarcane beyond the first
displacement step are not included. For example, when sugarcane displaces corn,
a part of the displaced corn may displace soy, which may be reclaimed in natural
vegetation. This effect is not considered: only the first-step expansion of the
displaced corn on natural vegetation is taken into account.

19 May 2011 26


3 Crop/cattle displacements that cannot be explained using the prioritised relocation
in the same region do not cause interregional/international ILUC. The report
suggests that this is explained by yield increases in other regions but this is not
entirely clear. In our review we therefore assumed this was compensated for
through a combination of demand-induced agricultural intensification and/or
reduced demand in other sectors.

The above methodological choices have an even larger effect due two underlying
choices the authors have made:
1 All yield increases for crops other than sugarcane are attributed to reducing the
pressure caused by expansion of sugarcane. This is an unconventional choice as
other methodologies explicitly or implicitly exclude these business-as-usual yield
increases as they are not a result of the additional biofuel demand. This can for
example be done by using a baseline scenario in an equilibrium model study. In
addition, the applied business-as-usual yield increases found in the used 2005-
2008 period are much higher than historical trends.
2 Over the chosen timeframe of the study (2005-2008), contrary to other historical
timeframes and predictions for the future, Brazilian cropland for crops other than
sugarcane decreased significantly. As this reduction in cropland is not clearly
compensated for in the methodology by e.g. a loss of exports or a comparison
with a baseline scenario, this leaves significant room for accommodating the
sugarcane expansion. In addition, the reduction in cropland of other crops means
there was relatively low pressure of cropland expansion on deforestation, leading
to low deforestation numbers.

A few examples of the results of applying this methodology to the South-East region
of Brazil, which has the majority of all the sugarcane expansion, are:
1 Sugarcane partially expanded on corn cropland. First, a ~15% correction caused
by business-as-usual yield increase is used to reduce the amount of actually
displaced cropland. Then, it is observed that only 3% of the corn expansion took
place on natural vegetation, meaning that a very small share of the sugarcane
expansion on corn cropland is found to have led to deforestation: ~2.5%. No
further knock-on effects of the displaced corn displacing other crops/cattle are
taken into account. In total, the relatively low number is a result of the
methodological choices of regional prioritization, business-as-usual yield
increases, not considering of knock-on effects and the used timeframe.
2 Sugarcane partially expanded on soy cropland. First, a ~20% correction caused by
business-as-usual yield increase is used to reduce the amount of actually
displaced cropland. Then the authors conclude that the amount of new land taken
into production for soy in the region is only ~15% of the amount of soy cropland
displaced by sugarcane in the region. Following calculations of conversion of
natural vegetation therefore only take into account this small amount of new land
taken into production for soy. This implicitly means that there is a sharp reduction
in regional soy production, which in other methodologies usually leads to reduced

19 May 2011 27


exports and/or increased imports in turn leading to interregional or international
displacement, but this is not taken into account.

The mentioned methodological choices lead to a relatively very low value for cropland
expansion due to additional biofuel demand: 8% of the additional demand is covered
by cropland expansion into natural vegetation or only 0.03 ha/toe of biofuel. This
implies that the other 92% must have come from either yield increases on cropland
and/or pastures caused by the additional biofuel demand or a reduction in
consumption in other sectors20. Although the GHG emissions per hectare of this crop
expansion are relatively high, which can be expected as only high carbon stock
forested biomes were considered, the resulting value for GHG emissions from (I)LUC is
relatively low at 8 gCO2eq/MJ fuel.

IFPRI
performed by IFPRI. However, in the case of sugarcane indicative trends can be
derived from the EU RED mandate scenario calculated in the same IFPRI study. This is
because ethanol from sugarcane provides 69% of the increased demand for biofuels
from energy crops in that scenario. Therefore, it can be assumed that trends observed
in that scenario might be indicative of those in the pathway-specific sugarcane
scenario21. However, these intermediate values, displayed where available in Table 2
- 5 should only be interpreted as an indication.

The values show that the end result of the sugarcane pathway-specific calculation and
that of the total EU mandate calculation are equal: 18 gCO2eq/MJ. In both scenarios,
biofuel-induced agricultural intensification and reduced demand in other sectors play a
significant role in reducing the need for cropland expansion, although less for
sugarcane than for the total mandate.

The amount of cropland expansion per unit of biofuel and GHG emissions per
converted hectare are unknown for the pathway-specific calculation. As the end result
is the same as in the total mandate scenario, their product should be equal to that of
the total mandate scenario. We can speculate that, due to the relatively high yield of
sugarcane, the sugarcane cropland expansion per unit of fuel will not be much higher
than that of the total mandate. This means that the GHG emissions per converted
hectare are likely to be somewhat similar to the 133 tCO2eq/ha found in the total
mandate scenario, which is a low value compared to other studies. This is potentially
caused by the absence of a knock-on effect of pastures. This means that the potential

20
All land that is not considered natural vegetation in the study is considered being either cropland or
pasture. This means that any expansion of cropland not leading to a change in natural vegetation should
lead to a reduction in pasture. This reduction in pasture should be accomodated by pasture yield increases
or reduced demand for products coming from the pasture.
21
This implies that we assume that increased sugarcane production in the EU mandate scenario has no or
limited interaction with the production of other crops.

19 May 2011 28


reclamation of pastures that are displaced by (sugarcane) cropland may not be
comprehensively included in the IFPRI study.

Öko-Institut/IFEU (ILUC Factor)
The ILUC Factor is a simple and transparent methodology for calculating GHG
emissions associated with ILUC. It derives a standard value of 270 tCO2eq/ha for
average emissions of one hectare of crop expansion, based on historic trade and land-
use change patterns and literature carbon stock values. Then it makes an assumption
on the “ILUC risk level”: how likely it is that one hectare of displaced crops will lead to
land-use change, as other effects like demand-induced yield increases, reduced
demand in other sectors and use of “set-aside and abandoned land” will not cause
ILUC according to the methodology.

There is no set value for this risk level in the methodology, it is only reported that it
will be between 25-75%, with 50% being the “medium level”. We have chosen to use
this medium level values throughout this report. Furthermore, the ILUC values change
slightly over time as yield and trade data change. We report the 2010 ILUC values.
Finally, we report the ILUC values for the case where the crop is grown on land
previously used as cropland. The latter choice ensures that the reported values are
independent of the direct LUC effect that takes place in growing the biofuel crop,
making it suitable for comparison with the results of the other studies.

After all these values have been determined, the only difference between pathways is
the yield per hectare of the crop used as feedstock. As sugarcane is a relatively high
yielding crop it has, for the ILUC Factor, relatively low GHG emissions from land-use
change. It should be noted that the actual pathway-specific value is very dependent
upon the choice of the ILUC risk level, for which little argumentation is provided.

the differences
There is a significant range in GHG emissions associated with (I)LUC for ethanol from
sugarcane, with the results spanning a 4 – 46 gCO2eq/MJ ethanol fuel.

The EPA and the ICONE study find the lowest values: 4 and 8 gCO2eq/MJ ethanol fuel
respectively. On an abstract level this has the same cause. Both studies find that
sugarcane expands mainly on low carbon stock grassland in Brazil and find that this
expansion on pasture leads to very little knock-on effects in high carbon stock areas
such as primary forest. Due to a difference in methodology, this shows in different
ways in Table 2 - 5. For ICONE, conversion of natural vegetation in forest biomes to
cropland is considered and is found to be very low in size (0.03 ha/toe) but high in
emissions per hectare (380 tCO2eq/ha). For EPA, a much higher expansion number
(0.14 ha/toe) is found, but this is mainly expansion on grasslands. Therefore, the
average emissions per hectare number is low (34 tCO2eq/ha). Both are a
representation of a similar effect.

19 May 2011 29


The IFPRI study, especially when it is considered that they use a 20-year time horizon,
also has a low value of (I)LUC emissions of 18 gCO2eq/MJ fuel. The lack of detailed
data makes it difficult to analyse the causes. Estimations from the total mandate
scenario in that study, where sugarcane plays a large role, indicate that a relatively
low number for average emissions per hectare cropland expansion (estimate from
total mandate: 133 tCO2eq/ha) could play a role. This could be caused by the absence
of a cattle displacement knock-on effect. A recent report by JRC (2010) indicates that
IFPRI might have significantly underestimated the loss of below- and aboveground
carbon.

The E4tech study is in the middle of the range of the found values with emissions of
22 gCO2eq/MJ fuel. This follows from the fact that most assumptions chosen as input
for the causal-descriptive model are in the middle of the range of those of other
studies. Note that the E4tech study accidentally used the wrong carbon stock numbers
in Brazil. Correcting for this error would lead to a higher ILUC value, as discussed in
section 2.6.8.

The ILUC Factor and CARB studies are at the higher end of the range. When both are
compared on a 30-year time horizon basis, they have emissions of 28 and 46
gCO2eq/MJ fuel respectively. The ILUC Factor result is very transparent, but heavily
depends on the assumption of the ILUC risk level. We chose to use the medium level
here. The CARB result is mainly high because of a relatively high number for cropland
expansion at 0.29 ha/toe. This likely has to do with a limited effect of biofuel-induced
agricultural intensification and reduced demand in other sectors, compared to EPA and
ICONE, but this could not be derived from the data. Combined with the middle of the
range value for average emissions per hectare cropland expansion (202 tCO2eq/ha)
this leads to the highest emission value per unit of biofuel.

2.5.3 Biodiesel from soy

The table on the next pages summarizes the key results and assumptions of the
various studies.

19 May 2011 30


Table 2 - 6 Pathway-specific results from different studies on GHG emissions from (I)LUC for biodiesel from soy. Please refer to the text for detailed

Öko-Institut/IFEU
BIODIESEL FROM SOY Unit CARB (LCFS) E4tech EPA (RFS2) IFPRI
(ILUC Factor)
Included, with Included, but Included, as gross
- Implicitly included.23 a 55% Implicitly included.23 replacement rates yield numbers are
effect.22 are unclear. used.24
Included for both
Biofuel-induced agricultural Included, but extent
% 7% crops and
intensification Included, with a unclear.
pastures.26
significant but 50%27
Included. (Estimate
unquantifiable effect.25 Included, but extent
Reduced demand in other sectors % 0% for total mandate:
unclear.
~10%)28

Included, but extent Included, but extent Included, but extent
Cropland expansion % 94% 50%
unclear. unclear. unclear.

22
This means: 55% of the emissions otherwise attributed to the pathway are compensated by the emission savings caused by the co-product.
23
In the EPA and CARB models, the vegetable oil market and the oil meal market are separate. The model extracts soybean oil from the vegetable oil market for
production of biodiesel. This can lead to extra soy production, including additional soy meal, but could also be compensated for by e.g. an increase in production of
palm oil.
24
The ILUC Factor methodology uses gross cropland yields. This means that the yield of all products from one hectare, including co-products is used for the
calculations. The GHG effects are thus automatically allocated to both products and co-products on an energy content basis.
25
From communication between Ecofys and the authors we know this is included. They estimated that these effects would be significant, which is supported by the
sensitivity of the final outcome of the study to changes in assumptions in these areas. Unfortunately no conclusive quantification of these effects could be made from
the available data.
26
data, but could very well be significant. As an example: international pasture reduction equals 19% of soy cropland expansion. As this effect could also partially be
caused by reduced demand for cattle products, it can not only be attributed to pasture intensification.
27
This number is derived from the “medium ILUC risk level” used in the ILUC Factor methodology. In addition to biofuel-induced agricultural intensification and
reduced demand in other sectors, it also includes use of “set-aside and abandoned land” which the authors deem to have no associated ILUC.
28

19 May 2011 31


Öko-Institut/IFEU
BIODIESEL FROM SOY Unit CARB (LCFS) E4tech EPA (RFS2) IFPRI
(ILUC Factor)
Cropland expansion ha/toe 0.31 0.16 0.94 Unspecified. 0.21


Weighted average of emissions from
tCO2eq/ha 253 46529 43 Unspecified. 270
crop expansion
Project horizon for emissions from
Years 30 30 30 20 20
(I)LUC
gCO2eq/MJ
Emissions from (I)LUC 62 66 32 75 67
fuel
Emissions from (I)LUC (assuming a
gCO2eq/MJ
standardized project horizon of 20 93 99 48 75 67
fuel
years; for comparison only)

29
This value is for the actual expansion of palm oil and rapeseed oil needed to substitute the soy oil used for biofuels. It excludes any influence of the co-products.
When these are taken into account, the value is 505 tCO2eq/ha.

19 May 2011 32


CARB
The oil meal co-product is included in the modelling as a replacement for livestock
feed. However, it is not clear what effect this has on the amount of cropland
expansion needed for biofuel production.

The modelling done in the study also accounts for the effects of biofuel-induced
agricultural intensification and reduced demand in other sectors. It is unclear how
large these effects are.

being converted to cropland as either “forest land” or “pasture land”. For soy it is
found that ~30% of cropland expansion occurs on forest land and ~70% on pasture
where displaced pastures are reclaimed in other areas. All emissions are allocated to
a 30-year time period.

E4tech
[We reviewed scenario 1 of the soy oil scenarios of E4Tech, please refer to Section 2.4
for details.]

The E4tech study assumes that all soy oil needed for biofuel production will come from
soy oil previously used for food applications. As E4tech did not include the potential for
reduced demand in other sectors in their study, this soy oil is completely substituted
by other vegetable oils. Their analysis finds this to be 75% of palm oil and 25% of
rapeseed oil. For the effects of these substitution they thus use a weighted average of
the effects found in their palm oil and rapeseed oil analysis (refer to Section 2.5.4 for
details on these analyses). Therefore, the soy oil (sub)results are heavily influenced
by those of palm oil.

This means that:
• The co-product effect is large. It compensates for 55% of the land use emissions
caused by increased palm and rapeseed oil production. This is because both palm
oil and rapeseed oil receive significant credit for their co-products that displace
other vegetable oils and animal feed as discussed in more detail in sections 2.6.3
and 2.6.4.
• The additional yield effect is modest at 7%, as E4tech assumes no biofuel
demand-induced yield increase for palm oil.
• The required cropland expansion is low compared to the other studies at 0.16
ha/toe as palm oil has a high yield.
• The emissions from this crop expansion are high compared to the other studies at
505 tCO2eq/ha as E4tech assumes that the palm expansion takes place in
Indonesia and Malaysia including expansion on forest land.

19 May 2011 33


• The end result of 66 gCO2eq/MJ biofuel is comparable to that found by most other
studies.

EPA
The EPA study includes a co-product markets for soy, but in an implicit way. In the
model, the vegetable oil market and the oil meal market are separate. The model
extracts soybean oil from the vegetable oil market for production of biodiesel. This can
lead to extra soy production, including additional soy meal, but could also be
compensated for by e.g. an increase in production of palm oil.

It is also clear that biofuel-induced agricultural intensification and reduced demand in
other sectors are included in the model. Both cannot be quantified exactly, but could
very well play a significant role. For example, the study provides data for global
reduced food consumption for the total RFS2 mandate. Based on this data we made a
rough estimate on the size of this reduction relative to the additional production of
land-using biofuels being ~10%. Biofuel-induced agricultural intensification occurs
both on cropland and on pastures in the EPA model. The magnitude of this effect is
not clear from the data, but could very well be significant. As an example:
international pasture reduction equals 19% of cropland expansion in the soy scenario.
As this effect could also partially be caused by reduced demand for cattle products, it
cannot only be attributed to pasture intensification.

The resulting cropland expansion value of 0.94 ha/toe is very high compared to other
studies. Due to the lack of detailed data on the effects that reduce the need for
cropland expansion, this cannot be explained. The total cropland expansion value
consists of separately quantified US (“domestic”) and non-US (“international”)
cropland expansion. This is due to the fact that EPA used a combination of a model
specific for the US and an international model for their calculations. For soy 53% of
the expansion is domestic, 47% is international.

The GHG emission effect of cropland expansion found in the EPA study is extremely
low compared to the other studies. The US model endogeneously calculates the GHG
emission effect of domestic cropland expansion. For soy, negative emissions of -8
gCO2eq/MJ are found, meaning that the land-use changes lead to carbon
sequestration. The causes for this effect are unclear because they occur within the
model and are not explained in detail by the authors.

The international model cannot endogeneously calculate the GHG emission effect of
international cropland expansion. Therefore, satellite data is used to predict which
land types are converted and Winrock coefficients for carbon stocks of these land
types are used to calculate the associated emissions. Analogously, the EPA
methodology takes into account the 30-year carbon sequestration provided by
cropland and pasture that is abandoned from agricultural use (e.g. as a result of
biofuel-induced pasture intensification). The 2007 average land cover, as determined
by the same satellite data, is used as a proxy for the land type that abandoned land

19 May 2011 34


reverts to. No analysis is provided to test the latter assumption but it has a significant
and reducing effect on the ILUC emissions value for biodiesel from soy because these
two different uses of satellite data can lead to large differences between the
emission/sequestration factors for converted/reverted land in the same region. For
example, when 1 hectare of cropland or pasture expansion occurs in the Brazilian
Amazon, it is assumed that 54% of this expansion takes place in forested areas. When
1 hectare of cropland or pasture is abandoned in the same region, 83% of it is
expected to revert back to forest.

In the case of soy, a very large reduction of pasture area in the Amazon biome is
found. This leads to large amounts of carbon sequestration in the Amazon. This factor,
combined with the negative emissions associated with domestic land-use change, lead
to a value of 43 tCO2eq/ha, which is extremely low compared to the other studies.

As the amount of land-use change at 0.94 ha/toe is large, the total emission effect is
still significant at 32 gCO2eq/MJ biofuel using a time horizon for the emissions of 30
years.

IFPRI
performed by IFPRI. Therefore, no breakdown or discussion of their 75 gCO2eq/MJ
value can be provided. The exception is the projection on time horizon for the
emissions at 20 years, where several of the other studies use 30 years. If a 30-year
horizon were to be used, the value would be 50 gCO2eq/MJ.

Öko-Institut/IFEU (ILUC Factor)
The methodology of the ILUC Factor is explained in detail in Section 2.5.2. It is argued
there that, after choosing the same reference year 2010 and the same ILUC risk level
of 50%, the crop yield is the only remaining variable.

From the results for soy, we see that the assumed crop yields from soy are lower than
those of sugarcane, as the GHG emissions associated with (I)LUC are higher at 67
gCO2eq/MJ fuel. The projection on time horizon for the emissions is 20 years, where
several of the other studies use 30 years. If a 30-year horizon were to be used, the
value would be 45 gCO2eq/MJ.

the differences
All studies find significant GHG emissions associated with (I)LUC for biodiesl from soy,
When all results are compared on a 30-year time horizon basis, they span a range of
32 – 62 gCO2eq/MJ biodiesel fuel.

EPA finds the lowest value at 32 gCO2eq/MJ. In a way this is remarkable, as EPA finds
by far the highest number for cropland expansion at 0.94 ha/toe. This is completely

19 May 2011 35


compensated by the extremely low average value for emissions per hectare of
cropland expansion: 43 tCO2eq/ha. This value is largely influenced by two effects.
First, a significant part of the cropland expansion occurs in the US which leads to a
negative emission effect, meaning that the land-use changes lead to carbon
sequestration. As this happens endogeneously in the model, its precise reason cannot
be identified. In addition, international land-use change leads to a very large reduction
of pasture area in the Amazon biome. The EPA methodology assumes that this area
previously used as pasture will, when it is not used as cropland, largely revert to
forest30. This leads to large amounts of carbon sequestration in the Amazon.

The values for IFPRI and the ILUC Factor are relatively close: 67 and 75 gCO2eq/MJ
with a 20-year time horizon for emissions respectively (values would be 45 and 50
gCO2eq/MJ with a 30-year time horizon). As IFPRI provide very little detailed data on
their modelling, their value cannot be explained further. The ILUC Factor again, as
with sugarcane, heavily depends on the assumption of the ILUC risk level. We chose
to use the medium level here.

The CARB result is a bit higher, when compared based on the 30-year time horizon, at
62 gCO2eq/MJ. The amount of cropland expansion is about three times lower than in
the EPA study (0.31 instead of 0.94 ha/toe), but the associated emissions per hectare
are almost six times higher (253 instead of 43 tCO2eq/ha). This latter fact is caused
by the absence in the CARB study of the effect of pasture reduction in the Amazon
found in the EPA study.

The end result of the E4tech study is in line with that of the CARB study at 66
gCO2eq/MJ. However, this could be coincidental as E4tech assumes substitution of soy
oil by palm oil and rapeseed oil and thus uses the values found for those two oils to
calculate the value for soy oil.

2.5.4 Other pathways
The IFPRI study and the E4tech study have performed pathway-specific analysis for
more crops than corn, sugarcane and soy. These are discussed here.

IFPRI
The IFPRI study has run pathway-specific model runs for more crops than corn,
sugarcane and soy. No detailed intermediate results are available for these pathway-
specific calculations performed by IFPRI. However, there are end results for GHG
emissions associated with (I)LUC based on a 20-year project horizon. These are
presented in Table 2 - 7.

30
Please refer to the detailed discussion of the EPA methodology in this section for more details on this
effect.

19 May 2011 36


Table 2 - 7 Results from pathway-specific model runs in the IFPRI study other than corn,
sugarcane and soy.

Pathway GHG emissions associated with (I)LUC
(gCO2eq/MJ fuel)
Ethanol from:
Sugar beet 16
Wheat 37
Biodiesel from:
Palm oil 50
Rapeseed oil 54
Sunflower oil 61

The lack of detailed intermediate results makes it difficult to discuss the found values.
What is clear though is that sugarbeet (16 gCO2eq/MJ) scores about equal to
sugarcane (18 gCO2eq/MJ). It is also clear that relatively high yielding crops – sugar
beet, sugar cane, palm oil – score well within their respective fuel type category.
Finally, in this study, ethanol pathways score significantly better than biodiesel
pathways, although this can not be explained from the limited information provided.

E4tech
E4tech has done detailed additional pathway-specific causal-descriptive analysis for
palm oil, wheat and rapeseed oil. The results are presented in Table 2 - 8 and then
discussed per pathway.

Table 2 - 8 Pathway-specific results on GHG emissions from (I)LUC for biofuel from palm oil,
wheat and rapeseed oil in the E4tech analysis. Please refer to the text for detailed
discussion of these results.

Palm Wheat Rapeseed
E4tech Unit
oil oil
Effect of co-products as share of total emissions
% 50% 318% 67%
Biofuel-induced agricultural intensification % 0% 78% 25%
Reduced demand in other sectors % 0% 0% 0%

Cropland expansion % 100% 22% 75%

Cropland expansion ha/toe 0.11 -0.32 0.32

19 May 2011 37


Palm Wheat Rapeseed
E4tech Unit
oil oil
Weighted average of emissions from crop expansion31 tCO2eq/ha 834 107 101
Project horizon for emissions from (I)LUC Years 30 30 30
gCO2eq/MJ
Emissions from (I)LUC 81 -36 17
fuel
Emissions from (I)LUC (assuming a standardized project gCO2eq/MJ
122 -54 26
horizon of 20 years; for comparison only) fuel

Palm oil
[We reviewed scenario 4 of the palm oil scenarios of E4Tech, please refer to Section
2.4 for details.]

E4tech assumes that all palm oil used for biofuels will lead to additional palm oil
production, as palm oil is currently the cheapest vegetable oil on the market. This
expansion is assumed to take place in regions that are currently already producers of
palm oil: mainly Indonesia and Malaysia. No biofuel induced agricultural intensification
is foreseen based on expert opinions by stakeholders.

The palm oil is found to have a strong co-product effect, that compensates for 50% of
the emissions of palm expansion. The main driver for this is that E4tech assumes that
the co-product palm kernel oil will displace low yield coconut oil. Unfortunately, this is
not readily supported by historic data, as we explain in Section 2.6.4.

This large co-product effect, combined with the high yield of oil palms, leads to a
relatively low cropland expansion of 0.11 ha/toe. However, as this expansion takes
place in Indonesia and Malaysia, including significant amounts of forested land as well
as peatland, the emissions associated with this expansion are extremely high at 947
tCO2eq/ha. This is partially caused by the fact that the scenario reviewed by us
includes 33% of expansion on peatland, which has very high carbon stocks. This
causes the resulting (I)LUC emissions to be high at 81 gCO2eq/MJ.

Wheat
[We reviewed scenario 4 of the wheat scenarios of E4Tech, please refer to Section 2.4
for details.]

In most E4tech scenarios, including the one reviewed by us, 78% of the wheat needed
for biofuels comes from additional yield increases, not requiring any additional land.
This is the results of the use of the Lywood (2009a) method, discussed further in
Section 2.6.1. Any additional fertiliser emissions needed for these yield increases are
not included in the calculations. This is discussed in more detail in section 2.6.2.

31
The values in this row are for the actual expansion of the main crop considered and exclude any influence
of the co-products. When these are taken into account, values are: 956, 144 and 69 tCO2eq/ha respectively.

19 May 2011 38


The remaining 22% of the wheat needed for biofuels is met by extra land use for
wheat production in the EU. This is assumed to lead to avoided abandoning of
cropland in the EU with no impacts outside the EU. We critically discuss this
assumption in Section 2.6.5. The yields assumed for wheat on this otherwise
abandoned cropland are assumed to be the EU average. Section 2.6.6 discusses this
assumption and its effects in more detail.

Furthermore, this cropland expansion on otherwise abandoned land leads to low
emissions as the avoided reversion of cropland to natural land has low associated
foregone carbon sequestration. This is partially caused by an error in the underlying
carbon stock data used in the study, as discussed in Section 2.6.7.

Finally, the high yielding wheat also produces significant amounts of DDGS co-product
which is assumed to be used as animal feed. DDGS is assumed to replace significant
amounts of soy meal, leading to avoided conversion of natural land to low yield
soybean cropping in South America. This avoided soy area expansion more than
compensates for any wheat area expansion in the EU, leading to a net area reduction
of -0.32 ha/toe. (Note that this is party explained by the fact that only 22% of the
wheat for biofuels causes additional land to be taken into wheat production, the other
78% of the wheat comes from biofuel induced yield increases.) At the same time,
emission savings due to avoided conversion of natural land to soy cropland in South
America are relatively high, as this concerns land with relatively high carbon stocks.
This leads to a very high co-product credit of 318% for wheat ethanol. Please refer to
Section 2.6.3 for more discussion on this.

The resulting total (I)LUC emissions for ethanol from wheat are thus negative: -36
gCO2eq/MJ. In interpreting this value, it should be noted that all assumptions critically
discussed in Sections 2.6.1-2.6.7 have a reducing effect on the (I)LUC emissions for
ethanol from wheat as noted in the text above. In the other wheat to ethanol
scenarios analysed by E4tech some of these assumptions have been varied, but in
each scenario only one of the assumptions is varied. Therefore, while most of these
scenarios find a higher ILUC value than scenario 4, none of them gives an indication of
the ILUC value in which all issues described in section 2.6 are taken into account.

Rapeseed oil
[We reviewed scenario 1 of the rapeseed oil scenarios of E4Tech, please refer to
Section 2.4 for details.]

Rapeseed oil use for biofuels is assumed to be met by extra production in the EU first.
As no biofuel induced yield increases are foreseen in the EU, this is assumed to lead a
chain of agricultural displacement events, ultimately leading to avoided abandoning of
EU cropland compared to the baseline scenario. As with wheat, this is subject to the
considerations in Section 2.6.5 and 2.6.6. However, E4tech finds an insufficient land

19 May 2011 39


availability in the EU to produce all needed rapeseed. Therefore EU imports of
rapeseed from the Ukraine are increased.

The additional demand for rapeseed from Ukraine leads to significant biofuel induced
yield increase in the Ukraine: 25% of the total biofuel demand for rapeseed. We
discuss this in Section 2.6.1. The remaining demand is met by area expansion of
rapeseed in the Ukraine on natural land.

This extra land use for rapeseed in the EU and the Ukraine are compensated by the
fact that the co-product rapeseed meal is used as animal feed. As with wheat DDGS
this leads to avoided conversion of natural land to low yield soybean cropping in South
America. This leads to a large co-product credit of 67% for rapeseed. Please refer to
Section 2.6.3 for more discussion.

Even though there is a large co-product credit, the low (biofuel) yields of rapeseed still
lead to a net area expansion of 0.32 ha/toe. However, the emissions associated with
this LUC are found to be low: 69 tCO2eq/ha. As with wheat this low number is the
result of an error in the underlying carbon stock data used by E4tech, see Section
2.6.7.

The resulting total (I)LUC emissions for biodiesel from rapeseed oil are 17 gCO2eq/MJ.
In interpreting this value, it should be noted that all assumptions critically discussed in
Sections 2.6.1-2.6.7 have a decreasing effect on the (I)LUC emissions for biodiesel
from rapeseed as noted in the text above. In the other rapeseed to biodiesel scenarios
analysed by E4tech some of these assumptions have been varied, but in each scenario
only one of the assumptions is varied. Therefore, while most of these scenarios find a
higher ILUC value than scenario 1, none of them gives an indication of the ILUC value
in which all issues described in section 2.6 are taken into account.

Comparison of IFPRI and E4tech results
Unfortunately it is not possible to compare the IFPRI and E4tech results for biofuel
from palm oil, wheat and rapeseed oil in detail, as the IFPRI study does not provide
the necessary detail.

2.6 Discussion on key assumptions in E4tech-study

One of the key benefits of the causal descriptive ILUC modelling performed by E4tech
is its transparency. Almost all key assumptions are documented in the report, making
it transparent how the various ILUC numbers were obtained. This is in sharp contrast
to most agro-economic equilibrium models, in which assumptions on key parameters
are often hidden in the mathematical functions of the model where they are usually
invisible to people other than the modellers.

The transparency of the E4tech study allowed this review to reflect on some of the key
assumptions made in that study in more detail, which we have done in this section. In

19 May 2011 40


general, the comments we make with respect to these assumptions should not be
taken as a weakness of the study relative to the studies based on agro-economic
equilibrium models as the lack of transparency in these models did not permit a fair
comparison within the scope of this report.

The next sections describe the key assumptions made in the E4tech scenarios
reviewed by us in more detail. In their study, E4tech analysed different scenarios with
different assumptions to assess the impact of variations in assumptions on the end
result. As explained in Section 2.4, we chose the E4tech scenario with the most
common denominators in its assumptions for our review. Where relevant, we discuss
the results of the variations in the assumptions in these scenarios in this section.

In discussing the assumptions, we have tried to, where relevant, indicate: whether the
assumption on a parameter is in line with historical developments for that
parameter32; whether the assumption is significantly different from historical
developments and what reasons are given for that in the report; or whether the
assumption is based on an error in the report or the underlying data.

2.6.1 Demand induced yield increases
Why it matters - optimistic assumption. All biofuel demand that can be met through
an additional increase in yields, that results from the additional demand for biofuel
feedstock, does not require additional land and therefore does not cause land use
change. Therefore, the size of the demand induced yield increase is a key parameter
in any ILUC study. This is also discussed in Section 2.3.

Assumption in E4tech scenario analysed in this report: for most crop-scenarios
the E4-tech study based their assumptions on a paper by Lywood (2009a) which
establishes a relationship between output growth and yield and area growth. This
leads to a very large demand-induced yield increases for especially EU wheat, and to a
lesser extent for sugarcane and Ukrainian rapeseed. For wheat, 78% of total biofuel
demand is met by additional yield increases in the scenario reviewed by us.

Discussion: the wheat-yield growth numbers resulting from the assumption in E4tech
(2010) for EU crops are very high compared to wheat-yield growth numbers in the EU
since 1990. Scenarios 4-8 for wheat in E4tech (2010) assume an annual yield increase
of 1.35% between 2008 and 2020. It is furthermore assumed that two-thirds of this is
caused by the increased biofuel demand (without the biofuel demand, the yield is
assumed to grow by only around 0.45% annually). For comparison, the actual annual
yield growth in the EU-27 between 1990 and 2009 was 0.7%. Between 1999 and 2009
it was 1.0%. In summary, for EU wheat E4tech (2010) assumes a future annual yield
growth that is significantly higher than it has been in the last 10 to 20 years and
attributes the majority of it to the increased biofuel demand. This results in the

32
All historical numbers on yields, area and production are from FAOSTAT unless otherwise stated.

19 May 2011 41


assumption that 78% of all additional wheat needed for biofuel is met by additional
yield increase (and therefore has zero land use impacts). A lower demand induced
yield increase would lead to more LUC and therefore a higher ILUC value. This is
explored in scenario 3 of the E4tech wheat analysis. In that case, with the lower yield
effect being the only difference compared to the scenario reviewed by us, the resulting
GHG emissions are 31 gCO2eq/MJ fuel higher.
It is worth noting that the high yield growth prediction in E4tech (2010) compared to
historic numbers cannot be explained by a stronger growth in production in 2008-
2020 compared to historic growth in production. Annual growth in EU-27 wheat
production in scenarios 4-8 in E4tech (2010) amounts to 1.1%, the same as the actual
average annual production growth in EU-27 wheat production realized between 1990
and 2009.
The assumed yield growth for Ukrainian rapeseed and Brazilian sugarcane are in line
with historical yield growth numbers but what stands out is that the vast majority of
the yield increase is assumed to be driven by the additional biofuel demand. For
Brazilian sugarcane, 67% of total yield growth between 2008 and 2020 is assumed to
be the result of additional biofuel demand. In other words, without this additional
biofuel demand yields would grow only a third of what they do in the biofuel scenario.
The same number for rapeseed from Ukraine is even higher, at 93%. What is
remarkable about the rapeseed number is that in de baseline production is increasing
by 3.2% annually and yields only grow 0.2% annually, a ratio of 0.2:3.2 = 0.08. In
the biofuel scenario the yield suddenly grows 12 times faster, at 2.9% annually. While
the growth in production is also stronger in the biofuel scenario, 8.6% annually, the
ratio of growth in yield to growth in production now is 2.9:8.6 = 0.34. In other words,
in the biofuel scenario yields respond more than four times as strong to growth in
production than in the baseline scenario.

2.6.2 Emissions from additional fertilizer
Why it matters – optimistic assumption. In E4tech (2010) the additional feedstock
for biofuels partly comes from additional yield increases. Especially for wheat this
fraction is very high: 78% of the wheat for ethanol stems from additional yield
increases. Some parties have argued that such yield increases will, at least in part,
stem from additional nitrogen fertilizer application and that this will therefore increase
GHG emissions from fertilizer production and application.

Assumption in E4tech-study: E4tech (2010) does not take into account GHG-
emission from additional fertilizer. (Note that most of the reviewed studies do not take
this into account). The argumentation given for this is that a) not all yield increases
will stem from increased fertilizer usage and b) emissions from additional fertilizer are
not expected to change the order of magnitude of ILUC emissions. The latter is
illustrated with a simple calculation of the rough size of such emission for wheat
ethanol, which they estimate to be 8.9 gCO2/MJ ethanol.

19 May 2011 42


Discussion: with respect to a) it seems plausible that certainly not all yield increases
will stem from additional fertilizer application. However, with respect to b) E4tech base
their conclusion on the size of emissions from additional fertilizer on a calculation
based on rough assumptions on the effect of additional fertilizer on yields. Based on
these assumptions E4tech (2010) finds a maximum value of 8.9 gCO2/MJ ethanol.
Using the same assumptions we find a maximum value of 12.9 gCO2/MJ ethanol. As
further explained in Appendix A, the difference is due to the fact that E4Tech (2010)
allocates the additional N-fertilizer emissions to all EU-wheat ethanol instead of to the
additional wheat ethanol resulting from demand-induced yield increase.

The underlying assumption used in E4tech’s calculation is that a 10% increase in N-
fertilizer in the UK would increase wheat yields by 0.5 t/ha. No reference is provided
for this. A report by the UK HGCA (HGCA, 2007) shows a much higher increase in
fertilizer needed to increase UK wheat yields by 0.5 t/ha, around 35%. Applying these
numbers leads to 106 gCO2/MJ ethanol resulting from additional fertilizer to achieve
the higher yields: a number 12 times larger than the number given by E4tech.
Therefore, the number of 8.9 gCO2/MJ ethanol appears to be the result of a
combination of 1) an optimistic assumption on wheat yield responses to higher N-
fertiliser inputs and 2) an optimistic assumption on the allocation of the emission to
the additional wheat-ethanol produced. This is discussed in more detail in Appendix A.
The higher number of 106 gCO2/MJ ethanol mentioned above surely is an
overestimate of the actual emissions from the additional yield increase, as not all yield
increase will come from increased N-fertilizer application, but it shows these emissions
could be significantly larger than 8.9 g CO2/MJ ethanol.

2.6.3 The role of co-products – DDGS and rapeseed meal
Why it matters – optimistic assumption. Additional availability of co-products from
biofuel production reduces the demand for the products that are replaced by these co-
products. The reduction in land needed to grow these replaced products depends on
what product is replaced and at what yields.

Assumptions in E4tech scenario analysed in this report: the E4-tech study based
their assumptions on a paper by Lywood (2009b). What stands out compared to other
ILUC studies is the high proportion of soy meal replaced by especially rapeseed meal
and wheat DDGS: every ton of DDGS or rapeseed meal replaces 0.6 ton of soy meal33.
Because of the low yields of soy and the deforestation in South America connected to
it, replacing soy meal results in a very large carbon credit. A lower soy meal
replacement rate is explored in scenario 5 of the E4tech rapeseed oil analysis. In that
case, with the lower replacement rates being the only difference compared to the

33
For both wheat and rapeseed, one scenario is included in which only 50% of the DDGS or rapeseed meal
is used to replace other products. This significantly increases the iluc numbers for wheat-ethanol and
rapeseed-biodiesel.

19 May 2011 43


scenario reviewed by us, the resulting GHG emissions are 11 gCO2eq/MJ fuel higher. In
the wheat analysis, no variation of soy meal replacement rates was analysed.

Because less soy is produced, less soy oil is produced as well. E4tech (2010) assumes
this is compensated by an increase in palm oil production. The GHG effect associated
with this increase is calculated from one of the palm oil scenarios calculated by
E4tech. It is worth noting that the outcome of the chosen palm oil scenario is in the
low end of the range found for indirect palm oil emissions, as this scenario considers
only 5% expansion on peatland.

Discussion on soy meal replacement: the replacement ratio of 0.6 ton of soy meal
per ton of DDGS or rapeseed meal is high compared to some other ILUC studies. E.g.
CARB assumes that DDGS from corn only replaces corn, not soy meal. However, there
are significant differences between the US feed market and the EU feed market that
can explain different values for the two markets. Also, the values used in E4tech
(2010) are within the range given by Croezen et al. (2008).

Discussion on soy oil compensation by palm oil
The chosen scenario for substitution by palm oil assumes that only 5% of the palm oil
produced to compensate for the reduction in soy oil production comes from peatland.
This 5% is low compared to numbers cited by peatland experts after the E4tech study
was completed. The University of Leicester (2010) estimates that current palm
plantations on peatland already make up 30% of total planted area and that this
number will be higher for new plantations. These observations are also in line with
recent satellite findings by Sarvision (2011). As illustrated by palm oil scenarios 3 and
4 in the E4tech study, increasing the percentage of palm on peat from 5% to 33%
increases palm oil iluc emissions by around 130%.

2.6.4 The role of co-products – Palm kernel oil
Assumption in E4tech-study – optimistic assumption. E4tech (2010) assumes that
the co-product from palm oil production, palm kernel oil (PKO), will replace coconut
oil. This assumption clearly impacts the ILUC number for biodiesel from palm oil, but it
also impacts biodiesel from rapeseed and soy, and ethanol from wheat (because the
co-products of these biofuels replace soy, leading to less soy oil production, which is
then compensated by an increase in palm oil production.)

Discussion: the assumption that PKO replaces coconut oil has a very significant
positive effect. This is caused by the low yield of coconut oil. This can be seen in figure
9 of E4tech (2010) (p. 47) where the credit for avoided coconut oil production makes
up for half the total carbon footprint of palm oil production. This is remarkable as for
every ton of palm oil, only around 0.12 ton of PKO are produced. Furthermore, the
assumption that increased PKO production in Indonesia would lead to a reduced
growth in Indonesian coconut oil production is not supported by historic figures.

19 May 2011 44


As can be seen from the figure below, growth in Indonesian coconut oil has been
relatively stable since the 70s. Indonesian PKO production was all but non-existent
until the mid 80s when it took off to become more than three times as large as
coconut oil production today. Global numbers show a similar development. These
numbers illustrate that it is unlikely that PKO replaces large amounts of coconut oil. If
they would, the very strong historic growth in PKO production would have had a more
pronounced effect on historic coconut oil production. A smaller land credit for PKO
would lead to higher ILUC for palm oil, and therefore higher ILUC for biodiesel from
palm oil, but also for biodiesel from rapeseed and soy as well as ethanol from wheat
as explained above.
Coconut oil Coconut oil
Palm kernel oil Palm kernel oil
2,500 6,000

5,000
2,000
Indonesian production (kton)

Global production (kton)
4,000
1,500

3,000

1,000
2,000

500
1,000

0 0
1961

1966

1971

1976

1981

1986

1991

1996

2001

2006

1961

1966

1971

1976

1981

1986

1991

1996

2001

2006
2.6.5 Area expansion in EU takes place on abandoned cropland
Why it matters - optimistic assumption. The type and the location of land on which
expansion for biofuel crops take place has an effect on the carbon emissions this
causes. E.g. an expansion on abandoned cropland in the EU will cause lower LUC
emissions than an expansion on forested area in tropical regions. This is caused by
two elements. First, the carbon stock per hectare differs for different land types.
Second, yield will differ per location and this determines the amount of hectares
needed to grow the required quantities of crop.

Assumptions in E4tech scenario analysed in this report:
• E4tech (2010) assumes that the additional EU-wheat area and the additional EU-
rapeseed area34 needed for biofuels would not lead to new land being converted to
cropland, but to less land being abandoned in the EU compared to the baseline
scenario35. This is primarily explained by the assumption that in the baseline the

34
Note that part of the additional rapeseed is assumed to be imported from Ukraine, where it does lead to a
cropland expansion.
35
Note that in wheat scenario 3 a small variation to this assumption is studied by E4tech. In this scenario,
85% of additional EU wheat area is still realised on otherwise abandonned cropland, but in this scenario
15% of the total increase in EU wheat area compared to the baseline takes place “new” cropland. While
scenario 3 finds a significantly higher ILUC number than most other wheat scnearios, this should largely be

19 May 2011 45


total area used for wheat cultivation is decreasing by 2.2 Mha between 2008 and
2020.
• In addition, it is acknowledged that the area no longer used for wheat cultivation
in the baseline would not actually be abandoned but would be used for other crops
as wheat is typically grown on good agricultural land. It is then assumed that this
land would have been used by other crops in the baseline, but by wheat and/or
rapeseed in the biofuel scenarios. It is finally assumed that the additional wheat
and/or rapeseed in the biofuel scenario will displace these other crops to other
areas within the EU and that the area that they are displaced to is of a similar size
as the area they are displaced from (now used for wheat and/or rapeseed).

Discussion:
• The fact that the EU wheat area declines by 2.2 Mha between 2008 and 2020 in
the baseline in E4tech (2010) is the result of an explicit assumption on the
baseline. The baseline in E4tech (2010) is based on predictions by FAPRI (2009)
but all feedstock produced for biofuels beyond the 2008 levels are taken out in the
baseline. This includes EU 1.4 Mha of wheat for biofuels in the EU.36 Thereby,
E4tech (2010) implicitly assumes that the decline in the EU wheat area in the
baseline, which is the result of a modification to the actual IFPRI scenario,
automatically translates into a similarly sized decline in EU-cropland. However, it
is also possible that (part of) the area no longer used for wheat would have been
used for an increase in the area used for another crop. (Note in this respect that
E4tech (2010) also assumes a 1.2 Mha increase in EU rapeseed between 2008 and
2020.) In that case, (part of) the land would not have been abandoned and (part
of) the additional wheat/rapeseed area for biofuels would either have to come
from an expansion in cropland in the EU or will displace other crops that will then
have to be produced outside the EU (where yields may be lower, leading to more
LUC, and where carbon stocks may also be higher, leading to higher emissions per
hectare). This would likely to lead to higher ILUC emissions for ethanol from
wheat and biodiesel from rapeseed.
• E4tech (2010) argues that, in the baseline, the land that is taken out of wheat
production is not expected to be abandoned but used for other crops because
wheat is typically grown on high quality agricultural land. However, this seems to
contradict with the assumption in E4tech 2010 stated above: that if a certain size
area is not taken out of wheat production in the biofuel scenario, this would lead
to a similar size of land not being abandoned somewhere else in the EU (because
the crops that would have taken the place of wheat in the baseline scenario,
because of which land would have been abandoned elsewhere in the EU, now still

attributed to the lower yield increase assumption compared to other scenarios. The effect of having 15% of
wheat expansion for biofuels on “new” cropland in stead of otherwise abandonned cropland, keeping
everying else constant, has not been quantified. No scenarios are included in E4tech in which (part of) the
wheat expansion displaces EU exports, possibly leading to cropland expansion outside the EU.
36
FAPRI predicted only a reduction of around 0.8 Mha in EU wheat by 2020 compared to 2.2 in the E4tech
(2010) baseline.

19 May 2011 46


need to be produced elsewhere in the EU). This latter assumption assumes that
the yields of these other crops would have been the same on the high quality land
used for wheat as on the land on which they would have otherwise been grown.
This contradicts with the statement that taking an area out of wheat production
does not cause that land to be abandoned because the quality of the land is so
good that other crops will move onto that good land. If yields of such other crops
are indeed lower on the areas they are displaced to if the good land is used for
wheat for biofuels, the total amount of cropland not being abandoned in the EU in
the biofuels scenario is larger than the area used for wheat or rapeseed
production for biofuels. This would increase the total LUC for biofuels from EU
wheat and EU rapeseed as well as the associated emissions from LUC.

2.6.6 Yield on abandoned agricultural land
Why it matters – optimistic assumption. If new land is taken into production for
biofuel feedstock production, the amount of land needed for this depends on the yields
on that land. There has been debate on weather yields on land that are taken into
production in the future will be lower than the average yield. Some parties have
argued that this would be likely as the best land is already in use. As E4tech (2010)
discusses, other parties have argued that this depends very much on the region and
crop considered.

Assumption in E4tech-study: E4tech (2010) assume average yields on lands taken
into production in the future and they state various reasons for this. Our focus here is
on the assumption that yields on abandoned agricultural land are also equal to the
average yields. This is relevant primarily for the EU crops wheat and rapeseed which
E4tech (2010) assumes to be grown on abandoned agricultural land in the EU (or land
that would have stopped growing the wheat in the absence of biofuel demand) and for
which they assume the yields to be equal to the average yields.

Discussion: the assumption that for a certain crop the yields on abandoned
agricultural land in the EU-27 are equal to the average yields of that crop in the EU-27
seems to be contradicted by the following observations:
• Between 1999 and 2008 11 countries in the EU-27 saw their wheat area decline
by a total of 1.6 Mha (other countries showed an increase in wheat area.) The
average 2007-2009 yields of wheat in these countries, weighted by the size of the
abandoned wheat area in each country, amounts to 4.2 t/ha. The EU-27 average
wheat yield in 2007-2009 was 5.3 t/ha. This implies that, even if the yields on the
abandoned areas where the same as the national average, the average yield on
the land that was taken out of wheat production between 1998 and 2008 was
22% lower than the EU-27 average EU wheat yield.
• Yields within a country show a certain range due to varying qualities of land within
a country. Since yields are a key determinant of profitability of wheat production it
is expected that “land in low yielding regions is likely to go out of production
sooner than in higher yielding regions.” (University of Cambridge, 2006). This

19 May 2011 47


would mean that the difference between the average EU-27 wheat yields and the
yields on abandoned land is even larger than the above mentioned 22%.
• Finally, there are also yield differences between the various fields within a single
farm. A farm survey in the UK showed the wheat yields on the worst fields of a
farm, where 37% lower than the average wheat yields on that farm. (Defra xx).
Combined, the three effects discussed above imply that the yields on abandoned
cropland in the EU could be significantly lower than the average yields in the EU. in
that case the amount of (I)LUC and the associated emissions would increase for both
biodiesel from rapeseed and ethanol from wheat.

2.6.7 Forest reversion and carbon sequestration in the EU
Why it matters – error in underlying data. In E4tech (2010) most of the cropland
expansion for additional wheat and rapeseed for biofuels is assumed to come from EU
cropland. It is assumed that this leads to a reduction in the amount of cropland
abandoned in the EU by 2020. To determine the GHG-effects of this, the relevant
question is what the carbon stock on this land would have been if the land would
indeed have been taken out of cropland use.

Assumption in E4tech-study: E4tech (2010) uses the Winrock data prepared for the
EPA for RFS-2 on land conversion (for cropland expansion) and land reversion (for
abandoned cropland) as well as Winrock data on the carbon stocks on the various land
types that abandoned cropland in the EU would revert to. The same numbers are used
by several of the other studies reviewed in this report, namely RFS-2 and IFPRI.

Discussion: the Winrock numbers used in E4tech (2010) for carbon sequestered in
abandoned cropland that reverts back to forest contain an error. The number for
carbon stored in biomass on land reverting back to forest was accidentally set to zero
for the first 20 years. The carbon sequestration in biomass for land reverting to forest
in the US is 9 t CO2/ha/y. Using the same number for the EU, the average carbon
sequestration on abandoned cropland that would revert back to forest increases from
29 t C/ha to 78 t C/ha. Because only 34% of abandoned cropland is assumed to revert
back to forest, the average carbon sequestration on abandoned cropland increases
somewhat less, from 29 t C/ha to 51 t C/ha. Nonetheless, this still means that the
total emissions from land use change inside the EU increase by 75%37. This increases
the ILUC number for rapeseed-biodiesel by 25 gCO2/MJ and the ILUC number for
wheat-ethanol by 13 gCO2/MJ in the scenarios analysed in this report. The smaller
increase for wheat is explained by the fact that 78% of the wheat for biofuel is
assumed to come from additional yield increases, and the higher carbon stock
numbers therefore only affect 22% of the wheat for ethanol.

37
Note that the total ILUC number for rapeseed-biodiesel and wheat-ethanol is determined by more factors
than emissions from LUC in the EU alone.

19 May 2011 48


In addition to the above error some parties have commented that care must be taken
in interpreting the Winrock numbers when determining the fraction of abandoned land
that will revert to forest, or the number may be underestimated. The reason for this is
that the analysis of Winrock is based on satellite images from 2001 and 2007 and that
land that was cropland in 2001 and abandoned after 2001 will often not show up as
forest yet on the satellite image of 2007 but rather as one of the ‘intermediary’ land
types such as “grassland”, “mixed”, “savannah” or “shrub land”. It has been argued
that some of the abandoned cropland that reverted to one of these intermediary land
types by 2007 would still revert to forest but would simply need more time for this,
especially if the cropland was abandoned close to 2007. If this is correct, the actual
foregone carbon sequestration, as a result of additional demand for rapeseed or
wheat, would be higher.

2.6.8 Carbon stocks losses from conversion to pasture in Brazil
Assumption in E4tech-study - error in Brazilian sugarcane scenarios. E4tech (2010)
studied various scenarios for Brazilian sugarcane in which different assumptions are
made on the knock-on effect of sugarcane displacing cattle. In the scenario reviewed
in this report (scenario 7), 1.0 ha of pasture displaced by sugarcane leads to 0.84 ha
of pasture expansion elsewhere in Brazil. Because most of the pasture expansion
takes place in the north of Brazil, E4tech intended to use carbon stock numbers for the
north of Brazil as described on page 130 of E4tech (2010). However, accidentally the
wrong carbon stock numbers were used – those for the average of entire Brazil, not
specifically for the north of Brazil.

Discussion: the error in the sugarcane scenarios means that the average carbon
stock loss for lands converted to pasture, as a result of sugarcane expansion on
pasture lands, should be 91 t C/ ha in stead of 56 t C / ha, a 62% increase. Correcting
for this error results in a 50% increase in the ILUC emissions from Brazilian
sugarcane. Because not all sugarcane ethanol comes from Brazil in the E4tech
scenarios, the effect on the average ILUC emissions from sugarcane ethanol are
smaller: between 3 and 10 gCO2/MJ depending on the scenario according to the
authors.

2.6.9 Conclusion - impacts of assumptions in E4tech study on various crops
The points raised above with respect to some of the key assumptions in the E4tech
study have different impacts on different biofuel pathways, as illustrated in the table
below. In many cases these impacts are associated with the main crop considered, but
in some cases the effect takes place through the calculated co-product credit. The
biofuel pathways that are most affected are wheat ethanol and rapeseed biodiesel as
all the above raised points, except the last one, have an effect on these pathways.
It should be noted again that the E4tech study analysed several scenarios for each
crop and that some of the assumptions critically discussed here have been varied in
other scenarios. However, in doing so only one of the assumptions is varied each time,
and some of the above discussed assumptions or errors have not been varied in other

19 May 2011 49


scenarios. Therefore, these alternative scenarios do not give an indication of what the
ILUC value would be if all the above issues are taken into account.
Table 2 - 9 Overview of the issues identified in the reviewed E4tech scenarios and the crops for
which they impact on the ILUC value.

Biodiesel Ethanol
Assumption PO SO RO Wheat SC
Demand induced yield increase X X X X
Emissions from additional fertilizer X X X X
Co-products: role of DDGS and rapeseed meal X X
Co-products: PKO replaces coconut oil X X X X
EU area expansion on abandoned cropland X X
Yields on abandoned cropland X X
C-sequestering on abandoned EU cropland X X
C-stock losses from conversion to pasture X

2.7 Conclusion

We reviewed the quantification of pathway-specific indirect effects of biofuel
production for biofuel from corn, sugarcane, soy and other feedstocks using seven
studies.

Within each pathway, there is no clear consensus on the size of total emissions from
direct and indirect land-use change. A slight trend can be seen where sugarcane
generally has the lowest emissions from land-use change (4-46 gCO2eq/MJ), followed
by corn (27-103 gCO2eq/MJ, with the second highest value being 54 gCO2eq/MJ),
followed by soy (32-75 gCO2eq/MJ). In the other pathways, a similar trend is visible:
ethanol pathways score better than biodiesel pathways. However, due to the large
ranges in the results it would be premature to draw firm conclusions, based on the
studies reviewed in this report, on the (I)LUC from ethanol versus biodiesel.

While the results vary per study, most studies find the emissions for most pathways to
be significant when compared to e.g. a fossil reference value of 80-90 gCO2eq/MJ fuel.
An exception to this rule is the ethanol from wheat scenario in the E4tech study, that
finds negative (I)LUC emissions. However, the assumptions and error made in that
scenario (see Section 2.5.4 for details) were critically discussed in Section 2.6 of this
review.

But even when studies find comparable numbers this does not always imply they
share a common understanding of ILUC: there may simply be various differences in
the intermediate outcomes that cancel each other out. For example, EPA and CARB
find comparable numbers for ethanol from US corn while they differ markedly in their
findings on the amount of (I)LUC and the GHG emissions per unit of (I)LUC. This

19 May 2011 50


illustrates the importance of looking at the underlying numbers in addition to the end
results.

The differences between specific studies within a certain pathway can be made
reasonably clear by our framework for quantitative comparison of the intermediate
results. In many instances the data to derive these intermediate results are not
available and, more importantly, a clear explanation of the causes of the found
differences is not given by the authors. Sometimes these data are in principle
available, but are not reported in the studies presented by the quantification
initiatives. In other instances, the data cannot be extracted from the quantification at
all, mostly due to the complex setup of the models and methodologies used. An
exception to this rule was the E4tech study which contained a very transparent and
well documented causal-descriptive approach. However, in general, a more
comprehensive documentation of assumptions and intermediary results would allow
for a more detailed comparison between models, their similarities and differences.

19 May 2011 51


Box 2: ILUC impacts on biodiversity

All studies reviewed in this report focus on the GHG emissions from (I)LUC
resulting from an additional demand for biofuels. None of the studies includes a
detailed analysis of the impacts of (I)LUC on biodiversity. However, as explained in
Section 2.2.1 indications on the impacts on biodiversity are available from the
existing studies. After all, all of the reviewed studies, except the ILUC-factor from
the Öko -Institute, provide the following information:
The amount of (I)LUC that takes place, usually on a country/region level (step
2 in the 4-step process described in Section 2.2.1). This information is
summarised for each crop in each study in the tables of Section 2.5 in ha/toe.
The type of land converted as a result of (I)LUC (step 3 in the 4-step process
described in Section 2.2.1. Combined with the information on the
country/region where the (I)LUC takes place, this provides information on the
type of biome or ecosystem that is converted.
Therefore, even though the reviewed studies did not explicitly analyse the impacts
on biodiversity, they contain valuable information on the potential biodiversity
impacts of additional biofuel demand.
Another indication of the impacts on biodiversity of additional biofuel demand is
given by the recent “Rethinking Global Biodiversity Strategies” (PBL 2010). While
this study is not specifically focussed on the impacts of biofuels or bioenergy it
does find in its results that increasing the amount of bioenergy, without additional
measures to minimise impacts on biodiversity, has a negative impact on global
biodiversity by 2050 even if it reduces global warming. Interestingly the study also
finds that if additional measures are taken alongside bioenergy, such as additional
yield increases, the total impacts on biodiversity are positive.
To increase the understanding of the impacts of biofuels on biodiversity, it is
recommended that future biofuel/bioenergy (I)LUC studies attempt to quantify
these biodiversity impacts next to the impacts on GHG emissions.

19 May 2011 52


3 Mitigation of unwanted indirect effects

The previous chapters focused on the quantification of indirect effects of bioenergy
production. This chapter focuses on how unwanted indirect effects can be prevented.
It will be argued that different but complimentary solutions are needed for the short
and long term respectively. Surprisingly, most work on ILUC continues to focus on a
quantification of the effects, with still relatively little attention to concrete mitigation
options. The few mitigation options that have been developed, or are under
development, are discussed.

3.1 Possible ways to mitigate unwanted indirect effects from bioenergy

3.1.1 The bigger picture: Global versus project-level mitigation measures
In theory, three types of mitigation measures are available to prevent or minimise
unwanted indirect effects from bioenergy. The first two concern global mitigation
measures, while the third describes project-level mitigation measures:
1 Prevent unwanted direct LUC, globally and for all sectors. Unwanted ILUC from
bioenergy manifests itself through unwanted direct LUC for the production of
agricultural products for other sectors such as the food and feed sectors.
Preventing unwanted direct LUC, for example through better land-use planning
and corresponding enforcement, would thus eliminate unwanted ILUC altogether.
Note that because of the international characteristics of ILUC and the competition
for land between different sectors, this mitigation measure requires global
implementation for all land-based sectors to be effective. While a worthy
mitigation measure for the longer term, this mitigation measure is unlikely to
materialise fully in the short to medium term and is largely outside of the
influence of the bioenergy sector.
2 Reduce pressure on land from the agricultural sector as a whole by increasing
yields, supply chain efficiencies and/or a reduction in consumption, for example
through increased R&D. This could reduce the need for expanding the area used
for agricultural production. However, a globally constant or shrinking agricultural
area alone does not necessarily prevent unwanted LUC. Shifts in land used for
agricultural production (without a net increase in the total area) can still cause
unwanted LUC. Also, this mitigation measure is unlikely to materialise sufficiently
in the near future, with projections from leading agricultural institutions indicating
an expanding agricultural area during the next decades. Also this option lies
largely outside of the influence of the bioenergy sector.
3 Practical production models that prevent indirect impacts at a project level. While
the other two mitigation measures take a more macro approach (in which
governments will be key actors) this approach focuses on the role individual
producers – acting alone or in cooperation with others inside or outside of the
bioenergy sector – can play (in the absence of the above two mitigation

19 May 2011 53


measures). This includes mitigation measures such as the much debated
production on “unused land”. Such mitigation measures are able to lend
themselves to a certification approach as they focus on individual producers.

3.1.2 What individual producers can do to prevent unwanted indirect effects
Four main solutions have been put forward for producers to expand biomass usage for
energy purposes without causing unwanted indirect effects (Ecofys 2007a, Ecofys
2008, Ecofys, 2009, RFA 2008):
1 Biomass production on “unused land” – land that does not provide provisioning
services38,39. Because this does not displace other human uses of the land, it does
not cause any indirect effects. Clearly, expanding production on unused land does
lead to a direct LUC. The big advantage is that direct LUC is controllable (e.g.
through certification) and can be limited to those areas where effects are
acceptable, while the effects of indirect LUC are largely uncontrollable40.
2 Introducing energy crop cultivation without displacing the original land use
through increased land productivity or integration models. Especially in developing
countries there is a significant potential for yield improvements. Potential negative
environmental or social impacts from intensification models have to be taken into
consideration for this type of solution. Note that integration models can stretch
beyond the scale of individual farms, for example when bagasse from sugarcane is
used to feed cattle of surrounding cattle farmers.
3 Bioenergy production from residues. Current functions and uses of these residues
must be well understood, otherwise displacement, and the associated indirect
effects, may still occur.
4 Bioenergy production from aquatic biomass such as algae currently not used for
other purposes. Specific sustainability aspects for such production would need to
be taken into account.

3.2 Summary of existing mitigation initiatives

This section gives a summary of the existing mitigation initiatives for unwanted
indirect effects of bioenergy production. First, the two main characteristics on which
each initiative is analysed are presented. Then, a table gives a summary of the various
initiatives and provides an initial analysis on the two main characteristics. In sections
3.3 through 3.6 the individual initiatives are analyzed in detail on a number of
important characteristics.

38
The Millennium Ecosystem Assessment distinguishes four categories of ecosystem services: Provisioning
services, regulation services, cultural services and supporting services. Provisioning services are defined as
harvestable goods such as fish, timber, bush meat, genetic material, etc.
39
Also referred to as “degraded land”, “marginal land”, “waste land” or “abandoned land”.
40
Often an area is not completely “unused” and a sliding scale exists between this “unused land” concept
with the “intensification” concept, see next bullet.

19 May 2011 54


3.2.1 Main characteristics used in analysis of mitigation initiatives
• Scope: Is the measure focused on GHG effects only or also on other measures
such as biodiversity and food consumption?
• Behavioural change: Does the measure provide concrete incentives for
behavioural change by the actors involved in biofuel production and consumption?
This can be relevant on two levels. First: are actors driven to choose a certain
feedstock with a lower risk of indirect effects? Second: are actors that are
committed to a certain feedstock driven to make choices in their production
process that eliminate or minimise risks on indirect effects?

3.2.2 Summary of the various mitigation initiatives and their characteristics
Table 3 - 10 shows a summary of the various initiatives that have proposed or are
developing proposals for measures to mitigate indirect effects of biofuels. Detailed
analysis is provided in sections 3.3 through 3.6.

Measure Scope Drives behavioural
change
Feedstock For a
choice given
feedstock
RFS – US Renewable GHG-factor GHG
+ -
Fuels Standard
LCFS – Californian Low GHG-factor GHG
+ -
Carbon Fuel Standard
LIIB – Low Indirect Preventing GHG
Impact Biofuels displacement by Biodiversity
(Ecofys et al.)41 expanding on land Land rights + +
without provisioning Food
services consumption
Impact Biofuels displacement through Biodiversity
(Ecofys et al.) agricultural Land rights + +
intensification Food
consumption
Impact Biofuels displacement through Biodiversity
(Ecofys et al.) using wastes or Land rights + +
residues Food
consumption
EU RED – EU
Renewable Energy Various policy options are being considered
Directive

41
Formerly known as Responsible Cultivation Areas (RCA, Ecofys et al. 2010).

19 May 2011 55


Table 3 - 10 Summary of the various initiatives that have proposed or are developing proposals for
measures to mitigate indirect impacts from biofuels. For each initiative the main
measure and its scope are given. Also, it is indicated with a +/- score whether the
initiative is likely to drive behavioural change of actors as described in section 3.2.1.
Detailed analysis is provided in sections 3.3 through 3.6.

3.3 Detailed analysis of mitigation initiative: RFS

3.3.1 RFS and indirect effects
The Renewable Fuels Standard is a federal biofuel obligation in the United States that
consists of various components for different “types” of biofuels. In the RFS different
pre-defined biofuel chains (e.g. corn ethanol) are categorised based on their feedstock
and GHG performance. The GHG performance is calculated with a life cycle analysis
that includes a pre-determined amount of emissions from ILUC thus including ILUC in
the characterisation of a particular pathway for a biofuel.

3.3.2 Scope
The RFS focuses on GHG savings and for ILUC the policy includes only the GHG
effects. While this may have close links with effects on biodiversity and food
consumption, measures to mitigate unwanted effects on these aspects are not
explicitly included in the RFS.

3.3.3 Incentives for behavioural change
The emissions from ILUC have a significant impact on the GHG emissions of a biofuel
pathway in the RFS, and thereby on the type of biofuel the pathway is categorised
into. As the RFS requires that a certain part of the total target is met though biofuels
with a high GHG saving, it provides a concrete incentive for biofuel types with little or
no emissions from ILUC, such as biofuels from residues. In other words, the RFS
contains an incentive for producers to choose a feedstock with little or no emissions
from ILUC.

However, for a given feedstock, producers cannot prevent or lessen the GHG effect
from ILUC by taking additional measures to prevent or reduce the risk of ILUC,
because it is a standard, pre-determined amount coming from the life cycle
calculations done within the RFS. Thereby, for a given feedstock the RFS does not
provide any incentives for producers to change their behaviour such as to minimise
the risk of ILUC.

3.4 Detailed analysis of mitigation initiative: LCFS

3.4.1 LCFS and indirect effects
California’s Low Carbon Fuel Standard (LCFS) adopted by the Air Resources Board on
23 April 2009 requires a 10% reduction in the average greenhouse gas emission

19 May 2011 56


intensity of the State’s transportation fuels by 2020. Biofuels are expected to play a
major role in achieving these targets.

The GHG savings of biofuels compared to the fossil reference fuels are determined
through an LCA of pre-defined biofuel chains. Emissions from ILUC are included in this
LCA. Thereby the scheme provides incentives for biofuels that cause no or less ILUC.
Currently, calculations have been done for a few different pre-defined biofuel chains
fed by energy crops.

3.4.2 Scope
The focus of the LCFS is on GHG emissions and therefore only GHG effects from ILUC
are currently within the scope of the LCFS. Discussions are ongoing on including wider
social and environmental sustainability aspects.

As for the RFS, the LCFS provides an incentive for biofuel producers to use feedstocks
that have little or no emissions from ILUC. Currently, like the RFS, the LCFS does not
give clear guidelines that a biofuel producer can follow at the project level, after a
feedstock has been chosen, to prevent or mitigate unwanted indirect effects. As a
result, no incentive for implementing effective ways to mitigate indirect effects exists
other than feedstock choice.

3.5 Detailed analysis of mitigation initiative: EU Renewable Energy Directive

The EU Renewable Energy Directive (RED) contains a 10% target for renewable
energy in transport, in which biofuels are expected to play an important role. Only
biofuels that meet certain sustainability criteria count towards this target. These
sustainability criteria primarily cover GHG emissions from the entire fuel chain, and
carbon stocks and biodiversity effects from direct LUC. The RED currently does not
contain explicit measures aimed at reducing unwanted indirect impacts. However, the
European Commission (EC) published a report in December 2010 in which it indicated
that it would, by July 2011, publish an Impact Assessment that would look into four
possible policy options to address indirect impacts of biofuels:

1 take no action for the moment while continuing to monitor;
2 increase the minimum greenhouse gas saving threshold for biofuels,
3 introduce additional sustainability requirements on certain categories of biofuels,
4 attribute a quantity of greenhouse gas emissions to biofuels reflecting the
estimated indirect land-use impact.

The EC might, in its Impact Assessment, indicate which policy option it believes is
most appropriate and might publish a policy proposal accompanying the Impact
Assessment to amend the RED and FQD accordingly.

19 May 2011 57


3.6 Detailed analysis of mitigation initiative: Low Indirect Impacts Biofuels

3.6.1 Low Indirect Impacts Biofuels
The Low Indirect Impacts Biofuels (LIIB) initiative started in 2008 as a private sector
initiative coordinated by Ecofys in collaboration with NGOs such as WWF and
Conservation International and industrial parties such as Shell, BP and Neste Oil with
the overarching goal to:
Identify areas and/or production models that can be used for environmentally and
socially responsible energy crop cultivation, without causing unwanted
displacement effects. (Ecofys 2010)

What is now called LIIB was originally called Responsible Cultivation Areas (RCA).
Currently, the initial RCA initiative is being developed into a Certification Module for
Low Indirect Impact Biofuels – in a consortium including WWF, Roundtable on
Sustainable Biofuels, Ecofys, DNV and several pilot organisations, funded by the
Global Sustainable Biomass programme of NL Agency. The aim of this Certification
Module is allow policy makers, voluntary certification schemes, producers and other
stakeholders to credibly distinguish biofuels that were produced in a way that
minimizes the risk of unwanted indirect effects.

The central principle under the LIIB initiative is that of the need for additional
production. Indirect effects of additional energy crop production are the result of a
displacement of other productive functions of the land. For example, existing palm oil
production that was previously used for the food sector is now used for biodiesel
production. Displacement of existing production is therefore at the heart of the
concept of indirect effects. Preventing displacement, by realising additional production
instead of displacing existing production, is therefore at the heart of the solution to
minimise the risk of indirect effects.

19 May 2011 58


The current project that develops a Certification Module for Low Indirect Impact
Biofuels focuses on providing project-level solutions for producers that want to
minimise the risk of ILUC. Thereby, the focus is on three of the options introduced in
section 3.1.2:
1 Realising additional production without displacement by expanding production on
land without provisioning services;
2 Realising additional production without displacement by increasing land
productivity. Two sub-options are distinguished here:
a. Yield increases on existing biofuel feedstock farms/plantations such as
palm oil or rapeseed.
b. Integration models in which food and fuel production are combined in
such a way that biofuel feedstock is produced without displacing food
production.
3 Realising additional biofuel production through the use of previously unused
residues.
It is acknowledged that this list of solution types is not necessarily complete and that
other solution types could be added at a later stage.

Two concrete examples are included in the textboxes on the following pages. More
elaborate information on the potential for the LIIB production models and their main
barriers can be found in “Mitigating indirect impacts of biofuel production - Case
studies and Methodology” (Ecofys 2009a) and the three reports of the three RCA pilot
studies in Indonesia and Brazil (Conservation International 2010a, Conservation
International 2010b, WWF 2010). Other studies that show the potential and the
environmental and social benefits of the sustainable production models that fall under
LIIB include “Bioenergy and food production for local development in Brazil: inputs for
policy-making (Sparovek et al, 2010), and Smallholder Oil Palm Production Systems in
Indonesia: Lessons from the NESP Ophir Project (Jelsma et al, 2009).

The Certification Module for Low Indirect Impact Biofuels is currently being tested in
four pilot locations (Brazil, Indonesia, Mozambique and South Africa) and opportunities
for further piloting, with increased focus on EU countries, are being investigated.

3.6.2 Scope
The central concept of the LIIB initiative is to expand agricultural production for
biofuels without displacing other provisioning services of the land.42 This would
prevent all the potential consequences, such as effects on biodiversity or carbon
stocks, of such displacement. Unwanted effects on food consumption will also largely
be prevented as no food or feed production is displaced, thereby preventing shortages
in the food/feed sector. Provisions are also included that prevent the use of
agricultural land for biofuels in areas where such agricultural land is scarce.

42
Within the development of the RCA, but also in other developments, a debate has been raised on the
effectiveness of expanding on land without provisioning services or ‘idle land’. A more elaborate discussion
of this debate and its links to the RCA initiative are presented in Appendix B.

19 May 2011 59


The LIIB concept includes concrete incentives for producers to change their behaviour.
Producers are requested to cultivate their feedstock either on lands without
provisioning services (while also meeting biodiversity, carbon stocks and land right
criteria) or to increase the productivity of the land, e.g. by integrating food and fuel
production (see concrete examples in Boxes below).

The LIIB concept is feedstock neutral. This means that, in principle, all feedstocks
could meet the LIIB criteria.

3.7 Conclusion on current mitigation measures

Three main conclusions can be drawn on current mitigation measures for indirect
effects of biofuel production:
• The amount of mitigation measures that currently exists is small. In addition, of
this small amount of measures most are not yet fully operational.
• Most of the mitigation measures focus only on GHG effects of biofuels by
incorporating an ILUC factor in the general life cycle analysis of feedstock-based
biofuel pathways. This has the inherent limitation that there is no incentive for
options to mitigate indirect effects at the project level, given a certain feedstock.
• The LIIB initiative is the only initiative to the authors’ knowledge to work on
pragmatic solutions for biofuel feedstock production that has a minimised risk of
indirect effects by preventing displacement effects from occurring at the project
level. The concept is currently being developed into an operational Certification
Module for Low Indirect Impact Biofuels, which policy makers and other
stakeholders could use to distinguish biofuel produced in a way that minimises the
risk of unwanted indirect effects.

19 May 2011 60


Box 3: Options for energy crop production without displacement from the LIIB initiative (I/II)

Expanding oil palm production on “unused land”

• Expanding production without ILUC: Casson (2007) describes how carbon emissions
from the oil palm sector can be reduced by redirecting oil palm expansion away from
forested areas and peat lands to degraded lands. Planting oil palm on Imperata
Grassland could lead to an increase in carbon stocks as well.

• Potential: Casson (2007) cites numbers on degraded land from the Indonesian Ministry
of Forestry, which has classified over 23 million hectare as degraded land. Garrity et al.
(1997) estimated the total area of Imperata Grassland in Asia at 35 million hectare
(8.5 million hectare in Indonesia). This compares to roughly 10 million hectare of
globally harvested oil palm plantations today.

• Risks: Not all degraded land will be available. Some of it will not be suitable for oil
palm production. Furthermore, degradation is often caused by the presence of people
and degraded areas are therefore often populated and the local population may be
occupying some of the lands. In addition, degraded land can still contain high
conservation values.

• Economic viability: Generally feasible. Some additional costs in the case of Imperata
Grassland for herbicides treatment in the early years of establishment. Fairhurst et. al.
(2009) find that Oil Palm plantations on grasslands are more profitable than
plantations on secondary forest.

• Added value from carbon benefits: Ecofys (2007b) finds that the GHG-performance of
biofuel from oil palm can be significantly improved if plantations are established on
Imperata Grassland. This could lead to a higher economic value as mechanisms such
as the EU Renewable Energy Directive and EU Fuel Quality Directive reward higher
GHG savings.

19 May 2011 61


Box 4: Options for energy crop production without displacement from the LIIB initiative (II/II)

Integration of sugarcane and cattle

• Expanding production without ILUC: Sparovek et al. (2007 and 2010) present an
integrated sugarcane and cattle production model in which hydrolysed bagasse is used
as animal feed. The additional feed would allow for more cows per hectare, freeing up
part of the pasture land for sugarcane. As a result the same land that used to support a
certain number of cattle, now supports the same amount of cattle while also producing
ethanol from sugarcane. In other words, sugarcane production is expanded on pasture
areas without displacing the original cattle production. This could reduce the migration
of ranchers to remote areas in the Cerrado and the Amazon region.

• Potential: The authors do not give estimates for the total potential. Not all pasture land
will be suitable for sugarcane. Total permanent meadow and pastures, both natural and
cultivated, in South America amount to over 450 million hectare, with 200 million
hectare in Brazil (FAO 2009). Total sugarcane area equals 8 million hectare (6.7 in
Brazil), suggesting a significant potential for the integration model. Also the RCA pilot
study on sugarcane-cattle integration in Brazil finds a significant potential (CI 2010b).

• Risks: The integration model requires close interaction between two very different
sectors. Diverting part of the bagasse from electricity generation to animal feed has
only a minimal impact on the direct emissions of the sugarcane to ethanol chain (<1%
reduction in the GHG-savings compared to fossil fuels.)

• Economic viability: The authors state that the model is feasible at current market
conditions.

• Added value from carbon benefits: Policies to promote GHG-savings through biofuels in
the EU and US are expected to include emissions from ILUC in the near to medium
future. Projects that can demonstrate to prevent ILUC, such as the integration model,
would then be recognised to achieve higher GHG-savings and may therefore obtain a
higher value.

19 May 2011 62


Reference sources

CARB, 2009. California’s Low Carbon Fuel Standard (LCFS) Final Statement of
Reasons, December 2009

Casson A, Tacconi L, and Deddy K, 2007. Strategies to reduce carbon emissions from
the oil palm sector in Indonesia. Paper prepared for the Indonesian Forest Climate
Alliance, Jakarta. 2007.

Conservation International, 2010a. Identification of Responsible Cultivation Areas in
Pará Brazil – pilot report. A report prepared by CI with support from Ecofys.

Conservation International, 2010b. Identification of Responsible Cultivation Areas in
São Paolo Brazil – pilot report. A report prepared by CI with support from Ecofys.

Croezen H, Brouwer F, 2008. Estimating indirect land use impacts from by-products
utilization. Contribution to the Gallagher Review of the indirect effects of biofuel
production.

Dehue B, 2006. Palm Oil and its by-products as a renewable energy source, potential,
sustainability and governance. July 2006.

Ecofys, 2009a. Mitigating indirect impacts of biofuel production - Case studies and
Methodology. Report prepared for the UK Renewable Fuels Agency.

Ecofys, 2009b. Summary of approaches to accounting for indirect impacts. Report
prepared for VROM.

Ecofys, 2010. Responsible Cultivation Areas - Identification and certification of
feedstock production with a low risk of indirect effects.

EPA, 2010. Renewable Fuel Standard Program (RFS2) Final Rule, March 2010
ICONE, An Allocation Methodology to Assess GHG Emissions Associated with Land Use
Change, Final Report, September 2010

HGCA, 2007. Project report 417: Optimising nitrogen applications for wheat grown for
the biofuels market. Available at:
http://www.hgca.com/cms_publications.output/2/2/Publications/Publication/Optimisin
g%20nitrogen%20applications%20for%20wheat%20grown%20for%20the%20biofuel
s%20market.mspx?fn=show&pubcon=3958

HGCA, 2009. Nitrogen for winter wheat – management guidelines. Available at:
http://www.hgca.com/document.aspx?fn=load&media_id=5684&publicationId=7231

19 May 2011 63


IFPRI, 2010. Global trade and environmental impact study of the EU biofuels mandate,
March 2010

Jelsma I, Giller K, Fairhurst T, 2009. Smallholder Oil Palm Production Systems in
Indonesia: Lessons from the NESP Ophir Project.

JRC, 2010. Biofuels: a New Methodology to Estimate GHG Emissions from Global Land
Use Change - A methodology involving spatial allocation of agricultural land demand
and estimation of CO2 and N2O emissions.

Lywood, W., Pinkney, J. & S. Cockerill, 2009a. The relative contributions of changes in
yield and land area to increasing crop output. In Global Climate Biology Bioenergy 1:
360-369 Lywood, W., Pinkney, J. & S. Cockerill, 2009b. Impact of protein concentrate
coproducts on net land requirement for European biofuel production. In Global Climate
Biology Bioenergy 1: 346-359

Öko-Institut and IFEU, 2009. The ILUC Factor contained in Sustainable Bioenergy:
Current Status and Outlook, March 2009

PBL, 2010. Rethinking Global Biodiversity Strategies: Exploring structural changes in
production and consumption to reduce biodiversity loss.. A report by the Netherlands
Environmental Assessment Agency (PBL).

Sarvision, 2011. Impact of oil palm plantations on peatland conversion in Sarawak
2005-2010.

Searchinger et al., 2008. Use of U.S. Croplands for Biofuels Increases Greenhouse
Gases Through Emissions from Land Use Change, Sciencexpress, February 2008

Sparovek G, Berndes G, Egeskog A, Mazzaro de Freitas F L, Gustafsson S, Hansson J,
2007. Sugarcane ethanol production in Brazil: an expansion model sensitive to
socioeconomic and environmental concern. September 2007.

Sparovek G, Berndes G, Barretto A, Martins SP, Maule R, Burgi R, Smorigo J, 2010. Bioenergy
and food production for local development in Brazil: inputs for policy-making.

University of Leicester, 2010. Carbon Emissions from Drained Tropical Peatlands. Contribution to
ILUC workshop hosted by the EC’s Joint Research Center, “Critical issues in estimating ILUC
emissions", 9-10 November 2010, Ispra (Italy).

University of Cambridge, 2006. Update of “CAP Reform: Arable Decoupling” using data from the
Special Study Cereal Production (2004/05) and the Farm Business Survey (2004/05). Final
Report for the Department for Environment Food and Rural Affairs. September 2006.

19 May 2011 64


USDA, Foreign Agricultural Services, 2010. Production, Supply and Distribution Online. Accessed
in November 2010.

WWF, 2010. RCA pilot reports, consisting of three parts:
• Budiman, A.; Smit, H. (2010). Identification of Responsible Cultivation Areas in
West Kalimantan Indonesia - Phase I: Preliminary Assessment. A report prepared
by WWW with support by Ecofys

• Smit, H.; Budiman, A.; Yaya, A.(2010). Identification of Responsible Cultivation
Areas in West Kalimantan Indonesia - Phase II: Desk-based analysis. A report
prepared by WWW with support by Ecofys

• Smit, H.; Budiman, A.; Yaya, A. (2010). Identification of Responsible Cultivation
Areas in West Kalimantan Indonesia - Phase III: Field verification. A report
prepared by WWW with support by Ecofys

19 May 2011 65


Appendix A Emissions from additional fertilizer

E4tech (2010) states that the increased emissions from applying additional N-fertilizer
to wheat, in order to increase the wheat yields and thereby expand production without
expanding the land, amount to 8.9 gCO2/MJ ethanol. Using the same assumptions we
find a value of 12.9 gCO2/MJ ethanol. Using yield response numbers from a previous
report from the same organization that E4tech uses for the yield response figure given
in their report, The UK HGCA, we even find a number of 106gCO2/MJ. The number of
8.9 gCO2/MJ ethanol therefore appears to be the result of a combination of 1) an
optimistic assumption on wheat yield responses to higher N-fertiliser inputs and 2) an
optimistic assumption in the allocation of the emission to the additional wheat-ethanol
produced.
• E4tech (2010) states on p28 that it assumes UK N-fertiliser application rates are
183 kg N / ha with corresponding wheat yields of 7.76 ton / ha. This translates
into an average N application of 23 kg N / ton wheat. It is furthermore assumed
that “a 10% increase in fertilizer input would lead to an increase in yield of 0.5 t
wheat / ha.” This amounts to 36 kg N per ton of additional wheat, an increase of
13 kg N per ton of wheat compared to the current average number. This number
was used for illustrative purposes only but it is unclear where this number is
based on. The same section shows a graph from HGCA (2009) on the relationship
between N supply and grain yield but numbers on the response of grain yields to
additional N fertilizer cannot be directly derived from that figure as it only gives N-
numbers relative to the optimal N supply, not relative to the absolute N supply.
Based on a figure in a report by the same organization (HGCA, 2007) a rough
number can be estimated, at around 130 kg N per ton of additional wheat. This is
an increase of 107 kg N per ton of additional wheat compared to the current
average number 0f 23 kg N per ton of wheat: a difference of a factor 8 with the
E4tech (2010) assumption.
• E4tech (2010) states on p 29 that the additional N-fertilizer, of 36 kg N per ton of
additional wheat, causes a total increase in fertilizer related emissions of 8.9
gCO2/MJ ethanol. We calculate the number to be 12.9 gCO2/MJ ethanol, see table
below. The difference is caused by the fact that E4tech allocates the additional
emissions from additional N-fertiliser to all EU wheat-ethanol production, and we
allocate it purely to the wheat-ethanol resulting from demand-induced yield
increases – after all, the wheat produced on additional wheat land will have its
own emissions from fertilizer application. This effect is partly compensated by the
fact that E4tech does not allocate any of the emissions to DDGS, while we allocate
40% of the emissions to the additional DDGS produced as a result of the yield
increase (allocation based on energy content).

We therefore find that if the UK would increase its yield purely through the use of
additional N-fertilizer, the emission from additional N-fertilizer application would

19 May 2011 66


amount to 106 gCO2/MJ ethanol, not 8.9 gCO2/MJ ethanol43. While this number surely
is an overestimate of the actual emissions from the additional UK wheat yield
increase, as not all yield increase will come from increased N-fertilizer application, it
shows these emissions are more important than suggested by the number of 8.9 g
CO2/MJ ethanol in E4tech (2010).

Table A - 1 Emissions from additional fertilizer input. Recalculation of the numbers cited by
E4tech (2010). Note that the last row shows the maximum additional emissions
associated with additional fertilizer to produce the additional wheat for biofuels.
This is a maximum figure as it assumes that all yield increase stems from
additional N-fertilizer application. Also note that this number does not include the
business as usual average fertilizer application per ton of wheat as such emissions
are already included in the LCA of the direct emissions of biofuel pathways.
Therefore the values in the last row are the result of penultimate row minus
average value of 23 kg N per ton of wheat.
The calculations are based on the same emission factors as used in E4tech (2010)
and used default conversion efficiencies and allocation to co-products from
Biograce.
+10% N +10% N
Current practice E4tech (2010) HGCA (2007)
as assumed in numbers on yield numbers on yield
E4tech (2010) response response
N-application (kg N/ha) 183 201.3 248
Yield (t wheat/ha) 7.76 8.26 8.26
kg N / t wheat - average 23.6 24.4 30.0
kg N / t wheat - marginal 36.6 130.0
additional kg N / t wheat for marginal wheat compared
to average wheat 13.0 106.4
kg CO2 emissions / kg N
(fertilizer production and soil emissions) 12.9 12.9 12.9
g CO2 from N / kg wheat - average 304 314 387
g CO2 from N / kg wheat - marginal 472 1677
MJ of ethanol / kg wheat (from Biograce) 7.8 7.8 7.8
allocation to DDGS (from Biograce) 40% 40% 40%
allocation to ethanol (from Biograce) 60% 60% 60%
g CO2 from N / MJ ethanol - average 23 24 30
g CO2 from N / MJ ethanol - marginal 36 129
g CO2 from N / MJ ethanol - marginal less average 12.9 106

43
Note that the correct number here is the additional fertilizer needed to produce one additional ton of
wheat in which all the additional fertilizer is allocated to the increase in wheat yield, see table A-1 for
calculations.

19 May 2011 67


Appendix B The debate on the effectiveness of expanding
on ‘idle land’ and its links to the RCA initiative

This appendix serves as an extension to the information presented on expanding on
land without provisioning services in section 3.6.

Discussions with experts and stakeholders have primarily raised concerns on the
effectiveness of using land without other provisioning services. Some parties claim
that agricultural land is scarce and will become more scarce in the future and
therefore should not be used for biofuels at all. In other words, while expanding
production in these areas may not cause displacement effects today, there may be
displacement effects in the future – as the land would otherwise have been taken into
production for food in the future.

The validity of this argument depends strongly on the future land requirements for
food, feed and fibre production on the one hand and the availability of agricultural land
on the other hand. Both are subject to large uncertainties. On the positive side, the
doubling in world food production in the past decades was met almost entirely by
agricultural intensification, with only 10-15% of the increase in production coming
from an expansion in cropland. With a slowdown in population growth, future growth
rates in food demand are expected to decline. In terms of land availability, a recent
study by IIASA shows that several hundreds of millions of hectare of land suitable for
rain-fed biofuel crop production exist that are not used for cropland today and are not
under forest cover or in protected areas. On the negative side, yields may not grow as
strong as they did in the past and climate change may have negative impacts. On the
large potential land availability found by studies such as (IIASA 2009), large
uncertainties exist on what these lands are actually used for today and to what extent
these areas can be taken into agricultural production. Further analysis of this topic is
beyond the scope of this study.

The coordinators of the initiative state that the solution is primarily meant as an
intermediate solution, initially aimed at the period up to 2020/2022 – the period for
which the EU and the US have set biofuel mandates, until global efforts to control
unwanted direct LUC are effectively implemented, thereby eliminating unwanted
indirect LUC, and that the risk of structural land shortages in the medium term is
small. In addition, the coordinators state that today’s biofuel energy crop feedstocks
can switch easily between food and fuel markets and therefore using areas without
current provisioning services for biofuels does not pose irreversible risks – the crops
and the areas on which they are cultivated could be reverted to food market relatively
easily.

19 May 2011 68


110530 ecofys gbep iluc report with e4tech

More Related Content

What's hot

Viewers also liked

Similar to 110530 ecofys gbep iluc report with e4tech

More from ralphbrieskorn

Recently uploaded

110530 ecofys gbep iluc report with e4tech