Energy 76 (2014) 336e344

Contents lists available at ScienceDirect

Energy
journal homepage: www.elsevier.com/locate/energy

Marginal abatement cost curves and abatement strategies: Taking

option interdependency and investments unrelated to climate change
into account
F. Levihn*, C. Nuur, S. Laestadius
€gen 30, Stockholm SE-100 44, Sweden
Industrial Engineering and Management, Royal Institute of Technology (KTH), Lindstedsva

a r t i c l e i n f o a b s t r a c t

Article history: Firms usually have optimization tools for evaluating various investment options; policymakers likewise
Received 4 March 2014 need tools for designing economically efficient policies. One such tool is the MACC (marginal abatement
Received in revised form cost curve), used to capture the least-cost sequence of abatement options. Such curves are also used for
2 August 2014
understanding the implications of government policies for markets and firms. This article explores dy-
Accepted 6 August 2014
namic path-dependent aspects of the Stockholm district heating system case, in which the performance
Available online 6 September 2014
of some discrete options is conditioned by others. In addition, it proposes adding a feedback loop to
handle option redundancy when implementing a sequence of options. Furthermore, in an energy system,
Keywords:
MACC
actions unrelated to climate change abatement might likewise affect the performance of abatement
CO2 abatement options. This is discussed together with implications for climate change policy and corporate investment
Investment optimization optimization. Our results indicate that a systems approach coupled with a feedback loop could help
overcome some of the present methodological limitations.
© 2014 Elsevier Ltd. All rights reserved.

1. Introduction
The mechanisms used to mitigate climate change generally
entail decision-making regarding technology investment and
adopting a least-cost planning approach to sustain competiveness.
In the context of energy systems, this involves both interaction with
present systems and the performance of discrete options. The long
technological life of many energy-related investments does not
make analysis easier.
One important instrument for policymakers and industry when
studying possible abatement options is MACCs (marginal abate-
ment cost curve). MACCs are used to analyse investments and the
impact of policy measures in order to find the least-cost options to
achieve a certain target, such as climate change mitigation. MACCs
are also used to analyse market reactions to economic policy, such
as the effects of the Kyoto Protocol, and to assess policies, such as
the Europe 2020 strategy [23,24,34].
This paper aims to develop the methodology of expert-based
MACCs for least-cost investment planning in circumstances in
which the options are not only interdependent but also dependent
on developments external to the climate change discourse. While
the existing literature on MACCs discusses the influence of path-
* Corresponding author.
E-mail addresses: fabian.levihn@indek.kth.se, levihn@kth.se (F. Levihn).
dependency and local contexts see, e.g., [5,9,34], it does not take
interdependencies fully into account [22]. claim that it is unclear to
what extent interactions between abatement options are included
in present MACCs based on discrete options, and recommend
applying a systems approach. All else being equal, major events,
which are sometimes external to the climate change mitigation
discourse, that condition and influence the properties of the op-
tions considered remain relevant. The magnitude of such effects
would vary between contexts. In larger energy systems, such as the
Nordic or European power markets, changes related to individual
power plants would have a smaller effect than would similar
changes in a DH (district heating) network. On the other hand, the
aggregated changes that are often the focus of broader MACCs [30]
provide conditions in which interdependencies exist. This research
gap will be addressed here.
The contribution of this article lies in exploring the dynamic,
path-dependent aspects whereby certain abatement alternatives
condition others, including the effects of option sequence and
major events not traditionally included in MACC analyses.
To provide an illustrative example, we have chosen to analyse
planned developments and possible abatement options for the south/
central Stockholm DH network. This energy system is small enough to
be predictable but large enough to enable analysis and discussion to
reach general conclusions valid for other energy systems.

http://dx.doi.org/10.1016/j.energy.2014.08.025
0360-5442/© 2014 Elsevier Ltd. All rights reserved.
 F. Levihn et al. / Energy 76 (2014) 336e344
2. Marginal abatement cost curves
At a national level, abating GHG (greenhouse gas) emissions at
the lowest cost to society is among the most important tasks when
designing climate change policy [6]. In the same way, a profit-
maximizing firm aims to find the lowest-cost GHG abatement op-
tions [5]. This task includes many possible strategies with regard to
installed capacity and technology, as well as changes in production
processes.
Recently increased interest in MACCs stems largely from reports
from the consultancy firm McKinsey & Company [23]. In the USA
and the European Union, these curves have been important tools
for assessing the costs of emission reductions, supporting recom-
mendations by many NGOs, such as the World Bank. At the Euro-
pean level, MACCs are used to estimate the price of a certain
amount of emission allowances within the EU ETS (European Union
Emission Trading Scheme). [13] discuss an example that treats the
possibility of including road transport in the EU ETS to meet
abatement targets. Another use of MACCs is to analyse what
abatement options are possible given a specific allowance price
[37]. The model is also used for more local assessments, such as
developments in single countries, industries, and technologies see,
e.g., [2,26,36,38,42,45]. Studies analysing the MAC (marginal
abatement cost), but without using a curve, are also common, for
example, the cost analysis of carbon capture under different policy
scenarios [35] and the optimization of NGCC (natural gas combined
cycle) power plants [4]. Outside Europe, MACCs have been used, for
example, in research into optimizing the Chinese cement industry
[44] and evaluating climate change policy in Brazil [33]. The curves
have also been applied to the abatement of emissions other than
CO2, such as SOx emissions in the USA [41].
Although MACCs might be regarded as recent tools used to
appraise climate change mitigation, their application started in the
1970s when the global economy was hit by the oil crises. Referred
to as savings curves or CSCs (conservation supply curves), their goal
was initially to evaluate energy efficiency and provide options for
improvements at both the plant and policy levels see, e.g., [32,39].
The model is also used in this way today, for example, in the
research of [16] and [7].
In production theory, the explanation behind the curves is
centred on the fact that if part of a process is viewed as inappro-
priate, the curves represent the marginal loss in profit of changed
production, or the MC (marginal cost) of achieving a target [24]. As
investments may provide a positive financial return, a negative MC
is also common. In essence, devising and analysing MACs provide
insights into how to achieve a target through possible options, as
they correspond to the additional costs or benefits of possible ac-
tions to fulfil a unit of the target.
At the plant level, MACCs can be used to link the individual
plant's emission levels to the cost of additional emission abate-
ment, i.e., MA (marginal abatement), relating to specific actions and
technologies. In a CO2 context, MACCs are usually illustrated with
tonnes of avoided CO2 emissions on the x-axis and costs per
reduced tonne of CO2 on the y-axis [9,23]. The total abatement cost
of a set target then corresponds to the total area under the MACC
from zero to that target on the x-axis, so the total cost depends on
both the reduction target and the shape of the MACC [9,34].
A MACC could be created using either an expert or model-based
approach depending on the dataset available [19,23]; however, it is
worth noting that other classifications might also be used see, e.g.,
[10,24,43].
Of the model-based approaches, the topedown approach is
usually a macroeconomic approach, generally implemented using
equilibrium models [9,24,34], such as the EPPA (Emission Predic-
tion and Policy Analysis) model developed at MIT, and other energy
337
system models analysing the relationship between environmental
policy and the impact of technical change [3]. The topedown
approach is concerned with how markets respond to exogenous
pressures, such as an assumed or pending policy intervention and
its implications for a system. In this context, MACCs could cover the
economy as a whole [20]. Analysing the exogenous mechanisms
that influence the system would allow abatement costs to be
calculated from a welfare perspective [9]. From this perspective, it
has been argued that all taxes and subsidies should be excluded
from MAC analyses, as they are only transferred between groups
within society [23]. In addition, the topedown approach could be
used at the industry level, where it would often include the detailed
and exhaustive analysis of issues such as policy tilt pricing and
production costs [24]. In the other model-based approach, the
bottomeup approach, the marginal costs are derived from energy
system optimization models [9]. A recent example of this approach
applied in the CSC context is that of [15].
The other dominant approach is the expert-based approach,
which is underpinned by an engineering or expert mindset incor-
porating the detailed analysis of various discrete options. [10,43],
and [41] categorize this as a type of bottomeup approach. The
difference between these authors and [23] and [_Taylor_2012]
seems to lie in their categorization between expert-based as a
bottom-up approach on the one hand, and bottom-up as a model
based on the other. All the relevant literature, however, agrees on
the definition of the expert-based approach, which is generally
based on individual estimations of particular alternatives. The
approach seeks to find the best possible (i.e., lowest cost) options
for achieving a target within a given context. From a corporate
perspective, expert-based MACCs answer the questions of how the
market will or should adapt to policy and what the best available
options are for future investments and actions. The expert-based
approach has been the most common approach in MACC analysis
[23]. If a firm is a price taker, prices might be treated as fixed and
could thus mirror the actual price scenario a firm expects under this
approach.
The local context of technological lock-ins, path dependency,
and other barriers could affect a firm's decisions [1]. It may turn out
that the paths entered into by firms are incompatible with the rest
of the system [5]. An important aspect of MAC analysis is that
structural factors might have a huge impact, and the results of
similar abatement options might vary between different regional
and local contexts. Initially, a low market price for energy relative to
the introduced cost of CO2 emissions would mean increased effects
of the abatement policy [9].
The supply structure of the energy industry affects the marginal
cost of adopting a substitute. The costs of abatement options are
significantly lower in an energy system where direct substitutions
of natural gas for coal are possible, compared with a system, for
example, in which such a switch has already taken place.
The original CSC literature took option interaction into account
through an iterative process wherein the lowest-cost option would
Fig. 1. The iterative methodology of the original CSC approach in a MAC optimization
context.
338 F. Levihn et al. / Energy 76 (2014) 336e344
always be chosen (see Fig. 1). In every step, the costs of all available
options were recalculated and the new lowest-cost option was
added to the curve. The aim of the curve was then to find the
lowest-cost sequence of investments in energy conservation in
relation to investments in new electric power production capacity
[32]. In its most basic form, the iterations were simplified to esti-
mate the impact of earlier options when assessing the cost of later
options when there was obvious option interdependency [7].
implemented this as a boundary to rule out impossible combina-
tions of energy conservation measures for the German steel
industry.
Although MACCs are a key tool in environmental economics,
their analytical properties are rarely researched outside this area
[31]. Top-down MACCs are often derived from complex economic
models used to predict the emissions and costs of various policies,
so they represent a compromise and are not as good as empirically
estimated relationships [12]. Research in this field is increasing, but
detailed studies of specific technologies and their associated costs
are usually confidential or unpublished [5]. Although MACCs are
frequently used, their usability depends on the assumptions un-
derlying them: one curve might be suitable for answering some
questions, but unsuitable for others. We still lack in-depth, critical
analyses of MACCs as a tool in the academic world [23].
Some such analyses do, however, exist. [46] discuss the ranking
of options with negative marginal cost, for example, options with a
positive financial return. [10] examine the robustness of MACCs in
relation to the combination of economic policy, fuel prices, and
load. Other examples include [34]; who evaluate the use of MACCs
in relation to welfare costs, and [40]; who discuss how MACCs
could be used in designing abatement strategies. Research into
MACCs' relative path dependency has also used the concept to
describe whether a certain MACC is dependent on previous eco-
nomic policy [34].
Expert-based studies of MACCs have been criticized for many
reasons, two of which are that they do not take systemic effects into
account and that they rely on the interaction of different mitigation
measures [23]. In addition, there is currently no consensus on how
technical change affects MACCs. This has been highlighted by [3];
who argue that empirical evidence of how technical change affects
MAC would be beneficial for technologically detailed analytical
models.
2.1. A proposed approach to improving MACC analysis
To address some of the problems of using bottomeup/expert-
based CSCs/MACCs in least-cost planning, a new approach based on
a systems perspective is proposed here. We do this by building on
the iterative approach of [32] and the constraints used by Ref. [7] to
check for incompatible options.
The proposed approach starts from the assumption that each
selected option forms part of a new system of which the other
approaches are part. As such, each selected option both changes the
hosting system and potentially contributes to path dependency, as
all other options are affected by the option adopted. Such a view
necessitates, in addition to an iterative approach, the inclusion of a
feedback loop to check whether earlier options are affected in such
a way that they become redundant or uneconomic.
The dynamic transformation of energy systems is not only
dependent on climate change abatement and energy conservation
options; in addition, options and events that increase energy con-
sumption or GHG emissions could become part of evolving systems
and thus affect the performance of abatement options. If such
changes are part of a potential scenario, or of the known or ex-
pected transformation of a system, it is possible to include these in
a CSC/MACC. Of course, this raises the question of option timing, for
example, in the case that one option could severely alter system
performance if adopted after certain other options [29].
3. The Stockholm district heating system
In Stockholm, Sweden's capital, the DH network, constructed in
1953, is currently the main supplier of heat and hot water. The
system consists of two larger networks, the south/central DH sys-
tem being the most relevant from a climate change abatement
perspective. In the south/central DH system, on average, more than
9700 GWhth (the lower-case “th” indicates thermal energy) of heat
is generated annually. Furthermore, more than 1800 GWhe (the
lower-case “e” indicates electric power) of electricity is generated
from the CHP (combined heat and power) units in the network.
Within the network, the utility company, Fortum Va €rme samma €gt
med Stockholms stad AB (hereafter “Fortum”), is the main actor and
in practice has a monopoly, although some smaller utilities are also
present.
On the production side, the south/central DH system is based on
a variety of fuels and production technologies (see Fig. 2). Peak
production excluded, the DH system consists mainly of CHP pro-
duction and large-scale electricity-driven HPs (heat pumps), which
create interesting dynamics if the energy balance of base- and
middle-load production is altered, for example, in the case of new
base-load production CHP, which is currently being expanded in
the network. A new waste incineration CHP unit would, for
example, create additional production, at the same time as all other
units behind it in merit order would be used less. As a result, the
generation of electricity and heat from all CHP plants behind it in
merit order would decrease. On the other hand, the large share of
HPs would also reduce production, directly lowering electric power
consumption. The peak-load capacity, based mainly on fossil- and
bio-oil HOBs (heat-only boilers), would also be used less.
The CHP units in the system use a wide range of fuels and
processes, including the combustion of biomass, fossil oil, coal, and
waste. The coal CHP plant, KVV6, located at Va €rtaverket, is notable,
as it is based on the highly efficient PFBC (pressurized fluidized bed
combustion) technology and produces a large amount of electricity
relative to the heat generated. PFBC is regarded as a “clean coal”
technology [14], and the plant operates at a power-to-heat ratio (a)
of 0.58, which is markedly high and as such is one of the benefits of
the technology. Each year, more than 1700 GWhth of heat and 800
GWhe of electricity are generated by this plant. To meet GHG
reduction goals, measures to co-fire biomass are now being
developed at KVV6 with the aim of reaching a 30% share by 2016.
KVV6, built as a fully functioning demonstration plant for PFBC
technology by the firm ABB and becoming operational in the early
1990s, has been the object of much research throughout its oper-
ational life. In recent years, online pilot experiments with CCS
(carbon capture and storage) technology have been conducted at
the plant [8,17] for demonstration purposes.
The installed capacity of HPs equals approximately 550 MWth.
During a normal year, approximately 770 GWhe of electricity is
used to generate 2200 GWhth of heat. Electricity is also used in
other units, and the network consumes a total of approximately
1100 GWhe electricity per year. The fossil-oil-based peak-load HOB
capacity equals approximately 1500 MWth, but is only responsible
for approximately 80 GWhth of heat generation each year.
4. Methodology
The analysis was performed using a combination of empirically
established operational data and calculations with Optima invest-
ment optimization and analysis software. The study was conducted
from 2011 to 2014 in cooperation with Fortum, the major actor in
 F. Levihn et al. / Energy 76 (2014) 336e344
Fig. 2. Duration chart of heat produced, energy balance, and merit order of the south/central Stockholm DH network, 2009.
the network, which made comprehensive operational data avail-
able to us. Optima software was developed in 2004 by Fortum to aid
investment decision making and simulations. Fuel and policy-
related prices were based on the average prices for the
2010e2030 period, based on the International Energy Agency's
(IEA's) World Energy Outlook 2010 “new policies scenario” [18] to
aid transparency and facilitate the publication of the results. The
use of the price scenario favoured by Fortum would have made the
results more relevant to the affected decision makers, but would
also have resulted in the MACCs of this study becoming confiden-
tial. Other price scenarios could also be considered, but as this
study is methodological in nature, the use of the scenario is pref-
erable to forecasting.
4.1. Optima software
The Optima software has recently been used in research into
changes in DH production in Stockholm [28,29]. It is similar to the
HEATSPOT/EUROSPOT model see, e.g., [11,25], in that it uses an
annual load-duration curve to describe the system. Other, more
advanced software, such as MARKAL or customized linear pro-
gramming as used by [27]; could have been considered, but since
access to the Optima software together with technological and
operational data were granted by Fortum, it was the best option
given the purpose of this study. In addition, data on previously
investigated options and considered investments were available,
including lessons learnt from online experiments with CCS.
The software makes calculations by establishing an energy bal-
ance and merit order using annual variable production costs together
with technical parameters such as availability. As the program has
been in use since 2004, the parameters relating to different plants
have been updated in line with actual plant performance. The sim-
ulations include considerable technical detail with more than 800
parameters describing various plants and potential options in the
system. The system boundary for the calculations was production
connected to the south/central DH system in Stockholm.
339
All variable costs are attributed to heat generation, while in-
come from electric power generation and subsidies for renewable
production are deducted when applicable. Fixed costs are handled
separately and do not affect the energy balance or merit order,
which is a valid assumption for optimizing a DH network in which
all production-related expenses concern one corporation. The fixed
costs were added by the software when calculating total system
cost (TC). To calculate MC and MA from the simulation output, the
following two equations were used:
MCa ¼ TCa  TCb (1)
MAa ¼ Eb  Ea (2)
The MC of an option was calculated by subtracting the TC before
implementation (b) from the TC after implementation (a). MA was
calculated by subtracting the amount of emissions after imple-
mentation from the emissions before implementing the option. The
MAC was then calculated by dividing equation (1) by equation (2),
giving equation (3).
TCa  TCb
MACa ¼ (3)
Eb  Ea
All costs were calculated in Swedish Kronor (SEK). At the time of
writing, 1 EUR (Euro) equaled approximately SEK 9. The MC was
given in millions of SEK (MSEK), MA in thousands of metric tonnes
or kilotonnes (kt), and MAC in SEK per metric tonne (SEK/t). MC
could be either positive or negative, a positive value indicating
increased costs and a negative value decreased costs and thus a
positive financial return on the investment.
As discussed by [10]; the MACCs included in the present paper
are likely to display low robustness towards other price scenarios.
As a result, they indicate the abatement potential and costs corre-
sponding to the particular price scenario used. Understanding the
system behaviour under other price scenarios would necessitate
340 F. Levihn et al. / Energy 76 (2014) 336e344
further calculations while adding little value to the methodological
insights we seek to illustrate here.
Options such as the construction of new plants affect the MC
and MA in an energy system. The changed MC and MA could be
attributed to both the operation of the new plant and to the
changed operation of other plants in the system. Decomposition
analysis helps bring such details into the analysis [21]. Discussions
of what contributed to changed MC and MA, either directly by the
option itself or indirectly by the operation of other production
units, are treated in the “Results” section of this article.
4.2. Model application
To calculate MC and MA, a baseline was established with pro-
duction mirroring 2009 levels. To enable the analysis, MACCs were
calculated in three ways: the “straw man”, “iterative”, and “system”
approaches.
 The straw man approach simulated each option against the base
scenario of 2009. MC and MA were calculated for each option
using equations (1) and (2) for all possible options. The ones
with the lowest MAC were simply added, one at a time, until
total abatement was achieved. While such an approach should
never be undertaken when there are dynamics present between
different abatement options, MACCs calculated using this
approach are unfortunately not uncommon. As such, the straw
man provides the worst-case scenario for MACC analysis and the
choice of name refers to the analogy of “hitting on a straw man”.
 The traditional iterative approach was constructed as described
by [32]; where MC and MA were recalculated for all remaining
options every time the option with the lowest MAC was selected
(see Fig. 1). Thus, TCb and Eb were based on the previously
selected abatement option on which the next iterations were
performed, except for the first iteration, which was related to
the 2009 base scenario.
 The proposed system approach was based on the traditional
iterative approach but taking account of system development,
including major developments unrelated to climate change
abatement. The goal was to identify abatement options that
become redundant after implementing other options, and
abatement options whose abilities alter significantly when new
options are applied. The implementation of earlier abatement
options was analysed when a new option was added and thus
transformed the analysed system. Furthermore, major activities,
such as the inclusion of new production with no effect on
reducing CO2 emissions, were included, as these might affect the
ensuing option selection and the performance of other abate-
ment options.
A MACC was constructed from each of these three calculations.
As this is a typical expert-based or engineering approach, the study
is limited to specifically analysing the properties of these particular
kinds of MACCs.
4.3. Abatement options
The abatement options considered here were options that had
been analysed by Fortum as possible future investments. Some of
these are part of the BPP (base production plan), which is the plan
for the long-term development of the network, including new
plants to be built, major modifications to present plants, and the
decommissioning of major plants. Table 1 shows the possible in-
vestments included in the analysis.
Of the options shown in Table 1, four were included in the BPP:
efficiency improvements of KVV6 through a new FGC (flue gas
condensation facility); co-firing 30% biomass with the coal in KVV6
(target to be achieved in 2016); a new biomass fluidized bed
combustion (FBC) CHP plant, KVV8 (under construction); and a
waste incineration CHP plant, KVV7 (planned for 2016). If a normal
lifespan of 40 years applies to the coal-based KVV6, it could be
expected that either the plant would need to be decommissioned
around 2030 or major investments would be necessary to prolong
its technical life.
Although the BPP exists, both the straw man and iterative ap-
proaches assumed that these options were not under construction
or decided on. In contrast, the systems approach took account of
these expected events at each step of the MACC, as each would
transform the system.
4.4. CO2 emissions accounting and co-generation
The question of how to calculate CO2 emissions in relation to
measures affecting plants operating in multi-energy-carrier sys-
tems is important. One such issue is how to allocate the effect of
changed power generation from CHP plants. This issue was thor-
oughly discussed recently for the Nordic context by [28]; who
emphasized the need to adopt different allocation methods
depending on the nature of the study. The present study has
included only the direct, local emissions of the potential decision
maker. It should be noted, however, that including the change in
global CO2 emissions could be important given a MACC with
different characteristics and a different purpose.
5. Results
To structure the analysis, the results of applying each of the
three MACC methodologies will be discussed individually. The re-
sults emphasize MC and MA because of the use of the IEA's “new
policies scenario” from 2010 [18]. Consequently, the actual results
in terms of absolute cost levels and abatement could change for
other price scenarios depending on the robustness of the MACCs.
Our goal is to illustrate how different results could be obtained
depending on which of the three methodologies is applied, which is
why the robustness of this MACC for other price scenarios is outside
the scope of this article.
5.1. Straw man MACC
Starting with the straw man, the MAC is calculated without
considering the interaction between discrete options. When the
coal PFBC plant, KVV6, is provided with FGC, it corresponds to the
lowest MAC option. This option contributes to a financial return
Table 1
List of considered abatement options.
Options Comments
KVV6 flue gas Commissioned in 2010
condensationa
KVV6 50% biomass 50% mixture of biomass in the fuel of KVV6
KVV6 30% biomassa 30% mixture of biomass in the fuel of KVV6
KVV6 heat only Convert KVV6 from CHP to HOB
KVV6 CCS Installing CCS technology at KVV6
KVV6 decommissioned KVV6 decommissioned without other
measures taken
KVV8 parallel KVV6a A new biomass FBC CHP unit at Va €rtaverket;
under construction
KVV8, KVV6 Decommissioning KVV6 when KVV8
decommissioned is commissioned
KVV7 Ho €gdalena New waste combustion CHP; planned
Biofuel conversion Converting peak HOBs from crude oil to biofuels
of fossil HOBs
a
Part of the base production plan.
 F. Levihn et al. / Energy 76 (2014) 336e344 341

Fig. 3. The straw man MACC: the curve is constructed using an approach in which
MAC is calculated for all options, after which the option with the lowest MAC is
selected.
through cost savings in the range of MSEK 80 annually under this
price scenario, but with a relatively limited MA at 8 kt annually. The
resulting MAC is approximately SEK e9500/t CO2. Of the remaining
options, biofuel conversion of fossil peak HOBs has the second
lowest MAC at SEK e2500/t CO2. The corresponding annual MA and
MC are approximately 20 kt and MSEK e40, respectively (Fig. 3).
The third least expensive abatement option is the new biofuel
FBC plant, KVV8. This plant reduces costs by approximately MSEK
140 and CO2 emissions by approximately 150 kt annually. This
reduction is largely because KVV8 will be ahead of the coal-fired
KVV6 and the fossil peak HOBs in merit order, which is why the
output of the fossil plants in the DH system is reduced.
The last option is fitting KVV6 with CCS. This option has an MC of
approximately MSEK 110 but reduces CO2 emissions by 700 kt
annually, which is a relatively large MA. If decomposed, this
reduction is attributable partly to the actual capturing of CO2 and
partly to the fact that KVV6 would be used less as other units move
ahead of it in merit order when production costs at the plant in-
crease. As a result, CCS would increase the use of biomass and
electric power in the production mix. The corresponding MAC
relating to CCS is approximately SEK 160/t CO2.
Combining the results of the sequence of four options to be
implemented in the straw man approach results in a cumulative
abatement of 878 kt of CO2 and MSEK 153 in cost savings annually,
which gives a corresponding MAC of SEK e175/t CO2.
5.2. Iterative MACC
Fig. 4. MACC constructed using an iterative process: the option with the lowest MAC
was selected for each iteration.
The effect of constructing KVV8 on CO2 abatement is reduced
when the interdependencies between options are taken into ac-
count. As efficiency is increased by installing FGC at KVV6, with
more kWhth generated at KVV6 per t CO2 and the fossil peak HOBs
already converted from fossil fuels, at approximately 150 kt annu-
ally, marginally less CO2 emissions would be reduced by con-
structing the KVV8. Furthermore, the accumulation of fixed costs
and variable operational cost savings for plants after KVV8 in merit
order also reduce the MC of constructing the plant to approxi-
mately MSEK e120 annually. In other words, the financial return of
using other plants less when a new plant is constructed is reduced,
as previous options applied have increased those plants' efficiency
within the climate change context. The total effect of constructing
KVV8 is a MAC increased to approximately SEK e790/t CO2. Table 2
compares the straw man and iterative approaches for these first
three options.
The fourth option differs altogether between the iterative and
straw man approaches. After KVV8 is completed, the coal-fired
plant KVV6 will be used less in production each year, making it
more economical to apply the option of co-firing 30% biomass in
the plant instead of implementing CCS. The MC for the option is
MSEK e13 and the MA 181 kt CO2, which results in a MAC of
SEK e74/t CO2.
To achieve the same levels of abatement as found using the
straw man approach with CCS, the remaining option with the
lowest MAC would be to decommission KVV6. After the other op-
tions are implemented, this would result in an additional MA of
Table 2
Comparison of three abatement options common to both the straw man and iter-
ative MACCs.

MC [SEK million] MA [kt CO2] MAC

If Meir's (1983) originally proposed iterative approach is instead [SEK/t CO2]
applied (Fig. 4), some differences are immediately evident. While
Straw man KVV6 FGC 79 8 9488
the installation of FGC is the same first option, the influence of
Biofuel 41 20 2067
option interdependency becomes evident for the remaining conversion
options. KVV8 136 149 911
After FGC is installed in KVV6, the heat production from the Total 255 177 1441
plant increases, reducing the need to use the fossil HOBs. Conse- Iterative KVV6 FGC 79 8 9488
quently, KVV6 is in production less, giving fewer production hours Difference* 0% 0% 0%
by which to divide fixed costs and in which to realize the benefits of Biofuel 20 18 1066
conversion
lowered variable costs. The result is a decreased MA of 2 kt annually
Difference* 105% 6% 94%
e a relatively small reduction, but equaling a 6% decrease in climate KVV8 117 148 791
change abatement potential for this option. Costs are affected more Difference* 17% 1% 15%
and the marginal cost reduction declines to MSEK e20 annually, Total 215 174 1235
equaling a 105% overestimation of the straw man MACC. The cor- Difference 19% 2% 17%

responding MAC is reduced to SEK e1066/kt CO2. * Difference between iterative and straw man calculations.
342 F. Levihn et al. / Energy 76 (2014) 336e344
489 kt CO2 annually. The cost would increase by approximately
MSEK 160 annually, which results in a corresponding MAC of SEK
334/t CO2 abatement.
The combined effect of the iterative MACC is a cumulative
reduction of 845 kt CO2 and MSEK 65 in reduced production costs
annually. In other words, the cost reductions calculated using the
straw man approach are overestimated by 135% when option in-
terdependencies are considered and KVV6 is decommissioned
rather than provided with CCS, which becomes uneconomical after
adopting the other options. The combined MAC of the sequence is
SEK e74/t CO2.
5.3. Systems MACC
For the last MACC approach (Fig. 5), we consider what the DH
energy system would be like when each option of the iterative
approach is implemented. While the iterative methodology of
Meier (1983) does not incorporate a feedback loop to consider the
eventual redundancy of previous adopted options, such redun-
dancy is evident after FGC and KVV8 are constructed. The result is a
reduction in demand for the biofuel-converted HOBs to the extent
that they would constitute only reserve capacity and not be used
under normal operating conditions, making that option redundant.
This is similar to the results of [10] with regard to the fuel switching
context, who demonstrated that once a technical maximum is
reached, the abatement potential of substituting one type of
existing production for another is exhausted. This redundancy is
managed by removing the option from the iterative process.
If the process is restarted with the biofuel conversion option
excluded, constructing the KVV8 biofuel CHP unit becomes the
option with the second-lowest MAC.
Versus when constructed after the biofuel conversion option, the
cost savings from constructing KVV8 first increase to approximately
MSEK 140 annually. The MA increases to approximately 160 kt CO2
annually, which is even higher than that achieved with the straw
man calculations. This is attributable partly to the fossil HOBs being
used less with KVV6, as these will be behind it in merit order, and
partly to larger CO2 emission reductions from KVV6, as this plant
plays a larger part in the DH production mix after implementing FGC.
The base production plan includes plans to construct an addi-
tional waste incineration CHP unit, KVV7. This plant would be con-
structed after KVV8 and marginally increase CO2 emissions.
Fig. 5. MACC constructed using the iterative approach with a feedback loop in which
the effect of a future abatement option on previously selected options is included in
the optimization. In addition, the construction of a plant that transforms the system
and thus affects the abatement option is included, although it is not an abatement
option in itself.
Illustrated in a MACC, this would result in an overlap in which some
of the gains from KVV8 in terms of reduced CO2 emissions are
countered if KVV7 is constructed. While the increase in MA is only
7 kt annually, this is comparable to the decrease from the FGC option.
On the other hand, the DH system production costs would be
reduced by MSEK 110 annually. The combined effect of a relatively
marginal increase in emissions and relatively large financial return
under this price scenario results in a low negative cost per increased
unit of emissions at SEK 16,040/t CO2 for constructing KVV7.
KVV7 is included in the analysis because waste incineration CHP
is high in merit order due to its negative variable operational costs
(e.g., the system is paid for waste management), which means that
KVV7 will affect the MA and MC of most other possible options.
After KVV7 is constructed, the reduced production from KVV6
reduces the cost savings of the 30% biomass co-firing option for
KVV6 to MSEK 12 annually. As KVV6 and other plants would be
used less, the MA decreases to 173 kt CO2 annually. The combined
effect is a MAC of SEK e68/ton CO2.
The last option is to decommission KVV6. One could claim that
this option is similar to the case of the biofuel conversion option
becoming redundant and that this would also make the 30% co-
firing option redundant. As this has already been exemplified and
the scenario in which KVV6 is used with 30% biomass co-firing until
the end of its technical lifespan is possible, it is reasonable to keep it
in the analysis. If abatement goals require that KVV6 be decom-
missioned in the near future, the 30% co-firing option should be
excluded.
The effect of closing KVV6, given the interdependencies with the
other options, including KVV7, is to reduce the MC to MSEK 143
annually, i.e., a 12% cost saving. This is attributable to the reduced
use of the plant versus in the iterative calculations, which is why
decommissioning has less effect on the overall DH system. Like-
wise, the MA is reduced to 445 kt CO2 annually with a corre-
sponding MAC of SEK 322/t CO2.
It would be interesting to compare the overall emission re-
ductions from the systems and iterative MACCs. While KVV7 is
planned, but outside the normal CO2 abatement discourse, it has a
large impact on the MAC within the option sequence. Given the
system interdependencies, the MA is reduced by 9%e777 t CO2
annually if KVV7 is constructed, while the financial return increases
threefold from an MC of MSEK e65 to 195 annually. Put differ-
ently, options external to the climate change discourse, such as
many cases of waste incineration CHP, can increase the financial
return of climate change abatement when considered as part of a
package of options. In this case, the MAC of the sequence of options
decreases by 69% to SEK e251/t CO2.
It should be noted that further abatement would require mea-
sures targeting waste incineration CHPs. Furthermore, to secure the
future supply of DH, it is advisable to build additional production
capacity if KVV6 is to be decommissioned. Moreover, it should be
noted that the impact of decommissioning CHP capacity in relation
to lost electric power production could increase CO2 emissions
given a multi-energy systems perspective [28].
5.4. Discussion and recommendations
The present exercise of comparing alternatives was undertaken
not only to demonstrate that the better approach is just better. For
an actor, such as an energy utility, that knows or expects the
transformation of a technological system, it is beneficial to include
this in MACC analysis. This is not currently done. In the case of the
DH in Stockholm and the utility firm Fortum, we have demon-
strated how such transformation could be included.
Table 3 lists the combined results of the sequence of options for
two approaches used here, assuming the information needed for
 F. Levihn et al. / Energy 76 (2014) 336e344
Table 3
Comparison of the total cost, total abatement, and combined MAC of the sequence of
abatement options of the iterative and systems approach.
MC [SEK million] MA [kt CO2] MAC [SEK/t CO2]
Iterative 65 845 74
Systems 195 777 251
the systems approach, i.e., that the biofuel conversion of the fossil-
oil peak boilers becomes redundant after a large biofueled CHP
plant is constructed. The plans to build a new production plant (i.e.,
waste incineration CHP) that results in a small increase in CO2
emissions, but large cost savings, should also be included in the
analysis as it affects the system of which all other options are part.
Given the traditional iterative approach, but without consid-
ering the above-mentioned two effects on the energy system, the
abatement potential of the sequence of abatement options would
be overestimated by 9%. This would also result in unnecessary in-
vestments in the conversion of the peak boilers. Together with not
considering the economic benefits of new waste incineration, the
financial return/cost savings would be underestimated by 67%.
If no interdependencies are taken into account the situation
would be even more problematic, as the resulting investments
would be in abatement options that are not financially sound when
interdependencies are accounted for. The cost and abatement po-
tential would also be greatly overestimated.
In general, we make the following recommendation: If changes
to a technological system that do not result in climate change
abatement are known or expected, or if there is a desire to know
the effects of such changes, these should be included in the anal-
ysis. In practice, this could be done using an overlay in which the
negative abatement options create an overlap in the MACC.
Furthermore, we need to understand the effects of later abatement
options on previously adopted options in order to rule out redun-
dant investments.
6. Conclusions
The aim of this article was to demonstrate how the CSC/MACC
model methodology could be developed to better account for op-
tion interdependency, including changes unrelated to progress to-
wards the analysed goal. These interdependencies between options
affect expert-based MACCs, as a sequence of abatement options
potentially influences the cost and abatement potential of each
included option, and of possible future options. In every step of
creating an expert-based MACC, path-dependent aspects are
created as the system of which the options are part evolves.
Particular options therefore condition or affect others. These as-
pects are coupled not only to previous economic policy, as dis-
cussed by [34]; but also e in the case of expert-based MACCs e to
each selected discrete option. The abatement and costs associated
with a particular abatement option depend on the other options
previously adopted.
Approaches in which options are selected based on the MAC
without iterative calculations should be avoided, as they risk
overestimating the abatement potential and underestimating the
costs. They could also result in the adoption of less efficient options.
The use of an iterative approach also potentially avoids the problem
of incompatible options as discussed by [7]; because the depen-
dence of future options on those previously adopted can be
captured together with other interdependencies, as system trans-
formation is accounted for.
In summary, to improve the methodology for creating expert-
based/bottomeup MACCs, some general implications of this study
are as follows:
343
1. There is a need to extend the iterative approach by adding a
feedback mechanism in which previously adopted options are
monitored for redundancy. This calls for a greater understand-
ing of the system of which the considered options are part, but
avoids potentially unnecessary costs.
2. Considering activities not related to climate change abatement
is important, as such activities affect the system of which the
abatement options are part and could affect the performance of
abatement options. Consequently, it is important to include
them in the MACC process, which might otherwise create
overlap between abatement options.
The inclusion of options and events (and their results) external
to climate change abatement in the MACC sequence also improves
the applicability of the model. In this article, the example illustrated
how waste incineration CHP could help reduces the total MAC by
increasing the financial return of the MACC sequence.
This alleviates some of the drawbacks of MACCs based on
discrete options identified by Ref. [22]. The application of a systems
approach is both possible and advisable. When considering in-
vestment options, an iterative approach with a feedback control
loop checking for interaction with previously adopted options re-
duces optimization inefficiencies. With regard to timing [40], and in
addition to the possible need to implement more expensive options
ahead of cheaper ones, options external to the climate change
discourse should also be considered if they affect the performance
of abatement options. This necessitates the inclusion of options
with small or negative effects on climate change. Of course this
depends on the nature of planned modifications to the system that
are not directly related to reaching the desired goal.
The present results are generalizable to other least-cost plan-
ning contexts, such as water conservation, in which CSC/MACC
evaluation models are used. Of course, this depends on what in-
formation is available or can be collected. Nevertheless, it is
important to know about the consequences of various approaches
and their potential effects on various future abatement/conserva-
tion scenarios.
For policymakers and corporate leaders, the article raises the
concern that understanding discrete actions is not enough, as the
sequence of actions and their interactions with surrounding sys-
tems are also important. It also raises the idea of considering op-
tions that do not reduce CO2 emissions, but only if these affect the
performance of the abatement sequence.
Acknowledgements
This study was financed by the Swedish Energy Agency
through the project Investments in energy efficiency and climate
change abatement: revising marginal cost curves as an optimiza-
tion model. Direct emissions and net power production simula-
tions were undertaken in cooperation with the district heating
€rme sama
utility Fortum Va €gt med Stockholms Stad AB. Special
thanks are extended to Ove Åsman and Anna Vidlund for their
help.
References
[1] Arthur WB. Competing technologies, increasing returns, and lock-in by his-
torical events. Econ J 1989;99:116e31.
[2] Aydin L, Acar M. Economic and environmental implications of Turkish
accession to the European Union: a CGE analysis. Energy Policy 2010;38:
7031e40.
[3] Baker E, Clarke L, Shittu E. Technical change and the marginal cost of abate-
ment. Energy Econ 2008;30:2799e816.
[4] Bernier E, Mare chal F, Samson R. Optimal greenhouse gas emissions in NGCC
plants integrating life cycle assessments. Energy 2012;37:639e48.
344 F. Levihn et al. / Energy 76 (2014) 336e344
[5] Beumont NJ, Tinch R. Abatement cost curves: a viable management tool for
the achievement of win-win waste reduction strategies. J Environ Manag
2004;71:207e15.
[6] Broberg T, Samakovlis E, Sjo €stro €
€ m M, Ostblom G. En samh€ allsekonomisk
granskning av Klimatberedningens handlingsplan fo €r svenska klimatpolitik.
National Institute of Economic Research in Sweden; 2008. Specialstudie No. 8.
[7] Brunke J-C, Biesl M. A plant-specific bottom-up approach for assessing the
cost-effective energy conservation potential and its ability to compensate
energy-related costs in the German iron and steel industry. Energy Policy
2014;67:431e46.
[8] Bryngelsson M, Westermark M. CO2 capture pilot test at a pressurized coal
fired CHP plant. Energy Procedia 2009;1:1403e10.
[9] Criqui P, Mima S, Viguier L. Marginal abatement cost of CO2 emission re-
ductions, geographical flexibility and concrete ceilings: an assessment using
the POLES model. Energy Policy 1999;27:585e601.
[10] Delarue ED, Ellerman AD, D'haeseleer WD. Robust MACCs? The topography of
abatement by fuel swithiching in the European power sector. Energy 2010;35:
1465e75.
[11] Egeskog A, Hansson J, Berndes G, Werner S. Co-generation of biofuels for
transportation and heat for district heating systems: an assessment of the
nation possibilities in EU. Energy Policy 2009;37:5260e72.
[12] Ellerman AD, Decaux A. Analysis of post-Kyoto CO2 emissions trading using
marginal abatement curves. Cambridge, MA: MIT; 1998.
[13] Flachsland C, Brunner S, Edenhofer O, Creutzig F. Climate policies for road
transport revisited (II): closing the policy gap with cap-and-trade. Energy
Policy 2011;39:2100e10.
[14] Franco A, Diaz AR. The future challenges for clean coal technologies: joining
efficiency increase and pollutant emission control. Energy 2006;34:348e54.
2006.
[15] Ghaderi A, Parsa Moghaddam M, Sheikh-El-Eslami MK. Energy efficiency
resource modeling in generation expansion planning. Energy 2014;68:529e37.
[16] Hasanbeigi A, Morrow W, Sathaye J, Masanet E, Xu T. A bottom-up model to
estimate the energy efficiency improvement and CO2 emissions reduction
potentials in the Chinese iron and steel industry. Energy 2013;50:315e25.
[17] Hetland J, Christensen T. Assessment of a fully integrated SARGAS process
operating on coal with near zero emissions. Appl Therm Eng 2008;16:2030e8.
[18] IEA. Paris: ISBN: 978-92-64-08624-1. World energy outlook 2010; 2010.
[19] Kesicki F. Marginal abatement cost curves for policy making e expert-based
vs. model-derived curves. Presented at the 33rd IAEE International Confer-
ence 6e9 June 2010. Rio de Janeiro, Brazil.
[20] Jacoby HD. The uses and misuses of technology development as a component
of climate policy. Cambridge, M.A: MIT; 1998.
[21] Kesicki F. Marginal abatement cost curves: combining Energy system
modelling and Decomposition analysis. Environ Model Assessments 2013;18:
27e37.
[22] Kesicki F, Ekins P. Marginal abatement cost curves: a call for caution. Climate
Policy 2011;12:219e36.
[23] Kesicki F, Stranchan N. Marginal abatement cost (MAC) curves: confronting
theory and practice. Environmental Science Policy 2011;14:1195e204.
[24] Klepper G, Peterson S. Marginal abatement cost curves in general equilibrium:
the influence of world energy prices. Resour Energy Econ 2006;28:1e23.
[25] Knutsson D, Sahlin J, Werner S, Ekvall T, Ahlgren EO. HEATSPOT: a simulation
tool for national district heating analysis. Energy 2006;31:278e93.
[26] Kok R, Annema JA, van Wee B. Cost-effectivness of greenhouse gas mitigation
in transport: a review of methodological approaches and their impact. Energy
Policy 2011;39:7776e93.
[27] Komiyama R, Fujii Y. Assessment of massive integration of photovoltaic sys-
tem considering rchargable battery in Japan with high time-resolution
optimal power generation mix model. Energy Policy 2014;66:73e89.
[28] Levihn F. CO2 emissions accounting: whether, how, and when different allo-
cation methods should be used. Energy 2014;68:811e8.
[29] Levihn F, Nuur C. Biomass and waste incineration CHP: co-benefits of primary
energy savings, reduced emissions and costs. WIT Transactions on Ecology
and the Environment 2014;190:127e38.
[30] McKinsey&Company. The carbon productivity challenge: curbing climate
change and sustaining economic growth. McKinsey Global Institute; 2008.
[31] McKitrick R. A derivation of the marginal abatement cost curve. J Environ Econ
Manag 1999;37:306e14.
[32] Meier AK. Supply curves of conserved energy. California: Lawrence Berkeley
National Laboratory; 1982.
[33] de Melo CA, Jannuzzi G, Tripodi AF. Evaluating public policy mechanisms for
climate change mitigation i Brazilian buildings sector. Energy Policy 2013;61:
1200e11.
[34] Morris D, Paltsev S, Reily J. Marginal abatement costs and marginal welfare
costs for greenhouse Fas emissions reductions: results from the EPAA model.
Environ Model Assess 2012;17:325e36.
[35] Nemet GF, Baker E, Jenni KE. Modeling the future cost of carbon capture using
experts' elicited probabilities under policy scenarios. Energy 2013;56:218e28.
[36] Nordrum S, Lieberman D, Colombo M, Gorski A, Webb C. Assesment of green-
house Gas mitigation options and costs for California Petroleum Industry Fa-
cilities: the shape of Things to Come. Energy Procedia 2011;4:5729e37.
[37] Stankeviciute L, Kitous A, Criqui P. The fundamentals of the future interna-
tional emission trading system. Energy Policy 2008;36:4272e86.
[38] Stranchan N, Pye S, Kannan R. The iterative contribution and relevance of
modelling to UK energy policy. Energy Policy 2009;37:850e60.
[39] Stoft S. The economics of conserved-energy supply curves, power; 1995
[Berkley California].
[40] Vogt-Schilb A, Hallegatte S. Marginal abatement cost curves and the optimal
timing of mitigation measures. Energy Policy 2014;65:645e53.
[41] Vijay S, DeCarolis JF, Srivastava RK. A bottom-up method to develop pollution
abatement cost curves for coal-fired utility boilers. Energy Policy 2010;38:
2255e61.
[42] Wada K, Akimoto K, Sano F, Oda J, Homma T. Energy efficiency opportunities
in the residential sector and their feasibility. Energy 2012;48:5e10.
[43] Wa €chter P. The usefulness of marginal CO2-eabatement cost curves in Austria.
Energy Policy 2013;61:1116e26.
[44] Yang X, Teng F, Wang G. Incorporating environmental co-benefits into climate
policies: a regional study of the cement industry in China. Appl Energy
2013;112:1446e63.
[45] Zakkour P, Dixon T, Cook G. Financing early opportunity CCS projects in
emerging economics through the carbon market: mitigation potential and
costs. Energy Procedia 2011;4:5692e9.
[46] Taylor S. Marginal abatement cost curves for policy making e expert-based vs.
model-derived curves. Energy Policy 2012;48:142e51.