Assessing uncertainty in the agricultural marginal
abatement cost curves
V. Eory1*, C. F. E. Topp1, D. Moran1, A. Butler2
1
Research Division, SRUC, West Mains Road, Edinburgh EH9 3JG, UK
Biomathematics & Statistics Scotland, JCMB, The King's Buildings, Edinburgh, EH9 3JZ, UK
* Corresponding author. Tel.: +44 131 535 4301; fax: +44 131 535 4345. E-mail address:
[email protected]
2
Abstract
Information on the uncertainty of quantitative results feeding into public decision making is
essential for designing robust policies. However, this information is often not available in relation to
the economics of greenhouse gas (GHG) mitigation in agriculture. This paper analyses the
uncertainty of the mitigation estimates provided by a Marginal Abatement Cost Curve (MACC). The
case study is based on the GHG MACC developed for UK agriculture, restricting the analysis to
Scottish soils. The statistical uncertainty of the inputs was propagated through the model via Monte
Carlo analysis using three uncertainty scenarios. The results show that the uncertainty in the
economically optimal abatement in Scottish agriculture is high with the medium and high
uncertainty scenarios (mean ± 2 SD being 55%-101% and 98%-140% of the mean), while
assuming low uncertainty in the inputs results in 24%-68% uncertainty in the output. However, the
ranking of the measures are relatively robust with all three uncertainty scenarios, especially in
terms of which options have cost-effectiveness below the carbon price threshold. These results
imply that although there is a large uncertainty in abatement potential estimates, we have higher
certainty in which mitigation options should be implemented on farms.
Keywords: marginal abatement costs curves, uncertainty, greenhouse gases, agriculture, GHG
mitigation
Introduction
The complexity in uncertainty assessment often negatively impacts on the knowledge exchange
between scientist and policy makers, resulting in limited integration of uncertainty in the decision
making process (Knaggard 2013). These problems can be – at least partially – overcome by
applying decision making methodologies which incorporate uncertainties.
In certain fields of climate change research integrating uncertainty in the analysis has become the
norm, particularly in the physical sciences like climate modelling, but also to some extent in
economic research. Peterson (2006) gives an extensive overview of the economic models on
climate change integrating uncertainty in their assessments. These exercises either target the
global economy or the energy system, usually reporting on the uncertainties in GHG emission,
damage costs or mitigation costs. These results are particularly valuable at high level policy
decisions. However, being global or regional representations, they are limited in advising policy
development at the national level, where information on specific mitigation options, locations and
sectors are needed.
One particular decision making tool widely used today for assessing GHG mitigation policy is the
marginal abatement cost curve (MACC) analysis. It has been deployed in numerous instances to
estimate the optimal level of mitigation effort and to prioritise mitigation actions in terms of their
1
cost-efficiency (i.e. the cost of reducing GHG emission), often feeding into the policy process, for
example in the EU, US and UK (Kesicki and Strachan 2011). The MACCs’ popularity with policy
makers can be partly explained by its high visuality: it is able to convey condensed information in a
relatively simple way. However, this power has to be used with caution, as it can increase the risk
of overconfidence in the results – especially if uncertainty is not represented, which is often the
case. One of the noted shortcomings of most of the MACC analyses by date is the lack of
uncertainty assessment (Kesicki and Ekins 2012). Nevertheless, it is possible to address this
shortcoming of the MACC analysis.
Marginal cost
In theory, uncertainty in the marginal abatement cost curve and in the marginal damage cost curve
specifies the uncertainty of the economically optimal abatement level, which is defined as the
intercept of the two curves (Figure 1). In addition to the uncertainty of the economic optimum, an
engineering MACC, assessing alternative technologies as mitigation options, is likely to have
uncertainty in the abatement potential and cost of each option and in the ranking of the mitigation
options. This information becomes highly relevant when a MACC is feeding into the policy process
where the aim is to stimulate the uptake of selected mitigation options.
Marginal damage
cost curve
Marginal abatement
cost curve
Optimal abatement
Abatement
Figure 1. Effect of uncertainty on the optimal abatement (based on Smith and Stern (2011))
However, authors discussing the economics of GHG mitigation in agriculture rarely feature
uncertainty analysis in their results. Some exceptions include uncertainty analysis of mitigation
potential of biogas production in Germany (Meyer-Aurich et al. 2012), and farm level mitigation
potential and cost estimates on a UK farm with uncertainty reported on the total emissions and on
one mitigation option (Gibbons et al. 2006).
Here we make a systematic attempt to analyse the uncertainty of the cost-effectiveness of GHG
mitigation in agriculture and provide policy recommendations in the wake of the information about
the uncertainties.
Data and Methods
The analysis revisits data used to derive the GHG MACC developed for UK agriculture (Moran et al.
2011), restricting the analysis to Scottish soils. Moran et al. (2011) estimated the cost and
abatement potential of options applicable in UK systems, and calculated the ratio of these to attain
the cost-effectiveness of the options (£ tCO2e-1) for the years 2012, 2017 and 2022, considering
2
interactions between the options. Four uptake scenarios were modelled reflecting different
assumptions on the future policy environment: low feasible potential (LFP) assuming voluntary
uptake facilitated by information provision, central feasible potential (CFP) assuming voluntary
uptake based on financial incentives, high feasible potential (HFP) assuming compulsory regulation,
and maximum technical potential (MTP) assuming 100% uptake to explore the agronomic limits of
abatement. In the current exercise the focus was particularly on cropland options.
Statistical uncertainty of the inputs was propagated through the model via Monte Carlo analysis.
Given the lack of data on uncertainties, three uncertainty scenarios were constructed: High,
Medium and Low uncertainty, and accordingly three sets of probably density functions (PDFs) were
assigned to the inputs(Table 1).
Table 1. Characteristics of the three PDFs assigned to the inputs of the MACC model
Input
category
N2O GWP
Activity levels
Applicability
Uptake
Interaction
factors
Abatement rate
Net cost
a
Description and unit
100 year GWP [kg CO2e
(kg N2O)-1]
Areas of land under
different type of crops
(four crop categories) [ha]
Biophysical feasibility of
applying a option on a land
category [-]
Level of implementation of
a option by farmers across
Scotland, on land areas
where the option is
applicable [-]
Factor assigned to each
possible pairs of options,
describing the synergies
and antagonisms in the
GHG effectiveness of the
options [-]
High
UCa
Medium
UCa
Low
UCa
Mean±30%
Mean±20%
Mean±10%
Mean±30%
Mean±20%
Mean±10%
Mean±0.5
Mean±0.3
Mean±0.1
Mean±0.5
Mean±0.3
Mean±0.1
Mean±0.5
Mean±0.3
Mean±0.1
GHG effectiveness of the
options [t CO2e ha-1 year1
]
Mean-100%
mean+200%
Mean±200%
Mean-100%
mean+100%
Mean±100%
Difference between the
gross margin of the farm
with and without the
option applied, calculated
with a profit maximising
farm model [£ ha-1 year-1]
Mean±200%
Mean±100%
Mean-50% mean+50%
Mean±50%
Mean±50%
The PDF boundaries define the two standard deviation (95.45%) range (mean ± 2 SD)
Results
When propagating the uncertainties of all the inputs, the uncertainty (mean ± 2 SD) of the
economically optimal abatement rate was 26%, 63% and 106% of the mean with the Low, Medium
and High uncertainty scenario, respectively, for the 2022 CFP scenario (Figure 2). Across all years
and uptake scenarios it ranged from 24% to 140% of the mean, the lowest uncertainty existing for
the HFP in 2022 with the Low uncertainty scenario, and the highest uncertainty existing for the LFP
in 2012 with the High uncertainty scenario (Figure 3). In general, the uncertainty of this output
metric is decreasing with the increasing level of uptake in the uptake scenarios from LFP to MTP,
3
and also as the results are projected further in the future – this latter finding can also be explained
by the assumption of a linearly increasing level of uptake through time.
Probability
No PDFs
Narrow PDFs
Medium PDFs
Wide PDFs
GHG abatement (kt CO2e y-1)
Figure 2. Uncertainty of the economically optimal GHG abatement on croplands in Scotland, 2022, CFP. No
PDFs: the economically optimal abatement potential without uncertainty propagation
140%
140%
120%
120%
100%
100%
80%
80%
60%
60%
40%
40%
20%
20%
Low UC
0%
2012
Medium UC
2017
2022
High UC
a)
0%
LFP
CFP
HFP
MTP
b)
Figure 3. Uncertainty (mean ± 2 SD) of the economically optimal GHG abatement on croplands in Scotland.
a) as a function of year and uncertainty scenario for the CFP uptake scenario, b) as a function of uptake
scenario and uncertainty scenario for the year 2022
The uncertainty of the inputs results uncertainty also in the ranking of the options due to the
uncertainty in their cost-effectiveness and in the interaction factors. Figure 4 reveals that this
uncertainty can be relatively high for some options, for example ‘Improving the timing of slurry and
poultry manure application’, ‘Using reduced tillage and no-till techniques’ and ‘Avoiding nitrogen
4
application in excess’. Other options have lower uncertainty in their ranking, like ‘Separating slurry
applications from fertiliser applications by several days’ and ‘Introduction of new species (including
legumes)’. In spite of the uncertainty in the individual ranking of the options, the set of options
which are estimated to be cost-effective are relatively stable, having only a few options which
crosses the SPC threshold form either side.
0.6
Economically optimal
abatement
Improving the timing of mineral nitrogen
application
Adopting plant varieties with improved N-use
efficiency
0.5
Improving land drainage
Improving the timing of slurry and poultry
manure application
Using reduced tillage and no-till techniques
0.4
Probability
Avoiding nitrogen application in excess
Using manure nitrogen to its full extent
0.3
Using composts, straw-based manures in
preference to slurry
Separating slurry applications from fertiliser
applications by several days
0.2
Using nitrification inhibitors
Introduction of new species (including legumes)
Using controlled release fertilisers
0.1
Reducing N nitrogen fertiliser
Adopting systems less reliant on inputs
0
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
Using biological fixation to provide nitrogen
inputs
Ranking
Figure 4. Uncertainty (mean ± 2 SD) in the ranking of the cropland mitigation options, Scotland, 2022, CFP,
Medium uncertainty scenario
The contribution of the uncertainty in each input category to the uncertainty of the economically
optimal abatement was examined by propagating the uncertainty of all the inputs in one input
category at a time, for all the three years, four uptake scenarios and three uncertainty scenarios.
Figure 5 shows the contribution of the inputs’ uncertainty for the year 2022 with uptake scenario CFP. The
uncertainty in abatement was the most important contributor in most of the cases, with the
exception of the scenarios with low uptake rates – as is the case for the year 2012 and for the
uptake scenario LFP –, where the uncertainty of the uptake inputs is the most important driver. On
the other side, uncertainty in activity levels and costs adds the least to the uncertainty of the
optimal abatement, with the contribution of the uncertainties in the interaction factors, applicability
inputs and GWP of N2O being slightly higher.
5
120%
Uncertainty range (mean ± 2SD)
100%
All
80%
Abatement
Uptake
60%
Applicability
IFs
GWP
40%
Activity level
Costs
20%
0%
No UC
Low UC
Medium UC
High UC
Figure 5. The contribution of the inputs’ uncertainty to the uncertainty of the economically optimal abatement
potential (mean ± 2 SD), cropland mitigation, Scotland, 2022, CFP
Discussion and conclusions
The work presented in this paper systematically assesses the statistical uncertainties in the GHG
abatement potential in agriculture by an example of engineering MACC analysis of crops and soil
related mitigation options in Scotland. The statistical uncertainty was assessed by creating three
uncertainty scenarios and looking at how do the output uncertainties change and how do the inputs
contribute to the outputs’ uncertainties.
The uncertainty in the economically optimal abatement becomes high in the medium and high
uncertainty scenarios (mean ± 2 SD being 55%-101% and 98%-140% of the mean), while
assuming low uncertainty in the inputs results in 24%-68% uncertainty in the output. However, the
ranking of the measures are relatively robust, especially in terms of which options have costeffectiveness below the carbon price threshold. These results imply that although there is a large
uncertainty in abatement potential estimates, we have higher certainty in which mitigation options
should be implemented on farms. This finding corresponds to Gibbons et al. (2006), who found that
the total emissions from the farms are very uncertain, the mitigation options’ effects (expressed as
a proportion of total farm emissions) had a lower degree of uncertainty.
Looking at the contribution of the uncertainties in the input categories to the uncertainty in the
economically optimal abatement potential, abatement rate and uptake rate are the most important
input categories. At the same time these two and another three input categories (applicability rate,
net costs and interaction factors) have the largest extent of uncertainty. Inputs which both have
high uncertainty and contribute highly to the output uncertainty are the key factors to be
addressed if we are to reduce uncertainty in the outputs (Heijungs 1996). Regarding to the optimal
abatement potential from Scottish farms’ soil emissions the two key issues are uptake rate and
abatement rate (Figure 6.).
6
High
KEY ISSUES
Uptake rate
Abatement rate
NOT KEY ISSUES
Low
Contribution to output (optimal
abatement) uncertainty
POTENTIALLY
KEY ISSUES
POTENTIALLY
KEY ISSUES
Applicability rate
Activity level
Net costs
GWP
Interaction factors
Low
High
Level of uncertainty
Figure 6. Uncertainty assessment of the input categories
There are opportunities to reduce the uncertainty in agricultural GHG mitigation, though when
considering these opportunities the effort needed to reduce the uncertainties must be weighed
against the benefits gained from more robust predictions. First of all, the data gaps in the
uncertainties of the inputs are very large, both in the biophysical and in the socio-economic inputs.
Improving scientific reporting practice to include quantified data about the statistical uncertainty in
underlying research would be one of the most efficient ways to reduce uncertainty. Those inputs
with high uncertainty are for consideration for improvement. Ongoing research on the biophysical
aspects of the mitigation options is constantly providing more data on the abatement rate and at a
limited level about interaction factors – again, these results are most useful if accompanied by
uncertainty estimates. Similar improvements needed in the economic analyses, to reduce
uncertainties in cost estimates and also to improve the robustness of future changes in agriculture
and land use. Uncertainty in uptake rate can be improved by better understanding of behavioural
processes and the effects of policy instruments on farmers’ choices. Applicability rates are
ultimately based on agronomic experts’ opinion, elicitation of uncertainty in this case is also
possible, though resource intensive. Overall, it is likely that the uncertainties in biophysical and
economic modelling will become more explicit in the future, reducing the extent of uncertainty in
integrated modelling. However, improving our knowledge about the uncertainty in applicability
rates and uptake rates requires even more effort. Nevertheless, emphasis should be put on
supporting the ongoing research about abatement rates and about farmers’ behaviour.
There are certain limitations of the analysis presented. Firstly, as a MACC is using – amongst other
sources – other models’ results as inputs where the uncertainty is not reported, there is always
scope to go beyond the MACC model and do the uncertainty analysis in the underlying models as
well (e.g. in the farm financial model). However, as the uncertainty data regarding the other inputs
were neither available (with the exception of GWP), an uncertainty assessment was carried out
instead of an uncertainty analysis – the difference being in the inputs’ PDFs, as an uncertainty
analysis requires knowledge about the PDFs, while an uncertainty assessment – with using
assumptions on the PDFs – aims to explore the most important features of the uncertainty in the
7
modelling, without putting much emphasis on quantified uncertainty results. Second, given that
uncertainties of the different inputs often share the same source, i.e. there is correlation between
the uncertainties (e.g. the abatement rate of some options), specifying joint distributions would
improve the analysis. The scope of this analysis did not allow for exploring these correlations,
resulting in a potentially minor overestimation of the uncertainties of the outputs.
MACCs, like other integrative tools, accumulate uncertainties. Input data might include statistics,
meta-analysis of field experiments, results from biophysical and financial models, results from
expert elicitation exercises, or assumptions based on the judgment of the researchers. These
inputs all have their underlying uncertainties, partly quantifiable statistical uncertainties, partly
unquantifiable uncertainties. However, assessing the importance of these uncertainties along with
what extent they can be reduced is an important piece of information for designing more robust
policies.
Acknowledgement
This research was undertaken within the Scottish Government Rural Affairs and the Environment
Portfolio Strategic Research Programme 2011-2016. Specifically with funding provided to
ClimateXChange. For more information please see:
http://www.scotland.gov.uk/Topics/Research/About/EBAR/StrategicResearch/future-researchstrategy/Themes/ThemesIntro. Further funding was provided by the AnimalChange project which
received funding from the European Community's Seventh Framework Programme (FP7/ 20072013) under the grant agreement n° 266018.
References
Gibbons,J.M., S.J.Ramsden, and A.Blake. 2006. "Modelling uncertainty in greenhouse gas
emissions from UK agriculture at the farm level." Agriculture Ecosystems & Environment.
112:347-355.
Heijungs,R. 1996. "Identification of key issues for further investigation in improving the reliability
of life-cycle assessments." Journal of Cleaner Production. 4:159-166.
Kesicki,F. and P.Ekins. 2012. "Marginal abatement cost curves: a call for caution." Climate Policy.
12:219-236.
Kesicki,F. and N.Strachan. 2011. "Marginal abatement cost (MAC) curves: confronting theory and
practice." Environmental Science & Policy. 14:1195-1204.
Knaggard,A. 2013. "What do policy-makers do with scientific uncertainty? The incremental
character of Swedish climate change policy-making." Policy Studies.1-18.
Meyer-Aurich,A., A.Schattauer, H.J.Hellebrand, H.Klauss, M.Pl+Âchl, and W.Berg. 2012. "Impact of
uncertainties on greenhouse gas mitigation potential of biogas production from agricultural
resources." Renewable Energy. 37:277-284.
Moran,D., M.MacLeod, E.Wall, V.Eory, A.McVittie, A.Barnes, R.M.Rees, C.F.E.Topp, G.Pajot,
R.Matthews, P.Smith, and A.Moxey. 2011. "Developing carbon budgets for UK agriculture,
land-use, land-use change and forestry out to 2022." Climatic Change. 105:529-553.
Peterson,S. 2006. "Uncertainty and economic analysis of climate change: A survey of approaches
and findings." Environ Model Assess. 11:1-17.
Smith,L.A. and N.Stern. 2011. "Uncertainty in science and its role in climate policy." Philosophical
Transactions of the Royal Society A-Mathematical Physical and Engineering Sciences.
369:4818-4841.
8