Академический Документы
Профессиональный Документы
Культура Документы
a r t i c l e i n f o a b s t r a c t
Article history: In the United States, the food industry alone accounts for approximately 19% of the total energy
Received 16 January 2017 consumed. The objective of this study is to forecast the food-related energy consumption and evaluate
Received in revised form policy alternatives for reducing food related energy consumption in the medium and long term. Influ-
7 July 2017
ential factors of energy consumption in the food industry include population growth rate, GDP, agri-
Accepted 12 July 2017
Available online 14 July 2017
culture's share of GDP, food waste, technologies, and agricultural investment. Based on the life cycle of
the food industry, five life stages were considered: agricultural production, industrial processing and
packaging, transportation, wholesale and retail, and household. In this study, based on the causal re-
Keywords:
Food industry
lationships and feedback loops between these factors and the energy consumed for each life stage, a
System dynamics system dynamics (SD) model was designed to simulate the situation for the US food industry. Policy
Energy consumption options of reducing waste percentage and improving industrial energy productivity are incorporated for
Policy analysis building the energy use mitigation scenarios. By implementing the integrated mitigation scenarios, a
9.43% reduction in total food system energy consumption can be attained in 2030.
© 2017 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.jclepro.2017.07.093
0959-6526/© 2017 Elsevier Ltd. All rights reserved.
14 Y. Xu, J. Szmerekovsky / Journal of Cleaner Production 165 (2017) 13e26
Mattias, 2009). resource savings and energy efficiency. In other efforts to increase
In this study, based on the causal relationships and feedback energy efficiency and save energy, Nunes et al. (2014) devised a
loops between these factors and energy consumed for each life predictive tool for assessing the energy performance of cold storage
stage of the food system, an SD model integrated with life cycle in agrifood industries, and Seck et al. (2013, 2015) developed a
analysis (LCA) was built to capture the constant shifts and changes bottom-up model to analyze the impact of heat recovery using heat
occurring in the US food system. The model was validated on data pumps in industrial processes. Lin and Xie (2015) studied the
from 1980 to 2014 and used to predict the trends of total US food substitution relationships between energy and other input factors
energy consumption through 2030. The objective here is to esti- like capital and labor and the rebound effect of the food industry in
mate potential energy consumption in the future and to assess the China. Most of this research provides tools for measuring energy
effects of two potential policy approaches for saving energy in the efficiency development in a single stage of the food industry.
US food industry: specifically, reducing the percentage of food However, we are unaware of any study that has comprehensively
waste and improving industrial energy productivity. examined the energy conservation status in all of the five life stages
Many papers have discussed energy conservation in the food in the food system.
industry in different parts of the world, but none has comprehen- Energy conservation or carbon emission reductions have always
sively examined the energy conservation status in all of the five life been a hot topic for researchers focusing on diverse industries.
stages in the food system. This paper addresses the following Methods employed in such studies can be divided into three main
important issues: categories: top-down, bottom-up, and hybrid approaches. The top-
down approach, such as Stochastic Frontier Analysis (SFA) and Data
the impacts of population and GDP growth on food demand; Envelopment Analysis (DEA), refers to the decomposition of a
the inherent structures of energy consumption under diverse scenario into a set of equations. DEA has been applied in the
scenarios for each life stage and the whole food system; transport sector (Park et al., 2016), the power industry (Hampf and
the impact of investments in public agricultural research and Rødseth, 2015), and the iron and steel industry (Yang et al., 2016).
development (R&D) on agricultural energy intensity; SFA has been used in the electricity sector (Chen et al., 2015), the
the impact of agriculture's share of GDP on agriculture energy chemical industry (Lin and Long, 2015), and the thermal power
use; and industry (Lin and Long, 2015). Bottom up approaches, such as
analysis of potential energy savings attained by two policies: conservation supply curve (CSC) and regression, analyze the global
reducing waste percentage and improving energy productivity impacts associated with an industry by assembling the various local
in industrial processing. disaggregated influences. Chen et al. (2017) applied CSC to analyze
the costs and the potential of energy conservation in China's coal-
The rest of this paper is organized as follows: Section 2 presents fired power industry. CSC has also been used in India's cement,
literature relevant to this study and how this research is derived; iron and steel industries by Morrow et al. (2014). A hybrid approach
Section 3 discusses the methodology, the modeling framework of combines both bottom-up and top-down approaches.
this study and the model validation; Section 4 develops the sensi- For the food industry, Gowreesunker and Tassou (2015) sum-
tivity analysis, the policy analysis and the result discussion; and marized many different approaches for modeling the food energy
Section 5 provides the conclusion and the limitations of this paper. flow, including life cycle analysis (LCA), MARKAL model, regression
analysis, input-output (IO) analysis and SD model. Among them,
2. Literature review LCA, MARKAL model and regression analysis belong to bottom-up
approaches, but IO analysis and SD model are top-down ap-
Much research has analyzed food related topics, especially is- proaches. MARKAL is a demand-driven model used to compare
sues like food security (Godfray et al., 2010; Maxwell, 1996; multi-period energy-related systems at the national level to
Schmidhuber and Tubiello, 2007) and food quality (Cen and He, represent its evolution. It is often used to determine the trade-offs
2007; Grunert, 2005). Since the food industry is not an energy- between different objectives such as costs and environmental in-
intensive industry and cost for energy inputs is only approxi- dicators (Gowreesunker and Tassou, 2015). Regression is a simple
mately 3% of the total cost of production (Muller et al., 2007), much model to reflect the causal relationships between the dependent
less literature deals with energy consumption in the food industry. and independent variables in the food system. Using statistical
Energy conservation in the food industry has been studied by methods or empirical observations to determine suitable influen-
several authors in different parts of the world, such as the US tial variables is especially important for regression analysis
(Azzam, 2012; Canning et al., 2010; Specht et al., 2014), China (Lin (Mendenhall and Sincich, 2012). Among these five methods, LCA
and Xie, 2015, 2016; Ma et al., 2012), the UK (Hall and Howe, and IO analysis are the most popular quantitative approaches. IO is
2011), France (Seck et al., 2013, 2015), the Netherlands (Ramírez a technique that represents the interdependencies between
et al., 2006), Switzerland (Muller et al., 2007), Portugal (Nunes different branches of economic sectors. Food-energy IO analysis is
et al., 2014, 2016), and Sweden (Wallgren and Mattias, 2009). usually used at a product specific level. This method always needs a
Ramírez et al. (2006) developed feasible indicators to monitor wide range of data presented in tabular form and can be attained
energy efficiency developments in the food industry based on from national statistical offices on a yearly basis. Canning et al.
physical production data at the company level. To our knowledge, (2010) employed the IO model to analyze the energy flow in the
this was the first time that energy usage per unit of physical output US food system between 1997 and 2007, and Zhang et al. (2012)
was examined in a non-energy intensive industry at a high level of built a multi-regional IO model to track the use of energy for
aggregation. Hall and Howe (2011) analyzed the influence of various sectors in China in 2007. LCA, also known as cradle to grave
improving different kinds of energy intensive technologies on en- analysis, is a technique for evaluating environmental impacts
ergy savings for the food processing industry. Azzam (2012) associated with all the processes or activities throughout the life
answered the question of how much of US energy consumption cycle of a product. LCA analysis was developed in the 1960s. Since
goes to the US food system by compiling diverse previous research then, more and more researchers have used LCA to monitor energy
findings into a table and provided several energy efficient actions. flows in the food industry, including Monforti-ferrario et al. (2015)
Specht et al. (2014) introduced an environmentally friendly build- and Dutilh and Kramer (2000). Another powerful approach to
ing named Zfarming for future urban food production promising model a complex system is SD, a computer aided approach to
Y. Xu, J. Szmerekovsky / Journal of Cleaner Production 165 (2017) 13e26 15
understanding the behavior of a system over time. SD models energy consumption and CO2 emissions in Beijing, China. They
provide an effective way to predict dynamic implications of separated the urban system into six subsystems, including agri-
policies. cultural consumption. The methodology of using agricultural en-
SD was first proposed by Forrester (1961). Since then, system ergy intensity and agricultural GDP to measure agriculture energy
dynamics has been widely used in various fields of study such as use in this research is derived from Feng et al. (2013). According to
sociology, ecology, engineering, and medical sciences. For example, Suh (2015), agricultural energy intensity is mainly influenced by
Hsu (2012) developed an SD model to assess the effects of policies agricultural budget and output. Since public research and devel-
like capital subsidies and feed-in-tariffs on solar PV installations. opment (R&D) investment is a major factor in the agricultural
Using this model to simulate the effect of applying different com- budget (Heisey et al., 2011), public R&D investment and agricultural
binations of promotion policies under budget limitations and CO2 output were chosen as two parameters affecting the agricultural
emission reduction goals in the long term, policymakers can carry energy intensity in this paper.
out cost/benefit analysis and decide the best policy to adopt in real
world. SD was also used by Saysel and Hekimog lu (2013) to build a
3. Methodology
quantity model to explore the options for carbon mitigation in the
Turkish electric power industry. After simulating the industry on an
SD is a powerful method for analyzing and simulating complex
annual basis for 30 years, the analysis revealed that adopting policy
systems and is often used for policy analysis. Vensim, a simulation
options such as feed-in-tariffs, investment subsidies and carbon
software for improving the performance of real systems, was used.
taxes could reduce CO2 emissions by 50%. Ansari and Seifi (2012)
This software has causal tracing, Monte Carlo sensitivity, optimi-
presented an SD model to study the effect of subsidy reform of
zation and sub-scripting capabilities (Zhang et al., 2012).
energy prices in Iran on energy consumption in the Iranian iron and
An SD model consists of feedback and causal loops, stock, flow
steel industry, and Ansari and Seifi (2013) presented another SD
and auxiliary variables. A feedback loop is a closed sequence of
model to explore the impact of this reform on production and en-
causes and effects. Stock is the state of an element that accumulates
ergy consumption for the Iranian cement industry. In both models,
the influences it receives over time. Flow is the increase or decrease
energy consumption in the industry was estimated under various
in the value of a stock during a period of time. The auxiliary vari-
production and export scenarios while taking into account new
ables are the rates of change for the stocks by determining the
energy prices. Corrective policies were suggested to achieve the
values of flow for a period of time. After defining and testing these
most energy savings based on the simulated outlook of energy
components, the system was simulated for a reasonable time
demand for each scenario. Other SD models have been employed to
period with respect to initial baseline values of past data. Then, by
examine semiconductor industry development (Chen and Jan,
comparing the simulation output with past data, the model was
2005), natural gas industry in the UK (Chi et al., 2009) and in
validated. Finally, a series of scenario analysis for the variation of
China (Li et al., 2011), and highway sustainability (Egilmez and
key input parameters was conducted to examine the system's
Tatari, 2012). SD has often been used to explain the social and
responses.
corporate behavior of a complex system over time, thereby making
it useful for assessing the effects of energy conservation and carbon
emission reduction policies in diverse industries. 3.1. Reference mode
In this research, SD modeling is used to estimate the total energy
consumption for the entire food sector instead of considering each In SD, the reference mode is simply another name for a plot of
food type or category in detail. In addition, to better illustrate this the behavior of key variables over time. Reference mode, based on
complex system, SD is integrated with some concepts of LCA for a historical information, gives clues to appropriate model structure
better understanding of the process. IO and MARKAL are impossible and can be used to check plausibility once the model is built. The
to apply without vast and accurate data which is not available for reference mode used in the SD model in this paper is the increasing
our problem. Regression is useful to get a rough indication of policy energy consumption trend in the US food industry. There is a sig-
impacts for a simple case, but errors and uncertainties increase a lot nificant amount of data on energy use and economic activity in
when applying it to a complex system. Dutilh and Kramer (2000) different parts of the food sector that can be found from the
employed LCA to compare energy requirements for different cate- Department of Energy, Department of Agriculture and Department
gories of food in each production and consumption stage and of Commerce, but there isn't any official government figure on the
concluded that there is a remarkable relationship between energy total amount of US energy consumption specifically for the food
requirements and market value throughout the production chain. industry. Indeed, because of the complexity of the modern food
They demonstrate which life stages should be incorporated and industry, isolating food activity from other aspects of life is nearly
how to estimate energy demand when modeling energy con- impossible, since food energy use is distributed across the resi-
sumption. However, they only provide a rough estimate for the dential, industrial, commercial, and transportation sectors (Minn
energy requirements for some major groups of food products. In and Seager, 2009). In this study, we use the proportion of energy
contrast, all types of food products will be considered as a whole in used for food and data of total US energy consumption to estimate
this research to provide existing and future information of energy the energy consumption in the US food sector. Annual Data for the
consumption for the entire food sector. This makes assessing which period 1980e2014 for total US energy consumption was attained
energy saving procedures should be used in the food industry and from the US Department of Energy, EIA (2016). According to the
their effectiveness more clear. Roy et al. (2009) indicate that agri- food energy analysis models of Steinhart 1974 and Minn 2009, food
cultural production has been the hotspot for researchers' focus energy took 12.7% of total US energy consumption in 1970 and
during the lifetime of food products. Some other LCA studies on 17.8% in 2007 (Minn and Seager, 2009). Based on these data, we
agricultural products include both agricultural production and in- estimate that the percentage of total US energy consumption that is
dustrial processing as priorities (Audsley et al., 1997; Berlin et al., attributable to the food system grows at a rate of (17.8%e12.7%)/
2007). Thus, in this research, the life stages of agricultural pro- (2007e1970) ¼ 0.14% per year, starting at 12.7% in 1970. This per-
duction and industrial processing and packaging are important centage is then multiplied by total US energy consumption to es-
focus points. Feng et al. (2013) used an SD model to analyze urban timate US food system energy consumption from 1980 to 2014. The
reference mode for US food energy consumption is shown in Fig. 1.
16 Y. Xu, J. Szmerekovsky / Journal of Cleaner Production 165 (2017) 13e26
3.2. Stock and flow diagram year). Correlation and regression analysis were conducted
between food requirements and agricultural output. The
A stock and flow diagram is the algebraic representation of a results shown in Table 1 indicate that there is a strong cor-
model based on the causal relationships of parameters. In this relation between food requirements and agricultural output
section, each parameter in the stock and flow diagram is explained. (0.957). Thus, the regression equation is given by
Then, a sub-model to estimate food demand is introduced. Finally, a
comprehensive stock and flow diagram is constructed. Agricultural output ¼ ð0:139food requirement þ 36830:8Þ=1000
Data source of variables and comprehensive regression equa- R2 ¼ 0:92; p value ¼ 0:000
tions used in the model are shown in Appendix. All variables (1)
measured in dollars are in real terms of 2009. Customer price index
(CPI), which was found from Bureau of Labor Statistics (2016), was
used to convert nominal dollar value into constant dollar value in (3) Agriculture energy use: Direct energy consumption, farm
2009 to exclude the effects of inflation. For example, the CPIs for fuel, and indirect energy consumption, like farm chemicals,
2005 and 2009 are 195.3 and 214.5, respectively, then the value of machinery and irrigation, were used to calculate the total
261.5 billion dollars for agricultural output in 2005 would be equal energy input of agriculture production. The idea of using the
to 287.2 billion dollars in 2009. This is calculated as 261.5*(214.5/ product of agricultural energy intensity and agricultural GDP
195.3). to measure agriculture energy use was derived from Feng
et al. (2013). Agricultural energy intensity was mainly
3.2.1. Exogenous variables influenced by agricultural budget and output effects (Suh,
GDP and population are two important exogenous variables 2015). Public R&D investment is a major factor in the agri-
which directly affect the demand for food (Ansari and Seifi, 2012, cultural budget, and it has a significant impact on agricultural
2013). The GDP directly reflects the economic development of a productivity (Heisey et al., 2011). Thus, a regression model
country. Population growth is closely related to the increase of food was built using agricultural energy intensity, public R&D
demand. Existing data for the GDP growth rate and the population investment, and agricultural output. The related index data
growth rate for 1980e2014 were averaged and then used to show was obtained from the relationship among these variables.
the pattern in the model. The average GDP growth rate is 0.272% Agriculture share of GDP measures the contribution of agri-
and average population growth rate is 0.998%. culture sector to the national economy. Relevant data from
1980 to 2014 were gathered from the Bureau of Economic
3.2.2. Endogenous variables Analysis (BEA, 2016) and an average value of agriculture
share was thus calculated and used in the model.
(1) Food requirement: Food requirement measures the quantity
of food needed to be produced to satisfy total food demand Agriculture energy use ¼ agricultural energy intensity
(2)
and post-consumer food waste. Market equilibrium, where agricultural GDP
food supply is exactly equal to food demand, is a prerequisite
for this research. Food waste takes a large proportion of the
food supply and can't be ignored. According to Buzby et al.
(2014), the amount of food loss and waste is between 30 Table 1
and 40 percent of total food supply from food industry. Thus, Correlation analysis between food requirement and agricultural output.
food requirement was calculated by summing total food
Food requirement Agricultural output
demand and estimated food waste.
(2) Agricultural output: Agricultural output measures the value Food requirement 1 0.957
Agricultural output 0.957 1
of agricultural products produced during a certain period (a
Y. Xu, J. Szmerekovsky / Journal of Cleaner Production 165 (2017) 13e26 17
Equations used in Vensim are shown in the Appendix. consumption data for the 1980e2014 period. Both data have no
The SD model was simulated for 35 years and the total food outliers (p-value > 0.05). Estimated data is normally distributed
system energy consumption obtained from the model at yearly because p-value is bigger than 0.05. However, p-value for the actual
intervals was compared with the real data. The comparison of the data is smaller than 0.05, indicating it's not normally distributed.
actual and the SD model's outputs is shown in Fig. 4. The R-square is Fortunately, ANOVA is pretty tolerant of violations of normality.
0.907. According to the output of two-sample variance test, both data
One-way ANOVA statistical hypothesis test was used to compare have equal variances (p-value > 0.05).
the actual energy consumption trend with the trend obtained with Since all three assumptions are met, the one-way ANOVA test
the proposed SD model. Some assumptions should be met before was employed for the comparison. The result of the ANOVA test is
using the one-way ANOVA test: a. normal distribution; b. equal shown in Table 2. According to these results, there is no statistically
variance; c. no outliers. significant difference between two samples (p-
Minitab was used to test these assumptions. Normality test and value ¼ 0.732 > 0.05). Thus, the reference mode provided by the
outlier test were applied to both actual and estimated energy proposed SD model is valid.
20
18
Quadrillion Btus
16
14
12
10
8
1980 1985 1990 1995 2000 2005 2010 2015
Table 2 intensity, and agriculture's share of GDP directly influence the en-
Results of the one-way ANOVA test. ergy use by agriculture sector. Thus, population, GDP, public R&D
Source of Variation SS df MS F P-value F crit investment and agriculture share were chosen for sensitivity
Between Groups 0.79039 1 0.79039 0.11758 0.73272 3.98189
analysis in this study.
Within Groups 457.0771 68 6.72172
Fig. 6. The estimated total food demand trends for baseline and G1.
20 Y. Xu, J. Szmerekovsky / Journal of Cleaner Production 165 (2017) 13e26
Fig. 7. The estimated total food energy consumption trends for baseline and RD1.
system. Total food demand is positively related to total food system 4.1.4. Impact of agriculture share on total food system energy
energy consumption, thus, GDP has a positive impact on the total consumption
food system energy consumption. Agriculture's share of GDP has a positive impact on agricultural
GDP, and agricultural GDP has a positive impact on the agriculture
energy use. And agriculture's share of GDP can be temporally
changed and is not connected to the year in this model. Thus, ag-
4.1.3. Impact of public R&D investment on total food system energy riculture's share of GDP, as an exogenous variable, should be tested
consumption for its influence on the total food energy consumption. In addition
As discussed above, agriculture energy intensity is mainly to the 0.852% share of the GDP given to agriculture (Baseline), an
influenced by agricultural budget and output effects. Public R&D additional scenario with a 1% share of the GDP for agriculture (AS1)
investment, as the major factor in the agricultural budget, is starting in 2015 was investigated. As shown in Fig. 8, agriculture's
consequently an important factor in agricultural energy use. share of GDP has a positive impact on the total food system energy
However, public R&D investment differs every year, so its sensi- consumption.
tivity should be analyzed. A 5% yearly growth rate of public R&D
investment starting from 2015 (RD1) was compared with the
baseline constant R&D investment. As shown in Fig. 7, the public 4.2. Scenario generation
R&D investment has a negative impact on the total food system
energy consumption, but the impact is not very big. In this study, the main objective is to reduce total energy
Fig. 8. The estimated total food energy consumption trends for baseline and AS1.
Y. Xu, J. Szmerekovsky / Journal of Cleaner Production 165 (2017) 13e26 21
consumption in the food supply chain. To realize this goal, two intervals, and annual reduction of waste percentage is 0.0109
potential strategies were considered and implemented in the SD (0.35*0.5/16) for WP1 and 0.0055 (0.35*0.25/16) for WP2.
model: reducing waste percentage (WP) and improving energy (2) Policy 2; improving energy productivity (EP): Energy pro-
productivity (EP). Two action plans were implemented based on ductivity is measured by output divided by energy input. The
each policy. Different scenarios were thus generated under the energy efficiency in food processing and packaging facilities
broad categories of baseline scenario and alternative scenarios. can be improved by upgrading existing equipment or intro-
ducing new energy-saving equipment such as refrigeration,
4.2.1. Baseline scenario pumps or fan systems. Research indicates that 10e15% in-
The baseline scenario was generated for the existing growth rate dustrial energy savings could be realized by implementing
of population and GDP without any policy interventions. Public new technologies (MnTAP, 2009). In other words, imple-
R&D investment remains constant as of the year 2014 and agri- menting new technologies could achieve an 11.1e17.6% in-
culture's share of GDP remains 0.852%. As mentioned earlier, the crease in energy productivity. According to equation (5),
population growth and GDP have positive effects on the total food these are calculated as (1/(1e10%) e 1) and (1/(1e15%) e 1).
demand as well as total food system energy consumption. The Two plans were applied to improve energy productivity. Plan
impacts of public R&D investment and agriculture's share of GDP 1 (EP1) attains a 17.6% increase in energy productivity by
on total food energy consumption were tested through sensitivity 2016, and plan 2 (EP2) achieves an 11.1% increase in energy
analysis. Thus, these four parameters should be controlled for the productivity by 2016. The expected energy productivity
baseline scenario and alternative scenarios. The projected trends of trends are shown in Fig. 10. Fig. 10 shows that energy pro-
total food energy consumption for the baseline were used as ductivity improves significantly in 2015 for EP1 and EP2,
feedback for applying policy options of reducing waste percentage then, the growth rates are the same as the baseline starting
and improving energy productivity. in 2016.
Fig. 11. The estimated total food energy consumption trends for all scenarios.
4.3.2. Alternative scenarios baseline scenario were tested and results were compared with each
Two policies were applied in the SD model and each policy has other. Results in terms of energy consumption in quadrillion Btus in
two plans. WP1 attains 50% food waste reduction by 2030 and WP2 2030 are show in Fig. 12.
attains 25% food waste reduction by 2030; EP1 achieves a 17.6% There are four possible combinations of these four plans: 1)
increase in energy productivity by 2016 and EP2 attains an 11.1% reducing waste percentage by 50% and at the same time attaining a
increase in energy productivity by 2016. These four plans and the 17.6% increase in energy productivity (WP1þEP1); 2) reducing
Y. Xu, J. Szmerekovsky / Journal of Cleaner Production 165 (2017) 13e26 23
Fig. 12. The food energy consumption results of single plans in 2030.
waste percentage by 50% and attaining an 11.1% increase in energy amount of energy before the year 2020, but WP1 consumes the
productivity (WP1þEP2); 3) reducing waste percentage by 25% and least amount of energy for the years 2020 onwards.
attaining a 17.6% increase in energy productivity (WP2þEP1); and From the results in Fig. 12, in 2030, estimated total food energy
4) reducing waste percentage by 25% and attaining an 11.1% in- consumption is 26.4418 quadrillion Btus if no action is taken. WP1
crease in energy productivity (WP2þEP2). These four combinations saves the most energy compared with the baseline; 7.15%
were tested and results are shown in Fig. 13. ((26.4418e24.5509)/26.4418) energy reduction will be attained by
From the results in Fig. 11, projected total food energy con- reaching the 50% food waste reduction goal. This suggests a greater
sumption under WP1 and WP2 show increasing trends for the opportunity for energy savings from reducing food waste than from
period 2015e2030, but the growth rates are lower than the base- improving energy productivity in the processing and packaging
line scenario since reducing waste percentage is applied at yearly stage in the mid run.
intervals. However, under EP1 and EP2, the estimated total food From the results shown in Fig. 13, the lowest energy consump-
energy consumption decreases substantially for the year 2016, then tion is reached by implementing both WP1 and EP1. In 2030, a
it increases with the same growth rate as the baseline scenario 9.43% ((26.4418e23.9495)/26.4418) energy consumption reduction
starting from 2017. There is a sudden decrease in the estimated is achieved by this combination compared with the baseline.
total food energy consumption in 2016 because new technologies In the above scenario analysis, the population growth rate, the
for improving energy productivity would be introduced that same GDP growth rate, the public R&D investment and agriculture's
year. WP1 has the lowest growth rate, and EP1 has the lowest value share of GDP remain constant. Population and GDP are major fac-
for projected total food energy consumption in 2016. As shown in tors in controlling total food demand, which in turn is the primary
Fig. 11, the lines for WP1 and EP1 intersect in the year 2020. Thus, determinant of agricultural production and total food system en-
when only considering single plans, EP1 consumes the least ergy consumption. Population stabilization or a decrease in GDP
Fig. 13. The food energy consumption results of plan combinations in 2030.
24 Y. Xu, J. Szmerekovsky / Journal of Cleaner Production 165 (2017) 13e26
would decrease the food demand in the food system and thus EP1 and WP1. Population growth rate, GDP growth rate and agri-
reduce total food energy consumption. Public R&D investment has culture's share of GDP have positive impacts on total food system
a negative impact on total food system energy consumption and energy consumption and public R&D investment has a negative
agriculture's share of GDP has a positive impact on total food en- impact on total food energy consumption. Therefore, targets for
ergy consumption. An increase in public R&D investment or a energy consumption must allow for future changes of these
decrease in agriculture's share of GDP would reduce the total food variables.
energy consumption. A 9.43% energy consumption reduction could The food industry is a very complex system which includes
be attained in 2030 by a combined scenario by reducing waste many different food types. Different types of food experience
percentage by 50% and attaining a 17.6% increase in energy pro- diverse life cycles and consume different degrees of energy for each
ductivity. Population growth rate, GDP growth rate, public R&D stage. Future work can specifically address the difference of energy
investment and agriculture's share of GDP are important exogenous consumption among different food types into modeling. Also, other
variables that typically change over time. Therefore, targets for influential factors such as household affluence and machinery
energy consumption must allow for future changes. could be considered as parameters. Although household energy use
consumes a large percentage of total food energy, this part is a
small part of this model since the focus of this research is on the
5. Conclusion and future work agricultural sector. Further research can further address the
household sector.
A system dynamics model for energy consumption in the food
industry was developed. The model was built to make projections
Acknowledgments
of food related energy consumption in the United States over a time
of 50 years. The objective was to evaluate policy alternatives for
I thank Megan Even and Yong shin Park for critically reading the
reducing total food energy consumption. Two potential strategies
manuscript and helpful discussions. This research did not receive
were considered and implemented in the SD model: reducing
any specific grant from funding agencies in the public, commercial,
waste percentage and improving energy productivity. By reducing
or not-for-profit sectors.
waste percentage at yearly intervals, the growth rate of the total
food energy consumption decreases for future years. By employing
new technologies to improve energy productivity in 2016, there is a Appendix A. Supplementary data
sudden decrease in total food energy consumption for the year
2016, then it increases with the same growth rate as the baseline Supplementary data related to this article can be found at http://
scenario starting from 2017. When only considering single plans, dx.doi.org/10.1016/j.jclepro.2017.07.093.
EP1 consumes the least amount of energy before the year 2020, but
WP1 consumes the least amount of energy for the years 2020 on- Appendix
wards. The final results show that a 9.43% reduction in total food
energy consumption can be attained in 2030 by combining both
Table 1
Main variables and parameters of SD model.
Table 1 (continued )
Research Funding in the Public and Private Sectors. http://www.ers.usda.gov/ Benchmark Input-Output Accounts (Various Years). http://www.bea.gov/
data-products/agricultural-research-funding-in-the-public-and-private- industry/index.htm#benchmark_io. (Accessed 10 June 2010).
sectors.aspx. (Accessed 10 June 2010). U.S. Department of Energy, Energy Information Administration. Annual Energy
U.S. Department of Agriculture, Economic Research Service, 2015. Per Capita Food Review (various years), www.eia.doe.gov/aer/. Accessed 10 June 2016.
Expenditure. http://www.ers.usda.gov/data-products/food-expenditures.aspx. U.S. Department of Labor, Bureau of Labor Statistics (BLS), 2016. Customer Price
(Accessed 10 June 2010). Index (CPI). http://data.bls.gov/pdq/SurveyOutputServlet. (Accessed 10 June
U.S. Department of Agriculture, Economic Research Service, 2016. Price Spread. 2010).
http://www.ers.usda.gov/media/308011/aer780h_1_.pdf. (Accessed 10 June Wallgren, C., Mattias, H., 2009. Eating energy d identifying possibilities for reduced
2010). energy use in the future food supply system. Energy Policy 37, 5803e5813.
U.S. Department of Agriculture, 2015. USDA and EPA Join with Private Sector, Yang, W., Shi, J., Qiao, H., Shao, Y., Wang, S., 2016. Regional technical efficiency of
Charitable Organizations to Set Nation's First Food Waste Reduction Goals. Chinese iron and steel industry based on bootstrap network data envelopment
Retrieved January 1, 2016, from. http://www.usda.gov/wps/portal/usda/ analysis. Socio Econ. Plan. Sci. 57, 1e11.
usdahome?contentid¼2015/09/0257.xml&navid¼NEWS_ Zhang, X.C., Ma, C., Zhan, S.F., Chen, W.P., 2012. Evaluation and simulation for
RELEASE&navtype¼RT&parentnav¼LATEST_RELEASES&edeployment_ ecological risk based on emergy analysis and pressure-state-response model in
action¼retrievecontent. a coastal city, China. Procedia Environ. Sci. 13, 221e231.
U.S. Department of Commerce, Bureau of Economic Analysis (BEA), 2016.