Int. J.

Production Economics 181 (2016) 441–449

Contents lists available at ScienceDirect

Int. J. Production Economics

journal homepage: www.elsevier.com/locate/ijpe

Environmental efficiency and energy consumption of highway

transportation systems in China
Malin Song a,n, Wanping Zheng a, Zeya Wang b
a
School of Statistics and Applied Mathematics, Anhui University of Finance and Economics, Anhui, Bengbu 233030, PR China
b
Department of Statistics, Rice University, United States

art ic l e i nf o a b s t r a c t

Article history: With the rapid development of the Chinese economy and urbanization process, an increasingly large
Received 9 June 2015 urban transport system has led to increasingly serious resource and environmental problems in China.
Accepted 23 September 2015 We combined the super-efficiency slack-based measure model, including undesirable outputs, with the
Available online 3 October 2015
window data envelopment analysis model, which managed panel data, to calculate the environmental
Keywords: efficiencies of highway transportation systems in regions of China. This measured the level of sustainable
Highway transportation systems development in China’s highway transportation systems. Furthermore, we calculated the consumption
Environmental efficiency redundancy of gasoline and diesel, and the excess emissions of nitrogen oxides and particulate matter.
Energy consumption We analyzed the correlativity among environmental efficiency, nitrogen oxide efficiency, and particulate
Air pollution
matter efficiency to research the state of energy consumption and atmospheric pollution on the highway
transportation systems in China. The results showed that the overall level of environmental efficiency of
highway transportation systems in China was not optimal, with great differences between regions. Most
regions had problems with excessive energy consumption and motor vehicle pollution. The Chinese
government needs to control both energy consumption and pollutant emissions to manage the envir-
onmental problem of atmospheric pollution.
& 2015 Elsevier B.V. All rights reserved.

1. Introduction consumption decrease of more than 3.1% and to decrease major

Since China’s reform and opening-up, the Chinese economy has
developed rapidly, but energy and environmental problems have
developed owing to high energy consumption. This high energy
consumption seriously restrains China’s economic and social sus-
tainable development. The issues of how to alleviate climate pol-
lution, such as haze, and to guarantee the population’s health have
become important subjects for the Chinese people and the Chinese
economy. The main sources of atmospheric pollution in China are
coal and oil. The most important tools to combat haze are energy
conservation, emission reduction, energy structure adjustment,
and structural reform.
To cope with the change in the global climate, China has
reinforced resource conservation and management, strengthened
environmental protection, and implemented the “green and low-
carbon” initiative in the 12th Five Year Plan for 2011–2015.
Recently, its government stated that the main goal of China’s
economic and social development would be an energy
n
Corresponding author. Tel.: þ 86 15805521822; fax: þ86 05523173360.
E-mail address: songml@aufe.edu.cn (M. Song).
pollutants consistently.
To adhere to this new directive, China will have to identify and
tackle its environmental issues. Oil-fired pollutants are one of the
main sources of atmospheric pollution. Most oil-fired pollutants
come from the energy consumption of motor vehicles. With the
rapid development of the Chinese economy and urbanization
process, larger transport systems have increased the influence of
transportation on resources and the environment exponentially.
Among the various types of transportation, highway transporta-
tion is the largest contributor to resource consumption and
environmental pollution. The main resources consumed by high-
way transportation are land use and energy consumption. Envir-
onmental pollution is mainly caused by vehicle emissions of
atmospheric pollutants and noise pollution caused by road traffic.
In order to measure the sustainable development of highway
transportation systems and further analyze the energy consump-
tion and atmospheric pollution of regions’ highway transportation
systems, we used the super-efficiency slack-based measure (SBM)
model to include undesirable outputs and calculate the environ-
mental efficiency of highway transportation systems in the regions
of China. We focused on the problems of energy consumption and
pollutant emissions, which are the major problems of highway

http://dx.doi.org/10.1016/j.ijpe.2015.09.030
0925-5273/& 2015 Elsevier B.V. All rights reserved.
442 M. Song et al. / Int. J. Production Economics 181 (2016) 441–449
transportation; these are also the most serious threats for atmo-
spheric pollution.
In what follows, we review some of the most relevant literature
on the energy and environmental problems of transportation
systems in China (Section 2). The indicator selection and data
sources are presented in Section 3, and a case study of the envir-
onmental efficiency of highway transportation systems in regions
of China in Section 4. There is further discussion of the state of
energy consumption and pollutant emissions in Section 5, and we
draw conclusions in Section 6.
2. Literature review
Issues of sustainable consumption and production have been
explored in recent years, and some researchers have identified the
major factors for improving consumption and production (Tseung
et al., 2013). In the 1990s, scholars from all fields began to pay
attention to the energy consumption and environmental pollution
of transportation systems.
Regarding the evaluation of environmental pollution, although
the number of vehicles in China is not large, as pointed out by
Walsh (1999), the pollution caused by transportation cannot be
ignored. Some researchers have explored the current status of the
road transportation industry and forecast future trends of oil
demand and CO2 emissions in China (He et al., 2005). There are
several crucial problems and challenges for the development of
Chinese transportation (Zhou and Szyliowicz, 2006). In recent
years, more and more attempts have been made to explore the
status and tendencies of the transportation industry in China. Hu
et al. (2010) indicated that by 2030, the vehicle population in
China will be 400 million, and fuel demand will be 350 million
tons if the current pattern continues. Measuring sustainable pro-
duction indicators has been identified as a very important envir-
onmental activity (Tseng, 2013). For example, Zhou et al. (2010)
researched the effects of temporary transportation control mea-
sures on urban motor vehicle emissions during the 2008 Olympic
Games in Beijing. The results showed that reasonable traffic sys-
tem improvement strategies could effectively control total motor
vehicle emissions in Beijing after the Olympic Games. Yan and
Crookes (2010) presented the current status of China’s road
transport industry in terms of vehicles, infrastructure, energy use,
and emissions and analyzed the implemented mitigation mea-
sures and future measures. Zhang et al. (2011) analyzed the cur-
rent status of transportation energy consumption in China and
identified the relationships between transportation energy con-
sumption and its impacted factors.
In order to facilitate change, strong national leadership is
required to help integrate accessibility and mobility objectives into
the urban development of expanding Chinese cities (Yang and
Gakenheimer, 2007). The exergy method of assessing the energy
utilization of the transportation industry in China was proposed by
Ji and Chen (2006). The exergy method is used to assess the
transportation system waste gases CO2, NOx, and SO2 emitted from
fossil fuel consumption in China (Ji et al., 2009). Methodologically,
the dynamics model is useful in the assessment of passenger
transport. By developing a system dynamics model for policy
assessment and CO2 mitigation potential analysis, Han and Haya-
shi (2008) concluded that the major potential for emissions
reduction exists in inter-city passenger transport. Jifeng et al.
(2008) used a system dynamics approach to evaluate urban
transportation system performance in Dalian, China, concluding
that Dalian should restrict the total number of vehicles to improve
the sustainability of its transportation systems. Life cycle analysis
(LCA) is also a new technique (Yan and Crookes, 2009) used to
analyze the energy use and greenhouse gas emissions of road
transportation fuels in China. Hence, borrowed quantitative tech-
niques are significant for the analysis of transportation systems.
Based on China’s 2008 statistical transport data 2008, analyzed the
trend of China’s multimodal transport markets and provided some
policy measures to optimize the operations of China’s intercity
transport network. A calculation model of transportation energy
consumption based on vehicle driving was developed (Shunping et
al., 2010) to calculate China’s transportation energy consumption
level, which was found to be rising. A non-radial SBM model
(Chang et al., 2013) was used to analyze the environmental effi-
ciency of China’s transportation industry in 2009.
From these quantitative methods, we find that the method of
data envelopment analysis (DEA) has been widely applied by many
scholars to the assessment of transportation systems. Combined
with the Charnes, Cooper, Rhodes (CCR) model and principal
component analysis (PCA), DEA has been used to evaluate the
performance of the transportation industry of China (Ding et al.,
2011). Recently, a revised DEA method was proposed with sensi-
tivity analysis of indexes to evaluate the performance of bus routes
within a public transportation system in Beijing and put forward
some improvement suggestions (Li et al., 2013). Based on the DEA
model, a method of three-stage virtual frontier version was pro-
vided to evaluate the transportation energy efficiency in each
region of China (Cui and Li, 2014). A non-radial DEA model was
combined with multidirectional efficiency analysis (MEA) invol-
ving undesirable outputs to measure the energy and environ-
mental efficiency of China’s transportation system during the
period 2006 to 2010 (Bi et al., 2014).
In addition to the studies mentioned, there is some research
using a DEA approach to explore transportation systems in
developed regions. The DEA model has been used to analyze the
efficiency and partial productivity of European railway freight
transportation in European countries during the period 1980–
2003 (Hilmola, 2007). A novel slack-based measure network data
envelopment analysis (SBM-NDEA) approach has been proposed
to measure both the technical efficiency and service effectiveness
of airlines (Tavassoli et al., 2014). In addition, the DEA Malmquist
productivity index provided significant results when used to
measure the productivity growth of transportation industries in
the United States (Choi et al., 2015).
We combined the super-efficiency SBM model and the window
DEA model to measure the environmental efficiency of highway
transportation systems in 30 regions of China and the level of
sustainable development in China’s highway transportation sys-
tem; then, we further analyzed the energy consumption and
atmospheric pollution of highway transportation systems to pro-
vide scientific advice for the sustainable development of highway
transportation systems in China. The SBM model can effectively
solve the problem of undesirable outputs; further, it can distin-
guish the efficiency levels of efficient decision-making units
(DMUs); the window DEA model can increase the quantity of
DMUs and solve problems related to the efficiency evaluation of
panel data effectively. We will explore these problems with a
super-SBM-window combined model.
This research is different from prior work in four ways that
reflect its unique contributions. First, this research focused on
highway transportation systems rather than the whole transpor-
tation system. Recent research about the environmental efficiency
of China’s transportation system, like the work of Chang et al.
(2013), has focused on China’s whole transportation sector,
including transport, storage, and post. Recently, the air pollution
caused by motor vehicle emissions has become one of the most
 M. Song et al. / Int. J. Production Economics 181 (2016) 441–449 443

severe environmental problems in China. Thus, we focused on the
highway transportation system to make our research more specific
and targeted.
Second, we selected relatively comprehensive indicators based
on the features of the highway transportation system. The occu-
pation of land resources, energy consumption, air pollution, and
noise pollution are the main influences of highway transportation
system on resources, energy, and the environment. We selected
four inputs from the three aspects of land resources, labor, and
energy consumption; four outputs from the aspect of the service
the highway system provides; and three undesirable outputs from
the two aspects of air pollution and noise pollution. These 11
indicators reflect the total input–output of the highway system
and the main impacts on energy and the environment. The
research of Chang et al. (2013) only used labor, capital, and energy
as inputs, value-added as an output, and carbon dioxide as an
undesirable output.
Third, we chose the super-SBM-window combined model,
which can solve the problems related to undesirable outputs, the
quantity of DMUs, and the panel data. Further, the super-efficiency
model can distinguish the efficiency levels of efficient DMUs to
make the results more detailed and accurate. Chang et al. (2013)
also used the SBM-DEA model to solve the undesirable output
problem, but the efficiency values of all efficient DMUs were 1, so
they could not distinguish between efficient DMUs.
Finally, in the context of the situation of China, this research
has practical implications. Energy and environmental problems
now seriously restrain China’s economic and social sustainable
development. Further, the rapid development of the Chinese
urbanization process is strengthening the influence of highway
transportation on energy and the environment. Evaluating and
studying China’s energy and environmental problems related to
the highway system is of importance for China’s reformation and
development. This study not only calculated the environmental
efficiency of the highway system, but also went into energy con-
sumption and air pollution.
3. Indicators and data sources
We selected data from 30 provinces of the Chinese Mainland
from 2011 and 2012. (Tibet was excluded because no energy
consumption data were available.) Motor vehicle pollution emis-
sion data have been included in China’s Statistical Yearbooks since
2011. Refer to Table 1 for the original data of the descriptive
statistics.
Highway mileage, employed population (the population
employed in the transportation sector), gasoline consumption, and
diesel consumption were considered input indicators; passenger
capacity, passenger turnover, freight volume, and freight turnover
were considered desirable output indicators; and nitrogen oxide
emissions, particulate matter (PM) emissions, equivalent sound
level (LEQ) of road traffic noise were considered undesirable out-
put indicators to calculate the environmental efficiency of the
highway traffic systems in the regions of China. Highway mileage
measures the distance of the transportation routes; it also indir-
ectly measures the land resources occupied by highway trans-
portation. Employed population reflects labor input, while gaso-
line and diesel consumption measure energy consumption. Pas-
senger capacity/turnover and freight volume/turnover reflect
highway traffic transportation services for society. Nitrogen oxide
emissions and PM emissions measure atmospheric pollution,
while LEQ of road traffic noise measures noise pollution.
The data for gasoline and diesel consumption came from the
Energy Statistics Yearbook of China, and the data for nitrogen
oxide emissions and PM emissions came from the Environment
Statistics Yearbook of China; all other data came from China’s
national data network. The data for energy consumption from the
Energy Statistics Yearbook of China measured the whole trans-
portation industry, while gasoline and diesel consumption mainly
came from motor vehicle consumption for highway transporta-
tion. We selected these two indicators as measurements of the
energy consumption of highway transportation since gasoline and
diesel are two of the largest types of energy consumed. In addition,
the LEQ of road traffic noise pollution is mainly derived from
highway transportation noise, so we used the LEQ of highway
transportation noise to represent the LEQ of road traffic noise.

Table 1
Descriptive statistical analyses of variables.Data source: Statistics from China.

Indicators Variable Unit Year Average value Maximum Minimum Standard deviation

Input Highway mileage 10 10

km 2011 13.48 1.21 28.33 7.22
2012 13.91 1.25 29.35 7.43
Employed population 1 person 2011 56,249 9013 186,881 37,269
2012 92,478 14,237 301,012 68,239
Gasoline consumption 104 ton 2011 143.43 5.76 575.23 130.43
2012 153.4 5.48 598.24 138.80
Diesel consumption 104 ton 2011 323.64 57.8 1082.28 226.18
2012 345.83 60.76 1282.30 249.87

Desirable output Passenger capacity 104 2011 109,419 3477 493,618 102,882
2012 118,442 3748 556,510 114,072
Passenger turnover 109 passenger kilometer 2011 557.93 55.64 2082.68 467.82
2012 614.81 59.50 2470.11 533.19
Freight volume 104 ton 2011 93,971 8952 279,380 68,820
2012 106,248 9700 296,754 77,561
Freight turnover 109 ton kilometer 2011 1711.59 97.14 6624.35 1839.83
2012 1983.57 109.35 7266.77 2102.15

Undesirable output Nitrogen oxides emission ton 2011 211,309 29,637 561,766 141,719
2012 212,009 31,224 547,971 139,476
PM emission ton 2011 20,830 3089 56,129 14,896
2012 21,047 3080 65,932 15,626
LEQ dB(A) 2011 68.6 67.0 70.0 .9
2012 68.6 66.9 70.1 .9
444 M. Song et al. / Int. J. Production Economics 181 (2016) 441–449

Further, we adopted provincial capitals to represent their pro- outputs was as follows:
vinces when we analyzed the LEQ of road traffic noise. P
m
After calculating and comparing the summary statistics of all 1þm
1
si =xik
i¼1
the input and output indicators for all regions for 2011 and 2012, min ρ¼  
P
q1 P
q2
we can understand the overall conditions of highway traffic sys- 1q 1
þ q2 srþ =yrk þ sbt  =btk
1
tems in the 30 provinces of Mainland China. r¼1 t¼1

We found the average value of most the indicators, except LEQ, X

n
s:t: xij λj  si rxik
to increase a little from 2011 to 2012, as shown in Table 1. The j ¼ 1;j a k
average values of the four input indicators increased as follows: X
n
3.20% for highway mileage, 64.40% for employed population, 6.96% yrj λj þ srþ Zyrk
for gasoline consumption, and 6.86% for diesel consumption. We j ¼ 1;j a k

found that employed population increased the most among the X

n

input indicators, while the growth rates for consumption of
gasoline and diesel were almost the same. The average value of the
four desirable output indicators increased as follows: 8.25% for
passenger capacity, 10.20% for passenger turnover, 13.06% for
freight volume, and 15.89% for freight turnover. We found that the
growth rate of freight was higher than that of passenger vehicles.
As for undesirable outputs, nitrogen oxide emission increased .33%
whereas PM emission increased 1.04%, which showed a small
increase of pollution emissions from motor vehicles.
Table 1 also shows that the standard deviation of most indi-
cators increased a little, except for nitrogen oxide emissions,
which decreased by 1.58%. This shows that the difference in
nitrogen oxide emissions in different regions decreased while
other the differences for other indicators increased. The table
shows that the standard deviation of the four input indicators
increased by 2.84% for highway mileage, 83.10% for employed
population, 6.41% for gasoline consumption, and 10.47% for diesel
consumption. According to the value of the standard deviation of
employed population, we found the difference in employed
population of transportation increased substantially among dif-
ferent regions. The difference in diesel consumption showed a
larger increase than that of gasoline consumption among the
regions. The standard deviation of the four desirable output indi-
cators increased as follows: 10.88% for passenger capacity, 13.97%
for passenger turnover, 12.70% for freight volume, and 14.26% for
freight turnover. The differences in the rates of increase among
these four indicators were quite small. As for the undesirable
outputs, nitrogen oxide emissions increased  1.58%, PM emis-
sions increased 4.90%, and the LEQ increased 2.74%. The differ-
ences in PM emissions among regions increased from 2011 to
2012, which is consistent with the increased consumption of
diesel. This proves that diesel vehicles were the major cause of PM
emissions. Although the average value of the LEQ remained the
same, the differences increased among regions.
btj λj sbt  r btk
j ¼ 1;j a k
!
1 X q1 Xq2
1 srþ =yrk þ sbt  =btk 4 0
q1 þ q2 r ¼ 1 t¼1
λ; s  ; s þ Z 0
i ¼ 1; 2; ⋯; m; r ¼ 1; 2; ⋯; q1 ; t ¼ 1; 2; ⋯; q2 ; j ¼ 1; 2; ⋯; nðj akÞ ð1Þ
In Eq. (1), the objective function ρ refers to the environmental
efficiency to be measured; si ,srþ , and sbt  refer to slacks of input,
desirable output, and undesirable output, respectively; m,q1 , andq2
refer to the factor types of input, desirable output, and undesirable
output, respectively; and xij , yrj , and btj are the ith input, the rth
desirable output, and the tth undesirable output, respectively, of
DMU j.
Window DEA is a common method used for panel data analy-
sis. It sets all DMUs within a certain period (window) as a refer-
ence, creating a data point multiplied in the reference set. The
number of DMUs will increase if each data point is considered as a
DMU, so window DEA can solve the problem of an insufficient
number of DMUs effectively (Charnes et al., 1985).
The window DEA method considers the data for each DMU in
each period of time as relatively independent of “DMU” (“DMU”
here means generalized DMU). Suppose there are n DMUs and p
periods of time; then, the total “DMU” will be np. If the width of
the window d is set to 1, then the window’s “DMU” is equal to the
actual number of DMUs and is equivalent to analyzing the periods
of “DMU.” If the width of the window d is set to p, then the
window’s “DMU” gathers all periods of “DMU” into a reference set
and then evaluates the set.
This paper uses a combined super-SBM-window model to
evaluate the environmental efficiency of highway transportation
systems in 30 regions of China. The super-SBM-window model
solves three problems: efficiency evaluation of undesirable out-
puts, distinguishing efficient DMUs, and the evaluation of multi-
period panel data.

4. Measure of environmental efficiency 4.2. Environmental efficiency

4.1. Methods 4.2.1. Comparative analysis

While DEA models measure inefficiency, there is no slack
variable. To solve this problem, let us consider the SBM model
(Tone, 2001), which can effectively solve the problem of environ-
mental efficiency measurement with undesirable outputs. It is
noticeable that the SBM needs to be reduced with undesirable
outputs (Cooper et al., 2007). The super-efficiency model was first
put forward by Andersen and Petersen (1993). The maximum
efficiency in the standard-efficiency model is 1, which is con-
sidered to be censored data. Thus, the core of the super-efficiency
model is to solve the problem of censored data. Normally, the
value of efficient DMUs in the super-efficiency model is more than
1, allowing efficient DMUs to be distinguished. This paper used the
super-SBM model to calculate environmental efficiency consider-
ing undesirable outputs. The super-SBM model with undesirable
We adopted four input indicators, seven output indicators, and
30 DMUs to calculate the environmental efficiency of the different
regions in China for two years. In order to increase the quantity of
DMUs, the window width was set to 60, and the “DMU” for two
years was gathered as a reference set, which solved the problem of
disunity of multi-period frontiers. We combined the super-
efficiency SBM model and the window DEA model, and we cal-
culated the environmental efficiency of highway transportation
systems in 30 regions of China for 2011 and 2012 (Table 2, Fig. 1)
based on the selection of four input indicators, four desirable
output indicators, and three undesirable output indicators. We
found the value of environmental super-efficiency of efficient
DMU to be more than 1 using the super-SBM-window model. This
model can not only distinguish between the efficient DMUs, but
also sort the DMUs of the 30 regions effectively based on efficiency
 M. Song et al. / Int. J. Production Economics 181 (2016) 441–449 445

Table 2 Table 3
Environmental efficiency of highway transportation system in each region. Regions’ environmental efficiency rankings.Data source: Table 2.
Data source: Calculated using super-SBM-window model.
Region Year 2011 Year 2012 Region Year 2011 Year 2012
Region Year 2011 Year 2012 Region Year 2011 Year 2012
Beijing 2 4 Henan 11 12
Beijing 1.1269 1.0403 Henan .5318 .5429 Tianjin 17 14 Hubei 23 22
Tianjin .3781 .5284 Hubei .2759 .3045 Hebei 16 17 Hunan 12 11
Hebei .4294 .4727 Hunan .5228 .6138 Shanxi 28 29 Guangdong 4 3
Shanxi .1739 .1679 Guangdong 1.0487 1.0541 Inner Mongolia 25 26 Guangxi 15 15
Inner Mongolia .2232 .2292 Guangxi .4533 .4876 Liaoning 20 18 Hainan 7 10
Liaoning .3402 .4418 Hainan 1.0299 .8681 Jilin 19 21 Chongqing 5 5
Jilin .3540 .3265 Chongqing 1.0458 1.0378 Heilongjiang 30 30 Sichuan 6 6
Heilongjiang .1482 .1451 Sichuan 1.0346 1.0347 Shanghai 3 8 Guizhou 26 25
Shanghai 1.0710 1.0073 Guizhou .2228 .2362 Jiangsu 13 13 Yunnan 29 28
Jiangsu .5144 .5428 Yunnan .1678 .1684 Zhejiang 9 9 Shaanxi 21 20
Zhejiang 1.0140 1.0013 Shaanxi .3246 .3331 Anhui 1 1 Gansu 14 16
Anhui 1.1412 1.1084 Gansu .4772 .4842 Fujian 22 23 Qinghai 24 24
Fujian .2871 .2685 Qinghai .2310 .2519 Jiangxi 18 19 Ningxia 8 2
Jiangxi .3684 .4017 Ningxia 1.0260 1.0570 Shandong 10 7 Xinjiang 27 27
Shandong 1.0046 1.0118 Xinjiang .2191 .1944

environmental efficiency for the remaining 20 regions were .5729

Fig. 1. Environmental efficiency of each region.
value (Table 3). At the same time, we also compared the efficiency
values for 2011 and 2012 for the different regions.
Based on Tables 2 and 3, and Fig. 1, it is clear that between 2011
and 2012, the values of environmental efficiency for the highway
transportation system in 30 regions (hereafter referred to as
environmental efficiency) were almost the same; they changed
insignificantly. Nine regions—Anhui, Beijing, Guangdong, Ningxia,
Shandong, Zhejiang, Shanghai, Sichuan, and Chongqing—showed
environmental efficiency for these two years. Hainan’s value
reached the efficient level of 1.0299 in 2011 and dropped to the
non-efficient level of .8681 in 2012. The average values of
for 2011 and .5787 for 2012. The level rose slightly, but not sig-
nificantly. The standard deviations were .36 for 2011 and .34 for
2012, which means the differences between regions decreased
very little. The efficiency value of the Anhui region stayed at the
highest level among regions, at 1.1412 for 2011 and 1.1084 2012.
The Heilongjiang area stayed at the lowest levels of .1482 and
.1451, respectively. Hainan dropped to the non-efficient level of
.8681 in 2012 was because of the imbalance between inputs and
outputs for the highway transportation system in Hainan in 2012.
Specifically, highway mileage redundancy was 39.33 kilometers,
the employed population redundancy was 2164 persons, and
diesel consumption redundancy was 35,700 tons, while passenger
turnover deficiency was 7.1 hundred million passenger kilometers
and freight volume deficiency was 75.136 hundred thousand tons.
The equivalent sound level (LEQ) exceeded 3.24 dB. Specifically,
comparing the values of environmental efficiency, the 10 regions
that regressed from 2011 to 2012 were Beijing, Shanxi, Jilin, Hei-
longjiang, Shanghai, Anhui, Fujian, Hainan, Chongqing, and Xin-
jiang. However, the environmental efficiency values of the
remaining 20 regions improved in 2012. Hainan and Beijing were
the two regions in which the efficiency value decreased most
significantly; the reduced values were .1618 and .0866, respec-
tively. Tianjin and Liaoning were the two regions in which effi-
ciency value increased the most; the increased values were .1503
and .1016, respectively.
In terms of regional rankings, 11 regions—Beijing, Hebei,
Shanxi, Inner Mongolia, Jilin, Shanghai, Fujian, Jiangxi, Henan,
Hainan, and Gansu—regressed from 2011 to 2012. The rankings of
the other 19 regions remained unchanged or showed progression.
Shanghai regressed the most, although the value of efficiency
remained at the same level for the two years. Ningxia made the
most progress, from eighth to second.
Twenty regions retained non-efficient environmental efficiency
levels in both 2011 and 2012 because of an imbalance in the input–
output systems of their highway transportation systems. For
example, the region of Heilongjiang, whose environmental effi-
ciencies stayed at the lowest levels of .1482 and .1451, had severe
redundancy of inputs and undesirable outputs, and deficiency of
outputs. Specifically, the highway mileage redundancies were
1.2 and 1.206 hundred thousand kilometers in 2011 and 2012,
respectively; the redundancies of the employed population were
50,977 and 62,728 persons, respectively; and the redundancies of
gasoline and diesel consumption were 157.09 and 117.66 ten
thousand tons in 2011 and 177.09 and 134.29 ten thousand tons in
2012, respectively. While the deficiencies of passenger capacity,
freight volume, and freight turnover in 2011 and 2012 were 32.53
446 M. Song et al. / Int. J. Production Economics 181 (2016) 441–449

Table 4
Pearson correlation of input and output variables.Data source: Calculated using SPSS.

HM EP GC DC PC PT FV FT NOE PME LEQ

HM 1
EP .26* 1
GC .43** .66** 1
DC .52** .52** .68** 1
PC .50** .73** .77** .60** 1
PT .59** .69** .80** .66** .94** 1
FV .65** .49** .54** .74** .64** .77** 1
FT .51** .30* .26* .61** .38** .54** .90** 1
NOE .62** .50** .59** .70** .54** .64** .77** .74** 1
PME .57** .45** .54** .71** .49** .61** .79** .81** .97** 1
LEQ .02 .20 .16 .24 .08 .06 .11 -.03 -.05 -.05 1

HM¼ highway mileage, EP ¼employed population, GC¼ gasoline consumption, DC¼ diesel consumption, PC¼ passenger capacity, PT ¼passenger turnover, FV¼ freight
volume, FT¼ freight turnover, NOE¼ nitrogen oxides emissions, PME¼ PM emissions.
*
Significant correlation at the .05 level (two-tailed).
**
Significant correlation at the .01 level (two-tailed).

and 46.97 million persons, 77.77 and 91.00 million tons, and Table 5
612.86 and 649.14 hundred million passenger kilometers, respec- Gasoline and diesel consumption redundancy.Data source: Calculated using super-
SBM-window model (104 tons).
tively. The excess of nitrogen oxide emissions and PM emissions
were 202,196 and 20,132.73 tons in 2011 and 197,000 and Region Gasoline redundancy Diesel redundancy
19,565.52 tons in 2012.
Year 2011 Year 2012 Year 2011 Year 2012

4.2.2. Further discussion Tianjin  45.53  25.96  75.76  43.28

Overall, 33% of regions achieved reached relatively efficient Hebei  42.79  29.84  204.11  167.27
environmental efficiency. We wondered whether there might be Shanxi  83.40  85.50  263.20  264.14
too many input and output indicators relative to the number of Inner Mongolia  126.93  120.68  466.71  426.03
Liaoning  231.48  188.59  385.32  274.11
DMUs. There were four input indictors, seven output indictors, and Jilin  15.01  16.47  127.45  146.56
60 DMUs reached by processing the window DEA model, which Heilongjiang  157.09  177.09  117.66  134.29
met the DMU guiding principles (Cooper et al., 2007), that is Jiangsu  288.06  295.08 0.00 0.00
  Fujian  67.02  66.99  179.06  190.02
n Z max m  q; 3  ðm þqÞ ð2Þ Jiangxi  46.20  50.15  113.18  98.78
Henan  85.64  80.67  101.53  93.33
Furthermore, we checked whether the number of efficient Hubei  213.10  234.35  157.99  114.89
Hunan  57.36  59.40  155.23  13.32
DMUs would decrease if we decreased the quantity of indicators.
Guangxi  64.73  63.71  189.84  182.59
We calculated 11 input and output indictors of Pearson correlation Hainan 0.00 0.00 0.00  3.57
(Table 4): highway mileage, employed population, gasoline con- Guizhou  82.09  81.32  130.18  174.41
sumption, diesel consumption, passenger capacity, passenger Yunnan  83.70  86.08  342.63  366.06
Shaanxi  106.06  106.25  243.87  246.50
turnover, freight volume, freight turnover, nitrogen oxide emis-
Gansu 0.00 0.00  67.53  88.08
sions, PM emissions, and LEQ. The results show that passenger Qinghai  3.29  1.42  42.08  43.94
capacity and turnover, freight volume and turnover, and emissions Xinjiang  57.10  67.29  124.67  140.50
of nitrogen oxides and PM were highly correlated. The coefficients
of Pearson correlation were as high as .94, .90, and .97. This is
5. Energy consumption and air pollution
because passenger turnover and freight turnover include not only
information related to passenger capacity and freight volume, but
We obtained the slack variablessi ,stþ , and st of the input
also average transport distance. Thus, we removed the output
indicators and desirable and undesirable output indicators of
indicators of passenger capacity and freight volume, and then
DMUs using super-SBM-window model calculation. Thereby, we
combined nitrogen oxide and PM emissions as an undesirable
could get the inputs, redundancy of undesirable outputs, and
output indicator to represent the pollutant emissions of motor
deficiency of output of non-efficient DMUs. Thus, as we concerned
vehicles. The results of environmental efficiency we calculated ourselves with energy consumption and pollutant emissions of
were almost the same as the previous results after we removed environmental pollutants, we only discuss the redundancy of
and combined the indicators. Thus, we used the original 11 indi- gasoline and diesel consumption, and nitrogen oxide and PM
cators to calculate environmental efficiency since the information emissions. We need to reduce gasoline and diesel consumption
had more integrity. Further, we have better data for analysis if we (refer to Table 5) and reduce nitrogen oxide and PM emissions
distinguish between nitrogen oxide and PM emissions. (refer to Table 6; the minus sign indicated the need to reduce) in
In addition, the environmental efficiency of Beijing ranked best order to reach the production frontier with 21 non-efficient
in China, which went against the fact that there was serious pol- regions, including Tianjin and Hebei.
lution in Beijing. The environmental efficiency calculated in this
paper was the input–output efficiency of highway transportation 5.1. Redundancy of gasoline and diesel consumption
systems considering pollution emissions, so the value of environ-
mental efficiency reflected not only environmental pollution pro- From Table 5, we gather that redundancy of gasoline and diesel
blems, but also other aspects of inputs and outputs. The environ- consumption in non-efficient regions is greater than 0, except for
mental efficiency we calculated reflected the sustainable devel- redundancy of diesel consumption in Jiangsu, redundancy of
opment level of China’s highway transportation system. gasoline consumption and diesel consumption in Hainan in 2011,
 M. Song et al. / Int. J. Production Economics 181 (2016) 441–449 447

Table 6 Table 7
Excessive emissions of nitrogen oxides and PM.Data source: Calculated using super- Nitrogen oxide efficiency and PM efficiency.Data source: Table 6.
SBM-window model (tons).
Region Nitrogen oxide efficiency PM efficiency
Region Nitrogen oxides Particulate matter
Year 2011 Year 2012 Year 2011 Year 2012
Year 2011 Year 2012 Year 2011 Year 2012
Tianjin 0.50 0.926 0.368 0.604
Tianjin  27,047.43  4024.31  4102.46  2610.14 Hebei 0.35 0.416 0.383 0.471
Hebei  367,765.44  319,990.01  34,621.02  28,342.76 Shanxi 0.25 0.290 0.324 0.371
Shanxi  202,782.78  185,759.21  15,662.58  14,253.65 Inner Mongolia 0.44 0.529 0.411 0.498
Inner Mongolia  139,407.91  115,800.12  17,070.83  14,533.36 Liaoning 0.56 0.609 0.555 0.571
Liaoning  123,865.60  108,952.32  14,016.05  12,844.10 Jilin 0.43 0.475 0.461 0.515
Jilin  104,975.41  92,669.26  10,183.41  8778.42 Heilongjiang 0.21 0.229 0.230 0.249
Heilongjiang  202,195.91  197,000.04  20,132.73  19,565.52 Jiangsu 0.72 0.775 0.848 0.960
Jiangsu  94,278.57  76,336.53  4394.51  1149.05 Fujian 0.88 0.855 1.000 1.000
Fujian  11,624.59  15,104.15 0.00 0.00 Jiangxi 0.47 0.532 0.420 0.489
Jiangxi  114,079.05  103,658.40  15,638.39  13,657.11 Henan 0.45 0.443 0.482 0.356
Henan  271,900.20  279,540.79  26,624.26  42,434.33 Hubei 0.77 0.792 0.992 1.000
Hubei  41,622.49  38,789.55  120.81 0.00 Hunan 0.91 0.932 1.000 1.000
Hunan  14,595.93  11,843.46 0.00 0.00 Guangxi 1.00 0.996 0.936 0.935
Guangxi 0.00  596.82  1059.58  1107.65 Hainan 1.00 1.000 1.000 1.000
Hainan 0.00 0.00 0.00 0.00 Guizhou 0.89 0.950 0.995 1.000
Guizhou  10,352.90  4847.25  49.42 0.00 Yunnan 0.43 0.458 0.603 0.636
Yunnan  110,317.81  11,0161.58  6112.14  5895.33 Shaanxi 0.72 0.756 0.615 0.648
Shaanxi  50,638.91  43,145.34  8878.58  8083.83 Gansu 0.53 0.521 0.910 0.881
Gansu  53,075.50  56,453.69  708.10  1005.40 Qinghai 0.45 0.473 0.515 0.552
Qinghai  17,348.04  17,124.47  1498.86  1378.49 Xinjiang 0.20 0.216 0.302 0.315
Xinjiang  227,747.30  233,708.53  15,069.11  15,500.52

5.3. Environmental efficiency and nitrogen oxide efficiency and PM

and redundancy of gasoline consumption of Gansu. There are
always phenomena of excessive energy consumption in these non-
efficient regions. The total amounts of excessive gasoline con-
sumption were 18.5660 million tons for 2011 and 18.3682 million
tons for 2012 for all regions. The most serious examples of
excessive consumption were found in Jiangsu, Hubei, Liaoning,
Heilongjiang, Inner Mongolia, and Shaanxi (excessive consump-
tion over one million tons). The total amounts of excessive diesel
consumption were 34.8799 million tons for 2011 and 32.1167
million tons for 2012 for all regions. The most serious excessive
consumption took place in Inner Mongolia, Liaoning, Yunnan,
Shanxi, Shaanxi, and Hebei (excessive consumption over two
million tons; Hebei dropped to 1.6727 million tons in 2012). Inner
Mongolia and Liaoning had serious excess issues for both gasoline
and diesel consumption.
5.2. Excess of nitrogen oxide and PM emissions
Table 6 shows that the excess amount of nitrogen oxide and PM
emissions in non-efficient regions is greater than 0, except for
excess PM emissions in Fujian and Hunan, the excess of PM
emissions for 2012 in Hubei, the excess of nitrogen oxide emis-
sions for 2011 in Guangxi, and the excess of nitrogen oxide and PM
emissions in Hainan. There are always phenomena of excessive
pollutant emissions in these non-efficient regions. The total
amounts of excessive nitrogen oxide emissions were 2,185,621.79
tons for 2011 and 2,015,505.84 tons for 2012 for all regions. The
most serious producers of excessive nitrogen oxide emissions are
Hebei, Henan, Xinjiang, Shanxi, Heilongjiang, Inner Mongolia,
Liaoning, Jilin, Jiangxi, and Yunnan (excessive emissions over
100,000 tons; Jilin dropped to 92,669.26 tons in 2012). The total
amounts of excessive PM emissions were 195,942.84 tons for 2011
and 191,139.66 tons for 2012 for all regions. The most serious
excessive PM emissions were in Hebei, Henan, Heilongjiang, Inner
Mongolia, Shanxi, Jiangxi, Xinjiang, Liaoning, and Jilin (excessive
emissions over 10,000 tons; Jilin dropped to 8778.42 tons in 2012).
The nine regions mentioned have serious excess issues for both
nitrogen oxide and PM emissions.
efficiency
In order to further confirm the correlation between environ-
mental efficiency and pollutant emissions, we calculated nitrogen
oxide efficiency and PM efficiency (Table 7). Nitrogen oxide effi-
ciency is the ratio of goal nitrogen oxide emissions to actual
nitrogen oxide emissions; PM efficiency is the ratio of goal PM
emissions to actual PM emissions. Goal nitrogen oxide emissions is
the difference between actual nitrogen oxide emissions and
redundant nitrogen oxide emissions; goal PM emissions is the
difference between actual PM emissions and redundant PM
emissions. The calculation principle above is the reference carbon
efficiency formula from Chang et al. (2013).
The redundancy of nitrogen oxide and PM emissions in efficient
regions is 0, so the nitrogen oxide and PM efficiency is 1. Table 7
shows the nitrogen oxide and PM efficiency of non-efficient
regions. If the value of efficiency is 1, there are no excessive
emissions; the lower the efficiency, the larger the amount of
excess emissions.
We calculated the coefficient of Pearson correlation of envir-
onmental efficiency, nitrogen oxide efficiency, and PM efficiency;
we found these three efficiencies have remarkable correlativity.
The correlated coefficients of environmental efficiency with
nitrogen oxide and PM are .77 and .68, respectively. This proves
that environmental efficiency and nitrogen oxide/PM are highly
correlated. In other words, the environmental efficiency of high-
way transportation systems has direct correlation with the pollu-
tant emissions of motor vehicles. Sustainable development of
China’s highway transportation system is closely linked with the
pollutant emissions of motor vehicles.(Table 8)
6. Conclusions
Overall, the levels of environmental efficiency of highway
transportation in the 30 regions of China are quite different. In
2011, 10 regions achieved relative efficiency, and nine regions did
so in 2012. However, the environmental efficiency values of 17
regions in 2011 and 16 regions in 2012 were lower than .5, and the
448 M. Song et al. / Int. J. Production Economics 181 (2016) 441–449
Table 8
Pearson correlation between three efficiencies.Data source: Calculated using SPSS.
Environmental Nitrogen oxide
efficiency efficiency
Environmental 1
efficiency
Nitrogen oxide .770** 1 Andersen, P., Petersen, N.C., 1993. A procedure for ranking efficient units in data
efficiency
PM efficiency .683** .925**
**
Significant correlation at the .01 level (two-tailed).
efficiency value we calculated was a relative efficiency value. The
regions that achieved efficient status had not reached absolutely
efficient levels of environmental efficiency. The results show that
the input–output efficiency of highway transportation systems in
China is not very good when considering the effect of environ-
mental pollution. The sustainable development of China’s highway
transportation systems should be improved.
With regards to energy consumption, the regions with the most
serious excessive gasoline consumption were Jiangsu, Hubei,
Liaoning, Heilongjiang, Inner Mongolia, and Shaanxi (excessive
consumption over one million tons). The regions with the most
serious excessive diesel consumption were Inner Mongolia,
Liaoning, Yunnan, Shanxi, Shaanxi, and Hebei (excessive con-
sumption over two million tons). Inner Mongolia and Liaoning had
serious excess issues for both gasoline and diesel consumption.
These regions should enhance their measures of energy con-
servation. For example, they should popularize cars that use green
energy, compel people to take public transit instead of personal
cars, and adopt policies that limit the number of cars allowed to
pass through the system in major cities based on whether the last
number on the license plate is odd or even.
As for the pollutant emissions of motor vehicles, the regions
with the most serious excessive nitrogen oxide emissions were
Hebei, Henan, Xinjiang, Shanxi, Heilongjiang, Inner Mongolia,
Liaoning, Jilin, Jiangxi, and Yunnan (excessive emissions over
100,000 tons). The most serious excessive PM emissions were in
the nine regions mentioned above, except Yunnan (excessive
emissions over 10,000 tons). These regions need to strengthen
their measures of energy conservation and reduce emissions. They
must not only undertake the measures mentioned above, but also
improve the standard and quality of oil, and they should ban
vehicles that cannot meet the requirements.
Environmental efficiency has high correlation with nitrogen
oxide efficiency and PM efficiency, and nitrogen oxide and PM
efficiency have high correlation with gasoline and diesel con-
sumption. China needs to continue its journey toward green
transportation system construction, improve the sustainable
development of its highway transportation systems, and solve its
atmosphere pollution problems. The country needs to control
energy consumption and pollutant emissions while finding a way
of lowering energy consumption and lowering emissions. China
needs to improve its efficiency, and should reduce pollution.
Acknowledgments
This work was supported by Major Projects in Philosophy and
Social Science Research of the Ministry of Education of China,
under Grant no. 14JZD031; the Program for New Century Excellent
Talents in University, under Grant no. NCET-12-0595; the National
Natural Science Foundation of China, under Grant nos. 71171001
and 71471001; and the Key Projects in Philosophy and Social Sci-
ence of Anhui, China, under Grant no. AHSKZ2014D01.
PM efficiency
References
envelopment analysis. Manag. Sci. 39 (10), 1261–1264. http://dx.doi.org/
10.1287/mnsc.39.10.1261.
1
Bi, G., Wang, P., Yang, F., Liang, L., 2014. Energy and environmental efficiency of
China's transportation sector: a multidirectional analysis approach. Math. Probl.
Eng. . http://dx.doi.org/10.1155/2014/539596, Article ID 539596
Chang, Y.T., Zhang, N., Danao, D., Zhang, N., 2013. Environmental efficiency analysis
of transportation system in China: a non-radial DEA approach. Energy Policy
58, 277–283. http://dx.doi.org/10.1016/j.enpol.2013.03.011.
Charnes, A., Clark, C.T., Cooper, W.W., Golany, B., 1985. A developmental study of
data envelopment analysis in measuring the efficiency of maintenance units in
US air forces. Ann. Oper. Res. 2, 95–112. http://dx.doi.org/10.1007/BF01874734.
Choi, J., Roberts, D.C., Lee, E.S., 2015. Productivity growth in the transportation
industries in the United States: an application of the DEA Malmquist pro-
ductivity index. Am. J. Oper. Res. 5 (01), 1–20. http://dx.doi.org/10.4236/
ajor.2015.51001.
Cooper, W.W., Seiford, L.M., Tone, K., 2007. Data Envelopment Analysis: A Com-
prehensive Text with Models, Applications, References and DEA-Solver Soft-
ware. Springer Science & Business Media, Berlin, Germanyhttp://dx.doi.org/
10.1007/978-0-387-45283-8).
Cui, Q., Li, Y., 2014. The evaluation of transportation energy efficiency: an applica-
tion of three-stage virtual frontier DEA. Transp. Res. D – Transp. E 29, 1–11.
http://dx.doi.org/10.1016/j.trd.2014.03.007.
Ding, X.D., Xu, L., Yao, Z.G., 2011. Performance evaluation of Chinese transportation
industry based on DEA method. Wuhan Ligong Daxue Xuebao (J. Wuhan Univ.
Technol.) 33 (3), 77–81 (in Chinese).
Han, J., Hayashi, Y., 2008. A system dynamics model of CO2 mitigation in China’s
inter-city passenger transport. Transp. Res. D – Transp. E 13 (5), 298–305. http:
//dx.doi.org/10.1016/j.trd.2008.03.005.
He, K., Huo, H., Zhang, Q., He, D., An, F., Wang, M., Walsh, M.P., 2005. Oil con-
sumption and CO2 emissions in China's road transport: current status, future
trends, and policy implications. Energy Policy 33 (12), 1499–1507. http://dx.doi.
org/10.1016/j.enpol.2004.01.007.
Hilmola, O.P., 2007. European railway freight transportation and adaptation to
demand decline: efficiency and partial productivity analysis from period of
1980-2003. Int. J. Prod. Performance Manag. 56 (3), 205–225. http://dx.doi.org/
10.1108/17410400710731428).
Hu, X., Chang, S., Li, J., Qin, Y., 2010. Energy for sustainable road transportation in
China: challenges, initiatives and policy implications. Energy 35 (11),
4289–4301. http://dx.doi.org/10.1016/j.energy.2009.05.024.
Ji, X., Chen, G.Q., 2006. Exergy analysis of energy utilization in the transportation
sector in China. Energy Policy 34 (14), 1709–1719. http://dx.doi.org/10.1016/j.
enpol.2005.01.012.
Ji, X., Chen, G.Q., Chen, B., Jiang, M.M., 2009. Exergy-based assessment for waste gas
emissions from Chinese transportation. Energy Policy 37 (6), 2231–2240. http:
//dx.doi.org/10.1016/j.enpol.2009.02.012.
Jifeng, W., Huapu, L., Hu, P., 2008. System dynamics model of urban transportation
system and its application. J. Transp. Syst. Eng. Inf. Technol. 8 (3), 83–89. http:
//dx.doi.org/10.1016/S1570-6672(08)60027-6.
Li, J., Chen, X., Li, X., Guo, X., 2013. Evaluation of public transportation operation
based on data envelopment analysis. Procedia – Soc. Behav. Sci. 96, 148–155.
http://dx.doi.org/10.1016/j.sbspro.2013.08.020.
Shunping, J.I.A., Baohua, M.A.O., Shuang, L.I.U., Qipeng, S.U.N., 2010. Calculation and
analysis of transportation energy consumption level in China. J. Transp. Syst.
Eng. Inf. Technol. 10 (1), 22–27. http://dx.doi.org/10.1016/S1570-6672(09)
60020-9.
Tavassoli, M., Faramarzi, G.R., Saen, R.F., 2014. Efficiency and effectiveness in airline
performance using a SBM-NDEA model in the presence of shared input. J. Air
Transp. Manag. 34, 146–153. http://dx.doi.org/10.1016/j.jairtraman.2013.09.001.
Tone, K., 2001. A slacks-based measure of efficiency in data envelopment analysis.
Eur. J. Oper. Res. 130 (3), 498–509. http://dx.doi.org/10.1016/S0377-2217(99)
00407-5).
Tseng, M.L., 2013. Modeling sustainable production indicators with linguistic pre-
ferences. J. Clean. Prod. 40, 46–56. http://dx.doi.org/10.1016/j.jclepro.2010.11.019.
Tseng, M.L., Tan, R.R., Siriban-Manalang, A.B., 2013. Sustainable consumption and
production for Asia: Sustainability through green design and practice. J. Clean.
Prod. 40, 1–5. http://dx.doi.org/10.1016/j.jclepro.2012.07.015).
Walsh, M.P., 1999. Transportation and the environment in China. China Environ.
Ser. 3, 28–37.
Yan, X., Crookes, R.J., 2009. Life cycle analysis of energy use and greenhouse gas
emissions for road transportation fuels in China. Renew. Sustain. Energy Rev. 13
(9), 2505–2514. http://dx.doi.org/10.1016/j.rser.2009.06.012).
Yan, X., Crookes, R.J., 2010. Energy demand and emissions from road transportation
vehicles in China. Prog. Energy Combust. 36 (6), 651–676. http://dx.doi.org/
10.1016/j.pecs.2010.02.003.
Yang, J., Gakenheimer, R., 2007. Assessing the transportation consequences of land
use transformation in urban China. Habitat. Int. 31 (3), 345–353. http://dx.doi.
org/10.1016/j.habitatint.2007.05.001.
 M. Song et al. / Int. J. Production Economics 181 (2016) 441–449 449

Zhang, M., Li, H., Zhou, M., Mu, H., 2011. Decomposition analysis of energy con-
sumption in Chinese transportation sector. Appl. Energy 88 (6), 2279–2285.
http://dx.doi.org/10.1016/j.apenergy.2010.12.077.
Zhou, W., Szyliowicz, J.S., 2006. The development and current status of China's
transportation system. World Transp. Policy Pract. 12 (4), 10–16.
Zhou, Y., Wu, Y., Yang, L., Fu, L., He, K., Wang, S., Hao, J., Chen, J., Li, C., 2010. The
impact of transportation control measures on emission reductions during the
2008 Olympic Games in Beijing, China. Atmos. Environ. 44 (3), 285–293. http:
//dx.doi.org/10.1016/j.atmosenv.2009.10.040.