T RANSPORT CARBON EMISSIONS REDUCTION EFFICIENCY AND ECONOMIC GROWTH : A PERSPECTIVE FROM NIGHTTIME LIGHTS REMOTE SENSING

: The carbon emissions are essential for climate change and 26% of the world's carbon emissions are related to transport. But focusing only on fewer carbon emissions might be biased at times. In order to keep a balance between economic growth and carbon emissions reduction, this paper evaluated the performance of carbon control by considering the input factors and output factors together, which is more comprehensive and reliable. Firstly, this paper has calculated the transport carbon emissions reduction efficiency (TCERE) based on the model of super SBM with undesirable outputs. The input factors include capital stock, labor force and fossil energy consumption. And the output factors include gross domestic product and carbon dioxide emissions. Then the influencing factors of TCERE were analyzed using econometric models. The economic growth, transport structure, technology level and population density were posited as influencing factors. This paper creatively proposed the per capita nighttime lights brightness as a new indicator for economic growth. An empirical study was conducted in East China from 2013 to 2017, and this study has found that the relationship between TCERE and economic growth shows an U-shape. Besides, transport structure and technology level both show a positive impact on TCERE. The implications of our findings are that: (a) The TCERE declines slower in East China, giving us reason to believe that the improvement of TCERE is predictable; (b) When economic growth exceeds the turning point, economic growth is conducive to the improvement of TCERE. We could develop the economy boldly and confidently; (c) Increased investment in railway and waterway transportation infrastructure projects is needed to strengthen the structure of the railway and waterway transportation systems. Furthermore, the general public and businesses should be encouraged to prefer rail or river transportation; (d) Investment in scientific and technological innovation should be enhanced in order to produce more efficient energy-use methods.


Introduction
It is acknowledged that climate change is a global issue and carbon emissions play a significant role in it (Zhang et al., 2018). The Chinese government has enacted laws and implemented programs to develop a low-carbon economy and reduce carbon emissions in order to implement the Paris Agreement on Climate Change and the United Nations 2030 Agenda for Sustainable Development. 26% of the world's carbon emissions are related to transport (Cui et al., 2015). Reducing transport-related carbon emissions is crucial for low-carbon economy, and many nations have put limitations on transport-related carbon emissions, including Canada and Brazil (Cui et al., 2015). Some researchers also have studied the variation of transport carbon emissions (Blesl et al., 2007;Xu et al., 2014). But focusing only on fewer carbon emissions might be biased at times. In order to keep a balance between economic growth and carbon emissions reduction, we need to evaluate the performance of carbon control by considering the input factors and output factors together, which is more comprehensive and reliable. The input factors include capital stock, labor force and fossil energy consumption . And the output factors include gross domestic product (GDP) and carbon dioxide (CO2) emissions (Zhao et al., 2020;. Efficiency is an indicator that assesses a decision-making unit's (DMU's) ability to maximize outputs given a certain set of inputs, which is commonly applied in assessing the relationship between inputs and outputs . Thus, this paper uses the transport carbon emissions reduction efficiency (TCERE) to evaluate the performance of transport carbon control. With the purpose of better analyzing the impact of influencing factors (especially economic growth) on TCERE in East China, this paper intend to conduct an research using nighttime lights data and super slacks-based measure (SBM) model with undesirable outputs. First, a series of raw data in East China will be obtained through yearbooks, remote sensing images, etc. Next, the input and output variables for determining efficiency will be calculated using raw data. Then, the TCERE will be calculated using a super SBM with undesirable outputs model. Besides, the per capita brightness of nighttime lights, as well as other influencing factors, will be calculated using raw data. Lastly, we will establish an econometric model and intend to examine how influencing factors affect the TCERE. The main contributions and novelty of our work are that: − We adopt an improved DEA model named the super SBM with undesirable outputs model to calculate TECRE, which is more reasonable. − When analyzing the influencing factors of TCERE, we creatively use a new indicator derived from remote sensing images to represent economic growth. The remaining parts are structured as follows. Section 2 is the literature review. Section 3 describes the processing methods. Section 4 describes the data. Section 6 presents the results and discusses the findings. Lastly, the conclusions are presented.

Literature review
In general, methods to evaluate efficiency can be divided into parametric and non-parametric models (Fan et al., 2017). Unlike parametric models, nonparametric models do not require to specify function forms (Cullinane et al., 2006;. Thus, non-parametric models can avoid causing model specification errors (Fan et al., 2017). Among various non-parametric models, the data envelopment analysis (DEA) is a common one. DEA employs a linear programming approach to determine the efficiency of each DMU, which has been extensively applied in various fields (Zhao et al., 2020). Tsolas (2022) applied DEA to assess regional entrepreneurship in Greece and demonstrated the viability. Wang  Neglecting undesirable outputs might lead to inaccurate analysis (Wang et al., 2012). Thus, Tian et al. (2022) used the SBM model with undesirable outputs to assess the urban green development efficiency of 277 Chinese cities. Park et al. (2018) adopted the SBM model with an undesirable output (CO2) to analyze the environmental efficiency of US transport sector. However, the maximum value of a standard SBM model is 1.0, whereas many DMUs might reach the upper bound and cannot make a valid comparison . Therefore, we adopt an improved DEA model, i.e., the super SBM with undesirable outputs model. In this model, the efficiency value can exceed 1 so that the subsequent analysis is more effective (Zhang et al., 2022). Many researchers have studied the influencing factors of TCERE (Cui et al., 2015). Understanding the influencing factors is important, because the government can use this information to formulate appropriate policies, balancing the economic development against TCERE. TCERE is an important indicator of environmental quality. According to the environmental Kuznets curve (EKC) hypothesis the economic growth is associated with environmental quality (Churchill et al., 2018). Thus, our paper mainly aims to examine the relationship between economic growth and TCERE. When testing the EKC hypothesis, many studies use per capita GDP to characterize the economics growth (Yan et al., 2020). However, it is unavoidable that certain mistakes may occur during the collection and compilation of GDP statistics (Xu et al., 2015;Wang et al., 2020). Additionally, the GDP data might be overstated because of the strive for economic growth success (Qin et al., 2019). Therefore, more and more researchers are using nighttime lights data to replace GDP when studying the economy. Chen et al. (2021) studied the relationship between light intensity and per capita income and found it is feasible to estimate living standards using nighttime lights. Gu et al. (2022) used DMSP/OLS and NPP/VIIRS nighttime lights data to predict economic development. Thus, we also use per capita nighttime lights brightness instead of GDP to characterize economic growth. The benefits of this approach are many. First, nighttime lights offer a consistent and long-lasting data source (Tan et al., 2018). Moreover, nighttime lights objectively capture economic activity and can represent information that typical GDP statistics cannot express (Qin et al., 2019). In this paper, the TCERE will be calculated using a super SBM model with undesirable outputs. Besides, the per capita brightness of nighttime lights, as well as other influencing factors, will be used to examine the impact on TCERE.

Input and Output Variables
In order to calculate TCERE, this article posits capital, labor force and energy as the inputs, gross domestic product (GDP) as the desirable output and carbon dioxide (CO2) emissions as the undesirable output, in accordance with previous studies (Zhao et al., 2020;. Definitions of these variables are summarized in Table 1. The detailed meanings of these variables are explained as follows. (1) Capital The indicator we chose to characterize capital is the capital stock of transport sector. Capital stock is measured in 10 8 CNY, which is calculated using the following formula: where denotes the capital stock of transport sector in year , denotes the depreciation rate which is set to 8.76%  and denotes the investment in fixed assets of transport sector.
(2) Labor force The indicator we choose to characterize labor force is the total number of transport employees in urban non-private units, private enterprises and self-employed individuals at year-end.
(3) Energy The indicator we chose to characterize energy is the total energy consumption of transport measured in 10 4 tons of coal equivalent, which is calculated using the following formula: where t stands for year; i stands for energy type; Et' is the total energy consumption of transport measured in coal equivalent; Eit is the energy consumption of transport measured in original physical unit; and ci is the conversion factors from physical unit to coal equivalent, which is from the China Energy Statistical Yearbook, as shown in Table 2. (4) GDP The indicator we used to characterize GDP is the GDP of transport sector, which has been transformed into 2010 price using deflators.
The indicator we chose to characterize CO2 is the CO2 emissions of transport sector, which is calculated using the following formula: where k=29.307 GJ/tonCE and αi stands for carbon dioxide emission factors, which is from the 2006 IPCC Guidelines for National Greenhouse Gas Inventories, as shown in Table 3.
Yan, Y., Li, J., Archives of Transport, 66(2), 21-32, 2023 25 Suppose DMUo is an alternative decision making unit where o ranges over 1, 2, ..., n. Let xo be the input data for DMUo, yo g be the desirable output data for DMUo and yo b be the undesirable output data for DMUo, which are defined below.
In order to evaluate the super-efficiency value * of DMUo, a convex optimization in , , , is given below. * = , , , where m=3, s1=1, s2=1 in this paper. This model is called the super SBM with undesirable outputs. In this model, the efficiency value can be higher than 1. To implement a plausible approach of evaluating TCERE, we adopt the super SBM with undesirable outputs model in accordance with previous studies (Zhang, 2018), because it can be more effective.

Influencing factors
This paper also posits economic growth, transport structure, technology level and population density as influencing factors. Definitions of these influencing factors are summarized in Table 4. In these influencing factors, PN (per capita nighttime lights brightness) is a new indicator for economic growth in individual provincial region.
Here the PN value is the total DN (Digital Number) value of nighttime lights image in individual provincial region divided by population.

Econometric Models
The EKC model provides a theoretical framework to study the relationship between economic growth and environment (Yan et al., 2020). Thus, this paper extends the EKC model and establishes the following four econometric models: where i denotes the data for the i-th region and t for the t-th year; denotes the constant term; 1 − 5 denotes a series of unknown parameters; denotes an entity fixed effect; denotes a time fixed effect and is the error term.

Technical Route
The process of calculating the TCERE and analyzing its influencing factors mainly consists of the following steps. 1. Acquiring raw data from relevant yearbooks, remote sensing images, etc. 2. According to Section 3.1, the raw data are processed to obtain the input and output variables. 3. The super SBM with undesirable outputs model is then used to evaluate the TCERE. 4. As described in Section 3.3, the raw data are processed to obtain the influencing factors. Analyzing the relationship between TCERE and influencing factors based on econometric models. It is also the most economically and culturally developed region in China, with well-connected railways, rivers, roadways, and shipping. Around 8.7% of China's total land area is located in East China (Kang, 2022). East China also ranked first among the seven major geographical regions of China in terms of GDP (43.8 trillion CNY) and GDP per capita (103,000 CNY) in 2021 (Kang, 2022). Fig. 1 shows the map of study area. The shapefiles of maps used in this paper are sourced from National Geomatics Center of China.

Data Sources
The raw data for calculating input and output variables are collected from the China Statistical Yearbook, China Price Statistical Yearbook, China Energy Statistical Yearbook, etc., in each year. Descriptive statistics for input and output varia-bles are reported in Table 5.
In this paper, the PN (per capita nighttime lights brightness) is a new indicator for the influencing factor of economic growth. To calculate this influencing factor, the nighttime lights remote sensing images of the Visible and Infrared Imaging Suite (VIIRS) were used here. The information about VIIRS nighttime lights images is available at (https://eogdata.mines.edu/products/vnl/) accessed on 10 June 2022. As for the product type, we select the "Annual VNL V2: average-masked". The relevant images can be downloaded at (https://eogdata.mines.edu/nighttime_light/annual/v20/) accessed on 10 June 2022. Fig. 2 shows the nighttime lights brightness images, and the PN value in individual provincial region is presented in Fig. 3.   Table 6.

TCERE
According to the super SBM with undesirable outputs model described in Section 3.2, the TCERE is computed. Fig. 4 depicts the spatial distribution of TCERE in East China from 2013 to 2017. As illustrated in Fig. 4, TCERE decreased almost completely between 2013 and 2017. TCERE in Shanghai has always been at its lowest level, regardless of the year.The statistics for TCERE are shown as box plots in Fig. 5, where we can see that the mean TCERE has been declining from 2013 to 2017, but the trend of decline has got slower, indicating a sign of improvement.

Influencing Factors Analysis
After calculating the TCERE, this paper also conducts an econometric analysis to study the relationship between TCERE and its influencing factors.
Four different econometric models are considered, and the estimation results are reported in Table 7. As shown in Table 7, the log-likelihood (Log-L) value of OLS2 is greater than others, and the AIC value is less. Hence, the OLS2 model seems more suitable to describe the relationship. Turning to estimation results of OLS2 model in Table 7, the coefficient of PN 2 is positive and statistically significant (p<0.05), and the coefficient of PN is negative and also statistically significant (p<0.05). It suggests that the relationship between TCERE and economic growth shows an U-shape. Thus, promoting the economic growth is a good way to improve the TCERE. The coefficient of TS is positive and statistically significant, indicating that a better transport structure tends to raise the TCERE. The coefficient of TL is also positive and statistically significant, expressing that a higher technology level lead to an increase in the TCERE. The coefficient of PD is not statistically significant, implying that the population density might not be the main influencing factors of TCERE.

Conclusion
The transport carbon emissions reduction efficiency (TCERE) is essential for climate change. In order to analyze the TCERE more effectively, this paper has calculated the TCERE based on the model of super SBM with undesirable outputs, and then proposed econometric models to analyze its influencing factors. Previous researchers have hardly studied these influencing factors using nighttime lights images, and that is the main contribution of our work. An empirical study was conducted in East China from 2013 to 2017, and this study has found that the relationship between TCERE and economic growth shows an U-shape. Besides, transport structure and technology level both show a positive impact on TCERE. The implications of our findings are given below. − The TCERE declines slower in East China, giving us reason to believe that the improvement of TCERE is predictable. − When economic growth exceeds the turning point, economic growth is conducive to the improvement of TCERE. We could develop the economy boldly and confidently. − Increased investment in railway and waterway transportation infrastructure projects is needed to strengthen the structure of the railway and waterway transportation systems. Furthermore, the general public and businesses should be encouraged to prefer rail or river transportation. − Investment in scientific and technological innovation should be enhanced in order to produce more efficient energy-use methods.
The preceding conclusions are also subject to some limitations. For example, we avoid the epidemic factor. More research is needed to better understand how epidemics affect the TCERE. In the future, we could also investigate the practical applications of our research in carbon-reduction activities.