Time-Space Evolution and Drivers of CO₂ Emissions from Land Utilization in Xinjiang from 2000 to 2020

Abstract

One of the key elements influencing carbon emissions is changes in land use, which affects the roles of carbon sources and sinks. We calculated the carbon emissions from land use in Xinjiang, constructed a Geographically and Temporally Weighted Regression model (GTWR), and investigated the spatio-temporal evolutionary trajectory and heterogeneity of carbon emissions based on the land use data of three periods from 2000 to 2020 and the socio-economic and energy data of the same period. The results show the following: ① in Xinjiang, the area of water, forests, grasslands, and idle land declined between 2000 and 2020, while the area of construction and agricultural land increased; ② Xinjiang’s land-use-related carbon emissions are rising annually. The primary driver of this expansion is the region’s growing area of construction land; ③ the degree of economic development, the amount of land used for building, and the degree of urbanization are the primary factors influencing carbon emissions in Xinjiang; and ④ In Xinjiang, there is a clear spatial heterogeneity in the factors that influence carbon emissions. Additionally, different influencing factors have different effects on emissions, depending on the region and stage of development, indicating that the level of resources and the region’s economic center of gravity have different effects on emissions.

1. Introduction

One of the biggest issues facing humanity in the twenty-first century is climate warming, and addressing climate change has drawn attention from all throughout the world. The IPCC report from 2021 states that carbon dioxide emissions are the main cause of the unmistakable warming caused by human activity [1].

Cities with high population densities are a significant contributor to high energy use and carbon emissions. Global carbon emissions have expanded dramatically as a result of the acceleration of urbanization; the carbon emissions produced by cities, which make up just 2% of the world’s land, account for 75% of all carbon emissions globally [2].

Based on figures published by the International Energy Agency, China surpassed all other countries in 2009 to become the largest contributor of carbon emissions globally [3]. Consequently, China’s carbon emissions have garnered significant international scrutiny.

Land serves as the conduit for the release of carbon from terrestrial ecosystems and human activities, and is a significant determinant of carbon emissions [4]. Both land utilization and alterations in land utilization structure significantly influence carbon absorption and emission [5,6]. Unscientific land utilization diminishes the ability of soil to sequester carbon [7], resulting in an escalation of carbon emissions.

Structural land-use change is now the second most significant factor in the rise of global carbon emissions, ranking just below fossil fuel combustion [8]. Between 1850 and 1998, it is estimated that one-third of the total carbon emissions induced by human activities were attributed to land use changes [9]. In the future, human activities that produce greenhouse gas emissions will increasingly affect the land space [10]. As a result, the land system will play a crucial part in the feedback loop of human society.

In recent years, researchers, both domestically and internationally, have conducted extensive studies on carbon emissions resulting from land use. These studies primarily concentrate on methods for accounting for regional carbon emissions, factors that influence carbon emissions, the spatial and temporal characteristics of carbon emissions, and the establishment of regional carbon compensation zones.

In the field of regional studies, researchers used the carbon emission coefficient method to estimate the carbon emissions from land use in Hubei Province. Their findings revealed that the primary sources of carbon emissions are construction land and cropland. They also observed that the spatial fluctuation of carbon emissions decreases from east to west and from north to south [11]. The analysis of land-use conversion data reveals that in 2020 almost 66% of Brazil’s carbon dioxide emissions are attributed to land-use change, significantly affecting the carbon footprint of Brazilian agricultural products [12]. Furthermore, certain studies have utilized the “Thin Record” model to approximate the carbon emissions from terrestrial ecosystems in China throughout the last 300 years [13]. Other studies have employed the STIRPAT model to examine the impact of population, affluence, and technology on carbon emissions in 125 countries. These studies have determined that affluence is the primary factor influencing carbon emissions on a global scale, with technology and population being secondary factors [14].

In research methodology, the researchers employed the emission factor approach to compute carbon emissions, developed a GIS-based kernel density model, and utilized the outcomes of carbon emission accounting and land use zoning to create a multilayer perceptron model [15]. Furthermore, they compiled a specific inventory of countries, processes, greenhouse gases, and products emitted from global land use over time. Additionally, the researchers extensively utilized GIS and RS technology to measure the spatial and temporal changes in carbon emissions from various types of land use [16].

Previous studies have demonstrated that carbon emissions are significant when forest land is converted to other land use types. Conversely, the transformation of other land use types into forest land results in the highest carbon sink. Several scholars have examined the factors that influence carbon emissions in Austria and Czechoslovakia using Kaya’s constant equation. The primary factors considered are population, energy intensity, and energy composition [17]. Additionally, researchers have analyzed the relationship between land use change and energy carbon emissions using the GWR model to demonstrate how reducing the energy carbon footprint can help alleviate environmental pollution [18].

Overall, the research on the patterns and changes in carbon emissions from land use primarily examines national and provincial levels. The factors influencing these emissions are predominantly related to socio-economic aspects such as the economy, population, and industry. To analyze these factors, researchers commonly employ methods like the STIRPAT model, the Kaya equation, the LMDI decomposition method [19], and GTWR analysis [20]. Nevertheless, the aforementioned studies fail to encompass the carbon emission attributes of all land use categories. Moreover, domestic researchers have primarily focused on investigating carbon emissions from land use in the central and southern regions of China, as well as the three northeastern provinces. Consequently, there is a scarcity of research on the spatial and temporal patterns of carbon emissions from land use and the factors that influence land use in the arid zones’ interior [21].

With the in-depth implementation of the “Western Development” and “One Belt, One Road” strategy, the economy of the Xinjiang region (Figure 1) in China has been developing rapidly, but this rapid economic growth is mainly supported by the traditional economic growth mode of three highs (high investment, excessive energy consumption and significant pollutant levels) in terms of growth to support. This kind of rough growth not only caused the waste of energy, but also made the resources and environmental pressure increase dramatically. In the current situation of vigorously developing a low-carbon economy, ensuring Xinjiang achieves sustainable economic development while simultaneously reducing carbon dioxide emissions is crucial, not only for the region’s economic and environmental coordination, but also for the overall realization of China’s low-carbon economy [22].

This paper aims to analyze carbon emissions from land use in Xinjiang by utilizing three-period land use data from 2000 to 2020, along with socio-economic and energy data from the same period. The study will involve constructing a GTWR model to investigate the changing patterns in land use, the spatio-temporal evolution of carbon emissions, and the spatial and temporal variations in the factors influencing them. The following is a diagram of the research structure of this paper (Figure 2).

2. Materials and Methods

2.1. Data Sources

The digital elevation model is created from 18 topographic maps on a scale of 1:10 million (1:6 million for specific regions) that span the entire world. These maps were produced by the International Hydrographic Organization (IHO) and the Intergovernmental Oceanographic Commission (IOC) in collaboration with the relevant countries, and have a spatial resolution of around 500 m.

The primary data needed for this study consist of land-use raster data and socio-economic data from the years 2000, 2010, and 2020. The land-use raster data were acquired from the Resource and Environmental Science and Data Center of the Chinese Academy of Sciences, with a spatial resolution of 30 m.

The land use categories in the study region were categorized into six main classes: farmland, forest land, grassland, watershed, construction land, and unutilized land, based on the standard “Current Land Use Classification” (GB/T21010-2017) [23].

The Statistical Yearbook (2000~2020) produced by the Xinjiang Bureau of Statistics and the China Energy Statistical Yearbook (2000~2020) released by the National Statistical Department provided the socio-economic data and energy consumption data, respectively.

The data needed to calculate carbon emissions from arable land include agricultural fertilizer usage, crop area, agricultural machinery power, and irrigated area. For carbon emissions from land used for construction, the data needed include mineral energy consumption like petroleum, coal, and natural gas. To analyze carbon emission indicators, the data needed include total population, total production value, and total energy consumption.

The administrative demarcations of each prefecture in Xinjiang Uygur Autonomous Region are established according to the year 2021.

2.2. Research Methodology

Various land use categories exhibit distinct exploitation behaviors, resulting in varying carbon emission impacts. This study calculates carbon emissions based on the utilization characteristics of various land use types and categorizes land use carbon emissions into two main groups: carbon sources and carbon sinks.

Construction land and cropland are identified as carbon sources, while cropland, forest land, grassland, watersheds, and unutilized land are identified as carbon sinks.

Cropland and building land are significant contributors to carbon emissions, mostly due to human production and operation activities and industrial energy consumption. However, their carbon emissions need to be indirectly accounted for, unlike other land use types which may be directly accounted for.

2.2.1. Direct Carbon-Emission-Accounting Models

Referring to the existing studies and considering the latitude and longitude position and geographic location status of Xinjiang, this paper adopts the fixed coefficient method to estimate the carbon sinks of cropland, forest, grassland, waters, and unutilized land (

Table 1. Carbon emission factors for each land use type (unit: tons of carbon emissions per square hectare).

Land Use Types	Carbon Emission Factor tC/(hm2-a)	References
cropland	0.007	Chen Panqin, Wang Xiaoke et al. [24]
forest	0.644	Niu Yawen, Zhao Xianchao et al. [25]
grassland	−0.021	Xu Ying, Guo Nan et al. [26]
water	−0.253	Sun He, Liang Hongmei et al. [27]
unutilized land	−0.005	Lai Li [28]; Sun Shuangmei [29]

), and adopts the direct accounting method to estimate the carbon sink capacity of the study area, using the subsequent mathematical equation:

$$C_i={\displaystyle \sum _{\mathrm{i}=1}^5{S}_i}{f}_i$$

(1)

where $C_i$ denotes the carbon emissions associated with various land use kinds; $S_i$ denotes the area of the ith specific land use type; and $f_i$ denotes the carbon emission factor of the ith land use type.

2.2.2. Indirect Carbon-Emission-Accounting Models

Cropland manifests itself as a source of carbon when it is used as a carrier for human activities in agricultural production and operation, such as agricultural fertilizer application, use of agricultural machinery, and irrigation [30], and is calculated by the following formula:

$$E_t=E_f+E_m+E_i$$

(2)

where ${ E}_t$ is the emissions from cropland and ${ E}_f$, $E_m$, and ${ E}_i$ are the carbon emissions from the production of fertilizers for farmland, the agricultural machinery production and utilization, and the irrigation process, respectively.

Equation (2) can be decomposed as:

$$E_t=G_{f}A+(A_{m}B+W_{m}C)+A_{i}D$$

(3)

where A, B, C, D are conversion coefficients and their values refer to the study of Zhang Junfeng et al. [31], which are 857.54 kg/mg, 16.47 kg/hm², 0.18 kg/kW, and 266.48 kg/hm²; $G_f,A_m,W_m, {\mathrm{a}\mathrm{n}\mathrm{d} A}_i$ are agricultural fertilizer use, crop cultivation area, the aggregate power of agricultural machines, and the irrigated area, respectively.

The carbon emissions associated with building land primarily stem from the consumption of industrial energy, as it serves as a significant carbon source. The carbon emissions from building land are evaluated using the carbon emission coefficient of energy consumption provided in the 2006 IPCC Guidelines for National Greenhouse Gas Inventories [32]. The calculation formula is as follows:

$$E_y=\frac{44}{12}\times {\displaystyle \sum E_i\times {\theta }_i \times }{\mu }_i$$

(4)

where $E_y$ denotes the energy consumption’s carbon emissions; ${ E}_{i }$ is the consumption of the ith energy source;${ \theta }_i$ is the standard coal conversion factor of the ith energy source; ${ \mathrm{a}\mathrm{n}\mathrm{d} \mu }_i$ is the carbon emission coefficient of the ith energy source. The standard coal conversion coefficient and carbon emission coefficient (

Table 2. Standard Coal Conversion Factors and Carbon Emission Factors for Different Energy Types (unit: kilograms of standard coal; tons of standard coal).

Type of Energy	Conversion Factor for Standard Coal kgce/kg	Carbon Emission Factor t/tce	Type of Energy	Conversion Factor for Standard Coal kgce/kg	Carbon Emission Factor t/tce
raw coal	0.714	0.756	gasoline	1.471	0.571
refined coal	0.900	0.756	diesel fuel	1.457	0.592
other coal washing	0.285	0.216	fuel oil	1.429	0.619
coke	0.971	0.855	liquefied petroleum gas	1.714	0.504
coke oven gas	0.571	0.355	refinery dry gas	1.571	0.460
blast furnace gas	0.129	0.355	other petroleum products	1.400	0.586
petroleum	1.330	0.448	thermodynamic	0.034	0.670
crude oil	1.429	0.586	electrical power	0.345	0.272
diesel	1.471	0.554	other fuels	0.645	0.756

) mainly refer to the China Energy Statistical Yearbook and the IPCC Carbon Emission Calculation Guidelines (2006).

2.2.3. Indicator Dimensionless Processing

Since the evaluation indicator units are not uniform and all the indicators need to be combined to jointly evaluate their impact on carbon emissions in the study area during the study period, it is necessary to standardize the study indicators, and the formula for calculating the Z-score transformation is as follows:

$$Z=\frac{X-\epsilon }{\sigma }$$

(5)

where $Z$ denotes the standardized value, $X$ denotes the observed value, $\epsilon$ denotes the average of the entire dataset, and $\sigma$ denotes the statistical measure of the dispersion or variability of the entire dataset.

2.2.4. Geographically and Temporally Weighted Regression Model (GTWR)

The Geographically and Temporally Weighted Regression model is a specific multiple linear regression model for identifying and quantifying the effects of spatial and temporal factors on a variable, which is proposed on the basis of the geographically weighted regression model. The regression parameters of the independent variables vary with space in the GWR model, whereas they vary with location in space-time in the GTWR.

Thus, this model provides a more accurate depiction of the spatial and temporal correlation between factors and dependent variables compared to GWR.

Since the carbon emission in Xinjiang presents uneven spatio-temporal distribution characteristics and carbon emission has a spatio-temporal non-stationarity, it is necessary to explore the spatio-temporal heterogeneity of carbon emission locally when studying the effect of carbon emission in Xinjiang. The GTWR model can be expressed as follows [33]:

$$y_i={\beta }_0(u_i,v_i,t_i)+{\displaystyle \sum {\beta }_k(u_i,v_i,t_i)}{X}_{ik}+{\epsilon }_i$$

(6)

where $\left(u_i,v_i\right)$ denotes the latitude and longitude of the ith sample point; ${ t}_i$ denotes the time of observation; ${ y}_i$ denotes the value of the dependent variable for the ith sample point; and $X_{ik}$ denotes the kth explanatory variable for the ith sample point. ${ \epsilon }_i$ is the model error term. ${ \beta }_0\left(u_i,v_i,t_i\right)$ represents the regression constant for the ith sample point, and ${ \beta }_k\left(u_i,v_i,t_i\right)$ represents the regression coefficient of the kth explanatory variable for the ith sample point. In this study, carbon emissions are considered as the dependent variable, while six categories of factors, specifically the industrial structure, economic development level, urbanization level, population agglomeration level, land use for construction ratio, and technology level, are chosen as independent variables [30].

3. Results

3.1. Spatio-Temporal Dynamic Analysis of Land Use Change in Xinjiang from 2000 to 2020

The findings of the examination on the fluctuating alteration of land utilization in Xinjiang between 2000 and 2020 are presented in

Table 3. Changes in degree of dynamics for land-use-type dynamics (in %).

Year	Sports Event	Cropland	Forest Land	Grassland	Waters	Constructed Land	Unutilized Land
2000~2010	rate of change	35.83	−25.94	1.77	−35.53	83.70	−0.48
	dynamic degree	3.58	−2.59	0.18	−3.55	8.37	−0.05
2010~2020	rate of change	11.62	−2.56	−0.61	3.45	12.66	−0.79
	dynamic degree	1.16	−0.26	−0.06	0.35	1.27	−0.08
2000~2020	rate of change	51.62	−27.83	1.15	−33.30	106.96	−1.27
	dynamic degree	2.58	−1.39	0.06	−1.67	5.35	−0.06

and Figure 3. From 2000 to 2010, the study area experienced an overall increase in the area of cropland, grassland, and construction land, while the area of forest land, waters, and unutilized land decreased. Notably, the largest increase was observed in construction land, with a significant rate of change of 83.70% over the 10-year period; the degree of dynamics was 8.37%, increasing at a rate of 2.91 × 10⁴ hm²·a. the area of waters had the biggest decrease, with a rate of change of −35.53% during the 10-year period and a motivation of −3.55%, with a rate of decrease of 14.34 × 10⁴ hm²·a.

From 2010 to 2020, the cropland, waters, and constructed land all showed an increasing trend, and the forest land, grassland, and unutilized land showed a decreasing trend, among which the constructed land had the biggest increase, with a rate of change of 12.66% during the 10-year period. The dynamic degree was 1.27%, increasing at the rate of 0.83 × 10⁴ hm²·a, and the forest land had the largest decrease, with the rate of change over 10 years of −2.56%, and a dynamic degree of −0.26%, decreasing at the rate of 0.58 × 10⁴ hm²·a.

Compared with 2010~2020, the area change of each type of land use during 2000~2010 is large, mainly manifested in the inflow and outflow of grassland and unutilized land; about 12% of unutilized land becomes grassland, and about 23% of grassland becomes unutilized land.

During 2000~2020, the cropland, grassland, and construction land show an increasing trend, and the forest land, waters, and unutilized land show a decreasing trend. Among them, the increase in construction land is the most obvious, with the rate of change of 106.96% over 20 years. The degree of dynamics is 5.35%, increasing at the rate of 1.87 × 10⁴ hm²·a; the decrease in water area is larger, with the rate of change of −33.30% over 20 years, and the degree of dynamics is −1.67%, decreasing at the rate of 6.72 × 10⁴ hm²·a.

Figure 3 and Figure 4 illustrates the changes in land use types in Xinjiang between 2000 and 2020. The most prominent spatial change during this period was the transformation of unutilized land into grassland, primarily occurring in the Altai Mountains, Tianshan Mountains, and Kunlun Mountains.

Additionally, there was a conversion of grassland into cropland and unutilized land, with the majority of the unutilized land conversion zones concentrated in the southern region of Xinjiang, surrounding the Tarim Basin.

The majority of unused land conversion areas are concentrated in specific regions of Aksu, Kashgar, Hotan, Bayangol Mongol Autonomous Prefecture, and Changji Hui Autonomous Prefecture.

Cropland conversion areas are primarily found in the north-central part of Kashgar Region, the north-central part of Aksu, and the southern part of Tacheng Region.

Areas where water is converted to unused land are mainly located in the northern and southern slopes of the Tianshan Mountain range and the western part of the Kunlun Mountain range.

Between 2010 and 2020, the most noticeable change in land use types was the transformation of unused land into grassland. This change was primarily observed in the eastern part of Altay Region, the southeastern part of Changji Hui Autonomous Prefecture, and the southeastern part of Bayangol Mongol Autonomous Prefecture.

Additionally, there was a significant conversion of grassland into cropland, mainly occurring in the northeastern part of Kashgar Region, the central part of Aksu Region, the southern part of Tacheng Region, and the central part of Changji Hui Autonomous Prefecture.

The spatial transformation of land use in Xinjiang between 2000 and 2020 can be summarized as the conversion of unused land into grassland, grassland into cropland, and unused land and water into unused land. The changes in land use during the period from 2000 to 2010 were more significant compared to the period from 2010 to 2020. The areas that remained unchanged in terms of land use were primarily unutilized land, located in the Taklamakan Desert, Gurbantunggut Desert, and Kumutag Desert.

3.2. Analysis of Carbon Emissions from Land Use in Xinjiang from 2000 to 2020

The carbon sources, sinks, and net carbon emissions in Xinjiang from 2000 to 2020 were determined based on Equations (1), (3), and (4). Analysis of

Table 4. Carbon emissions from various land uses in the study area (unit: 10⁴ tons of carbon emissions).

Particular Year		2000	2010	2020
Carbon Sinks	cropland	−4.15	−5.64	−5.64
	forest	−245.01	−181.46	−176.82
	grassland	−100.92	−101.79	−101.17
	waters	−130.94	−84.41	−87.34
	unutilized land	−50.01	−49.76	−49.37
	carbon absorption total amount	−530.13	−423.07	−420.33
Carbon Sources	cropland	88.26	116.61	139.23
	constructed land	1195.27	3224.23	14,149.60
	carbon emission total amount	1283.53	3340.84	14,288.84
Net Carbon Emission		753.40	2917.77	13,868.51

reveals that there is a direct correlation between the year and net carbon emissions, with a consistent and substantial increase over time. Specifically, the net carbon emissions rose from 7,534,000 t in 2000 to 138,685,100 t in 2020, resulting in a net increase of 131,151,100 t. This represents a staggering 18-fold increase, with the most significant growth occurring between 2010 and 2020, during which the net carbon emissions surged by 109,507,400 t. This translates to an average annual growth rate of 37.53%.

From a carbon source perspective, construction land is the primary contributor to carbon emissions, accounting for over 90% of the total emissions. These emissions have been rapidly increasing over the years. On the other hand, while carbon emissions from cropland have shown a steady increase, their proportion has decreased from 6.9% to 0.97%. This suggests that the role of cropland as a carbon source is gradually diminishing, with construction land emerging as the primary source of carbon emissions from land use.

The carbon sinks in Xinjiang had a consistent fall of approximately 1,098,000 tons between 2000 and 2020. This loss was particularly notable between 2000 and 2010, with a decline rate of 25.31%. Of all the land types, forest land has the highest capacity to absorb carbon, accounting for approximately 40% or more. However, over time, this capacity gradually decreases. This is primarily because around 45% of forest land was converted to grassland between 2000 and 2010, and the overall area of forest land has been decreasing. As a result, the carbon sink capacity has also decreased. The carbon sink of the watershed saw a net decline of 436,000 metric tons during the period from 2000 to 2010. Additionally, approximately 37% of the water resources were converted to unutilized land. The carbon storage capacity of farmland, grassland, and unutilized land remained largely constant.

The spatial distribution of carbon emissions and carbon sinks was determined based on the direct mapping of land use categories. The density of carbon emissions and carbon sinks was then estimated by considering the area of each land use category. The corresponding results can be observed in Figure 5 and Figure 6.

The findings depicted in Figure 5 indicate that during the initial phase of the study, areas with high carbon emissions from land use were primarily concentrated in the central urban regions of Urumqi and other Xinjiang prefectures. Subsequently, these areas gradually expanded towards the outskirts of the urban centers after 2010. In 2010, there was an 83.49% increase in the number of high carbon emission areas compared to 2000.

Notably, these areas were predominantly located in the periphery of the urban centers. Throughout the study period, the carbon emission density of cropland exhibits a fluctuating pattern, initially decreasing and subsequently increasing.

Conversely, the carbon emission density of construction land experiences a significant increase, of 472.15%, over the same period. This leads to the conclusion that the carbon emission resulting from construction land surpasses the contribution of cropland to carbon emission.

Observing the spatial change of the carbon sink density in Xinjiang (Figure 6), the results showed that high carbon-sink land-use patches were mainly distributed in the Altay Mountains, and on the north and south slopes of the Tianshan Mountains and the Kunlun Mountains in Xinjiang. During the study period, there was no significant change in carbon sink density as a whole; the total area of carbon sink patches decreased by 0.27%, and the average carbon sink density decreased by 0.08%.

The spatial-distribution and change trajectories of carbon emissions from land use in Xinjiang from 2000 to 2020 are shown in Figure 7. It can be observed that, in the case of carbon emissions increasing year by year from 2000 to 2020, the spatial-distribution pattern radiating from the central and northern regions to the surrounding cities is presented.

The areas of high carbon emissions and changes in carbon emissions are centrally distributed in the city of Urumqi and the Changji Hui Autonomous Prefecture. Low-value areas are concentrated in the Kizilsu Kirgiz Autonomous Prefecture, Hotan and Kashgar regions.

3.3. Analysis of Carbon Emission Drivers

3.3.1. Data Testing

This work employs the Geographically and Temporally Weighted Regression model (GTWR) to investigate the influential factors behind carbon emissions from land use and their geographical dynamics in Xinjiang. The six categories of factors were initially assessed for covariance and subsequently examined for significance. The results revealed an anomaly, which can be attributed to the excessively high value of the urban development level in Karamay. Upon excluding Karamay City from the analysis, it was observed that the six types of factors significantly influence carbon emissions. Furthermore, the variance inflation factor (VIF) was found to be less than 10, indicating the absence of multicollinearity.

Table 5. Research indicators and their interpretation.

Indicator Name	Hidden Meaning	VIF Value
Urbanization Level (of a city)	urban resident population/total population	3.757
Industrial Structure	secondary sector output/gross output	1.648
Level of Economic Development	GDP per capita	2.171
Degree of Population Concentration	permanent population per unit area	1.801
Technical Level	energy consumption per unit of GDP	1.567
Percentage of Constructed land	proportion of constructed land to urban area	2.870

displays the variance inflation factor and index explanation for each variable.

The GTWR_Beta plug-in, developed by Prof. Bo Huang’s team at the Chinese University of Hong Kong, utilizes the ArcGIS platform to automatically determine the ideal bandwidth and set the spatio-temporal distance-parameter ratio to 1. The regression coefficients of the affecting factors were evaluated and computed. The AIC criterion and the goodness-of-fit R² were chosen as the evaluation metrics for the model’s confidence. The findings may be found in

Table 6. GTWR-related parameter settings.

Model	Bandwidth	Residual Squares	Sigma	AICc	R2	R2 Adjusted
GTWR	0.196	0.607	0.125	238.549	0.984	0.981

. Regarding the goodness-of-fit, both R² and the R² adjusted exceeded 0.95, indicating that this model captures well the impact of the explanatory factors on the dependent variables.

3.3.2. Analysis of Temporal Heterogeneity of Regression Coefficients of Factors Affecting Carbon Emissions from Land Use

The regression coefficients of the factors affecting carbon emissions in each state of Xinjiang were plotted as box plots according to the year (Figure 8 and Figure 9) to explore the temporal heterogeneity.

During the study period, there was a consistent increase in carbon emissions in Xinjiang as a result of urbanization. The outcome is the conversion of a limited portion of agricultural and ecological land, including forests, grasslands, and unused land, into construction land. This conversion diminishes the capacity of these areas to absorb carbon, resulting in a reduction of carbon sinks.

The amount of economic development throughout time has a large and positive impact on carbon emissions, much like the level of urbanization.

The level of population concentration has a direct correlation with carbon emissions. Over time, most regions have experienced an increase in population concentration coefficients.

The degree of technology has a predominantly beneficial impact on carbon emissions, leading to a progressive increase over the course of the study period.

The ratio of developed land has a predominantly favorable impact on carbon emissions, exhibiting a consistent pattern of gradual growth over time.

Throughout the study period, there was an observed pattern of carbon emissions being influenced by the industrial structure, with an initial upward trend followed by a subsequent increase.

3.3.3. Analysis of Spatial Heterogeneity of Regression Coefficients of Factors Affecting Carbon Emissions from Land Use

The regression coefficients of the influencing factors were graphically represented using ArcGIS, resulting in spatial- and temporal-distribution maps for the six categories of influencing factors.

• Urbanization level

The findings indicate that there is a correlation between the degree of urbanization and carbon emissions, as depicted in Figure 10. Specifically, the data reveal a trend of higher carbon emissions in the southwest region and lower emissions in the middle east region.

The primary factor contributing to carbon emissions in Kizilsu Kirgiz Autonomous Prefecture, and Kashgar and Hotan regions is the amount of urbanization, whereas Urumqi and Changji Hui Autonomous Prefecture have comparatively lower levels of urbanization and thus lesser contributions to carbon emissions. The correlation coefficients of most cities decreased over time, transitioning from positive to negative.

2.

Industrial structure

The regression coefficients of industrial structure (Figure 11) exhibit a pattern of high values on both ends and low values in the middle from 2000 to 2010. In 2020, there is a distinct pattern of high values in the western region and low values in the eastern region.

The regression coefficients for Urumqi, the capital of Xinjiang Province, and the adjacent Changji Hui Autonomous Prefecture are lower. The Aksu and Hami regions are primarily affected by the industrial structure, with the Turpan and Bayangol Mongol Autonomous Prefectures following suit.

3.

Economic development level

The correlation between the degree of economic development and carbon emissions (Figure 12) demonstrates a gradual increase from the western regions to the eastern regions. The Bayangol Mongol Autonomous Prefecture has the most significant influence on carbon emissions in terms of economic development, followed by Urumqi and Changji Hui Autonomous Prefecture.

4.

Population agglomeration

The relationship between the degree of population concentration and carbon emissions, as depicted in Figure 13, exhibits a pattern of high values in the east and low values in the west.

The Altay region is most affected by population agglomeration, followed by the Turpan and Hami regions. With the exception of Kashgar, Hotan, Aksu and Kizilsu Kirgiz Autonomous Prefecture, where the regression coefficients decrease annually, the regression coefficients of the remaining prefectures increase as the years progress.

5.

Technology level

The relationship between technological level and carbon emissions (Figure 14) has high values on both ends and low values in the middle. The carbon emissions in Kashgar are mostly influenced by the level of technology, with Kizilsu Kirgiz Autonomous Prefecture following closely behind.

The emissions are comparatively lower in Aksu and Hotan. The coefficients of most cities have been gradually increasing over time. The regression coefficients for Kashgar and Kizilsu Kirgiz Autonomous Prefectures have changed from negative to positive.

6.

Proportion of construction land

The relationship between the proportion of construction land and carbon emissions (Figure 15) demonstrates a clear geographical pattern, with lower levels observed in the western regions and higher levels in the eastern regions.

The allocation of land for construction has the most significant influence on carbon emissions in the Altay region, with the Tacheng region and Changji Hui Autonomous Prefecture following suit. At the beginning of the study, construction land had a negative effect on carbon emissions in Aksu, Ili Kazakh Autonomous Prefecture, and the Hotan region.

4. Discussion

This paper examines the carbon emissions resulting from land use in Xinjiang over three time periods, categorized by different types of land use. It analyzes the changes in land use dynamics in Xinjiang over the past 20 years. This study looks into the patterns of spatial distribution of carbon emissions as well as the temporal and spatial fluctuations in different contributing components.

A study revealed notable variations in carbon emissions among the states in Xinjiang. Furthermore, there was a consistent annual increase in carbon emissions, with the most substantial rise observed in the carbon emission intensity of building land.

Due to the utilization of the fixed coefficient approach in this study to calculate carbon sinks, the size of each land use category remains relatively stable during the study period, resulting in few changes in carbon sinks.

4.1. Study of Variations in Carbon Emissions Resulting from Land Use

Between 2000 and 2020, the primary source of carbon emissions is construction land, mainly because of the fast economic growth in Xinjiang. This growth has resulted in the utilization and development of a substantial amount of unused land. Simultaneously, the consumption of fossil fuels in land development and utilization is expanding at a rapid pace, resulting in a notable increase in carbon emissions from construction sites.

The lifestyle and living standards of residents in cities in the northern border region, such as Urumqi, have been improved due to industrialization and urbanization, resulting in a significant rise in carbon emissions from building sites.

The southern border region is situated in a remote area with limited transit access. Its economic growth is less advanced, focusing mostly on animal husbandry, resulting in fewer carbon emissions compared to the northern border.

4.2. Variability in Factors Affecting Carbon Emissions from Land Use over Space and Time

At the start of the research period, Xinjiang was in the initial phases of urbanization, marked by land urbanization and haphazard economic development. Over time, urbanization has been speeding up due to strong economic growth, resulting in a higher consumption of fossil energy per unit of GDP growth. This has led to a rise in carbon emissions due to the increased demand for energy resources.

Some prefectures avoided haphazard land expansion during early development stages. However, over the years, issues such as excessive land expansion and low land-utilization rates have emerged. Advancements in production technology are leading to a transformation in the industrial structure and optimization of the energy consumption structure. This has resulted in a decrease in energy consumption per unit of GDP, leading to a progressive reduction in carbon emissions.

Population growth leads to higher energy consumption, impacting the region’s carbon emissions. However, as industries concentrate due to population agglomeration, resource utilization efficiency improves, resulting in reduced regional carbon emissions over time. The early phase of the study saw a quick increase in building land due to intensive use, resulting in a more efficient utilization of building land in the later phase.

As production activities in the secondary and tertiary industries continue to increase, intensifying the consumption of resources such as oil, coal and natural gas, the secondary industry is mostly concentrated in the southern border and other prefectures, indicating that these prefectures have undertaken more energy-consuming production activities, resulting in higher carbon emissions. However, with the transformation of the industrial structure and the country’s advocacy of a low-carbon economy, the center of gravity of the industry has gradually been shifted from the secondary to the tertiary industry, resulting in a gradual decline in carbon emissions.

5. Conclusions

This paper utilizes land use data and energy consumption data to analyze the estimation of carbon emissions, land use changes, and spatio-temporal differentiation of carbon emissions in Xinjiang from 2000 to 2020. The objective is to uncover the relationship between land employment and spatio-temporal shifts of carbon emissions in the study area. The findings indicate the following:

(1) From 2000 to 2020, there has been a notable shift in land use patterns in Xinjiang. Specifically, there has been a significant expansion in the areas designated for cultivation and building, while the areas of forest land, grassland, waterways, and unutilized land have experienced a decline. Grasslands and waters are primarily transformed into unused land, while unused land is primarily transformed into grasslands and areas for construction.

(2) The carbon emissions resulting from land usage in Xinjiang are consistently increasing on an annual basis. The primary contributor to carbon emissions is land used for construction, which accounts for over 90% of total emissions and exhibits significant growth over time. Conversely, forested land serves as the primary carbon sink, although its capacity to absorb carbon steadily diminishes with time.

Regarding spatial distribution, during the initial phase of the study, there was a notable concentration of land plots with high carbon emissions in the inner urban regions of Urumqi and other Xinjiang prefectures. However, after 2010, these emissions steadily extended towards the outskirts of urban centers. The distribution of high-carbon-sinking land use plots is mostly concentrated in the Altai Mountains, on the northern and southern slopes of the Tianshan Mountains, and in the Kunlun Mountains in Xinjiang.

(3) Between 2000 and 2020, urbanization, economic development, and the amount of construction land had a rising influence on carbon emissions. Conversely, population concentration and technological advancements had a declining effect on carbon emissions. The impact of industrial structure on carbon emissions will exhibit fluctuating changes.

The regression coefficients of the influencing elements of carbon emissions in Xinjiang exhibit significant regional variation, with the industrial organization being the most prominent influencing factor, followed by the proportion of construction land and technology. The effects of different influencing factors on carbon emissions vary across regions and developmental phases, suggesting that regional development disparities contribute to variations in the primary drivers of regional carbon emissions.

(4) The aforementioned analysis demonstrates that the primary determinants of carbon emissions in various states undergo noticeable fluctuations during different stages of development and at different points in time, indicating the presence of uneven development. Therefore, when devising strategies to reduce carbon emissions and allocating regional carbon emission reduction targets, it is imperative to consider the actual conditions and developmental characteristics of each region.