Spatio-Temporal Relationship and Transition Patterns of Ecosystem Service Value and Land-Use Carbon Emissions on the Loess Plateau

Abstract

Ecosystem services play a vital role in human well-being, with land-use changes exerting substantial influence on ecosystem service value (ESV) and land-use carbon emissions (LUCEs). Understanding the spatio-temporal relationship and transition dynamics between ESV and LUCEs is essential for promoting high-quality ecological development aligned with the “dual carbon” objective. This study takes the Loess Plateau as the research object. Based on five-phase land-use data from 2000 to 2020, the ESV and LUCEs are calculated. Exploratory spatio-temporal data analysis is used to explore their spatio-temporal relationship and transition paths, and the quadrant model is introduced to analyze the transition patterns from the perspective of ecological quality. The results indicate the following: (1) From 2000 to 2020, the ESV of the Loess Plateau increased from CNY 579.032 billion to CNY 582.470 billion, with an overall increase of only 0.15%. Among the changes in land use, changes in forest and grassland significantly affected the ESV. (2) The LUCEs from land use on the Loess Plateau increased from 137.15 Mt to 458.43 Mt, with an average annual growth rate of 6.22%. Affected by industrialization and urbanization, the LUCEs showed significant spatial differences at the provincial and county scales. (3) There was a certain positive spatial correlation between ESV and LUCEs. The distribution of significantly correlated areas did not change significantly from 2000 to 2020, and the relationship characteristics were mainly characterized by Type IV transitions. (4) At the county scale, ESV and LUCEs exhibited temporal stability, with most counties situated in the general ecological category, indicating substantial potential for enhancing regional ecological quality. These research outcomes offer a foundational framework for devising tailored regional carbon emission reduction strategies.

1. Introduction

Ecosystem services refer to various benefits directly or indirectly provided by ecosystems for human survival and development [1]. The United Nations Millennium Ecosystem Assessment classified ecosystem services into four major categories, namely provisioning services, regulating services, supporting services, and cultural services [2], clarifying the close relationship between ecosystem services and human well-being [3]. As an effective indicator for measuring ecosystem services, ESV plays an important role in conducting ecosystem monitoring and management and formulating ecological environment protection policies [4,5]. With economic development, the large-scale emission of anthropogenic greenhouse gases has intensified global warming [6,7], posing huge challenges to the sustainable development of the natural ecological environment and the economy and society [8,9]. Against the background of climate and environmental changes, LUCEs have become a hot topic of global concern [10,11]. As a key factor affecting environmental change, land-use changes not only directly affect ecosystem services [12] but also indirectly affect the process of LUCEs [13,14]. The feedback relationship between ESV and LUCEs reflects the synergy between regional ecological functions and carbon dynamics [15]. As the world’s largest carbon emitter, China’s carbon emissions reached 13.92 Mt in 2021, with carbon dioxide contributing up to 85.84% [16]. To mitigate carbon emissions and address climate change, the Chinese government has set the goals of achieving a “carbon peak” by 2030 and “carbon neutrality” before 2060. Therefore, under China’s “double carbon” strategy, balancing ecological conservation and development is crucial for the high-quality advancement of the Loess Plateau. In this study, we conduct a spatio-temporal quantitative assessment of ESV and LUCEs, examining their correlation over time and space. The findings provide critical insights for adjusting land-use structures, planning control schemes, and formulating differentiated ecological protection policies [17,18].

In recent years, the impact of land-use change on ESV and LUCEs has emerged as a significant focus in ecological research [19]. Studies on ESV have employed quantitative assessments using methods such as the equivalent factor method, the benefit transfer method, and the ESV index method [20]. These studies have concentrated on elucidating the feedback mechanisms between ecosystem services and land use, analyzing their interrelationships, and predicting future scenarios [21,22,23]. For instance, Rudolf et al. [22] demonstrated substantial variations in service value across different land-use types by estimating the ESV of ten major biomes globally. Research on LUCEs in China has predominantly examined various spatial scales, including the national [24,25], provincial [26,27], municipal [28,29], and county levels [30,31], focusing on analyzing the spatio-temporal evolution characteristics, influencing factors, action mechanisms, and effects of LUCEs [32,33]. Scholars have also investigated the relationship between LUCEs and economic development, leading to a shift from a unidimensional to a multidimensional and even interdisciplinary approach [34]. Although the research on ESV and LUCEs is relatively comprehensive, the knowledge of the correlation between the two based on their different spatio-temporal relationship is still relatively insufficient. In particular, there is an obvious gap in the knowledge of the transition patterns of the spatio-temporal relationship between them. Many of the existing studies focus on the static description level [35,36], and few have conducted dynamic analyses from the perspective of transition patterns. Therefore, dynamic analysis based on transition patterns should be strengthened to deepen the understanding of the spatio-temporal relationship between ESV and LUCEs and provide support for the improvement of relevant theoretical systems.

As the largest loess-covered area globally, the Loess Plateau faces environmental challenges such as soil aridity, erosion, limited vegetation, and water scarcity, rendering it one of China’s most ecologically fragile regions [37]. Urban expansion-driven land-use alterations have notably diminished the ecosystem services of the Loess Plateau, perpetuating strain on the local environmental system. Taking the Loess Plateau as an example, this study, based on five periods of land-use and socio-economic data from 2000 to 2020, calculates the ESV and LUCEs of 340 counties (districts) in the region. Spatial autocorrelation and a spatio-temporal transition matrix based on LISA (Local Indicators of Spatial Association)–Markov are used to explore the relationship and transition paths between the two. A four-quadrant model between ecosystem service value intensity (ESVI) and land-use carbon emissions intensity (LUCEI) is constructed to deepen the research on transition patterns from the perspective of ecological quality, aiming to provide theoretical support for ecosystem protection and carbon emission reduction on the Loess Plateau.

2. Materials and Methods

2.1. Study Area

The Loess Plateau, situated in arid and semi-arid regions of China (33°41′–41°16′ N, 100°52′–114°33′ E), is a vital area for national soil and water conservation and ecological development. It spans Shanxi, Shaanxi, Gansu, Qinghai, and Henan Provinces, Ningxia Hui Autonomous Region, and parts of the Inner Mongolia Autonomous Region, covering a total of 7 provinces, 45 cities, 340 counties, and approximately 640,000 km² (Figure 1). As of the 7th National Population Census in 2020, the Loess Plateau is home to 115 million permanent residents, representing 8.15% of the national population. Urbanization in the region surged from 34.9% to 63.8% between 2000 and 2020, with the GDP and industrial output value in 2020 soaring to 14.35 times and 10.86 times their 2000 levels, respectively. Despite being an area with significant population, resource, and environmental challenges, the fragile natural conditions on the Loess Plateau, coupled with intensive human activities, have led to ongoing ecosystem degradation, resulting in severe soil and water loss locally and globally. Changes in land-use patterns have triggered notable shifts in the regional ecosystem and carbon emissions [38]. Notably, initiatives such as the “Grain for Green Project” and the “Natural Forest Protection Project”, have driven a continuous increase in forest and grass coverage on the Loess Plateau from 2000 to 2020, with vegetation coverage growing at an average rate of 0.0095 per year, enhancing ecosystem services [39]. However, serious challenges such as soil and water loss, water scarcity, air pollution, and elevated carbon emissions persist due to the complex human–land dynamics in the region. Aligned with the national strategy of “Ecological Protection and High-quality Development in the Yellow River Basin,” urgent actions are needed to optimize land use and regulate resource–environment interactions to achieve ecological security and sustainable regional development.

2.2. Data

This study selected land-use data from the Loess Plateau for five periods (2000, 2005, 2010, 2015, and 2020). The data were sourced from the Data Center for Resources and Environmental Sciences, Chinese Academy of Sciences (www.resdc.cn, accessed on 15 May 2025), with a spatial resolution of 30 m × 30 m. In combination with the current land-use classification standard, the land-use/land-cover data were re-classified into six primary categories: cultivated land, forest land, grassland, water area, construction land, and unused land. The data on the area, yield, and price of the main crops grown on the Loess Plateau involved in this study were derived from the statistical yearbooks of each province and the Compilation of National Agricultural Product Cost–Benefit Data. After unit conversion and calibration, the data were aggregated and calculated to measure the agricultural output level and economic benefits. The energy and carbon emission data were mainly sourced from the China Energy Statistical Yearbook, which also provided the standard coal conversion coefficients for different energy sources. The IPCC Guidelines for National Greenhouse Gas Inventories provided the carbon emission coefficients for various fossil fuels. The energy consumption data required for calculating LUCEs were obtained from the statistical yearbooks of each province within the study area, the 7th National Population Census Bulletin, and the China Energy Statistical Yearbook. To ensure the temporal consistency and continuity of the data, all statistical data were selected for the five time points of 2000, 2005, 2010, 2015, and 2020. For the missing year data, linear interpolation was used for completion. The detailed data sources and corresponding time periods are summarized in

Table 1. Datasets and sources used in this study.

Data	Year	Source	Resolution
Land use data	2000, 2005, 2010, 2015, 2020	Resources and Environmental Science and Data Center, Chinese Academy of Sciences (https://www.resdc.cn, accessed on 15 May 2025)	30 m × 30 m
Crop area, yield and price	2000, 2005, 2010, 2015, 2020	Compilation of National Agricultural Product Cost (https://www.ndrc.gov.cn/xwdt/ztzl/ncpdc70zn/wap_index.html, accessed on 15 April 2025) China Statistical Yearbook (https://www.stats.gov.cn/, accessed on 1 May 2025)	/
Energy data	2000, 2005, 2010, 2015, 2020	China Energy Statistical Yearbook (http://www.tjnjw.com/hangye/n/zhongguo-nengyuan-tongjinianjian.html, accessed on 15 April 2025)	/
Carbon emission data	2000, 2005, 2010, 2015, 2020	IPCC Guidelines for National Greenhouse Gas Inventory (https://www.ipcc-nggip.iges.or.jp/public/2006gl/chinese/vol4.html, accessed on 15 April 2025)	/
Population data	2000, 2005, 2010, 2015, 2020	The seventh national census bulletin (https://www.gov.cn/guoqing/2021-05/13/content_5606149.htm, accessed on 20 April 2025)	/

.

2.3. Methods

To explore the spatio-temporal relationship and transition patterns between the ESV and LUCEs on the Loess Plateau (Figure 2), we first quantitatively estimated these values at the county scale based on five-phase land-use data from 2000 to 2020. Second, the bivariate spatial autocorrelation analysis method was employed to identify the spatial relationship and agglomeration types of the two at different stages. Third, the LISA time path model and a spatio-temporal transition matrix were introduced to depict the relationship types and transition paths of ESV and LUCEs. Finally, a four-quadrant model of ESV intensity and LUCE intensity was constructed to deepen our understanding of the transition patterns of ESV and LUCEs in different periods from the perspective of ecological quality and further predict the sustainable development trend in the Loess Plateau.

2.3.1. ESV Measurement Model

To quantitatively evaluate the ESV of the Loess Plateau, we utilized the equivalent factor method of ESV per unit area [40], first proposed by Costanza et al. [41]. Based on the Millennium Ecosystem Assessment, it divides ecosystem services into four major categories—provisioning, regulating, cultural, and supporting—and further subdivides them into 17 types. It has now become an internationally used classification framework. Subsequently, Xie Gaodi et al. [42] constructed a localized equivalent factor system containing 11 types of service functions in combination with China’s ecological environment characteristics and socio-economic statistical data, which more effectively meets the assessment needs in the Chinese context. Based on the value coefficients of terrestrial ecosystem services proposed by Xie Gaodi et al., we made partial corrections according to the actual situation in the Loess Plateau: In terms of ecosystem types, the study area was divided into five categories: cultivated land, forest land, grassland, water area, and unused land. The ESV of construction land was assigned a value of 0, which is consistent with the treatment methods of Costanza et al. and Xie Gaodi et al. In terms of service function classification, according to the characteristics of the semi-arid climate, severe soil erosion, and high ecological vulnerability of the Loess Plateau, the 11 types of service functions of Xie Gaodi et al. were integrated into 10 types, and, finally, an equivalent table of ESV per unit area of the Loess Plateau ecosystem was created (

Table 2. ESV equivalent per unit area.

Class of ESV	Cropland	Forest	Grassland	Waters	Unutilized Land
Raw materials	0.40	0.54	0.34	0.37	0.03
Water supply	0.02	0.28	0.19	5.44	0.02
Gas regulation	0.67	1.76	1.21	1.34	0.11
Climate regulation	0.36	5.27	3.19	2.95	0.10
Environmental purification	0.10	1.57	1.05	4.58	0.31
Hydrological regulation	0.27	3.81	2.34	63.24	0.21
Soil disposition	1.03	2.14	1.47	1.62	0.13
Nutrient cycle	0.12	0.16	0.11	0.13	0.01
Biodiversity	0.13	1.95	1.34	5.21	0.12
Aesthetic landscape	0.06	0.86	0.59	3.31	0.05

).

The equivalent factor method for assessing ESV utilizes the economic value of grain production per unit area of farmland under natural conditions as a reference point, known as the “1 standard equivalent factor.” This method evaluates the relative contributions of different ecosystem and service types through various techniques such as ratio analysis, weighting, and correction. In order to enhance the precision of estimating ESV in the Loess Plateau region, we performed a regional calibration of the standard equivalent factor. Specifically, by analyzing data on the planting area, yield, and market price of three primary crops (wheat, corn, and soybeans) from provincial statistical yearbooks spanning 2000 to 2020, the average crop yield over the study period was determined to be 2394.11 kg/hm², with an average purchase price of CNY 2.71/kg. Consequently, the economic value of grain production per unit area of farmland was computed at CNY 648.04/hm². Following the principle that “the equivalent coefficient of the economic value of 1 standard equivalent is 1/7 of the economic value of food production per unit area of farmland” [43], the economic value of the equivalent factor for ESV per unit area of farmland on the Loess Plateau was established at CNY 926.86/hm², and the ESV and the ESVI value were calculated as follows:

$$V_{es}=\sum \left(A_k\times V{C}_k\right)$$

(1)

$${\overline{V}}_{esj}={\displaystyle \frac{V_{esj}}{S_j}}$$

(2)

where $V_{es}$ is the value of ecosystem services, $A_k$ is the area of land-use type, $V{C}_k$ is the ESV coefficient, and $S_j$ is the area of the $j$th county.

2.3.2. LUCE Measurement Model

LUCEs include direct carbon emissions and indirect carbon emissions. The former refers to the carbon emissions caused by current land use, while the latter refers to the anthropogenic carbon emissions carried by land-use types [44]. The Loess Plateau involves six types of land use, including cultivated land, forest land, grassland, water area, construction land, and unused land. The carbon source and carbon sink capacities of different land-use types vary. This study comprehensively considered the direct and indirect carbon emissions of land use and calculated the LUCEs for the Loess Plateau based on the carbon emission coefficients of land-use types [45]. Specifically, the direct carbon emission coefficient method was used to measure the carbon emissions of cultivated land, forest land, grassland, water area, and unused land, and the consistency and rationality of the coefficients were verified. The calculation formula is as follows:

$$E_k=\sum E_i=\sum T_i\times {\delta }_i$$

(3)

where $E_k$ represents the direct carbon emissions, $E_i$ denotes the carbon emissions of different land-use types, $T_i$ is the area of each land-use type (hm²), and ${\delta }_i$ is the carbon emission coefficient of each land-use type (

Table 3. Carbon emission coefficient of different land-use types (t·hm⁻²).

Land-Use Type	Cropland	Forest	Grassland	Waters	Unutilized Land
Carbon emission coefficient	0.422	−0.644	−0.021	−0.218	−0.005
Reference	Zhao, X.C. et al. [46]	Huang, H.Y. et al. [47]	Fang, J.Y. et al. [48]	Huang, H.Y. et al. [47]	Huang, H.Y. et al. [47]

).

An indirect estimation method was used to calculate the carbon emissions of construction land. Since it is difficult to accurately estimate the carbon emission coefficient of construction land due to the influence of human activities, using the carbon emission coefficients of energy consumed by construction land for calculation is an effective assessment method. In this study, the standard coal conversion coefficients and carbon emission coefficients of various energy sources were obtained from the China Energy Statistical Yearbook and the IPCC Guidelines for National Greenhouse Gas Inventories (

Table 4. Standard coal coefficient and carbon emission coefficient for different types of energy.

Energy Types	Coal	Hard Coke	Crude Oil	Fuel Oil	Gasoline	Kerosene	Natural Gas
Standard coal coefficient	0.7143	0.9714	1.4286	1.4286	1.4714	1.4714	1.3301
Carbon emission coefficient	0.7559	0.8550	0.5857	0.6185	0.5538	0.5714	0.4483

). The energy consumption was converted into tons of standard coal according to the standard coal conversion coefficients, and then the carbon emissions of construction land were calculated according to the carbon emission coefficients. The calculation formula is

$$E_t=\sum E_{ti}=\sum E_{ni}\times {\theta }_i\times f_i$$

(4)

where $E_t$ represents the carbon emissions of construction land, $\sum E_{ti}$ represents the carbon emissions generated by various energy consumptions, $\sum E_{ni}$ represents the consumption of various energy sources, ${\theta }_i$ represents the standard coal conversion coefficients of various energy sources, and $f_i$ represents the carbon emission coefficients of various energy sources. The calculation formula for the LUCEI is:

$${\overline{E}}_i={\displaystyle \frac{E_{ai}}{S_i}}={\displaystyle \frac{E_{ki}+E_{ti}}{S_i}}$$

(5)

where ${\overline{E}}_a$ represents the LUCEI of the $i$th administrative unit, $E_{ai}$ is the total carbon emissions, $E_{ki}$ and $E_{ti}$ are the sum of carbon emissions from cropland, forest, grassland, water, and unutilized land and the carbon emissions from construction land, respectively, and $S_i$ is the area of the $i$th administrative unit.

2.3.3. Analysis of Spatial Relationship and Dynamic Evolution

In this study, the bivariate Moran’s I global spatial autocorrelation method was used to systematically analyze the spatial relationship between the ESV and LUCEs at the county scale on the Loess Plateau [49]. The spatial weight matrix was based on the Queen adjacency method, which means that if two spatial units share a common edge or a common point, they are considered adjacent units. The definition is as follows:

$$w_{\mathrm{ij}}=\left\{\begin{array}{cc}1,& if units i and j are adjacent\\ 0,& if units i and j are not adjacent\end{array}\right.$$

(6)

where $w_{\mathrm{ij}}$ represents the spatial weight of unit $i$ and $j$. On this basis, the bivariate Moran’s I global index is used to measure the spatial correlation between the ESV of the $i$th spatial unit and the LUCEs of the $j$th spatial unit in its neighborhood. The calculation formula is as follows:

$$I_{se}={\displaystyle \frac{n{\sum }_{i=1}^n {\sum }_{j=1}^n W_{ij}\left(y_{i,s}-{\overline{y}}_s\right)\left(y_{i,e}-{\overline{y}}_e\right)}{(n-1){\sum }_{i=1}^n {\sum }_{j=1}^n W_{ij}}}$$

(7)

where $I_{se}$ is the bivariate global spatial autocorrelation Moran’s index of ESV and LUCEs. $y_{i,s}$ and $y_{i,e}$ are the ESV and LUCEs of the $i$th evaluation unit.

This paper introduced the LISA cluster map for local spatial autocorrelation analysis and classified five types, namely HH, LL, LH, HL, and NN, based on the LISA values and significance levels. The GeoDa software 1.14 is used to conduct spatial correlation tests on the ESV and LUCEs during different periods. To ensure the reliability of statistical inference, 999 Monte Carlo random permutations were performed with a significance level set at 95% (p < 0.05). Generally, when $I_{se}$ > 0, the HH/LL-type indicates that the attribute values of this spatial unit are higher/lower than those of the surrounding areas, and the comprehensive spatial differences are small; when $I_{se}$ < 0, the LH/HL-type indicates that the spatial units with lower/higher attribute values are higher/lower than the surrounding provinces and the comprehensive spatial differences are large; and the NN-type indicates no significant correlation.

Considering that the spatial pattern has dynamic evolution characteristics, this study introduced the LISA time path model in the Exploratory Time–Space Data Analysis (ESTDA) method to quantify the spatio-temporal trajectory characteristics of each evaluation unit in multi-period LISA clustering [50]. The formula is as follows:

$$T=\frac{n{\sum }_{t=1}^{T-1} d\left(L_{i,t},L_{i,t+1}\right)}{{\sum }_{i=1}^n {\sum }_{t=1}^{T-1} d\left(L_{i,t},L_{i,t+1}\right)}$$

(8)

$$Q=\frac{{\sum }_{t=1}^{T-1} d\left(L_{i,t},L_{i,t+1}\right)}{d\left(L_{i,1},L_{i,T}\right)}$$

(9)

where $T$ represents the relative length, $Q$ represents the curvature, $n$ represents the quantity, and $d\left(L_{i,t},L_{i,t+1}\right)$ represents the moving distance of the research unit $i$ between year t and year t + 1. If both $T$ and $Q$ are greater than 1, it indicates that the research unit has a dynamic local spatial structure and direction. To elucidate the evolution mechanism of this spatial structure further, this study developed a spatio-temporal transition matrix by integrating LISA time paths and Markov chains to investigate the dynamic trajectory characteristics of the local spatial relationships’ strength between the region itself and neighboring units. By analyzing changes in spatial morphology from the initial period (t) to the final period (t + 1), this study categorized four types of spatio-temporal transitions (

Table 5. Spatio-temporal transition type.

Type	Connotation	Expression Formula
Type I	Only the unit itself undergoes a transition	HHt→LHt+1, LLt→HLt+1, LHt→HHt+1, HLt→LLt+1
Type II	Only the adjacent unit undergoes a transition	HHt→HLt+1, LLt→LHt+1, LHt→LLt+1, HLt→HHt+1
Type III	Both the unit and its adjacent unit undergo transitions	HHt→LLt+1, LLt→HHt+1, LHt→HLt+1, HLt→LHt+1
Type IV	Both the unit and its adjacent unit remain unchanged	HHt→HHt+1, LLt→LLt+1, LHt→LHt+1, HLt→HLt+1

).

2.3.4. Four-Quadrant Model

As a commonly used multi-index combination zoning method, the four-quadrant method has been widely applied in studies focusing on areas such as ecological function zoning, land-use conflict identification, and ecological risk assessment [51,52]. It can intuitively reveal the spatial differentiation characteristics of each index, providing scientific support for formulating differentiated strategies for ecological protection and low-carbon development. To further verify the applicability of the transition path and conduct an in-depth analysis of the evolution characteristics of the ecological environment quality on the Loess Plateau, we construct a four-quadrant model with the ESVI on the horizontal axis and the LUCEI on the vertical axis to explore the transition mode of the Loess Plateau at the county scale. In the model, the ESVI refers to the ecosystem service value carried per unit area, which is used to measure the spatial distribution differences of regional ecosystem service functions. The LUCEI represents the carbon emission level generated by land use per unit area, reflecting the regional development and utilization intensity and the resulting ecological environment pressure. It is worth noting that forests, grasslands, and water areas have significant carbon sink functions, and their carbon emission coefficients are negative. Therefore, when the proportion of such land use is relatively high and the energy consumption level is low, the comprehensively calculated carbon emission intensity may be negative, indicating that the region is in a net absorption state in the carbon cycle. This model uses the natural break point classification method to classify the ESVI and LUCEI, respectively. It reasonably determines the high- and low-level thresholds according to the data distribution and is widely used in land use and ecosystem service research [53]. Based on the classification results, a four-quadrant division system is constructed [53] (Figure 3 and

Table 6. Four-quadrant classification table of regional ecological quality levels.

Indicator	First Quadrant	Second Quadrant	Third Quadrant	Fourth Quadrant
Ecosystem service value intensity	10,847.35~81,586.24	0.00~10,847.35	0.00~10,847.35	10,847.35~81,586.24
Land-use carbon emission intensity	1194.41~8368.32	1194.41~8368.32	−0.26~1194.41	−0.26~1194.41

): The first quadrant represents the ecological good area, characterized by a high ESVI and high LUCEI. The second quadrant pertains to the ecological poor area, featuring a low ESVI and high LUCEI. The third quadrant encompasses the ecological general area, showcasing a low ESVI and low LUCEI. Lastly, the fourth quadrant encompasses the ecological quality area, distinguished by a high ESVI and a low LUCEI.

3. Results

3.1. Analysis of Spatio-Temporal Changes in ESV

Table 7. Unit-area ESV coefficients of different land-use types on the Loess Plateau (CNY/hm²).

Class of ESV	Cropland	Forest	Grassland	Waters	Unutilized Land
Raw materials	370.74	497.41	318.22	338.30	27.81
Water supply	18.54	256.43	176.10	5042.12	18.54
Gas regulation	621.00	1631.27	1118.41	1237.36	101.95
Climate regulation	333.67	4881.46	2956.68	2729.60	92.69
Environmental purification	92.69	1452.08	976.29	4240.38	287.33
Hydrological regulation	250.25	3531.34	2165.76	58,609.99	194.64
Soil disposition	954.67	1986.57	1362.48	1501.51	120.49
Nutrient cycle	111.22	151.39	105.04	115.86	9.27
Biodiversity	120.49	1810.47	1238.90	4828.94	111.22
Aesthetic landscape	55.61	794.01	546.85	3067.91	46.34
Sum	2928.88	16,992.43	10,964.75	81,711.98	1010.28

shows the ESV coefficients per unit area of different land-use types on the Loess Plateau, and

Table 8. ESV of different land-use types on the Loess Plateau from 2000 to 2020 (billion CNY).

	2000	2005	2010	2015	2020
Cropland	60.492	59.311	58.297	58.079	56.733
Forest	157.865	161.637	163.287	162.912	163.790
Grassland	285.323	284.167	285.842	285.163	284.570
Waters	71.016	72.420	68.559	69.254	73.214
Unutilized land	4.336	4.469	4.087	4.058	4.164
Sum	579.032	582.005	580.072	579.466	582.470

shows the ESV amounts of different land-use types. The results indicate that the ESV on the Loess Plateau experiences periodic fluctuations, but the overall trend remains stable. From 2000 to 2020, the ESV increased from CNY 579.032 billion to CNY 582.470 billion, with an increase of CNY 3.438 billion and an average annual growth rate of 0.15%. This suggests that the growth rate of the ESV on the Loess Plateau is relatively low, and the regional ecosystem service supply capacity is relatively stable. In terms of land-use types, the ESV of cultivated land decreased from CNY 60.492 billion in 2000 to CNY 56.733 billion in 2020, a decrease of 6.22%; the value of grassland decreased from CNY 285.323 billion to CNY 284.570 billion, remaining basically stable; and the value of unused land decreased from CNY 4.336 billion to CNY 4.164 billion, a decrease of 3.97%. In contrast, the ESV of forest land and water areas increased by CNY 5.925 billion and CNY 2.198 billion, respectively. This is mainly related to the implementation of ecological protection policies on the Loess Plateau [54], which led to the conversion of other land-use types into forest land and water areas. Among them, forest land and grassland, as the main components of the ESV, account for 76.95% of the total ESV. Although the value coefficient of water areas is the highest among the six land-use types, due to its small area proportion, it only accounts for 12.21% of the total ESV.

In terms of the spatial pattern, the ESV of the Loess Plateau shows significant spatial differences (Figure 4). Areas with a high ESV are mainly distributed in western and southern Inner Mongolia, central and northern Shaanxi Province, and other regions with good forest and grass coverage. Taking Yulin City in Shaanxi Province as an example, its total ESV reached CNY 34.298 billion in 2020, maintaining a relatively high level for 20 consecutive years. Areas with a low ESV are mainly distributed in central Shanxi Province, western Shaanxi Province, the Guanzhong Plain in the central and southern part of the study area, and the Weihe River Basin and other areas with dense cultivated land and construction land. The county-scale further reveals the changes in the spatial differences of ESV. From 2000 to 2020, the ESV of 161 counties increased to varying degrees, with an average growth rate of 10.40%. These counties are mainly distributed in the central hilly belt of the study area, mountainous areas in southern Shanxi and southern Shaanxi, and the southern margin of Inner Mongolia. In these regions, the vegetation cover, including in woodlands and grasslands, is extensive, and human activity is minimal. Long-term ecological management has yielded significant results. Meanwhile, the ESV of 171 counties decreased, with an average decline of 7.83%. A typical example is Erqi District in Zhengzhou City, Henan Province. Its ESV decreased from about CNY 0.25 billion in 2000 to about CNY 0.06 billion in 2020, a decline of 76%. In addition, the reductions in ESV in Laocheng District and Xigong District of Luoyang City and other districts and counties all exceeded 50%. These areas have flat terrain, and the accelerated urbanization process has led to the expansion of construction land [18], reducing ecological land and, thus, significantly weakening the regional ecosystem service value (Figure 5).

3.2. Analysis of Spatio-Temporal Changes in LUCEs

As shown in

Table 9. LUCEs on the Loess Plateau region from 2000 to 2020 (million metric tons, Mt).

Year	Cropland	Forest	Grassland	Water	Unutilized Land	Construction Land	Net Carbon Emissions
2000	8.72	−5.98	−0.55	−0.19	−0.02	135.17	137.15
2005	8.55	−6.13	−0.54	−0.19	−0.02	241.33	243.00
2010	8.40	−6.19	−0.55	−0.18	−0.02	358.55	360.01
2015	8.37	−6.17	−0.55	−0.18	−0.02	413.69	415.14
2020	8.17	−6.21	−0.55	−0.20	−0.02	457.24	458.43

, the LUCEs on the Loess Plateau showed an increasing trend from 2000 to 2020. They increased from 137.15 Mt in 2000 to 458.43 Mt in 2020, with an average annual growth rate of 6.22%. Among the different land-use types, construction land was the main land-use type driving the growth of carbon emissions. Its LUCE increased from 135.17 Mt in 2000 to 457.24 Mt in 2020, accounting for over 97% of the total increment during the same period. Since resource- and energy-intensive industries have a large demand for construction land and generate a large amount of carbon dioxide during production processes, the carbon emissions from land use have been increasing year by year [55]. Meanwhile, the total carbon emissions from cultivated land were relatively small and showed a slow downward trend, decreasing by 0.55 Mt from 2000 to 2020. This change was mainly affected by factors such as the reduction in cultivated land area and the decrease in energy consumption during agricultural production. As the main carbon sink in the region, the carbon absorption of forest land increased from 5.98 Mt in 2000 to 6.21 Mt in 2020. Limited by the sharp increase in carbon emissions from construction land, the carbon sink of forest land could only offset 0.07% of the increment, indicating that under the current land-use pattern, although the carbon sink of forest land increases due to the expansion of its area, the total carbon sink is far less than the carbon source. The ability to offset high-intensity carbon emissions is limited, resulting in high net carbon emissions. Moreover, the net carbon emissions grow at the same rate as the carbon source. In addition, the total carbon absorption of grassland, water area, and unused land changed little overall, with annual averages of 0.55 Mt, 0.19 Mt, and 0.02 Mt, respectively. Although their carbon sink effects were limited, they still had significance in maintaining the functions of the regional ecosystem.

There are significant spatial differences in LUCEs among counties on the Loess Plateau (Figure 6). At the provincial scale, the total LUCEs on the Loess Plateau decrease in the order Shanxi > Inner Mongolia > Shaanxi > Henan > Ningxia > Gansu > Qinghai, which not only reflects the differences in industrial structure, energy consumption intensity, and land-use patterns among provinces but also reveals the profound influence of policy regulation on LUCEs. In some provinces relying on policy controls like the delineation of ecological protection red lines and the construction of key ecological function areas [56], such as Shaanxi and Ningxia, the disorderly expansion of high-carbon industries and the blind growth of construction land have been effectively curbed, leading to a significant slowdown in the growth rate of LUCEs in some counties. In contrast, traditional resource-based regions like Shanxi and Inner Mongolia have room for adjustment in terms of industrial transformation and policy implementation, and their LUCEs remain at a high level. At the county scale, high-value areas of LUCEs are mainly distributed in regions with high urbanization levels and high construction land densities, such as the Hetao Plain and the Guanzhong Plain. These areas have a flat terrain, a convenient transportation network, and a large proportion of secondary industries, with significantly higher land-use carbon emission intensities than other regions. The top 50 counties contribute 50.70% of the total LUCEs in the study area. Among them, Lingwu City, Yinchuan City, Ningxia Hui Autonomous Region, has an average annual carbon emission of 9.13 Mt, which is 10.89 times the regional county average. In contrast, in the southwestern part of the Loess Plateau, land use is mainly forest and grassland, with low development intensity and a high proportion of ecological land, resulting in relatively low LUCEs.

3.3. Spatial Relationship and Dynamic Evolution Analysis of ESV and LUCEs

Table 10. Bivariate global Moran’s I statistic of the Loess Plateau from 2000 to 2020.

Years	2000	2005	2010	2015	2020
Moran’s I	0.054	0.045	0.08	0.079	0.079
P	<0.05	<0.05	<0.05	<0.05	<0.05
z	2.2679	1.9266	3.4090	3.3615	3.3884

shows the results of the bivariate global Moran’s I spatial autocorrelation analysis of ESV and LUCEs on the Loess Plateau. Moran’s I values for each year from 2000 to 2020 are all positive, and the p-values are less than 0.05, indicating a positive spatial correlation between ESV and LUCEs on the Loess Plateau. Specifically, Moran’s I value showed a slight decline in 2005 but rebounded rapidly after 2010 and tended to be stable, indicating that the spatial correlation between ESV and LUCEs continued to strengthen. This trend may be comprehensively influenced by land-use policies, industrial structure adjustment, and ecological restoration projects [57]. With the support of policies enacting the return of farmland to forest, natural forest protection, and the construction of key ecological function areas, the ecological functions in some regions have been significantly enhanced, increasing the ESV. Meanwhile, the continuous expansion of resource-based cities has intensified LUCEs, resulting in the co-existence of high ESV and high LUCEs in space.

Figure 7 displays the bivariate LISA map illustrating the relationship between ESV and LUCEs on the Loess Plateau. In 2000, the region comprised 17 HH-type areas, 32 LL-type areas, 10 LH-type areas, 24 HL-type areas, and 302 NN-type areas. By 2020, these numbers changed to 19 HH-type areas, 33 LL-type areas, 18 LH-type areas, 22 HL-type areas, and 293 NN-type areas. These alterations suggest fluctuations in the aggregation patterns of ESV and LUCEs over the two-decade period, albeit without significant overall changes. Geospatially, HH-type areas primarily cluster in transition zones between urban developments and mountainous forests, notably in Fugu County, Hongsibu District, Pingluo County, and the Ordos Plateau. Here, the co-occurrence of energy extraction, industrial activities, and ecological functions results in high ESV and LUCEs [33]. LL-type areas concentrate in the northern Weihe River region and the border of Shaanxi and Shanxi provinces, with a sporadic presence in Linxia County and Gangu County in Gansu Province, and Gonghe County and Guinan County in Qinghai Province. These areas, characterized by underutilized land and low land-use intensity, exhibit low ESV and LUCEs. LH-type areas are dispersed across the study area, particularly in central urban zones of cities like Baotou, Datong, and Ordos. These regions have a high population and industry concentration, more active economies, and significant development potential. They are key areas for urbanization construction, but the carbon emissions from land use in these locations are far higher than their ecosystem service capacity [18]. HL-type areas predominantly lie in ecological barrier zones in the south and west, which are rich in forest and grass resources. Construction land is sporadically distributed in a dot-like pattern, resulting in low LUCEs and robust carbon sequestration capabilities.

To explore the dynamic evolution of the spatial relationship between ESV and LUCEs,

Table 11. Space–time transition matrices of the Loess Plateau from 2000 to 2020.

t/t + 1	HH	LL	LH	HL
HH	IV (71, 0.9861)	0	0	0
LL	0	IV (127, 0.9338)	0	I (1, 0.0074)
LH	0	0	IV (59, 0.9833)	0
HL	0	I (1, 0.0101)	0	IV (92, 0.9293)

presents the spatio-temporal transition matrix from 2000 to 2020 constructed based on the LISA time path and Markov chain. The results show that the spatial relationship pattern in the study area has significant spatio-temporal stability. The vast majority of counties and their neighboring areas maintain a Type IV spatio-temporal transition, reflecting a high degree of spatial path dependence and local structural stability. Among them, the stability rate of HH-type areas is the highest, reaching 98.61%, and the stability rate of LL-type areas also exceeds 93%. The stability rates of LH-type and HL-type areas are 98.33% and 92.93%, respectively, indicating the persistence of the relationship in most counties. A small number of Type I transitions indicate that local adjustments of clustering types have occurred in some counties or their neighboring areas, specifically manifested as the transformation from the LL-type to the HL-type or from the HL-type to the LL-type. This change may be related to local land-use changes and ecological restoration policies.

3.4. Transition Modes Based on the Four-Quadrant Model

Based on the transition matrix analysis, the quadrant diagrams of the ecological environmental quality of the Loess Plateau in 2000 and 2020 were constructed (Figure 8) to further deepen the research on the transition mode from the perspective of ecological quality. In the study area, the number of counties in the general ecological area was 291 in 2000 and 285 in 2020, with a total area of 48.71 × 10⁴ km² and 48.11 × 10⁴ km², respectively, which is consistent with the high stability shown by the HH- and LL-type areas in the transition matrix. This indicates that the ecosystems of most counties are in a relatively stable state but have not yet reached a virtuous cycle, reflecting the long-term, slow evolution characteristics of the Loess Plateau ecosystem. The cycles of soil conservation, vegetation restoration, and ecological function improvement are relatively long, and the areas with significant short-term changes are limited. Most county-level ecological functions still have great potential for improvement [58]. Meanwhile, the number of counties in the good ecological area increased from one in 2000 to three in 2020, and the area increased from 0.75 km² to 4.99 km², showing a slight increase. This confirms the rationality of the few positive conversion paths in the transition matrix, indicating that some local areas have achieved phased results in ecological restoration, land management, or natural restoration. The number of counties in the quality ecological area changed little, from 93 in 2000 to 92 in 2020, which corresponds to the local conversion path from the HL-type to the LL-type in the transition matrix. However, the occupied area increased by 0.59 × 10⁴ km², indicating that some areas with a good ecological foundation face potential risks of functional weakening or system degradation. In contrast, there were a total of five counties in the poor ecological area in 2020, with a total area of 54.72 km². Although this number is small, there is a visible trend of ecological risks in local areas, which requires high-level vigilance. In terms of spatial distribution, the spatial differentiation of ecological quality is significantly correlated with land-use types. The quality ecological area and the good ecological area are mainly distributed in counties where forest land and grassland are the main land-use types, featuring superior natural conditions and relatively low human interference. On the other hand, the general ecological area and the poor ecological area are mainly concentrated in the valley plain areas dominated by cultivated land and construction land, showing significant spatial aggregation characteristics. The intensity of human activities in this region is high [59]. The expansion of construction land and the high-intensity development of resources have weakened the ecosystem’s stability, resulting in a relatively low ecological service capacity.

To assess ecological quality trends across counties on the Loess Plateau, changes were categorized into five groups: significant improvement, improvement, no change, decline, and significant decline (Figure 9). The findings revealed that 318 counties maintained stable ecological quality from 2000 to 2020, while 8 counties demonstrated notable improvement, and 14 counties exhibited varying degrees of decline. This pattern of overall stability with pockets of enhancement suggests advancements in ecological management and natural restoration over the past two decades. Nonetheless, most regions are still in the early stages of ecological recovery, with no immediate substantial progress. Areas experiencing declining ecological quality include Huimin District, Kundulun District, and Kangbashi District in the Inner Mongolia Autonomous Region; Dengfeng City and Hubin District in Henan Province; Dongxiang Autonomous County in Gansu Province; and Jinyuan District in Shanxi Province. These declines are linked to increased construction activities, intensive resource exploitation, and diminished green spaces amid rapid urbanization. Thus, interventions are necessary to curb carbon emissions and enhance ecological conditions.

4. Discussion

4.1. Spatio-Temporal Relationship and Transition Patterns Between ESV and LUCEs

This study reveals the spatio-temporal relationship and transition mode between the ESV and LUCEs on the Loess Plateau from 2000 to 2020. Our research found that the net carbon emissions on the Loess Plateau showed a rapid growth trend from 2000 to 2020. The sharp increase in carbon emissions from construction land was the dominant factor in the change in LUCEs, which is consistent with the research results of Jiaying Peng et al. [60] based on Chinese counties. Since the 21st century, with the acceleration of China’s industrialization and urbanization, the carbon emission level on the Loess Plateau has been continuously rising. From 2000 to 2020, the average annual growth rate of LUCEs on the Loess Plateau was 6.22%, higher than the national level of 5.36% [61] during the same period. Shanxi, Shaanxi, and Inner Mongolia, as regions with concentrated coal resources, contributed about 60–70% of the carbon emissions during the coal mining and utilization process [62]. In contrast, the change in the ESV was not significant, but the spatial pattern showed significant changes. On the one hand, the forest area in the study area increased from 8.48 × 10⁴ km² in 2000 to 9.82 × 10⁴ km² in 2020, a growth of 15.86%. The water area increased from 1.02 × 10⁴ km² to 1.96 × 10⁴ km², which was concentrated in the central and eastern regions with significant hilly and mountainous landforms, and to some extent enhanced the supply and regulation functions of the regional ecosystem. On the other hand, after 2000, the Loess Plateau entered a period of rapid urbanization. The area of construction land increased by 92.25% in 20 years, mainly occurring in the Weihe Valley and Fenhe Valley in the southeast and the Hetao Plain in the northwest of the Loess Plateau. These areas are densely populated, with large food demand and energy consumption. Human activities have a great impact on the ecological environment, and areas with general and poor ecological conditions are mostly distributed here. It is worth noting that there is a certain degree of positive correlation between the ESV and LUCEs in some regions. This phenomenon is related to the superposition of human activities such as ecological governance and urban expansion, indicating that the ecological restoration of territorial space in some areas has not effectively reduced the carbon emission intensity.

4.2. Policy Implications

(1)

Enhancing the safeguarding of cultivated land and optimizing land resource management are essential for bolstering the foundational support for ecosystem services. The Loess Plateau exhibits a declining trend in the value of cultivated land ecosystem services, particularly in regions with intensive human activities like the Weihe River Valley and the Hetao Plain, characterized by dense cultivated land and relatively weak ecosystem stability, leading to high carbon emission intensity. Hence, it is imperative to advance the establishment of high-quality farmland in areas with a substantial proportion of cultivated land, such as Shanxi and Shaanxi, while intensifying the monitoring and differentiated management of cultivated land quality. Additionally, the adoption of cultivated land rotation and ecological reseeding techniques is recommended to shift cultivated land from solely providing grains to serving as a multifunctional ecosystem supporting water conservation, carbon sequestration, and oxygen release. These measures aim to enhance the service capacity of agricultural ecosystems, mitigate carbon emissions, and promote sustainable land-use practices.

(2)

The optimization of land use and industrial layout must be promoted to improve resource utilization efficiency and the level of environmental governance. The use of high-efficiency energy-saving equipment and water-saving technologies must be promoted to reduce energy and water consumption; industrial emissions should be strictly supervised, and advanced pollution control technologies should be adopted to reduce emissions of harmful substances; urban spatial structure optimization should be enhanced; and construction land expansion must be regulated. Land-use structure optimization should also be enhanced through integrated planning, fostering compact, efficient, and sustainable urban spatial systems. It is also important to prioritize the development of urban greenways, water connectivity projects, and ecological corridors in ecologically valuable and high-carbon-emission areas like Fugu County and Hongsibao District to bolster ecosystem resilience, regulatory capacity, and mitigate urban heat island effects and carbon emissions.

(3)

Enhancing the restoration of forest and grassland resources is crucial to bolstering the ESV of the Loess Plateau. Carbon sinks, including forest land and grasslands, play a significant role in offsetting the substantial carbon emissions associated with construction activities. Therefore, initiatives like converting farmland back to forest and grassland and safeguarding natural forests must be sustained. Particularly in the hilly and mountainous regions of the central and eastern areas, such as southern Shaanxi and southern Shanxi, there should be intensified efforts to rehabilitate forests and grasslands. However, it is imperative to prevent ecological overcompensation issues, such as the formation of “soil dry layers” and the growth of “stunted trees” due to excessive afforestation [63]. Enhancing the hydrological resilience of ecological projects and establishing a robust and effective ecological buffer system are essential measures to address these challenges.

(4)

Decision-makers should seek to enhance the ecological protection compensation system by enhancing the understanding of ecosystem value realization pathways and enhancing the ecological compensation standards and horizontal compensation mechanisms, transitioning from reactive to proactive compensation models. It is also important to bolster policy support for farmers and marginalized regions, encourage societal involvement in ecological conservation, and foster internal system restoration and transformation capabilities.

(5)

Differentiated management should be implemented across various zones. Good ecological areas should prioritize maintaining high ecological service levels while optimizing industrial structures, promoting clean energy, and enforcing total carbon emission regulations to reduce land-use carbon intensity. In poor ecological areas, construction land should be limited, the transformation and upgrade of traditional high-emission industries should be promoted, and the land-use efficiency should be enhanced. For the general ecological areas, guided by ecological compensation and regional planning, ecologically friendly industries should be introduced to stimulate economic growth and gradually boost the ecosystem supply capacity. The quality ecological areas should focus on preserving natural resources and continuously enhancing their ecosystem service capacity. Distinctive ecological tourism projects can be developed by leveraging protection policies and improving infrastructure.

4.3. Innovations and Limitations

Land-use changes significantly impact ESV and LUCEs. While numerous studies have addressed this issue, previous research has primarily focused on independent administrative units, limiting the exploration of the relationship between ESV and LUCEs. This study introduces two key innovations. Firstly, it examines the Loess Plateau as a case study area due to its complex terrain, abundant gullies, and location in a transitional coastal–inland zone, offering diverse natural geographical conditions suitable for county-scale research. Moreover, the rapid urbanization and substantial land-use changes on the Loess Plateau since the 21st century underscore the urgency of investigating the interplay between ecosystems and carbon emissions to inform ecological protection and low-carbon policy development. Secondly, this study delves into county-scale analysis, systematically investigating the spatio-temporal relationship and transition patterns between ESV and LUCEs. It dynamically assesses transition quantities, patterns, and the evolution of regional ecological quality to enhance the scientific basis and comprehensive planning of ecological environment protection and carbon emission reduction policies.

This study has several limitations. Firstly, the ESV of various land-use types on the Loess Plateau was estimated using the equivalent factor method, with adjustments made to the equivalent factor table to align with the Plateau’s specific conditions. However, the types of natural ecosystems here are far more complex and diverse than the five land-use types selected in the study. Assigning a value of 0 to construction land fails to fully reflect its potential negative impacts on the ecosystem; therefore, the accuracy of the estimation results may be limited. Secondly, LUCEs at the county level were calculated by allocating emissions for each county based on data from the 7th National Population Census, due to data collection constraints. This approach may overlook variations in carbon emissions among regions stemming from differences in industrial structure, consumption patterns, income levels, etc. Meanwhile, the utilization of standard coal conversion and carbon emission coefficients derived from the “IPCC Guidelines for National Greenhouse Gas Inventories” may not be suitable for the Loess Plateau’s energy framework. Thus, the development of regional carbon emission factors requires refinement. Additionally, for construction land, this study mainly estimates land-use carbon emissions indirectly through energy consumption, without fully considering the potential negative impacts of different industrial types, resource extraction and processing technologies, water resource consumption, and industrial chemical emissions on soil, water, and the atmosphere. This may lead to an underestimation of the potential pressure of construction land on ecosystem functions. Therefore, future research should comprehensively measure the carbon emissions from construction land use by integrating indicators such as water resource utilization, pollutant emissions, and industrial structure based on land-use carbon emissions, and better evaluate its impacts on ecosystem functions. As a typical ecologically fragile area, more attention should be paid to the resilience and critical response mechanism of the Loess Plateau’s ecosystem. Subsequently, the integration of system dynamics and scenario simulation techniques will be employed to formulate an ecosystem regulation model for risk mitigation, thereby enhancing the predictive capacity concerning the evolutionary trajectory of complex systems and abrupt disturbances.

The standard coal conversion and carbon emission factors employed are derived from the “IPCC Guidelines for National Greenhouse Gas Inventories,” which may not suit the Loess Plateau’s energy structure. Regional carbon emission factors require further refinement. This study primarily estimates carbon emissions from construction land via energy consumption, neglecting the potential impacts of industrial types, resource extraction, processing technologies, water use, and chemical emissions on soil, water, and the atmosphere. This could underestimate the pressure that construction land exerts on ecosystem functions. Future research should integrate water resource use, pollutant emissions, and industrial structure with land-use carbon emissions to comprehensively assess construction land’s impact on ecosystem functions. Given the Loess Plateau’s ecological fragility, its ecosystem’s resilience and response mechanisms warrant closer attention. System dynamics and scenario simulations will be employed to develop an ecosystem regulation model for risk adaptation, enhancing predictions of complex system evolution and responses to sudden shocks.

5. Conclusions

This study used land-use data and socioeconomic data from five periods in 2000, 2005, 2010, 2015, and 2020 to calculate the ESV and LUCEs in the Loess Plateau at different scales. Based on an exploratory spatio-temporal data analysis and the quadrant model, the spatio-temporal relationship and transition patterns between ESV and LUCEs were explored. The conclusions are as follows:

(1)

From 2000 to 2020, the ESV of the Loess Plateau showed a gradually increasing trend, rising from CNY 579.032 billion in 2000 to CNY 582.470 billion in 2020, with an overall increase of only 0.15%. Areas with a high ESV were mainly distributed in the western and southern parts of Inner Mongolia, as well as the central and northern parts of Shaanxi Province. Areas with a low ESV were mainly distributed in the central part of Shanxi Province, the western part of Shaanxi Province, and the surrounding areas, such as the Guanzhong Plain and the Weihe River Basin in the central and southern parts.

(2)

From 2000 to 2020, the LUCEs in counties across the Loess Plateau exhibited a notable upward trajectory. In particular, the surge in carbon emissions from construction land, driven by energy consumption, emerged as the primary driver behind the rapid escalation in LUCEs, intricately linked to economic and social progress. Moreover, substantial spatial variations in LUCEs were observed, with the top 50 counties accounting for 50.70% of the overall carbon emissions.

(3)

A positive correlation exists between ESV and LUCEs on the Loess Plateau. The increasing bivariate global Moran’s I suggests a strengthening spatial relationship between the two factors, leading to a more concentrated distribution. Analysis using bivariate LISA indicates a consistent relationship between ESV and LUCEs from 2000 to 2020 on the Loess Plateau, with spatial distribution closely linked to land-use types. The spatio-temporal transition matrix reveals that most counties and their adjacent areas predominantly exhibit Type IV transitions, demonstrating a strong spatial path dependence.

(4)

From 2000 to 2020, the intensity of ESV and the intensity of LUCEs on the Loess Plateau remained generally stable, with differences in the transition patterns. The general ecological areas were dominant and recovered slowly, while the quality ecological areas were limited. The number of counties in good and poor ecological areas was the smallest, and HH-and LL-type transition paths accounted for the vast majority. There was no significant change in the ecological quality of most counties; counties with declining ecological qualities were mainly concentrated in areas with accelerated urbanization, showing local degradation from the HL-type to the LL-type, and this was highly correlated with land-use types and human activities.