Mechanisms of Coordinated Evolution and Spatial Responses in the Human–Land System During Urban–Rural Integration in Karst Mountainous Areas: A Case Study of Guiyang City

Abstract

The traditional urbanization path based on scale expansion is unsustainable in karst mountainous regions due to fragmented topography and ecological fragility. Taking Guiyang City as a case study, this paper constructs two evaluation indicator systems for urban–rural development and environmental support. Employing the entropy method, coupled coordination degree model, Grey relational analysis, Geodetector, and multi-source spatial analysis methods to examine the evolutionary trajectory, driving mechanisms, and spatial responses of the human–land system from 2000 to 2024. The results show three main findings. First, the comprehensive score of Guiyang’s urban–rural human–land system increased from 0.054 to 0.826, and the coupling coordination degree rose from 0.223 (relative imbalance) in 2000 to 0.903 (high-quality coordination) in 2024, while the environmental support system deviated from the classic environmental Kuznets curve. Second, the driving force has shifted from economic scale to green well-being. The interaction analysis using Geodetector shows that all interaction types fall under the category of two-factor enhancement, among which the interaction coefficient between the number of broadband internet subscribers and other driving factors has the highest explanatory power, with a q-value of 0.949. Third, spatially, the light center distribution stabilized after 2015, and the land use ecological transition index dropped from 0.162 to 0.050 while the D-value continued rising, showing a significant negative correlation (r = −0.89, p < 0.05). Construction land was concentrated in low-slope (0–6°) and mid-elevation (1000–1400 m) basin areas, overlapping with high-quality farmland, and the synchronization rate between economically active areas and construction expansion was 50%. These findings reveal a digital–ecological co-evolution path in karst regions and provide an empirical basis for urban–rural integration governance.

1. Introduction

Since the mid-20th century, the polarization effects of urbanization have caused widespread rural decline and urban expansion [1,2]. Academic discourse has shifted from a unidirectional “rural support for cities” model [3] to frameworks that emphasize functional complementarity and spatial symbiosis [4,5]. China’s urban–rural integration strategy goes beyond previous policies by promoting two-way factor flows, functional complementarity, and institutional synergy [6,7,8,9], with the aim of achieving sustainable integration of land, economy, society, and environment [10,11,12].

Urban–rural integration is essentially a spatial restructuring of the human–land relationship [13]. Wu Chuanjun’s theory of the regional system of the human–land relationship views human society and the geographical environment as a dynamic, coupled structure [14]. Recently, building on this theory, scholars have developed conceptual frameworks such as the rural regional system and the urban–rural integration system [15], emphasizing that cities and rural areas function as coupled subsystems whose sustainable development depends on human–land coordination [16,17,18,19]. However, coupling has a threshold effect; beyond critical values, degradation replaces coordination [20,21]. Achieving coordination is especially critical in ecologically fragile regions with scarce geographic capital [22].

Existing studies have measured urban–rural environment coupling using the entropy method and the coupled coordination degree model [23,24,25,26] and have identified driving factors via regression analysis, Geodetector, and Grey relational analysis [27,28,29,30]. Research subjects have focused primarily on coastal cities and eastern plain regions [31,32,33,34]. The Pressure-State-Response (PSR) framework [35] and its derivatives provide a classic paradigm for deconstructing causal chains between human activities and the environment [36]. However, these frameworks assume linear transmission and treat information flows as secondary, posing adaptability challenges in the digital age [37]. In recent years, the enabling effects of digital technologies on urban and rural development have gradually emerged as a research frontier [38,39]. Existing research indicates that digital infrastructure can reshape regional spatial structures by reducing transaction costs and overcoming spatiotemporal barriers, and that the development of digital villages has a significant promotional effect on rural industrial integration [40,41]. Yet most studies treat digital elements as general drivers, lacking theoretical refinement on how digital technologies might compensate for geographic capital deficits in ecologically fragile areas.

Existing research on karst mountainous areas has focused on rocky desertification control [42,43], ecological resettlement [44], and land use monitoring [45]. The fundamental contradiction in southwestern karst regions is resource constraints arising from population–land tension and fragmented topography [46]. Some recent studies have examined coupling between rural settlement restructuring and land use transformation [47] or between urban–rural integration and rural resilience [48]. However, existing work shows a divergence between ecological conservation and urbanization pathways. Few studies integrate urban–rural development and environmental carrying capacity into a unified system for long-term tracking, especially under overlapping multiple strategies in karst areas. Located in the core karst landscape of the Yungui Plateau [49], Guiyang faces rigid spatial constraints due to fragmented topography and arable land scarcity. As a National Big Data Comprehensive Pilot Zone and ecological civilization model city, Guiyang has undergone human–land restructuring under multiple strategic initiatives over two decades. At the same time, the coexistence of rapid digital economic development and strict ecological red lines makes it possible to examine how digital technologies reduce spatial and temporal distances and how early environmental governance influences the trajectory of human–land coordination. Therefore, by establishing a comprehensive evaluation system that encompasses urban and rural development as well as environmental support, and by constructing an integrated ‘evolution–mechanism–space’ analytical chain combining the entropy method, the coupled coordination degree model, Grey relational analysis, Geodetector, and multi-source spatial analysis methods, we go beyond the common practice of using these methods separately and enable a systematic understanding of how temporal evolution translates into spatial responses. Theoretically, we reveal an ‘early governance’ pathway for the human–land system in karst mountainous areas, which deviates from the classic environmental Kuznets curve (EKC). Unlike plain regions that often experience a ‘pollute first, treat later’ trajectory, Guiyang initiated intensive ecological governance as early as 2012, when the coupling coordination degree had not yet reached 0.6, thereby avoiding severe environmental degradation. Moreover, we identify a two-factor enhancement effect between digital technology and other driving factors, extending the traditional PSR framework into the digital era. Practically, we quantify the topographic constraints on construction land expansion in karst areas: expansion is concentrated on 0–6° gentle slopes and 1000–1400 m mid-elevation zones, overlapping with high-quality farmland, while steep slopes (>25°) and high-altitude areas are strictly protected. We also discover a ‘semi-decoupling’ phenomenon between economic growth and new land occupation, with a synchronization rate of only 50%. These findings provide spatially explicit guidance for urban–rural integration governance in similar mountainous regions.

2. Materials and Methods

2.1. Theoretical Framework

This study is grounded in Wu Chuanjun’s theory of the human–land relationship system [14] and the urban–rural integration theories of Liu Yansui [13] and Fang Chuanglin [50], which collectively inform an overall objective: sustainable development of the human–land system and urban–rural integration. In response to the karst mountainous context, characterized by fragmented topography, ecological fragility, rocky desertification risk, and rigid spatial constraints imposed by ecological red lines. To systematically examine how the human–land system evolves, what drives its changes, and how these changes manifest spatially, a three-stage analytical chain is constructed that links temporal evolution, driving mechanisms, and spatial responses (Figure 1). The first stage calculates the composite scores of urban–rural development (U1) and environmental support (U2), together with the coupling coordination D-value, over the period 2000–2024, which is divided into four policy phases. The second stage applies Grey relational analysis to identify temporally synchronous drivers and Geodetector (Excel version, http://www.geodetector.cn/) to quantify the explanatory power of individual factors and their interactions, thereby revealing phase-specific shifts in the dominant drivers. The third stage maps spatial responses through nightlight center migration, land use ecological transition index, topographic zoning, and synchronization rate between economic activity and construction expansion. The findings ultimately support policy implications tailored to karst regions, including enhanced ecological governance, human settlement improvement, and optimized industrial selection.

2.2. Study Area

Guiyang is a typical karst mountainous city in southwest China, covering an area of 8034 km², with over 85% of its terrain consisting of karst landforms and an average elevation of 1100 m (Figure 2). It administers six urban districts, three counties, and one county-level city, with an urbanization rate exceeding 80% in 2023. Urban and rural construction land is concentrated within a limited basin area, imposing strict spatial constraints on the evolution of human settlements and land use. Rural areas, isolated by topography and lacking adequate infrastructure and public services, exhibit a marked urban–rural divide. Since 2000, Guiyang has implemented multiple overlapping initiatives, including the Western Development Strategy, the National Big Data Comprehensive Pilot Zone, the National Ecological Civilization Model City, and the “Provincial Capital Enhancement” initiative. This makes it an ideal case study for examining the relationship between geographical constraints and integrated urban–rural development.

2.3. Data Sources and Preprocessing

This study established a multi-source data system covering the period from 2000 to 2024, including socioeconomic statistics, remote sensing imagery, and basic geographic information (

Table 1. Data types and data sources.

Data Type	Time Range	Resolution	Data Source	Preprocessing
Indicator panel data	2001–2025	—	Guiyang Statistical Yearbook; Guizhou Provincial Statistical Yearbook; statistical bulletins and data released by relevant departments.	Interpolation; normalization.
Nighttime light data	2000–2024	500 m	National Qinghai–Tibet Plateau Science Data Center: Extended-time-series NPP-VIIRS-like artificial nighttime light dataset for China (1986–2024). https://data.tpdc.ac.cn.	Projection; resampling; cropping.
Land use data	2000–2020	30 m	Institute of Geographic Sciences and Natural Resources Research, Chinese Academy of Sciences (LUCC): China Multi-period Land Use Remote Sensing Monitoring Dataset. www.resdc.cn.	Projection; reclassification; cropping.
Digital elevation data	—	30 m	Geospatial Data Cloud: ASTER GDEM V3 30 m Resolution Digital Elevation Data. www.gscloud.cn.	Projection; cropping; slope and elevation extraction.
Administrative boundary data	—	1:250,000 km	Map Approval No. GS (2024)0650 Standard Base Map.	Projection; cropping.

).

Indicator data were drawn from the Guiyang and Guizhou Statistical Yearbooks (2001–2025). Missing values in trend indicators were imputed using linear regression, while missing values in volatility indicators were estimated using provincial growth rates. Negative indicators were normalized.

Nighttime light data were sourced from the “China Regional Extended-Time-Series NPP-VIIRS-like Artificial Nighttime Light Dataset (1986–2024)” released by the National Qinghai–Tibet Plateau Science Data Center [51]. Annual light images for Guiyang (2000–2024) were extracted and total and average light values were calculated. Data validity was verified using Pearson and Spearman correlation (Section 3.5.1).

Land use data (30 m resolution) for 2000, 2005, 2010, 2015, and 2020 were obtained from the Chinese Academy of Sciences (www.resdc.cn). The data were reclassified into six categories: built-up, cropland, forest, grassland, water, and unutilized land.

DEM data were obtained from ASTER GDEM V3 (30 m), from which slope and elevation were extracted. Slope was classified into four levels—0–6°, 6–15°, 15–25°, and >25°—with 25° serving as the threshold for restricting cultivation on steep slopes. Elevation was classified into four levels based on urban vertical differentiation: <1000 m, 1000–1200 m, 1200–1400 m, and >1400 m.

Administrative boundaries were extracted from the standard base map based on map review number GS (2024)0650. All raster data were uniformly projected to WGS 1984 UTM Zone 48N, resampled to 30 m resolution, and cropped using administrative boundaries as a mask.

2.4. Indicator System Construction

Following Wu Chuanjun’s theory of human–environment interaction [14], we deconstruct Guiyang’s urban–rural human–land system into two target layers: urban–rural development (U1) and environmental support (U2). U1 reflects human development intensity and urban–rural integration level, while U2 reflects the karst environment’s carrying capacity and ecological constraints. The two subsystems are mutually coupled and jointly determine human–land coordination (

Table 2. Development of the indicator system.

Target Layer	Criteria Layer	Indicator Layer	Code	Unit	Attribute	Weight
Urban–rural development	Population development	Urbanization rate	X1	%	+	0.079
		Population density	X2	km2/person	+	0.083
	Agricultural modernization	Agricultural mechanization rate	X3	%	+	0.158
	Land and space	GDP per unit of land	X4	10,000 CNY/km2	+	0.054
		Built-up area	X5	km2	+	0.094
	Economic development	Per capita GDP	X6	10,000 CNY	+	0.082
		Share of tertiary industry output	X7	%	+	0.062
	Transportation connectivity	Per capita public bus rides	X8	trips	+	0.040
		Highway mileage	X9	km	+	0.046
	People’s livelihood	Ratio of per capita disposable income between urban and rural permanent residents	X10	%		0.075
		Ratio of urban to rural consumer spending	X11	%	−	0.071
	Information connectivity	Number of broadband Internet subscribers	X12	Households	+	0.096
		Number of mobile phone subscribers	X13	Households	+	0.059
Environmental support	Pollution pressure	Average fertilizer application per mu of arable land	X14	kg	−	0.085
		Industrial wastewater discharge	X15	10,000 t	−	0.176
		Total industrial sulfur dioxide emissions	X16	10,000 t	−	0.239
	Environmental management	Comprehensive utilization rate of industrial solid waste	X17	%	+	0.046
		Urban domestic sewage treatment rate	X18	%	+	0.106
		Environmental protection investment	X19	10,000 CNY	+	0.063
	Ecological foundation	Forest coverage rate	X20	%	+	0.164
		Air quality excellence rate	X21	%	+	0.017
		Per capita Park green space area in built-up areas	X22	m2/person	+	0.103

).

The U1 system encompasses seven dimensions, population, agricultural modernization, land and space, economy, transportation, livelihoods, and information exchange, comprising 13 indicators. The selection of indicators was based on mainstream paradigms for evaluating urban–rural integration [50,52], with targeted enhancements tailored to the characteristics of karst mountainous regions. Specifically, per-unit-area GDP was added to measure the economic output efficiency of scarce terraced land in basin areas; road mileage was retained to characterize transportation accessibility costs in mountainous areas; and the number of broadband internet subscribers and mobile phone users was included to capture the reshaping effect of digital technology on the flow of urban–rural factors.

The U2 system encompasses three dimensions: pollution pressure, environmental governance, and ecological foundation. On the pressure side, we select fertilizer application rates, industrial wastewater discharge, and sulfur dioxide emissions, which together constitute the primary anthropogenic environmental stresses in karst regions [53]. On the governance side, we select the comprehensive utilization rate of solid waste, sewage treatment rates, and environmental protection investment, reflecting the intensity of environmental regulation. On the ecological foundation side, forest coverage serves as the core proxy indicator for the effectiveness of rocky desertification control [54], supplemented by the rate of good air quality and per capita park green space area to characterize urban ecological well-being.

The indicator system presented in Table 2 is intended for human-influenced karst landscapes where natural baseline conditions and anthropogenic pressures co-exist. In purely natural karst regions with negligible human disturbance, the system may be adapted by omitting the anthropogenic indicators (X14–X19) and retain only the natural baseline indicators (X20–X22). Hence, the full set of indicators is most appropriate for regions characterized by active human–land interactions.

2.5. Research Methods

This study comprehensively employs the entropy method, coupling coordination degree model, Grey relational analysis, Geodetector, and spatial analysis methods. Entropy method and coupling coordination degree model are used to measure the temporal evolution of the system; Grey relational analysis and Geodetector are used to identify driving mechanisms from two complementary dimensions—temporal synchrony and hierarchical explanatory power, respectively; while the shift in the center of gravity of nighttime lights, land use transition matrices, and topographic zoning statistics are employed to reveal the spatial response characteristics of coordination evolution. Through the progressive integration of these methods, an “evolution–mechanism–spatial” analytical chain is constructed.

2.5.1. Entropy Method and Coupling Coordination Degree Model

The entropy method is based on the principle of information entropy and assigns objective weights based on the dispersion of indicator data. Compared to subjective weighting methods such as the Analytic Hierarchy Process (AHP), the entropy method avoids cognitive biases introduced by expert scoring and is particularly suitable for the comprehensive evaluation of long-term, multi-indicator systems [55]. Composite scores were calculated for U1 and U2 via data standardization, information entropy calculation, and weight determination.

$${\mathrm{U}}_{\mathrm{k}}={\displaystyle \sum _{\mathrm{j}=1}^{\mathrm{nUk}}}{\mathrm{W}}_{\mathrm{j}}^{\mathrm{Uk}} \times { \mathrm{Z}}_{\mathrm{ij}}^{\mathrm{Uk}}\left(\mathrm{k}=1,2\right)$$

(1)

Here “$\mathrm{nUk}$” is the number of indicators in the corresponding subsystem, ${\mathrm{W}}_{\mathrm{j}}^{\mathrm{Uk}}$ is the weight of the jth indicator in the kth subsystem, and ${\mathrm{Z}}_{\mathrm{ij}}^{\mathrm{Uk}}$ is the standardized indicator value.

The coupling degree (C) and coupling coordination degree (D) were introduced to measure the mutual coupling intensity and coordination level between U1 and U2:

$$\mathrm{C} = \sqrt{{\displaystyle \frac{\mathrm{u}1 \times \mathrm{u}2}{{\left(\frac{\mathrm{u}1 + \mathrm{u}2}{2}\right)}^2}} }= 2 \times \sqrt{{\displaystyle \frac{\mathrm{u}1 \times \mathrm{u}2}{{(\mathrm{u}1 + \mathrm{u}2)}^2}}}$$

(2)

$$\mathrm{D}=\sqrt{\mathrm{C} \times \mathrm{T}}$$

(3)

$$\mathrm{T}=\mathsf{\alpha }\mathrm{U}1+\mathsf{\beta }\mathrm{U}2$$

(4)

Here, $\mathrm{C}$ represents the coupling degree; $\mathrm{D}$ represents the coupling coordination degree; $\mathrm{T}$ represents the comprehensive coordination index; and $\mathsf{\alpha }$ and $\mathsf{\beta }$ are undetermined weights that reflect the relative importance of the two subsystems. Existing research generally holds that urban–rural development and the ecological environment hold equal importance in sustainable development. $\mathsf{\alpha }$ and $\mathsf{\beta }$ are set to 0.5.

Referring to existing studies [56,57], the value ranges for the coupling coordination degree (D) and the coupling degree (C) are divided into five levels: serious imbalance (0–0.2), relative imbalance (0.2–0.4), low coordination (0.4–0.6), good coordination (0.6–0.8), and high-quality coordination (0.8–1).

2.5.2. Grey Relational Analysis (GRA)

Grey relational analysis (GRA), suitable for small samples without distributional assumptions [58], was applied to time-segmented analysis to address Geodetector’s limitation in reflecting temporal dynamics. Using D-value as the reference sequence and each driving factor as a comparison sequence, the correlation coefficient r is calculated after initialization and dimensionless processing:

$${\mathrm{R}}_{\mathrm{i}} = {\displaystyle \frac{\mathrm{i}}{\mathrm{n}}}{\sum }_{\mathrm{k}=1}^{\mathrm{n}}{\mathsf{\xi }}_{\mathrm{i}}\left(\mathrm{k}\right)$$

(5)

The value of R_i ranges from 0 to 1; a higher value indicates that the trend of this factor is more closely aligned with that of the D-value and that its driving effect is stronger. The correlation coefficients for the entire period (2000–2024) and for each of the four policy phases is calculated.

2.5.3. Geodetector

Geodetector is a statistical method for detecting spatial heterogeneity and identifying driving factors [59]. By extending this method to the temporal dimension [60], it uses years as the stratification unit to examine the extent to which differences in factors across years explain variations in the D-value; this extension has already been applied in studies of long-term human–land relationships [61]. The explanatory power of the factors is measured by the q-value:

$$\mathrm{q} = 1 - {\displaystyle \frac{{\sum }_{\mathrm{k}=1}^{\mathrm{L}}\mathrm{Nk}{\mathsf{\sigma }}_{\mathrm{k}}^2}{\mathrm{N}{\mathsf{\sigma }}^2}} = 1 - {\displaystyle \frac{\mathrm{SSW}}{\mathrm{SST}}}$$

(6)

In the equation, ${\mathsf{\sigma }}^2$ represents the variance of the selected factor; N represents the number of spatial units; L represents the number of factor types; $\mathrm{Nk}$ and ${\mathsf{\sigma }}_{\mathrm{k}}^2$ represent the sample size and variance of spatial unit k, respectively. q ranges from 0 to 1; a higher q-value indicates greater explanatory power of the influencing factor for the spatiotemporal evolution of human–land coordination.

The interaction detector can identify the explanatory power of the combined effect of two different factors on human–land harmony. Specifically, by comparing the relationship between univariate and bivariate Q-values, it classifies interaction types into five categories: nonlinear attenuation, single-factor nonlinear attenuation, two-factor enhancement, independence, and nonlinear enhancement.

From the indicator system, 12 factors were selected for Geodetector analysis, spanning both the urban–rural development system (U1) and the environmental support system (U2) (

Table 3. Driving factors selected for Geodetector.

Code	Driving Factor	Unit	Source System
X1	Urbanization rate	%	U1
X6	Per capita GDP	10,000 CNY	U1
X7	Share of tertiary industry output	%	U1
X9	Highway mileage	km	U1
X10	Ratio of per capita disposable income between urban and rural permanent residents	%	U1
X12	Number of broadband Internet subscribers	households	U1
X14	Average fertilizer application per mu of arable land	kg	U2
X15	Industrial wastewater discharge	10,000 t	U2
X18	Urban domestic sewage treatment rate	%	U2
X19	Environmental protection investment	10,000 CNY	U2
X20	Forest coverage rate	%	U2
X22	Per capita park green space area in built-up areas	m2/person	U2

). The selection followed three criteria: (1) representation of key dimensions in urban–rural human–land coordination; (2) data availability for the entire study period (2000–2024); and (3) relevance to policy-driven evolutionary mechanisms in Guiyang. The selected factors are listed in Table 3. These 12 factors encompass the core dimensions of both urban–rural development and environmental support, enabling a comprehensive assessment of the driving mechanisms behind D-value evolution. It is worth noting that although several drivers used in Geodetector are derived from the U1 and U2 assessment systems, this does not constitute a circular argument. First, the dependent variable in Geodetector is the D-value, which is a nonlinear aggregation of all 22 indicators, whereas the drivers are the original raw indicators, which are mathematically distinct; second, Geodetector captures the spatial hierarchical heterogeneity of the D-value across years, while the entropy method captures the temporal variation in each subsystem; third, the comparative analysis of Geodetector and the Grey relational analysis significantly enhances the robustness of the research results.

2.5.4. Spatial Analysis

To reveal the spatial mapping characteristics of the coordinated evolution of humans and the environment, the following analyses were conducted using integrated data on nighttime lights, land use, and DEM: (1) Shift in light centers of gravity. The weighted average center method was used to calculate the annual light centers of gravity, with pixel brightness values serving as weights, to analyze the spatial migration trajectories of economically active areas. (2) Land use transitions matrix and land use ecological transition index. Based on five-period land use data, transitions matrices between adjacent periods were generated. Adverse and beneficial transitions were defined (

Table 4. Directions of land use conversion.

Conversion Direction	Type	Criteria
Arable land → construction land, forest land → construction land, grassland → construction land, water bodies → construction land, unutilized land → construction land	Adverse transitions	Degradation of ecological functions or occupation of high-quality arable land, detrimental to the harmony between people and land
Construction land → arable land, construction land → forest land, construction land → grassland, construction land → water bodies	Beneficial transitions	Restoration of ecological functions or land reclamation, conducive to the harmony between people and land

), and the land use ecological transition index (LETI) was calculated:

$$\mathrm{LETI} = {\displaystyle \frac{{\mathrm{A}}_{\mathrm{Beneficial}}}{{\mathrm{A}}_{\mathrm{Adverse}}}}$$

(7)

to measure the human–land coordination effects in terms of land-use conversion trends. (3) Topographic zoning statistics. Slope data (0–6°, 6–15°, 15–25°, >25°) and elevation data (<1000 m, 1000–1200 m, 1200–1400 m, >1400 m) were extracted from the DEM to calculate the proportion of newly added construction land in different topographic categories for each time period. (4) Overlap analysis. Using the 2020 nighttime high-intensity light zones to represent economically active areas, and the union of new construction land from 2000 to 2020 to represent expansion zones, the synchronization rate between the two was calculated:

$${\mathrm{R}}_{\mathrm{sync}}={\displaystyle \frac{{\mathrm{A}}_{\mathrm{sync}}}{{\mathrm{A}}_{\mathrm{light}}}}\times 100\%$$

(8)

to assess the spatial compatibility of industry–city integration.

3. Results

3.1. Division of Evolutionary Periods

Based on policy milestones and observed U1/U2/D trajectories, we divided 2000–2024 into four phases (

Table 5. Time Period Classification and Characteristics.

Phase	Year	U1 Range	U2 Range	D Range	Core Policies
I	2000–2009	0.033–0.253	0.076–0.363	0.223–0.550	Western development strategy
II	2010–2014	0.290–0.528	0.456–0.503	0.603–0.718	Pilot demonstration zone for ecological civilization, big data pilot phase
III	2015–2019	0.569–0.688	0.520–0.623	0.737–0.809	Ecological civilization model city, big data comprehensive pilot zone
IV	2020–2024	0.703–0.953	0.688–0.744	0.834–0.903	“Strengthening the Provincial Capital” initiative, digital vitalization + ecological development

).

3.2. Evolutionary Characteristics of System Scores

3.2.1. Urban–Rural Development System (U1): Accelerated Growth and Fluctuations in Growth Rates

The U1 score rose from 0.033 in 2000 to 0.953 in 2024 (Figure 3a). The growth trajectory exhibits a fluctuating pattern of slow–fast–slow–fast, with annual growth rates varying across different phases: from 0.022 during the infrastructure-driven phase to 0.048 during the policy-driven phase, 0.024 during the structural transformation phase, and 0.050 during the aggregation-driven phase. Overall, U1 growth has gradually shifted from being driven by factor inputs to being propelled by a dual-engine of policy dividends and structural optimization.

3.2.2. Environmental Support System (U2): Shallow U-Shaped Evolution and Early Intervention Effects

The U2 score followed a fluctuating upward trend, rising from 0.076 in Phase I to 0.744 in Phase IV (Figure 3b). In the first stage, the U2 score fluctuated between 0.076 and 0.363. In Phase II, the U2 score rose from 0.456 to 0.503. In Phase III, the U2 index rose from 0.520 to 0.623. Forest coverage exceeded 55%, and good air quality stabilized at over 95%. In Phase IV, the system entered a high-level plateau, with various environmental indicators tending toward saturation, and the system shifted from environmental restoration to steady-state maintenance.

The improvement in the U2 score is attributable to Guiyang’s environmental protection policies, such as the implementation of ecological protection red lines, which have set the city on a path where economic development and environmental protection advance in tandem. Guiyang initiated ecological civilization construction before the D-value reached 0.6, thereby preventing severe environmental degradation in areas prone to rocky desertification.

3.2.3. Urban–Rural Human–Land Integrated System Score (U): From Divergence to Synergy

From 2000 to 2024, the total score (U = U1 + U2) of Guiyang’s urban–rural human–land system rose from 0.054 to 0.826, an approximately 15-fold increase; however, the sources of growth varied significantly across different phases (Figure 3c). In Phase I, the U score grew by an average of 0.028 annually, driven primarily by infrastructure and industrial investment, while U2 fluctuated considerably. In Phase II, the growth rate accelerated to 0.036. The construction of the Pilot Demonstration Zone for Ecological Civilization stabilized U2 and accelerated U1, with the two systems beginning to form a “dual-engine drive.” In Phase III, the average annual growth rate was 0.028. The combined effects of policies for the National Ecological Civilization Model City and the Big Data Comprehensive Pilot Zone narrowed the growth gap between U1 and U2, and digital empowerment and ecological governance began to integrate deeply. In Phase IV, the average annual growth rate reached 0.032. The “Strengthening Provincial Capitals” initiative drove high growth in U1, while U2 entered a plateau phase without declining, resulting in a “simultaneous improvement in both quantity and quality” of the total score. Overall, since 2015, the growth curves of U1 and U2 have shifted from “divergence” to “convergence,” with enhanced synergy between development and the environment; however, U2 has approached saturation. In the future, it will be necessary to leverage digital technology and institutional innovation to tap into environmental resilience in order to maintain high-level coordination.

3.3. Evolution and Transition of the D-Value

3.3.1. Phases in the Evolution of the D-Value

The D-value rose from 0.223 in 2000 to 0.903 in 2024, with the coordination level undergoing a complete transition from relative imbalance–low coordination–good coordination–high-quality coordination (Figure 4,

Table 6. D-value of urban–rural human–land systems at each stage.

Phase	Year	D-Range	Coordination Level	Dominant Driving Factors
I	2000–2009	0.223–0.550	Relatively imbalanced to low-level coordination	X4, X6, X15
II	2010–2014	0.641–0.718	Good coordination	X1, X10, X22
III	2015–2019	0.737–0.809	Good coordination to high-quality coordination	X1, X2 X7
IV	2020–2024	0.834–0.903	High-quality coordination	X2, X10, X22

). The coupling degree (C) consistently remained above 0.91, indicating a stable interdependence between U1 and U2.

In Phase I, the D-value grew by an average of 0.036 annually. It surpassed 0.4 in 2005, entering the low-level coordination stage; however, the foundation of coordination was not solid, and the fluctuating rise in U2 signaled the continuous accumulation of environmental pressure. In Phase II, the annual growth rate declined to 0.029, yet coordination quality improved: following the approval of the Ecological Civilization Demonstration Zone in 2012, the decline in U2 narrowed, and the driving force of coordination shifted from a single-engine U1 to a dual-engine U1 + U2 model. In Phase III, the D-value exceeded 0.8 in 2019, marking the first entry into high-quality coordination. The expansion of the digital industry and ecological governance formed a synergistic force, significantly reducing the environmental cost per unit of GDP. In Phase IV, the growth rate of the D-value further slowed to 0.017, but continued to optimize at a high plateau. While the strong provincial capital drove factor agglomeration, U2 did not decline, indicating that the system has acquired the regulatory capacity to protect the environment amidst agglomeration.

3.3.2. Karst-Specific Characteristics of the Transition Path

Guiyang’s path of coordinated transition differs significantly from that of the eastern plains. The “pollute first, treat later” sequence implied by the classic environmental Kuznets curve (EKC) was compressed and even partially reversed in Guiyang. The construction of an ecological civilization demonstration zone was initiated as early as 2012, when the D-value had not yet reached 0.6. The bottom of the U-shaped curve for U2 (0.520) was higher than the initial value (0.076), thereby avoiding a phase of severe environmental deterioration. This strategy of advancing the timeline benefited from the policy-driven effect of rigid constraints imposed by the ecological baseline of the karst mountainous region; the irreversibility of rocky desertification control necessitated the early intervention of ecological red lines. The Guiyang case demonstrates that ecologically fragile areas can achieve an alternative path that bypasses the EKC inflection point by advancing the timeline of governance.

3.4. Driving Mechanisms

3.4.1. Identification of Driving Factors Across All Time Periods

Based on data spanning the entire period from 2000 to 2024, the q-values of Geodetector and the R-values of GRA for each factor influence on the D-value were calculated. Both methods identify highway mileage, forest coverage, per capita park green space, and tertiary industry share as high-impact factors (

Table 7. Top 10 driving factors for the entire period based on GRA and Geodetector.

Rank	GRA	R-Value	Geodetector	Q-Value	p-Value
1	X4	0.991	X9	0.914	<0.001
2	X17	0.987	X20	0.907	<0.001
3	X5	0.985	X22	0.868	<0.001
4	X19	0.984	X18	0.862	<0.001
5	X9	0.983	X7	0.862	<0.001
6	X15	0.983	X6	0.842	<0.001
7	X22	0.979	X10	0.836	<0.001
8	X11	0.976	X1	0.807	<0.001
9	X2	0.976	X12	0.745	<0.001
10	X20	0.974	X19	0.539	<0.001

Note: Geodetector lists only factors with p < 0.001. GRA is based on initialization calculations.

, Figure 5a). Their high explanatory power reflects karst-specific constraints: transportation overcomes fragmentation, forest coverage is a rigid constraint, and service-oriented growth lowers environmental costs.

The divergence between the two methods is equally insightful, revealing different dimensions of the driving forces’ effects. An example of this divergence is the number of broadband Internet subscribers: it has low Grey correlation (r = 0.696) but high q-value (0.745). This discrepancy arises from different principles: GRA captures temporal trend consistency, while Geodetector’s hierarchical processing identifies structural effects. More importantly, the results of the interaction analysis using Geodetector tool show that all interaction types fall under the category of two-factor enhancement. the interaction between the number of broadband Internet subscribers and other driving factors has a maximum explanatory power of q = 0.949 (Figure 5b), indicating digital compensation for geographic capital through synergy with agglomeration.

Furthermore, environmental governance factors exhibit high correlation but low q-values, indicating that while they have improved in tandem with coordination over the long term, their explanatory power regarding annual fluctuations has been diluted by continuously growing factors such as transportation and ecology; pollution emission factors, having fallen to low levels and stabilized since 2012, no longer exert significant explanatory power for annual fluctuations in the D-value (q > 0.05), yet they maintain a high degree of temporal correlation. The application of this complementary methodology effectively enhances the robustness of driver identification.

3.4.2. Phased Evolution of Driving Forces: From Economic Scale to Green Well-Being

An analysis covering the entire period may obscure the shifting roles of driving forces across different developmental stages. Therefore, GRA was calculated separately for each of the four policy phases to track the dynamic changes in the ranking of core factors (

Table 8. Top-eight driving factors by period based on GRA.

Rank	2000–2009	2010–2014	2015–2019	2020–2024
1	X6	X10	X1	X22
2	X4	X22	X2	X2
3	X15	X1	X7	X10
4	X19	X18	X9	X14
5	X17	X11	X10	X4
6	X16	X2	X20	X7
7	X5	X9	X18	X9
8	X8	X16	X5	X17

).

In Phase I, the driving force system exhibited distinct characteristics of scale expansion. Per capita GDP and GDP per unit of land ranked first and second, respectively, while industrial wastewater discharge ranked third, forming a driving model characterized by “growth and pollution occurring in tandem.” During this period, industrialization and infrastructure investment driven by the Western Development Strategy were the core drivers of improved coordination, while environmental governance, though already underway, remained largely limited to end-of-pipe measures.

In Phase II, the driving system underwent significant changes. The urban–rural income ratio rose to the top position, with per capita park green space and urbanization rates ranking second and third, respectively. This marked a shift following the launch of the National Ecological Civilization Pilot Demonstration Zones in 2012, where narrowing the urban–rural gap and enhancing green well-being began to replace mere economic expansion, becoming the new engines driving improved coordination. Although pollution indicators such as industrial sulfur dioxide emissions remained highly correlated, the synchrony between governance investments and improvements in coordination had clearly strengthened, and the system entered a transitional tug-of-war phase characterized by “pressure governance.”

Phase III witnessed the synergistic effects of digital empowerment and ecological governance. The dominant positions of urbanization rates and population density indicate that the continued concentration of population in urban areas remains the core driving force; the share of the tertiary sector’s output value entered the top three for the first time, reflecting the emerging structural driving role of the digital economy, such as the big data industry. At the same time, the correlation of ecological indicators, such as forest coverage and sewage treatment rates, remained high, marking the concentrated release of achievements in the construction of ecological civilization model cities. In this phase, digital technology and green governance began to form synergies, driving a gradual reduction in the environmental cost per unit of GDP.

In Phase IV, the system of driving forces was reshaped. Per capita park green space in built-up areas surged to the top of the rankings, followed closely by population density and the urban–rural income ratio, while traditional economic indicators fell out of the top five. This clearly indicates that in the new phase of high-quality development, green well-being and agglomeration benefits have become the key factors in maintaining high levels of coordination. The driving forces of endogenous development factors, such as improvements in urban spatial quality and the equalization of public services, have already surpassed mere economic growth in total volume. It is worth noting that the rankings for agricultural mechanization rates and the number of broadband internet subscribers were relatively low. The former is constrained by the limited application scenarios resulting from the fragmented karst terrain, while the latter is due to saturation in penetration rates and a mismatch between the growth curve and the pattern of the coordination index. However, this does not negate the foundational role that both play through their interactive effects.

3.4.3. Karst-Specific Characteristics and Theoretical Implications of the Evolution of Driving Mechanisms

Taken as a whole, the driving mechanism behind the coordinated development of the human–land system in urban and rural Guiyang has followed a unique path: from being driven by economic scale, through structural optimization and governance, to being driven by green well-being. This evolutionary trajectory aligns closely with policy directions and more profoundly reflects the intrinsic laws governing the human–land relationship in karst mountainous regions. The rigid constraints of the ecological baseline have prompted a significant advance in the timing of environmental governance interventions. Consequently, urban–rural equity and green well-being emerged as dominant factors as early as Phase II, thereby preventing severe pollution and shortening the deterioration phase of the EKC. The geographical disadvantage of fragmented terrain amplifies the marginal contribution of transportation infrastructure, making road mileage a high-impact factor spanning all stages. Meanwhile, the two-factor enhancement effects of digital technology provide a new driving mechanism for the integration of towns and villages in mountainous regions with scarce geographic capital. It does not merely add linearly to traditional factors, but rather reshapes development efficiency through deep interaction with population and industry, serving as the driving force behind the “digital aggregation synergy” model. This indicates that, within the PSR framework, digital factors do not belong to a single stage but play a unique role in facilitating full-chain interaction across all stages.

3.5. Spatial Responses and Coordination Analysis

3.5.1. Spatiotemporal Evolution of Nighttime Lighting

Nighttime light intensity, as a proxy indicator of economic activity, increased steadily from 2000 through 2024 (Figure 6). Total light intensity is highly correlated with the U1 and D-values (

Table 9. Correlation between total light output and U1 and D-values.

Year	Pearson		Spearman’s
	Light vs. D	Light vs. U1	Light vs. D	Light vs. U1
2000–2024	0.946 **	0.958 **	0.946 **	0.966 **
2000–2009	0.867 **	0.823 **	0.939 **	0.927 **
2010–2014	0.917 *	0.904 *	0.9 *	0.9 *
2015–2019	0.839	0.838	0.8	0.821
2020–2024	−0.052	−0.308	−0.1	−0.3

** Significant at the 0.01 level. * Significant correlation at the 0.05 level.

). After the third phase, this correlation weakened, reflecting a shift from urban and rural economic development toward digital and service industries, as well as a transition in urban and rural land use from extensive expansion to the renewal of existing stock; consequently, the increase in lighting was driven more by urban lighting projects than by overall economic growth.

The light center migrated southeast at ~1.1 km/year during 2000–2010, slowed to ~0.5 km/year during 2010–2015, and stabilized after 2015 (annual migration < 0.2 km) near the Guanshanhu–Yunyan district boundary (Figure 7). This stabilization coincided with D-value exceeding 0.8 (high-quality coordination), suggesting an intrinsic link between spatial stability and human–land coordination.

3.5.2. Land Use Changes and the Land Use Ecological Transition Index (LETI)

Changes in land use reflect the relationship between humans and the land. The rate of expansion of built-up areas surged from 1.62% in the first stage to 19.28% in the fourth stage. The area of land converted to non-agricultural use increased from 16.90 square kilometers to 280.26 square kilometers, a 15-fold increase. LETI decreased from 0.162 to 0.050, a 69% decline (

Table 10. Expansion of construction land and land conversion.

Period	Average Annual Expansion Rate (%)	Area of Beneficial Conversion (km2)	Area of Adverse Conversion (km2)	LETI
2000–2005	1.620	2.737	16.897	0.162
2005–2010	4.660	4.186	48.178	0.087
2010–2015	3.740	2.948	46.437	0.063
2015–2020	19.280	13.961	280.259	0.050

, Figure 8).

LETI showed a significant negative correlation with the D-value (r = −0.89, p < 0.05). This discrepancy reveals that non-land factors, such as road networks, forest cover, the share of the tertiary sector, and digital technologies, partially offset the negative pressure caused by adverse land conversion. However, risks persist: between 2015 and 2020, areas subject to adverse land conversion accounted for over 70% of the total across the four periods, and were primarily concentrated on slopes of 0–6°, which constitute high-quality farmland in karst regions. With per capita arable land in Guiyang at less than one-third of the national average, the depletion of low-gradient farmland is irreversible. There is a time lag in the loss of water conservation and carbon sequestration functions resulting from forest land conversion; the current U2 plateau may not fully reflect the cumulative ecological costs of recent large-scale development.

3.5.3. Expansion of Construction Land Under Topographic Constraints

The fragmented karst topography imposes strict constraints on spatial expansion. In the study area, specially protected lands are defined as areas with slope > 25° or elevation > 1400 m, where construction is strictly prohibited to prevent rocky desertification and preserve ecological functions. These thresholds are derived from the Guizhou Ecological Red Line policy and the national standard for limiting cultivation on steep slopes. Conversely, lands that can be converted for development under certain conditions are those with gentle slopes (0–6°) and mid-elevation (1000–1400 m), corresponding to the basin areas where most urban and rural construction is concentrated. Conversion of such lands is permitted only when arable land compensation is fulfilled. Slopes between 6° and 25° at other elevations are classified as restricted development lands, where only infrastructure and public facilities are allowed. These standards ensure that ecological protection and development are balanced in the karst mountainous context.

Newly developed land on steep slopes (>25°) remained below 0.5% in all periods (

Table 11. Proportion of new construction land by slope grade.

Construction Land by Slope Grade (%)	2000–2005	2005–2010	2010–2015	2015–2020
0–6°	0.363	1.379	1.384	7.555
6–15°	0.264	0.639	0.626	3.893
15–25°	0.102	0.220	0.176	1.392
25–90°	0.030	0.104	0.070	0.498

, Figure 9), indicating effective protection of areas at high risk of rock desertification. In contrast, the proportion of gentle slopes (0–6°) surged from 0.363% in Phase I to 7.555% in Phase IV, accounting for >60% of all newly developed land between 2015 and 2020, and spatially overlapping with high-quality farmland. The proportion of slopes with gradients of 6–15° rose from 0.264% to 3.893%, indicating that development has extended from valley bottoms to foothills. This area has shallow soil (<50 cm) and receives approximately 1100 mm of concentrated annual rainfall, increasing the risk of erosion.

Elevation analysis reveals intensive development in mid-elevation areas, limited development at low elevations, and cautious development in high-elevation areas (

Table 12. Proportion of new construction land by altitude.

Construction Land by Elevation (%)	2000–2005	2005–2010	2010–2015	2015–2020
0–1000 m	0.040	0.074	0.252	0.769
1000–1200 m	0.169	0.697	0.512	3.900
1200–1400 m	0.312	0.710	0.762	4.358
1400 m and above	0.043	0.485	0.272	1.073

, Figure 10). The mid-elevation zone (1000–1400 m) concentrates approximately 80% of the urban population, with cumulative new development accounting for 8.258% between 2015 and 2020. Low-elevation river valleys (<1000 m) consist primarily of high-quality farmland, with development intensity remaining stable at <0.769%. High-altitude areas (>1400 m) exceeded 1% for the first time in Phase IV (1.073%), covering an area of ~5.1 square kilometers. These are mostly ecological restoration zones following the protection of rocky desertification areas; they exhibit low stability, and development may trigger irreversible reversal of rocky desertification. Three cumulative risks emerge: accelerated depletion of high-quality basin farmland, increased soil erosion risk on gentle slopes, and rising disturbance risk in high-altitude ecological restoration areas.

3.5.4. Synchronization Rate of Economic Development and Construction Expansion

Using 2020 high-intensity nighttime light zones to represent economically active areas, and the union of newly added construction land from 2000 to 2020 to represent expansion zones, the overlap area was 94.48 km², with a synchronization rate of 50.104% (

Table 13. Synchronization rate of economic development and construction expansion.

Category	Area (km2)	Proportion (%)
High-lighting Areas	188.569	100.000
Total expansion area	388.818	—
Synchronization zone	94.481	50.104

, Figure 11). About half of the economically active areas lie within construction expansion zones; the other half are distributed in urban renewal zones or non-land-intensive industrial areas. It provides spatial evidence for the driving mechanism findings: the increase in D-value during Phase IV does not depend on construction land expansion but results from digital empowerment (interaction q = 0.956), industrial structure upgrading and green well-being improvements. The declining LETI alongside a high D-value indicates that about half of economic growth is not accompanied by new land occupation.

4. Discussion

4.1. Theoretical Framework for Urban–Rural Integration in Karst Mountainous Areas

Unlike China’s eastern plains, which follow an urban–rural integration process of “scale expansion → industrial clustering → population influx” and rely on vast tracts of land and a relatively homogeneous environment [13,62], the coordination of Guiyang’s urban–rural human–land system is largely attributable to improved transportation infrastructure, clear priorities for ecological governance, and the widespread adoption of digital technologies.

Several details clarify how this alternative pathway works. Digital technology partly compensates for geographic capital through spatiotemporal compression [63,64]. The interaction term between the number of broadband internet subscribers and other driving factors is of the two-factor enhancement type, with a maximum explanatory power of q = 0.949, exceeding that of any single factor. Typically, the environmental Kuznets curve implies environmental degradation before improvement [65,66]; however, Guiyang implemented intensive ecological governance even when its per capita GDP was only approximately $3900 and its D-value had not yet reached the threshold for high-quality coordination. At the same time, the spatial structure adapts to the topography: Guiyang exhibits a polycentric, clustered pattern. Compared to the mountainous terrain of Chongqing [67], the smaller settlements here and wider ecological buffer zones reflect the more fragmented nature of the karst landscape.

However, tensions remain. Between 2015 and 2020, the depletion of high-quality farmland on gentle slopes and the decline of the LETI to 0.050 indicate that this pathway relies heavily on land-driven development. The core test of this pathway’s long-term sustainability lies in whether the positive impacts of the digital dividend and ecological governance can continue to offset the cumulative costs of land development [68,69].

4.2. Extending the PSR Framework in the Digital Age: Three Effects of Digital Empowerment and Evaluation

The traditional PSR framework assumes that information flows serve as an adjunct to responses and that stress develops along a linear path [70]. The digital age challenges these assumptions: information flows have become an independent force that reshapes stress generation, while real-time sensing enables responses to shift upstream.

This study finds that digital technologies permeate all stages of pressure (P), state (S), and response (R), expanding the PSR framework into three effects (Figure 12). First, the pressure reduction effect: e-commerce and smart tourism reduce resource consumption and pollution per unit of GDP at the source of pressure. Second, the state-awareness compression effect: Guiyang’s environmental big data platform compressed the environmental feedback cycle from annual to near-real-time, creating a window for early intervention, which explains why Guiyang’s ecological governance preceded environmental degradation. Third, the response decision optimization effect: the interaction between the number of broadband internet subscribers and other driving factors is of the two-factor enhancement type, with a maximum explanatory power of q = 0.949, indicating that digital technologies have reshaped the efficiency of urban–rural factor flows and the precision of governance responses [71].

Guiyang’s four-stage evolution provides empirical validation. In Phase I, digital infrastructure was underdeveloped, and human–land relations followed the traditional PSR logic. In Phase II, digital technology emerged but served only as an auxiliary tool for the response phase. In Phase III, with the rapid increase in internet penetration, the three effects began to manifest systematically; digital empowerment and ecological governance achieved “mutual reinforcement” for the first time. In Phase IV, digital empowerment and response were deeply integrated, forming a closed-loop of intelligent sensing and decision-making. Digital-based responses began to partially replace engineering-based responses; the city’s ranking in environmental investment dropped from fourth in the first stage to eighteenth in the fourth stage, while the U2 index remained persistently high.

4.3. Sustainability Implications of the Divergence Between Ecological Land Use Conversion and Coordination

The significant negative correlation between LETI and D-value (r = −0.89, p < 0.05) indicates that, despite land use deviating from ecological friendliness, system coordination continues to increase. This finding can be explained by the distinction between weak sustainability and strong sustainability.

Weak sustainability assumes that natural capital and man-made capital are interchangeable; sustainability can be achieved as long as the total capital stock does not decline [72]. From this perspective, the growth in Guiyang’s transportation infrastructure, digital technology, and forest cover has offset the loss of high-quality arable land in the total capital account, driving the D-value upward.

Strong sustainability emphasizes that certain key natural capital stocks are irreplaceable and must be maintained independently [73]. From this perspective, the loss of high-quality farmland on gentle slopes is irreversible. Soil formation rates in karst regions are extremely low; once arable land in the basin is occupied by construction, its food production capacity and ecosystem services cannot be fully restored through off-site compensation [74]. A decline in the LETI to 0.050 can be regarded as a quantitative signal of compromised resilience. Therefore, focusing solely on area-equivalent compensation while ignoring the ecological and economic functions of different land types will lead to irreversible losses. Elevating compensation from area equivalence to functional equivalence is the theoretical requirement for advancing toward robust sustainable governance.

4.4. Mountainous Urbanization Under Topographic Constraints

Mountainous urbanization follows divergent paths due to geographical context. The Alpine region and Chongqing represent two distinct yet well-documented environments of urbanization in mountainous areas, with the Alps exemplifying long-term slope management in a European setting and Chongqing reflecting rapid polycentric expansion under topographic constraints in China. By comparing the mountainous urbanization processes in these two representative cases with Guiyang’s situation, we can better understand how Guiyang pursues a path of “digital–ecological synergy” and how karst constraints shape its governance trajectory. The Alpine region achieves balance through slope restrictions and compact development [75], while Chongqing uses a “multi-centric, cluster-based” structure to respond to parallel ridge–valley topography [67]. Guiyang’s expansion follows a pattern of protecting peripheries and developing central areas. Compared to the Alps, Guiyang’s steep slope restrictions are driven more by karst desertification control than by landscape risk management. Compared to Chongqing, Guiyang’s urban clusters are smaller with wider ecological buffer zones, reflecting the strong constraints of peak cluster depression landscapes on contiguous development.

This development process may be associated with three types of cumulative risks. First, farmland depletion in basin areas: the scarcest high-quality cropland is concentrated on 0–6° slopes at 1000–1400 m elevation. During 2015–2020, this zone accounted for 7.555% of new construction land. At current rates, developable space may saturate within 10–15 years. Second, soil erosion risk on slopes: in the 6–15° gentle slope zone, soil depth typically falls below 50 cm; surface hardening increases runoff coefficients, and combined with concentrated rainfall, erosion risk rises. Third, the risk of reversing rocky desertification at high altitudes: new development above 1400 m exceeded 1% in Phase IV. These are mostly post-remediation ecological restoration zones with low community stability; once disturbed, rocky desertification becomes irreversible [76]. These risk indicators suggest that the short-term economic rationality of the current expansion model comes at the cost of long-term natural capital depletion, and its sustainability depends on whether a transition from incremental expansion to stock optimization can be achieved before critical thresholds are reached.

4.5. Research Limitations and Future Directions

This study has several limitations. First, regarding data limitations, our analysis ends in 2024, and short-term changes in 2025–2026 are unlikely to significantly alter the main conclusions, as the key transitions occurred before 2015 and the D-value has plateaued since 2020. Nevertheless, future updates with longer time series would be valuable. Due to limited long-term remote sensing data, the study did not directly incorporate rocky desertification area but used forest cover as a proxy. Although the two are highly correlated, proxy use still results in information loss. Subsequent studies should use higher-resolution imagery to construct long-term rocky desertification datasets.

Second, methodological limitations arise regarding the potential overlap between driving factors and evaluation metrics. Some of the drivers used in the Geodetector analysis also appear as original metrics in the U1 and U2 evaluation systems. While we do not believe this constitutes circular reasoning, we acknowledge that complete independence would be methodologically clearer. However, given limitations in data availability, we were unable to compile a fully independent set of drivers over the 25-year study period. To partially address this concern, we employed cross-validation using GRA and Geodetector to enhance the robustness of our conclusions. In future research, we will use fully independent drivers to validate our findings.

Third, there are limitations regarding generalizability. Guiyang enjoys unique policy advantages; whether its “digital–ecological synergy” approach can be replicated in other karst regions with weaker policy support and underdeveloped digital infrastructure—such as Bijie, Southwest Guizhou, Northwest Guangxi, and Southeast Yunnan—requires verification through comparative case studies.

Fourth, the depth of micro-level mechanisms remains limited. While this study highlights the interaction between digital factors and population aggregation (q = 0.956), it has not yet delved into the micro-level mechanisms. Future research could combine farmer surveys and business interviews to explore the micro-level mechanisms of digital empowerment within human–land systems.

5. Conclusions

Taking Guiyang City as an example, this study integrates the entropy method, the coupled coordination degree model, Grey relational analysis, Geodetector, and multi-source spatial analysis methods to examine the evolutionary characteristics, driving mechanisms, and spatial responses of the human–land system during the urban–rural integration of the karst mountainous region from 2000 to 2024. The results are as follows:

First, the comprehensive score of Guiyang’s urban–rural human–land system increased from 0.054 to 0.826, and the coupling coordination degree rose from 0.223 (relative imbalance) in 2000 to 0.903 (high-quality coordination) in 2024, while the environmental support system deviated from the classic environmental Kuznets curve. This indicates that Guiyang, through ecological governance interventions initiated in 2012, avoided severe environmental degradation and maintained a balance between economic development and environmental protection.

Second, the driving forces underwent a structural shift from being primarily driven by economic scale to being primarily driven by green well-being. The interaction between the number of broadband internet subscribers and other driving factors is of the two-factor enhancement type, with a maximum explanatory power of q = 0.949, indicating that digital technology compensated for the scarcity of geographic capital through a two-factor synergy effect with population agglomeration.

Third, spatially, the center of light intensity stabilized after 2015, while the LETI decreased from 0.162 to 0.050, showing a significant negative correlation with the D-value. The expansion of construction land was concentrated on slopes with gradients of 0–6° and in areas at elevations of 1000–1400 m, overlapping with high-quality farmland, whereas steep slopes and high-altitude regions were strictly protected. The synchronization rate between economic development and land expansion was approximately 50%, indicating a reduced reliance of economic activities on land expansion.

These findings highlight the policy need to promote the alignment of digital infrastructure investment with the population distribution of mid-altitude communities, while shifting farmland compensation standards from area equivalence to functional equivalence.