Spatial Gradient Effects of Landscape Pattern on Ecological Quality Along the Grand Canal

Abstract

The Grand Canal serves as a vital water transportation route, a UNESCO World Cultural Heritage site, and an ecological corridor. It is currently undergoing coordinated transformation through infrastructure development, heritage preservation, and ecological restoration. However, existing research has primarily focused on either cultural heritage conservation or localized ecological issues, with limited attention to the spatial relationship between landscape patterns and ecological quality along the entire corridor. To address this gap, this study examines eight sections of the Grand Canal and develops a gradient analysis framework based on equidistant buffer zones. The framework integrates the Remote Sensing Ecological Index (RSEI) with landscape pattern indices to assess ecological responses across spatial gradients. A Multi-scale Geographically Weighted Regression (MGWR) model is applied to reveal the spatially heterogeneous effects of landscape patterns on ecological quality. From 2013 to 2023, landscape patterns showed a trend toward increasing agglomeration and regularity. This is indicated by a rise in the Aggregation Index (AI) from 91.24 to 91.38 and declines in both patch density (PD) from 8.45 to 8.20 and Landscape Shape Index (LSI) from 199.74 to 196.72. During the same period, ecological quality slightly declined, with RSEI decreasing from 0.66 to 0.57. The effects of PD and Shannon’s Diversity Index (SHDI) on ecological quality varied across canal sections. In highly urbanized areas such as the Tonghui River, these indices were positively correlated with ecological quality, whereas in less urbanized areas like the Huitong River, negative correlations were observed. Overall, the strength of these correlations tended to weaken with increasing buffer distance. This study provides a scientific foundation for the integrated development of ecological protection and spatial planning along the Grand Canal and offers theoretical insights for the refined management of other major inland waterways.

1. Introduction

The Grand Canal of China, the world’s longest and largest artificial waterway, has served both as a critical transportation route and a cornerstone of Chinese cultural heritage for over 2500 years [1,2]. In 2014, it was inscribed on the World Heritage List, affirming its globally recognized cultural significance [3]. In recent decades, ecosystems along the canal have received increasing attention due to their marked spatial heterogeneity and heightened sensitivity to anthropogenic disturbances. These ecosystems are characterized by low ecological resilience and limited restoration capacity, which distinguish them from typical linear ecological corridors [4]. Amid global climate change, rapid urbanization, and the pursuit of ecological civilization, the Grand Canal is undergoing a functional transformation from a traditional navigation route to a multifunctional spatial system that integrates cultural preservation, ecological rehabilitation, and regional development [5]. Accordingly, a key challenge is how to systematically assess and improve ecological quality while maintaining the canal’s cultural integrity [6,7].

In recent years, several ecological restoration initiatives have been launched along the Grand Canal [5,8,9], leading to initial improvements in ecological function and landscape structure in certain sections. Nevertheless, the canal still faces pressing issues, including intensified landscape fragmentation, increasing risks of ecological degradation, and an uneven spatial distribution of water resources [10,11]. More fundamentally, despite advances in heritage preservation, ecological rehabilitation, and urban planning, the spatial interplay between landscape pattern evolution and ecological quality remains insufficiently understood [12,13]. As a linear system with diverse functional demands, the Grand Canal requires a more integrated analytical approach that captures the relationship between landscape structure and ecosystem performance.

Landscape patterns, typically quantified using landscape pattern indices, describe the spatial arrangement and structural features of land-use types. These indices reflect the composition, configuration, and inter-relationships of landscape patches and are widely used in regional assessment and ecosystem restoration research [14,15]. In linear geographic units such as rivers and transportation corridors, they play a key role in revealing connections between land-use changes and ecological processes, including water quality, biodiversity, and ecosystem services [16,17,18]. For example, Han et al. identified spatial heterogeneity in ecological responses along the Qinhuai River using water quality metrics and landscape pattern indices [19]. Yang et al. examined the spatiotemporal evolution of ecosystem services along the Qinghai–Tibet Highway under the influence of climate factors [20]. Mohammadi et al. applied a multi-buffer method to explore spatial couplings between land-use patterns and pollution in the Talar watershed [21]. However, the Grand Canal presents additional complexities. As a highly artificial water transport system, it features frequent alterations between water bodies, green spaces, and urban construction areas, creating sharp landscape heterogeneity along its banks [22]. To effectively analyze the ecological responses of such fragmented spatial structures, a tailored, multi-scale analytical framework is required.

Traditional methods of ecological quality assessment, such as ecosystem service valuation [23,24], ecological risk indices [25], and carrying capacity evaluations [26,27] often rely on expert-defined weights. This subjectivity limits their objectivity and consistency, particularly in large-scale applications. In contrast, the Remote Sensing Ecological Index (RSEI), which integrates multi-source data using principal component analysis, offers a more objective, scalable, and computationally efficient approach [28]. It has been successfully applied in a variety of ecosystems [29,30,31] and, more recently, extended to linear geographic features. For instance, Zou et al. and Wang et al. applied RSEI to assess the ecological condition of the Qinghai–Tibet Railway and high-altitude highways [32,33], while Li et al. used it for the Grand Canal [4]. Despite its growing application, most studies focus on dynamic monitoring of ecological quality and lack systematic analysis of how landscape pattern transformations affect ecosystem conditions. Few have addressed the impacts of spatial changes, such as connectivity loss, habitat fragmentation, or intensified edge effects, on ecological succession [34,35]. Therefore, a comprehensive assessment framework is needed, one that integrates landscape structure indicators with ecological response mechanisms across multiple spatial scales.

Spatial heterogeneity is a defining feature of the relationship between landscape patterns and ecosystem dynamics. It provides a critical foundation for understanding localized ecological responses and for developing targeted management strategies [10]. Geographically Weighted Regression (GWR) is a commonly used spatial analysis method for identifying such local variations [36]. However, traditional GWR assumes a uniform spatial scale of all variables, making it less suitable for analyzing complex ecological systems with drivers operating at differing spatial extents [37]. The Multi-scale Geographically Weighted Regression (MGWR) model addresses this limitation by introducing variable-specific bandwidths, allowing for more precise detection of spatial non-stationarity [38]. MGWR has proven effective in studies related to urban heat islands, land-use transition, and ecosystem service distribution [39,40]. Nevertheless, MGWR does not fully account for distance sensitivity, which is essential in corridor studies. Therefore, combining MGWR with buffer-based gradient analysis allows for simultaneous integration of spatial location and ecological processes, helping to reveal how landscape patterns influence ecological quality at various distances.

In this context, this study takes the Grand Canal as a representative case and develops an integrated framework to examine the interactions among landscape patterns, ecological quality, and spatial responses from 2013 to 2023. The specific objectives are as follows: (1) To identify the spatial distribution and gradient evolution of landscape patterns using a buffer-based method; (2) to construct RSEI from multi-temporal remote sensing data to evaluate changes in ecological quality; (3) to apply the MGWR model to quantify spatial heterogeneity in the effects of landscape patterns on ecological quality; and (4) to assess spatial differentiation and effect intensity across buffer zones through gradient analysis. The results aim to clarify the spatiotemporal dynamics of ecological quality along the Grand Canal and provide scientific support for integrated ecological management and spatial planning in similar inland waterways.

2. Study Area and Data Sources

2.1. Study Area

The Grand Canal of China, extending nearly 3200 km, is the world’s longest and oldest artificial waterway [16]. It comprises three main sections: the Beijing–Hangzhou Grand Canal, the Sui–Tang Grand Canal, and the Zhedong Canal. Due to historical and hydrological changes, much of the Sui–Tang section has dried up, making full restoration impractical. Therefore, this study focuses on the Beijing–Hangzhou and Zhedong segments. As of 2024, the Beijing–Hangzhou Canal has been completely rewatered and partially reopened for navigation, although certain sections remain restricted by local hydrological conditions [41]. The Zhedong Canal maintains overall navigability and relatively stable water levels, though some localized improvements are still needed. Based on differences in historical development and geographical location, the study area includes eight representative sections arranged from north to south: Tonghui River, North Canal, South Canal, Huitong River, Zhong Canal, Huaiyang Canal, Jiangnan Canal, and Zhedong Canal [7]. These sections span approximately 1947 km (Figure 1).

2.2. Data Sources and Pre-Processing

Canal and river data were obtained from the National Catalogue Service for Geographic Information and processed according to the General Plan for the Conservation and Management of the Grand Canal (2012–2030). Following the approach of Yang et al. [18], a series of equidistant buffer zones was created using the canal centerline as the baseline. These zones extend 5 km on both sides at 500 m intervals to capture spatial gradients in landscape pattern and ecological quality. Land-use data were obtained from the China Land Cover Dataset (CLCD), which has an overall classification accuracy of 79.31% [42]. Landsat 8 OLI imagery was obtained from Google Earth Engine (GEE), selecting images from May to October to correspond with peak vegetation growth and minimize seasonal variation [43]. All spatial data were pre-processed using ArcGIS Pro 3.0, including projection transformation and spatial alignment. A uniform resolution of 30 m was applied under the WGS_1984_UTM_Zone_50N coordinate system. Details of the data are provided in

Table 1. Data sources.

Data	Resolution	Year	Data Source
Land-use data	30 m	2013–2023	https://doi.org/10.5281/zenodo.12779975 (accessed on 12 February 2025)
Landsat 8 OLI	30 m	2013–2023	https://www.usgs.gov/landsat (accessed on 13 February 2025)
Canals and rivers	−	2021	https://www.webmap.cn (accessed on 13 February 2025)

.

3. Methods

This study develops an integrated evaluation framework to examine the spatial interaction between landscape patterns and ecological quality along the Grand Canal (Figure 2). It first evaluates six key aspects of landscape configuration, including fragmentation, boundary complexity, shape irregularity, aggregation, connectivity, and diversity. Ecological quality is then quantified using the RSEI. A buffer-based gradient analysis is applied to identify landscape indices that are sensitive to distance from the canal and to reveal corresponding spatial response patterns. Finally, the MGWR model is employed to capture spatial heterogeneity and quantify the varying effects of landscape patterns on ecological quality across buffer gradients.

3.1. Landscape Pattern Metrics

To comprehensively characterize land-use structure along the Grand Canal, six representative landscape pattern indices were selected based on three core dimensions commonly used in landscape ecology: patch characteristics, spatial configuration and connectivity, and landscape composition and diversity [18,44]. Patch density (PD) and Edge Density (ED) measure fragmentation and boundary complexity. Landscape Shape Index (LSI), Aggregation Index (AI), and Patch Cohesion Index (COHESION) capture spatial structure and connectivity. Shannon’s Diversity Index (SHDI) represents the richness and evenness of land-use types. Collectively, these indices provide a comprehensive description of landscape patterns relevant to ecological processes and environmental quality [14,15]. All metrics were calculated at the landscape level using Fragstats 4.2, with detailed definitions presented in

Table 2. Landscape pattern indices.

Index	Description
Patch Density (PD)	Quantifies landscape fragmentation; a higher number of patches indicates greater fragmentation.
Edge Density (ED)	Captures boundary complexity; longer total edge length suggests more intricate patch boundaries.
Landscape Shape Index (LSI)	Measures patch shape irregularity, normalized to remove area influence.
Aggregation Index (AI)	Indicates the degree of clustering among patches of the same type; higher values imply stronger aggregation.
Patch Cohesion Index (COHESION)	Reflects spatial connectivity of similar patches; higher values denote more continuous landscapes.
Shannon’s Diversity Index (SHDI)	Represents landscape diversity in terms of richness and evenness of patch types, indicating overall heterogeneity.

.

3.2. Ecological Quality Assessment

The RSEI integrates four components: the Normalized Difference Vegetation Index (NDVI) for greenness, Land Surface Temperature (LST) for heat, Wetness Index (WET) for humidity, and the Normalized Difference Built-up and Soil Index (NDBSI) for dryness. These indices are combined using principal component analysis (PCA), enabling a comprehensive evaluation of surface vegetation, temperature, moisture, and aridity [45,46]. To reduce the influence of large water bodies on the calculation, the Modified Normalized Difference Water Index (MNDWI) was used to mask and exclude these areas before further processing [47,48].

Table 3. Formula of the four ecological indices.

Index	Formula
NDVI	$NDVI=({\rho }_{NIR}-{\rho }_{Red})/({\rho }_{NIR}+{\rho }_{Red})$
LST	$LST=0.02{LST}_0-273.15$
WET	$WET=0.1147{\rho }_{Red}+0.2489{\rho }_{NIR1}+0.2408{\rho }_{Blue}+0.3132{\rho }_{Green}-0.3122{\rho }_{NIR2}-0.6416{\rho }_{SWIR1}-0.5087{\rho }_{SWIR2}$
NDBSI	$NDBSI=(SI+IBI)/2$
	$SI=\left[\left({\rho }_{SWIR1}+{\rho }_{Red}\right)-\left({\rho }_{Blue}+{\rho }_{NIR}\right)\right]/\left[\left({\rho }_{SWIR1}+{\rho }_{Red}\right)+\left({\rho }_{Blue}+{\rho }_{NIR}\right)\right]$
	$IBI={\displaystyle \frac{2{\rho }_{SWIR1}/({\rho }_{SWIR1}+{\rho }_{NIR})-[{\rho }_{NIR}/({\rho }_{NIR}+{\rho }_{Red})+{\rho }_{Green}/({\rho }_{Green}+{\rho }_{SWIR1}\left)\right]}{2{\rho }_{SWIR1}/({\rho }_{SWIR1}+{\rho }_{NIR})+[{\rho }_{NIR}/({\rho }_{NIR}+{\rho }_{Red})+{\rho }_{Green}/({\rho }_{Green}+{\rho }_{SWIR1}\left)\right]}}$

Note: In the formulas, ${\rho }_i$ represents the surface reflectance of the corresponding band, and ${LST}_0$ refers to the original land surface temperature from the MOD11A2 product.

summarizes the formulas used for NDVI, LST, WET, and NDBSI [29].

PCA was applied to the four normalized indices (NDVI, WET, LST, and NDBSI) to compute the initial RSEI (${RSEI}_0$), with the first principal component (PC1) selected as the representative value [48,49]:

$${RSEI}_0=PC1\left[f\left(NDVI,WET,LST,NDBSI\right)\right]$$

(1)

The final RSEI was obtained through min–max normalization:

$$RSEI=({RSEI}_0-{RSEI}_{0\_min})/({RSEI}_{0\_max}-{RSEI}_{0\_min})$$

(2)

Based on prior studies [8,10], normalized RSEI values are classified into five ecological quality levels: excellent (0.8–1.0), good (0.6–0.8), moderate (0.4–0.6), fair (0.2–0.4), and poor (0.0–0.2).

3.3. Spatial Correlation Between Landscape Patterns and Ecological Quality

To evaluate the spatial relationship between landscape patterns and ecological quality, global spatial autocorrelation was first analyzed using Moran’s I, which measures the degree of clustering among spatial units. The formula for Moran’s I is as follows:

$$I={\displaystyle \frac{n}{{\sum }_{i=1}^n{\sum }_{j=1}^n{w}_{ij}}}·{\displaystyle \frac{{\sum }_{i=1}^n{\sum }_{j=1}^n{w}_{ij}\left(x_i-\overline{x}\right)\left(x_j-\overline{x}\right)}{{\sum }_{i=1}^n{\left(x_i-\overline{x}\right)}^2}}$$

(3)

where n is the number of spatial units, $x_i$ and $x_j$ are the ecological quality values of units i and j, $\overline{x}$ is the global mean, and $w_{ij}$ is the spatial weight. A positive value (I > 0) indicates clustering of similar values, a negative value (I < 0) indicates dispersion, and a value close to zero suggests a random distribution.

Following this, Ordinary Least Squares (OLS) regression was conducted to investigate the relationships between ecological quality and multiple landscape pattern indices. To mitigate multicollinearity and ensure the statistical validity of the model, variables with a Variance Inflation Factor (VIF) greater than 7.5 were excluded [38].

Considering the spatial resolution of remote sensing data, the heterogeneity of landscape patterns, and the perception scale of ecological processes, a 600 m regular grid was adopted as the spatial analysis unit [16]. The MGWR model was then constructed using MGWR 2.2 and is defined as follows:

$$Y_i={\beta }_0\left(u_i,v_i\right)+\sum _{j=1}^n{\beta }_j\left(u_i,v_i\right)X_{ij}+{ϵ}_i$$

(4)

where $Y_i$ is the ecological quality of grid unit i, $X_{ij}$ is the value of explanatory variable j in unit i, ${\beta }_j\left(u_i,v_i\right)$ is the spatially varying coefficient of variable j at location $\left(u_i,v_i\right)$, ${\beta }_0\left(u_i,v_i\right)$ is the spatially varying intercept, and ${ϵ}_i$ is the error term. The spatial scale of each explanatory variable is determined by its optimal bandwidth.

Finally, MGWR results were extracted based on equidistant buffer zones along the Grand Canal to assess the magnitude, direction, and spatial distribution of the effects of landscape pattern indices across buffer gradients.

4. Results

4.1. Trends in Landscape Pattern and Ecological Quality

4.1.1. Trends in Landscape Pattern Changes

To reveal temporal trends in landscape pattern evolution across canal sections, this study conducted a systematic analysis using six landscape pattern indices (Figure 3). At the whole-canal scale, the AI increased slightly from 91.24 to 91.38, indicating a general trend toward more clustered land-use. COHESION remained stable at 99.67, reflecting little change in structural connectivity. Meanwhile, ED decreased from 57.38 to 56.47 and PD declined from 8.45 to 8.20, reflecting a reduction in the number of patches and a corresponding increase in average patch size. The LSI fell from 199.74 to 196.72, suggesting a trend toward more regular patch configurations. SHDI increased slightly from 0.947 to 0.950 before declining to 0.945, implying a minor reduction in landscape type diversity.

Notable variations in landscape pattern changes were observed across different canal sections. In the North Canal, AI declined from 91.98 to 91.46, while ED increased from 52.99 to 56.47, and LSI rose from 49.06 to 52.13, indicating more fragmented patterns with increased patch complexity. SHDI also increased, suggesting slightly improved landscape diversity. In contrast, the Jiangnan Canal and the Tonghui River both showed increases in AI, rising from 88.89 to 89.79 and from 94.56 to 94.80, respectively. ED decreased from 73.29 to 67.29 in the Jiangnan Canal and from 36.16 to 34.50 in the Tonghui River, indicating more aggregated patches with simpler boundaries. However, SHDI declined in both sections, decreasing from 0.98 to 0.94 in the Jiangnan Canal and from 0.68 to 0.66 in the Tonghui River, indicating a reduction in landscape heterogeneity. In the Huaiyang Canal, both AI and SHDI increased slightly, reflecting a modest rise in patch aggregation and landscape richness. The South Canal experienced a slight decline in AI, from 93.62 to 93.50, while SHDI increased from 0.65 to 0.67, reflecting enhanced landscape type diversity despite reduced aggregation. The Zhedong Canal exhibited an increase in AI to 89.36 and a significant decrease in LSI, alongside a drop in SHDI, indicating a trend toward simplification of the landscape structure. In the Zhong Canal, AI decreased slightly, while SHDI rose from 0.88 to 0.89, pointing to a marginal improvement in landscape diversity.

4.1.2. Trends in Ecological Quality Changes

Before analyzing ecological quality changes, PCA was conducted on four remotely sensed ecological indicators from 2013 to 2023 to assess the robustness of the RSEI model. As shown in

Table 4. Principal component analysis results (PC1).

Year	Loading of NDVI	Loading of LST	Loading of WET	Loading of NDBSI	Contribution Ratio of PC1 (%)
2013	−0.605	−0.381	0.410	0.566	77.14
2014	−0.617	−0.386	0.380	0.571	78.79
2015	−0.603	−0.404	0.370	0.580	78.98
2016	−0.600	−0.431	0.353	0.573	80.78
2017	−0.600	−0.416	0.336	0.595	77.36
2018	−0.600	−0.426	0.357	0.576	74.50
2019	−0.597	−0.432	0.358	0.573	79.59
2020	−0.597	−0.393	0.401	0.574	80.11
2021	−0.582	−0.417	0.388	0.580	80.05
2022	−0.616	−0.395	0.352	0.583	77.70
2023	−0.599	−0.414	0.381	0.569	79.31

, the first principal component (PC1) consistently accounted for the largest proportion of total variance each year, averaging 78.57% and never falling below 70%, confirming its effectiveness in representing ecological quality dynamics.

To examine temporal trends in ecological quality, RSEI values from 2013 to 2023 were analyzed across eight canal sections and the whole-canal section (Figure 4 and Figure 5). Overall, the average RSEI declined from 0.66 to 0.57, indicating a general degradation in ecological quality. However, clear spatial disparities were observed among sections.

The Tonghui River exhibited the most substantial improvement, with RSEI increasing from 0.33 to 0.49 and showing a steady upward trend beginning in 2016. The Huaiyang Canal, Jiangnan Canal, and Zhedong Canal also showed moderate but consistent improvement. In contrast, the North Canal and South Canal followed a U-shaped trajectory, reaching their lowest ecological quality around 2016, followed by gradual recovery after 2018. More concerning were the trends observed in the Huitong River and Zhong Canal, where RSEI values declined sharply from 0.78 to 0.66 and from 0.82 to 0.57, respectively, indicating sustained ecological degradation. These trends may reflect varying levels of anthropogenic pressure, restoration initiatives, and land-use changes across different sections of the Grand Canal.

4.2. Landscape Pattern and Ecological Quality Responses Across Canal Buffer Gradients

According to Spearman’s correlation analysis (Figure 6), both landscape pattern indices and RSEI values exhibited significant spatial relationships with buffer distance from the canal. Among all indices, PD, SHDI, and RSEI showed the highest sensitivity to changes in distance. Across the entire study area, PD and RSEI were positively correlated with increasing distance from the canal, while COHESION showed a negative correlation. Other indices did not show statistically significant correlations at the overall level.

At the section level, the Tonghui River displayed the most pronounced gradient response: PD, ED, LSI, SHDI, and RSEI all increased with distance, whereas AI and COHESION declined. Similarly, RSEI showed positive correlations with distance in the South Canal, Huitong River, Huaiyang Canal, Jiangnan Canal, and Zhedong Canal. LSI also tended to increase with distance in both the Tonghui and Jiangnan sections. Conversely, COHESION decreased with distance in the Tonghui River, North Canal, and Jiangnan Canal, indicating reduced landscape connectivity farther from the canal. SHDI exhibited a negative correlation with distance in the South Canal and Huaiyang Canal, suggesting a potential decline in landscape heterogeneity in more peripheral zones, possibly due to the dominance of homogenous land-use types in these areas.

4.2.1. Landscape Pattern Index Responses Across Canal Buffer Gradients

To examine how landscape patterns vary across different buffer distances, this study focused on indices significantly correlated with canal distance and analyzed their spatial trends (Figure 7). Across the entire study area, COHESION and PD exhibited clearly contrasting trends. COHESION continuously declined beyond 2000 m, reaching its lowest value in the outermost buffers, suggesting a weakening of patch connectivity. In contrast, PD peaked around 3500 m, indicating increased landscape fragmentation at greater distances from the canal.

At the sectional level, COHESION showed a slight recovery following an initial decline in the Tonghui River, North Canal, Zhong Canal, and Jiangnan Canal, with minimum values observed at 4000 m, 3000 m, 2500 m, and 3000 m, respectively. In contrast, PD continued to rise in the Tonghui River and South Canal, while in the Jiangnan Canal, it peaked at 3500 m before declining slightly.

Further analysis revealed that ED, PD, LSI, and SHDI generally showed positive correlations with buffer distance in most sections, reflecting increasing fragmentation and landscape heterogeneity along the gradient. However, SHDI declined in the South Canal, Huaiyang Canal, and Zhedong Canal, indicating a reduction in landscape diversity with increasing distance in these regions. Notably, the South Canal exhibited the most pronounced increase in PD. The Jiangnan Canal, similar to the Tonghui River, exhibited a coordinated decline in AI and COHESION alongside increases in PD and LSI, reflecting a consistent spatial response pattern potentially driven by similar urban and ecological processes.

4.2.2. Ecological Quality Responses Across Canal Buffer Gradients

To further examine the spatial relationship between ecological quality and canal distance, this study analyzed variations in RSEI values across buffer zones along the study area (Figure 8). The results indicate a nonlinear trend: ecological quality generally increased with distance from the canal, peaked at mid-range buffers, and then declined. Specifically, the average RSEI across the full corridor peaked around 3500 m, suggesting that ecological conditions were most favorable in the middle buffer zones, possibly due to reduced development pressure and improved vegetation coverage.

Section-level analysis showed consistent yet nuanced trends. Both the Tonghui River and the South Canal followed the overall trend, with RSEI values peaking at 3500 m. The Huaiyang Canal and Zhedong Canal exhibited similar trends, with peaks occurring slightly farther out at 4000 m. In contrast, the Jiangnan Canal showed a delayed response, reaching its maximum ecological quality at 5000 m. Uniquely, the Huitong River showed a continuous increase in RSEI, peaking at 4500 m, suggesting a more sustained ecological benefit at greater distances from the canal.

4.3. Spatial Mechanisms of Landscape Pattern Effects on Ecological Quality

4.3.1. Spatial Effects of Landscape Pattern on Ecological Quality

To assess the spatial aggregation of ecological quality within each canal section, global spatial autocorrelation was measured using Moran’s I. As shown in

Table 5. Global spatial autocorrelation of ecological quality in canal sections.

Canal Sections	Moran’s I	Z-Score	p-Value
Tonghui River	0.756	42.803	0.000
North Canal	0.692	56.488	0.000
South Canal	0.762	105.278	0.000
Huitong River	0.696	116.113	0.000
Zhong Canal	0.750	95.144	0.000
Huaiyang Canal	0.726	70.513	0.000
Jiangnan Canal	0.732	71.388	0.000
Zhedong Canal	0.724	98.891	0.000

, all sections exhibited significant positive spatial autocorrelation, indicating clustered distributions of ecological quality. Z-scores exceeded the significance threshold substantially, and all p-values were 0.000, confirming the robustness of spatial clustering patterns.

Following OLS regression and VIF screening, PD and SHDI emerged as the most influential landscape indices, showing strong correlations with ecological quality and acceptable multicollinearity. These two variables were retained in the MGWR model. As shown in

Table 6. Comparison of OLS and MGWR models.

Canal Sections	OLS			MGWR
	AICc	R2	Adj.R2	AICc	R2	Adj.R2
Tonghui River	3899.508	0.389	0.388	2114.404	0.826	0.810
North Canal	9726.510	0.020	0.019	5491.568	0.780	0.750
South Canal	27,821.192	0.042	0.042	12,619.346	0.839	0.819
Huitong River	21,105.191	0.095	0.095	10,451.657	0.828	0.804
Zhong Canal	21,406.759	0.060	0.060	11,500.445	0.789	0.767
Huaiyang Canal	10,757.009	0.057	0.057	5263.762	0.834	0.808
Jiangnan Canal	26,015.883	0.037	0.037	13,281.209	0.817	0.790
Zhedong Canal	13,292.110	0.091	0.090	7308.020	0.799	0.771

, MGWR significantly improved model performance across all sections, with lower AICc values and greater explanatory power than the OLS model. This improvement highlights MGWR’s capacity to capture spatially non-stationary and local variations relationships between landscape patterns and ecological quality.

Regression coefficients derived from MGWR were further analyzed to interpret the spatial influence of landscape structure on ecological quality (

Table 7. MGWR regression results.

Canal Sections	Landscape Pattern Indices	Average Value	Minimum Value	Median Value	Maximum Value
Tonghui River	PD	0.222	−0.198	0.282	0.453
	SHDI	0.324	0.080	0.265	0.618
North Canal	PD	0.125	−1.003	0.132	1.392
	SHDI	−0.186	−0.789	−0.307	0.545
South Canal	PD	0.057	−0.051	0.047	0.177
	SHDI	−0.262	−0.958	−0.350	0.741
Huitong River	PD	−0.002	−0.707	0.029	0.503
	SHDI	−0.313	−0.985	−0.373	0.768
Zhong Canal	PD	0.071	0.068	0.071	0.075
	SHDI	−0.204	−1.022	−0.251	0.634
Huaiyang Canal	PD	−0.032	−1.065	−0.025	0.880
	SHDI	−0.105	−0.934	−0.150	1.087
Jiangnan Canal	PD	−0.102	−1.255	−0.069	0.867
	SHDI	−0.014	−1.048	0.026	0.797
Zhedong Canal	PD	−0.138	−0.497	−0.135	0.089
	SHDI	0.075	−1.150	0.076	1.166

). Notably, PD and SHDI exhibited marked heterogeneity across canal sections. In the Tonghui River and North Canal, average PD coefficients were positive, suggesting that a moderate level of fragmentation may enhance ecological quality, potentially by increasing patch-level diversity and microhabitat variability. Conversely, SHDI coefficients were mostly negative, implying that excessive diversity and irregularity could disrupt ecological continuity and reduce system stability. Coefficient ranges varied substantially, for example, PD in the North Canal ranged from −1.003 to 1.392, indicating strong spatial differentiation. Similarly, in the Huaiyang Canal and Zhedong Canal, SHDI coefficients fluctuated from negative to positive, further confirming the spatially non-stationary effects of landscape patterns on ecological quality.

Based on MGWR results, this study explored the spatial heterogeneity in the effects of PD and SHDI on RSEI across different canal sections (Figure 9). In the Tonghui River, South Canal, and Zhong Canal, PD generally showed a positive correlation with RSEI, although the strength of the association varied spatially. In the Tonghui River, the strongest positive effects of PD were concentrated within Beijing’s Third Ring Road, while SHDI had its greatest influence in Tongzhou District. These findings suggest that high-density green patches and moderate landscape diversity may jointly enhance ecological quality, possibly by reinforcing habitat continuity and buffering microclimatic extremes.

In the Zhong Canal, PD–RSEI correlations weakened downstream, and SHDI was largely negatively associated with RSEI, except in Suqian and Huai’an, where localized positive relationships were observed. This pattern indicates that landscape diversity shaped by urban spatial planning may enhance ecological quality under certain conditions. In the South Canal, PD–RSEI correlations intensified southward from Tianjin, peaking between Dezhou and Linqing. In the Jiangnan Canal, which spans highly urbanized zones, SHDI and RSEI were positively correlated in the Changzhou–Suzhou section, with PD also showing spatial continuity. This implies that even in highly developed areas, a well-structured landscape can sustain ecological quality.

Conversely, in the North Canal, Huaiyang Canal, and Huitong River sections, both PD and SHDI were negatively associated with RSEI, though the strength of these relationships varied across space. In the North Canal, the PD–RSEI relationship diverged sharply between urban and rural areas. SHDI exhibited negative correlations in most rural stretches, with exceptions in urban centers such as Beijing and Tianjin. In the Huaiyang Canal, positive effects were localized near Gaoyou Lake and the urban core of Yangzhou, while negative associations dominated the rest of the section. In the Huitong River, PD was negatively associated with RSEI across farmland, wetland, and forested areas. SHDI was positively correlated only in Liaocheng’s urbanized areas, and negatively correlated elsewhere, highlighting the region’s heightened sensitivity to landscape configuration in determining ecological quality.

4.3.2. Landscape Pattern Effects on Ecological Quality Across Buffer Gradients

To explore how landscape patterns influence ecological quality at varying distances from the canal, this study analyzed the correlation trends between PD and SHDI with RSEI across buffer gradients in each canal section (Figure 10). In the Tonghui River, North Canal, South Canal, and Zhong Canal, PD showed a consistently positive correlation with RSEI across the buffer distances, suggesting that increased patch density was generally associated with enhanced ecological quality. However, in the North Canal, the strength of this positive correlation declined notably beyond 1000 m, reaching a minimum at 4000 m before slightly recovering, a pattern indicative of spatial decay. In the Huitong River, PD was positively associated with RSEI within 2500 m but turned negative at greater distances, implying that the ecological benefits of patch density diminish with increasing distance from the canal. In contrast, PD remained negatively correlated with RSEI at all distances in the Huaiyang Canal, Jiangnan Canal, and Zhedong Canal, indicating that fragmentation in these areas may undermine ecological stability across the landscape.

The correlation between SHDI and RSEI was more variable and generally negative. A relatively stable negative relationship was observed in the Zhong Canal, North Canal, South Canal, Huaiyang Canal, and Huitong River, implying that excessive landscape heterogeneity may impair ecological integrity in these sections. In the Zhedong Canal, this negative correlation intensified with distance, reinforcing the notion that uncontrolled diversity can lead to functional fragmentation. In contrast, the Tonghui River exhibited a distinct pattern: both PD and SHDI were consistently and positively correlated with RSEI. This finding suggests that moderate fragmentation and diversity, when guided by effective planning and ecological infrastructure, may jointly promote ecological quality. Meanwhile, in the Huaiyang Canal, the persistent negative correlation between both PD and SHDI with RSEI indicates that high levels of fragmentation and heterogeneity may interactively contribute to ecological degradation.

5. Discussion

5.1. Mechanistic Analysis of the Evolution of Landscape Pattern and Ecological Quality

From 2013 to 2023, the landscape pattern along the Grand Canal exhibited a clear trend toward spatial centralization and structural simplification. Land-use types became more concentrated, and patch configurations tended to be more regular. Section-level analysis revealed notable fragmentation in the North Canal, marked by increased patch density and edge complexity, a trend likely driven by intensive land development activities [50]. In contrast, both the Tonghui River and Jiangnan Canal showed rising AI values and declining SHDI, suggesting that urbanized artificial landscapes tend to become more homogeneous under high-intensity development [51,52]. This pattern may reflect a planned distribution and accessibility of green spaces [53]. Conversely, the South Canal, Huaiyang Canal, and Zhong Canal exhibited increasing SHDI values, indicating enhanced landscape heterogeneity likely resulting from localized ecological restoration efforts [8]. These spatial differences appear to be closely linked to regional variations in land-use functions, water infrastructure conditions, and the stringency of ecological policies. Notably, a decline in fragmentation was observed in sections such as the Tonghui River and Jiangnan Canal, contradicting the conventional assumption that urbanization invariably leads to fragmentation [54,55]. This suggests that intensive anthropogenic intervention may instead promote landscape agglomeration and structural regularization. This pattern is consistent with Carlier et al.’s findings on infrastructure-driven landscape simplification [56], emphasizing the need to incorporate spatial logic and policy context when analyzing landscape evolution. In particular, landscape homogenization in well-managed urban cores may not signal ecological degradation but rather reflect a strategic reorganization of spatial structure to improve systemic efficiency and connectivity under planning constraints.

In terms of ecological quality, pronounced spatial heterogeneity was observed across sections (Figure 4). Its evolution was influenced not only by ecological restoration measures but also by the effectiveness of land-use regulation and policy coordination. In the Yangtze River Delta, the Huaiyang Canal, Jiangnan Canal, and Zhedong Canal achieved steady improvements in ecological quality, supported by the integration of wetland restoration, ecological corridor construction, and urban boundary controls. These coordinated strategies effectively reconciled development pressures with ecological goals. In contrast, ecological conditions in the Zhong Canal and Huitong River, located in Shandong and northern Jiangsu, deteriorated during the study period due to a combination of intensified shipping, urban expansion, and agricultural encroachment. The underlying challenges included delayed ecological compensation mechanisms and fragmented spatial governance, which undermined restoration efforts. While previous studies reported general improvements in ecological quality across canal-adjacent provinces and cities [57,58], our results highlight the need for section-specific analysis, as macro-level improvements may mask local ecological decline. This supports Jiang et al.’s recommendation for differentiated restoration strategies tailored to the specific spatial and developmental contexts of each section [59].

Further insights emerged from the buffer zone analysis, which highlighted the spatial gradients in landscape–ecosystem interactions. Indices such as PD, COHESION, and SHDI showed distinct responses across distance zones, with more urbanized sections exhibiting stronger gradient sensitivity (Figure 7). Near-canal areas typically experience tighter ecological controls and enforcement, while outer zones were more susceptible to land-use conversion and development pressures, leading to increased fragmentation and reduced connectivity [60]. For example, the consistent increase in fragmentation in the Tonghui River section reflects the strong influence of urban expansion on landscape structure, consistent with Yang et al.’s findings from the Yongding River, where fringe areas showed reduced structural complexity and increased aggregation, with a notable transition in landscape configuration occurring around 3500 m from the river [18].

Ecological quality similarly exhibited a nonlinear spatial pattern, peaking around 3500 m from the canal (Figure 8). This mid-range zone likely represents a transitional area where development pressures and ecological processes temporarily balance. The result echoes the “core–edge” theory, in which built-up areas dominate the canal edge, while ecological functionality is better preserved in peripheral zones [61]. However, peak locations varied by section. In the Tonghui River and South Canal, ecological quality peaked around 3500 m, indicating a dual-function zone where development and ecological function coexist. In contrast, the Jiangnan and Huaiyang Canals exhibited delayed peaks, possibly reflecting interactions between urban expansion and water network restoration [51]. The Huitong River showed a continuous increase in ecological quality with increasing distance, reflecting a relatively natural shoreline and limited anthropogenic disturbance. These patterns suggest that the relationship between landscape structure and ecological quality is shaped by spatial thresholds and functional trade-offs. Moderate fragmentation in well-planned urban cores may enhance edge effects and functional diversity, while excessive heterogeneity in poorly managed zones may disrupt ecological continuity and reduce system resilience.

5.2. Differential Responses of Ecological Quality to Landscape Pattern

Based on the MGWR results, this study identified spatially heterogeneous effects of PD and SHDI on ecological quality across different canal sections (Figure 9). Positive correlations were commonly observed in well-planned urban areas, where structured green spaces and moderate landscape diversity enhance ecological resilience. In contrast, negative correlations emerged in fragmented or poorly managed zones, where excessive heterogeneity undermined connectivity and ecological stability. These differences reflect variations in land-use types, restoration effectiveness, and policy enforcement, underscoring the context-dependent mechanisms through which landscape patterns affect ecological quality.

For instance, in the Tonghui River, the southern segment of the South Canal, and the urban core of the Suzhou–Wuxi–Changzhou metropolitan area, PD showed a significant positive association with RSEI, while SHDI also contributed positively to selected locations. These results suggest that deliberate spatial planning, such as the concentration of high-density green patches can enhance landscape connectivity and integrate ecological functions. In the Tonghui River, within Beijing’s Third Ring Road, long-standing implementation of green infrastructure plans and development boundaries, along with sustained ecological investment in Tongzhou District as Beijing’s sub-center, have collectively strengthened the relationship between moderate fragmentation and improved ecological quality [62,63]. Such efforts have also contributed to a more comfortable walking environment for residents [64]. These findings suggest that the ecological impacts of human activity depend not only on development intensity, but also on spatial configuration, planning continuity, and institutional coherence. In historically well-regulated cities, anthropogenic interventions may even promote ecological recovery by enhancing spatial structure and connectivity.

Conversely, in the North Canal, Huaiyang Canal, and Huitong River, PD and SHDI generally showed negative correlations with RSEI, accompanied by significant spatial variation. These areas, primarily located in rural–urban transition zones of the Beijing–Tianjin–Hebei region and southwestern Shandong, have historically featured dispersed township development and extensive agricultural land-use, but limited ecological intervention. The Huitong River, which traverses multiple prefecture-level jurisdictions, lacks a unified ecological management framework, resulting in inconsistent policy enforcement and uneven construction standards. Consequently, ecological spaces have been fragmented, and landscape structures fail to achieve sustainable configurations, limiting the ability of landscape diversity to support ecological quality [65].

Interestingly, localized shifts from negative to positive correlations between PD or SHDI and RSEI were observed in urban sections of the Zhong Canal (Suqian–Huai’an) and Jiangnan Canal (Changzhou–Suzhou). This trend may be attributed to increasing spatial complexity and the reorganization of ecological function zones within cities. In recent years, under national spatial planning initiatives, cities have promoted the development of green wedges and waterfront ecological corridors, facilitating structural and functional integration of previously fragmented green spaces. These efforts have contributed to improvements in ecological quality. This contrasts with the findings by Aurora et al., who reported that urban expansion in Chiba Prefecture, Japan (1990–2021), generally led to ecological degradation [66]. In contrast, this study observed synchronous trends of urban growth and ecological improvement in certain Grand Canal sections, consistent with Yang et al.’s findings on internal spatial optimization in Kunming, Nanjing, and Guangzhou, where urban core restructuring supported ecological recovery from 1990 to 2020 [35]. These examples highlight the importance of considering multiple interacting landscape spaces in an integrated manner to optimize urban ecological performance [67].

Analysis of buffer gradients further revealed that the influence of patch density on ecological quality tends to weaken with increasing distance from the canal. In most sections, PD is positively correlated with RSEI in nearshore zones but declined in more distant areas (Figure 10). This pattern likely reflects stronger ecological governance, green investment, and spatial planning controls near the canal. In contrast, foreshore zones often face regulatory gaps and intensified land-use conversion, where abrupt transitions from forest to construction land reduce ecological connectivity and resilience. In these zones, SHDI frequently showed negative correlations with RSEI, likely due to uncoordinated landscape diversity leading to function fragmentation. This scenario reflects an ecological risk profile characterized by high heterogeneity but low structural stability [14].

These spatial patterns must be interpreted within the broader historical and institutional context of the Grand Canal. As a major hydraulic and transportation corridor in ancient China, the canal profoundly shaped regional land-use and ecological processes. Historically, the construction and maintenance of the Grand Canal transformed natural hydrological systems, forming a linear landscape structure centered around waterways, docks, and settlement clusters [68]. These human interventions supported the expansion of irrigation networks, market towns, and transit hubs, driving long-term land-use changes [69]. Traditional uses such as irrigation and transport altered hydrological connectivity and aquatic ecosystems, while industrialization and modern urbanization introduced new pressures including water pollution and flow disruption [13]. These cumulative impacts fragmented ecosystems and impaired ecological function. In recent years, however, efforts in heritage conservation and ecological restoration have reshaped local landscape patterns and improved connectivity. This dynamic reflects a disturbance–response cycle, in which human activity affects hydrological flow, land-use intensity, and biodiversity, resulting in complex, nonlinear, and multi-scale feedback between landscape structure and ecological quality along the canal corridor [70].

5.3. Ecological Planning Strategies and Policy Implications

In recent years, policies promoting ecological protection and landscape construction along the Grand Canal have steadily advanced, gradually forming a development model that integrates cultural heritage preservation, ecological restoration, and functional coordination. The Outline of the Plan for the Protection, Inheritance, and Utilization of the Culture of the Grand Canal and the Plan for the Construction and Protection of the Grand Canal National Cultural Park have defined national-level implementation pathways for integrating culture and ecological values [71]. Provincial and municipal governments, including Jiangsu Province, Beijing, and Hangzhou, have actively implemented ecological restoration and spatial remediation efforts, offering valuable practical experience in landscape optimization and ecological reconstruction [72]. However, current policy frameworks still fall short in addressing the spatial coupling between landscape pattern evolution and ecological quality changes. First, special purpose planning often lacks alignment with the national territorial spatial planning system. Ecological governance frequently depends on fragmented, localized interventions that fail to incorporate spatial structure analysis. Second, many implementation efforts prioritize visible restoration outcomes, while neglecting the foundational role of spatial configuration in shaping ecological conditions. As a result, the relationships among ecological processes, spatial structure, and functional outcomes are often poorly captured. Third, canal navigation continues in many sections without effective coordination mechanisms to balance ecological protection with transport infrastructure. In certain sections, navigation facilities and heavy traffic still exert significant pressure on ecological spaces.

To address these challenges, spatial zoning strategies for regulating ecological quality should be refined based on landscape patterns’ characteristics. On the one hand, ecological control units should be delineated, targeting areas with high fragmentation and low connectivity for prioritized regulation and resource allocation [73]. Restoration-prone zones should be incorporated into these units to guide targeted ecological enhancements. Zoning-specific and hierarchical regulation objectives should then be established, including measures such as restricting high-intensity development, improving ecological nodes through greening, and enhancing patch connectivity. On the other hand, a multi-scale spatial coordination mechanism should be developed across canal sections, buffer zones, and subregional units. Findings from buffer gradient analyses should be integrated into restoration plans and land-use boundary frameworks, thereby clarifying the spatial logic underlying coordination between ecological redlines and development activities [74]. This would enhance the precision and operational feasibility of ecological governance. International examples, such as the Rideau Canal in Canada and the Canal du Midi in France, demonstrate the benefits of functional zoning for ecological heritage management. These systems achieved synergistic integration of cultural preservation and ecological protection through refined landscape unit management and targeted interventions at the land–water interface [75,76]. In parallel, ecological impact factors should be systematically identified, and a dynamic evaluation index system should be developed to support zoning assessments, prioritize interventions, and guide integrated management. Such strategies will facilitate the scientific and adaptive advancement of ecological spatial governance.

5.4. Limitations and Future Research

Although this study offers meaningful insights into the landscape patterns and ecological quality along the Grand Canal, several limitations remain. First, the accuracy of available spatial data constrains the depth of analysis over large-scale regions. The current resolution of remote sensing data limits the ability to precisely detect fine-scale landscape units and complex ecosystem structures. Future research should incorporate higher-resolution imagery to enhance data accuracy and enable more detailed assessments of landscape and ecological changes. Second, the RSEI has intrinsic limitations when applied to ecological assessments at large watershed scales. Future studies should explore enhanced indices, such as an updated Water Body Ecological Quality Index (WQI), to provide a more comprehensive evaluation of ecological conditions. Third, in the buffer zone analysis, this study relied on multi-year mean values of landscape and ecological indices due to spatial and temporal constraints. This limits the ability to capture temporal dynamics and may obscure short-term or seasonal ecological changes. Future research should adopt year-by-year comparative analyses and integrate interannual trend evaluations to more accurately reflect ecosystem responses over time and under different development or restoration scenarios.

6. Conclusions

This study systematically examined the evolution of landscape patterns and their spatial impact on ecological quality along the Grand Canal from 2013 to 2023. The results revealed a general trend toward increased landscape aggregation and structural regularization, as indicated by an increase in the AI from 91.24 to 91.38 and decreases in PD and LSI from 8.45 to 8.20 and from 199.74 to 196.72, respectively. These changes reflect a shift toward more centralized land-use and more regular patch configurations. However, distinct spatial differences emerged among canal sections. The North Canal exhibited increased fragmentation, whereas landscape diversity declined significantly along the Tonghui River and Jiangnan Canal, highlighting the heterogeneous nature of landscape transformation across regions. In terms of ecological quality, the overall RSEI slightly declined from 0.66 to 0.57. Nevertheless, several river sections, including the Tonghui River, Huaiyang Canal, Jiangnan Canal, and Zhedong Canal, showed continuous improvement. In contrast, the North and South Canals followed a U-shaped trajectory, reflecting initial degradation followed by gradual recovery. Buffer gradient analysis further demonstrated spatial variations in ecological responses to landscape structure. Both PD and RSEI tended to increase with distance from the canal, while COHESION decreased, with all indicators peaking around 3500 m. This suggests that mid-distance zones represent key interfaces where landscape disturbance and ecological processes intersect. Moreover, the effects of PD and SHDI on ecological quality varied significantly across canal sections. In urbanized areas such as the Tonghui River and South Canal, patch density and landscape diversity jointly contributed to ecological enhancement. Conversely, the North Canal, Huaiyang Canal, and Huitong River exhibited predominantly negative correlations, implying that unregulated development and excessive heterogeneity may compromise ecological function. Overall, this study reveals the nonlinear and spatially heterogeneous relationships between landscape pattern changes and ecological quality. These insights provide a scientific foundation for the conservation and sustainable development of the Grand Canal and offer theoretical guidance for the ecological management of other major inland waterways.