Comprehensive Monitoring of Construction Spoil Disposal Areas in High-Speed Railways Utilizing Integrated 3S Techniques

Abstract

High-speed railways are critical infrastructure in many countries, but their construction generates substantial spoil, particularly in mountainous regions dominated by tunnels and slopes, necessitating the establishment and monitoring of spoil disposal areas. Inadequate monitoring of spoil disposal areas can lead to significant environmental issues, including soil erosion and geological hazards such as landslides and debris flows, while also hindering the recycling and reuse of construction spoil, thereby impeding the achievement of circular economy and sustainable development goals for high-speed railways. Although the potential of geographic information systems, remote sensing, and global positioning systems in waste monitoring is increasingly recognized, there remains a critical research gap in their application to spoil disposal areas monitoring within high-speed railway projects. This study proposes an innovative framework integrating geographic information systems, remote sensing, and global positioning systems for monitoring spoil disposal areas during high-speed railway construction across three key scenarios: identification of disturbance boundaries (scenario 1), extraction of soil and water conservation measures (scenario 2), and estimation of spoil volume changes (scenario 3). In scenario 1, disturbance boundaries were identified using Gaofen-1 satellite data through processes such as imagery fusion, unsupervised classification, and spatial analysis. In scenario 2, unmanned aerial vehicle data were employed to extract soil and water conservation measures via visual interpretation and overlay analysis. In scenario 3, Sentinel-1 data were used to analyze elevation changes through the differential interferometric synthetic aperture radar method, followed by the estimation of spoil volume changes. The effectiveness of this integrated framework was validated through a case study. The results demonstrate that the framework can accurately delineate disturbance boundaries, efficiently extract soil and water conservation measures, and estimate dynamic changes in spoil volume with an acceptable error margin (15.5%). These findings highlight the framework’s capability to enhance monitoring accuracy and efficiency. By integrating multi-source data, this framework provides robust support for sustainable resource management, reduces the environmental impact, and advances circular economy practices. This study contributes to the efficient utilization of construction spoil and the sustainable development of high-speed railway projects.

1. Introduction

Currently, countries around the world generally regard sustainable development as a strategic goal, aiming to coordinate resource utilization, environmental protection, and climate change mitigation [1]. The rapid expansion of high-speed railways (HSRs), defined by speeds exceeding 200 km/h, has become a key infrastructure development project in many countries [2]. However, the construction of HSRs inevitably generates large amounts of waste, and the accumulation of this waste not only occupies significant land resources but also poses potential pollution and environmental damage [3], presenting a severe challenge to the sustainable development of HSRs.

The circular economy, as an emerging paradigm for promoting sustainable development, has gained widespread attention in recent years. The core concept of the circular economy is to create an “industrial system designed and systematically restored or regenerated” based on the principles of “reduce, reuse, and recycle” [4], aiming to minimize resource waste and environmental burdens [5]. In HSRs, the application of circular economy can improve resource utilization efficiency and reduce the environmental impact by procuring sustainable materials, maximizing material recycling, and minimizing waste generation. Additionally, the circular economy emphasizes avoiding the generation of unnecessary waste and preventing waste from being disposed of in landfills.

However, the overall utilization rate of construction spoil in HSRs is still low [6,7], which necessitates the establishment of large-scale spoil disposal areas (SDAs) to store the large amounts of spoil generated, thereby supporting material recycling and reuse. Meanwhile, the large volumes of spoil stored in SDAs are typically loose and fragmented, and long-term accumulation may lead to soil erosion, water loss, and further environmental degradation, including potential landslides and debris flows. Therefore, effective monitoring and management of SDAs are crucial for mitigating these environmental impacts and achieving the goals of the circular economy.

For waste monitoring, the importance of integrating 3S techniques, including geographic information systems (GIS), remote sensing (RS), and global positioning systems (GPS), is increasingly recognized. Researchers have explored their potential in monitoring areas such as olive oil mill SDAs [8], municipal open SDAs [9], mine SDAs [10], and hazardous SDAs [11]. Existing studies typically use individual 3S techniques or combinations of different techniques for various monitoring purposes, including settlement monitoring, deformation monitoring, temperature monitoring, and concentration monitoring.

However, there is currently a lack of research integrating 3S techniques for monitoring SDAs in HSR projects. On the one hand, SDAs in HSRs have their own specific characteristics. These areas are primarily used for the large-scale storage of spoil generated by excavation activities during the construction stage. Environmental protection requirements are particularly high, especially when passing through ecologically sensitive regions. Close attention needs to be paid to the stability of the spoil and its impact on the soil and water environment to prevent landslides, debris flows, and soil erosion. Therefore, directly applying 3S techniques from other monitoring fields (such as municipal and mine SDAs) is not sufficient for effectively monitoring SDAs in HSRs. On the other hand, current applications of 3S techniques in SDA monitoring typically use individual techniques for monitoring aspects such as settlement, temperature, and concentration, with a lack of systematic research on the integration of 3S techniques for comprehensive monitoring.

To address the research gap mentioned above, this study proposes an innovative integrated 3S framework for the comprehensive monitoring of key scenarios in SDAs during the construction stage of HSRs. The framework specifically targets the monitoring of disturbance boundaries, soil and water conservation measures, and spoil volume changes. The effectiveness and applicability of this integrated approach are validated through a focused case study. This research expands the application of 3S techniques and provides practitioners with a powerful tool to effectively monitor and manage SDAs in HSR projects, thus promoting the circular economy and sustainable development goals of HSRs.

2. Literature Review

The circular economy aims to conserve limited resources, reduce waste, and overcome the limitations of the traditional linear economy [12]. This paradigm has garnered widespread global attention. The transition to a circular economy has become an inevitable trend worldwide to effectively utilize resources and promote material reuse. Although circular economy initiatives in waste management are still in their early stages, their scientific contributions are progressively increasing and expanding. Current research on circular economy in the field of construction waste primarily focuses on technical aspects while also addressing broader implementation barriers and strategies. On the technical front, materials such as wood–cement composite materials [13], self-compacting concrete [14], green and sustainable concrete [15], high-strength concrete [16], eco-hybrid cement-based building insulation materials [17], and renewable bricks [18,19] have demonstrated feasibility. Additionally, some studies have explored the use of neural network technology to replace traditional manual processes in aggregate recycling [20,21]. In the implementation process, numerous barriers exist, including insufficient government incentives, inadequate policy support, lack of public education and awareness, and a deficient legal framework [22,23]. Regarding implementation strategies, researchers have investigated key success factors for a closed-loop circular economy in construction and demolition waste management in China [24,25]. Strategies such as procurement innovation [26], building design and construction strategies [27], digital construction tools [28,29], policy support [30], and incentive mechanisms [31] have been proposed to promote the widespread application of circular economy principles in construction waste management.

However, most existing studies are primarily focused on the construction industry, with limited attention given to the specific applications of circular economy in HSR projects. In the context of HSR projects, construction spoil refers to soil, silt, rocks, sludge, and other solid waste excavated or displaced during the construction process. This type of spoil constitutes the largest proportion of construction waste, particularly in mountainous corridors dominated by tunnels and slope construction [32]. Although construction spoil holds potential for recycling and reuse, the overall utilization rate of spoil in HSRs remains relatively low [6]. Therefore, establishing SDAs is not only essential for storing the vast amounts of spoil generated during the construction stage in HSRs but also serves as a critical prerequisite for recycling and reuse.

Moreover, the accumulation of spoil in SDAs poses risks of soil erosion and potential geological disasters such as collapses, landslides, and debris flows. Effective monitoring and management of SDAs are crucial to mitigate these risks. Such measures are not only necessary for ensuring environmental safety but also represent a key step in the circular economy framework for enabling material recycling and reuse. By employing advanced monitoring techniques and methods, spoil management and utilization efficiency can be significantly improved, thereby providing vital support for the sustainable development of HSRs.

In terms of waste monitoring, 3S techniques have increasingly demonstrated their importance. Initially, the applications of GIS, GPS, and RS in waste monitoring were primarily independent. GIS was mainly used to monitor potential landfill sites [33,34] and to identify areas at risk of illegal waste dumping [35]. Adamcová et al. [36] used GPS monitoring data to accurately estimate the surface displacement of landfills over time. RS applications are more extensive, covering landfills, mining sites, wastewater treatment plants, and olive oil factories, with a focus on monitoring subsidence [37,38,39], deformation [40], temperature [41,42,43], and concentration levels [10,44]. In particular, the use of unmanned aerial vehicles (UAVs) for RS has become an important research direction [45,46]. Mello et al. [47] assessed the accuracy and application of UAV photogrammetry in measuring the geometry and volume of sanitary landfills, comparing it with traditional measurement methods to demonstrate the effectiveness of UAVs in routine landfill monitoring.

Emerging research highlights the integration of 3S techniques and the application of new techniques in waste monitoring [48,49,50,51,52]. Gautam et al. [53] proposed a ranking algorithm using RS and GIS to identify and categorize suitable, moderately suitable, or unsuitable landfill sites in India. Biluca et al. [54] addressed the lack of methods for selecting appropriate landfills or recycling plants for construction and demolition waste in urban areas by combining compensatory and non-compensatory multi-criteria analysis with GIS. Joshi et al. [55] implemented a wireless personal area network and cloud-assisted architecture using the Internet of Things and Xbee communication protocol for real-time monitoring of solid waste in remote areas. Xu et al. [56] emphasized the development of an optical fiber sensing method for large deformation measurement in construction solid waste landfills, improving risk assessment and mitigation capabilities through accurate field monitoring and effective early warning systems. Bošković et al. [57] proposed a method to optimize waste collection processes by leveraging digital technologies such as GIS, the Internet of Things, and data analytics, significantly reducing travel distance and time in waste collection and transportation.

In summary, the potential of 3S techniques is increasingly recognized in monitoring olive oil mill SDAs, municipal open SDAs, mine SDAs, and hazardous SDAs. However, despite the critical importance of SDAs in HSRs, existing research has largely overlooked their monitoring. Additionally, more research emphasizes the application of a single technique rather than the integration of techniques. Moreover, existing research mainly focuses on the monitoring of sedimentation, temperature, and concentration. Existing research is not applicable to the monitoring focus of disturbance boundaries, soil and water conservation measures, and spoil volume changes in SDAs. Therefore, it is necessary to integrate 3S techniques to carry out comprehensive monitoring of SDAs during the construction stage of HSRs in order to support the recycling and reuse of construction spoil and contribute to the development of a circular economy in HSRs.

3. Research Method

To address the comprehensive monitoring challenges of SDAs during the construction of HSRs in the context of a circular economy, this study proposes a research framework based on 3S techniques (Figure 1). GPS provides spatial positioning and reference information for the research data, serving as the foundation for buffer analysis, spatial overlay, and imagery geolocation in GIS. RS is primarily used for data collection, including data from the Gaofen (GF)-1 satellite, UAVs, and Sentinel-1 satellite. GIS are mainly used for processing RS data and developing models in conjunction with other methods. Essentially, georeferenced RS imagery integrates both RS and GPS techniques—RS provides the structural information of the imagery, while GPS supplies the positional data, supporting the structural framework. GIS then completes the series of processes for data processing, analysis, and decision-making based on the RS imagery. Through the integrated application of these 3S techniques, this study focuses on analyzing disturbance boundaries, soil and water conservation measures, and spoil volume changes within SDAs of HSRs to achieve comprehensive monitoring objectives. Additionally, a case study is presented to validate the application and effectiveness of these 3S techniques in the comprehensive monitoring of SDAs in HSRs.

3.1. Data Collection

This study uses RS to collect data, including data for GF-1, UAV, and Sentinel-1, to serve the three main monitoring application scenarios of SDAs in HSRs.

3.1.1. GF-1 Satellite Data

This study utilizes data from the GF-1 satellite to identify disturbance boundaries of SDAs in HSRs, which was obtained from the Department of Water Resources of Guangdong Province. GF-1, a pivotal satellite in China’s high-resolution Earth observation system, was launched on 26 April 2013. The satellite payload specifications are shown in

Table 1. GF-1 satellite payload specifications.

Sensor	Band	Wavelength Range	Resolution (m)
Panchromatic multispectral camera	1	0.45–0.90	2
	2	0.45–0.52	8
	3	0.52–0.59
	4	0.63–0.69
	5	0.77–0.89
Multispectral camera	6	0.45–0.52	16
	7	0.52–0.59
	8	0.63–0.69
	9	0.77–0.89

. The satellite is equipped with high-resolution panchromatic and multispectral cameras, along with various multispectral cameras offering different spatial resolutions, enabling ground observations up to 800 km in width. GF-1’s extensive ground coverage and high spatiotemporal resolution allow for the frequent acquisition of data over the same area within short intervals, making it particularly effective for monitoring dynamic environmental changes. The high-resolution imagery provided by GF-1 is crucial for capturing fine surface details, while its temporal resolution enables continuous tracking of surface disturbances. These attributes make GF-1 an optimal choice for accurately identifying and monitoring disturbance boundaries in SDAs, especially in complex terrain regions.

3.1.2. UAV Data

This study uses UAV to extract information on soil and water conservation measures in SDAs of HSRs. UAV has the characteristics of rapidity, efficiency, stability, and convenience, which can effectively make up for the shortcomings of low resolution and slow update of traditional RS imagery. Through on-site investigation of the SDAs in HSRs, the take-off and landing location of the UAV was determined, its flight route was formulated, and the UAV data was extracted.

The DJI Phantom 4 RTK UAV (SZ DJI Technology Co., Ltd., Shenzhen, Guangdong, China) was utilized for an aerial survey of a SDA. The UAV has the following performance characteristics: a maximum take-off altitude of 6000 m, a maximum ascent speed of 6 m/s, a maximum descent speed of 3 m/s, and a maximum horizontal flight speed of 58 km/h. The hovering accuracy is ±0.1 m vertically and ±0.1 m horizontally. Equipped with a 20-megapixel imaging sensor, it offers a maximum photo resolution of 5472 × 3648 and a 4K video resolution of 3840 × 2160. The UAV meets the specifications of China’s GB/T 7930-2008 standard for 1:500 topographic aerial photogrammetry [58]. The maximum operational area per flight is approximately 1 km².

Using the automatic flight mode, the flight route range was scientifically designed, with appropriate route planning and measurement methods selected. The routes were reasonably segmented and merged to ensure that the final flight path was precisely adapted to the specific dimensions of the SDA, guaranteeing complete site coverage. Flight parameters were optimized to capture detailed imagery, including an elevation of 250 m for broad area coverage, a minimum side overlap of 70%, and a heading overlap of 80% for imagery consistency, complemented by a full 360° main course angle to capture all site aspects.

Following aerial imagery acquisition, the data is processed using Pix4Dmapper 4.4.12, a professional drone mapping software. The processing includes initial data preparation, aberration correction, aerial triangulation, and digital elevation modeling, followed by orthorectification, imagery mosaicking, cropping, and other key steps, culminating in the creation of an orthophoto map of the UAV survey area.

3.1.3. Sentinel-1 Satellite Data

The Sentinel-1 satellite data used in this study comes from the European Space Agency, is used to monitor elevation changes in SDAs, and serves as the basis for monitoring dynamic changes in earth and rock. The Sentinel-1 satellite features a C-band radar imaging system with capabilities spanning interferometric wide swath, strip map, extra-wide swath, and wave modes. Its advanced attributes, including dual-polarization, rapid revisit cycles, and quick product generation, significantly bolster its earth observation capabilities. This research utilizes quarterly-collected Sentinel-1 C-band synthetic aperture radar (SAR) imagery to monitor vertical changes in SDAs to accurately analyze and monitor terrain alterations within SDAs.

3.2. Model Development

The model development process corresponds to three major application scenarios: identification of disturbance boundaries, extraction of soil and water conservation measures, and estimation of spoil volume changes within SDAs in HSRs.

3.2.1. Monitoring Scenario 1: Identification of Disturbance Boundaries

During the spoil accumulation process, it is inevitable that the land resources within the SDAs will be encroached upon, and potential soil erosion issues may further exacerbate this encroachment. Therefore, it is necessary to regularly identify the disturbance boundaries of SDAs to ensure that the spoil piling process is being carried out according to protocol and to promptly detect any illegal land occupation. The process of identifying disturbance boundaries in SDAs primarily involves techniques and methods such as imagery fusion and cropping, unsupervised classification, manual verification, and spatial analysis, which are executed in four main steps.

• Step 1: Imagery fusion and cropping

This study utilized the band information from the GF-1 panchromatic multispectral camera, with the panchromatic band offering a spatial resolution of 2 m and the multispectral bands offering a spatial resolution of 8 m. To avoid imagery blurring due to mismatched spatial resolutions during subsequent unsupervised classification, data fusion was performed. Using ENVI 5.1, the multispectral data underwent radiometric calibration, atmospheric correction, and orthorectification, while the panchromatic data underwent radiometric calibration and orthorectification.

Radiometric calibration was conducted to eliminate nonlinear sensor response errors and ensure the accuracy of reflectance data. This process can be conducted by opening the “radiometric correction/radiometric calibration” module in the software toolbox and loading the required multispectral data. Following the step-by-step instructions, the calibration type should be set to reflectance, and an appropriate scale factor must be defined, thus completing the radiometric calibration of the multispectral dataset. For panchromatic data, the same process is applied by specifying the calibration type as reflectance and setting the scale factor accordingly.

Atmospheric correction was applied to remove the effects of atmospheric conditions on the imagery, ensuring spectral consistency. This process can be conducted by opening the “radiometric correction/atmospheric correction module/FLAASH atmospheric correction” in the toolbox and selecting the already radiometrically corrected multispectral data. Key parameters include geolocation (latitude and longitude), sensor type, altitude, pixel size, and the time of image acquisition. An appropriate atmospheric model and aerosol model must then be chosen based on regional climate conditions, after which the atmospheric correction process is completed.

Orthorectification was performed to eliminate geometric distortions caused by terrain undulations and sensor viewing angles, thereby providing accurate geospatial positioning for the imagery. This process can be conducted by opening the “geometric correction/orthorectification/RPC orthorectification workflow” in the toolbox, selecting the radiometrically and atmospherically corrected multispectral and panchromatic images, and setting the desired output pixel size and image resampling parameters. Upon completion, the orthorectified images offer accurate spatial information, facilitating more reliable spatial analysis and data fusion in further research.

The fusion results achieved a unified spatial resolution of 2 m, preserving the spectral characteristics of the multispectral imagery while enhancing the spatial details of the imagery. Subsequently, the fused GF-1 data were processed using ArcGIS 10.8.1. An appropriate spatial extent was delineated based on the dimensions of the study area, and the corresponding spatial reference information was established. This approach aimed to minimize the influence of other land features on the disturbance boundary identification process and to reduce the workload associated with subsequent imagery processing. During the cropping process, precise spatial reference information, including geographic coordinate systems and projection parameters, was set to ensure consistency between the imagery and other geographic data.

• Step 2: Unsupervised classification

To efficiently extract valuable information from GF-1 satellite data for identifying disturbance boundaries in SDAs, both supervised and unsupervised classification methods can be employed. Supervised classification achieves precise information extraction from GF-1 satellite data by leveraging pre-labeled training samples. In contrast, unsupervised classification rapidly extracts imagery information through automatic clustering methods, particularly in scenarios lacking prior knowledge. Unsupervised classification is especially effective in identifying clusters or categories with similar spectral characteristics within the data, making it highly suitable for delineating disturbance boundaries of SDAs in HSR projects.

Common unsupervised classification techniques include K-means and the iterative self-organizing data analysis technique (ISODATA). The K-means algorithm partitions data into a predefined number of clusters by minimizing the variance within each cluster, thereby achieving classification. However, the K-means algorithm presents certain limitations in practical applications, primarily due to the necessity of predefining the number of clusters. This requirement can be a significant constraint in unsupervised classification, as the exact number of categories in the sample data is often unknown. Additionally, K-means is sensitive to the initial placement of cluster centroids, which can result in convergence to local optima and affect the stability and accuracy of the classification results.

To address these limitations, the ISODATA method enhances the K-means algorithm by incorporating “merge” and “split” operations during the classification process, thereby producing more robust and adaptable classification outcomes. Specifically, ISODATA dynamically adjusts the number of clusters based on intra-cluster variance and inter-cluster distances in each iteration, enabling the algorithm to merge similar clusters or split overly large clusters as necessary. This dynamic adjustment mechanism significantly improves the robustness and accuracy of the classification, particularly when dealing with complex terrains and diverse spectral characteristics in the dataset. Consequently, this study employs the ISODATA method for clustering the processed GF-1 imagery. The procedure involves selecting the “classification/unsupervised classification/ISODATA classification” tool, importing the dataset, and configuring key parameters (e.g., number of classes, maximum iterations, and change threshold) in the “ISODATA parameters” dialog. Through iterative testing, the optimal number of clusters and the number of iterations are determined. By grouping pixels with similar spectral, color, texture, spatial, and morphological characteristics within the GF-1 imagery, the unsupervised classification process is effectively completed, facilitating accurate identification of disturbance boundaries in SDAs.

• Step 3: Manual verification

Since unsupervised classification methods categorize imagery pixels without directly assigning real-world attributes to objects, post-classification optimization is necessary. Additionally, due to variations in weather, incident angles, and shadowing, identical features may appear differently in GF-1 imagery, while different features might exhibit similar color and texture characteristics. Therefore, after completing unsupervised classification, expert experience is required to optimize and adjust the classification categories, correct classification patterns, and verify specific features through multiple rounds of cross-checking and iterative refinement. In this study, expert knowledge was utilized to determine the physical attributes of each land category, merging those with identical attributes to accurately delineate SDAs and non-SDAs.

• Step 4: Spatial analysis

After manual verification, the unsupervised classification imagery may still contain two types of “noise”. The first type is “noise” within the identified SDAs, where some internal regions of the SDAs are incorrectly classified as other land types. The second type involves regions initially classified as SDAs, which are sparsely distributed within vegetated areas, resulting in numerous speckles (noise). These types of “noise” can be removed using spatial filtering. Subsequently, GIS-based buffer analysis and overlay analysis techniques can be employed to further reduce misclassification. This step ensures the accurate representation of real-world attributes, thereby achieving precise imagery classification, ultimately yielding accurate disturbance boundaries of the SDAs.

3.2.2. Monitoring Scenario 2: Extraction of Soil and Water Conservation Measures

To ensure the safe operation of SDAs and to mitigate their environmental impact and potential soil erosion risks, specific engineering and vegetative measures, known as soil and water conservation measures, are implemented around the SDAs. As construction progresses, it is crucial to monitor the current state of these soil and water conservation measures as part of effective SDAs management. The monitoring process primarily involves three steps: visual interpretation, manual verification, and spatial overlay analysis, using a combination of methods and techniques.

• Step 1: Visual interpretation

This study employs visual interpretation of UAV imagery to preliminarily extract information on soil and water conservation measures within SDAs. UAV imagery provides high-resolution visual data that clearly delineate roads, construction vehicles, vegetation, and the soil and water conservation structures established around SDAs. By analyzing various features present in UAV imagery, including color, texture, structure, and size, ArcGIS is utilized to precisely delineate the spatial extent of these conservation measures through manual digitization. This manual mapping process involves directly tracing the identified conservation structures onto GIS layers, ensuring the accurate spatial representation of these measures.

• Step 2: Manual verification

Although the method of information extraction through visual interpretation is relatively straightforward, it nevertheless requires substantial professional expertise and experience. Once the initial extraction is completed, multiple rounds of review are essential to ensure both the comprehensiveness and accuracy of the identified soil and water conservation measures. This iterative verification process is first conducted by experts with relevant domain knowledge, who meticulously examine the UAV imagery to detect and correct any potential omissions or misclassifications. For instance, these experts repeatedly confirm whether the extracted soil and water conservation structures correspond to actual field conditions, ensuring that all critical measures, such as drainage systems, slope protection structures, and vegetated areas, are accurately identified and recorded. In addition, auxiliary data sources—such as high-resolution topographic maps and historical construction records—are employed for cross-validation, further enhancing data reliability and consistency. Through this multi-tiered iterative review mechanism, human errors and subjective biases can be effectively minimized, ultimately ensuring that the final information on soil and water conservation measures achieves a high level of accuracy and credibility.

• Step 3: Overlay analysis

By conducting spatial overlay analysis on information extracted from multiple time periods, this study comprehensively evaluates and monitors the dynamic changes in soil and water conservation measures within SDAs. Specifically, visual interpretation methods are employed to extract information on soil and water conservation measures from UAV imagery across different periods, including key elements such as drainage systems, slope protection structures, and vegetated areas. Subsequently, using the overlay analysis functions of ArcGIS, these multi-temporal data layers are spatially superimposed. Through a systematic comparison of the spatial positional differences and changes in spatial relationships of the extracted conservation measures across different periods, this study identifies trends in the addition, expansion, or reduction of soil and water conservation measures. Furthermore, spatial overlay analysis reveals the spatial distribution patterns of these measures at different time points, aiding researchers in gaining a deeper understanding of their maintenance and management effectiveness. By employing statistical analysis methods to quantitatively assess these changes, this study accurately evaluates the maintenance status of soil and water conservation measures and identifies critical areas that require further strengthening or optimization.

3.2.3. Monitoring Scenario 3: Estimation of Spoil Volume Changes

To ensure the stability of SDAs and prevent potential soil erosion, water contamination, and landslides, dynamic monitoring of the spoils is essential. This study utilizes the differential interferometric synthetic aperture radar (D-InSAR) method to focus on monitoring the elevation and volume changes of spoil, which involves five key steps: imagery coregistration and cropping, interferogram pairing and baseline estimation, interferogram formation and phase unwrapping, surface deformation calculation, and dynamic estimation of spoil volume changes.

D-InSAR utilizes two interferometric SAR imagery of the same region acquired at different times. The first interferogram is generated from SAR imagery captured before and after deformation, containing only topographic information. The second interferogram, similarly derived from SAR imagery taken before and after deformation, includes both topographic phase information and additional phase variations resulting from surface deformation. By performing differential processing—specifically, subtracting the first interferogram from the second interferogram—D-InSAR effectively eliminates the topographic phase component. This subtraction operation isolates the phase changes solely attributable to surface deformation, thereby enabling the precise measurement of subtle ground movements.

• Step 1: Imagery coregistration and cropping

Coregistering Sentinel-1 SAR imagery from different time periods to ensure spatial consistency is a crucial step for subsequent analysis. The coregistration process is completed by setting the master and slave imagery, subswath, satellite orbit files used, and auxiliary digital elevation model data. Additionally, to enhance processing speed and efficiency, the relevant swath and burst parameters are extracted based on the latitude and longitude data of the study area, and the appropriate imagery range is cropped accordingly.

• Step 2: Interferogram pairing and baseline estimation

The master and slave imagery for both ascending and descending orbit experiments are selected. The multilook factors to the correct ratio and perform multilook processing on the interferometric pairs are set. Based on the resolution principles, the time baseline threshold and spatial baseline threshold for both the ascending and descending orbit experiments are set, ultimately forming multiple interferometric pairs. Additionally, baseline estimation to determine whether the baseline of the interferometric pairs is below the critical value and whether good interferometric results can be achieved is performed.

• Step 3: Interferogram formation and phase unwrapping

Interferometric processing on the imagery is performed based on the spatio-temporal baseline distribution formed by the interferometric pairs to generate differential interferograms, removing terrain phase effects using external digital elevation model data. Then, filtering and enhancement are applied to the differential interferograms, and the minimum cost flow method for phase unwrapping is used to obtain the true phase values in the study area. By reviewing each generated coherence map, filtered interferogram, and unwrapped phase map, interferometric pairs with poor quality are manually removed.

• Step 4: Surface deformation calculation

A model is constructed on pixels with high coherence, using singular value decomposition to invert the deformation rate estimates, and residual terrain influences. Then, phase residuals are calculated and residual separation is performed. This process ultimately provides the surface subsidence and uplift data within the study area.

• Step 5: Dynamic estimation of spoil volume changes

Based on the concept of mathematical segmentation, SDA in HSRs can be decomposed into irregular shapes composed of multiple rectangular prisms with square bases. Specifically, the disturbance boundary of the SDA can be considered as consisting of multiple pixels, where the base size of each pixel is determined by the spatial resolution of SAR data, with a base area of $d^2$. The dynamic changes in spoil volume can be calculated using Formula (1):

$${∆}_{\mathrm{V}}={\sum }_i^n∆h_i\times d^2$$

(1)

where ${∆}_{\mathrm{V}}$ is the total volume change of spoil, $n$ and $i$ are the total number of pixels with SAR elevation changes contained in the SDAs disturbance boundaries and the $i$th pixel, $∆h_i$ is the SAR-based elevation change of the $i$th pixel, and $d$ is the pixel size of SAR.

The average elevation change can be calculated as shown in Formula (2).

$$\overline{h}={\displaystyle \frac{1}{n}}{\sum }_i^n∆h_i$$

(2)

where $\overline{h}$ is the average elevation change obtained using the D-InSAR technique.

Based on the average elevation change and dynamic volume change obtained using the D-InSAR technique, as well as the actual measured values, the errors in average elevation change and dynamic volume change are calculated.

4. Case Study

To validate the scientific robustness and practical applicability of the proposed integrated 3S monitoring framework, this study selected the representative SDAs located in the Guangdong (GD) section of the Ganzhou–Shenzhen (GS)-HSR as the case study. The monitoring focuses on three key aspects during the construction process: identification of disturbance boundaries, extraction of soil and water conservation measures, and estimation of dynamic spoil volume changes. This case study comprehensively demonstrates the application methods, technical processes, and practical effectiveness of the monitoring framework across various scenarios. Additionally, it evaluates the framework’s capacity to enhance monitoring accuracy and efficiency.

4.1. Background of the Case Study

This article takes the GD section of the GS-HSR (Figure 2) as an example to conduct research on the site selection of the spoil site. The GD section of GS-HSR spans from the border of Jiangxi and Guangdong provinces in China. The total length of this section is 297.031 km, which includes 39.612 km of railway line, 207 bridges spanning 130.892 km, and 89 tunnels totaling 126.527 km in length. This extensive project traverses twenty critical environmentally sensitive areas, encompassing eleven forest parks (including two provincial, five municipal, and four county parks), and passes through four sections designated as strict land ecological control areas by Guangdong Province. It also impacts five drinking water source protection zones, highlighting the environmental sensitivity and the need for careful ecological preservation measures in the construction and operation of this case.

The GD section of GS-HSR includes a total of 69 SDAs, covering a spoil area of 1.5844 km² and a spoil volume of 13,680,000 m³. The three largest contributors to the spoil volume are tunnels, stations, and embankments, with respective spoil volumes of 8,513,800 m³, 2,618,000 m³, and 2,017,600 m³, accounting for 62.23%, 19.14%, and 14.75% of the total spoil volume, as shown in

Table 2. Statistics of spoil volume from different sub-projects of the GD section of GS-HSR (volume unit: 10,000 m³).

Sub-Project	Excavation Volume	Fill Volume	Integrated Utilization Volume	Spoil Volume	Proportion of Spoil Volume
Roadbed	11,008,300	1,270,800	7,719,900	2,017,600	14.75%
Bridge	5,213,100	1,992,900	2,688,200	532,000	3.89%
Tunnel	17,044,600	222,100	8,308,700	8,513,800	62.23%
Station	12,794,200	4,233,300	5,942,900	2,618,000	19.14%
Large temporary facilities	3,039,500	3,039,500	0.00	0.00	0.00%
Total	49,099,700	10,758,600	24,659,700	13,681,400	100.00%

.

4.2. Result Analysis

Based on the data from the case study, this section analyzes the application effectiveness of the integrated 3S framework across three core monitoring scenarios. The results demonstrate the precise identification of disturbance boundaries, efficient extraction of soil and water conservation measures, and accurate estimation of spoil volume changes, highlighting the reliability and precision of the outcomes.

4.2.1. Identification of Disturbance Boundaries

This study selects the SDAs (Figure 3) in the GD section of the GS-HSR (from distance kilometer 257 + 850 to 260 + 400) as an example for verification. The ISODATA method was used to perform unsupervised classification on the RS imagery of the SDAs. The preliminary results of the unsupervised classification were manually verified, followed by spatial analysis to remove “noise”, ultimately yielding accurate disturbance boundaries of the SDAs.

(1)

Unsupervised classification of RS imagery from SDAs

Unsupervised classification focuses on optimizing classification parameters, which is crucial for delineating spectral clusters in GF-1 satellite imagery. Using the ISODATA unsupervised classification method, repeated testing revealed that at least three iterations and more than thirteen classifications were optimal for accurately identifying SDAs. As shown in Figure 4, 14 categories were determined, and the SDAs exhibit clear spectral differences from the surrounding vegetation, primarily due to the contrast between vegetation and bare soil. This distinction facilitates the identification of the disturbed area within the SDAs.

(2)

Manual preliminary judgment of classification results

Because the unsupervised classification method relies solely on the spectral characteristics of surface features in GF-1 satellite imagery, misclassifications may occur, necessitating manual verification of the preliminary classification outcomes to accurately identify SDAs. As illustrated in Figure 3, SDAs exhibit a distinctive gray-brown tone in the RS imagery, contrasting sharply with the surrounding vegetation. Consequently, in Figure 4, most pixels categorized under class 14 correspond to SDAs. However, a portion of non-SDA features—such as the residential area in the lower right corner of Figure 3—is also included within class 14 and must be manually removed. In addition, due to the “same object, different spectra” phenomenon, certain areas that are actually SDAs were not assigned to class 14, instead being misclassified into other categories, requiring further manual judgment and correction.

Given that this study primarily focuses on the extraction of SDA regions, only the actual physical attributes of the SDA areas are interpreted, whereas the physical properties of other land classes are not evaluated. Specifically, through expert-based interpretation and consolidation, the initially identified 14 categories from Figure 4 are integrated and summarized into two categories: “Manually verified SDAs” and “Non-SDAs”. This approach merges unsupervised classification outcomes on the basis of manual verification, conferring real-world physical meaning exclusively to the classes identified as SDAs while abstaining from interpreting the physical attributes of the remaining classes. The preliminary identification results, as shown in Figure 5, ensure high-precision SDA extraction and provide robust data support for subsequent spatial analysis and environmental management activities.

(3)

Determination and analysis of disturbance boundaries of SDAs

During the preliminary identification of SDAs, this study identified three primary issues: first, certain internal regions of SDAs were misclassified as other land classes, resulting in noise within the identification outcomes; second, transportation land and urban construction land were grouped together with SDAs, causing classification confusion; and third, some areas initially identified as SDAs were sporadically distributed within vegetated regions, leading to numerous speckle noises.

To address these issues and achieve precise boundaries for SDAs, this study first applied a 7 × 7 spatial filtering technique to the preliminary identification results, effectively eliminating misclassifications caused by local noise and thereby resolving the first and third issues. Subsequently, the spatially filtered classification results were subjected to spatial overlay analysis with land use data from 2016 (the year preceding the commencement of the GS-HSR construction) to eliminate areas where SDAs were confused with transportation land, urban construction land, and other land classes due to misclassification. This approach effectively reduced the phenomenon of other land classes being erroneously classified as SDAs. Finally, based on the spatial distance relationship of SDAs relative to the railway engineering projects, further misclassified areas were excluded. Since SDAs are constructed within a 500 m range on both sides of the railway, the research team utilized GIS-based spatial buffer analysis and overlay analysis techniques to remove misclassified patches within the 500 m range on both sides of the GS-HSR. Through a series of manual judgments and analyses, this study ultimately obtained precise disturbance boundary delineations for soil and water erosion in SDAs, as shown in Figure 6.

4.2.2. Extraction of Soil and Water Conservation Measures

This study uses a relatively small SDA in the GD section of GS-HSR (Figure 7) to demonstrate the process of collecting information on water conservation measures through UAV imagery. Based on UAV imagery, through manual visual interpretation and manual verification, the soil and water conservation measures in SDAs were extracted, and then, through spatial overlap analysis, the changes in water conservation measures in different periods were analyzed.

(1)

Manual visual interpretation and verification of UAV imagery

As depicted in Figure 7, UAV imagery provides high-resolution visual data that clearly delineate critical infrastructural elements such as roads, construction vehicles, vegetation, and the soil and water conservation measures established around the SDA during the construction stage. Utilizing distinct features such as color, texture, structure, and size present in the UAV imagery, soil and water conservation measures were meticulously extracted through a manual digitization process using ArcGIS software. This process involved identifying and outlining specific conservation structures, such as drainage systems, slope protection barriers, and vegetative buffers, based on their unique spectral and spatial characteristics.

(2)

Spatial overlay analysis of UAV imagery from multiple periods

After the extraction of soil and water conservation measures, conducting a spatial overlay analysis on UAV imagery from multiple periods is essential for assessing the temporal dynamics and current status of these conservation measures within the SDA. This analysis involves the integration of multi-temporal UAV datasets to identify and quantify changes in the spatial distribution and effectiveness of conservation practices over time.

The spatial overlay analysis was performed using ArcGIS, where UAV imagery captured at different stages of the construction process was superimposed to facilitate a comparative evaluation. By aligning and overlaying these images, this study was able to detect variations in the extent, condition, and implementation of soil and water conservation measures. Key changes, such as the expansion or reduction of drainage systems, alterations in slope protection structures, and variations in vegetative cover, were identified and quantified through this comparative approach.

4.2.3. Estimation of Spoil Volume Changes

Using quarterly Sentinel-1 C-band SAR imagery, this study examines the Yagongshan SDA as an example, employing the D-InSAR method to analyze elevation changes. Combined with the analysis of the disturbance boundaries, the dynamic changes in volume of earth and rock from the SDA were estimated.

(1)

Elevation changes of SDAs

The design height of the Yagongshan SDA was set at 25 m. During the period from May 2018 to June 2018, the actual height of the SDA increased by an average of 3.77 m. However, the monitoring imagery generated using the D-InSAR method indicated an average height increase of 4.13 m. This shows that the D-InSAR method overestimated the elevation by 0.36 m, or 9.5%. Despite this overestimation, the method’s efficiency in estimating elevation changes in SDAs is noteworthy, demonstrating significant application potential.

(2)

Estimation of dynamic changes of earth and rock in SDAs

Through the disturbance boundary identification method proposed in this study, the disturbance area of the Yagongshan SDA was determined to be 16,600 m². Between May and June 2018, the elevation change estimated using D-InSAR was 4.13 m, resulting in a calculated earth and rock volume change of 68,600 m³. A review of the construction process records during this period showed that the actual cumulative amount of earth and rock volume was 59,400 m³. The estimated data exceeded the actual data by 9200 m³, an overestimation of 15.5%. Although the accuracy of estimating dynamic changes in earth and rock volume is limited, this represents a new technique application that will revolutionize SDA monitoring and management.

5. Discussion

This study proposes an innovative monitoring framework for SDAs during the construction of HSRs by integrating 3S techniques, showcasing the novel application of 3S techniques in new monitoring scenarios. Compared to traditional methods that rely solely on manual monitoring, this approach significantly improves observation efficiency. For disturbance boundary identification of SDAs, this study employs an unsupervised classification method for processing RS imagery, which, unlike supervised classification, does not require extensive manually labeled data and can automatically identify feature connections, making it more suitable for monitoring SDAs [52]. When performing unsupervised classification, the ISODATA method enhances accuracy over the traditional K-means method. However, as unsupervised classification only classifies imagery pixels without assigning real-world attributes, it requires repeated expert validation. Combining supervised and unsupervised classification methods could potentially yield even better results.

For the extraction of soil and water conservation measures, UAV imagery is employed, effectively addressing the limitations of low resolution and slow update rates of traditional RS imagery [47]. UAV offers high-resolution capabilities, reaching centimeter or even millimeter levels, compared to the meter-level resolution of traditional RS imagery. Additionally, UAV periodicity can be flexibly adjusted based on specific needs, with minimal operational costs. However, due to the limited imagery width, UAVs are more suitable for monitoring small-scale SDAs. For larger-scale SDAs, traditional RS imagery remains more appropriate due to its broader coverage.

For monitoring dynamic changes in the volume of earth and rock, using the Yagongshan SDA as an example, the D-InSAR method overestimated the elevation change by 0.36 m, equivalent to 9.5%. This error primarily stems from the coarse resolution of the SAR data. When calculating volume changes based on the disturbance boundary of the Yagongshan SDA extracted by the GF-1 satellite, the estimated volume was 92,000 m³ higher than the actual data, an overestimation of 15.5%. The discrepancy is further exacerbated by mismatches in the spatial registration of GF-1 optical data and SAR data [40]. Despite these accuracy limitations, several studies are currently addressing this issue, indicating that this new technology demonstrates considerable potential compared to traditional methods in the monitoring and management of SDAs.

Moreover, the monitoring framework established in this study holds significant implications for promoting a circular economy. By accurately monitoring the dynamic changes within SDAs, the framework enables optimized resource management, reduces environmental impacts of waste, and promotes effective resource reutilization. For instance, precise estimation of earth and rock volume changes facilitates the rational allocation of construction resources, preventing resource wastage, and by optimizing deposition and transportation processes, it reduces carbon emissions. Additionally, the effective implementation of soil and water conservation measures helps prevent soil erosion, protect the ecological environment, and achieve the dual objectives of economic benefits and environmental protection. This comprehensive monitoring approach not only enhances the sustainability of HSR projects but also provides practical evidence for the application of a circular economy in infrastructure development.

Despite the demonstrated effectiveness of the proposed framework, there remain several limitations that warrant attention in subsequent research: (1) Monitoring accuracy depends heavily on data resolution. UAV imagery provides high spatial resolution but limited coverage, whereas traditional RS imagery offers broader coverage with reduced detail. Ongoing efforts in multi-source data fusion could balance high resolution and large-area coverage, thereby enhancing both detail and scale. (2) Unsupervised classification currently requires substantial expert validation, which can be time-consuming and subject to human error. The integration of machine learning and artificial intelligence, particularly deep learning-based feature extraction and classification, could automate key steps in data processing and reduce the burden on human analysts. (3) While periodic data acquisition offers snapshots of SDA status, it may not capture rapid changes during critical construction phases. Incorporating real-time RS systems or low-cost ground-based sensors could facilitate continuous monitoring and higher temporal resolution, improving the responsiveness of interventions. (4) Fluctuating environmental conditions, such as weather and seasonal changes, can significantly affect data accuracy, especially for soil and water conservation measures. Robust preprocessing algorithms are needed to normalize these external factors, ensuring consistent data quality and comparability over time.

6. Conclusions

This study established an innovative monitoring framework integrating 3S techniques, addressing the application of 3S techniques in the comprehensive monitoring of SDAs during the construction of HSRs. This innovative approach effectively mitigates the limitations of traditional manual monitoring, which often suffers from insufficient observation accuracy, periodicity, and efficiency. The proposed framework holds considerable promise for advancing circular economy principles by enabling the accurate monitoring of dynamic changes within SDAs. This facilitates optimized resource management, minimizes the environmental impact of waste, and promotes the efficient reutilization of construction materials.

Specifically, GF-1 satellite data, combined with unsupervised classification, expert manual verification, and spatial analysis, accurately identified the disturbance boundaries of SDAs. UAV data, coupled with visual interpretation and spatial overlay analysis, facilitated the rapid and precise extraction of water conservation measures and assessment of their maintenance status, overcoming the limitations of traditional satellite RS imagery’s low resolution and slow update rates. Using Sentinel-1 satellite data, the D-InSAR method efficiently estimated dynamic changes in SDA’s elevation and volume of spoil. Despite a 15.5% error in volume changes estimation of spoil, the method demonstrates significant potential for future applications.

Challenges remain, particularly spatial data inaccuracies and resolution limitations affecting monitoring accuracy. Future research should focus on enhancing data resolution and leveraging advanced analytical techniques, such as machine learning, to improve processing efficiency and automation for better imagery classification and water conservation measure extraction. Additionally, improving D-InSAR monitoring accuracy is crucial. With technological advancements, 3S techniques are expected to be more extensively utilized in comprehensive SDA monitoring, aiming for superior data integration and optimized D-InSAR monitoring across various scenarios. These efforts will significantly contribute to the sustainable development of HSR projects.