Next Article in Journal
Sustainable Marketing: Can Retailers’ Profit-Motivated Consumer Education Enhance Green R&D and Production?
Previous Article in Journal
Dual Mechanisms of Digital Transformation in Sustaining Green Innovation: A Supply Chain Perspective on Capability–Motivation Dynamics
Previous Article in Special Issue
Surface Deformation Monitoring and Spatiotemporal Evolution Analysis of Open-Pit Mines Using Small-Baseline Subset and Distributed-Scatterer InSAR to Support Sustainable Mine Operations
 
 
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:
Background:
Article

30-Year Dynamics of Vegetation Loss in China’s Surface Coal Mines: A Comparative Evaluation of CCDC and LandTrendr Algorithms

1
College of Geoscience and Surveying Engineering, China University of Mining and Technology (Beijing), Beijing 100083, China
2
China Coal Green Energy Technology (Beijing) Co., Ltd., Beijing 100032, China
3
Satellite Application Center for Ecology and Environment, MEE, Beijing 100094, China
*
Author to whom correspondence should be addressed.
Sustainability 2025, 17(20), 9011; https://doi.org/10.3390/su17209011 (registering DOI)
Submission received: 16 August 2025 / Revised: 4 October 2025 / Accepted: 10 October 2025 / Published: 11 October 2025
(This article belongs to the Special Issue Application of Remote Sensing and GIS in Environmental Monitoring)

Abstract

Large-scale vegetation loss induced by surface coal mining constitutes a critical driver of regional ecological degradation. However, the applicability of existing change detection methodologies based on remote sensing within complex mining areas under diverse climatic conditions remains systematically unverified. To address this gap and reveal nationwide disturbance patterns, this study systematically evaluates the performance of two algorithms—Continuous Change Detection and Classification (CCDC) and Landsat-based Detection of Trends in Disturbance and Recovery (LandTrendr)—in identifying vegetation loss across three major climatic zones of China (the humid, semi-humid, and semi-arid zones). Based on the optimal algorithm, the vegetation loss year and loss magnitude across all of China’s surface coal mining areas from 1990 to 2020 were accurately identified, enabling the reconstruction of the comprehensive, nationwide spatio-temporal pattern of mining-induced vegetation loss over the past 30 years. The results show that: (1) CCDC demonstrated superior stability and significantly higher accuracy (OA = 0.82) than LandTrendr (OA = 0.31) in identifying loss years across all zones. (2) The cumulative vegetation loss area reached 1429.68 km2, with semi-arid zones accounting for 86.76%. Temporal analysis revealed a continuous expansion of the loss area from 2003 to 2013, followed by a distinct inflection point and decline during 2014–2016 attributable to policy-driven regulations. (3) Further analysis revealed significant variations in the average magnitude of loss across different climatic zones, namely semi-arid (0.11), semi-humid (0.21), and humid (0.25). These findings underscore the imperative for region-specific restoration strategies to ensure effective conservation outcomes. This study provides a systematic quantification and analysis of long-term, nationwide evolution patterns and regional differentiation characteristics of vegetation loss induced by surface coal mining in China, offering critical support for sustainable development decision-making in balancing energy development and ecological conservation.

1. Introduction

As China’s primary energy source, coal serves as a fundamental pillar of national energy security [1,2]. Compared to underground mining, surface mining offers significant advantages in operational safety, production efficiency, and resource recovery rates [3,4,5]. Consequently, the proportion of surface coal mining in China’s total output has increased steadily. However, surface coal mining activities directly disrupt topsoil and native vegetation, causing significant harm to regional ecosystems [6,7,8].
Vegetation is pivotal to ecological reclamation in mining areas. Extensive research has been conducted to monitor disturbances utilizing remote sensing. Existing methods can be categorized into three categories. The first utilizes vegetation indices and statistical approaches (e.g., linear regression [9,10], trend analysis [11,12], and coefficient of variation analysis [13,14]) to characterize spatiotemporal vegetation dynamics. In the studies of surface mining disturbances, this category has been demonstrated to effectively capture long-term vegetation change trends (e.g., gradual vegetation degradation over a period of 5–10 years due to continuous mine expansion). However, its capacity to respond to short-term changes is significantly constrained, which is a critical requirement for the effective monitoring of open-pit mines. This limitation results in an inability to precisely identify the specific year of short-term mining induced disturbance, and further fails to link discrete disturbance signals to individual mine extension activities. The second category extracts disturbance through land cover classification, mapping transitions to disturbance types [15,16]. This approach discretizes continuous spectral information, fails to quantify disturbance magnitude, and propagates errors through multi-temporal classifications [17]. The third category is concerned with the analysis of vegetation index time-series trajectories to simultaneously capture spatial extent, timing, and magnitude of disturbance. This approach succeeds in overcoming the limitations of prior methods. Among these time-series algorithms, the Landsat-based Detection of Trends in Disturbance and Recovery (LandTrendr) [18] and Continuous Change Detection and Classification (CCDC) [19] are two prominent and widely adopted representatives [20,21]. LandTrendr employs a temporal segmentation strategy, fitting piecewise linear functions to annual image composites to identify both abrupt changes and gradual trends [22]. Conversely, CCDC employs a harmonic regression model on all available images in the time series, thereby accounting for seasonal cycles and facilitating the detection of more subtle and continuous changes [23]. The efficacy of both algorithms in mapping forest disturbances caused by logging and fire has been well-documented. In recent times, there has been an increase in the utilization of these devices in mining environments, indicating a potential for the identification of associated vegetation loss [24,25]. However, despite the common purpose shared by these approaches, fundamental differences in modeling methodology can lead to considerable discrepancies in output for the same study area [26]. A recent comparative study highlighted such differences in forested regions [27], but similar evaluations remain scarce in the context of mining related disturbances. Although some scholars have analyzed the applicability of LandTrendr and CCDC in the Shendong coal base (arid and semi-arid climate zone) [28], their conclusions have clear regional limitations. A growing number of recent studies have emphasized the necessity for systematic evaluations of these algorithms across diverse geographic and ecological settings.
To address this gap, this paper applies LandTrendr and CCDC to China’s surface coal mines across semi-arid, semi-humid, and humid climates. The objectives of this study are threefold: (1) A rigorous comparison is made of their applicability and performance differences under these distinct climatic conditions; (2) Employing the optimal algorithm, this study extracts vegetation loss year and magnitude for all mining areas from 1990 to 2020; (3) The subsequent calculation of cumulative vegetation loss area and interannual loss dynamics is then undertaken. This work provides robust methodological validation for complex climatic conditions and, more importantly, pioneers the discovery of comprehensive nationwide evolution patterns and regional differentiation characteristics of mining induced vegetation loss of China. This quantification of mining impacts provides a scientific basis for measuring and monitoring the environmental dimension of sustainability in the mining sector.

2. Materials and Methods

2.1. Study Areas and Data

This paper encompasses all surface coal mining areas in China. Spatial boundaries were delineated using 2020 mining right data from China’s Ministry of Natural Resources, covering 329 surface coal mines across 15 provinces (Figure 1). The total study area spans approximately 4100 km2. Spatially, Inner Mongolia exhibits the highest concentration of mines (66%), followed by Xinjiang (8%). Shanxi, Yunnan, and Shaanxi provinces also hold significant resources (7%, 5%, and 4%, respectively). These mines exhibit pronounced climatic heterogeneity, with 56% located in semi-arid zones (annual precipitation < 400 mm; natural vegetation: grassland), 28% in arid zones (<200 mm; desert), 12% in humid zones (>800 mm, with maximum annual precipitation typically exceeding 1600 mm, particularly in mountainous areas of the southeastern regions; natural vegetation: forest), and 4% in semi-humid zones (>400 mm, generally ranging between 400 and 800 mm; natural vegetation: forest-steppe/shrub-steppe). Mines in arid zones were excluded due to sparse vegetation and undetectable disturbances. Consequently, the study areas were categorized into semi-arid, semi-humid, and humid zones.
The data consist predominantly of 16,188 Landsat images and 852 Sentinel-2A images acquired from Google Earth Engine (GEE) between 1990 and 2021. Additionally, 813 high-resolution remotely sensed imagery from GF satellites were obtained from the China Centre for Resources Satellite Data and Application.

2.2. Change Detection Algorithms of Vegetation

The overall methodological workflow for detecting and analyzing vegetation loss is summarized in Figure 2. The process involves data preprocessing, the application of both CCDC and LandTrendr algorithms, and accuracy assessment. In this paper, “vegetation disturbance” is used as a broader term referring to changes in vegetation caused by various factors (both natural and anthropogenic), manifesting as fluctuations in Normalized Difference Vegetation Index (NDVI). “Vegetation loss” is specifically used to denote the complete removal of vegetation from surface mining activities.

2.2.1. Principles of LandTrendr

LandTrendr is a temporal segmentation algorithm for detecting changes in satellite image time-series. It segments and fits annual time-series data to model pixel-level trajectories. The key steps include: spike removal, vertex identification, trajectory fitting, model simplification, and optimal model selection [29,30,31].
In this study, NDVI was used to monitor vegetation loss in surface coal mining areas across China’s climatic zones. NDVI is a robust and widely recognized indicator of vegetation vitality [32,33]. Figure 3 shows a representative example of NDVI time-series segmentation using this algorithm. Surface mining activities typically cause a severe and persistent decline in NDVI, primarily due to the removal of vegetation and the exposure of soil. The algorithm is designed to identify disturbances by detecting such significant and sustained deviations in the spectral trajectory. The loss magnitude is derived from the outputs of model, calculated as the difference between the model-fitted NDVI value at the start of the disturbance segment and the model-fitted value at its nadir (lowest point).
The standard Google Earth Engine (GEE) implementation of LandTrendr (developed by Oregon State University) was utilized. This implementation is optimized for Landsat data and does not require modification for the specific application under consideration. The LandTrendr algorithm requires setting 8 main parameters to ensure identification accuracy. To ensure a fair comparison of the inherent performance of CCDC and LandTrendr, a single parameter set was applied across all climatic zones. This parameter set had been determined through preliminary sensitivity analysis to be optimal at the national scale. This design controls for the parameter variable, thereby more clearly revealing the robustness of the algorithmic models themselves across different ecological environments. The Parameter settings applied in this study are listed in Table 1.

2.2.2. Principles of CCDC

CCDC is an algorithm designed for land change monitoring, with the objective of detecting changes in land cover by analyzing satellite data time-series. The methodology uses all available Landsat images within a specified time range to establish a temporal spectral characteristic model. This model encompasses the following categories of spectral characteristics: seasonal, trend, and abrupt [34,35]. The CCDC consists of two main components: Continuous Change Detection (CCD) and classification. The CCDC detects breakpoints in time-series data through harmonic fitting and a dynamic RMSE threshold.
This study monitors vegetation loss in surface coal mining areas across China based on the CCDC. Figure 4 presents a representative example of NDVI time-series curve segmentation at the pixel level by the algorithm. By analyzing remotely sensed imagery and the curve, we observed that the vegetation pixel was disturbed due to mining activities in 2011. For CCDC, the loss magnitude is calculated based on the model-predicted values, representing the difference between the predicted NDVI value immediately before the breakpoint and the predicted value of the subsequent segment. The standard CCDC implementation available in GEE was employed. The application of a single parameter set was conducted across all climatic zones. This parameter set had previously been determined through preliminary sensitivity analysis to be optimal at the national scale. CCDC requires setting six main parameters, whose meanings and values are listed in Table 2.

2.3. Accuracy Verification

To quantitatively evaluate the accuracy of the LandTrendr and CCDC algorithms in identifying vegetation loss, a reference dataset was constructed based on Sentinel-2 imagery, high-resolution remote sensing image interpretation and field surveys (Figure 5). Accuracy assessment was performed by calculating confusion matrices [36].
The validation sample design followed established practices, employing a stratified random sampling approach that considered both the three climatic zones (semi-arid, semi-humid, humid) and the key classes of interest: the semi-arid zone (3879 points), the humid zone (1605 points), and the semi-humid zone (729 points). For each validation sample pixel, all available temporal remotely sensed imagery spanning 1990–2020 was systematically examined. For the early years of study (1990–2000) with limited high-resolution data, the principal method of data analysis was visual interpretation of Landsat time-series, in order to identify the characteristic spectral-temporal signature of vegetation loss induced by mining activities. Vegetation loss from surface coal mining (e.g., topsoil stripping, overburden dumping) typically lasts years, exhibiting continuous, significant vegetation loss in imagery. The distinctive characteristics of the vegetation loss were utilized to determine the specific year of vegetation loss for each sample pixel. A single consolidated confusion matrix was constructed for the entire study period. From this, Overall Accuracy (OA) was calculated for the complete classification. User’s Accuracy (UA) and Producer’s Accuracy (PA) were derived for each annual change class and for the undisturbed class. The complete confusion matrix is provided in Appendix A due to its size.

3. Results

3.1. Accuracy Comparison

3.1.1. Identification Accuracy of Loss Year in China’s Surface Coal Mines

As illustrated in Figure 6, the comparative performance analysis of LandTrendr and CCDC in identifying vegetation loss year reveals significant differences. Overall, the CCDC demonstrates markedly superior global accuracy relative to LandTrendr. The OA for loss year identification were 0.31 and 0.82, highlighting CCDC’s enhanced robustness within the complex mining environment.
Temporal analysis revealed pronounced spatiotemporal limitations in the LandTrendr algorithm. A significant decline in PA occurred between 2008 and 2011, indicating substantial omission errors for loss events during this period. This finding indicates that LandTrendr demonstrates an inadequate level of sensitivity to particular temporal disturbance signals. Furthermore, LandTrendr’s PA was consistently and significantly lower than its UA, underscoring critical deficiencies in detection capability. This pattern indicates a conservative detection strategy prioritizing avoidance of false positives (commission errors) at the expense of a high false negative (omission error) rate. In contrast, the CCDC algorithm exhibited consistent stability in PA and UA values ranging from 0.7 to 0.9 across all years, thereby substantiating its remarkable temporal consistency. The characteristic of PA being marginally lower than UA reflects its high detection rate accompanied by a moderate level of commission errors. While the algorithm effectively captures mining induced disturbance signals, it also incorporates certain vegetation fluctuations attributable to other disturbances within its results.
Spatial validation within representative mining areas (see Figure 7) further corroborates these findings. The spatial continuity and boundary delineation accuracy of loss areas identified by CCDC were significantly superior to those generated by LandTrendr. This advantage was particularly pronounced in complex mining features such as waste dumps and open pits, where CCDC achieved enhanced precision. LandTrendr results exhibited a fragmented spatial distribution, characterized by numerous omissions within active mining areas (e.g., pit edges) and erroneous inclusions of undisturbed areas. This spatial inconsistency severely constrains its utility for large-scale mapping applications.

3.1.2. Identification Accuracy in Loss Year of Difference Zones

Analysis reveals that the predominant vegetation loss events within surface coal mines across China’s diverse climate zones occurred between 2003 and 2005 (Figure 8). Accuracy validation process has demonstrated that the CCDC consistently exhibits superior capability in capturing genuine loss events across all climate zones, thus maintaining robust performance. The CCDC achieved OA of 0.86, 0.77, and 0.81 in the humid, semi-humid, and semi-arid zones, respectively. This indicates remarkable stability in its performance with minimal inter-regional variation.
In contrast, the LandTrendr algorithm yielded OA of 0.47 and 0.57 in the semi-humid and humid zones, respectively. In stark contrast, its performance declined significantly in the semi-arid zone, achieving an OA of only 0.18, markedly lower than in the other two zones. Across all climate zones, both algorithms exhibited UA values higher than PA. This pattern suggests a common tendency towards commission errors in detecting subtle vegetation loss events. Notably, vegetation fluctuations attributable to natural factors were frequently detected alongside disturbances from mining activities.
The two algorithms demonstrated a consistent spatial differentiation pattern in terms of identification accuracy. Accuracy systematically decreased along the climate gradient, with the highest performance observed in the humid zone, followed by the semi-humid zone, and the lowest accuracy in the semi-arid zone. This spatial pattern closely correlates with baseline vegetation cover conditions characteristic of each climate zone. High baseline vegetation coverage in humid zones produces more pronounced NDVI anomalies following loss, facilitating effective algorithm detection. Conversely, the inherently sparse vegetation in semi-arid regions creates strong spectral similarity between undisturbed areas and bare soil. Following loss, the reduction in NDVI is typically subtle and spectrally ambiguous, as the reflectance signal transitions from a low vegetation condition to a bare soil state. This phenomenon results in the presence of faint breakpoint signatures within the NDVI time-series. Consequently, the algorithm frequently fails to identify a sufficiently distinct spectral change, resulting in a higher rate of missed detections and reduced overall accuracy. This inherent signal attenuation primarily accounts for the LandTrendr algorithm’s exceptionally low accuracy in semi-arid environments.
Overall, the CCDC algorithm demonstrates consistently robust identification performance across diverse climatic conditions, while LandTrendr’s performance exhibits greater dependence on baseline vegetation states.

3.2. Spatio-Temporal Characteristics of Vegetation Loss in China’s Surface Coal Mines

Figure 9 illustrates the interannual variation in vegetation loss area attributed to surface coal mining within China’s semi-arid zone from 1990 to 2020. Over these three decades, the cumulative vegetation loss area across China’s surface coal mining areas reached 1429.68 km2. Notably, anomalously low regional NDVI values in 2000 and 2001, resulting from climatic anomalies, caused algorithms to misclassify vegetation fluctuations in non-mining areas as disturbance. Consequently, the recorded loss areas for these two years are abnormally elevated. Excluding the influence of these anomalous years, the overall vegetation loss area within China’s surface coal mines exhibited an increasing trend starting in 2003, coinciding with accelerated mining activities, and continued until 2013. A distinct low period in loss area occurred between 2014 and 2016, a fluctuation pattern consistent with the trajectory of regulatory changes within the Chinese coal industry.
Analysis of loss patterns across climate zones reveals the semi-arid zone exhibited the largest cumulative loss area over the thirty-year period, reaching 1240.42 km2 and accounting for 86.76% of the national total loss area. Vegetation loss in this zone was concentrated primarily after 2005. Its peak loss area occurred in 2010, measuring 89.33 km2. The semi-humid zone accumulated a loss area of 116.87 km2 during 1990–2020, representing 8.17% of the national total. Loss events here mainly occurred after 2003. Its peak loss area was observed in 2013 at 11.04 km2. The humid zone accumulated a loss area of 72.39 km2, constituting 5.06% of the national total. After excluding an anomalous value in 1991, its loss area peaked in 2009 at 5.51 km2.

3.3. Spatio-Temporal Characteristics of Vegetation Loss Magnitude in China’s Surface Coal Mines

As illustrated in Figure 10, there is a considerable disparity in the magnitude of vegetation loss resulting from surface coal mining across China’s varied climate zones. Statistical analysis indicates that the semi-arid zone exhibits the lowest overall loss magnitude, with a mean value of 0.11. It is noteworthy that 83.28% of pixels fall within the low-magnitude range (0–0.20), thereby underscoring the relatively limited absolute vegetation impact from mining activities in this zone. The semi-humid zone displays a significantly higher mean loss magnitude of 0.21. The pixel magnitude distribution in this area displays a clear bimodal pattern, with approximately 47.32% of pixels falling within the 0–0.20 range and 44.43% within the moderate magnitude interval (0.20–0.40). This reflects a more complex and widespread degree of vegetation loss. The humid zone demonstrates the highest mean loss magnitude of 0.25. The pixel distribution pattern of the area under consideration is shown to resemble that of the semi-humid zone, albeit with higher overall magnitude; 40.88% of pixels are in the 0–0.20 range, while the largest proportion (43.05%) is concentrated in the moderate loss interval, thus confirming that vegetation in this zone experiences the most profound impact from mining activities.
Vegetation loss magnitude increases significantly with greater regional humidity. This spatial differentiation is profoundly linked to baseline ecosystem conditions. Humid zones are distinguished by luxuriant native vegetation, characterized by elevated baseline NDVI. Consequently, intense anthropogenic losses, such as surface coal mining, induce greater absolute reductions in vegetation cover, manifesting as higher calculated loss magnitude values. This underscores the vulnerability of high-productivity ecosystems to human loss. In contrast, the semi-arid zone is characterized by sparse native vegetation and inherently low baseline NDVI. It has been demonstrated that even when subjected to mining activities of comparable magnitude, the absolute magnitude of vegetation cover change is relatively smaller, resulting in lower loss magnitude values. This finding suggests a coupling effect between ecosystem fragility and loss response, rather than implying lesser loss severity.

4. Discussion

4.1. Applicability of CCDC and LandTrendr in Different Climatic Conditions

This study applied the CCDC and LandTrendr to surface coal mining areas across China’s diverse climatic zones, utilizing NDVI time-series curves [37,38] with parameters adjusted for regional contexts. The results demonstrate that CCDC consistently outperformed LandTrendr in identifying vegetation loss across all three climate zones. This finding is consistent with the findings of several recent comparative studies conducted in other geographical and thematic contexts. For instance, a global assessment by Taku and Narumasa (2025) demonstrated that CCDC achieved a higher level of accuracy (F1 score of 78.14%) compared to LandTrendr (F1 score of 71.29%) in the detection of urban land cover change [26]. This superiority was attributed to CCDC’s ability to fully utilize spectral information and explicitly model seasonal dynamics. A similar observation was reported in a study focusing on forest disturbance in Southeastern China by Ding and Li (2023), who found that CCDC outperformed LandTrendr on most accuracy metrics, with the caveat that CCDC is more sensitive to sub-annual changes [39]. This performance disparity stems from fundamental differences in their curve-fitting methodologies. CCDC utilizes all available remotely sensed imagery within the temporal range during its curve-fitting process. Harmonic regression is utilized to explicitly model seasonal dynamics alongside long-term trends. Conversely, LandTrendr constructs its time-series curve by deriving a single annual value, typically through maximum value compositing, thereby failing to capture intra-annual phenological variation. Despite the provision of higher-frequency data (such as monthly observations), LandTrendr’s piecewise linear fitting method would inherently smooth out intra-annual phenological variations. When allowing for a ±1 year temporal tolerance, CCDC’s OA for identifying loss year improved significantly: nationally from 0.82 to 0.88, in the semi-arid zone from 0.81 to 0.87, in the semi-humid zone from 0.77 to 0.89, and in the humid zone from 0.86 to 0.91. LandTrendr also exhibited improvements but to a markedly lesser degree: nationally from 0.31 to 0.37, semi-arid zone from 0.18 to 0.22, semi-humid zone from 0.47 to 0.58, and humid zone from 0.47 to 0.64. It is important to note that the present study was conducted within the spatial boundaries of mining permits, which may include other large-scale disturbances such as infrastructure development. While this approach minimizes the inclusion of non-mining related disturbances (e.g., forest management, agriculture), some commission errors may still occur. However, the distinctive spatial pattern of surface mining (characterized by large and contiguous disturbance patches) provides a relatively reliable signature for identification. It is recommended that future studies consider incorporating additional morphological parameters with a view to further improving the discrimination between mining and non-mining disturbances.
Furthermore, both algorithms achieved their highest identification accuracy within the humid zone. This is attributed to the region’s high baseline vegetation coverage, where mining activities induce strong disturbance signals with pronounced spectral characteristics. In semi-humid mining zones, the magnitude of vegetation loss is relatively smaller, and vegetation dynamics are significantly influenced by climatic variability. When the magnitude of mining loss becomes comparable to climatic fluctuations, distinguishing causative factors becomes challenging, leading to increased identification errors for both algorithms and consequently reduced performance relative to the humid zone. Semi-arid surface mining areas feature sparse, low-coverage baseline vegetation, resulting in inherently smaller absolute loss magnitudes. LandTrendr is particularly prone to oversimplifying NDVI curves into linear trends within this context, effectively removing critical phenological features. The findings of the present study contribute to the growing body of international literature on remote sensing-based loss monitoring by providing a comprehensive evaluation of two widely used algorithms in the specific context of surface coal mining across diverse climatic zones.
In this study, a unified parameter set and NDVI were utilized for a controlled comparison of algorithmic performance. However, it is imperative to recognize the inherent limitations of this approach. The findings demonstrate the performance of CCDC and LandTrendr under a specific configuration that has been meticulously selected for the purposes of this study. It is conceivable that the absolute performance of one or both algorithms, particularly in specific environmental contexts such as semi-arid zones, could be enhanced by utilizing a different vegetation index (e.g., one more sensitive to moisture stress) or a locally optimized parameter set. Consequently, the observed performance disparity between the algorithms may be more or less pronounced under different configurations. This observation does not contradict our primary finding that CCDC exhibits enhanced robustness under a standardized configuration. It is therefore recommended that future research explore the potential of ensemble approaches that leverage multiple vegetation indices and parameter spaces. This exploration has the potential to further enhance the accuracy and generalizability of large-scale vegetation loss monitoring.

4.2. Mining Induced Vegetation Loss Characteristics of China

The spatio-temporal patterns of vegetation loss within China’s surface coal mines from 1990 to 2020, revealed by this study, profoundly reflect the complex interplay between coal resource exploitation and natural geographical constraints. Temporally, the dynamics of loss area exhibit high synchronicity with national coal industry policies. The sustained upward trend observed from 2003 to 2013 corresponds directly to the expansionary “golden decade” of coal production. This period of rapid growth is well-documented in studies analyzing China’s energy policy and industrial development [40]. Conversely, the distinct low period between 2014 and 2016 resulted from stringent policy-driven development controls implemented during that phase [41].
The core mechanism driving the identified pattern of increasing vegetation loss magnitude along the climatic gradient (humid zones > semi-humid zones > semi-arid zones) lies in the inherent differences in baseline ecosystem productivity. This phenomenon is further exemplified by the predominant vegetation types, which undergo a transition from forests in humid regions to grasslands in semi-arid areas. Understanding these gradients is essential for defining region-specific sustainability thresholds and setting scientifically sound restoration goals. The high NDVI native vegetation of humid zones possesses a larger biomass base; consequently, mining-induced losses in absolute vegetation cover are more substantial, manifesting as higher calculated loss magnitude. This observation confirms the heightened sensitivity of high-productivity ecosystems to anthropogenic disturbance. In stark contrast, the inherently low baseline NDVI characteristic of semi-arid zones constrains the potential absolute loss of vegetation cover, resulting in lower loss magnitude values. Crucially, this lower metric does not signify a lesser ecological impact. Instead, it highlights the high vulnerability of semi-arid ecosystems, where the low productivity and severe environmental conditions impose significant constraints on the potential for recovery. Consequently, even losses of low magnitude have the potential to cause disproportionate long-term ecological damage, thereby pushing these systems towards critical thresholds and irreversible degradation. This underscores the profound vulnerability of these low-productivity ecosystems where even loss of relatively low magnitude may approach critical ecological thresholds, posing significant risks of triggering irreversible degradation cascades.
These findings are crucial for developing targeted sustainability strategies, ensuring that ecological restoration efforts are effective and commensurate with the specific vulnerabilities of different ecosystems: robust assessment of mining impacts must explicitly incorporate regional baseline ecosystem conditions. This aligns with the sustainability principle of context-specific solutions and the integrated approach to development. The higher loss magnitude values in humid zones require prioritizing high-magnitude restoration to recover lost functions and biomass. Conversely, lower magnitude values in semi-arid zones indicate system fragility, necessitating the most stringent environmental access standards, proactive conservation, and tailored restoration strategies to prevent ecosystem collapse.
In order to implement these differentiated management strategies in an effective manner, concrete policy measures are proposed. (1) The establishment of climate zone specific regulatory standards is imperative, with these standards recognizing varying ecological vulnerabilities. It is essential that stricter permitting requirements are implemented for semi-arid regions. (2) The implementation of a national mining loss monitoring platform is also crucial, with automated change detection algorithms being used for the timely identification of unauthorized operations. (3) Furthermore, it is necessary to strengthen environmental protection laws to include cumulative impact assessments and clear liability for ecological damage, particularly in vulnerable semi-arid ecosystems. The application of these targeted approaches provides a substantial theoretical and practical foundation for understanding differential ecological responses to anthropogenic disturbance across diverse ecosystems and for promoting sustainable socio-economic development in mining regions.

4.3. Implications for Future Work

This study applied the CCDC and LandTrendr to identify vegetation loss in surface coal mining areas across China’s diverse climatic zones. The results demonstrate that CCDC demonstrates superior detection capability for mining-induced loss across large-scale surface mining areas compared to LandTrendr. However, it is important to note that the application of either algorithm in isolation is inadequate for full representation of the impacts of mining, as both algorithms detect spectral changes but lack the capacity to determine the underlying cause of these changes. This has the potential to engender confusion between mining-related vegetation loss and changes caused by natural disasters (e.g., drought, fire), resulting in commission errors. Consequently, the central objective of our future research is to develop, validate, and rigorously test a novel algorithm specifically designed for robust, large-scale identification of vegetation loss in surface coal mining environments. A core innovation focus will be effectively disentangling mining-induced loss from natural ones, thereby significantly improving detection reliability and accuracy.

5. Conclusions

This study systematically compared the performance of CCDC and LandTrendr in identifying vegetation loss across China’s surface coal mining regions under different climatic zones. Using the optimal algorithm, we precisely extracted loss year and magnitude across all mining areas from 1990 to 2020, quantifying cumulative loss area and interannual dynamics over 30 years. The main conclusions are as follows:
  • This study provides the first empirical evidence across China’s multiple climate zones demonstrating CCDC’s superior robustness in complex mining environments. CCDC’s overall accuracy for loss year identification (OA = 0.82) significantly exceeded LandTrendr (OA = 0.31). Crucially, CCDC maintains consistent performance across climate gradients, providing critical guidance for selecting climate-adaptive algorithms in global mining environments.
  • The cumulative vegetation loss area in China’s surface coal mines (1990–2020) reached 1429.68 km2, showing strong spatial aggregation. Semi-arid zones accounted for 86.76% of the total, reflecting coal distribution constraints. Loss area increased continuously from 2003 to 2013, declining distinctly during 2014–2016 due to policy-driven regulation.
  • Vegetation loss magnitude increased significantly along the moisture gradient: semi-arid zone (0.11), semi-humid zone (0.21), and humid region (0.25). This requires regionally differentiated restoration strategies: humid zones need high-magnitude restoration to recover functionality, while lower values in semi-arid zones signal ecological fragility, demanding specific corporate accountability mechanisms.
This study utilizes a multidisciplinary approach, integrating remote sensing with ecological analysis, to underscore the operational monitoring capabilities that are imperative for achieving a balance between resource development and ecosystem protection. This contributes to the scientific foundation for sustainable mining practices.

Author Contributions

W.L.: Methodology, Formal analysis, Writing—original draft. Y.X.: Formal analysis, Writing—review and editing. H.X.: Investigation, Validation. H.Z.: Data curation, Validation. L.G.: Data curation, Validation. J.L.: Conceptualization. C.Z.: Conceptualization, Writing—review and editing. All authors have read and agreed to the published version of the manuscript.

Funding

This research was funded by the National Key Research and Development Program of China [grant number 2022YFF1303301]; the National Natural Science Foundation of China [grant number 42271480, 42371347]; and the Science and Technology Development Plan Project of the Silk Road Economic Belt Innovation-Driven Development Pilot Zone and the Urumqi-Changji-Shihezi National Innovation Demonstration Zone Grant [grant number 2023LQY02].

Data Availability Statement

The data will be made available by the authors on request.

Acknowledgments

The authors would like to acknowledge the developers of the LandTrendr and CCDC methods for making their algorithms, which were instrumental in the data processing and analysis phases of this research, publicly available.

Conflicts of Interest

Author Wanxi Liu was employed by the company China Coal Green Energy Technology (Beijing) Co., Ltd. The remaining authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Abbreviations

The following abbreviations are used in this manuscript:
LandTrendrLandsat-based detection of Trends in Disturbance and Recovery
CCDCContinuous Change Detection and Classification
NDVINormalized Difference Vegetation Index

Appendix A

Table A1. Confusion matrix for the identification of vegetation loss year by LandTrendr in China’s surface coal mines. C0 denotes the undisturbed category.
Table A1. Confusion matrix for the identification of vegetation loss year by LandTrendr in China’s surface coal mines. C0 denotes the undisturbed category.
YearC0<2003200320042005200620072008200920102011201220132014201520162017201820192020TotalUA
C0609812215894823773332284272851871531331402412561849834670.02
<200323104101300613321030352741900.55
2003001710220000000000000220.77
200402062000000000000000100.60
2005160046510000000000000590.78
20061200093612020000200001090.85
20070142022609102001000000930.65
20080200220415124000000000680.60
2009020002044641000001010610.75
20103120200344571200000100800.56
201100002020021094001001001210.90
201222020020011174580010001980.88
2013100000000043251122410001070.48
2014100002000101231635202312040.80
201530000111320001715012185212160.69
20160000000002000013416110640.64
201710002001302014522227171713030.75
20182000110220102112261953072730.71
2019510051221141317116122547266154340.61
20204700010512472420009861340.64
Total10724143311252061584714153155845282873673142305315275202136213
PA0.560.430.400.190.370.450.380.090.110.140.190.330.180.440.480.180.430.370.510.40OA = 0.31
Table A2. Confusion matrix for the identification of vegetation loss year by LandTrendr in semi-arid zone. C0 denotes the undisturbed category.
Table A2. Confusion matrix for the identification of vegetation loss year by LandTrendr in semi-arid zone. C0 denotes the undisturbed category.
YearC0<20052005200620072008200920102011201220132014201520162017201820192020TotalUA
C036352947693502891903942221591051161142112311708728540.01
<200500000413314300001424/
20050000000000000000000/
200600010000000000000011.00
200700001000000000000011.00
200800000500000000000051.00
20090000041100000001010170.65
201000000313000000000070.43
20110000000110100100100140.71
201200000000120000000030.67
20130000000049160101000310.52
201400000000001162000221780.79
2015300111300001075312111020.74
20160000000000000101110130.77
20170000013020011211355311730.78
20180011022010111226841041360.62
2019403122114131576122447190103420.56
20201001051247242000544780.56
Total44353352733963252124342491991871991523803733841523879
PA0.820.000.00 0.02 0.01 0.01 0.03 0.01 0.02 0.01 0.08 0.33 0.38 0.07 0.36 0.23 0.49 0.29 OA = 0.18
Table A3. Confusion matrix for the identification of vegetation loss year by LandTrendr in semi-humid zone. C0 denotes the undisturbed category.
Table A3. Confusion matrix for the identification of vegetation loss year by LandTrendr in semi-humid zone. C0 denotes the undisturbed category.
YearC0<2003200320042005200620072008200920102011201220132014201520162017201820192020TotalUA
C0113516106181119164212158121516872410.00
<20030255271062011191643161581520181272890.08
20030011000000000000000020.50
20040202200000000000000060.33
20050000251000000000000080.25
2006100001021200000000000160.63
200700000075100000000000130.54
20080000000410000000000050.80
2009000000001401000000000150.93
2010000000001175400000100280.61
2011000000000011300000000140.79
2012000000000005500001000560.98
201300000000000211810100000500.36
2014000000000001060520000680.88
2015000000000000011780100270.63
2016000000000000000191000200.95
2017000000000000000049810580.84
2018100000000000000001710190.89
2019000000000000000000210211.00
202016000000000000000007140.50
Total44665112516303036331273486314471453514729
PA0.250.540.17 0.40 0.18 0.40 0.44 0.13 0.47 0.47 0.33 0.43 0.53 0.70 0.55 0.43 0.69 0.38 0.60 0.50 OA = 0.47
Table A4. Confusion matrix for the identification of vegetation loss year by LandTrendr in humid zone. C0 denotes the undisturbed category.
Table A4. Confusion matrix for the identification of vegetation loss year by LandTrendr in humid zone. C0 denotes the undisturbed category.
YearC0<2003200320042005200620072008200920102011201220132014201520162017201820192020TotalUA
C02360710233779331917211633914159643720.06
<2003237910020000000200000201180.67
2003001600220000000000000200.80
20040004000000000000000041.00
2005160044000000000000000510.86
2006020008240002000020000920.89
20070142022524002001000000790.66
20080200220324124000000000580.55
2009020002002140000000000290.72
2010312020002252800000000450.56
2011000020200188000000000930.95
201222020020010117580000001390.84
2013100000000002172040000260.65
20141000020001001241000010580.71
20150000000002000658117210870.67
20160000000002000013124000310.39
20171000200000001341434130720.60
201810000000000010000941931180.80
2019110020000024000010555710.77
2020210000000000000000435420.83
Total59170371681129694560671171525494843480109101471605
PA0.390.460.43 0.25 0.54 0.64 0.75 0.71 0.35 0.37 0.75 0.77 0.31 0.44 0.69 0.35 0.54 0.86 0.54 0.74 OA = 0.57
Table A5. Confusion matrix for the identification of vegetation loss year by CCDC in China’s surface coal mines. C0 denotes the undisturbed category.
Table A5. Confusion matrix for the identification of vegetation loss year by CCDC in China’s surface coal mines. C0 denotes the undisturbed category.
Year C0<2003200320042005200620072008200920102011201220132014201520162017201820192020TotalUA
C092102010002241610100001220
<20031218724151113784775910101193490.62
2003003310000012032400030490.67
2004002265121211300101112500.52
200500101063533001001101001250.85
200608006171730111101111102040.84
200701000411751441212120001450.81
2008320014154104251013234554700.87
2009040011313341330001140013760.91
20100200031536235910333352343270.72
201100002222171749123676544545970.82
201200000003410254321083163505100.85
201300000001091136219102368213080.71
2014000000020125543003141203390.88
201521000010245049258695313100.83
201600000024202056616676222100.79
201740000003031133415423161014870.87
2018120000020131237834428635010.85
2019410001110242163693045375310.85
202001000012012310103781732030.85
Total10724143311252061584714153155845282873673142305315275202136213
PA0.860.900.77 0.84 0.85 0.83 0.74 0.87 0.82 0.75 0.84 0.82 0.76 0.82 0.82 0.72 0.80 0.81 0.87 0.81 OA = 0.82
Table A6. Confusion matrix for the identification of vegetation loss year by CCDC in semi-arid zone. C0 denotes the undisturbed category.
Table A6. Confusion matrix for the identification of vegetation loss year by CCDC in semi-arid zone. C0 denotes the undisturbed category.
YearC0<20052005200620072008200920102011201220132014201520162017201820192020TotalUA
C03700100022480010000550.67
<2005134231111690104648101371300.26
20050026053000000110100370.70
20060033533011010111110520.67
20070002441012001202000550.80
2008001493431231002133553830.90
20090011311276330001130013040.91
20100103153515211322342342220.68
201100022210123581655543544240.84
20120000030724214443153502730.78
20130000010892715110357212150.70
2014000002095021541121201790.86
2015000010045025158155311900.83
2016000024202055410635221420.75
20171000030210323122958813390.87
20181000010030227732297233570.83
20194001110240122462032643780.86
2020000012012110102771191440.83
Total44353352733963252124342491991871991523803733841523879
PA0.840.970.79 0.67 0.60 0.87 0.85 0.72 0.82 0.86 0.76 0.82 0.79 0.70 0.78 0.80 0.85 0.78 OA = 0.81
Table A7. Confusion matrix for the identification of vegetation loss year by CCDC in semi-humid zone. C0 denotes the undisturbed category.
Table A7. Confusion matrix for the identification of vegetation loss year by CCDC in semi-humid zone. C0 denotes the undisturbed category.
YearC0<2003200320042005200620072008200920102011201220132014201520162017201820192020TotalUA
C020000000000081000000110.18
<2003043000211021403133124710.61
20030051000000100200000090.56
20040014100000000000000060.67
200500009100100100000000120.75
2006010012020000100000000250.80
2007000002132110000010000200.65
2008100000025300001010100320.78
2009000000022300000000000250.92
2010000000001274200000000340.79
2011000000001325900100100400.63
201200000000011101010010001050.96
2013000000000016269101000440.59
2014000000000103064201000710.90
2015000000000000022631000320.81
2016000000000000010344100400.85
2017100000000000000257420660.86
2018000000000100010023440420.81
2019010000000000010003270320.84
2020010000000000000010010120.83
Total44665112516303036331273486314471453514729
PA0.500.930.83 0.80 0.82 0.80 0.81 0.83 0.77 0.75 0.76 0.80 0.76 0.74 0.84 0.77 0.80 0.76 0.77 0.71 OA = 0.77
Table A8. Confusion matrix for the identification of vegetation loss year by CCDC in humid zone. C0 denotes the undisturbed category.
Table A8. Confusion matrix for the identification of vegetation loss year by CCDC in humid zone. C0 denotes the undisturbed category.
YearC0<2003200320042005200620072008200920102011201220132014201520162017201820192020TotalUA
C0531020000000000000000560.95
<2003015172310002100001200001790.84
2003002800000000000200000300.93
2004001123110200300000000230.52
2005001071200200000000000760.93
200607002116200000000000001270.91
2007010000602022120000000700.86
2008220000642002000100000550.76
2009040000004200000001000470.89
2010010000000564701101000710.79
2011000020006210813020000001330.81
201200000000420117630000001320.89
2013000000000113420100100490.86
2014000000000202282001000890.92
2015210000002000227423000880.84
2016000000000000002260000280.93
2017200000000101011171400820.87
20180200000100010001097001020.95
201900000000000203123710031210.83
2020000000000002000000144470.94
Total59170371681129694560671171525494843480109101471605
PA0.90 0.890.76 0.75 0.88 0.90 0.87 0.93 0.70 0.84 0.92 0.77 0.78 0.87 0.88 0.76 0.89 0.89 0.99 0.94 OA = 0.86

References

  1. Soulard, C.E.; Acevedo, W.; Stehman, S.V.; Parker, O.P. Mapping extent and change in surface mines within the United States for 2001 to 2006. Land Degrad. Dev. 2016, 27, 248–257. [Google Scholar] [CrossRef]
  2. Miller, S.M.; Michalak, A.M.; Detmers, R.G.; Hasekamp, O.P.; Bruhwiler, L.M.P.; Schwietzke, S. China’s coal mine methane regulations have not curbed growing emissions. Nat. Commun. 2019, 10, 303. [Google Scholar] [CrossRef]
  3. Guo, L.; Li, J.; Zhang, C.Y.; Xu, Y.L.; Xing, J.H.; Hu, J.Y. Evaluating the Impact of Human Activities on Vegetation Restoration in Mining Areas Based on the GTWR. ISPRS Int. J. Geo-Inf. 2024, 13, 132. [Google Scholar] [CrossRef]
  4. Zhou, Y.; Li, C.Z.; Yang, W.L. Dynamic assessment of the eco-environmental effects of open-pit mining: A case study in a coal mining area (Inner Mongolia, western China). Sustainability 2025, 17, 5078. [Google Scholar] [CrossRef]
  5. Hosseinpour, M.; Osanloo, M.; Azimi, Y. Evaluation of positive and negative impacts of mining on sustainable development by a semi-quantitative method. J. Clean. Prod. 2022, 366, 132955. [Google Scholar] [CrossRef]
  6. Silvia, F.; Talia, V.; Matteo, M.D. Coal mining and policy responses: Are externalities appropriately addressed? A meta-analysis. Environ. Sci. Policy 2021, 126, 39–47. [Google Scholar] [CrossRef]
  7. Xu, Y.L.; Li, J.; Zhang, C.Y.; Raval, S.; Guo, L.; Yang, F. Dynamics of carbon sequestration in vegetation affected by large-scale surface coal mining and subsequent restoration. Sci. Rep. 2024, 14, 13479. [Google Scholar] [CrossRef] [PubMed]
  8. Liu, D.X.; He, Z.X.; Shi, H.D.; Zhao, Y.; Liu, J.B.; Liu, A.; Li, L.; Zhu, R.F. Drivers of vegetation cover and carbon sink dynamics in abandoned Shaoyang city open-pit coal mines. Sustainability 2025, 17, 7816. [Google Scholar] [CrossRef]
  9. Li, H.K.; Xu, F.; Li, Q. Remote sensing monitoring of land damage and restoration in rare earth mining areas in 6 counties in southern Jiangxi based on multisource sequential images. J. Environ. Manag. 2020, 267, 110653. [Google Scholar] [CrossRef] [PubMed]
  10. Lobo, F.L.; Costa, M.P.F.; Novo, E.M.L.M. Time-series analysis of Landsat-MSS/TM/OLI images over Amazonian waters impacted by gold mining activities. Remote Sens. Environ. 2015, 157, 170–184. [Google Scholar] [CrossRef]
  11. Li, S.J.; Wang, J.M.; Zhang, M.; Tang, Q. Characterizing and attributing the vegetation coverage changes in North Shanxi coal base of China from 1987 to 2020. Resour. Policy 2021, 74, 102331. [Google Scholar] [CrossRef]
  12. Xiao, W.; Guo, J.W.; He, T.T.; Lei, K.G.; Deng, X.Y. Assessing the ecological impacts of open-pit coal mining in Qinghai-Tibet Plateau-a case study in Muli coal field, China. Ecol. Indic. 2023, 153, 110454. [Google Scholar] [CrossRef]
  13. Yang, H.F.; Chen, W. Spatio-temporal pattern of urban vegetation carbon sink and driving mechanisms of human activities in Huaibei, China. Environ. Sci. Pollut. Res. 2022, 29, 31957–31971. [Google Scholar] [CrossRef]
  14. Zhang, C.Y.; Li, F.Y.; Li, J.; Zhang, K.; Ran, W.Y.; Du, M.H.; Guo, J.T.; Hou, G.F. Assessing the effect, attribution, and potential of vegetation restoration in open-pit coal mines’ dumping sites during 2003–2020 utilizing remote sensing. Ecol. Indic. 2023, 155, 111003. [Google Scholar] [CrossRef]
  15. Almeida-Filho, R.; Shimabukuro, Y.E. Digital processing of a Landsat-TM time series for mapping and monitoring degraded areas caused by independent gold miners, Roraima State, Brazilian Amazon. Remote Sens. Environ. 2002, 79, 42–50. [Google Scholar] [CrossRef]
  16. Townsend, P.A.; Helmers, D.P.; Kingdon, C.C.; McNeil, B.E.; de Beurs, K.M.; Eshleman, K.N. Changes in the extent of surface mining and reclamation in the Central Appalachians detected using a 1976–2006 Landsat time series. Remote Sens. Environ. 2009, 113, 62–72. [Google Scholar] [CrossRef]
  17. Nguyen, L.H.; Joshi, D.R.; Clay, D.E.; Henebry, G.M. Characterizing land cover/land use from multiple years of Landsat and MODIS time series: A novel approach using land surface phenology modeling and random forest classifier. Remote Sens. Environ. 2020, 238, 111017. [Google Scholar] [CrossRef]
  18. Kennedy, R.E.; Yang, Z.G.; Cohen, W.B. Detecting trends in forest disturbance and recovery using yearly Landsat time series: 1. LandTrendr—Temporal segmentation algorithms. Remote Sens. Environ. 2010, 114, 2897–2910. [Google Scholar] [CrossRef]
  19. Zhu, Z.; Woodcock, C.E. Continuous change detection and classification of land cover using all available Landsat data. Remote Sens. Environ. 2014, 144, 152–171. [Google Scholar] [CrossRef]
  20. Jiang, Y.; Liu, S.G.; Liu, M.C.; Peng, X.P.; Liao, X.; Wang, Z.; Gao, H.Q. A systematic framework for continuous monitoring of land use and vegetation dynamics in multiple heterogeneous mine sites. Remote Sens. Ecol. Conserv. 2022, 8, 793–807. [Google Scholar] [CrossRef]
  21. Yang, Y.J.; Erskine, P.D.; Lechner, A.M.; Mulligan, D.; Zhang, S.L.; Wang, Z.Y. Detecting the dynamics of vegetation disturbance and recovery in surface mining area via Landsat imagery and LandTrendr algorithm. J. Clean. Prod. 2018, 178, 353–362. [Google Scholar] [CrossRef]
  22. Liu, Y.X.; Xie, M.M.; Liu, J.Y.; Wang, H.H.; Chen, B. Vegetation disturbance and recovery dynamics of different surface mining sites via the LandTrendr algorithm: Case study in Inner Mongolia, China. Land 2022, 11, 856. [Google Scholar] [CrossRef]
  23. Arévalo, P.; Bullock, E.L.; Woodcock, C.E.; Olofsson, P. A suite of tools for continuous land change monitoring in google earth engine. Front. Clim. 2020, 2, 576740. [Google Scholar] [CrossRef]
  24. Xiao, W.; Deng, X.Y.; He, T.T.; Chen, W. Mapping annual land disturbance and reclamation in a surface coal mining region using google earth engine and the LandTrendr algorithm: A case study of the Shengli coalfield in Inner Mongolia, China. Remote Sens. 2020, 12, 1612. [Google Scholar] [CrossRef]
  25. Xiao, W.; Deng, X.Y.; He, T.T.; Guo, J.W. Using POI and time series Landsat data to identify and rebuilt surface mining, vegetation disturbance and land reclamation process based on Google Earth Engine. J. Environ. Manag. 2023, 327, 116920. [Google Scholar] [CrossRef]
  26. Taku, M.; Narumasa, T. Comparative Global Assessment and Optimization of LandTrendr, CCDC, and BFAST Algorithms for Enhanced Urban Land Cover Change Detection Using Landsat Time Series. Remote Sens. 2025, 17, 2402. [Google Scholar]
  27. Pasquarella, V.J.; Arévalo, P.; Bratley, K.H.; Bullock, E.L.; Gorelick, N.; Yang, Z.Q.; Kennedy, R.E. Demystifying LandTrendr and CCDC temporal segmentation. Int. J. Appl. Earth Obs. 2022, 110, 102806. [Google Scholar] [CrossRef]
  28. Li, J.; Zhang, Y.C.; Zhang, C.Y.; Xie, H.Z.; Zhang, C.Y.; Du, M.H.; Wang, Y.Y. Applicability Analysis of LandTrendr and CCDC Algorithms for Vegetation Damage Identification in Shendong Coal Base. Metal Mine 2023, 01, 55–64. [Google Scholar] [CrossRef]
  29. Liu, Q.H.; Liu, L.S.; Zhang, Y.L.; Wang, Z.F.; Guo, R.Z. Seasonal fluctuations of marsh wetlands in the headwaters of the Brahmaputra, Ganges, and Indus Rivers, Tibetan Plateau based on the adapted LandTrendr model. Ecol. Indic. 2023, 152, 110360. [Google Scholar] [CrossRef]
  30. Liu, W.J.; Guan, H.D.; Hesp, P.A.; Batelaan, O. Remote sensing delineation of wildfire spatial extents and post-fire recovery along a semi-arid climate gradient. Ecol. Inform. 2023, 78, 102304. [Google Scholar] [CrossRef]
  31. Tu, Y.W.; Liao, K.P.; Chen, Y.X.; Jiao, H.B.; Chen, G.S. Optimized Parameters for Detecting Multiple Forest Disturbance and Recovery Events and Spatiotemporal Patterns in Fast-Regrowing Southern China. Remote Sens. 2024, 16, 2240. [Google Scholar] [CrossRef]
  32. Rumpf, S.B.; Gravey, M.; Brönnimann, O.; Luoto, M.; Cianfrani, C.; Mariethoz, G.; Guisan, A. From white to green: Snow cover loss and increased vegetation productivity in the European Alps. Science 2022, 376, 1119–1122. [Google Scholar] [CrossRef]
  33. Camps-Valls, G.; Campos-Taberner, M.; Moreno-Martínez, Á.; Walther, S.; Duveiller, G.; Cescatti, A.; Mahecha, M.D.; Muñoz-Marí, J.; García-Haro, F.J.; Guanter, L.; et al. A unified vegetation index for quantifying the terrestrial biosphere. Sci. Adv. 2021, 7, eabc7447. [Google Scholar] [CrossRef]
  34. Moraes, D.; Barbosa, B.; Costa, H.; Moreira, F.D.; Benevides, P.; Caetano, M.; Campagnolo, M. Continuous forest loss monitoring in a dynamic landscape of Central Portugal with Sentinel-2 data. Int. J. Appl. Earth Obs. 2024, 130, 103913. [Google Scholar] [CrossRef]
  35. Zhou, Q.; Wang, L.; Tang, F.; Zhao, S.Y.; Huang, N.; Zheng, K.Y. Mapping spatial and temporal distribution information of plantations in Guangxi from 2000 to 2020. Front. Ecol. Evol. 2023, 11, 1201161. [Google Scholar] [CrossRef]
  36. Olofsson, P.; Foody, G.M.; Herold, M.; Stehman, S.V.; Woodcock, C.W.; Wulder, M.A. Good practices for estimating area and assessing accuracy of land change. Remote Sens. Environ. 2014, 148, 42–57. [Google Scholar] [CrossRef]
  37. Yang, Z.; Li, J.; Zipper, C.E.; Shen, Y.Y.; Miao, H.; Donovan, P.F. Identification of the disturbance and trajectory types in mining areas using multitemporal remote sensing images. Sci. Total Environ. 2018, 644, 916–927. [Google Scholar] [CrossRef]
  38. Yang, Z.; Shen, Y.; Li, J.; Jiang, H.Y.; Zhao, L.K. Unsupervised monitoring of vegetation in a surface coal mining region based on NDVI time series. Environ. Sci. Pollut. Res. 2022, 29, 26539–26548. [Google Scholar] [CrossRef]
  39. Ding, N.; Li, M. Mapping forest abrupt disturbance events in southeastern China—Comparisons and tradeoffs of Landsat time series analysis algorithms. Remote Sens. 2023, 15, 5408. [Google Scholar] [CrossRef]
  40. Hou, D.; Liang, Z.X. Dynamic supervision strategy of green mining of coal enterprises in China based on evolutionary game. Coal Eng. 2020, 52, 186–191. [Google Scholar]
  41. Shi, X.P.; Rioux, B.; Galkin, P. Unintended consequences of China’s coal capacity cut policy. Energy Policy 2018, 113, 478–486. [Google Scholar] [CrossRef]
Figure 1. Geographical distribution of surface coal mines in China.
Figure 1. Geographical distribution of surface coal mines in China.
Sustainability 17 09011 g001
Figure 2. Methodological flowchart.
Figure 2. Methodological flowchart.
Sustainability 17 09011 g002
Figure 3. Example of pixel time-series segmentation by LandTrendr. Based on remotely sensed images and curve analysis, the vegetation pixels experienced a major disturbance in 2011 due to mining activities.
Figure 3. Example of pixel time-series segmentation by LandTrendr. Based on remotely sensed images and curve analysis, the vegetation pixels experienced a major disturbance in 2011 due to mining activities.
Sustainability 17 09011 g003
Figure 4. Example of pixel time-series segmentation by CCDC.
Figure 4. Example of pixel time-series segmentation by CCDC.
Sustainability 17 09011 g004
Figure 5. Field survey documentation of representative surface coal mining sites.
Figure 5. Field survey documentation of representative surface coal mining sites.
Sustainability 17 09011 g005
Figure 6. Accuracy for the identification of vegetation loss year by LandTrendr and CCDC in China’s surface coal mines.
Figure 6. Accuracy for the identification of vegetation loss year by LandTrendr and CCDC in China’s surface coal mines.
Sustainability 17 09011 g006
Figure 7. Vegetation loss year identified by LandTrendr and CCDC.
Figure 7. Vegetation loss year identified by LandTrendr and CCDC.
Sustainability 17 09011 g007
Figure 8. Accuracy for the identification of vegetation loss year by LandTrendr and CCDC in different climate zones. (a) Semi-arid zone; (b) Semi-humid zone; (c) Humid zone.
Figure 8. Accuracy for the identification of vegetation loss year by LandTrendr and CCDC in different climate zones. (a) Semi-arid zone; (b) Semi-humid zone; (c) Humid zone.
Sustainability 17 09011 g008aSustainability 17 09011 g008b
Figure 9. Annual vegetation loss area from 1990 to 2020. (a) China; (b) Semi-arid zone; (c) Semi-humid zone; (d) Humid zone.
Figure 9. Annual vegetation loss area from 1990 to 2020. (a) China; (b) Semi-arid zone; (c) Semi-humid zone; (d) Humid zone.
Sustainability 17 09011 g009
Figure 10. Vegetation loss magnitude. (a) Semi-arid zone; (b) Semi-humid zone; (c) Humid zone.
Figure 10. Vegetation loss magnitude. (a) Semi-arid zone; (b) Semi-humid zone; (c) Humid zone.
Sustainability 17 09011 g010
Table 1. Parameter settings for the LandTrendr.
Table 1. Parameter settings for the LandTrendr.
Parameter NameData TypeValueParameter Description
maxSegmentsInteger6The maximum number of segments in time-series segmentation.
spikeThresholdFloating point0.9The threshold for suppressing spikes (1.0 indicates no suppression).
vertexCountOvershootInteger3The number of vertices by which the initial model can overshoot.
preventOneYearRecoveryBooleanTrueWhether to prevent events of recovery within one year.
recoveryThresholdFloating point0.5If a segment’s recovery rate is faster than 1/recovery threshold (per year), that segment is not allowed.
pvalThresholdFloating point0.05If the p-value of the fitted model exceeds this threshold, the current model is discarded.
bestModelProportionFloating point0.75The model with the highest proportion of vertices having p-values is selected from among the models with the lowest p-values, up to this specified proportion.
minObservationsNeededInteger9The minimum number of observations required to execute the fitted model.
Table 2. Parameter Settings for CCDC Algorithm.
Table 2. Parameter Settings for CCDC Algorithm.
Parameter NameData TypeValueParameter Description
NumSegmentsInteger10Time-series fitting
min ObservationsInteger6The minimum number of observations to trigger a breakpoint
chiSquareProbabilityFloat0.99The chi-square probability threshold for change detection [0, 1]
minNumOfYearsScalerFloat1.33The minimum year for applying the new fitting
dateFormatInteger1The type of date used for model fitting
maxIterationsInteger10,000Maximum number of iterations
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

Share and Cite

MDPI and ACS Style

Liu, W.; Xu, Y.; Xie, H.; Zhang, H.; Guo, L.; Li, J.; Zhang, C. 30-Year Dynamics of Vegetation Loss in China’s Surface Coal Mines: A Comparative Evaluation of CCDC and LandTrendr Algorithms. Sustainability 2025, 17, 9011. https://doi.org/10.3390/su17209011

AMA Style

Liu W, Xu Y, Xie H, Zhang H, Guo L, Li J, Zhang C. 30-Year Dynamics of Vegetation Loss in China’s Surface Coal Mines: A Comparative Evaluation of CCDC and LandTrendr Algorithms. Sustainability. 2025; 17(20):9011. https://doi.org/10.3390/su17209011

Chicago/Turabian Style

Liu, Wanxi, Yaling Xu, Huizhen Xie, Han Zhang, Li Guo, Jun Li, and Chengye Zhang. 2025. "30-Year Dynamics of Vegetation Loss in China’s Surface Coal Mines: A Comparative Evaluation of CCDC and LandTrendr Algorithms" Sustainability 17, no. 20: 9011. https://doi.org/10.3390/su17209011

APA Style

Liu, W., Xu, Y., Xie, H., Zhang, H., Guo, L., Li, J., & Zhang, C. (2025). 30-Year Dynamics of Vegetation Loss in China’s Surface Coal Mines: A Comparative Evaluation of CCDC and LandTrendr Algorithms. Sustainability, 17(20), 9011. https://doi.org/10.3390/su17209011

Note that from the first issue of 2016, this journal uses article numbers instead of page numbers. See further details here.

Article Metrics

Article metric data becomes available approximately 24 hours after publication online.
Back to TopTop