Waterlogging Stability Identification: Ray-Based Model Application in Mining Areas with High Groundwater Levels—A Case Study of Huainan Coal Field

Abstract

Surface subsidence and water accumulation are common consequences of underground coal mining in areas with high groundwater levels, leading to waterlogged zones. Predicting the stability of these subsidence-induced water bodies is critical for effective land reclamation, yet current methods remain inadequate, particularly when mining data are limited. This study addresses this gap by introducing a new approach to evaluate the stability of subsidence waterlogging zones. We developed a novel method based on the ray model to assess waterlogging stability in coal mining areas. Rays were cast from origins at 1° intervals to measure changes in water accumulation boundaries over time, using metrics like the Expansion Ratio Index and stability duration. The proposed method was applied to the Huainan coal field, a typical mining area with high groundwater levels in China. We studied 41 subsidence water patches, selecting ray origins for each patch and constructing a total of 14,760 rays at 1° intervals. (2) Out of all effective rays, 4250 (32.6%) were identified as stable. (3) Stability analysis classified 32.6% as “stable”, 66.4% as “observation required”, and 1.6% as “expanding.” Specific reclamation suggestions include filling shallow stable areas and developing permanent projects in larger stable zones.

1. Introduction

Coal plays a major role in global primary energy production and consumption. Global demand for coal is strong and growing. As a major energy-producing country, China accounts for about 70% of its primary energy composition through coal [1]. In China, over 90% of coal comes from underground mining. Surface mining directly removes surface rocks to form large pits, causing significant changes in the surface over a short period. In contrast, underground mining is a slow and gradual process that does not form large pits but can lead to gradual surface subsidence. Ground disturbance caused by underground mining is a global problem [2,3,4,5,6,7]. Statistics show that underground mining of ten thousand tons of raw coal can form 0.20–0.33 hm² of subsidence land. The number of subsidence areas has been increasing at a rate of 21 × 10⁴ hm² per year in China [8]. Subsidence is accompanied by dynamic changes in the mining face, which may produce continuous changes to land, soil, and moisture [9,10]. This problem is especially prominent in mining areas with high groundwater levels [11].

Underground mining causes the overlying rock layers to collapse, potentially disrupting the original underground water balance [12]. In mining areas with high water levels, groundwater flows into the mine shafts, causing the water level to rise, which can easily lead to water accumulation [13,14]. This situation becomes particularly evident as the mining depth increases. Mining activities can cause local subsidence in certain areas of the ground. These terrains are prone to water accumulation during rainfall or rises in groundwater levels. Ground subsidence areas present regular changes with the advancement of the mining face, and similar changes occur in subsidence water accumulation areas [15,16]. The dynamic changes in water accumulation disrupt the ecological balance [17,18], impacting the environment (decreased soil quality [19], degradation of ecosystem structure and function [20,21], food security (reduced or failed corn yields [22]), and national security. The Huainan coal field is one of the major coal production regions in China and a typical high groundwater level mining area. Due to prolonged underground mining activities, this region faces severe ground subsidence issues, resulting in numerous waterlogged subsidence areas. These areas easily accumulate water during heavy rainfall or rising groundwater levels, causing dynamic changes in surface water bodies and posing significant threats to local ecosystems and farmland safety [23,24]. Additionally, the complex topography of the Huainan region and the increasing mining depth present unique challenges for monitoring and managing subsidence waterlogging [25]. Conducting research on subsidence waterlogging in the Huainan coalfield not only contributes to a better understanding of the expansion characteristics of waterlogged areas but also provides scientific support for ecological restoration and land reclamation in the region [26].

At present, studies on subsidence waterlogging mainly focus on extraction and identification, area change, and landscape index change, ranging from field investigations to remote sensing satellite data applications. Satellite remote sensing is well-established in water body information extraction, using the water spectral index (WI) combined with thresholds [27,28], object-oriented methods [29] and long-time series data [26,30] to study the location, size, shape and rate of change in subsidence waterlogging [31,32,33]. Many studies have used methods such as annual average change amplitude, annual average change speed and contingency matrix of water body transfer to other land use types [34,35] to describe the spatial and temporal changes in waterlogging. However, unlike ordinary waterlogging, subsidence waterlogging dynamically changes as mining progresses. Moreover, subsidence waterlogging is not only a unique form of land use in mining areas but also a manifestation of surface subsidence and a major disturbance caused by mining. Therefore, monitoring subsidence waterlogging in mining areas should not only consider coverage information but also the expansion direction and stability issues caused by mining. Understanding the dynamic changes and stable expansion of subsidence water accumulation in time and space is crucial for environmental protection, risk assessment, and land reclamation in mining areas. However, there is still insufficient research to provide references for understanding the intensity and direction of subsidence water changes.

Traditional studies on the spatiotemporal changes in subsidence waterlogging in mining areas have primarily focused on the overall region rather than individual subsidence waterlogging patches [36,37]. As the coal mining face expands and the mining depth increases, the spatial distribution of these patches evolves into various irregular morphological combinations [38]. Additionally, the boundary changes in each direction are not entirely the same but in different states of expansion, making the change trends difficult to predict. More accurate methods need to be introduced to quantitatively study the boundary changes in each direction of subsidence waterlogging. Referring to studies on urban spatial changes and coastline direction changes, image models can be established to monitor urban boundary changes, analyze expansion characteristics, and driving factors [39]. In the study of coastline changes, the Digital Shoreline Analysis System (DSAS) [40,41,42] monitors boundary changes through baselines and is widely used in coastline change research. The ray-based method, derived from DSAS, can effectively monitor boundary changes in subsidence waterlogging. This method involves selecting ray origins and constructing rays to provide detailed analysis of boundary expansion characteristics and trends. Additionally, models such as the ellipticity index, shape ratio, and compactness can provide technical support for studying subsidence waterlogging. The Boyce–Clark shape index and fractal dimension are also useful models [43,44]. Moreover, construction land density and the migration of urban land gravity centers influence the scale and spatial distribution of urban areas. Quantitative indicators include changes in coastline length and area increases or decreases [45]. Other important indicators encompass change rates, shape variations, and transfers of shoreline types [46,47]. By utilizing remote sensing images, the DSAS method, and the ray-based method to monitor boundary changes in subsidence waterlogging, combined with common quantitative indicators, we can analyze the expansion intensity and stability of subsidence waterlogging.

The aims of the present study are (i) to develop a method to determine the stability of subsidence waterlogging in the Huainan mining area using Landsat data from 1989 to 2016 on the Google Earth Engine (GEE) platform, and (ii) to apply remote sensing technology for monitoring and interpreting the dynamic changes in subsidence waterlogging in mining areas. Specifically, we resolved the following research questions: (1) How to determine the stability of subsidence waterlogging without relying on coal mining-related information? (2) How can this method be used to monitor ecological restoration? (3) What are the potential benefits of applying this method in developing ecological restoration plans?

2. Materials

2.1. Overview of the Study Area

The Huainan coal field, located in Anhui Province, China (Figure 1), is a typical mining area with high groundwater levels in the eastern part of the country. The topography is mainly flat, with ground elevations ranging between 30 and 50 m. It has a warm temperate, humid, and semi-humid monsoon climate. The average annual precipitation is 887.1 mm, and the average annual temperature is 16.4 °C, with rain and heat occurring in the same season. The study area features a developed water system, including natural water bodies such as the Huai River, Weishan Lake, and Taibai Lake. The groundwater level is relatively high, with unconfined groundwater at a depth of about 1.5 m. When the ground subsides to this depth, waterlogging may occur. The Huainan coal field has a mining history of over 110 years, with exploitable coal reserves of 15 billion tons at depths exceeding 1 km. The number of minable coal seams is large, with an average total coal seam thickness of 20 to 30 m. Due to long-term mining activities, which have been ongoing since 1980, large water areas have formed, severely affecting the lives and production activities of residents. The study area includes towns such as Xieqiao, Guqiao, Panji, and Luji, along with numerous villages. This region is one of the most severely affected by coal mining-induced surface subsidence in Huainan, with many buildings and farmlands submerged by subsidence waterlogging.

2.2. Data Sources

The study utilized subsidence waterlogging data from the Huainan coal field, covering the period from 1989 to 2017. Data extraction and processing were conducted on the Google Earth Engine (GEE) platform to ensure accuracy and reliability. A total of 7523 Landsat images from Landsat 5, 7, and 8, spanning from 1986 to 2018, were selected. These images underwent atmospheric correction, and high-cloud pixels were filtered to retain only high-quality data. Water pixels were identified using a threshold-based index method that combined vegetation indices (NDVI, EVI) and a modified water index (MNDWI), with criteria set as [(MNDWI > EVI or MNDWI > NDVI) and EVI < 0.1]. Based on this, we calculated the annual water frequency index (AWFI) by analyzing the frequency of water pixel occurrences each year. AWFI was determined as the ratio of water pixel observations to total observations for each year, allowing classification into permanent and seasonal water bodies. Permanent water bodies (AWFI ≥ 0.75) indicated areas with sustained water presence, such as rivers and lakes, while seasonal water bodies (0.25 ≤ AWFI < 0.75) indicated intermittent water presence; pixels with AWFI < 0.25 were considered non-water. We applied a four-year sliding window to detect land-to-water and water-to-land transitions to further isolate dynamic water bodies caused by mining-induced subsidence or artificial excavation. Ultimately, we obtained subsidence waterlogging data for a total of 29 years, spanning from 1989 to 2017. The data from 1989 to 2016 were selected for studying spatial changes in the location of subsidence water bodies, while the 2017 subsidence waterlogging data were used as a ground truth reference for an accuracy assessment. This method was advantageous, as it reduced the influence of extreme weather and other transient factors, enhancing the stability of detected water body changes. However, a limitation was that changes occurring at the start or end of the time window could not be captured, resulting in missing data for these periods. Despite this limitation, the data source and processing steps were reliable, validated by previous studies, providing a robust foundation for analyzing the waterlogging stability in the Huainan coal field [26,30].

3. Methods

This study investigates the stability of water accumulation using the spatiotemporal changes in water boundaries. Firstly, necessary preprocessing is performed on the water accumulation data to ensure accuracy and usability. Secondly, the ray method is introduced to study the changes in subsidence waterlogging in various directions. We developed a method to assess the stability of subsidence waterlogging by identifying directional changes on the GEE platform. The uniqueness of this method lies in its automation, which can automatically determine the starting point and interval angles of the rays, converting the subsidence waterlogging into distance data. Finally, we introduced the expansion coefficient and expansion indexes to measure the extent of expansion in the subsidence waterlogging in each direction. Combining the stability threshold, we extracted stable subsidence waterlogging. Finally, for stable subsidence water areas, a series of reclamation suggestions are proposed. The technical route of the study is shown in Figure 2.

3.1. Data Preprocessing

This study focuses on the changes in the outermost boundary of subsidence waterlogging. We select a circular kernel with a radius of 2 pixels to perform morphological opening operations on the initial data. This involved erosion followed by dilation to eliminate small patches and supplement missing water body data caused by natural water body masking.

3.2. Construction of the Ray-Based Model

This study uses the ray method to identify directional changes in waterlogging. The basic steps are: (1) Select an origin and set the due east direction as the 0° line; (2) Choose a certain interval angle and successively emit rays in a counterclockwise direction; (3) Record the distance data from the origin to the intersection of the rays with the subsidence waterlogging boundaries for each year.

3.2.1. Selection of Ray Origins

For each subsidence waterlogging patch, the selection of the ray origin followed two key principles to ensure accuracy and relevance in capturing the directional changes in the boundary. The selected ray origin produced fitting results that closely matched the observed evolution of the waterlogging patch over time. This principle ensured that the model represented the actual physical changes in the area. The study began with the year the waterlogging patch first appeared, approximating the original mining location as accurately as possible. Starting from the initial occurrence year ensured that the selected origin reflected the starting conditions of waterlogging development. For each specific block, we observed the formation of small initial waterlogged patches. The center points of these small patches, which eventually evolved and merged into larger waterlogged areas, were chosen as the ray origins. We used these center points to construct distance lists for each ray, and the point that provided the best fit to the observed boundary evolution was selected as the final ray origin for that patch.

3.2.2. Selection of Ray Interval Angles

The interval angle between two adjacent rays was critical in the ray method, as it affected the model’s ability to fit curved subsidence waterlogging boundaries using linear segments. Smaller interval angles allowed for a finer, more accurate fit along curved boundaries, but required more computation. We initially selected five large subsidence waterlogged areas and constructed rays at 1° intervals, spanning from each area’s occurring year to 2016, to find the appropriate interval angle. We calculated the distance differences between adjacent rays, creating a dataset of 18,360 points, which we organized in ascending order. In this dataset, a sudden increase in distance differences was observed when the difference exceeded 103.2 m. Distances below 103.2 m were considered normal changes, while distances exceeding this threshold were treated as anomalies, indicating significant boundary shifts. The presence of anomalies necessitated a finer resolution in ray spacing to capture these changes accurately. We then examined several candidate interval angles—0.25°, 0.5°, 1°, 2°, and 4°—by calculating both the total distance difference between adjacent rays and the number of anomalies (distance differences > 103.2 m) for each angle. To systematically assess the effectiveness of different interval angles, we grouped them into pairs: (0.25°, 0.5°), (0.5°, 1°), (1°, 2°), and (2°, 4°). In each group, we compared the number of anomalies to determine the reduction in anomalies when decreasing the interval angle. For example, if the anomaly count was “n” at 0.5° and “m” at 0.25°, the reduction in anomalies, represented by “n-m”, indicated the improvement in boundary fit when decreasing the angle. A substantial increase in the total distance difference between adjacent rays signaled the need for a finer resolution, while a notable reduction in anomalies with a smaller interval angle reflected efficient anomaly elimination. Ultimately, we chose the interval angle that achieved the best balance by maximizing anomaly reduction and minimizing the total distance difference.

The distance data from the origin to the intersection of the rays with the subsidence waterlogging boundaries was recorded annually for each ray. This data collection in-volved measuring the straight-line distance between the origin and each boundary point intersected by the ray. Each recorded distance was logged into a dataset, allowing for a year-by-year comparison of boundary changes and enabling trend analysis of the water-logging progression.

We evaluated both user accuracy and producer accuracy to assess the reliability of the boundary delineation and ray method analysis. User Accuracy measured how accurately the model identified true waterlogged areas within the mapped boundaries, indicating its reliability in distinguishing these areas from other landscape features. This was assessed by comparing model-predicted boundaries with ground truth data from field observations or high-resolution images. Producer Accuracy, on the other hand, evaluated whether the model correctly represented all actual waterlogged areas, ensuring no areas were missed. Similar to User Accuracy, it was determined by comparing the model results with ground truth data, confirming that the model closely matched real conditions.

3.3. Construction of the Stability Model

In the process of extracting basic data and analyzing directional changes in subsidence waterlogging using the ray method, some errors typically occur. Therefore, to accurately identify stable sections of subsidence waterlogging, this study proposes the following methods and standards.

Initially, we utilized expansion coefficients and expansion force indices to determine the extent of expansion in various directions of the subsidence waterlogging. These metrics aid in quantifying the expansion of the waterlogged areas. Subsequently, we select appropriate thresholds based on the changes in distance to filter out expansions caused by errors. This ensures that we focus only on areas with minimal expansion, which are more likely to represent stable sections of subsidence waterlogging. Additionally, we incorporate thresholds based on the stable years of subsidence waterlogging to establish a comprehensive criterion for assessing their stability. By comparing the stable years at each location with the threshold, we can identify parts of the waterlogged area that have achieved a stable state.

3.3.1. Expansion Analysis of Subsidence Waterlogging

The Expansion Ratio Index (ERI) quantifies the rate of expansion in subsidence waterlogged areas. It is derived from the observed changes in distance over the previous two years. A higher ERI indicates a faster rate of increase in the ray distance in that direction, suggesting rapid development of the subsidence and a more rapid increase in the ray distance, indicating accelerated subsidence progression. The formula for calculating the ERI is as follows:

$$\mathrm{E}\mathrm{R}\mathrm{I}={\displaystyle \frac{{\mathrm{d}}_{\mathrm{n}}-{\mathrm{d}}_{\mathrm{l}}}{{\mathrm{d}}_{\mathrm{l}}}}$$

(1)

$d_n$ is the distance from the ray origin to the boundary of the subsidence waterlogged area in the following year, and $d_l$ is the distance from the ray origin to the boundary in the previous year. The ERI is calculated annually. A larger value indicates faster expansion and development of the subsidence waterlogging. Expansion coefficients for all directional rays of the subsidence waterlogging within the study area are obtained and classified. The grading standards for the expansion coefficients are detailed in

Table 1. ERI and EMI levels.

ER	Levels	EMI	Levels
0 < ERI < 0.2	slow	|EMI| < 0.2	slow momentum
0.2 ≤ ERI < 0.4	moderate	0.2 ≤ |EMI| < 0.4	moderate momentum
0.4 ≤ ERI	fast	0.4 ≤ |EMI|	high momentum

, dividing them into three levels: “slow”, “moderate”, and “fast”. These categories reflect the directional changes in distance and are derived from statistical analyses.

The Expansion Momentum Index (EMI) quantifies the prevailing trend in subsidence waterlogging by indicating the rate at which the expansion speed is changing. Derived from observing the acceleration of distance changes over the past three years, a higher EMI value denotes a more pronounced increase in the distance in each angular direction, reflecting a stronger expansion trend of the subsidence waterlogging. The formula for calculating the EMI is:

$$\mathrm{E}\mathrm{M}\mathrm{I}={\displaystyle \frac{({\mathrm{d}}_{\mathrm{n}}-{\mathrm{d}}_{\mathrm{m}})-({\mathrm{d}}_{\mathrm{m}}-{\mathrm{d}}_{\mathrm{l}})}{{\mathrm{d}}_{\mathrm{m}}-{\mathrm{d}}_{\mathrm{l}}}}={\displaystyle \frac{{\mathrm{d}}_{\mathrm{n}}+{\mathrm{d}}_{\mathrm{l}}-2{\mathrm{d}}_{\mathrm{m}}}{{\mathrm{d}}_{\mathrm{m}}-{\mathrm{d}}_{\mathrm{l}}}}$$

(2)

$d_n$ is the distance from the ray origin to the boundary of the subsidence waterlogging in the following year; $d_l$ is the distance in the previous year; $d_m$ is the distance in the middle year. The Expansion Momentum Index (EMI) is calculated once every three years. A larger absolute value of EMI indicates a more intense expansion trend of the subsidence waterlogging. The classification standards for the Expansion Momentum Index are shown in Table 1. The index is divided into three levels: “high momentum”, “moderate momentum”, and “slow momentum”. These are determined based on the speed of distance changes in various directional rays and are statistically categorized.

3.3.2. Methods for Determining the Stability of Subsidence Waterlogging

Based on previous studies on water body extraction, the average time required for water accumulation to transition from a seasonal to a permanent state is approximately 3.1 years [26,30]. Therefore, we selected three years as the threshold for stable water bodies. We analyzed historical subsidence data, focusing on ray distance variations. Variations within 10 m indicated natural fluctuations, while changes beyond 10 m suggested boundary expansion. Related studies also supported 10 m as a reasonable stability threshold, providing a solid basis for our analysis [48,49]. During preprocessing, it is essential to consider the presence of errors and to select appropriate thresholds to filter these errors. The ray method is used to analyze changes in subsidence waterlogging in various directions. (1) Based on the area of the subsidence waterlogging and the amount of distance change in various directional rays, a stability threshold is established to select parts where ray distance changes are minor and stable. (2) The stability of subsidence waterlogging is determined by considering the interannual distance changes in various directions. (3) The stability of subsidence waterlogging is comprehensively judged by considering ERI, EMI, and the threshold of ray distance changes. Specifically, rays with low ERI and EMI values in

Table 2. Rays’ classification basis of subsidence waterlogging.

Types	ERI	EMI	Distance Constant Duration
stable	0 < ERI < 0.2	0 ≤ |EMI| < 0.2	over 3 years
expending	0.4 ≤ ERI	0.4 ≤ |EMI|
observed required	others

, whose stability duration exceeds the threshold for stable years, are classified as “stable”. Rays with high ERI and EMI values, indicating ongoing expansion of the subsidence water body, are classified as “expanding”. Remaining combinations are classified as “observation required” requiring further analysis with data from subsequent years.

4. Results

4.1. Results of Ray-Based Model Construction

By 2016, subsidence waterlogging in the Huainan coal field had merged into 41 significant zones of subsidence waterlogging. Each of the 41 subsidence waterlogging areas exhibited a determination coefficient greater than 0.6. The weighted average determination coefficient is 0.84, indicating a good fit for the regression equation. Angles of 0.25°, 0.5°, 1°, 2°, and 4° are considered, with their respective total sum of adjacent ray distance differences and the count of differences exceeding 103.2 m (Figure 3). The 1° interval resulted in the most substantial reduction in the sum of adjacent ray distance differences and effectively minimized outliers, leading to its selection as the optimal ray interval angle.

4.2. Results of Waterlogging Direction Recognition

The ray method provided a list of distances from the ray origin to the boundaries of subsidence waterlogging in each direction, by subtracting the distance values of the previous year from those of the following year, obtaining the distance changes in each direction, which characterized the spatial changes in the subsidence waterlogging. In this study, we specifically chose the years 2013, 2014, and 2015 to illustrate the expansion dynamics in a representative subsidence waterlogging area within the DJ mining area, which initially formed in 2013. This subsidence waterlogging expanded under six rays at 0°, 60°, 120°, 180°, 240°, 300° during the years 2013, 2013–2014, and 2014–2015 (Figure 4). The 0° ray expanded 152 m in 2013, 704 m during 2013–2014, and 245 m during 2014–2015.

After selecting 1° as the interval angle between adjacent rays and determining their respective origins, the extracted distance data were compared and analyzed against the actual subsidence waterlogging data, and their respective areas were calculated, using the area of each year as a weighting factor to compute the area-weighted accuracy for the entire study area. The reverse construction accuracy for all years of subsidence waterlogging exceeded 96%, with an area-weighted accuracy of 96.35%, indicating high precision.

4.3. Results of Stability Identification for Subsidence Waterlogging

4.3.1. Results of Expansion Analysis

It shows the distribution of the ERI for directional rays in the study area in Figure 5. ZhangJ and PB are indicated as rapidly expanding, suggesting that subsidence waterlogging in these areas is developing quickly in these directions. GB and GQ are shown as slowly expanding, indicating that subsidence waterlogging in these areas has fully developed and possesses strong stability. Other rays are depicted at a medium speed, indicating a moderate level of development in subsidence waterlogging in these directions. Figure 5b displays the distribution of the Expansion Momentum Index for directional rays in the study area. PB and PT are indicated as high momentum, showing that subsidence waterlogging in these areas has a continuing development trend. GB and ZhangJ are shown as low momentum, indicating that geological activities in these areas are relatively stable. The remaining rays display a medium momentum, indicating a moderate level of development in subsidence waterlogging in these areas.

4.3.2. Determination of Stability for Subsidence Waterlogging

The stability threshold of 10 m is set for the annual change in ray distances within subsidence waterlogging. Areas where distance changes are below this threshold are considered stable for the corresponding year. The proportions of subsidence waterlogging stable for o year is 29.96%, for two years is 38.24%, for three years is 18.47%, for four years is 7.64%, for five years is 1.92%, and for six or more years is 3.77%. Parts that remained stable for three years or less accounted for 86.67%. It shows that most subsidence waterlogging areas experienced expansion again after maintaining stability for up to three years. For parts of subsidence waterlogging that remained stable for more than three years, the probability of experiencing further expansion events is very low. And these are considered to have reached a thoroughly stable state in the study. Therefore, three years has been selected as the threshold to distinguish between stable and expanding states of subsidence waterlogging. The part where the distance remains unchanged for more than three years is regarded as the subsidence waterlogging has reached stability. And the part where the distance remains stable for less than or equal to three years is regarded as the expansion state.

The study applies one-dimensional linear median filtering on adjacent ray types to eliminate “noise”. We select a sliding window of length 5 based on the number of similar ray types requiring filtering between them. The classification results for directional rays of subsidence waterlogging in the Huainan mining area are shown in Figure 6.

Stable rays suggest that subsidence waterlogging is in a stable developmental state. Rays marked as “observed required” require further analysis with subsequent years’ results. Expansive rays suggest that subsidence waterlogging is expanding quickly and is unstable. The stability of subsidence waterlogging in the Huainan coal field is as shown in Figure 6. Of these, 4250 rays are effectively stable, accounting for 32.6% of the total effective rays. The stability assessment results for rays in various directions show that 32.6% are “stable”, 66.4% are “observed required”, and 1.6% are “expanding”.

The analysis presented in Figure 5 and Figure 6 mainly focuses on the expansion state observed in recent years, rather than a sequence description of stability characteristics over a longer period of time in the past.

5. Discussion

5.1. Models of Subsidence Waterlogging Evolution

Examining the extensive time-series data on water bodies resulting from independent coal mining subsidence in the Huainan mining area, the spatial evolution of these subsidence waterlogging primarily falls into two categories (Figure 7): independent and combined evolution. The independent evolution type of subsidence waterlogging is characterized by the initial expansion from the originally formed patches with gradual spreading outwards based on the patches themselves. Conversely, the merged evolution type of subsidence waterlogging is characterized by an initial scattered distribution that coalesces into a single area within the same annual cycle. The developmental process of merged subsidence waterlogging also includes independent evolutionary processes. We selected two typical waterlogging patches that clearly represent the independent to illustrate these evolutionary patterns, and merged evolution types over the period from 2002 to 2016—a timeframe chosen for its consistent data availability and clarity in capturing these distinct evolutionary processes. Statistically, over 95% of the subsidence waterlogging areas in the Huainan mining area are of the merged evolution type, with the remaining small portion that has emerged in recent years being of the independent evolution type. This classification helps to clarify the different patterns of waterlogging formation and evolution in the area, highlighting the predominance of merged waterlogging evolution and its impact on landscape stability.

Within mining subsidence areas, some lands transition from terrestrial to aquatic environments due to waterlogging expansion events. These areas often maintain a state of continuous expansion for a period, which may stabilize after reaching a certain extent. Factors limiting further expansion include land reclamation measures and expansion reaching roads where further expansion is not feasible.

The observation of the “expansion-stabilization” process of mining subsidence waterlogging reveals four types of evolution for independent subsidence waterlogging. (1) Starting as terrestrial types, becoming subsidence waterlogging through waterlogging events, and being in the waterlogging expansion phase, termed the “expansion” type; (2) Starting as terrestrial types and becoming subsidence waterlogging through waterlogging events, going through an expansion phase, and ultimately reaching a stable state, termed the “expansion-stabilization” type; (3) Starting as terrestrial types and becoming subsidence waterlogging through waterlogging events, undergoing an expansion and stabilization process, and then experiencing further expansion, termed the “expansion-stabilization-expansion” type; (4) Starting as terrestrial types and becoming subsidence waterlogging through waterlogging events, undergoing two cycles of expansion and stabilization, and ultimately reaching a stable state, termed the “expansion-stabilization-expansion-stabilization” type. Combining “expansion-stabilization” events of mining subsidence water bodies, a method for predicting the boundaries of subsidence water bodies has been developed by fitting the interannual distance changes and years of directional rays. It supports long-term planning for land reclamation and ecological restoration.

5.2. Prediction of Subsidence Waterlogging Boundaries

Distances data from rays are fitted with years to construct fitting equations, extract directional changes in subsidence water bodies, and perform trend analysis. Before fitting the interannual changes in subsidence water bodies, a correlation test must be conducted between the two variables. For the entire ray network, there is a strong correlation between the distance from the origin to the boundary of the subsidence water body and the corresponding years, ensuring the reliability of subsequent fitting analysis conducted in the study. Based on the statistically determined length of years during which subsidence water bodies remain stable within the study area, directions in which the distance remains unchanged for more than three years are identified as developmentally stable sections. Following this rule, observations are made progressively backward from the current year. The distance changes in the rays over four years are calculated; if the change is zero, the expansion trend in that direction is zero. If not, all changes are zero, indicating an ongoing waterlogging expansion process; these distances and years are inputted into the fitting equation. This fitting approach, starting from the current evolutionary characteristics of the subsidence water bodies, uses three years as the threshold to distinguish between stable and expanding states, offering moderate sensitivity to changes in distance. Using this method, the distance data and years of all directional rays of subsidence water bodies in the Huainan mining area are fitted, thereby conducting a study of the directional interannual changes in subsidence water bodies. Taking a specific subsidence waterlogged area as an example, its regression rate is discussed. Figure 8 shows the linear regression rates of directional rays in a specific mining subsidence waterlogged area within mining area 1. The regression rates of rays within the 0° to 26° range are greater than 50 m/year, indicating the fast expansion and development of the subsidence water body; rates between 27° and 154° are less than 50 m/year, indicating slower development; rates between 155° and 310° are between 50 m/year and 150 m/year, indicating faster development; and rates between 311° and 359° are between 150 m/year and 310 m/year, indicating rapid development.

For each independent mining subsidence waterlogged area, predictions for the next year’s distance data are made based on the fitting equations constructed for rays at various directional angles. Additionally, using the directional angles of the rays, the corresponding boundary points of the subsidence water body for the next year are calculated. These points are connected in a counterclockwise order, from the first point calculated from the 0° ray to the last point from the 359° ray, creating a closed figure that represents the predicted boundaries of the subsidence water body for the following year. The predicted results are shown in Figure 9.

The predicted results for the next year’s subsidence waterlogging are overlaid with the actual subsidence waterlogging for that year to conduct an overlay analysis, obtaining the areas of intersection and non-intersection, and calculating both the predicted and actual areas of subsidence waterlogging. The accuracy evaluation results for each sub-mining area within the study region are shown in

Table 3. Evaluation index table. Evaluation index table comparing mining area, predicted area, true area, intersection area, user’s accuracy, producer’s accuracy, and overall accuracy percentages.

Mining Area	Predicted Area/m2	True Area/m2	Area of Intersection/m2	Users Accuracy/%	Producer Accuracy/%	Overall Accuracy/%
XQ	1489.91	1659.11	1476.50	99.10	88.99	89.30
ZhangJ	2005.23	2328.22	1956.68	97.58	84.04	82.33
GB	318.16	474.46	311.05	97.77	65.56	64.59
GQ	1070.10	1192.04	1052.02	98.31	88.25	86.94
DJ	549.25	569.94	521.74	94.99	91.54	87.33
P3	1214.55	1401.76	1185.72	97.63	84.59	82.88
ZhuJ	122.50	153.11	107.78	87.98	70.39	64.22
PB	263.72	260.13	246.03	93.29	94.58	88.56
P2	505.81	654.51	499.42	98.74	76.30	75.57
P1	1559.38	1773.29	1479.73	94.89	83.45	79.86
Total	9098.62	10,466.57	8836.68	97.12	84.43	82.51

.

5.3. Land Reclamation Suggestions

For reclamation of subsidence waterlogging in regions with high water levels, it is necessary to adopt appropriate restoration methods based on the principle of reclamation after the subsidence waterlogging has stabilized [50,51]. Reclamation methods for subsidence waterlogging include backfilling waterlogged areas, developing them into aquaculture businesses, or transforming them into tourism and leisure parks. Combining detailed hydraulic connectivity assessments can help stakeholders tailor reclamation measures based on specific environmental conditions in each region, thereby improving ecological and economic benefits. For example, in hydraulic connection areas with high pollution diffusion risks, advanced water treatment systems should be installed before initiating land reclamation. This preventive measure ensures effective control of pollution risks during the reclamation process and promotes sustainable ecological restoration. Drawing on prior research in ecological restoration for mining areas [52,53,54], this study situates its findings within broader adaptive management strategies designed to respond to the continuous environmental changes in these regions. Such strategies are essential for effective restoration, as they emphasize resilience and adaptability within fluctuating ecological conditions. By aligning our approach with these foundational principles, our recommendations remain both scientifically robust and innovative, providing a solid basis for sustainable reclamation practices. Regarding the reclamation of subsidence waterlogging in the Huainan coal field, the discussion follows the classification results of the directional rays of the subsidence waterlogging as mentioned above:

(1)

For subsidence waterlogging with stable rays: In smaller areas, such as the two independent areas in the northernmost and southernmost parts of mining area XQ shown in Figure 6, a fill reclamation method is adopted. During drier periods, soil from deeper waterlogged areas is used to fill shallower areas, which are subsequently reclaimed for agriculture. Deeper areas are converted into fishponds with tree planting and grass seeding on the pond slopes, using a “deep excavation and shallow filling” reclamation method. In contrast, larger subsidence waterlogging, like those in mining area ZhangJ (Figure 6), are developed into scenic water surfaces that enhance urban green spaces, wetland parks, and specialty tourism, alongside agricultural development in adjacent rural areas. In regions with stable subsidence and well-defined hydraulic connectivity, the introduction of vegetation restoration and soil improvement methods can enhance reclamation outcomes. For instance, introducing drought-tolerant native plants and arbuscular mycorrhizal fungi (AM fungi) can improve soil moisture retention and accelerate the natural recovery of vegetation, establishing a more resilient ecosystem [55].

(2)

For subsidence waterlogging with rays to be observed: Reclamation may proceed, but results may be incomplete, necessitating potential secondary reclamation. Thus, plans should consider costs and ecological impacts, with comprehensive reclamation measures deferred until stabilization. Possible approaches include dynamic wetland construction and edge lotus planting, with permanent structures avoided. In unstable or observation-required subsidence zones, dynamic wetland construction and vegetative boundaries can create temporary ecological balance. Additionally, methods such as nitrogen injection during ecological restoration can activate sulfate-reducing bacteria (SRB) in subsided areas, effectively reducing water pollution risks associated with mining operations [53].

(3)

For subsidence waterlogging with expanding rays: These rapidly expanding and recently formed waterlogging are difficult to stabilize and expensive to reclaim. At this stage, reclamation is not recommended. Instead, these areas are ideally suited for aquaculture. More intensive interventions are advised, such as constructing natural barriers or implementing controlled drainage systems. These measures can help stabilize the area, control water spread, and reduce potential environmental impacts, ultimately directing water expansion in a way that supports future ecological restoration efforts.

5.4. Limitations and Future Prospects

5.4.1. Analysis of Analytical Errors

The resolution and accuracy of remote sensing data significantly influenced our study’s outcomes. The use of lower resolution data may not adequately capture small-scale variations in subsidence waterlogging, potentially leading to oversimplifications that do not accurately reflect localized conditions. Furthermore, potential errors in satellite image processing or misinterpretations of spectral signatures could introduce inaccuracies in identifying and delineating waterlogged areas, affecting the reliability of our findings.

Our models, essential for simulating waterlogging dynamics and predicting future conditions, depend heavily on precise calibration using historical data. Inaccuracies in this calibration process, due to either insufficient or erroneous baseline data, can lead to significant errors in future predictions. The absence of extensive validation with independent datasets may further weaken the model’s predictive capabilities.

The algorithms used for data analysis and predictions are predicated on assumptions about data normality, homogeneity, and continuity. Non-linear and complex environmental interactions that deviate from these assumptions can introduce biases into our results. Additionally, computational constraints restrict the complexity of models that can be implemented, necessitating simplifications that may exclude relevant variables or interactions.

5.4.2. Addressing Temporal and Subsurface Data Limitations

Our analysis is constrained by the availability of data, which currently extends only up to 2016. This limitation restricts our ability to analyze long-term trends and the evolution of subsidence waterlogging over extended periods, potentially omitting significant fluctuations and events occurring outside the observed timeframe. The study primarily focuses on surface observations without an in-depth analysis of subsurface processes. This approach results in a potentially incomplete portrayal of the interactions between surface water dynamics and underground geological changes, which are crucial for a comprehensive understanding of subsidence dynamics.

5.4.3. Future Research Directions

Future studies will aim to integrate higher resolution and more recent datasets, extending the analysis period to include data up to the present day. This will enable a more detailed and accurate assessment of subsidence waterlogging trends. We propose developing more sophisticated hydrological models that account for both surface and subsurface processes, enhancing our understanding of the complex interactions that drive subsidence waterlogging. Engaging in collaborations with hydrologists and climate scientists will allow us to enrich our dataset and analytical depth. These partnerships are crucial for developing holistic models that incorporate a wider range of environmental factors, including the impact of extreme weather events and changes in nearby water systems.

6. Conclusions

In mining areas with high groundwater levels in Eastern China, mining subsidence lowers ground elevation below the water table, causing subsidence waterlogging. These extensive waterlogged areas inundate crops, damage high-quality farmland, and affect national initiatives like farmland protection and food security. The expansion of mining subsidence waterlogging is a key indicator of land stability, necessitating continuous monitoring. Timely land reclamation and ecological restoration should be implemented in stable areas, while targeted measures are required for unstable zones, advancing land reclamation and ecological restoration in mining regions. This study developed automatic identification and expansion prediction methods for mining subsidence waterlogging on the GEE platform, offering methodological and data support for environmental monitoring and land reclamation planning.

(1)

The study proposed a mechanism for automatically identifying directional changes in subsidence waterlogging using ray-casting. This includes the selection of ray origins, determination of adjacent ray intervals, and identification methods for directional changes in subsidence waterlogging, implemented through programming on the GEE platform.

(2)

The Huainan coal field was chosen as the study area with 41 ray origins and 1° intervals between rays for automatic directional change extraction. The distance from ray origins to waterlogging boundaries strongly correlated with the corresponding years, with over 86.33% of rays having a correlation coefficient above 0.8 and 67.45% above 0.9.

(3)

A stability identification method for coal mining subsidence waterlogging was developed as follows: First, the expansion coefficient and expansion force index were introduced to assess the expansion degree of subsidence waterlogging (low expansion intensity region: 0 < ERI < 0.2, 0 ≤ |EMI| < 0.2). Second, the stability was determined by the duration of the consistent interannual distance variation in the rays. Finally, comprehensive criteria were established, setting the stability threshold for interannual distance changes at 10 m and for stable years at 3 years. Using these criteria, 4250 stable rays were identified, representing 32.6% of the total effective rays and thereby delineating stable areas.

(4)

By assessing the stability of directional rays of subsidence waterlogging in the Huainan coal field, these rays were classified into three categories: “stable” (32.6%), “observation required” (66.4%), and “expanding” (1.6%). Specific reclamation suggestions were proposed accordingly. For shallow subsidence areas, a “deep excavation and shallow filling” method was used, while for large waterlogging areas, permanent projects such as aquaculture and tourism were recommended.

In the absence of detailed mining information for the area, this study utilizes remote sensing, GIS, geometric knowledge, and mathematical modeling to investigate the expansion and stability identification of subsidence waterlogging. It offers a new approach to understanding the spatial positional changes in mining subsidence waterlogging using remote sensing image data.