A Low-Cost Web Application System for Monitoring Geometrical Impacts of Surface Subsidence

Abstract

This paper develops a low-cost web application system for monitoring geometrical impacts of surface subsidence. In many of the developing countries, the method of extraction of minerals such as coal is often impractical and uneconomical, especially with surface mining. With global warming, rapid population growth, and fast-growing urbanization with a disregard for sustainability, the overall subsidence risk has significantly increased. Despite the maturity of Differential Interferometric Synthetic Aperture Radar (DInSAR) for timely monitoring of subsidence hazards, the potential of SAR constellations has been under-exploited, as most applications focus mainly on mapping unstable areas. The developed web application system exploits Sentinel-1 SAR constellation and Small-BAseline Subset (SBAS-DInSAR) technique, to provide new streamlines of information for monitoring solutions and improve disaster risk decision making. We illustrate the model by investigating and measuring potential surface subsidence caused by underground hard coal mining activities and exponential urban population growth within a major coalmine in Nigeria. Results of the yearly cumulative amount of horizontal and vertical deformation between 2016 and 2020 range from −25.487 mm to −50.945 mm and −24.532 mm to −57.161 mm, for high and low risks, respectively. Under the influence of external factors such as rising poverty and fast-growing urbanization, the destruction of in situ stress distributions will likely increase nonlinear deformations.

1. Introduction

Geometrical deformation occurs within the Earth rocks when they are continually being subjected to forces that tend to bend, twist, or fracture them [1,2]. When rocks bend, twist, or fracture, they become deformed. This change in shape or size due to subsurface movement of earth materials leads to surface subsidence [2,3]. Deformation of rock mass due to subsidence may be either elastic, plastic, brittle, or any combination of these processes [1]. This time dependent process usually causes a displacement of surface points or objects in a horizontal or vertical direction [3]. The worldwide need for energy resources required increased production of coal and other fossil fuels [3,4]. A large amount of this production will eventually come from underground mining in areas where surface mining is impractical or uneconomical [5]. Most of the major subsidence areas around the world have developed in the past half-century at accelerated rates due to the rapidly increasing exploitation of ground water, coal, oil, and gas [3]. The severe impacts created by these subsidence hazards are mostly life threatening in terms of damages to surface utilities and structures, changes in surface and underground water conditions, and environmental degradation [1,3].

Over the years, Nigeria has witnessed land deformation in response to various natural and anthropogenic factors, such as earth movement, indiscriminate withdrawal of minerals like coal, oil and gas exploration, and excessive withdrawal of ground water [5]. This is evidenced by ground fissures, landslides, and seismic hazards in various parts of the country [5,6,7]. With continuous increase in urban population growth and rapid development, it has become clear, from poorly controlled mining operations of the past [3], that we no longer have the luxury of mineral exploration without regard to present and future land use [5,7]. However, due to rising poverty and unemployment, the rapid urban population growth and developments are characterized by high levels of informalities [6,8]. With global warming and sea level rise, associated with storm surge, this high disregard for sustainability has increased the frequency of coastal flooding in most coastal areas [7,8]. All these forces combine and apply stress on already fragile socio-ecosystems, making them extremely vulnerable to an increased land surface deformation and subsidence, with adverse socio-economic impacts [5,8]. Although it is almost impossible to completely neutralize the damage due to this challenge, it is possible to minimize the potential risks by developing comprehensive disaster early warning strategies and preparing innovative developmental plans to provide resilience to such disasters [7]. Thus, monitoring such fast-changing environments is crucial to identifying key risks areas and planning adaptation responses to mitigate such disasters [1,6,8].

In the past, traditional methods were employed to monitor subsidence phenomena [3,9]. However, these methods are costly, time consuming, rigorous, and impractical to perform in inaccessible areas [10]. Furthermore, the development of land depressions may often be much faster, reaching 1 to 3 cm per 24 h; thus the real extent of the subsidence is seldom regularly surveyed onsite [6,9]. Differential Interferometric Synthetic Aperture Radar (DInSAR), has matured and offers a remedy to the limitations of traditional methods by enabling the timely monitoring of different types of natural hazards such as earthquakes, flooding, and subsidence with millimeter accuracy [3]. With the recent incoming of various satellite constellations, delivering high-resolution Synthetic Aperture Radar (SAR) images, it is now possible to detect surface changes with fine spatial details and with a short revisiting [8,11,12,13]. However, despite the maturity of the various DInSAR techniques, the potential of SAR constellations has not been fully exploited [6,8,9]. So far, most of the applications have aimed at assessing the use of SAR data sources for mapping unstable areas, rather than providing new streamlines of information for monitoring solutions [6,8,12,13,14].

In Nigeria, for example, little or no attention has been given to a land deformation hazard monitoring system which could evolve to early warning of land deformation scheme [5,8,15]. There is no record of a comprehensive, nationwide land deformation hazard system to help in the protection of critical facilities such as nuclear power plants for electricity generation, dams, rail lines, high rise buildings, and roads [6,13]. Previous studies have mainly assessed the physical [16], geological [17], and chemical [5] impacts of past mineral exploration and activities on the environment, while very little has been done to measure or monitor the land deformation on a geospatial scale. More specifically, the geometrical impacts of past coal mining and other mineral exploration on the environment, as well as their implications for future development and urbanization, is hugely underscored [5,6]. This paper addresses these challenges by developing a low-cost surface subsidence monitoring system that provides early warning information and improves disaster risk decision making. A web application model is illustrated by investigating and measuring potential surface subsidence caused by underground hard coal mining activities and exponential rapid urban population growth within a major coal mine in Nigeria known as Onyeama mine field. In line with the Sendai Framework for Disaster Risk Reduction (SFDRR 2015–2030), the model offers a solution for routine monitoring of subsidence hazards. The key feature of the application is that it combines the underlying complexity of different IT resources, to bring a scalable, reliable, sustainable, ready-to-use, on-demand, and cost-effective solution to the general public [18]. This simple, low-cost early warning system will improve preparedness to respond effectively and improve the dissemination of information about early warning signs of subsidence hazards at local and national levels.

2. The Study Area

The study area is Onyeama Mine Field Enugu State, Nigeria. It lies within latitudes 6°25′ N and 6°29′ N, and longitudes 7°25′ E and 7°30′ E, south-east, Nigeria (Figure 1). It is a rectangular area with dimension of 9248 m × 7341 m (area bounded in red). Hills and lowlands characterize the area. Minor streams and rivers drain the area, with River Ekulu draining toward the northern part and Asata River draining from the southern part. Geologically, the area is comprised of three formations: Ajali Sandstone, Mamu Formation, and Enugu Shale [17]. Mamu formation and Ajali sandstone underlie the hilly parts, while Enugu Shale underlies the lowlands. According to [17], the oldest, Enugu Shale, is exposed at the extreme eastern part of the area. It is generally fissile grey shale, dipping in direction of 2500 SW, and strike in 1500 ES–3300 NW direction. It is faulted (Normal Fault) and graded upward to Mamu Formation. It is characterized by four major lithology units: shale, sandstone, coal, and siltstone. The sandstone is fine to medium grained and moderately to weakly consolidate. The coal is generally black, comprising three thin coal seams that vary in thickness from 0.1 m to 0.9 m, which are intercalated by fine sandstone and shale beds. There are flakes of pyrite mineral within the grains of sandstone beds that make contact with the coal seams. Ajali Sandstone, which overlies Mamu Formation, is well exposed in the western part of the study area. Ajali Sandstone is whitish to faint brown with patches of iron stains, highly friable, clayey, poorly sorted coarse grained sandstone, and characterized by pronounced herring-bone cross bed. The three formations have a normal fault with other associated fractures.

3. The Web Application Model

This web application model is structured into three components: system design, data processing, and statistical analysis of the results (Figure 2).

3.1. Web Application System Design

The system architecture for our web application design is a connection of multiple components and broken down into three-tiered architecture (Figure 3). The selected architecture for the application follows the model view controller (MVC) pattern [18]. The first milestone in web application design process and task of the architect is the selection of views. This helps to identify how users will interact with the application and what will be the result of such interaction. The potential number of technologies to implement web-mapping projects is almost infinite. Therefore, any programming environment, programming language, and server-side framework can be used to implement web mapping projects [18]. For this web mapping application, we show the technologies implemented below:

(i)

Content: This defines the contents and structure of the web application page. HTML (Hyper Text Markup Language) was used to develop the content.

(ii)

Style: This defines the style and layout of the web application page. CSS (Cascading Style Sheet) was used to develop the style.

(iii)

Behavior: This controls how the web application will dynamically respond to actions and interactions. Java Script (JS) was used to develop the behavior.

(iv)

Database: This will store, organize, and manage the application data. We made use of the PostGIS Database.

(v)

Functionality: This back-end code of the application controls access to the data and how they are used. We used Geoserver to deploy this.

(vi)

Access: The application was Host on the web, so that other users may access the information online and make informed decisions. Apache Tomcat Platform was used to deploy it.

To achieve the interactive objectives of this web application, responsibilities were distributed to four controllers. The Attribute Controller is responsible for interaction between the executed request information from displayed vector and raster data. The Chart Controller is responsible for the display of the interactive line chart with statistical information. The Map Controller is responsible for rendering the various maps in raster and vector format. The Navigation Controller is responsible for the interactions with external services for the base map and, finally, the Report Controller renders the deformation statistical information. The name of the application is called “Geospatial Surface Subsidence Monitoring System’’. The choices made based on the objectives of this application were deemed appropriate and the technique fundamental for conveying the desired message to a user or a user community (Figure 4).

3.2. Web Application User Interface

The web application user interface is a layout view containing a single panel for design and clearly defines the different roles in the application. The panel contains five views (Figure 5). The top left of the interface is the Attribute Panel View where a query task can be executed to request information of the average rate of deformation over the years as well as buildings and road networks under risk. Below the Attribute Panel View is the Legend Panel View. This displays the legends of the raster information (horizontal and vertical time series deformation maps with the level of risks) and the legends of other vector information (investigation locations, building and road networks, State and LGA Boundaries etc.). The center of the application contains the Map View. The map view displays all the raster information (horizontal and vertical time series deformation maps with the level of risks) and the legends of other vector information (investigation locations, building and road networks, State and LGA Boundaries etc.). It also displays a Layers Switcher Button to switch on and off any layer at will. It also displays a Measurement Button to take measurement of Lengths and Area of various information directly on display. Further, it displays a Get Feature Info Button to select information content. A user may highlight this information with a click of the mouse. Again, the map view also contains the Reference Base Map (Satellite Image and Open Street Map) that supports the visibility of the theme on display. On the top right is the Horizontal Chart View, where the average horizontal deformation over the period of investigation of the fourteen investigation locations is displayed as a line graph. On the bottom right is the Vertical Chart View, where the average vertical deformation over the period of investigation of the fourteen investigation locations is displayed as a line graph. The statistical information is interactive and displays some background information (the investigation location and the average rate of deformation for that year) for the content that is highlighted when prompted by users. Finally, at the bottom of the interface is the Attribute Table Information. This information only comes up when a query is executed using the Attribute View Panel. It gives an overview of the attribute information of the average rate of deformation over the years as well as buildings and road networks at risk within the investigation area.

3.3. Web Application Services

Services are a distinct part of functionality that are provided through an interface. For our web application, a REST service was used and runs on the web server. The web server processes the script and generates the HTML pages that are then returned to the web browser. The service scripts contain CSS, Java Scripts, Chart.Js, and SQL statements that generate geometrical deformation and statistical information from the PostGIS database. These back-end services run behind the application, tell the web server to process the server scripts, and extract the required information for display on the web portal to the browser. In our case, it retrieves the base map, raster data (horizontal and vertical time series deformation maps with the level of risks), and vector layers (of road networks, building networks, investigation locations, and State and L.G.A boundaries) within the study area and the statistical information used to display the line chart.

3.4. Dataset Preparation

The primary source of the dataset used in this research is from the Copernicus Sentinel-1 Earth Observation Satellite SAR Data Archives, for measuring, monitoring, and mapping the surface deformation over the study area (Figure 6A). Sentinel-1 data exhibit some favorable characteristics: regional-scale mapping capability, systematic and regular SAR observations, and rapid product delivery (typically in less than 3 h from data acquisition) [13]. A total of 60 Terrain Observation with Progressive Scans SAR (TOPSAR), Single Look Complex (SLC) Sentinel 1 SAR images were acquired over a period of five (5) years, from January 2016 to December 2020. The size of each image was about 4.6 GB, making a total of approximately 276 GB for all images used. A secondary dataset was used to validate our research objectives (Figure 6B). The secondary data were acquired from two different sources: (1) GPS (X,Y,Z) field data collected from fourteen investigation locations and (2) vector shapefile of buildings and road networks which was digitized from a georeferenced high resolution Google Earth Image of 2020, within the study area.

3.5. Data Processing Using SBAS-DInSAR Technique

Among several DInSAR techniques, we adopted a parallel computing solution of [11], the Small-Baseline Subset (SBAS-DInSAR), for the processing of the archived Sentinel-1 images. The technique is an advanced DInSAR processing chain for the generation of Earth deformation time series (TS) along the satellite line of sight (LOS) and the estimation, with millimetric accuracy, of the mean yearly velocity maps (in mm/yr). The software is used for applying the SBAS-DInSAR processing chain, including Sentinel Toolbox (SNAP—Open Source), ArcGIS (Licensed), R for Spatial Statistics (Open Source), and Virtual Machine Player (VMware—Open Source). According to [11], some general notes are in order for the SBAS-DInSAR procedure. First, the processing steps from block A to G (Figure 7) are performed at full spatial resolution, whereas the subsequent steps work on multi-looked data (block H to L of Figure 7). This first step was performed using the SNAP software. Second, a common storage is assumed available to all the processing phases, i.e., each step gains access to the same common storage for reading inputs and writing outputs. This second phase was performed using SNAP and VMware software. ArcGIS and R were both used for visualization and statistical analysis, respectively. The full sequential steps of the SBAS-DInSAR processing chain through widely used metrics (such as speedup, efficiency, and load balance) are shown in Figure 7 below. The SBAS-DInSAR processing was performed stand alone (Figure 7), while statistical analysis and prediction was performed using Holt–Winter (Figure 2). Both results were deployed on the interactive web model.

4. Application of the Web Model

The web application model is illustrated by investigating and measuring potential surface subsidence caused by underground hard coal mining activities and exponential rapid urban population growth within a major coal mine in Nigeria known as Onyeama mine field. Generally, the severe effects of land subsidence in Enugu State are mostly around Onyeama mine and its environment (Figure 1). Deformation of the rock mass due to this subsidence within the area is either elastic, brittle, or a combination of these processes.

4.1. Average Absolute Horizontal Deformation Results

We obtained the horizontal deformation maps for every InSAR pair after the conversion from the residual phase correction, phase to deformation (Figure 7), masking area of low coherence, and geo-coding the products to have absolute geographical coordinates. According to the amount of deformation within each pair, the settlement is classified with three different colors (red, white, and blue), representing (high risk, medium risk, and low risk) deformations, respectively.

Figure 8A shows the 2016 average absolute horizontal deformation (mm) and investigation locations, with typical representative characteristics of the study area. Figure 8B shows the 2016 average absolute horizontal deformation (mm), the road, and building networks, which are under threat within the study area. Both images show the yearly cumulative amount of deformation ranging from −25.487 = low risk, −35.126 = medium risk, and −44.775 = high risk, respectively.

Table 1. Showing annual yearly average rate of horizontal deformation (mm) for the period of study.

S/N	Yearly Image of Study Area	Average Rate of Horizontal Deformation (mm)
1	Sentinel-1 SLC SAR 2016	−25.487 = low	−35.126 = medium	−44.775 = high
2	Sentinel-1 SLC SAR 2017	−20.893 = low	−33.072 = medium	−45.251 = high
3	Sentinel-1 SLC SAR 2018	−27.596 = low	−37.660 = medium	−47.722 = high
4	Sentinel-1 SLC SAR 2019	−24.636 = low	−37.778 = medium	−51.115 = high
5	Sentinel-1 SLC SAR 2020	−28.134 = low	−39.539 = medium	−50.945 = high

summarizes the annual yearly average horizontal deformation (mm) over a period of five years (2016–2020) of our study area with the associated level of risks. The low risk level ranges from −20.893 to −28.134, the medium risk level ranges from −33.072 to −39.539, and the high-risk level ranges from −44.775 to −51.115. The year with the lowest level of risk is 2020, with −28.134, while the year with the highest level of risk is 2016 with −44.775. This high risk in 2016 is likely due to new low-income dwellers, who relocated to urban centers due to rising insecurity in vulnerable rural areas [6]. They settled in at risk areas and live in houses that cannot resist hazard shocks.

4.2. Average Absolute Vertical Deformation Results

Similarly, we obtained the horizontal deformation maps for every InSAR pair after the conversion from the residual phase correction, phase to deformation (Figure 7), masking areas of low coherence and geo-coding the products to have absolute geographical coordinates. According to the amount of deformation within each pair, the settlement is classified with three different colors (red, white, and blue), representing high risk, medium risk, and low risk deformations, respectively.

Figure 9A shows the 2020 average absolute vertical deformation (mm) and investigation locations, with typical representative characteristics of the study area. Figure 9B shows the 2020 average absolute vertical deformation (mm), the road, and building networks, which are under threat within the study area. Both images show the yearly cumulative amount of deformation ranging from −27.791 (low risk), −42.476 (medium risk), and −57.161 (high risk), respectively.

Table 2. Showing annual yearly average rate of vertical deformation (mm) for the period of study.

S/N	Yearly Image of Study Area	Average Rate of Horizontal Deformation (mm)
1	Sentinel-1 SLC SAR 2016	−24.532 = low	−36.922 = medium	−49.312 = high
2	Sentinel-1 SLC SAR 2017	−18.665 = low	−34.308 = medium	−49.950 = high
3	Sentinel-1 SLC SAR 2018	−28.008 = low	−40.927 = medium	−53.846 = high
4	Sentinel-1 SLC SAR 2019	−26.821 = low	−43.785 = medium	−60.750 = high
5	Sentinel-1 SLC SAR 2020	−27.791 = low	−42.476 = medium	−57.161 = high

below summarizes the annual yearly average vertical deformation (mm) over a period of five years (2016–2020) of our study area with the associated level of risks. The low risk level ranges from −18.665 to −28.008, the medium risk level ranges from −34.308 to −43.785, and the high-risk level ranges from −49.312 to −60.750. The year with the lowest level of risk is 2018 with −28.008, while the year with the highest level of risk is 2016 with −49.312.

The results derived from (Table 1 and Table 2) show that all fourteen investigation locations within the study area (Figure 8A and Figure 9A) experienced significant levels of subsidence from 2016 to 2020. However, some parts of the study area (in the northern part, some parts in the center and a small area on the east side) (Figure 8A and Figure 9A) were stable for the period of investigation. The yearly cumulative amount of horizontal deformation ranged from −25.487 = low risk, −35.126 = medium risk, and −44.775 = high risk, respectively, while the yearly cumulative amount of vertical deformation ranged from −24.532 = low risk, −36.922 = medium risk, and −49.312 = high risk, respectively. Additionally, Figure 9B show additional information on roads and building networks (vector shapefile overlaid on horizontal and vertical deformation maps of 2016 and 2020). A total of 19,781 buildings (both old and new) and 1434 roads (tarred and untarred) were likely to have been affected by subsidence over the years. The level of impact on these roads and building infrastructure may vary across the study area due to the geography of the environment, the geological conditions, population within the area, and rate of extraction of fluids (ground water) over the years [6,17,19]. Again, because most parts of the study area were abandoned coal mine, it is likely that goafs formed by past mining activities and pressure from rapid urban expansion disturbed the overlying strata of these goafs over the years [19]. As a result, the aquifer in their thick surface soil may have lost water and the overlying stratum may have been compacted, thus resulting in an increase in vertical ground subsidence [19]. Similarly, due to the destruction of in situ stress distributions within the study area, the overall structure of the surrounding rock is likely to have been affected over time [10]. Under the influence of some external factors (mostly rapid urban population growth and development), nonlinear horizontal deformations of the ground is experienced [19]. These impacts are similarly likely to increase overtime with adverse effects of global warming, climate change, increased poverty rate, ever-growing population, and fast-growing urbanization with a disregard for sustainability (Figure 10A).

4.3. Holt–Winter Analysis of Absolute Horizontal Deformation Results

Here, we show results of the pattern and trend. We also make predictions of the horizontal deformation across the fourteen (14) locations (Figure 10A) investigated in this study. Data on average yearly rate of horizontal deformation from January 2016 to December 2020 were used. The forecasts were for January 2021 to December 2025 (60 months) (Figure 11).

4.4. Holt–Winter Analysis of Absolute Vertical Deformation Results

Similarly, we show results of the pattern and trends. We also make future predictions of the vertical deformation across the fourteen (14) locations (Figure 12A) investigated in this study. Data on average yearly rate of horizontal deformation from January 2016–December 2020 were used. The forecasts were for January 2021 to December 2025 (60 months) (Figure 13).

4.5. Validation of Prediction Model

We validated how well the prediction obtained from Holt–Winter performs against the original data. Holt–Winter generally makes short-term forecasts. However, due to the peculiarity of our dataset, it was chosen for a five-year forecast. The horizontal and vertical deformation time series data are not stationary, as they contain trends and the seasonal variations are not constant through the series. Again, the seasonal component is expressed in absolute terms in the scale of the observed series. Finally, the model does not make any assumptions about correlations between successive values of the time series data between successive months over the years. To perform the validation for both horizontal and vertical deformation results, we split the time series data into a training set (60%) and a test set (40%). We used the training set of the original data from 2016 to 2018, and predicted the test set of the original data, which is 2019–2020. Next, we made a correlation plot of the predicted 2019–2020 results from the model with the original 2019–2020 test set data set (Figure 10D and Figure 12D). This helped us to visually evaluate and determine how well our model predicted the original data. The validation accuracy was assessed based on the Mean Absolute Percentage Error (MAPE). The best-predicted investigation locations, based on MAPE, for both horizontal and vertical deformation were Ukaku at 36.048% and Okwojo-Ngwo at 43.524%, respectively. The results provide a reasonable forecasting accuracy for our model [20]. Because most of the subsidence values are negative (Figure 11 and Figure 13), MAPE puts a heavier penalty on negative errors than on positive errors [20]. Therefore, comparing the accuracy of predictions, it systematically selects values of forecasts that are relatively low.

5. Discussion

The geometrical impacts created by surface subsidence hazards within Onyeama mine and environment from 2016 to 2020 was measured using SBAS-DInSAR technique (Figure 7). TOPSAR Sentinel-1 SLC time series SAR data were employed for the study (Figure 6A). The results (Figure 8A,B and Figure 9A,B) show examples of horizontal and vertical land deformation velocity maps in mm/year during the period of investigation as identified in each time series. The absolute deformation results were calculated using a quantitative comparative analysis [19]. The outcome of block H to K (of Figure 7) is an unwrapped phase, which is a continuous raster, not yet in metric measure but radian units. To convert the unwrapped phase in radian units to absolute displacements, the Phase to Displacement operator in SNAP software is applied. It translates the phase into surface changes along the line-of-sight (LOS) in meters. The LOS is the line between the sensor and a pixel. Accordingly, positive values mean uplift and negative values mean subsidence of the surface (Figure 11 and Figure 13). The Phase to Displacement operator has no parameters and produces an output, which looks similar to the unwrapped phase, but now each pixel has a metric value indicating its displacement. The absolute deformation values in millimeters are further computed using logical expressions of band math’s tool in SNAP software. The wavelength of TOPSAR Sentinel-1 SLC SAR image is 5.6 cm or 56 mm. Therefore, applying the Expression: Horizontal_Displacement = (unwrapped phase * wavelength)/(−4 * PI), we obtain the absolute horizontal displacement. Similarly, if we apply the Expression: Vertical_Displacement = (unwrapped phase * wavelength)/(−4 * pi x cos (rad (incident angle))), we obtain the absolute vertical displacement [19]. The accuracy of monitored ground subsidence values is directly related to the coherence of the subsidence zones [19]. Hence, the coherence between the reference and the secondary image is estimated as an indicator of the quality of the phase information [21]. If the images have strong similarities, they are therefore usable for interferometric processing. We averaged the coherence coefficient of each map (based on level of risk) to determine the spatial distribution and variations in the subsidence values monitored [19]. The average coherence level for both horizontal and vertical deformation ranges between 0.45 and 0.47 across the time series image. We proceeded to mask out areas of low coherence using band maths and some logical expressions. For each yearly coherence image, we subtracted the minimum value from the maximum value and divided the outcome by two [19]. Next, we applied a logical Expression: IF (Coherence_Image >= outcome) THEN 1 ELSE NaN. This helps eliminate the areas of low coherence while leaving areas of high coherence. The outcome was a Coherence_Masked_Image. Finally, we multiplied the horizontal and vertical deformation image in millimeters in SNAP software band maths using the Expression: Horizontal_Displacement x Coherence_Masked_Image and Vertical_Displacement x Coherence_Masked_Image. The final deformation results of both horizontal and vertical components are shown in Figure 8A,B and Figure 9A,B, respectively. These results and analysis of both horizontal and vertical deformation velocity information can be visualized through the interactive web application (Figure 5).

The mechanism deployed to develop the web application system is seen in Figure 3, while the execution workflow is seen in Figure 4. The flexibility, maintainability, and scalability of the web application is, by the most part, determined by its overall architecture. The key feature of the application is that it combines the underlying complexity of different IT resources, to bring a scalable, reliable, sustainable, ready-to-use, on-demand, and cost-effective solution to the general public [18]. More specifically, the application displays 2D horizontal and vertical deformation maps (absolute data) as well as statistical information (relative data). Each map gives you the spatial extent and overview of surface subsidence within the study location, while statistical information allows for quantifying the impact or describing the rate of horizontal and vertical deformation at a glance. The Reference Base Map (Satellite Image and Open Street Map) supports the visibility of the theme on display, with basic geographic information on administrative boundaries and use case related information, required for the exploration of the time series dataset. The successful execution of queries task will help to obtain answers on the effectiveness, efficiency, and user satisfaction for the application (Figure 5).

Figure 10A and Figure 12A display the results of the constructed Holt–Winter for horizontal and vertical deformation prediction at the investigation location Abor. The monthly forecasts of absolute horizontal and vertical deformation (mm) 2021–2025 are shown as a blue line, and the dark grey and light grey shaded areas show 80% and 95% prediction intervals, respectively. We also investigated whether the predictive model can be improved upon by checking whether the in-sample forecast errors show non-zero autocorrelations at lags 1-20. This was done using a correlogram and by conducting the Ljung–Box test (Figure 10B and Figure 12B). From the literature reviews, if the predictive model cannot be improved upon, there should be no correlations between forecast errors for successive predictions [22,23]. The ‘in-forecast errors’ are calculated as the observed values minus predicted values, for each time point. The correlogram at investigation location Abor (Figure 10B and Figure 12B) shows that the sample autocorrelation for the in-sample forecast errors at lag 3, 5, and 8 exceeds the significance bounds. However, we would expect one out of 20 (5%) of the autocorrelations for the first twenty lags to be outside the 95% significance bounds by chance alone. The result of the Ljung–Box Test is 23.48 and the p-value is 0.266 for the absolute horizontal deformation, while the result of Ljung–Box Test for the absolutes vertical deformation is 19.86 and the p-value 0.4667 respectively. These values indicates that there are no autocorrelations in the in-sample forecast errors at lags 1-20; therefore, we suggest that the predictive model should not be improved upon. For the remaining 13 investigation locations, the result of Ljung–Box Test and the p-values varied accordingly; however, the values indicate that there is no evidence of non-zero autocorrelations [24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41].

In Figure 10C and Figure 12C, we investigate if the forecast errors have constant variance over time and are symmetric or normally distributed with mean zero by plotting a histogram (overlaid with a normal curve). From these histogram plots, it seems plausible that the forecast errors have constant variance over time, and are normally distributed with mean zero. Thus, there is little evidence of autocorrelation at lags 1-20 for the forecast errors (Figure 10B and Figure 12B). This suggests that the model plausibly provides a suitable prediction for all the fourteen locations within our study area. Furthermore, the assumptions upon which the prediction intervals were based are probably valid. Figure 10D and Figure 12D show the correlation plot test to validate how well the prediction obtained from exponential model performs against the original data. The plots display original 2019–2020 data vs. predicted 2019–2020 data and corresponding accuracy results, which validate the prediction from the model. It can be seen that though there is not a perfect fit of the original time series data, the model did reflect the pattern as well as correlate with the seasonal variations and trends of the time series 2019–2020 data.

From Figure 11 and Figure 13, we show a bar chart of yearly absolute horizontal and vertical deformation (mm) prediction from January 2021 to December 2025 for the fourteen (14) investigation locations using Holt–Winter. A weighted ranking analysis, which allocates scale based on a variety of attributes that the selected area possesses, was used [6,19]. By ranking the subsidence values according to scale, the less risky to most risky areas, or those prone to deformation, are highlighted. The rate of subsidence prediction from January 2021 to December 2025 was classified using natural breaks method in five categories of subsidence risk, namely very low, low, moderate, high, and very high.

6. Conclusions

In other to respond to climate change and ever increasing environmental burdens in urban cities, it is important to develop monitoring solutions and innovative plans on how to efficiently and effectively manage the natural environment and reduce the dependence on limited finite resources. This paper proffers a sustainable solution to mitigate these challenges, by developing a low-cost surface subsidence monitoring system to improve disaster risk decision making. The monitoring system exploits Sentinel-1 SAR constellation and Small-BAseline Subset (SBAS-DInSAR) technique, to investigate and measure potential surface subsidence caused by underground hard coal mining activities and exponential urban population growth within a major coalmine in Nigeria. Results of yearly cumulative amount of horizontal and vertical deformation between 2016 and 2020 range from −25.487 mm to −50.945 mm and −24.532 mm to −57.161 mm, for high and low risks, respectively. Under the influence of external factors such as rising poverty and fast-growing urbanization, the destruction of in situ stress distributions will likely increase nonlinear deformation. Before this study, the geometrical impacts, spatial extent, and vulnerability of the environment due to this hazard was hugely underscored. Though at this stage the web application system relies on post-processed deformation information, it can be viewed as a forward step in the area of a land deformation monitoring scheme in Nigeria. Our effort will limit the exposure of humans to environmental hazards and improve our responses to the impact of climate change. The perceived limitation of this web application system has to do with the reliability of the Internet, especially in developing countries where the Internet connections are not everywhere and there are network connectivity issues. However, this would likely improve with the advancement in 5G and IoT technologies. Furthermore, due to non-availability of data, this research did not consider the influences of overlying rock changes in the mining area, construction safety, urban expansion speed, ecological change, or climate conditions on surface deformation. This is highly recommended for further studies.